Nucl - Phys.B v.712

Nuclear Physics B 712 (2005) 319
Supersymmetry from boundary conditions

G. von Gersdorff a , L. Pilo b , M. Quirs c , A. Riotto b , V. Sanz d
a Department of Physics and Astronomy, Johns Hopkins University, Baltimore, MD 21218, USA
b Department of Physics and INFN, Sezione di Padova, via Marzolo 8, I-35131 Padova, Italy
c Instituci Catalana de Recerca i Estudis Avanats (ICREA) and Theoretical Physics Group, IFAE/UAB,
E-08193 Bellaterra (Barcelona), Spain

d Departament de Fsica Terica, IFIC, Universitat de Valncia, CSIC, E-46071 Valncia, Spain
Received 18 November 2004; accepted 5 January 2005

Available online 18 January 2005
Abstract
We study breaking and restoration of supersymmetry in five-dimensional theories by determining
the mass spectrum of fermions from their equations of motion. Boundary conditions can be obtained
from either the action principle by extremizing an appropriate boundary action (interval approach) or
by assigning parities to the fields (orbifold approach). In the former, fields extend continuously from
the bulk to the boundaries, while in the latter the presence of brane mass-terms cause fields to jump
when one moves across the branes. We compare the two approaches and in particular we carefully
compute the non-trivial jump profiles of the wavefunctions in the orbifold picture for very general
brane mass terms. We also include the effect of the ScherkSchwarz mechanism in either approach
and point out that for a suitable tuning of the boundary actions supersymmetry is present for arbitrary
values of the ScherkSchwarz parameter. As an application of the interval formalism we construct
bulk and boundary actions for super-YangMills theory. Finally we extend our results to the warped
RandallSundrum background.
2005 Elsevier B.V. All rights reserved.
PACS: 11.10.Kk
E-mail addresses: gero@pha.jhu.edu (G. von Gersdorff), luigi.pilo@pd.infn.it (L. Pilo), quiros@ifae.es
(M. Quirs), antonio.riotto@pd.infn.it (A. Riotto), veronica.sanz-gonzalez@durham.ac.uk (V. Sanz).
0550-3213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.004
G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319
1. Introduction
A common feature of five-dimensional (5D) supersymmetric models are fermions propagating in the bulk of the extra dimension. In order to extract physical predictions at low
energies, the four dimensional mass spectrum of those fermions has to be known. For instance supersymmetry breaking is determined by the mass spectrum of the gravitino and
the existence of a zero mode signals an unbroken supersymmetry. Similarly, when gauge
multiplets propagate in the bulk supersymmetry breaking is intimately linked to the existence of gaugino zero modes. In particular if supersymmetry breaking is implemented by
non-trivial twist conditions, or ScherkSchwarz (SS) mechanism [1], it acts in the same
way both in the gravitino and the gaugino sectors.
The aim of this paper is to study fermions propagating in a five-dimensional spacetime, with coordinates (x , y), where the compact fifth dimension (with radius R) has
two distinguished four-dimensional hypersurfaces located at y = 0 and y = R. Often
this space is constructed as the orbifold S 1 /Z2 , identifying points on the circle related
by the reflection of the fifth coordinate y y; in this approach y = 0 and y = R
are fixed points and the resulting spacetime is a singular space without boundaries. Fields
with odd parity with respect to the Z2 reflections are zero at the fixed points, while the
normal derivative of even fields is forced to vanish there. The treatment of fermions is
complicated in the presence of brane actions localized at the boundaries. In the orbifold
approach these brane actions are introduced with delta-function distributions peaked at the
location of the orbifold fixed points. The latter induce discontinuities in the wave functions
of the fermions, which take different values at the fixed points and infinitesimally close to
them [2,3]. A possible way to avoid these problems is working from the very beginning in
the fundamental region of the orbifold [0, R] and giving up the rigid orbifold boundary
conditions (BCs):1 the fields are then continuous and the BCs are determined by the action
principle applied to the bulk and boundary conditions. This is called the interval approach:
contrary to the orbifold approach the spacetime is not singular but it has two boundaries
at y = 0 and y = R. The interval approach leads to physically equivalent results to those
of the orbifold approach without any need of using, as the latter, singular functions.2 To
summarize in the orbifold approach one imposes fixed (orbifold) BCs while the brane
action induces jumps, whereas in the interval approach one imposes continuity and the
brane action induces the BCs.
In a previous letter [4] we followed the interval approach and showed how one can obtain consistent BCs by the use of the action principle:3 variation of the bulk action gives
rise both to a bulk term and a boundary term, while the variation of the boundary action is
always localized at the branes. Imposing the whole variation to vanish the bulk terms give
the usual bulk equations of motion (EOM) while the boundary terms give rise to the corresponding BCs (see [6] for a recent application to symmetry breaking). In [4] it was also
1 With some abuse of language we call boundary conditions the parity assignment for fields in the orbifold in
order to make contact to the interval approach.
2 The interval approach is sometimes called downstairs approach while the orbifold approach is called
upstairs approach.
3 For an alternative approach see [5].
shown that for consistent (nontrivial) BCs one needs to appropriately constrain the boundary action; in particular, a vanishing brane action leads to inconsistent BCs. The BCs can
be seen to represent a point in the Riemann sphere and hence are given by complex numbers zf , f = 0, . Explicit formulae for mass eigenvalues and eigenfunctions as functions
of zf can then be obtained. Whenever the BCs at the two branes are the same, z0 = z ,
a zero mode exists and N = 1 supersymmetry remains unbroken. An interesting phenomenon occurs when in addition a ScherkSchwarz [1] breaking is turned on. The latter can
be implemented as a vacuum expectation value (VEV) for the auxiliary vector field VM
which gauges the SU(2)R automorphism symmetry [7] as VM = M5 q , characterized
by the SS parameter and the SS direction (unit vector) q . Whenever q is aligned4 with
either z0 or z , the mass spectrum becomes independent on the SS parameter . In particular this means that if in addition z0 = z , supersymmetry remains unbroken whatever
the value of the SS parameter is. This was dubbed persistent supersymmetry in Ref. [4].
As an application of the interval formalism, we write down the action of super-YangMills
(SYM) theory. A boundary action is required both for consistent BCs as well as (global)
supersymmetry.
In the present work we also shed some light on the relation of the interval approach and
the more common orbifold picture in which BCs are fixed by the parity assignments of the
fields. We show how in the latter the mass spectrum can be computed by calculating the
highly generally nontrivial jump profiles across the branes. In order to consistently treat the
singularities arising from the presence of delta-functions we formally use regularized deltafunctions. The jumps determine the values of the wave functions an infinitesimal distance
away from the branes and thus can be used as new BCs in the general formulae for the
solution to the bulk EOM. The impact of fermionic brane mass terms on the spectrum
as well as its relation to the SS mechanism have been discussed before in a number of
papers [2,3,8,9], with sometimes contradictory results due to incorrect treatment of the
discontinuities. We believe that our present paper gives a clear and consistent treatment of
this slightly involved issue by employing the most general brane Lagrangian considered so
far.
The structure of the paper is as follows: Section 2 contains the results derived in [4]
on the interval approach. In Section 3 the action of the super-YangMills (SYM) theory is
derived. In Section 4 we give the treatment of the most general boundary mass terms in the
orbifold picture, as outlined above. Emphasis is put on the careful (regularized) treatment
of the singular profiles of the wave functions. In Section 5 we extend our results to the case
of warped (RS) geometry. Finally in Section 6 we draw our conclusions.
2. Fermions in the interval

In this section we recall some of the results of [4]. We will take the fermions to be
symplectic-Majorana spinors, although a very similar treatment holds for the case of fermionic matter fields associated to Dirac fermions. In particular we will consider the gaugino
4 See below for a precise definition of this term.
case, the treatment of gravitinos being completely analogous. The 5D spinors i satisfy
the symplectic-Majorana reality condition and we can represent them in terms of two chiral
4D spinors according to5
i
j
, i ij ,
i =
(2.1)
i
where ij = i(2 )ij and im j m = ji . Consider thus the bulk Lagrangian
i
i
Lbulk = i M DM = M DM DM M ,
(2.2)
2
2
where the last equation is not due to partial integration but holds because of the symplecticMajorana property, Eq. (2.1). The derivative is covariant with respect to the SU(2)R automorphism symmetry and thus contains the auxiliary gauge connection VM . The field VM is
nonpropagating and appears in the off-shell formulation of 5D supergravity [11]. A vacuum
expectation value (VEV)6
5
q , q2 = 1
VM = M
(2.3)
R
implements a ScherkSchwarz supersymmetry breaking mechanism [1] in the Hosotani
basis [7,12]. The standard form of the SS mechanism, originally introduced for circle compactification, can be recovered by a gauge transformation U that transforms away VM but
twists the periodicity condition for fields charged under SU(2)R on the circle. As a matter
of fact, in the interval a SS breaking term is equivalent to a suitable modification of the BCs
at one of the endpoints. The unitary vector q points toward the direction of SS breaking.
We supplement the bulk action by the following boundary terms at y = yf (f = 0, ) with
y0 = 0 and y = R

1
1
(f )
Lf = T (f ) + 5 V (f ) = i Mij j + h.c.,
(2.4)
2
2
where T (f ) and V (f ) are matrices acting on SU(2) indices,

M (f ) = i2 T (f ) iV (f )
(2.5)
and we have made use of the decomposition (2.1). Notice that the mass matrix is allowed
to have complex entries. Without loss of generality we take it to be symmetric, which
enforces T f and V f to be spanned by Pauli matrices.
The total bulk + boundary action is then given by

5
4
d x L0
d 4 x L .
S = Sbulk + Sboundary = d x Lbulk +
(2.6)
y=0
y=R
The variation of the total action gives the standard Dirac-like bulk equation of motion
provided that all the boundary pieces vanish. The latter are given by
i

(f )
ij + Mij j + h.c. y=y = 0.
(2.7)
f
5 We use the WessBagger convention [10] for the contraction of spinor indices.
6 Consistent with the bulk equation of motion d (
q V ) = 0 [11].
Since we are considering unconstrained variations of the fields, the BCs we obtain from
Eqs. (2.7) are given by

(f )
ij + Mij j y=y = 0.
(2.8)
f
These equations only have trivial solutions (are over-constrained) unless

(0)
()
det ij + Mij = det ij + Mij = 0.
(2.9)
Imposing these conditions we get the two complex BCs which are needed for a system of
two first order equations. Note that this means that an arbitrary brane mass matrix does not
yield viable BCs; in particular a vanishing brane action is inconsistent7 since det(ij ) = 0.8
However this does not imply that the familiar orbifold BCs 1 = 0 (2 = 0) cannot be
achieved; in the interval approach they correspond to M = 1 (M = 1 ).
The BCs resulting from Eqs. (2.8) are of the form
1 1

cf + cf2 2 y=y = 0,
(2.10)
f
where cf1,2 are complex parameters or, setting zf = (cf1 /cf2 )

2

zf 1 y=y = 0, zf C.
f
(2.11)
Physically inequivalent BCs span a complex projective space CP 1 homeomorphic to the

Riemann sphere. In particular zf = 0 leads to a Dirichlet BC for 2 , and the point at infinity
zf = leads to a Dirichlet BC for 1 . Notice that these BCs come from SU(2)R breaking
mass terms. Special values of zf correspond to cases when these terms preserve part of the
symmetry of the original bulk Lagrangian. In particular when both the SS and the preserved
symmetry are aligned those cases can lead to a persistent supersymmetry as we will see.
Once (2.9) is satisfied the values of zf in terms of the brane mass terms are given by
(f )
zf =
M11
(f )
1 + M12
(f )
1 M12
(f )
(2.12)
M22
where the second equality holds due to condition (2.9).

The mass spectrum is found by solving the EOM with the BCs (2.11). To simplify
the bulk equations of motion it is convenient to go from the Hosotani basis i to the SS
one i , related by the transformation

y
.
= U , U = exp i q
(2.13)
R
In the SS gauge the bulk equations read
i M M = 0.
(2.14)
7 In the sense that the action principle does not provide a consistent set of BCs as boundary equations of
motion.
8 Notice that this agrees with the methods recently used in Ref. [13].
We now decompose the chiral spinor i (x, y) in the Hosotani basis as i (x, y) =
i (y)(x), with (x) a 4D chiral spinor. Setting = U we get the following equations
of motion in the SS basis
d j
d i
= 0,
m j ij +
= 0.
m i ij
(2.15)
dy
dy
The parameter m in Eq. (2.15) is the Majorana mass eigenvalue of the 4D chiral spinor9
i = m,
i = m.
(2.16)
As a consequence of the transformation (2.13) the SS parameter manifests itself only in

the BC at y = R:10

2
0 1
= z0 ,
y=0

2
tan()(iq1 q2 iq3 z ) + z
,
=
1
(2.17)
y=R tan()(iq1 z + q2 z + iq3 ) + 1
where f are the BCs in the SS basis. In particular the BC is a function of , q and z .
From this it follows that we can always gauge away the SS parameter in the bulk Lagrangian going into the SS basis through (2.13). However now in the new basis reappears
in one of the BCs.
The bulk equations have the following generic solution

a cos(my) + z 0 a sin(my)
,
(y) =
(2.18)
a sin(my) + z0 a cos(my)
where a is a complex number given in terms of z0 and :
a=
1 + z0
z0
+
.
|z0 | |1 + z0 |
(2.19)
The solution (2.18) satisfies the BCs Eq. (2.17) for the following mass eigenvalues

z0
1
n

,
arctan
mn = +
(2.20)
R R
1 + z0
where n Z. When z0 = there is a zero mode and this corresponds to an unbroken
supersymmetry. Indeed, the only sources of supersymmetry breaking reside on the branes
(gaugino mass terms) and setting them to cancel each other, z0 = z , preserves supersymmetry [14]. Once supersymmetry is further broken in the bulk, an obvious way to restore it
is by determining z as a function of z0 and using the relation (2.17) with = z0 . This
will lead to an -dependent brane-Lagrangian at y = R. In this case we could say that
supersymmetry that was broken by BCs (SS twist) is restored by the given SS twist (BCs)
[9].
9 The bar acting on a scalar quantity, as, e.g., , and a chiral spinor, as, e.g.,
, denotes complex conjugation.
i
10 Notice that U (y = 0) = 1. The roles of the branes and hence of z and z can be interchanged by considering
0
the SS transformation U
(y) U (y R).
There is however a more interesting case: suppose the brane Lagrangian determines z
to be
q3
z = z(
(2.21)
q)
q1 iq2
with = 1. This special value of z is a fixed point of the SS transformation, i.e. f = zf .
q ) the spectrum becomes independent on . In other words, for this special subFor z = z(
set of boundary Lagrangians, the VEV of the field q V5 does not influence the spectrum.
The reason for this can be understood by going back to the Lagrangian which we used to
derive the BCs. From the relation (2.12) one can see that condition (2.21) is satisfied by
the mass matrix
()
M12 = q3 ,
()
M11 = (q1 + iq2 ),

()
M22 = (q1 iq2 )
(2.22)
which can be translated into a mass term at the boundary y = y along the direction of
q in the notation of Eq. (2.4). In particuthe SS term, i.e., V () = 0 and T () =
lar this brane mass term preserves a residual U (1)R aligned along the SS direction q . In
other words the SS-transformation U leaves both brane Lagrangians invariant and can
q ), i.e., V (0) = 0 and T (0) = T ()
be gauged away. When we further impose z0 = z(
the U (1)R symmetry is preserved by the bulk. In particular if z0 = z(
q ) supersymmetry remains unbroken, although the VEV of q V5 is nonzero. One could say that in
this case the theory is persistently supersymmetric even in the presence of the SS twist,
q ) the theory is persiswith mass spectrum mn = n/R. On the other hand if z0 = z(
tently nonsupersymmetric and independent on the SS twist: the mass spectrum is given by
mn = (n + 1/2)/R. In this case supersymmetry breaking amounts to an extra Z
2 orbifolding [15].
Actually (2.22) is the most general solution of Eq. (2.21). We can set T (f ) = tf and
(f
V ) = vf . The constraint (2.9) on the boundary mass-matrix translates into
tf2 v2f = 1,
tf vf = 0.
(2.23)
Consider now f = , using an SU(2) transformation we can always rotate v in the

z-direction. As a result without loss of generality we can take v = (0, 0, v3 ). Imposing
that = z and tf2 v2f = 1 we get
tf =
q + (v3 ),
2 = 1;
(2.24)
where is a vector which depends on v3 . The last constraint tf vf = 0 is satisfied only if
q .
v3 = 0, which gives (v3 = 0) 0. Then V () = 0 and T () =
3. Super-YangMills action in the interval

Up to now we have focused on the fermion sector spectrum. Adding the complete vector
multiplet does not invalidate our conditions for supersymmetry restoration as long as the
10
supersymmetry breaking brane mass terms are of the form given by Eq. (2.4). We would
like to show the invariance of the gaugino Lagrangian, Eq. (2.6), under (global) supersymmetry. To this end we will consider the super-YangMills multiplet containing the gauge
field BM with field strength GMN , the gaugino , the real scalar and the auxiliary
Clearly since we are not imposing a priori any BC on the fields in the
SU(2)R triplet X.
action, we have to worry about the total derivatives which arise in the variation of the bulk
action. The latter is given by

SYM
Sbulk
=
M
1
1
X

GMN GMN D2 DM DM + 2X
4
2

+ i M DM + igfABC A B C .
(3.1)
Here the sum over the adjoint indices of the fields is suppressed and D denotes the gauge
covariant derivative. Under a global supersymmetric transformation the Lagrangian transforms into a total derivative giving rise to the supersymmetry boundary-variation

SYM
Sbulk =
(3.2)
i 5 ,
M
where is given by

M DM 1 MN GMN 1 M DM .
= iX
4
2
(3.3)
To compensate for this we add the brane action

SYM
=
Sbrane

+ 1 T (f )
2T (f ) X
2
(3.4)
which transforms into

SYM
Sbrane
=
T (f ) .
(3.5)
Now the supersymmetry variation at each boundary is proportional to (1 + i 5 T (f ) )(yf ).

Denoting with (see Eq. (2.1)) the upper part of , whenever (T (f ) )2 = 1 these variations can cancel provided the transformation parameter satisfies the BCs 2 = z(T (f ) ) 1 .
The only possibility is that T (0) = T () , since is constant for global supersymmetry. Notice that according to Eqs. (2.9) and (2.5), this gives rise to the same BCs for the gaugino,
= = 0. The bottom line
2 = z(T (f ) )1 . The remaining EOM then fix the BCs G5 = X
of the off-shell approach is that, in the presence of a boundary, at most one supersymmetry
can be preserved. Global SUSY invariance for the action of a vector multiplet singles out
a special boundary mass term for gauginos such that z0 = z which is at origin of the zero
11
mode in the spectrum (see Eq. (2.20) for = 0 11 ). We expect there to be a locally supersymmetric extension of the action (3.1) + (3.4) for T (0) = T () . In this case the SU(2)R
auxiliary gauge connection VM from the supergravity multiplet gives an additional source
of supersymmetry breaking. Notice that for a globally supersymmetric vacuum there must
then be a solution to the Killing spinor equation
5 D5 (y) = 0,

2 (yf ) = z T (f ) 1 (yf ),
(3.6)
where DM is covariant with respect to SU(2)R . These equations coincide with the zero
mode condition for the gaugino considered above.
4. Fermions in the orbifold

In this section we will consider the same system of a symplectic Majorana spinor but
in the more common orbifold approach. As we have seen in the previous section, in the
interval framework once the action is given we obtain the bulk equations supplemented
with a consistent set of BCs. Let us now study what happens in the orbifold geometry for
an SU(2)R fermionic doublet i where SU(2)R is identified with the automorphism of
N = 2 supersymmetry algebra. We will consider the direct product of a flat 4D Minkowski
spacetime times the orbifold S 1 /Z2 , the (flat) geometry considered in Section 2. In the
S 1 /Z2 orbifold the fifth coordinate runs now along the circle y [R, R] and we
assign to the spinors the following parities
1 (x, y) = 1 (x, y),
2 (x, y) = 2 (x, y),
1 (x, R y) = 1 (x, R + y),

2 (x, R y) = 2 (x, R + y),
(4.1)
where = 1. The second condition is often replaced by demanding periodicity ( = +1)

or antiperiodicity ( = 1) and corresponds to an intrinsic parity for the inversion with
respect to y = R. The orbifold Lagrangian is then given by
(0)

L = i M DM + 2(y) Nij i j + h.c.
()

+ 2(y R) Nij i j + h.c. .
(4.2)
(0,)
Dirac-like mass terms mixing 1 and 2 , N12 , must have an odd profile as it is obvious
from the parity assignments in Eq. (4.1). If they are continuous at y = 0, R they do not
contribute to the brane mass Lagrangian in (4.2); therefore if they contribute they must
possess a discontinuity at the fixed points y = 0, R. The simplest ansatz is that near the
11 In the global theory on the interval all supersymmetry breaking is encoded in the T (f ) mass matrix. There
is no auxiliary field VM whose VEV could contribute to the breaking, nor can one choose a SS twist by hand,
since the BCs are uniquely fixed by the equations of motion.
12
fixed point at y = yf (f = 0, ) they behave as12

(f )
(f )
N12 = ND f (y),
(4.3)
where 0 = (y), = (R y) and the sign function (y) is defined as

y
(y) = 2
(z) dz.
(4.4)
In the simplest case where there is no mass localized at the orbifold fixed points (e.g.,
N (f ) = 0) there is a straightforward correspondence between the orbifold and the interval
approach: we can take the fields continuous across the orbifold fixed points. Then using
the parity assignment (4.1) and periodicity we have

2 R = 0.
2 0+ = 0,
(4.5)
Being the EOM the same we recover the interval result simply using as BC zf = 0. As
a result, when no mass term is present at the orbifold fixed points the parity assignment
in the orbifold is equivalent to the choice of BCs in the interval. As soon as we turn on a
mass term in the orbifold fixed point the correspondence is much more involved. First of
all we have to immediately face a technical problem inherent to the orbifold construction:
to give a meaning to the fixed point Lagrangian which contains the product of distributions
with overlapping singularities. In order to do that often implicit or explicit assumptions
regarding the continuity properties of the fields are made. We want to stress here that such
assumptions can lead to inconsistencies since solutions to the EOM might not exist. On the
other hand by regularizing the delta functions in (4.2) one can consistently assume all fields
to be smooth while only in the limit of a sharp delta the solutions to the EOM will develop
discontinuities. For a sharp delta function, (y) is simply the step function with value +1
(1) for y > 0 (y < 0). For a regularized delta function (y) is a regular odd function
that therefore satisfies the property (0) = 0. In all the expressions that follow delta and
epsilon functions should be considered as regularized with the sharp limit implicitly taken
at the end of the calculation.
Our strategy will thus be the following. In a first step we solve the EOM close to the
branes using the orbifold BCs (4.1). Taking the limit of sharp distributions we determine
the precise jump profile across the branes which will result in modified BCs, at 0+ and
, respectively. These modified BCs can again be encoded in complex numbers denoted
by z0+ , z which are functions of the brane masses. Since the bulk EOM are identical
to those in the interval approach we can directly use the results from Section 2, that is
Eqs. (2.18) to (2.20), with the only replacements z0 z0+ and z z . This establishes
the precise relation between the orbifold and interval approaches.
We use the notation of Section 2, where i (x, y) = i (y)(x) are the components of
the symplectic-Majorana spinors in the Hosotani basis. Going to the SS basis, i , =
12 Of course different anstze could be considered, as an odd power of (y) or any other odd function. We will
just consider in this paper the usual case of a linear behavior in (y).
13
exp (i q y/R) we obtain the following EOM in the orbifold geometry

dj
(f )
+ m i 2
ij
(y yf )Nij j = 0.
dy
(4.6)
f =0,
To determine the precise form of the discontinuities we consider the EOM close to the
fixed points at y = 0
(0)
d 1
(0)
+ 2 N22 2 + N12 1 (y) = 0,
dy
(0)
d 2
(0)
+ 2 N11 1 + N12 2 (y) = 0.

dy
(4.7)
(4.8)
Note that we have neglected the term m i as well as the one (y ) since they are
negligible close to y = 0. This approximation becomes more and more accurate the sharper
the distributions are taken, and it is in fact exact in the singular limit. At y = R the
equations are in the same form mutatis mutandi 0 by . From here it is easy to see that
making assumptions about continuity of the fields may fail. For instance, continuity at
(0)
y = 0 13 of both 1 and 2 is clearly consistent only with Nij = 0, while the weaker
(0)
(0)
(0)
(0)
assumption of only 1 smooth is inconsistent unless N22 = N12 = 0 or N11 = N12 = 0.

To avoid these difficulties we proceed as follows. We can solve Eqs. (4.7) and (4.8) by
assuming (close to the fixed point at y = 0) the functional dependence i = i ((y)) and
using the chain rule as
d i
= 2( i )
(y),
dy
where ( i )
d i /d. Using now (4.9) we can cast Eqs. (4.7) and (4.8) as
1
(0)
(0)
+ N22 2 + N12 1 = 0,
2
(0)
(0)
N11 1 N12 2 = 0.
(4.9)
(4.10)
(4.11)
Passing from Eqs. (4.7) and (4.8) to Eqs. (4.10) and (4.11) makes sense for any regularized
delta function for which we can solve our EOM. As we said we can consistently take the
sharp limit after solving the EOM. In that case, using (0) = 0 we can solve Eqs. (4.10)
and (4.11) with the BCs 2 (0) = 0, 1 (0) = 1. One gets for any regularized delta function
the exact analytical solution

(0)
(0) (0)
ND 2
N11 N22 1 (0) 2
1

(y)
(y) = e 2
(4.12)
; ; ND (y) ,
1 F1
(0)
2
4ND

(0)
(0) (0)
ND 2
N11 N22 3 (0) 2
(0)
2
(y)
2
(y) = N11 (y)e
(4.13)
; ; ND (y) ,
1 F1 1
(0)
2
4ND
13 To see this one typically integrates between 0 and 0+ . Note that a discontinuous even function has the same
value at 0 and 0+ but a different one at 0.
14
where 1 F1 is the Kummer confluent hypergeometric function.14 From Eq. (4.12) and the
definition of the Kummer function it is easy to check that the odd field is discontinuous at
(0)
(0)
(0) (0)
the brane unless N11 = 0. The even field makes a jump unless ND = 0 and N11 N22 = 0.
In fact the solutions (4.12)(4.13) have an interesting expression in the limit where the
(0)
Dirac mass ND 0. They are given by
(0) (0)

1
lim (y) = cos N11 N22 (y) ,
(4.14)
(0)
ND 0

(0)
N
(0) (0)

lim 2 (y) = 11
sin
N11 N22 (y) .
(0)
(0)
N22
ND 0
(4.15)
Similarly we can solve Eqs. (4.6) on the vicinity of R by assuming that near the
brane there is a functional dependence as i = i ( (y)) and using d (y)/dy = 2(y
R). The solutions can be easily read off from Eqs. (4.14) and (4.15) by simply changing
(0)
()
(0)
()
Nij Nij and 0 (for = 1) or by changing Nij Nij , 0 and
1 (y) 2 (y), i.e., by simply label changing 1 2 and 0 (for = 1).
We can now use the behaviour of the solutions close to the fixed points to read out the
jumps for odd and even fields caused by the presence of the delta-functions in the EOM.
We have at y = 0+

(0)
2 0+ = N11
1 F1

1
1 F1
(0)
(0)
N11 N22
(0)
4ND
(0) (0)
N11 N22
(0)
4ND
(0)
; 32 ; ND

1 0+ .

(0)
(4.16)
; 12 ; ND
Similar expressions at are obtained by doing the before-mentioned changes. Now we

can identify the values of the zf parameters entering in the mass-formula. From the jumps
at y = 0+ one obtains
(0)
1 F1
z0+ = N11

1
1 F1
(0)
(0)
N11 N22
(0)
4ND
(0) (0)
N11 N22
(0)
4ND
(0)
; 32 ; ND
(0)
(4.17)
; 12 ; ND
while z can be obtained by the above mentioned trivial substitutions. In the particularly
simple case of no Dirac mass one obtains the expressions

(0)
N
(0) (0)
tan
N11 N22 ,
lim z0+ = 11
(4.18)
(0)
(0)
N22
ND 0
14 The Kummer confluent hypergeometric function is defined by the series expansion
1 F1 [a; b; z] = 1 +

a(a + 1) (a + k 1) zk
.
b(b + 1) (b + k 1) k!
k=1

lim z = i
()
ND 0

N ( )11
i tan N ( )11 N ( )22
.
(
N )22
15
(4.19)
As for the solution of the EOM far from the branes, the form is identical to the one in
the interval case, Eq. (2.18), but now in the mass formula one has to use the quantities z0+
and z , which are given by different functions of the brane masses.
To cross-check the validity of this procedure we have done a numerical integration of
the full EOM, Eqs. (4.6). We indeed find that solutions only exist for the mass eigenvalues
given by Eq. (2.20), where zf are calculated from the mass parameters through Eqs. (4.17).
Let us summarize the orbifold approach. In this framework the BCs are uniquely fixed
by the parity assignments, while the boundary Lagrangian cause the wavefunctions to
jump. We have shown how these jumps can be computed by regularizing the delta functions, thereby obtaining smooth wave-functions. The results are manifestly independent
of the regularization, so this gives a well-defined procedure. The jumps define new BCs
at 0+ and which can be used to solve the bulk equations in the interval [0+ , ]. If
these BCs are aligned with the ScherkSchwarz direction, we again find the spectrum to
be independent of the SS parameter.
Finally let us stress that we have computed the orbifold mass eigenvalues and eigenfunctions for arbitrary brane mass terms. Our results generalize (and agree with) the particular
cases previously considered in the literature [2,3,9].
5. Warped geometry
In this section we extend our previous results in the interval approach to warped geometry. The SU(2)R automorphism group found in the off-shell bulk supergravity action in
flat space is broken by the warping down to two surviving U (1)s. One of them is gauged
by the graviphoton AM with gauge coupling g, and the other U (1) invariance becomes
redundant after the elimination of the auxiliary field V in terms of the graviphoton. The
gauged U (1) couples to the fermions via the off shell Lagrangian [11]
i
i
Lbulk = M DM DM M 2c t ,
(5.1)
2
2
where is a symplectic-Majorana fermion, the field t is an SU(2)R -triplet auxiliary-field
from the minimal supergravity multiplet, the covariant derivative is given by
1
i
DM = M AB M AB VM ,
(5.2)
4
2
and c is the bulk mass term fixed by supersymmetry in the warped space. In particular
c = 1 for gauginos.
The equations of motion then imply that [11]
1
t = g q,
4 3
while for VM one finds a relation with the graviphoton AM as
d(
q V 2gA) = 0,
q 2 = 1,
(5.3)
(5.4)
16
which gives the solution

q (CM + gAM ),
VM = 2
dC = 0.
(5.5)
The one form C is closed, dC = 0, but not necessarily trivial, C = 0. For nonzero g it
can however be absorbed into the graviphoton AM by a shift.15 All the ScherkSchwarz
breaking will then be encoded in the graviphoton VEV, A = dy. The graviphoton will
gauge the U (1)R SU (2)R subgroup defined by the unit vector q .
For the on-shell gaugino action we thus get
i
i
g
Lbulk = M DM DM M c q ,
2
2
2 3
(5.6)
where the covariant derivative is now given by

1
DM = M ABM AB i q gAM .
(5.7)
4
Let us now turn to the bosonic sector. In order to obtain a viable phenomenology we will
restrict our discussion to the RandallSundrum model [16], which leads to 4D Minkowski
spacetime. The warping in the bulk must then be balanced with the brane tensions ending
up with the tunings of the RandallSundrum model. Moreover local supersymmetry requires the bulk (AdS) cosmological constant to be related with the fermion coupling to the
graviphoton, = g 2 . The metric in conformal coordinates can be written as

a(u) = (ku)1 .
ds 2 = a 2 dx dx + du2 ,
(5.8)
The conformal coordinates (x, u) are related with the coordinates
in [16] by u(y) =
k 1 exp(ky). The constant k is related to the gauge coupling g as g = 3k. The graviphoton background in the conformal frame reads as
A = a du.
(5.9)
Concerning the boundary Lagrangian T (0) , V (0) will be the mass terms in the UV brane
located at u = uUV = 1/k and (uIR k)5 T () , (uIR k)5 V () the mass terms in the IR brane.
Proceeding as in Section 2, the BC at the IR brane is twisted by as given in (2.17).
Using the same notation and setting = /k, the bulk Dirac equation can be written in
the SS frame as

1
i M M a 2 + cka 3 q = 0.
(5.10)
2
The massive KK modes associated with the solutions of (5.10) can be expressed as combinations of Bessel functions of order = |c 1|/2, for q = (0, 0, 1), where the two signs
refer to the two gaugino components. Let us focus on the zero mode, its wavefunction is
given by

q log(ku)(i + c/2) k 1 .
(u) = a 2 (u) exp
(5.11)
15 Note that this does not happen in the flat case.
17
The condition for the existence of a zero mode coming from enforcing the BCs is the
following

(z0 ) cosh + q1 (z0 1) + iq2 (1 + z0 ) q3 (z0 + ) sinh = 0,
k
1
log(uIR /uUV ) =
.
(5.12)
2
2
In the absence of any fine tuning, the coefficients of sinh and cosh should vanish
separately and we get
=
z0 = =
q3
z(
q );
q1 iq2
(5.13)
the same condition as in flat space. There is however an important difference: the values of
zf are fixed by imposing that the SUSY transformations are fulfilled

zf 1 2 y = 0.
(5.14)
f
This will lead z0 , z to depend on the bosonic tensions on the respective branes 0 ,
and on the bulk cosmological constant g. When further imposing one of the Randall
Sundrum
z , and when enforcing the brane-bulk balance,
tunings, 0 = , one gets z0 = 16
0 = 24 g, one obtains z0 = z = z(
q ). Therefore, persistent supersymmetry is guaranteed. Any other BCs deviating from this value would explicitly violate local supersymmetry17 [17,18]. This is in contrast to the flat case, where any BCs for the fermions respect
local supersymmetry, and the breaking of N = 1 SUSY by nonaligned BCs is spontaneous.
On the other hand in the warped case, imposing local supersymmetry implies the BCs
(5.13) and supersymmetry is persistent whatever the VEV of q VM , Eq. (5.5), is. This fact
was already noticed in Refs. [1922]. In our formalism this persistence is a consequence
of the alignment of bulk and brane breaking. This alignment is due to the relation between
bosonic stability in the metric and the fermionic sector via supersymmetry transformations,
Eq. (5.14).
6. Conclusions
We have presented in this paper a detailed study of supersymmetry breaking and restoration when two sources of breaking are present, both ScherkSchwarz and boundary mass
terms. The mass spectrum of fermions is obtained through the equations of motion in the
bulk and brane boundary conditions. These boundary conditions are extracted either in the
interval approach, by extremizing the boundary action, or in the orbifold language, by assigning parities to the fields. We compare the two approaches and compute the nontrivial
jump profiles of the wavefunctions in the orbifold picture for general brane mass terms.
With our procedure of dealing with fermions we study in both approaches supersymmetry
16 The quantity z(
q ) was defined in Eq. (2.21).
17 The special case of flipped BCs [17] corresponds to z = z(
q ), z = = z(
q ).
0
18
breaking by adding the ScherkSchwarz breaking terms and computing the mass spectrum. We find out that for a suitable tuning of the boundary actions supersymmetry is
always present for arbitrary values of the ScherkSchwarz parameter. As an application of
the interval formalism, we construct bulk and boundary actions for supersymmetric Yang
Mills theories. Finally we extend our results to the warped RandallSundrum background.
In this case the SS-direction coincides with the gauged U (1)R subgroup, and local supersymmetry enforces alignment of the boundary conditions with the latter. Therefore any
misalignment is an explicit form of supersymmetry breaking and spontaneous SS breaking
becomes impossible. This conclusion agrees with the findings in Refs. [1922]. Within our
formalism, it follows from the alignment of bulk and brane breakings once one relates the
RandallSundrum tunings in the bosonic sector with the fermionic boundary conditions.
Acknowledgements
This work was supported in part by the RTN European Programs HPRN-CT-200000148 and HPRN-CT-2000-00152, and by CICYT, Spain, under contracts FPA 2001-1806
and FPA 2002-00748 and grant number INFN04-02. One of us (V.S.) thanks T. Okui and
J. Hirn for useful discussions and to the Particle Theory Department at Boston University
and the IPPP Physics Department at Durham University, where part of this work has been
done. Another of us (G.v.G.) thanks the Theory Division of the Deutsches ElektronenSynchroton (DESY) where part of this work was done. Finally three of us (L.P., A.R. and
V.S.) would like to thank the Theory Department of IFAE, where also part of this work has
been done, for hospitality.
References
[1] J. Scherk, J.H. Schwarz, Phys. Lett. B 82 (1979) 60;
J. Scherk, J.H. Schwarz, Nucl. Phys. B 153 (1979) 61.
[2] J.A. Bagger, F. Feruglio, F. Zwirner, Phys. Rev. Lett. 88 (2002) 101601, hep-th/0107128.
[3] A. Delgado, G. von Gersdorff, M. Quiros, JHEP 0212 (2002) 002, hep-th/0210181.
[4] G. von Gersdorff, L. Pilo, M. Quiros, A. Riotto, V. Sanz, Phys. Lett. B 598 (2004) 106, hep-th/0404091.
[5] C. Csaki, C. Grojean, J. Hubisz, Y. Shirman, J. Terning, Phys. Rev. D 70 (2004) 015012, hep-ph/0310355.
[6] C. Csaki, C. Grojean, H. Murayama, L. Pilo, J. Terning, Phys. Rev. D 69 (2004) 055006, hep-ph/0305237;
C. Csaki, C. Grojean, L. Pilo, J. Terning, Phys. Rev. Lett. 92 (2004) 101802, hep-ph/0308038.
[7] G. von Gersdorff, M. Quiros, Phys. Rev. D 65 (2002) 064016, hep-th/0110132.
[8] T. Gherghetta, A. Riotto, Nucl. Phys. B 623 (2002) 97, hep-th/0110022;
G. von Gersdorff, M. Quiros, A. Riotto, Nucl. Phys. B 634 (2002) 90, hep-th/0204041;
C. Biggio, F. Feruglio, A. Wulzer, F. Zwirner, JHEP 0211 (2002) 013, hep-th/0209046;
R. Rattazzi, C.A. Scrucca, A. Strumia, Nucl. Phys. B 674 (2003) 171, hep-th/0305184;
K.Y. Choi, H.M. Lee, Phys. Lett. B 575 (2003) 309, hep-th/0306232.
[9] J. Bagger, F. Feruglio, F. Zwirner, JHEP 0202 (2002) 010, hep-th/0108010;
K.A. Meissner, H.P. Nilles, M. Olechowski, Acta Phys. Pol. B 33 (2002) 2435, hep-th/0205166.
[10] J. Wess, J. Bagger, Supersymmetry and Supergravity, Princeton Univ. Press, Princeton, NJ, 1992.
[11] M. Zucker, Fortschr. Phys. 51 (2003) 899.
[12] Y. Hosotani, Phys. Lett. B 126 (1983) 309.
[13] I.G. Moss, hep-th/0403106.
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
P. Horava, Phys. Rev. D 54 (1996) 7561, hep-th/9608019.

R. Barbieri, L.J. Hall, Y. Nomura, Phys. Rev. D 63 (2001) 105007, hep-ph/0011311.
L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370, hep-ph/9905221.
T. Gherghetta, A. Pomarol, Nucl. Phys. B 602 (2001) 3, hep-ph/0012378.
J. Bagger, D. Belyaev, JHEP 0306 (2003) 013, hep-th/0306063.
J. Bagger, M. Redi, Phys. Lett. B 582 (2004) 117, hep-th/0310086.
Z. Lalak, R. Matyszkiewicz, Phys. Lett. B 583 (2004) 364, hep-th/0310269.
L.J. Hall, Y. Nomura, T. Okui, S.J. Oliver, Nucl. Phys. B 677 (2004) 87, hep-th/0302192.
T. Gherghetta, A. Pomarol, Nucl. Phys. B 602 (2001) 3, hep-ph/0012378.
19
Model building and phenomenology of flux-induced

supersymmetry breaking on D3-branes
Fernando Marchesano a , Gary Shiu a , Lian-Tao Wang a,b
a Department of Physics, University of Wisconsin, Madison, WI 53706, USA
b Jefferson Laboratory of Physics, Harvard University, Cambridge, MA 02138, USA
Received 1 December 2004; received in revised form 18 January 2005; accepted 27 January 2005
Abstract
We study supersymmetry breaking effects induced on D3-branes at singularities by the presence
of NSNS and RR 3-form fluxes. First, we discuss some local constructions of chiral models from D3branes at singularities, as well as their global embedding in flux compactifications. The low energy
spectrum of these constructions contains features of the supersymmetric Standard Model. In these
models, both the soft SUSY parameters and the -term are generated by turning on the 3-form NSNS
and RR fluxes. We then explore some model-independent phenomenological features as, e.g., the
fine-tuning problem of electroweak symmetry breaking in flux compactifications. We also comment
on other phenomenological features of this scenario.
PACS: 11.25.-w; 11.25.Mj; 11.30.Qc
1. Introduction
Despite the tremendous insight that string theory has provided us in understanding field
theory phenomena, most of the progress so far is based to a large extent on the assumption
of exact supersymmetry. Yet, to recover the observed low energy physics, supersymmetry
must somehow be broken. An important open problem in string theory is therefore how to
E-mail addresses: marchesa@physics.wisc.edu (F. Marchesano), shiu@physics.wisc.edu (G. Shiu),
liantaow@schwinger.harvard.edu (L.-T. Wang).
doi:10.1016/j.nuclphysb.2005.01.046
F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058
21
break supersymmetry while maintaining its beneficial features. The precise way in which
supersymmetry is broken is also of great phenomenological importance because current
and upcoming experiments at the Tevatron or the LHC will soon begin to probe the existence of superpartners and possibly the pattern of their soft masses. Over the past few years,
it has become clear that the phenomenological possibilities of string theory increase immensely with the inclusion of branes and fluxes. For example, in the brane world scenario,
supersymmetry can be broken either in the D-brane sector or in the bulk. In the latter case,
the D-brane sector preserves N = 1 supersymmetry, whereas supersymmetry breaking effects in the closed string bulk will be transmuted to the observable open string Standard
Model sector via gravitational interactions. From an effective field theory point of view, the
effects of supersymmetry breaking in the bulk felt by the Standard Model can be described
by a set of soft parameters in the resulting D = 4, N = 1 supergravity Lagrangian.
In this paper, we will focus on soft SUSY-breaking induced by type IIB 3-form NSNS
and RR fluxes [14], which is an example of bulk supersymmetry breaking. An appealing
feature of this scenario is that one can perform rather explicit string theoretical computations of the soft parameters [59]. More precisely, the 3-form fluxes couple to the
DiracBornInfeld (DBI) action describing the degrees of freedom on the D-branes and
hence one can deduce the soft parameters from the relevant operators of the DBI action. In
addition to inducing soft SUSY-breaking terms, flux compactifications have also shown to
play an important role in stabilizing moduli [2,4,10], as well as for constructing metastable
dS vacua [11] and inflationary models [12,13,15] from string theory.
As a first step towards studying the phenomenology of flux-induced SUSY breaking
scenario, we need to construct chiral models in flux compactifications whose low energy
spectrum contains semi-realistic features. Our framework to accomplish this task will be
type IIB orientifold compactifications involving D-branes. There are two known ways in
which non-Abelian gauge groups and chiral fermions charged under them can arise in
this context, either by considering D-branes localized at singularities [16] or magnetized
D-branes [1720]. The latter are related to intersecting D-branes constructions by T-duality
[21].1 In this work, we will restrict ourselves to D3-branes at singularities,2 since the soft
SUSY-breaking effects are much easier to work out. In fact, the soft SUSY-breaking terms
on D3-branes at singularities in a general background with fluxes have recently been derived in [6,7], following the approach in [5]. The reason for the simplicity of this setup
is that the Standard Model is localized on a single stack of D3-branes, filling four noncompact dimensions and being pointlike in an internal manifold M6 . Hence, one can
deduce the flux-induced soft parameters by performing a local expansion of the supergravity background around the location of the D3-branes.3 We will present some global
constructions of chiral models as explicit realizations of this scenario.
Another appealing feature of flux-induced SUSY-breaking is that all the soft parameters
and the -term are induced by different components of the 3-form flux. Hence, we have in
principle all the necessary ingredients to explore issues in electroweak symmetry breaking
1 Magnetized/intersecting D-branes and D-branes on singularities are not string theory constructions which
exclude each other. Indeed, as shown in [22] both can be combined to give new chiral models.
2 For recent work on constructing MSSM flux vacua from magnetized D-branes, see [23].
3 For recent developments on SUSY breaking soft-terms in systems of D3/D7-branes see [8,9].
22
(EWSB). However, it turns out that in the semi-realistic D3-brane models constructed in
this context so far [6,24,25], the would-be -term is absent. This is because the necessary
component of the fluxes which induce the -term is projected out by the symmetry of
the background (more specifically, the orbifold symmetry). Thus, the purpose of this work
is to construct and study some concrete examples of flux compactifications whose low
energy spectrum contains features of the Standard Model, and which allow for fluxes that
induce non-zero soft SUSY parameters and a non-vanishing -term. We also demand that
the models are free of extra chiral exotics charged under the SM gauge group and other
sources of supersymmetry breaking, so that the resulting phenomenological features are
due entirely to the fluxes. Note that unlike in the magnetized D-brane setup, in which a
three-generation MSSM model has been constructed [23], it is rather difficult to construct
a three-generation MSSM flux vacua from only D3-branes at singularities and which also
allows for a -term. In fact, the simplest model we found in this setup has only two chiral
families. Nonetheless, as long as we explore issues which are insensitive to the number of
families, we expect the results to be generic and applicable when more realistic models can
be constructed in the future.
This paper is organized as follows. In Section 2, we construct some chiral models of
flux compactifications from D3-branes at singularities. Our strategy is to first construct a
local singularity where a semi-realistic model can be localized, and which admits fluxes
leading to both the soft SUSY-breaking terms and the -term. We then embed such a local
singularity into a fully-fledged compactification. In Section 3, we explore some phenomenological features of flux-induced supersymmetry breaking and discuss the fine-tuning
problem of EWSB in this scenario. We conclude with some outlook in Section 4. Some details about fluxes and NSNS tadpoles, as well as the quantization conditions for the 3-form
fluxes are relegated to Apendices A and B.
2. Model building
In this section, we construct some simple models where the general flux-induced soft
SUSY-breaking scenario discussed in [6] (see also [7]) can be realized. In order to apply
the formalism developed in [6], we consider type IIB string theory configurations involving
both D3-branes and RR and NSNS 3-form fluxes. The D3-branes fill-up four non-compact
dimensions and sit at a particular point in a six-dimensional manifold M6 , whereas the
3-form fluxes are turned on transversally to the D3-brane worldvolume, i.e., the fluxes
have support on 3-forms in M6 (see Fig. 1). In this scenario, the open string degrees of
freedom associated to the D3-branes yield a supersymmetric gauge theory in their D = 4
worldvolume. The closed string background surrounding the D3-branes, on the other hand,
does not preserve the same supersymmetry as the D3s, and hence induces corrections
in the open string sector. From the point of view of the D3s worldvolume theory, these are
seen as SUSY-breaking soft terms.
A general derivation of the SUSY-breaking soft terms for D3-branes in a generic background was given in [6], with special emphasis on the case where the source of SUSY
breaking is given by the background 3-form fluxes. We will now search for specific models
where such general results can be applied. In order to have chiral N = 1, D = 4 gauge
23
Fig. 1. Flux-induced supersymmetry breaking scenario, for the particular case of D3-branes at singularities. The
D3-branes sit at a singularity of the internal space M6 , in order to achieve a chiral gauge theory. The geometrical background preserves N = 1 supersymmetry but the 3-form flux G3 surrounding the singularity breaks it,
inducing soft terms in the gauge theory.
theories from D3-branes we need to consider a compact manifold M6 containing singular

points, such as orbifold/orientifold singularities, and place the D3-branes on top of them.
The low energy spectrum of such gauge theory can be encoded in a quiver diagram [16],
and can lead to D = 4 chiral theories quite close to the Standard Model [26,27].
The consistency of these kind of models usually requires the presence of D7-branes
at this same singular point [26]. These new open string sectors will also introduce gauge
theories, and possibly extra chiral matter. One could then, in principle, build up a semirealistic model by using both D3- and D7-branes, as is usually the case in the literature. On
the other hand, the computation of soft terms is microscopically better understood in the
case of D3-branes. For this reason, we will focus on models where the gauge theory and
chiral matter of interest is fully embedded in the D3D3 sector of the model, and D7-branes
only play the role of a hidden sector in the low energy theory.
We then introduce non-trivial RR (F3 ) and NSNS (H3 ) background fluxes which will
induce the soft terms in the D3-brane worldvolume theory. These fluxes should be welldefined in the local orbifold geometry which the D3-branes see. The appearance of SUSYbreaking operators of lower dimension is specified by the local behaviour of the flux near
the D3-brane location [6]. We can then, in principle, consider a local orbifold/orientifold
model and perform a local analysis of the flux-induced SUSY-breaking, much in the spirit
of [5]. Indeed, the soft term structure can already be analyzed at this local level, without
knowing the specific embedding of the singularity in M6 .
Eventually, however, we would like M6 to be compact in order to recover D = 4 gravity
at low energies. The fluxes should then be properly quantized in terms of the homology/cohomology of M6 . Moreover, the whole construction must satisfy the full set of
consistency conditions known as RR tadpole cancellation conditions, many of them not
visible from the (local) point of view of the D3s gauge theory [28]. As we will see in
the next section, these global constraints allow us to study other kind of questions otherwise invisible from the local point of view, such as the distribution of phenomenologically
viable models in the space of well-quantized fluxes.
24
Global models of the type described above, based on toroidal orientifolds with nontrivial background fluxes, have already been built in [6,24]. These constructions are based
on D3-branes at C3 /Z3 orbifold singularities,4 which are the more appealing local models from the phenomenological point of view, since they naturally provide triplication of
families [26]. It turns out, however, that these Z3 orbifold geometries, and in particular
the global T6 /Z3 orientifold models constructed in [6,24], forbid any constant component
of the 3-form flux G3 other than the ones proportional to the holomorphic 3-form and its
complex conjugate. From the point of view of the general analysis in [6], this will imply
that only SUSY-breaking soft terms of the dilaton domination form will appear in the
effective theory and, e.g., no -term would be generated.
In this section we build examples of chiral N = 1 gauge theories based on D3-branes
at orbifold/orientifold singularities, which admit a more general pattern of flux-induced
soft terms. We will focus on a C3 /Z4 orbifold singularity, which is the simplest orbifold
example allowing for a flux-induced -term. Following the general philosophy in [26], we
first construct the local features of the D3-brane model, and then present an example of
how this local physics con be embedded in a global construction.
2.1. A local Z4 model
We now describe in some detail the construction of a local Z4 D3-brane model in the
background of SUSY breaking 3-form fluxes. The aim is to find a model which allows
for a semi-realistic gauge group and chiral fermions with appropriate quantum numbers,
both arising from strings ending just on D3-branes. This will allow for a more transparent
analysis of the effect of flux SUSY breaking, which can already be addressed at the local
level. We perform such analysis in Section 3. Although Z4 does not allow for more than
two generation models, the phenomenological analysis of Section 3 shows that the soft
SUSY breaking terms follow an universal pattern, and so we find that, in order to test the
phenomenological possibilities of flux-induced SUSY breaking, the number of generations
is not a crucial factor of the model.
In the following we will use the results and conventions in [26], to which the reader
is referred for more details and a general discussion on model building of D3-branes at
singularities.
2.1.1. Orbifolding
Let us consider a stack of N D3-branes sitting on top of a supersymmetric Z4 orbifold
singularity. Locally, we can represent such singularity by C3 /Z4 , where the generator of
the Z4 orbifold action acts geometrically on C3 as

: (z1 , z2 , z3 ) e2i/4 z1 , e2i/4 z2 , e2i/2 z3 = (iz1 , iz2 , z3 )
(1)
4 More precisely, these models are based on either D3-branes or anti-D3-branes at orbifold singularities,
surrounded by a 3-form flux which breaks supersymmetry. For a more complete discussion of the different possibilities within this scenario see [25].
25
and is encoded in the shift vector v = 14 (1, 1, 2).5 On the other hand, the action on the
D3s ChanPaton degrees of freedom is given by
,3 = diag(1n0 , i1n1 , 1n2 , i1n3 )
(2)
4
with i=1 ni = N . The massless spectrum of open strings starting and ending on D3branes can be obtained by considering an N N matrix 33 and imposing, on the Neveu
Schwarz sector, the conditions
(0) 1
(0)
33 = ,3 33 ,3
for (0)
33
1
33 = e2ivr ,3 33 ,3
(r)
(r)
|0,
12
(3)
(r)
r
for 33
1 |0,
(4)
where labels the D3-brane worldvolume dimensions, (3) representing D = 4 gauge

bosons, and r = 1, 2, 3 labels each of the complex scalars coming from each complex
plane in C3 . The conditions to be imposed in the Ramond sector are the same, the orbifold
being supersymmetric, and we obtain a massless spectrum completely specified by the four
integers {ni }4i=1
Vector multiplets
4
U (ni ),
i=1
Chiral multiplets
3
4

(ni , n i+4vr )
(5)
r=1 i=1
which the index i is to be understood mod 4. This spectrum can be encoded in the quiver
diagram of Fig. 2. This quiver construction can be generalized to arbitrary orbifold singularities, see, e.g., [29].
The quiver diagram shows clearly how the orbifold singularity Z4 provides a chiral
spectrum. Moreover, the chiral multiplets (ni , n i+1 ) come in pairs, which implies two
replicas of the same kind of particle. Of course, from a phenomenological point of view
the natural triplication of families is far more appealing, which can be obtained, e.g., by
means of Z3 orbifolds [26]. However, as explained above, these orbifolds impose too many
constraints on the background fluxes.
In general we can also consider D7-branes filling two complex dimensions in C3 /Z4
and fixed by the geometrical orbifold action. Let us, for instance, consider D7-branes not
wrapping the 3rd complex plane, denoted by D73 -branes, with the ChanPaton action of
the orbifold specified by
,73 = diag(1u0 , i1u1 , 1u2 , i1u3 ).
(6)
The quiver structure of Fig. 2 equally applies to this case. However, since the internal volume of the D7-branes is infinite, the gauge coupling constants of the corresponding gauge
5 We have introduced the shift vector v in order to simplify the discussion. In the language of [26], their
components vr correspond to ar /N .
26
Fig. 2. Quiver diagram of the Z4 orbifold. Here the nodes represent U (n) gauge groups and the arrows bifundamental (n, n ) representations of left-handed chiral multiplets between them, pointing in the sense of the
anti-fundamental.
groups vanish and they become global symmetries of the theory.6 The matter charged under
such global sector and the D3s gauge groups are open strings in the 373 and 73 3 sectors.
The orbifold projection is now
1
373 = eiv3 ,3 373 ,7
,
3
1
73 3 = eiv3 ,73 73 3 ,3
(7)
yielding a chiral spectrum which, in the case of Z4 is

Chiral multiplets
4

(ni , u i+1 ) + (n i , ui1 ) .
(8)
i=1
Finally, for consistency of the theory tadpoles must be cancelled. A local model is only
sensitive to twisted tadpole conditions, which in the case of D3/D7-branes at orbifold singularities read [26]
3
2 sin(kvr ) Tr k ,3 +
r=1
3

2 sin(kvr ) Tr k ,7r = 0.
(9)
r=1
In the case at hand we have

k = 1,
2 Tr 1 ,3 + Tr 1 ,73 = 0,
k = 2,
4 Tr 2 ,3 + Tr 2 ,73 = 0.
(10)
where 2 ,3 = 21 ,3 , etc. Notice that the case k = 0, corresponding to untwisted tadpoles,

identically vanishes. This is because the RR and NSNS tadpoles are suppressed by the
volume transverse to the D-brane where the closed string fields carrying the corresponding D-brane charge can propagate. In the case of untwisted charges the RR (or NSNS)
6 Of course, in a compact model we recover such gauge symmetries.
27
fields propagate all over C3 /Z4 , which is infinite volume. Actually, the same is true for
the twisted tadpole k = 2, since the rotation 2 leaves the third complex plane fixed, so
the corresponding twisted closed strings are not stuck at the orbifold singularity and can
propagate all over this complex plane. As a result, the k = 2 condition above need not be
taken into account in order to build a local Z4 orbifold model.
To summarize, from the point of view of the local physics at the singularity, and hence
the gauge field theory physics, the only relevant consistency condition is the twisted tadpole
for k = 1. Indeed, cancellation of chiral anomalies will be guaranteed by fulfilling such
consistency condition [28]. Thus, at this level of the construction, we only need to bother
about k = 1 twisted conditions. However, untwisted and k = 2 twisted conditions will be
relevant when embedding such singularity in a compact model, as we will show below.
Given this general spectrum we can think of building a semi-realistic local model with
two generations of quarks and leptons. To obtain standard-like or PatiSalam models, the
more appealing choices are
n0 = 3,
n1 = 2,
n2 = 1,
n3 = 1,
n0 = 3,
n1 = 2,
n2 = 2,
n3 = 1,
n0 = 4,
n1 = 2,
n2 = 1,
n3 = 0,
n0 = 4,
n1 = 2,
n2 = 2,
n3 = 0
(11)
all of these choices lead, however, to uncancelled k = 1 tadpoles that will require the presence of D73 -branes and will, eventually, lead to unwanted extra chiral matter in the model.
As pointed out in [30] this non-minimality of the spectrum is generic in orbifold models.
We thus turn to consider orientifolded versions of the above.
2.1.2. Orientifolding
Let us now consider type IIB string theory modded out by R, where R is a Z2 symmetry of the compactification manifold M6 . Such class of theories are indeed a natural
way to achieve non-trivial supersymmetric flux compactifications [2,4]. Let us, moreover,
consider that our previous Z4 orbifold singularity is also fixed by the action of R. In our
local model C3 /Z4 , we can take R : zi zi .7 Since the D3-branes are at a fixed point
of the new orientifold symmetry, we should project the open string degrees of freedom to
those invariant both under Z4 and R, in order to get the final gauge theory. The conditions (3) and (4) now become
1
(0) = ,3 (0) ,3
,
1
,
(r) = e2ivr ,3 (r) ,3
1
(0) = ,3 (0) ,3
,
(r) = ,3
(r) T
1
,3
,
(12)
(13)
where ,3 is the action of R on the ChanPaton degrees of freedom. Again, (0) corresponds to the gauge bosons and (r) to each complex scalar. String interactions are well
7 More precisely, we are considering the standard orientifold projection leading to O3-planes and preserving
N = 1 supersymmetry, which is of the form R, with R = R1 R2 R3 (1)FL . Here Ri acts as zi zi , and
FL is the left-handed world-sheet fermion number.
28
Fig. 3. Folding of the Z4 quiver diagram in Fig. 2 to the corresponding orientifold with vector structure. The
colored nodes represent pairs of U (n) gauge groups identified under the action of R, while the white nodes are
U (m) gauge groups projected to either SO(m) or USp(m).
defined if both conditions

T 1 = 1,
k = k
(14)
are satisfied by the unitary matrices , k , which restricts the possible form of to
a few cases [31]. In general, the orientifold symmetry will identify pairs of gauge groups
and chiral matter multiplets which were present in the orbifold theory, while some other
unitary gauge groups may be reduced to SO or USp gauge groups. Such action can be again
encoded in a folding of the orbifold diagram, as shown in Fig. 3 for the particular case of
interest in this paper.
The procedure of obtaining orientifold from orbifold quivers can be summarized in a
set of rules (for a general discussion of these rules, see, e.g., [32]). In this paper, we will
be interested in the orientifold projection first discussed in [31], and which in the closed
string sector gives rise to leftright symmetric RR states from the Z2 twisted sectors of the
theory. In the particular case of Z4 such projection imposes the constraint
= 2 T2
(15)
for both D3- and D7-branes. Together with (14) this implies
=
= 1, which is
8
known as an orientifold with vector structure. The two inequivalent choices of are
given by
1n0
n0
1n1
1n1
,b =
,a =
(16)
,
,
1n2
n2
1n3
1n3
4
,3
4
,7
8 At this stage, one may also consider Z orientifolds without vector structure. However, these orientifolds
4
possess exotic twisted tadpole conditions [33], which usually imply the presence of non-BPS D-branes in order
to find a consistent vacuum [34]. This would mean an extra source of supersymmetry breaking, and thus does not
seem very promising for our purposes.
29
Table 1
Spectrum of the Z4 orientifold with the choice ,a . Here A1 stands
for the antisymmetric representation of U (n1 )
Gauge group
Chiral multiplets
SO(n0 ) U (n1 ) SO(n2 )

1
2(n0 , n 1 ), 2(n1 , n2 ), (n0 , n2 ), A1 + A
0 1
. Moreover, in order
where n is a block-diagonal matrix with n blocks of the form 1
0
for to have a well-defined action on the ChanPaton matrices, we must impose n1 = n3 ,
u1 = u3 in (2) and (6), respectively.
One choice of or the other will depend on the RR charges of the O-planes placed
at the orientifold singularity. More specifically, an O3-plane with NSNS and RR untwisted
charges opposite to those of a D3-brane, denoted as O3(,) , will correspond to the choice
,3 = ,a , while that with the same charges, O3(+,+) , corresponds to ,3 = ,b .
Notice that the orientifold group element R 2 fixes the two first complex planes of
C3 , and hence also O7-planes will be present in our local model. Consistency conditions
imply that the charges of the O7-plane are opposite to those of the companion O3-plane.
We hence have the two possibilities (O3(,) , O7(+,+) ), which implies ,3 = ,a and
,73 = ,b , and (O3(+,+) , O7(,) ), which corresponds to the opposite choice. Finally,
in the particular case of Z4 such O-planes do not have any twisted charges, which implies
that the tadpole conditions (10) remain unchanged.
Of course, the D3-brane gauge theory and chiral spectrum will strongly depend on the
choice of ,3 , so let us discuss both possible choices (16) in turn.
O(,)
In this case the extra projection in (12) identifies the gauge groups U (n1 ) and U (n3 )
(recall that n1 = n3 ) and projects U (n0 ) and U (n2 ) into SO(n0 ) and SO(n2 ), respectively. The N = 1 chiral multiplets are summarized in Table 1 and Fig. 4.
O(+,+)
This projection again identifies the unitary gauge groups U (n1 ) and U (n3 ), while
U (n0 ) and U (n2 ) are now projected to USp groups. Notice that such group only makes
sense if both n0 and n2 are even integers. We again summarize the N = 1 gauge group
and chiral spectrum in Table 2 and Fig. 5.
These spectra apply to D7-branes as well, by just making the substitution ni ui .
(+,+)
More specifically, if we choose the projection (O3(,) , O73
) the D3-brane spectrum
will be given by Fig. 4 and D7-brane spectrum by Fig. 5, and the other way round if we
choose (O3(,) , O73(+,+) ). Finally, the orientifold projection identifies the 373 and 73 3
sectors, so in order to compute the D3D73 spectrum we only need to consider one of the
contributions in (7). The final result is given in Table 3.
2.1.3. Massive U (1)s
Before any attempt to build a model, it is important to know which of the above U (1)s
are actual gauge symmetries in the low energy theory. Indeed, given a configuration of
D3-branes at singularities the general gauge group will contain U (n) SU(n) U (1)
30
Fig. 4. Quiver diagram of the Z4 orientifold with the choice ,a .
Fig. 5. Quiver diagram of the Z4 orientifold with the choice ,b .

Table 2
Spectrum of the Z4 orientifold with the choice ,b . Here S1 stands
for the symmetric representation of U (n1 )
Gauge group
Chiral multiplets
USp(n0 ) U (n1 ) USp(n2 )

2(n0 , n 1 ), 2(n1 , n2 ), (n0 , n2 ), S1 + S 1
Table 3
Spectrum of the Z4 orientifold in the mixed 373 + 73 3 sector
Chiral multiplets
(n0 , u 1 ), (n1 , u2 ), (n2 , u1 ), (n 1 , u0 )
factors and many of these U (1)s can be seen to develop mixed U (1) non-Abelian gauge
anomalies from the chiral spectrum of the theory. Such field theory anomalies are cancelled
by a generalized GreenSchwarz mechanism mediated by closed string twisted modes [35].
As a result, these U (1) gauge bosons get a Stueckelberg mass, remaining in the effective
field theory as global symmetries.
Let us first consider the orbifold spectrum (5). By computing the field theory U (1)a
SU (nb )2 anomalies, a, b = 0, . . . , 3, is easy to see that the only non-anomalous U (1)s are
given by
Q1 =
Q0 Q2
+
,
n0
n2
Q2 =
Q1 Q3
+
,
n1
n3
31
(17)
where Qi corresponds to the generator of the ith U (1). Notice that Qdiag = Q1 + Q2 is
the diagonal U (1) which is always non-anomalous in this class of models [26].9
On the other hand, when consider an orientifold model, there is essentially one U (1)
factor, now inside U (n1 ), which can be seen to be anomalous. The only way to get a nonanomalous Abelian gauge factor seems to be to consider SO(2) U (1) gauge groups from
Fig. 4.
2.1.4. Constructing the local model
Let us now focus on the model building possibilities of the Z4 orientifolds described
above. Since the Abelian factor of U (n1 ) SU (n1 ) U (1) is anomalous (and hence does
not survive as a gauge symmetry at low energies), it turns difficult to find a natural candidate for the hypercharge or U (1)BL characteristic of leftright symmetric models. We
will then consider building PatiSalam models, which do not include any Abelian gauge
factor before gauge symmetry breaking. If we consider the choice of orientifold projection
(O3(,) , O7(+,+) ), the most realistic gauge groups that we can achieve in the D3-brane
worldvolume theory are given by the choices
n0 = 3,
n1 = 4,
n2 = 3,
n0 = 2,
n1 = 4,
n2 = 3
u0 u2 = 2,
(18)
where in the second choice we have imposed the twisted tadpole condition (10) for the
twist k = 1. Although these choices may lead to semi-realistic gauge groups by using the
identifications SO(3) SU (2) and SO(2) U (1), notice that the would-be Higgs and
quarks are triplets instead of doublets of SO(3).
A more appealing possibility is to consider the orientifold projection (O3(+,+) ,
O7(,) ). Indeed, we can now make use of the identity USp(2) SU (2) and achieve a
PatiSalam gauge group by the choice
n0 = 2,
n1 = 4,
n2 = 2
(19)
as shown in Fig. 6.
Notice that the k = 1 twisted tadpole condition in (10) is satisfied, so that in principle
we do not need the presence of D7-branes in our local model. Notice as well that the chiral
spectrum of this N = 1 PatiSalam model contains two replicas of chiral matter and only
one Higgs multiplet.
2.2. A global Z4 model
Let us now construct an example of a fully-fledged string compactification where
the Z4 orientifold model of Fig. 6 can be embedded. In such construction we will
9 This is generically the only non-anomalous U (1). In the particular case of Z we have an extra one, Q Q ,
4
1
2
basically because 2 leaves one complex plane fixed.
32
Fig. 6. PatiSalam Z4 orientifold model with two-generations.
Fig. 7. Representation of the fixed sets of the T6 /Z4 orbifold. The three squares represent the three T2 tori. Dots
are fixed points in the torus and a grid in a square corresponds to a fixed torus.
also include the background fluxes that will induce the soft terms in the supersymmetric D3-brane effective field theory. Considering a particular global model does not
only specify the gravity sector of the theory as well as possible open string hidden sectors but, as shown below, it also imposes some global constraints on the background
fluxes. As shown in Section 3, these constraints may be relevant when considering
certain phenomenological aspects such as electroweak symmetry breaking in this context.
The simplest compact model where C3 /Z4 singularities arise is the toroidal orbifold
6
T /Z4 , shown in Fig. 7. We are interested in orientifolds of such singular space containing
O3(+,+) s. This can be achieved by modding the theory by R, where R contains the Z2
geometrical action (z1 , z2 , z3 ) (z1 , z2 , z3 ) and satisfies the condition (15). Such
class of models (or rather T-dual versions) have been considered in [36,37], where anti-Dbranes were introduced in order to cancel RR tadpoles.10 In particular, we will consider a
T-dual version of the type I Z4 model in [37], obtained by T-dualizing in the two directions
of the third T2 . Moreover, we will consider the possibility of adding a discrete B-field b =
10 We are not particularly interested in introducing anti-D-branes at an orientifold point, since their effective
theory is already non-supersymmetric.
33
0, 1/2 in the third T2 before T-dualizing. After T-dualizing we obtain a model with either
64 O3(+,+) s (b = 0) or 48 O3(+,+) and 16 O3(,) (b = 1/2) [38,39]. The model will also
contain O73 -planes, namely 4 O73(,) for the choice b = 0 and 3 O73(,) + 1 O73(+,+)
for b = 1/2. Both O3- and O73 -planes contribute only to untwisted tadpoles with (NSNS,
RR) charges

O3 +32 (1 b), +32 (1 b) ,

O73 32 (1 b), 32 (1 b)
(20)
in units of D3- and D7-brane charges, respectively.
Let us now embed the local model described in the previous section in this setup.
More precisely, let us consider the D3-brane content (19) on top of the origin, which is
fixed both under Z4 and R. As we showed previously, the twisted tadpole condition for
k = 1 is automatically satisfied, without considering any extra D3-branes or D7-branes.
In this compact model, however, we also have to satisfy the k = 2 twisted tadpole in
(10). This can be easily done by adding a stack of D3 -branes with ChanPaton factors
,3 = diag(1R+2 , i1R , 1R+2 , i1R )
(21)
with R N. This stack must be placed in an orbifold singularity of the form (0, 0, P ),
where P = 0, 1/2, i/2, (1 + i)/2 is any of the Z2 fixed points in the third T2 (see Fig. (7)).
This is the case because the twisted RR field in the k = 2 sector propagates over the whole
third T2 , and hence the corresponding RR tadpole is blind to positions on the third twotorus.
Let us choose P = (1 + i)/2. That is, the extra stack of D3-branes is localized on the
point (0, 0, (1 + i)/2). Hence, they can be seen as a hidden sector with respect to the PSlike model at the origin, communicating to the latter only by means of gravitational effects.
Similarly, we can introduce a hidden sector of D73 -branes, wrapping the first and second
two-tori and localized at P = (1 + i)/2 in the third T2 . We choose these D7s to have
ChanPaton factors
,73 = diag(1S , i1S , 1S , i1S ),
(22)
where S N. Since these D-branes are in the regular representation of Z4 , they do not
generate any further twisted RR tadpoles.
The spectrum of this hidden sector will depend on the choice of discrete B-field on the
third T2 . In both cases (0, 0, (1 + i)/2) will be a Z4 orientifold singularity, but the O-plane
content will change for the choice of b above. If we choose b = 0, then the O-plane content
at this point is of the form (O3(+,+) , O7(,) ), whereas if we choose b = 1/2 it switches
to (O3(,) , O7(+,+) ). The gauge group in both cases is given by

D3 -branes SU(R) USp(R + 2) USp(R + 2),
b=0
D73 -branes SU(S) SO(S) SO(S),

1
D3 -branes SU(R) SO(R + 2) SO(R + 2),
b=
D73 -branes SU (S) USp(S) USp(S).
2
34
Since we are varying the charge of some orientifold planes by introducing a discrete Bfield, cancellation of untwisted tadpoles does also depends on the choice of b. In particular,
cancellation of RR D7-brane tadpoles reads
4 S = 32 (1 b)
(23)
so we must consider S = 8 for b = 0 and S = 4 for b = 1/2.

2.2.1. Introducing the flux
As the final stage of building the model, let us introduce the background 3-form flux,
which will cancel the untwisted D3-brane RR-charges, stabilize the dilaton/complex structure moduli and induce soft terms in the D3-brane worldvolume. In general we will have
both non-trivial RR and NSNS 3-form field-strength backgrounds, denoted as F3 and H3 ,
respectively. These fluxes must obey the Bianchi identities dF3 = dH3 = 0, as well as be
properly quantized with respect to any 3-cycle M6 [2,4]

1
1
(24)
F3 2Z,
H3 2Z.
2
2
Actually, in toroidal orientifold models there is an extra subtlety in flux quantization [40],
which imposes the integer number in (24) to be even whenever is passing through an
even number of O3(+,+) . This will be the case in the T6 /Z4 orientifold model discussed
above.
The two real 3-form fluxes are usually arranged into a complex 3-form flux as
G3 = F3 4 H3 ,
(25)
where 4 = a + i/gs is the usual type IIB axiondilaton coupling. This field-strength flux
is a source of the self-dual five-form, hence carries D3-brane RR and NSNS charge. The
RR charge is given by the topological number

3
i
G3 G
1
1
H
F
=
Qflux =
(26)
3
3
2 Im 4
(4 2 )2
(4 2 )2 2 Im
M6
M6
which is an integer number given the quantization conditions (24). The sign of this RR
charge depends on the particular decomposition of G3 in imaginary self-dual (ISD) and
imaginary anti-self-dual (IASD) components. In the particular case of T6 /Z4 , there are
four (untwisted) 3-forms surviving the orbifold projection. They are listed in Table 4,
with their decomposition in holomorphic indexes and their Hodge duality transformations.
In the following we will consider our flux G3 to be a linear combination of the IASD
3-forms in Table 4, which implies that they will carry negative RR D3-brane charge. Moreover, this will guarantee that the scalar potential generated by G3 will be at its minimum
(see the related discussion in Appendix B for more details), freezing the complex structure
moduli i i = 1, 2, 3, of T2 T2 T2 and the complex dilaton 4 to some specific values.11
11 Actually, the Z geometrical action already fixes the first two T2 s to be square tori, so we have = = i
4
1
2
and 3 , 4 as free parameters.
35
Table 4
Untwisted cohomology of T6 /Z4 . ISD stands for the 3-forms satisfying 6 = i, and IASD for those satisfying 6 = i
(0, 3)
(2, 1)
(3, 0)
(1, 2)
3-forms
d z 1 d z 2 d z 3
dz1 dz2 d z 3
dz1 dz2 dz3
d z 1 d z 2 dz3
ISD
IASD
More precisely, we will take the 3-form flux to be of the form

G3 = A dz1 dz2 dz3 + B d z 1 d z 2 dz3
(27)
with A, B C. There are two equivalent approaches in the literature for finding flux vacua.
One of them is considering some well quantized fluxes F3 and H3 and looking for points
in the moduli space of complex structures + dilaton where the scalar potential vanishes
[10]. The other is fixing G3 to be of the ISD or IASD form and then demanding the flux to
be well quantized, which is a condition that also depends on the complex structure moduli
and the complex dilaton [24].
In the present paper we will follow the second approach, imposing the quantization
conditions (24) in order to find solutions for a flux of the form (27). We leave the general
derivation and final expression of the flux quantization conditions to Appendix A. Here
we just present some simple examples that allow to complete the global T6 /Z4 construction. Indeed, let us choose 1 = 2 = 3 = 4 = i and impose A, B to be real numbers.
Conditions (24) reduce to

AB
A + B = 4n,
N
Z
(28)
n, n Z,
min
4 2
A B = 4n ,
where Nmin is the minimum amount of flux that has to be turned on, in order to satisfy
the quantization conditions of fractional cycles on Z4 . In our case Nmin = 4, where we
are also taking into account the subtleties mentioned above regarding toroidal orientifolds
(see Appendix A for more details). Finally, A = A/(4 2 ), B = B/(4 2 ), are flux
components conveniently normalized.
The D3-brane charge12 of such G-flux will be given by

Qflux = 4 A 2 + B 2 = 32 n2 + n 2 .
(29)
This negative D3-brane charge will help cancelling the untwisted D3-brane RR tadpole,
which is the only RR tadpole left to be cancelled in our global construction. Indeed, the
amount of positive D3-brane charge carried by the D3-branes and O3-planes in our model
is given by 16 + 4R + 32(b 1), and has to be cancelled by the contribution (29). This
amounts to

8 n2 + n 2 R = 4 + 8(b 1),
(30)
12 In our conventions each D3-brane in the covering space has charge +1. Hence in Eq. (26) M refers to the
6
covering space for the orientifold, i.e., T6 .
36
Table 5
Open string spectrum near the origin of the global T6 /Z4 model (visible sector), and on the fixed point at
(0, 0, (1 + i)/2) (hidden sector) for the choice of discrete B-field b = 1/2
Visible sector
Gauge group
D3D3
SU(4) SU(2) SU(2)
Hidden sector
Gauge group
Chiral multiplets
D3 D3
SU(R) SO(R + 2)2
D73 D73
SU(4) USp(4)2
1, R + 2)
2(R, R + 2, 1) + 2(R,
(1, R + 2, R + 2) + AR + A R
1, 4) + (1, 4, 4) + 10 + 10
2(4, 4, 1) + 2(4,
D3 D73
D73 D3
Chiral multiplets
1, 2)
2(4, 2, 1) + 2(4,
(1, 2, 2) + (10, 1, 1) + (10, 1, 1)
1, 1; 1, 1, 4)
(R, 1, 1; 1, 4, 1) + (R,
1, 1)
(1, R + 2, 1; 4, 1, 1) + (1, 1, R + 2; 4,
which can be satisfied for an infinite numbers of possibilities, some of them being b = 0,
n = 1, n = 1, R = 4, or b = 1/2, n = 1, n = 0, R = 0, etc.
2.2.2. Final spectrum
Let us summarize our results by writing down the final spectrum of our PatiSalam-like
model in Table 5. For definiteness, here we pick the particular choice of discrete B-field
b = 1/2. The visible part of the spectrum will concerns the D3-branes at the origin of
T6 /Z4 , just as in the local construction presented before. The extra D7-branes and D3branes introduced in order to cancel twisted k = 2 and untwisted tadpoles will give rise
to extra gauge groups and chiral fermions. However, since they are located in a different
orientifold fixed point, they will not introduce any extra massless particle charged under
the initial PatiSalam, and thus can be seen as a hidden sector of the theory.
Notice that the D-brane sector is totally supersymmetric, and so is the spectrum before
taking into account the effect of the flux. Indeed, the flux is the only source of SUSYbreaking, in contrast with the models in [36,37] which contained anti-D-branes in order to
cancel RR tadpoles. As a result, there are no unbalanced forces between pairs of D-branes,
which may induce effective potentials for the D-brane position moduli and extra sources
of SUSY breaking in the visible sector.13 Furthermore, the scalar potential generated by
the flux for the complex structure moduli fields is at its minimum, since we are turning on
IASD components only. Indeed, the only instability of the system comes from the excess
of D-brane tension, more specifically 2 (32 + 4R) units of D3-brane tension, which generates an NSNS tadpole typical of non-supersymmetric systems. Notice that this kind of
tadpole is the one present in the recent proposals for obtaining de Sitter vacua from string
theory [11,41], before all moduli have been stabilized. As discussed in Appendix B, the
precise effect of this tadpole strongly depends on the warp factors of the flux compactification.
13 In [36,37], there are anti-branes which are stuck at different fixed points. Therefore, the branes and anti-
branes do not annihilate even though there is an attractive force between them. However, this attractive force will
generate a potential for the Khler moduli of the orbifold.
37
3. Phenomenology
In the present section we study some phenomenological features of the flux-induced
SUSY breaking scenario. In particular, we focus on the T6 /Z4 orientifold model constructed in Section 2. Our main concern is the effective theory in the visible sector of
Table 5, i.e., the D3-brane system located at the origin of the orientifold. Notice that, in
the global non-supersymmetric model that we have built, the only explicit source of supersymmetry breaking is given by the IASD G3 flux. Hence, in principle we should not bother
about extra sources of SUSY breaking as, e.g., distant anti-D-branes.
In order to perform the phenomenological analysis, we first briefly review the results
in [6,7]. In particular, we will focus on the explicit expressions for scalar masses and soft
terms derived in [6], and then specialize to the case at hand. The soft-term expressions, as
well as the flux quantization conditions, allow us to address a series of phenomenological
issues. Some of these issues depend on the local details around the D3-brane location,
while some other on the global features of the configuration. As an example of the latter,
we consider the distribution of models in the space of well-quantized fluxes that allow for
a correct electroweak symmetry breaking pattern.
3.1. Soft Lagrangian for flux-induced supersymmetry breaking
For completeness and to set our notation, here we briefly review the results of [6,7]
regarding the soft Lagrangian for the flux-induced supersymmetry breaking scenario. In
particular, we follow the notation and expressions of [6]. We refer the readers to the original
papers for the derivation of these results. In the standard notation of the soft supersymmetry
Lagrangian (see, e.g., [42,43]), the dimensionful couplings on the worldvolume of a D3brane, and which are induced by the supergravity background around it are given by
m2i j = 2Ki j i j + gs (Im )i j ,
Bij = 2Kij ij + gs (Im )ij ,
1/2
gs
Aij k = hij k G123 ,
2

gs
C ij k = +hij l Slk (Alk ) ,
2 2
1/2
1/2
gs
M a = G123 ,
2
1/2
gs
ij = Sij ,
2 2
1/2
gs
Mgia = i jk Ajk ,
4 2
(31)
38
where the symmetric and anti-symmetric tensors Sij and Aij are constructed from the
3-form flux G3 as
1
Sij = (ikl Gj k l + j kl Gi k l),
2
1
Aij = (ik lGkl j jk lGkl i ),
2
and likewise for Sij and Aij . The tensors K, , and come from the power expansion of
the metric, dilaton and five-form flux, respectively, around the location of the D3-branes.
As such, they are constrained by the equations of motion of these fields. Combining these
constraints, one obtains the following sum-rule for the scalar masses [6]:
m21 + m22 + m23

gs
1
|G123 |2 +
=
|Sij |2 + |Aij |2
2
4
ij

1
1
Re G123 G1 2 3 + Slk Slk + Alk Alk .
4
4
(32)
Since the matter fields are localized on the brane which preserves N = 1 supersymmetry, their interactions presented above should also be understandable in terms of N = 1,
D = 4 supergravity. From this perspective, the fluxes could be viewed as sources of various
auxiliary fields in the supergravity Lagrangian. In particular:
The G123 component of the flux is proportional to the auxiliary component of the
dilaton,
rise to gaugino masses and trilinear couplings through the operators
2 FS . This gives
d SW W , and d 2 Si j k , respectively.
On the other hand, the flux Ak l is part of the D term of a vector supermultiplet.
The coupling Mgia has its origin in the operator d 2 f (S, )W W , where f is the gauge
kinetic function. From this term we obtain Mgia i f (S, )D. Comparing this with the
expression of Mgia in terms of flux, we can identify D i jk Ajk . There is also a term in
the scalar potential 1/2(Df (S, )D), where D i tij j . Expanding the gauge kinetic
function and using the identification we have just made, we obtain the term proportional to
Ajk in C ij k .
Turning on flux Sij does not break N = 1 supersymmetry on the D3 brane. This
is evident at the Lagrangian level. The effect of this flux is to generate a -term in the
superpotential. The |Fi |2 term derived from the superpotential is then responsible for the
term proportional to Sij in the coupling C ij k . Hence, C ij k contains both a supersymmetry
preserving part and a supersymmetry breaking part (proportional to the flux component
Ajk ).
Both Mgia and the supersymmetry breaking part of C ij k are somewhat atypical supersymmetry breaking terms. On the one hand, the terms C ij k are considered to be soft only
if there is no chiral multiplet which is a singlet of the gauge group. This is a condition that
the Minimal Supersymmetric Standard Model (MSSM) actually satisfies. However, those
terms are usually not included as part of the soft supersymmetry breaking Lagrangian of the
39
MSSM. On the other hand, gauge symmetry requires the existence of adjoint light matter
fields for a non-vanishing Mgia . Therefore, including Mgia in the soft Lagrangian generically
requires one to go beyond the matter content of the MSSM and postulate the existence of
some adjoint matter fields. In the case of flux-induced couplings, we see that the supersymmetry breaking part of C ij k and Mgia are proportional to the same flux Ajk . Therefore,
the SUSY-breaking part of C ij k would also vanish when we have only the MSSM matter
content.
Let us now focus on the soft terms present on a D3-brane placed on top of a C3 /Z4
singularity. The non-zero components of the G3 flux allowed by the orbifold symmetries
are G123 and G1 23
:
1
2
3
G3 = G123 dz1 dz2 dz3 + G1 23
d z d z dz .
(33)
Therefore, Aij = Aij = Sij = 0 and Sij = 0, except S33 = 2G1 23

. As a result, the scalar
masses are

gs
2
|G123 |2 + |G1 23
m21 + m22 + m23 =
(34)
| .
2
It is important to distinguish between scalar masses and soft masses. In Eq. (34), there
2
2
are two contributions to the scalar masses. One of them is proportional to |G1 23
| || ,
which is the supersymmetric mass term. The soft scalar masses contribution m0 , which
breaks supersymmetry and does not arise from the superpotential, corresponds to the piece
proportional to |G123 |2 .
The other soft terms are
1/2
gs
Aij k = hij k G123 ,
2
1/2
gs
M a = G123 .
2
(35)
Finally, we also have a non-vanishing -term

1/2
gs
ij = G1 23
i3 j 3 .
2
(36)
Notice that the spectrum presented in Eqs. (34) and (35), assuming universal soft
masses, is identical to the spectrum of dilaton dominated SUSY breaking scenario with
vanishing cosmological constant [44]. The soft
parameters in this scenario obey the simple relations Aij k = hij k M a and |M a | = 3m0 . This is not a surprise at all since the
supersymmetry breaking G123 flux can be understood as the source of the dilaton F -term
FS . In this sense, the soft term pattern of our string construction, as well as its associated
low energy physics, can be handled by means of a D = 4 effective Lagrangian approach.
Notice, however, that there is a remarkable fact of the flux-induced SUSY breaking scenario. Namely, we have an explicit source of supersymmetry breaking (in this case FS )
which is determined by the background flux instead of being a phenomenological input.
In the following we will exploit this fact, with the aim of learning some lessons from the
40
flux-induced SUSY-breaking scenario. As we will see, although we perform our analysis

on the particular T6 /Z4 model constructed in the previous section, the final conclusions
are rather model-independent.
3.2. Electroweak symmetry breaking
An important feature of the model constructed in Section 2 is that the background flux
G3 provides a source for both the soft parameters and the -term. As discussed below,
these essential quantities in electroweak symmetry breaking (EWSB) generically need to
be fine-tuned, at least at the level of one percent, in order to achieve a phenomenologically
viable electroweak symmetry breaking. It is thus interesting to see if the fact that these
parameters have the same microscopic origin (i.e., the flux) can give us some insight into
this fine-tuning problem involved in EWSB.
Our approach is different in spirit from the conventional studies of EWSB in a SUSYbreaking scenario. Typically, a particular SUSY-breaking scenario only provides a mechanism of generating soft SUSY-breaking terms. The -term is then treated as a free parameter, which is then fixed by requiring a correct electroweak symmetry breaking at low
energies. Now, when such SUSY-breaking scenario is described by an actual microscopic
theory, one may wonder if there are some constraints on the otherwise free parameters,
such as the -term. In the particular case of flux-induced SUSY breaking this turns out to
be the case.14 For instance, as explained in Section 2 the components of the flux G3 must
be properly quantized in terms of the internal geometry of M6 . This implies that, although
and M a are given by different components of the flux, in an actual model they are related
and constrained by the flux quantization conditions.
In the model at hand, a 3-form flux G3 has to satisfy eight different quantization conditions, which are presented in Appendix A. Notice that, in general, G3 is complex-valued,
and hence it will give rise to complex soft term in our model. In an scenario where the
superpartner masses are of the order of TeV scale, however, low energy measurements on
CP violating observables such as the electric dipole moment [46] generically constrain the
supersymmetry breaking parameters to be real. In order to satisfy this phenomenological
constraint, we will consider well-quantized fluxes G3 which have only real components.
This may seem to be a rather strong assumption in the space of fluxes. However, in our
analysis we will be mainly interested in how the conditions required for electroweak symmetry breaking are realized in this scenario. As will become obvious later, the mechanism
we discuss will only depend on the fact that the fluxes are quantized, and not on them being
real. Hence, it would also work in a more general scenario.15
With the above assumptions, the quantization conditions then reduce to Eq. (A.6) in
Appendix A, which in turn imply

1 m
1 m
Msoft ,
Msoft ,
|G1 23
|G123 | = 1 +
(37)
|= 1
gs n
gs n
14 Similar relations between the -term and the soft parameters also appear in the context of ScherkSchwarz
supersymmetry breaking in 5D orbifold models [45]. We thank Kiwoon Choi for pointing these references to us.
15 As, e.g., the one recently suggested in [47].
41
for some integers n and m. Here we have made

use of the fact that Im 4 = 1/gs , and we
have factored out the scale Msoft = 8n 2 / Vol(T6 /Z4 ). Notice that the values that the
fluxes can take are discretized, since gs is also fixed by the flux.
Electroweak symmetry breaking will be induced, as usual, by the radiative corrections
to the soft Higgs masses. There are two low energy outputs of the electroweak symmetry
breaking. They are v 2 = vu2 + vd2 , or equivalently m2Z , and tan = vu /vd . The only value
that is currently measured is m2Z . The value of tan (m2Hu + m2Hd + 2||2 )/B is determined by B which in our model depends on Kij , ij and ij , and so it is not constrained
by the local equations of motion derived in Ref. [6]. Therefore, we will take tan as a free
parameter at this stage.16
Roughly speaking, the condition for a correct electroweak symmetry breaking is that it
reproduces the measured Z-boson mass. As we commented above, since we have a prediction for , the condition becomes non-trivial. We will devote the rest of this section
studying how this condition can be satisfied. Including the effect of RGE running, the requirement can be written as [48]
m2Z C ||2 + C1/2 m21/2 + C0 m20 + CA A2t + CA At m1/2 ,
(38)
where m1/2 and At are the universal gaugino mass and trilinear coupling, respectively. The
numerical coefficients Ci s arise from integrating the renormalization equations for the soft
parameters and and minimizing the Higgs potential. As a result, they depend on several
other parameters, such as top Yukawa coupling (or mtop ), tan , and UV which is the scale
at which supersymmetry is broken. In our setup, string UV 1011 GeV [6]. For the
sake of concreteness, let us pick the MSSM values which are deduced from this UV and
tan = 5. The coefficients then read [48]
C 1.7,
C1/2 2.6,
C0 0.3,
CA 0.2,
CA 0.6. (39)
The general statement about the required fine-tuning for electroweak symmetry breaking is obvious here. Indeed, we could characterize the size of the soft parameters by some
scale Msoft . On the one hand, non-observation of superpartners at LEP-II puts various lower
bounds on Msoft already. Moreover, in the MSSM, a heavier Higgs mass which satisfies the
LEP lower bound [49] would almost necessarily require at least some of the soft parameters
to be several times larger than MZ [50]. On the other hand, the level of cancellation which
is required to achieve electroweak symmetry breaking increases quadratically with Msoft ,
which generically requires a 1% cancellation. More precisely we could use, as a measure
of the level of fine-tuning, the variation of the Z-boson mass resulting from a change in the
soft masses and the -term
2
m2i
mi
m2Z
.
=
C
(40)
i
m2Z
m2Z
m2i
16 In a global model such as the one presented here, generically there will be constraints on B from the equa-
tions of motion. Although tan is an important parameter for various aspects of supersymmetry phenomenology
[43], its precise value does not affect the tuning involved in electroweak symmetry breaking significantly. Therefore, for our purpose it is good enough to assume that tan is a free parameter. It would be interesting to study
the constraints on tan from the equations of motion in some explicit global models.
42
By requiring m2Z /m2Z 1, we obtain m2i /m2i m2Z /(Ci m2i ) 1% if mi TeV. This
shows the 1% tuning of the electroweak symmetry breaking in the MSSM, as well as the
quadratic grow of this fine-tuning.
Taking these facts seriously, there is a somewhat significant tension between electroweak symmetry breaking and the experimental Higgs mass bound within the framework
of the MSSM. However, it is important to keep in mind the assumptions underlying such
an argument. First, the above reasoning is based on the tuning of the soft parameters of an
effective theorythe soft SUSY Lagrangian. Given a microscopic theory of supersymmetry breaking, the soft terms are determined by fundamental quantities of this underlying
theory. A priori, this effective fine-tuning could look different when expressed in terms of
these fundamental parameters. In this case, the quantity m2i /m2i may not accurately characterize the likelihood of finding a correct EWSB solution. Second, we are supposing that
the -term, which is an important part of the Higgs mass, is a free parameter that can be
tuned.
For instance, the above assumptions may not apply when the soft parameters and the
-term are quantized. In fact, one of the main results of this section is to show that at least
in the class of string models with flux induced supersymmetry breaking that we studied,
the breaking of electroweak symmetry does not necessarily suffer from a quadratic finetuning. In some cases, the degree of fine-tuning may even decrease with a higher scale of
Msoft .
Let us be more explicit. Using Eqs. (34)(36), we have for this model

2
,
m2Z = 1.72 + 3.5 Msoft
(41)
where is the ratio of two components of the IASD flux

G1 23
1 gs n/m
.
=
G123
1 + gs n/m
(42)
2 m2 would require us to adjust (1.72 + 3.5) 0 in

Increasing soft masses Msoft
Z
order to achieve electroweak symmetry breaking. Therefore, n/m < 0. It also means that
for Msoft large, is almost fixed. Since we have the freedom of adjusting at will by
adjusting the ratio n/m, we expect to find models satisfying the requirement of Eq. (41)
for various values of Msoft .
We would like to assess further the level of difficulty in obtaining string vacua leading to
phenomenologically viable EWSB as we vary Msoft . To be precise, we measure the degree
of fine-tuning by the number of string vacua in which Eq. (41) is satisfied.17 In particular,
we are interested in the distribution of solutions as a function of Msoft . These results can
then be compared with the standard expectation of a quadratic fine-tuning, i.e., the number
2
of solutions going as Msoft
.
17 This stringy naturalness point of view on the fine-tuning problem is similar in spirit to [5153] on the scale
of supersymmetry breaking in the string theory landscape. For some other applications of the stringy naturalness
principle, see [5458].
43
Let us count the number of vacua with flux quanta n and m which satisfy approximately18 Eq. (41). It is clear from the definition that depends on (n, m). The functional
dependence of Msoft on the flux quanta is slightly more subtle. First, note that the value of
Msoft depends on the 3-form flux density, which in turn depends on the flux quantum n and
the volume of the corresponding 3-cycle. For our model at hand,

Msoft = 8n 2 / Vol T6 /Z4 nS .
(43)
In addition to the explicit n dependence, S will be in general a function of n. The precise
functional form of S depends on the mechanism which stabilizes the Khler moduli. For
instance, in the KKLT [11] scenario, the runaway behavior of the overall Khler modulus
due to uncanceled NSNS tadpoles is counteracted by non-perturbative effects, thus stabilizing the moduli at a finite value. Increasing the value of n would increase the NSNS
tadpoles and so we expect the stabilized value of Vol(T6 /Z4 ) to increase with n. In this
case, should be a decreasing function of n. More generally, let us parametrize the unknown mechanism that stabilizes the Khler moduli by writing
S n1/1 mX .
(44)
In other words,
Msoft = nS = n1/ mX ,
where mX is a n-independent mass scale. The tuning involved in EWSB is now

2

2/ 2
m2Z
2 n
2
2/ mX
log 1.7 + 3.5 n mX n n
.
=
+
n
n
m2Z
m2Z
(45)
(46)
Therefore, the degree of fine-tuning depends on the value of (hence the precise mechanism which stabilizes Khler moduli) and is not necessarily quadratic in Msoft . More
2
precisely, the leading behaviour of this fine-tuning on Msoft will be given by Msoft
. For
illustrative purpose, let us consider the following cases:
= 1. In this case, the volume modulus is stabilized at a value independent of
the flux quanta n. Although the number of flux vacua leading to a viable EWSB decreases with Msoft as depicted in Fig. 8, the tuning involved is linear rather than quadratic
in Msoft .
= 2. In this case, the number of flux vacua satisfying the EWSB condition
of Eq. (41) stays approximately constant, as shown in Fig. 9. Again, this is different
from the quadratic fine-tuning one would expect based on effective field theory arguments.
> 2. In this case, the number of solutions actually increases with increasing Msoft .
This means that it is easier to find vacua leading to a correct EWSB with a higher scale
of Msoft , which is exactly opposite to the conclusion one would have made based on the
18 To qualify as an approximate solution, the difference between the experimental value of the Z-boson mass
and the value calculated from Eq. (41), denoted as mZ , should be sufficiently smaller than mZ .
44
Fig. 8. Number of solutions as a function of Msoft if S is independent of n. The black (upper) line corresponds
to S = 10 GeV. The red (lower) line corresponds to S = 50 GeV.
conventional intuition of fine-tuning. Fig. 10 shows the likelihood of correct EWSB for
= 3.
Notice that the quadratic behaviour on Msoft is only recovered when 0. Quite
amusingly, this corresponds to the case where Vol(T6 /Z4 ) , that is, where there is no
stabilization mechanism for the Khler moduli that prevents the theory to be driven to the
decompactification limit.
It is also interesting to relate our results to the recent proposals of statistically favored
high scale supersymmetry breaking [52,53]. Suppose we have a set of supersymmetry
breaking order parameters Xi , i = 1, . . . , nX , after putting in the constraints of electroweak
symmetry breaking, the volume of parameter space is [53]

2
EW
2
Xi2
d Msoft d(Msoft ) d nX Xi Msoft
2
Msoft
nX 3
2EW Msoft
d(Msoft ),
(47)
where the -function impose the condition of all supersymmetry breaking scale. The number of states satisfy these constraints will be nsol d , where is the density of
states. For example, if =constant, there are more solutions at higher scales if nX 3.
One simple realization of this scenario involves two complex F fields as the supersymmetry breaking order parameter.
45
Fig. 9. Number of solutions as a function of Msoft for = 2, assuming mX = mZ .
In our case, we also have two auxiliary F fields. However, there are no large splittings between scalar and gaugino masses in this scenario due to the universal nature of
the dilaton coupling. Therefore, unlike the scenario proposed in Ref. [59], compatibility with gauge unification will favor a relatively low supersymmetry breaking scale
TeV. In this case, there are strong constraints on the CP violations from, e.g., EDM observables [46]. Generically, the phases of the supersymmetry breaking parameters are
constrained to be real. Therefore, effectively, nX = 2 in this scenario. On the other hand,
as commented earlier in this section, the distribution of states is not necessarily uniform but can be more densely populated at relatively higher scales. Therefore, the number of solutions would not necessarily decrease, and may even increase, with the scale
of Msoft .
It should be emphasized that this is by no means a scan of the string theory landscape
(since we are only exploring a tiny subspace of string vacua) nor do we attach any statistical interpretation to our results (which require assumptions about the precise way in which
all moduli are stabilized, and the measure on the landscape). Nevertheless, we provide an
explicit example to illustrate that the issue of fine-tuning in electroweak symmetry breaking may be viewed differently when the distribution of consistent string vacua are taken
into account. In particular, larger soft masses, at least those which is directly relevant to
electroweak symmetry, are not necessarily disfavored [52,53,59]. Moreover, as shown in
our analysis, this effect is only relying on the fact that fluxes are quantized (hence n and m
take on integer values) and is not sensitive to the explicit form of as a function of these
46
Fig. 10. Number of solutions as a function of Msoft for = 3, assuming mX = mZ .
parameters. Therefore, we expect this suppressed fine-tuning effect to be generic in a large

class of models with flux-induced electroweak symmetry breaking.19
4. Conclusions
In this paper, we have studied supersymmetry breaking effects induced on D3-branes at
singularities by the presence of NSNS and RR 3-form fluxes. We have constructed some
global models of flux compactifications as examples to illustrate the issues of model building and phenomenology involved in this scenario. Although the models presented are not
fully realistic, since they do not contain enough number of chiral families to account for
the SM flavor structure, our analysis serves as a template for future phenomenological
studies of this scenario. In any case, since the soft parameters induced by the fluxes on D3branes at singularities are universal and independent of the number of families, we expect
our conclusions on phenomenological features which are insensitive to them, such as the
fine-tuning involved in EWSB, to persist in more realistic models.
19 At the same time, we have not taken into account effects such as the next-leading order corrections, threshold
corrections, high energy input scale of RGE running, etc., on the Higgs potential. Changing tan will also change
the specific numbers used in the condition of electroweak symmetry breaking. We do not expect, however, the
suppressed fine-tuning effect to depend on the particular details of the Higgs potential.
47
In fact, the key feature of this SUSY breaking scenario that allows for a different behaviour from the usual EWSB fine-tuning is the fact that the soft parameters are quantized. In
our setup things are particularly simple, since both and soft terms come from the same
source, namely the fluxes. In general, there could be other sources for them. However, we
expect our arguments to apply as long as these parameters take on quantized values.
Other than the fine-tuning issue of electroweak symmetry breaking, we have not discussed in depth other phenomenological features of this SUSY breaking scenario. The
reason is that many phenomenological issues surrounding supersymmetry breaking, such
as flavor and CP observables, involve the family structure of the Standard Model which is
not realized in the models at hand. Clearly, it would be interesting to carry out a detailed
study of CP observables for flux vacua with realistic flavor structure. However, without a
detailed knowledge of the flavor structure, we could only anticipate that this type of models would produce signatures similar to a subclass of mSUGRA scenarios [60]. Notice
that the spectrum of this class of models is identical to the dilaton dominated supersymmetry breaking scenario. One of the generic features of this class of models is the soft
masses and trilinear couplings are universal.20 As a result, we expect the bino to be the
lightest superpartner. In addition, a naive estimate of the string scale in this class of models gives Ms 1011 GeV, and so string scale stable states (such as D-matter [62]) could
also be candidates for Wimpzillas dark matter [63]. Finally, there is no large splitting between gaugino and sfermion masses, which is very different from the split supersymmetry
scenario recently proposed in [59].
While this paper was written, Ref. [9] appeared where a Z4 orbifold model involving
D3- and D7-branes was constructed, and the flux-induced soft terms induced in such construction were computed. Although there are some similarities with the Z4 model presented
here, the constructions are not identical. It would be interesting to find a global embedding
of [9] and compare the resulting phenomenology with the global models discussed here.
Finally, in the magnetized D-brane setup, the Standard Model does not have to be localized completely at a singularity and so more realistic models can be constructed [23].
Some of the aforementioned challenges in constructing realistic flux vacua that appeared
in models including only D3-branes at singularities are a priori absent, and this seems a
promising direction for constructing realistic models with fluxes. It would be interesting
to apply the recent results of flux induced SUSY breaking terms for the D3/D7-system
computed in [8,9] to the three-generation MSSM flux vacua in [23].
Acknowledgements
We would like to thank Nima Arkani-Hamed, Vijay Balasubramanian, Alex Buchel,
Pablo G. Cmara, Kiwoon Choi, Tony Gherghetta, Luis Ibez, Gordy Kane, Hans-Peter
Nilles, and specially Lisa Everett and Angel Uranga for useful comments and discussions.
This work was supported in part by a DOE grant No. DE-FG-02-95ER40896. F.M. and
20 On the same footing, gaugino masses are also universal because the entire Standard Model is embedded on
a single stack of D-branes. This is in general not the case in the magnetized D-brane setup as, e.g., the models in
[23]. For some explicit computations which illustrate this fact see [61].
48
G.S. were also supported by NSF CAREER Award No. PHY-0348093, and a Research
Innovation Award from Research Corporation. L.W. was partially supported by the Wisconsin Alumni Research Foundation. We would also like to thank the Aspen Center for
Physics for hospitality.
Appendix A. T6 /Z4 flux quantization conditions
In this appendix we collect the flux quantization conditions for the particular case of the
toroidal orientifold T6 /Z4 . In a general geometry, these quantization conditions are given
by (24). In the particular case that M6 is an orbifold, one needs to integrate the 3-form
fluxes F3 and H3 over an integral basis of 3-cycles. Now, an integral basis of 3-cycles for
T2n /ZN or T2n /ZN ZM involves fractional cycles, i.e., those which are closed in the
orbifold but not in the covering space T2n . If we are not introducing any 3-form flux in the
collapsed cycles of T2n /ZN , however, the flux quantization conditions can be recasted in
terms of the homology of T2n . More precisely, we have

1
1
(A.1)
F3 2Nmin Z,
H3 2Nmin Z,
2
2
T2n ,
where is now a 3-cycle of

and Nmin is an integer which depends on the particular
Z4
orbifold. In a Z4 orbifold this number can be seen to be Nmin
= 2.21 Considering an orientifold of the previous construction implies that we have an extra Z2 action acting on T6 ,
and hence the previous number Nmin may be multiplied by 2. This turns out to be the case
in the T6 /Z4 model discussed in the main text,22 so that at the end of the day we have to
impose conditions (A.1) with being any 3-cycle of T6 and Nmin = 4.
These conditions can be written in a more specific way by using [24]
F3 =
and
Im(4 G3 )
,
Im 4
H3 =
(A.2)

dzj = ij ,
[ai ]
Im G3
,
Im 4
dzj = i ij ,
(A.3)
[bi ]
where i, j = 1, 2, 3 and [ai ], [bi ] is the basis of 1-cycles on the ith T2 . By using the explicit
expression
G3 = A dz1 dz2 dz3 + B d z 1 d z 2 dz3
(A.4)
21 Roughly speaking, N
2n
min can be computed from dividing the volume of a 3-cycle on T by its corresponding
fractional cycle on T2n /ZN . The integral homology of T6 /Z4 can be found in, e.g., [64].
22 The precise statement goes as follows [40]. We may either have that the flux quanta along the cycle are
orbi or 2nN orbi (n Z) depending on the fact that passes over an odd or even number
of the form (2n + 1)Nmin
min
(+,+)
, respectively. In the Z4 orbifold example considered in the main text we are in the second situation,
of O3
orbi = 2N orbi . See also the comments in [65] regarding this problem.
so we can effectively take Nmin
min
49
for the 3-form flux, conditions (A.1) are finally translated into
Im(A + B)
4Z,
Im 4
3]
Im[(A + B)
4Z,
Im 4
Re(A B)
4Z,
Im 4
3]
Re[(A B)
4Z,
Im 4
4 ]
Im[(A + B)
4Z,
Im 4
3 4 ]
Im[(A + B)
4Z,
Im 4
4 ]
Re[(A B)
4Z,
Im 4
3 4 ]
Re[(A B)
4Z,
Im 4
(A.5)
where we have conveniently normalized the flux components as A = A/(4 2 ), B =

B/(4 2 ). Notice that the complex structure of the first two-tori of T6 /Z4 do not enter
into these conditions, since the orbifold geometry fixes them to be 1 = 2 = i.
These previous equations are simplified by choosing some subspace of flux quanta. For
instance, as discussed in the main text, a phenomenologically interesting subspace is given
B to be real numbers. The quantization conditions (28) then read
by taking A,
A + B = 4n1 ,
t3 = 4n2 ,
(A + B)
t4
1 = 4n3 ,
(A B)
t4
r3
(A B) = 4n4 ,
t4
r4
(A B) = 4n5 ,
t4
(r
3 r4 + t3 t4 ) = 4n6 ,
(A B)
t4
r
(t
3 4 r3 t4 ) = 4n7 ,
(A + B)
t4
(A.6)
where 3 = r3 + it3 , 4 = r4 + it4 , and ni Z. These equations in turn imply

A = 2(n1 + n3 t4 ),
B = 2(n1 n3 t4 ),
t3 n 2
=
Q,
t4 n 1
n4
r3 =
Q,
n3
n5
Q,
r4 =
n3
n6 n4 n5
2 Q
t 3 t4 =
n4
n3
(A.7)
(A.8)
50
and
n7 n3 = n2 n5 n1 n4 .
(A.9)
From Eq. (A.7) and by using t4 = 1/gs we obtain
B
1 gs n1 /n3
=
A 1 + gs n1 /n3
(A.10)
which is Eq. (42) in the main text. On the other hand, Eq. (A.8) show that the closed string
moduli 3 , 4 are stabilized and can only take discrete values which depend on the lattice of
flux quanta {ni }7i=1 . Finally, Eq. (A.9) shows that this lattice is six-dimensional, matching
the number of free parameters in our problem.
Appendix B. Fluxes and NSNS tadpoles

Given a string theory compactification, there is a series of local and global constraints
that must be fulfilled in order to yield a consistent theory. In particular, the absence
of divergences in the open string sector imposes cancellation of RamondRamond and
NeveuSchwarz tadpoles. On the one hand, the presence of RR tadpoles is related to
gauge anomalies either in the low energy effective theory or in the worldvolume of suitable
D-brane probes [28]. On the other hand, NSNS tadpoles signal an instability of the vacuum configuration and usually involve effective potentials for some moduli. Whereas RR
tadpole cancellation conditions must always be imposed [66], NSNS tadpoles may remain
uncancelled, the divergences associated with them being cured by a FischlerSusskind
mechanism [67]. Some supergravity solutions in the presence of these global tadpoles have
been analyzed in the literature [6870] and, in general, they involve a redefinition of the
background where Poincar invariance is spontaneously broken.23 In the limit where the
supergravity approximation is valid, RR and NSNS tadpole conditions can be recasted as
imposing the Bianchi identities/equations of motion for the supergravity fields. This fact
was used in [4] in order to show the existence of warped compactifications arising as the
low energy supergravity limit of certain type IIB orientifolds.
It turns out that global NSNS tadpoles are indeed present in most of the semi-realistic
chiral flux compactification constructed previously [6,24,25,65]. In addition, they are a
necessary ingredient in any model involving flux-induced SUSY breaking on D3-branes.
Finally, they would also be present in the recent proposals for obtaining de Sitter vacua
from string theory [11,41]. Of course, these tadpoles are no longer present as such once all
the moduli of the compactification have been stabilized.
In order to have an estimate of the amount of NSNS tadpoles in a given type IIB string
compactification, let us then consider type IIB supergravity with action
23 See also [71] for recent progress in understanding NSNS tadpoles.

3
F52
M M G3 G
d x g R
12 Im
4 5!
2(Im )2

3
1
C4 G3 G
+
2
Im
8i10

p+1
Tpi g + pi
Cp+1 ,
d
+
1
SIIB = 2
210
51
10
Sp i
(B.1)
Sp i
where = a + i/gs is the usual type IIB axiondilaton coupling, G3 = F3 H3 stands

for the complexified RR + NSNS 3-form flux, F5 is the self-dual five-form field strength
and Cp+1 are the RR (p + 1)-form potentials. On the second line we have included the
contribution given by localized sources, such as Dp-branes with charge and tension pi
and Tpi , respectively, and wrapping a (p + 1)-submanifold Spi of the D = 10 target space.
Now, again in the limit where the supergravity approximation is valid, the amount of
NSNS tadpole can be estimated by performing a dimensional reduction of the action (B.1).
That is, we aim to compute the excess of tension coming from a compactification of D = 10
type IIB supergravity + localized sources. This computation has been performed in [72]
for the particular case of BPS-like warped compactifications of [4]. Let us now consider
the more general situation where no specific BPS-like supergravity solution is imposed.24
We will again consider a metric and five-form ansatze similar to the ones used in [4].
Namely,
2
(4)
= e2A(y) g
dx dx + e2A(y) g mn dy m dy n
ds10
(B.2)
which is of the usual warped form, with the warp factor A depending on the internal
coordinates y m of the compact six-manifold M6 , whereas being independent of the fourdimensional coordinates x . Similarly,

F5 = (1 + ) d d Vol(4) ,
(B.3)
where is an arbitrary function of the internal coordinates y m , and d Vol(4) is the volume
(4)
form of the four-dimensional unwarped metric g . We also consider Dp-branes filling the
D = 4 non-compact dimensions, that is, Spi = M4 i , where i is a (p 3)-cycle of
3 will be given by closed 3-forms
M6 . Finally, the field-strength 3-form fluxes G3 and G
on M6 , and the complex dilaton will only depend on internal coordinates = (y).
The contribution coming from the EinsteinHilbert and five-form terms is given by

F 2
g R 5
45

2

1
1
2 d 4 x g R(4) Vw1 d 6 y ge
10A e4A + ()2 ,
2
24
1
2
210
d 10 x
24 See [73] for related issues in flux compactification.
(B.4)
52
where we are contracting indices with the warped metric. Here, as in [72], we have defined
the warped volume Vw and four-dimensional Newton constant as

Vw
1
=
,
V
d 6 y ge
4A .
(B.5)
w
2
2
24
210
In order to simplify our discussion let us consider a compactification where only sources
of D3-brane and tension are present. By considering (B.4) as well as the terms for the G3
flux and the localized sources, we can rewrite the type IIB action (B.1) as25

1
eff
SIIB
= 2 d 4 x g R(4) Veff ,
(B.6)
24
1
3

2

G3 G
d y g e8A e4A + ()2 + e2A
2
12 Im

2
+ 210
T3 e2A
ti (yi )
Veff = Vw1
1

2
|iG3 6 G3 |2
d y g e10A e4A + e2A
2
24 Im

2
+ 210
ti (yi ) 3loc i ,
T3 e2A
= Vw1
(B.7)
(B.8)
where yi are the positions of the D3-branes/O3-planes in M6 , and ti are the units of tension
of each object, normalized such that a D3-brane has t = 1 in the covering space.26
It easy to see that Veff identically vanishes if we impose the conditions

T3 ti = T3 3loc i ,
(B.10)
6 G3 = iG3 ,
(B.11)
e4A =
(B.12)
for a definite choice of sign = . Eqs. (B.10)(B.12) with the choice = + are (part of)
the supergravity BPS-like solution found in [4]. The choice = corresponds to nothing
but a BPS-like solution with the opposite choice of supersymmetry inside the N = 2 parent
theory.
As explained in [4], conditions (B.10) comes from the saturation of a BPS-like bound
on the localized sources. Eq. (B.10) with = + is satisfied by objects with same charge
and tension, such as D3-branes and usual O3-planes, whereas objects with opposite charge
25 If we were considering more general compactifications, i.e., involving D7-branes, there would be extra
contributions to the action, as the one coming from the dilaton variation on the internal dimensions.
26 In order to relate (B.7) and (B.8) we have made use of the result
e10A m e4A m ie2A

3

G3 6 G
2 e2A T
210
3loc i
3
12 Im
i
deduced from the Bianchi identity of F5 . Here stands for equality up to integration on M6 .
(B.9)
53
and tension, such as anti-D3-branes, satisfy it for = . Similarly, (B.11) can be seen
as a BPS-like condition for the 3-form flux. More precisely, the ISD condition (6 G3 =
iG3 ) corresponds to G3 carrying the charge and tension of |Qflux | D3-branes, while an
IASD flux (6 G3 = iG3 ) carries the RR and NSNS charges of |Qflux | anti-D3-branes.27
Notice that given a well-quantized 3-form flux G3 , it may attain either the ISD or the IASD
condition, but never both.28
In general, if we consider flux compactifications with BPS-like localized objects and
fluxes, satisfying (B.10) and (B.11), but with different choice of for any of them, we will
get a compactification with non-vanishing, positive Veff , which indicates the presence of a
NSNS tadpole. In the following we will take the value of Veff as a measure of this excess
of tension.
Let us consider the particular case where only sources of D3-brane and tension are
present. The expression for Veff simplifies to
1
4A 2

2 + e4A G3 G3
+ ()
d y g e8A e
2
12 Im

2
i)
+ 210
T3 e4A
ti (y
Veff = Vw1
1

2
|iG3 6 G3 |2
= Vw1 d 6 y g e8A e4A + e4A
2
24 Im

2
i) ,
+ 210
T3 e4A
(ti qi )(y

(B.13)
where yi are the positions of the D3-branes and O3-planes in M6 , and ti , qi are the units
of tension and charge of these objects, normalized such that a D3-brane has t = q = 1 in
the covering space of the orientifold. Here the delta function is defined in terms of the
unwarped metric g mn , and the same for the scalar products, so that we are extracting all
the warp factor dependence.
27 There is some confusion in the literature regarding the role of IASD fluxes in flux compactification. In
principle there is nothing wrong or inconsistent in considering a compactification with a pure IASD flux. From
the supergravity point of view, we are introducing a background which carries the same charge and tension of a
set of |Qflux | anti-D3-branes. Hence, in order to see if the equations of motion/Bianchi identities of the five-form
F5 have a solution, we should only check that the total D3-brane charge in the compact space vanishes, that is,
the associated RR tadpole cancels. For instance, in the models of Section 2, Eq. (30) illustrates a cancellation of
D3-brane RR charge between D3-branes, O3(+,+) -planes and IASD fluxes, and thus each solution represents a
consistent string theory model.
28 More precisely, a flux G with Q
3
flux > 0 will attain the ISD condition at some point in the moduli space of
complex structures and complex dilaton, whereas it can never be a pure IASD flux. The opposite holds for a flux
such that Qflux < 0. From the point of view of the effective theory, reaching the ISD or IASD condition can be
IASD |, GISD
understood as a minimization of a scalar potential, which is roughly of the form Vsc | GISD
3 G3
3
IASD
and G3
being the decomposition of G3 on its ISD and IASD parts [4]. In the particular case of a G3 flux such
that Qflux = 0, this minimum is only reached in the boundary of the moduli space of complex structures.
54
In order to illustrate the use of this formula, let us first consider a system of a D3- and
anti-D3-brane, located at yD3 and yD3 . Eq. (B.13) reads

1
4A 2

1
2 + T3 e4A(yD3 ) + e4A(yD3 ) (B.14)

Veff = Vw
+ ()
d 6 y g e8A e
2

1

2
1
6
= Vw
(B.15)
d y g e8A e4A + 2T3 e4A(yD3 )
2

1

2
= Vw1 d 6 y g e8A e4A + + 2T3 e4A(yD3 ) .
(B.16)
2
Notice that the first contribution in these expressions is always non-vanishing, since the
relation (B.12) between the 5-form and the warp factors only holds in the BPS-like compactifications of [4]. On the other hand, its contribution is suppressed by the (warped)
volume of the compact manifold. The second contribution is also suppressed by the warp
factor e4A , which is usually minimized in the location of objects of positive tension.
In the case of an anti-D3-brane in a ISD 3-form flux background G3 carrying the charge
of a single D3-brane, we have instead

1
4A 2

w
2 + T3 Nflux
+ ()
+ e4A(yD3 )
Veff = Vw1 d 6 y g e8A e
(B.17)
2

1

2
= Vw1 d 6 y g e8A e4A + 2T3 e4A(yD3 )
(B.18)
2

1

2
w
= Vw1 d 6 y g e8A e4A + + 2T3 Nflux
,
(B.19)
2
where
w
=
Nflux
1
2
(4 )2

e4A
M6
3
G3 6 G
.
2 Im
(B.20)
A similar expression applies to a set of D3-branes in presence of a IASD G3 flux. We

hence see that D3-branes + IASD fluxes and anti-D3-branes + ISD fluxes, may be on equal
footing from the point of view of NSNS tadpoles, thus Einsteins equations of motion. At
least once the backreaction of the localized objects has been taken into account. In both
cases, the contribution to the excess of tension strongly depends on the warp factor of the
compactification.
References
[1] J. Polchinski, A. Strominger, New vacua for type II string theory, Phys. Lett. B 388 (1996) 736, hepth/9510227;
K. Becker, M. Becker, M-theory on eight-manifolds, Nucl. Phys. B 477 (1996) 155, hep-th/9605053;
J. Michelson, Compactifications of type IIB strings to four dimensions with non-trivial classical potential,
Nucl. Phys. B 495 (1997) 127, hep-th/9610151;
S. Gukov, C. Vafa, E. Witten, CFTs from CalabiYau four-folds, Nucl. Phys. B 584 (2000) 69, hepth/9906070;
S. Gukov, C. Vafa, E. Witten, Nucl. Phys. B 608 (2001) 477, Erratum.
55
[2] K. Dasgupta, G. Rajesh, S. Sethi, M-theory, orientifolds and G-flux, JHEP 9908 (1999) 023, hep-th/9908088.
[3] T.R. Taylor, C. Vafa, RR flux on CalabiYau and partial supersymmetry breaking, Phys. Lett. B 474 (2000)
130, hep-th/9912152;
B.R. Greene, K. Schalm, G. Shiu, Warped compactifications in M- and F-theory, Nucl. Phys. B 584 (2000)
480, hep-th/0004103;
G. Curio, A. Klemm, D. Lst, S. Theisen, On the vacuum structure of type II string compactifications on
CalabiYau spaces with H-fluxes, Nucl. Phys. B 609 (2001) 3, hep-th/0012213.
[4] S.B. Giddings, S. Kachru, J. Polchinski, Hierarchies from fluxes in string compactifications, Phys. Rev. D 66
(2002) 106006, hep-th/0105097.
[5] M. Graa, MSSM parameters from supergravity backgrounds, Phys. Rev. D 67 (2003) 066006, hepth/0209200.
[6] P.G. Cmara, L.E. Ibez, A.M. Uranga, Flux-induced SUSY-breaking soft terms, hep-th/0311241.
[7] M. Graa, T.W. Grimm, H. Jockers, J. Louis, Soft supersymmetry breaking in CalabiYau orientifolds with
D-branes and fluxes, hep-th/0312232.
[8] D. Lust, S. Reffert, S. Stieberger, Flux-induced soft supersymmetry breaking in chiral type IIB orientifolds
with D3/D7-branes, hep-th/0406092.
[9] P.G. Cmara, L.E. Ibez, A.M. Uranga, Flux-induced SUSY-breaking soft terms on D7D3 brane systems,
hep-th/0408036.
[10] S. Kachru, M.B. Schulz, S. Trivedi, Moduli stabilization from fluxes in a simple IIB orientifold, JHEP 0310
(2003) 007, hep-th/0201028.
[11] S. Kachru, R. Kallosh, A. Linde, S.P. Trivedi, De Sitter vacua in string theory, Phys. Rev. D 68 (2003)
046005, hep-th/0301240.
[12] S. Kachru, R. Kallosh, A. Linde, J. Maldacena, L. McAllister, S.P. Trivedi, Towards inflation in string theory,
JCAP 0310 (2003) 013, hep-th/0308055.
[13] J.P. Hsu, R. Kallosh, S. Prokushkin, On brane inflation with volume stabilization, JCAP 0312 (2003) 009,
hep-th/0311077;
H. Firouzjahi, S.H.H. Tye, Closer towards inflation in string theory, Phys. Lett. B 584 (2004) 147, hepth/0312020;
C.P. Burgess, J.M. Cline, H. Stoica, F. Quevedo, Inflation in realistic D-brane models, hep-th/0403119;
M. Berg, M. Haack, B. Kors, Loop corrections to volume moduli and inflation in string theory, hepth/0404087;
J.J. Blanco-Pillado, et al., Racetrack inflation, hep-th/0406230.
[14] G. Shiu, S.H.H. Tye, Some aspects of brane inflation, Phys. Lett. B 516 (2001) 421, hep-th/0106274.
[15] For attempts to construct brane inflationary models without a flat potential by considering velocitydependent effects [14], see, e.g., E. Silverstein, D. Tong, Scalar speed limits and cosmology: acceleration
from D-acceleration, hep-th/0310221;
M. Alishahiha, E. Silverstein, D. Tong, DBI in the sky, hep-th/0404084;
X.G. Chen, Multi-throat brane inflation, hep-th/0408084.
[16] M.R. Douglas, G.W. Moore, D-branes, quivers, and ALE instantons, hep-th/9603167;
C.V. Johnson, R.C. Myers, Aspects of type IIB theory on ALE spaces, Phys. Rev. D 55 (1997) 6382, hepth/9610140;
M.R. Douglas, B.R. Greene, D.R. Morrison, Orbifold resolution by D-branes, Nucl. Phys. B 506 (1997) 84,
hep-th/9704151.
[17] C. Bachas, A way to break supersymmetry, hep-th/9503030.
[18] R. Blumenhagen, L. Grlich, B. Krs, D. Lst, Noncommutative compactifications of type I strings on tori
with magnetic background flux, JHEP 0010 (2000) 006, hep-th/0007024.
[19] C. Angelantonj, I. Antoniadis, E. Dudas, A. Sagnotti, Type-I strings on magnetised orbifolds and brane
transmutation, Phys. Lett. B 489 (2000) 223, hep-th/0007090.
[20] D. Cremades, L.E. Ibez, F. Marchesano, Computing Yukawa couplings from magnetized extra dimensions,
JHEP 0405 (2004) 079, hep-th/0404229.
[21] M. Berkooz, M.R. Douglas, R.G. Leigh, Branes intersecting at angles, Nucl. Phys. B 480 (1996) 265, hepth/9606139.
[22] G. Aldazbal, S. Franco, L.E. Ibez, R. Rabadn, A.M. Uranga, D = 4 chiral string compactifications from
intersecting branes, J. Math. Phys. 42 (2001) 3103, hep-th/0011073.
56
[23] F. Marchesano, G. Shiu, MSSM vacua from flux compactifications, hep-th/0408059;

F. Marchesano, G. Shiu, Building MSSM flux vacua, hep-th/0409132.
[24] J.F.G. Cascales, A.M. Uranga, Chiral 4D N = 1 string vacua with D-branes and NSNS and RR fluxes,
JHEP 0305 (2003) 011, hep-th/0303024;
J.F.G. Cascales, A.M. Uranga, Chiral 4d string vacua with D-branes and moduli stabilization, hepth/0311250.
[25] J.F.G. Cascales, M.P. Garca del Moral, F. Quevedo, A.M. Uranga, Realistic D-brane models on warped
throats: fluxes, hierarchies and moduli stabilization, JHEP 0402 (2004) 031, hep-th/0312051.
[26] G. Aldazbal, L.E. Ibez, F. Quevedo, A.M. Uranga, D-branes at singularities: a bottomup approach to
the string embedding of the standard model, JHEP 0008 (2000) 002, hep-th/0005067.
[27] D. Berenstein, V. Jejjala, R.G. Leigh, The standard model on a D-brane, Phys. Rev. Lett. 88 (2002) 071602,
hep-ph/0105042;
L.F. Alday, G. Aldazbal, In quest of just the standard model on D-branes at a singularity, JHEP 0205
(2002) 022, hep-th/0203129.
[28] G. Aldazbal, D. Badagnani, L.E. Ibez, A.M. Uranga, Tadpole versus anomaly cancellation in D = 4, 6
compact IIB orientifolds, JHEP 9906 (1999) 031, hep-th/9904071;
A.M. Uranga, D-brane probes, RR tadpole cancellation and K-theory charge, Nucl. Phys. B 598 (2001) 225,
hep-th/0011048.
[29] A.M. Uranga, From quiver diagrams to particle physics, hep-th/0007173.
[30] L.E. Ibez, F. Marchesano, R. Rabadn, Getting just the standard model at intersecting branes, JHEP 0111
(2001) 002, hep-th/0105155.
[31] J. Polchinski, Tensors from K3 orientifolds, Phys. Rev. D 55 (1997) 6423, hep-th/9606165.
[32] R. Rabadn, Orientifolds, PhD thesis, Universidad Autnoma de Madrid.
[33] G. Aldazbal, A. Font, L.E. Ibez, G. Violero, D = 4, N = 1, type IIB orientifolds, Nucl. Phys. B 536
(1998) 29, hep-th/9804026.
[34] R. Rabadn, A.M. Uranga, Type IIB orientifolds without untwisted tadpoles, and non-BPS D-branes,
JHEP 0101 (2001) 029, hep-th/0009135.
[35] L.E. Ibez, R. Rabadn, A.M. Uranga, Anomalous U (1)s in type I and type IIB D = 4, N = 1 string
vacua, Nucl. Phys. B 542 (1999) 112, hep-th/9808139.
[36] G. Aldazbal, A.M. Uranga, Tachyon-free non-supersymmetric type IIB orientifolds via braneantibrane
systems, JHEP 9910 (1999) 024, hep-th/9908072.
[37] C. Angelantonj, I. Antoniadis, G. DAppollonio, E. Dudas, A. Sagnotti, Type I vacua with brane supersymmetry breaking, Nucl. Phys. B 572 (2000) 36, hep-th/9911081.
[38] E. Witten, Toroidal compactification without vector structure, JHEP 9802 (1998) 006, hep-th/9712028.
[39] M. Bianchi, G. Pradisi, A. Sagnotti, Toroidal compactification and symmetry breaking in open string theories, Nucl. Phys. B 376 (1992) 365;
M. Bianchi, A note on toroidal compactifications of the type I superstring and other superstring vacuum
configurations with 16 supercharges, Nucl. Phys. B 528 (1998) 73, hep-th/9711201;
Z. Kakushadze, G. Shiu, S.H.H. Tye, Type IIB orientifolds with NSNS antisymmetric tensor backgrounds,
Phys. Rev. D 58 (1998) 086001, hep-th/9803141.
[40] A.R. Frey, J. Polchinski, N = 3 warped compactifications, Phys. Rev. D 65 (2002) 126009, hep-th/0201029.
[41] A. Saltman, E. Silverstein, The scaling of the no-scale potential and de Sitter model building, hepth/0402135.
[42] A. Brignole, L.E. Ibez, C. Muoz, Soft supersymmetry-breaking terms from supergravity and superstring
models, hep-ph/9707209.
[43] For a recent review of soft SUSY-breaking terms, see D.J.H. Chung, L.L. Everett, G.L. Kane, S.F.
King, J. Lykken, L.T. Wang, The soft supersymmetry-breaking Lagrangian: theory and applications, hepph/0312378.
[44] V.S. Kaplunovsky, J. Louis, Model independent analysis of soft terms in effective supergravity and in string
theory, Phys. Lett. B 306 (1993) 269, hep-th/9303040;
A. Brignole, L.E. Ibez, C. Muoz, Towards a theory of soft terms for the supersymmetric Standard Model,
Nucl. Phys. B 422 (1994) 125, hep-ph/9308271;
A. Brignole, L.E. Ibez, C. Muoz, Nucl. Phys. B 436 (1995) 747, Erratum.
57
[45] H. Abe, K. Choi, K.S. Jeong, K.I. Okumura, ScherkSchwarz supersymmetry breaking for quasi-localized
matter fields and supersymmetry flavor violation, hep-ph/0407005;
K.W. Choi, N.y. Haba, K.S. Jeong, K.i. Okumura, Y. Shimizu, M. Yamaguchi, Electroweak symmetry breaking in supersymmetric gaugeHiggs unification models, JHEP 0402 (2004) 037, hep-ph/0312178.
[46] K. Abdullah, C. Carlberg, E.D. Commins, H. Gould, S.B. Ross, A new experimental limit on the electron
electric dipole moment, Phys. Rev. Lett. 65 (1990) 2347;
E.D. Commins, S.B. Ross, D. DeMille, B.C. Regan, Improved experimental limit on the electric dipole
moment of the electron, Phys. Rev. A 50 (1994) 2960;
M.V. Romalis, W.C. Griffith, E.N. Fortson, A new limit on the permanent electric dipole moment of Hg199 ,
Phys. Rev. Lett. 86 (2001) 2505, hep-ex/0012001.
[47] L.E. Ibez, The fluxed MSSM, hep-ph/0408064.
[48] G.L. Kane, S.F. King, Naturalness implications of LEP results, Phys. Lett. B 451 (1999) 113, hepph/9810374;
M. Bastero-Gil, G.L. Kane, S.F. King, Fine-tuning constraints on supergravity models, Phys. Lett. B 474
(2000) 103, hep-ph/9910506;
G.L. Kane, J. Lykken, B.D. Nelson, L.T. Wang, Re-examination of electroweak symmetry breaking in supersymmetry and implications for light superpartners, Phys. Lett. B 551 (2003) 146, hep-ph/0207168.
[49] LEP Higgs Working Group Collaboration, Searches for the neutral Higgs bosons of the MSSM: preliminary
combined results using LEP data collected at energies up to 209-GeV, hep-ex/0107030.
[50] For a recent review, see M. Carena, H.E. Haber, Higgs boson theory and phenomenology, V, Prog. Part.
Nucl. Phys. 50 (2003) 63, hep-ph/0208209.
[51] R. Bousso, J. Polchinski, Quantization of four-form fluxes and dynamical neutralization of the cosmological
constant, JHEP 0006 (2000) 006, hep-th/0004134.
[52] L. Susskind, Supersymmetry breaking in the anthropic landscape, hep-th/0405189.
[53] M.R. Douglas, Statistical analysis of the supersymmetry breaking scale, hep-th/0405279.
[54] M. Dine, E. Gorbatov, S. Thomas, Low energy supersymmetry from the landscape, hep-th/0407043.
[55] E. Silverstein, Counter-intuition and scalar masses, hep-th/0407202.
[56] J.P. Conlon, F. Quevedo, On the explicit construction and statistics of CalabiYau flux vacua, hepth/0409215.
[57] J. Kumar, J.D. Wells, Landscape cartography: a coarse survey of gauge group rank and stabilization of the
proton, hep-th/0409218.
[58] O. DeWolfe, A. Giryavets, S. Kachru, W. Taylor, Enumerating flux vacua with enhanced symmetries, hepth/0411061.
[59] N. Arkani-Hamed, S. Dimopoulos, Supersymmetric unification without low energy supersymmetry and signatures for fine-tuning at the LHC, hep-th/0405159.
[60] For recent studies, see, for example, H. Baer, C. Balazs, A. Belyaev, T. Krupovnickas, X. Tata, Updated reach
of the CERN LHC and constraints from relic density, b s and a in the mSUGRA model, JHEP 0306
(2003) 054, hep-ph/0304303.
[61] D. Lst, S. Reffert, S. Stieberger, MSSM with soft SUSY breaking terms from D7-branes with fluxes, hepth/0410074.
[62] G. Shiu, L.T. Wang, D-matter, hep-ph/0311228.
[63] S. Chang, C. Coriano, A.E. Faraggi, New dark matter candidates motivated from superstring derived unification, Phys. Lett. B 397 (1997) 76, hep-ph/9603272;
D.J.H. Chung, E.W. Kolb, A. Riotto, Superheavy dark matter, Phys. Rev. D 59 (1999) 023501, hepph/9802238;
D.J.H. Chung, E.W. Kolb, A. Riotto, Nonthermal supermassive dark matter, Phys. Rev. Lett. 81 (1998) 4048,
hep-ph/9805473;
D.J.H. Chung, E.W. Kolb, A. Riotto, Production of massive particles during reheating, Phys. Rev. D 60
(1999) 063504, hep-ph/9809453;
E.W. Kolb, D.J.H. Chung, A. Riotto, WIMPzillas!, hep-ph/9810361;
D.J.H. Chung, P. Crotty, E.W. Kolb, A. Riotto, On the gravitational production of superheavy dark matter,
Phys. Rev. D 64 (2001) 043503, hep-ph/0104100.
[64] R. Blumenhagen, L. Grlich, T. Ott, Supersymmetric intersecting branes on the type IIA T 6 /Z(4) orientifold, JHEP 0301 (2003) 021, hep-th/0211059.
58
[65] R. Blumenhagen, D. Lst, T.R. Taylor, Moduli stabilization in chiral type IIB orientifold models with fluxes,
Nucl. Phys. B 663 (2003) 319, hep-th/0303016.
[66] J. Polchinski, Y. Cai, Consistency of open superstring theories, Nucl. Phys. B 296 (1988) 91.
[67] W. Fischler, L. Susskind, Dilaton tadpoles, string condensates and scale invariance, 1, Phys. Lett. B 171
(1986) 383;
W. Fischler, L. Susskind, Dilaton tadpoles, string condensates and scale invariance, 2, Phys. Lett. B 173
(1986) 262.
[68] E. Dudas, J. Mourad, Brane solutions in strings with broken supersymmetry and dilaton tadpoles, Phys. Lett.
B 486 (2000) 172, hep-th/0004165.
[69] R. Blumenhagen, A. Font, Dilaton tadpoles, warped geometries and large extra dimensions for nonsupersymmetric strings, Nucl. Phys. B 599 (2001) 241, hep-th/0011269.
[70] R. Rabadn, F. Zamora, Dilaton tadpoles and D-brane interactions in compact spaces, JHEP 0212 (2002)
052, hep-th/0207178.
[71] E. Dudas, G. Pradisi, M. Nicolosi, A. Sagnotti, On tadpoles and vacuum redefinitions in string theory, hepth/0410101.
[72] O. DeWolfe, S.B. Giddings, Scales and hierarchies in warped compactifications and brane worlds, Phys.
Rev. D 67 (2003) 066008, hep-th/0208123.
[73] S.P. de Alwis, On potentials from fluxes, Phys. Rev. D 68 (2003) 126001, hep-th/0307084;
R. Brustein, S.P. de Alwis, Moduli potentials in string compactifications with fluxes: mapping the discretuum, Phys. Rev. D 69 (2004) 126006, hep-th/0402088;
S.P. de Alwis, Brane worlds in 5D and warped compactifications in IIB, Phys. Lett. B 603 (2004) 230,
hep-th/0407126.
Non-supersymmetric loop amplitudes

and MHV vertices
James Bedford, Andreas Brandhuber, Bill Spence,
Gabriele Travaglini
Department of Physics, Queen Mary, University of London, Mile End Road, London E1 4NS, United Kingdom
Received 21 December 2004; accepted 19 January 2005
Abstract
We show how the MHV diagram description of YangMills theories can be used to study
non-supersymmetric loop amplitudes. In particular, we derive a compact expression for the cutconstructible part of the general one-loop MHV multi-gluon scattering amplitude in pure YangMills
theory. We show that in special cases this expression reduces to known amplitudesthe amplitude
with adjacent negative-helicity gluons, and the five gluon non-adjacent amplitude. Finally, we briefly
discuss the twistor space interpretation of our result.
PACS: 11.30.Pb; 11.15.-q; 12.38.Bx
1. Introduction
Since Wittens discovery that topological string theory on super twistor space provides
a description of N = 4 super-YangMills (SYM) [1], considerable progress has been made
using twistor-inspired methods to study YangMills theories. An important factor in this
has been the proposal to use maximally helicity violating (MHV) diagrams, built using
MHV amplitudes as vertices, in order to derive amplitudes [2]. The MHV diagram conE-mail addresses: j.a.p.bedford@qmul.ac.uk (J. Bedford), a.brandhuber@qmul.ac.uk (A. Brandhuber),
w.j.spence@qmul.ac.uk (B. Spence), g.travaglini@qmul.ac.uk (G. Travaglini).
doi:10.1016/j.nuclphysb.2005.01.032
60
J. Bedford et al. / Nuclear Physics B 712 (2005) 5985
struction has the appealing feature that twistor space localisation is built in. The method
also carries the great practical advantage of tremendously simplifying the calculation of
amplitudes. It was quickly confirmed that MHV diagrams gave correct results at tree level
(see [3] for a review of the field up to August 2004). The prognosis at one-loop was initially poor, however, as general arguments indicated that it would be impossible to ignore
conformal supergravity fields propagating in the loops [4]. Separately to this, initial explorations of the differential equations satisfied by known one-loop amplitudes appeared to
suggest unexpected complications in their twistor-space localisation properties [5].
A direct derivation from MHV diagrams of the one-loop MHV scattering amplitudes in
N = 4 SYM was presented in [6]. The question of the localisation of amplitudes was then
revisited, and the complications previously found were seen to be due to the appearance
of additional inhomogeneous terms [7] in the differential equations obeyed by one-loop
amplitudes. Taking into account the corrections coming from this, one finds that indeed the
one-loop MHV amplitudes in N = 4 SYM localise on pairs of lines in twistor space [8],
as the direct construction of [6] suggests. This encourages one to conjecture that the whole
quantum theory of N = 4 SYM possesses simple twistor space localisation properties.
By studying the differential equations satisfied by the unitarity cuts of amplitudes, the
coefficients of the box functions in the next to MHV (NMHV) amplitudes in N = 4 SYM
have recently been shown to localise on planes in twistor space [9]. A direct MHV diagram
construction of these amplitudes has not yet been given however.
The study of the analytic properties of amplitudes, using the twistor-inspired approach,
has since been found useful in the general analysis of one-loop amplitudes in N = 4 Yang
Mills, with a recent derivation of the (++++) one-loop NMHV amplitude [10,11].
This coincides with one case of the general one-loop seven-gluon NMHV amplitude which
was also found recently [12] using the cut-constructibility approach. For N = 1 theories,
the twistor space structure of one-loop amplitudes was studied in [5,13,14] and it was found
that the holomorphic anomaly of unitarity cuts [7] leads to differential equations [13], in
contrast to algebraic equations for N = 4 [10], obeyed by the one-loop amplitudes.
MHV diagrams provide a well-defined prescription for the direct derivation of amplitudes. It is natural to ask whether the MHV diagram construction of the one-loop N = 4
MHV amplitudes of [6] can be generalised in other directionsin particular to theories
with less supersymmetry. This has been confirmed in recent work [15,16], where the MHV
diagram method was shown to correctly reproduce the known MHV amplitudes for the
N = 1 chiral multiplet. This result implies that one-loop MHV amplitudes for all supersymmetric gauge theories can be derived from MHV diagrams, and hence have simple
localisation properties in twistor space.
The close relationship between the MHV diagram construction and unitarity-based
methods [17], first seen in [6], and the success in applying this method to the N = 1 case,
encourages the belief that all cut-constructible amplitudes may be amenable to this new
approach. It is also of great importance to explore whether MHV diagrams can be used
at loop level in non-supersymmetric theories.1 These motivations lead one to consider the
one-loop MHV amplitudes in pure YangMills theory. These amplitudes consist of terms
1 The paper [5] discusses the twistor structure of some non-supersymmetric one-loop amplitudes and the possible role of additional vertices in these models. A recent paper [18] has also developed a generalised MHV
61
containing cuts, which we call the cut-constructible part of the amplitude, plus additional
rational terms. The amplitudes are of great interest, since they are an example of one-loop
n-point scattering amplitudes in QCD, where all external particles and the particle running
in the loop are gluons, and they can be decomposed as
glue
An
=4
= AN
4AnN =1,chiral + Ascalar
.
n
n
(1.1)
The first term describes the contribution of an N = 4 SYM multiplet to the amplitude.
The second is 4 times the contribution of an N = 1 chiral multiplet, and the third is a
non-supersymmetric amplitude with only complex scalars propagating in the loop. In this
paper we focus on the calculation of the final contribution since the other two are known.
A similar supersymmetric decomposition exists for one-loop gluon scattering amplitudes
with massless quarks or adjoint fermions running in the loop.
The one-loop MHV amplitude in pure YangMills is known only for two special
caseswhen the two negative-helicity external gluons are adjacent, the cut-constructible
part is known [19]; and, in the five-gluon case, the full amplitude, including rational parts,
has been calculated for arbitrary helicity configurations in [20]. In this paper, we will use
MHV diagrams to derive a compact expression for the cut-constructible part of the general
one-loop MHV multi-gluon amplitude when there are scalar particles in the loopthe last
term of (1.1). This generalises the known special cases with adjacent negative-helicity gluons, and the five-gluon non-adjacent amplitude. Moreover, this is the first example of the
application of the MHV diagram approach to non-supersymmetric loop amplitudes, and
provides further evidence that all cut-constructible (parts of) amplitudes may be derived
using standard MHV diagrams. Of course, it would be extremely interesting to extend the
MHV diagram method to obtain the rational pieces. This might require the construction of
suitable MHV vertices where the off-shell legs are continued to 4 2 dimensions or the
inclusion of additional effective vertices as proposed in [5].
The plan for the rest of the paper is as follows. In Section 2 we present the formal expression for the one-loop MHV diagrams with a complex scalar running in the loop, which
we use in Section 3 to rederive the known amplitude when the negative-helicity gluons
are in adjacent positions. In Section 4 we derive a compact expression for the amplitude
in the case where the negative-helicity gluons are in arbitrary positions. Our final result is
given by Eq. (4.22). We also briefly comment on the twistor space structure of our result.
Section 5 is devoted to some consistency checks of our general amplitude. Specifically, we
show that it correctly incorporates the adjacent case result [19], also directly reproduced
in Section 3, and the cut-constructible part of the five-gluon amplitude computed in [20].
Finally, we show that our expression has the expected infrared singularities.
For further related work on the string theory side, and on the gauge theory side, see
[2227] and [2834], respectively.
diagram construction for scattering amplitudes involving a Higgs boson and gluons. These amplitudes are described in terms of a tree-level, non-supersymmetric effective interaction which arises by integrating out a heavy
top quark in one-loop diagrams.
62
2. The scalar amplitude

In complete similarity with the N = 4 and N = 1 cases, see, e.g., [15], we can immediately write down the expression for the scalar amplitude in terms of MHV vertices
as

=
Ascalar
dM A l1 , m1 , . . . , i , . . . , m2 , l2
n
m1 ,m2 ,

A l2 , m2 + 1, . . . , j , . . . , m1 1, l1 ,
(2.1)
where the ranges of summation of m1 and m2 are

j + 1 m1 i,
i m2 j 1.
(2.2)
, for fixed m1 and m2 , is depicted in

The typical MHV diagram contributing to Ascalar
n
Fig. 1. The off-shell vertices A in (2.1) correspond to having complex scalars running in
the loop. It follows that there are two possible helicity assignments2 for the scalar particles
in the loop which have to be summed over. These two possibilities are denoted by in
(2.1) and in the internal lines in Fig. 1. It turns out that each of them gives rise to the same
integrand for (2.1),
iAtree
n
m2 m2 + 1m1 1 m1 il1 2 j l1 2 il2 2 j l2 2

.
ij 4 m1 l1 m1 1 l1 m2 l2 m2 + 1 l2 l1 l2 2
(2.3)
A crucial ingredient in (2.1) is the integration measure dM. This measure was constructed
in [6], using the decomposition L := l + z for a non-null four-vector L in terms of a null
vector l and a real parameter z. is a null reference vector, which disappears in the final
result. We refer the reader to Sections 3 and 4 of [6] for the construction of this measure
(also reviewed in Section 3 of [15]), and here we merely quote the result
dz 42
d
(2.4)
LIPS(l2 , l1 ; PL;z ),
z
where Li := li + zi , i = 1, 2 and z := z1 z2 . Thus the integration measure dM decomposes into the product of a Lorentz-invariant two-particle phase space measure and a
dispersive measure dz/z. The momentum PL;z flowing in the phase space measure is
dM =
PL;z := PL z.
(2.5)
The interpretation of dz/z as a dispersive measure follows at once when one observes that
[6]
2
dPL;z
dz
.
= 2
z
PL;z PL2
(2.6)
In order to calculate (2.1), we will first integrate the expression (2.3) over the Lorentz
invariant phase space (appropriately regularised to 4 2 dimensions), and then perform
2 For scalar fields, the helicity simply distinguishes particles from antiparticles (see, for example, [21]).
63
Fig. 1. A one-loop MHV diagram with a complex scalar running in the loop, computed in Eq. (2.1). We have
indicated the possible helicity assignments for the scalar particle.
the dispersion integral. For the sake of clarity, we will separate the analysis into two parts.
Firstly, we will present the (simpler) calculation of the amplitude in the case where the two
negative-helicity gluons are adjacent. This particular amplitude has already been computed
by Bern, Dixon, Dunbar and Kosower in [19] using the cut-constructibility approach; the
result we will derive here will be in precise agreement with the result in that approach.
Then, in Section 4 we will move on to address the general case, deriving new results.
3. The scattering amplitude with adjacent negative-helicity gluons

The adjacent case corresponds to choosing i = m1 , j = m1 1 in Fig. 1. Therefore we
now have a single sum over MHV diagrams, corresponding to the possible choices of m2 .
We will also set i = 2, j = 1 for the sake of definiteness, and m2 = m.
After conversion into traces, the integrand of (2.1) takes on the form

tr+ (/k1 k/2P/ L;z /l2 ) tr+ (/k1 k/2 /l2P/ L;z ) tr+ (/k1 k/2 k/m+1 /l2 ) tr+ (/k1 k/2 k/m /l2 )
,
(l2 m + 1)
(l2 m)
25 (k1 k2 )3 (l1 l2 )2
(3.1)
2 /2 by momentum conservation.
where we note that (l1 l2 ) = PL;z
The next step consists of performing the PassarinoVeltman reduction [35] of the
Lorentz invariant phase space integral of (3.1). This requires the calculation of the threeindex tensor integral

I (m, PL;z ) =
dLIPS(l2 , l1 ; PL;z )
l2 l2 l2
.
(l2 m)
(3.2)
This calculation is performed in Appendix A. The result of this procedure gives the following term at O( 0 ), which we will later integrate with the dispersive measure
64

k 1 k/2 k/mP/ L;z )]2

2 [tr+ (/
PL;z
A scalar
=
n
3
25 (k1 k2 )3

tr+ (/k1 k/2P/ L;z k/m )
2(k1 k2 )
(m m + 1),
+
(m PL;z )3
(m PL;z )2
(3.3)
and we have dropped a factor of 4Atree on the right-hand side of (3.3), where is
defined in (C.1). We can reinstate this factor at the end of the calculation. We also notice
that (3.3) is a finite expression, i.e., it is free of infrared poles.
An important remark is in order here. On general grounds, the result of a phase space
integral in, say, the P 2 -channel, is of the form

I() = P 2
(3.4)
f (),
where
f1
(3.5)
+ f0 + f1 + ,

and fi are rational coefficients. In the case at hand, infrared poles generated by the phase
space integrals cancel completely, so that we can in practice replace (3.5) by f ()
f0 + f1 + . The amplitude A is then obtained by performing a dispersion integral,
which converts (3.4) into an expression of the form
f () =

(P 2 )
g0
(3.6)
g() =
g0 log P 2 + g1 + O(),

where g() = g0 + g1 + , and the coefficients gi are rational functions, i.e., they are
free of cuts. Importantly, errors can be generated in the evaluation of phase space integrals
if one contracts (4 2)-dimensional vectors with ordinary four-vectors. This does not
affect the evaluation of the coefficient g0 := g( = 0), and hence the part of the amplitude
containing cuts is reliably computed; but the coefficients gi for i 1, in particular g1 , are in
general affected. This implies that rational contributions to the scattering amplitude cannot
be detected [19] in this construction. A notable exception to this is provided by the phase
space integrals which appear in supersymmetric theories. These are four-dimensional cutconstructible [19], in the sense that the rational parts are unambiguously linked to the
discontinuities across cuts, and can therefore be uniquely determined.3 This occurs, for
example, in the calculation of the N = 4 MHV amplitudes at one-loop performed in [6]. In
the present case, however, the relevant phase space integrals violate the cut-constructibility
criteria given in [19],4 since we encounter tensor triangles with up to three-loop momenta
in the numerator. Hence, we will be able to compute the part of the amplitude containing
cuts, but not the rational terms. In practice, this means that we will compute all phase space
integrals up to O( 0 ), and discard O() contributions, which would generate rational terms
that cannot be determined correctly.
After this digression, we now move on to the dispersion integration. In the center of
mass frame, where PL;z := PL;z (1, 0), all the dependence on PL;z in (3.3) cancels out, as
A() =
3 For more details about cut-constructibility, see the detailed analysis in Sections 35 of [19].
4 An example of an integral violating the power-counting criterion of [19] is provided by (A.3).
65
there are equal powers of PL;z in the numerator as in the denominator of any term. As a
consequence, the dependence on the arbitrary reference vector disappears (see [16] for
the application of this argument to the N = 1 case). Using (2.6) in order to re-express dz/z
in terms of the relevant dispersive measure, we see that we are left with dispersion integrals
of the form

2
ds
1
(s ) = csc() PL2 .
I PL :=
(3.7)
2

s PL
Taking this into account, the dispersion integral of (3.3) then gives

(PL2 ) [tr+ (/k1 k/2 k/mP/ L )]2
A scalar
=

csc()
n
3

25 (k1 k2 )3

tr+ (/k1 k/2P/ L k/m ) 2(k1 k2 )
+
(m PL )3
(m PL )2
(m m + 1).
(3.8)
The momentum flow can be conveniently represented as in Fig. 2, where we define

P := q2,m1 ,
(3.9)
Q := qm+1,1 = q2,m ,
p2
and qp1 ,p2 := l=p1 kl . We also have PL := q2,m = Q.
Now we wish to combine the terms in the first and second lines of (3.8) with those in
the third line. Since (3.8) is summed over m, we simply shift m + 1 m in the terms
of the second line. Let us now focus our attention on the second term in (3.3) (similar
manipulations will be applied to the first term). Writing the m m + 1 term explicitly, we
obtain a contribution proportional to

[tr+ (/k1 k/2 k/mP/ L )]2 [tr+ (/k1 k/2 k/m+1P/ L )]2

.
PL2
(3.10)
(m PL )2
((m + 1) PL )2
By shifting m + 1 m in the second term of (3.10), we convert its PL to PL q2,m1 =
P (whereas, in the non-shifted term, PL = Q). The expression (3.10) then reads

/ 2
[tr+ (/k1 k/2 k/mQ)]
Q2
,
P 2
(3.11)
2
(m Q)
/ = tr+ (/k1 k/2 k/mP/ ) and Q m = P m. Notice also that
where we used tr+ (/k1 k/2 k/mQ)
m Q = (1/2)(Q2 P 2 ).
Next we reinstate the antisymmetry of the amplitudes under the exchange of the indices
1 2 (which is manifest from Eq. (2.3)). Doing this we get
2

/
tr+ (/k1 k/2 k/mQ)
2
2
1
tr+ (/k1 k/2 k/mQ)
/
tr+ (/k1 k/2Q/
/ km)
2

= 2(k1 k2 )(m Q) tr+ (/k1 k/2 k/mQ)
(3.12)
/ tr+ (/k1 k/2Q/
/ km) .
Following similar steps for the first term in (3.8), we arrive at the following expression for
the amplitude before taking the 0 limit
A = A1, + A2, ,
(3.13)
66
Fig. 2. A triangle function contributing to the amplitude in the case of adjacent negative-helicity gluons. Here we
have defined P := qj,m1 , Q := qm+1,i = qj,m (in the text we set i = 1, j = 2 for definiteness).
where
A1, =

Atree
tr+ (/k1 k/2 k/mq/ m,1 ) tr+ (/k1 k/2q/ m,1 k/m ) T (m, q2,m1 , q2,m ),
[2] 6
t
1
A2,
2
Atree
tr+ (/k1 k/2 k/mq/ m,1 ) tr+ (/k1 k/2q/ m,1 k/m )
= [2]
3
3
(t1 )

2
tr+ (/k1 k/2 k/mq/ m,1 ) tr+ (/k1 k/2q/ m,1 k/m ) T(3) (m, q2,m1 , q2,m ),
(3.14)
and t1[2] follows from the definitions below Eq. (4.8). In order to write (3.14) in a compact
from, we have introduced -dependent triangle functions [15]
T(r) (p, P , Q) :=
1 (P 2 ) (Q2 )
,

(Q2 P 2 )r
(3.15)
where p + P + Q = 0, and r is a positive integer.5

We can now take the 0 limit. As long as P 2 and Q2 are non-vanishing, one has
lim T(r) (p, P , Q) = T (r) (p, P , Q),
0
P 2 = 0, Q2 = 0,
(3.16)
where the -independent triangle functions are defined by

log(Q2 /P 2 )
(3.17)
.
(Q2 P 2 )r
If either of the invariants vanishes, the limit of the -dependent triangle gives rise to an
infrared-divergent term (which we call a degenerate trianglethis is one with two massless legs). For example, if Q2 = 0, one has
T (r) (p, P , Q) :=
T (p, P , Q)|Q2 =0
1 (P 2 )
,

P2
5 For r = 1 we will omit the superscript (1) in T (1) .
0.
(3.18)
67
The two possible configurations which give rise to infrared-divergent contributions correspond to the following two possibilities:
2
2 = t [2] ;
(a) q2,m1 = k2 (hence q2,m1
= 0). In this case we also have q2,m
2
[2]
2 = 0). Therefore q 2
(b) q2,m = k1 (hence q2,m
2,m1 = tn .
We notice that infrared poles will appear only in terms corresponding to the triangle function T . Indeed, whenever one of the kinematical invariants contained in T (3) vanishes, the
combination of traces multiplying this function in (3.14) vanishes as well.
In conclusion, we arrive at the following result, where we have explicitly separated out
the infrared-divergent terms:6
Ascalar
= Apoles + A1 + A2 ,
n
(3.19)
where
1 [2] [2]
1
t2
Apoles = Atree
,
+ tn
6

n1

1
1
A1 = Atree [2]
6
t1 m=4
T (m, q2,m1 , q2,m ),
n1
2
1
1
A2 = Atree [2]
3
(t1 )3 m=4

2
tr+ (/k1 k/2 k/mq/ m,1 ) tr+ (/k1 k/2q/ m,1 k/m ) T (3) (m, q2,m1 , q2,m ).
(3.20)
More compactly, we can recognise that Apoles and A1 reconstruct the contribution of a
N = 1 chiral supermultiplet, and rewrite (3.19) as

1 N =1,chiral
1
m (3)
Ascalar
= A12
Atree
B12
T (m, q2,m1 , q2,m ),
n
3
3 12
n1
(3.21)
m=4
where

2
m
= tr+ (/k1 k/2 k/mq/ m,1 ) tr+ (/k1 k/2q/ m,1 k/m )
B12

2
tr+ (/k1 k/2q/ m,1 k/m ) tr+ (/k1 k/2 k/mq/ m,1 )
(3.22)
and
N =1,chiral
A12
= Atree
12
n

m=4
T (m, q2,m1 , q2,m ).
(3.23)
6 A factor of 4 will be understood on the right-hand sides of Eqs. (3.19), (3.21), (3.23), where is defined
in (C.1).
68
This is our result for the cut-constructible part of the n-gluon MHV scattering amplitude
with adjacent negative-helicity gluons in positions 1 and 2. This expression was first derived by Bern, Dixon, Dunbar and Kosower in [19], and our result agrees precisely with
this. A remark is in order here. In [19], the final result is expressed in terms of a function
L2 (x) :=
log x (x 1/x)
,
(1 x)3
(3.24)
which contains a rational part (x 1/x)/(1 x)3 . This rational part removes a spurious
third order pole from the amplitude, but with our approach we did not expect to detect
rational terms in the scattering amplitude, and indeed we do not find such terms.7 Furthermore, we do not find the other rational terms which are known to be present in the one-loop
scattering amplitude [20].
4. The scattering amplitude in the general case

The situation where the negative-helicity gluons are not adjacent is technically more
challenging. Our starting point will be (2.3), to which we will apply the Schouten identity
(see Appendix D for a collection of spinor identities used in this paper). Eq. (2.3) can then
be written as a sum of four terms:8
C(m1 , m2 + 1) C(m1 , m2 ) C(m1 1, m2 + 1) + C(m1 1, m2 ),
(4.1)
where
C(a, b) :=
il1 j l1 2 il2 2 j l2 iaj b

.
l1 al2 b
ij 4 l1 l2 2
(4.2)
The calculation of the phase space integral of this expression is discussed in Appendix B.
The result is

d 42 LIPS(l2 , l1 ; PL;z ) C(a, b)

1 tr+ (/ij/b/a/ )
tr+ (/ij/a/P/ L;z )
2(i j )
=
(a
b)
tr+ (/ij/P/ L;za/ )2
+
3 (a b)
(PL;z a)3
(PL;z a)2

2
1 tr+ (/ij/b/a/ ) tr+ (/ij/a/b/) tr+ (/ij/P/ L;za/ )
+
+
(a
b)
2
(a b)2
(PL;z a)2

tr+ (/ij/a/b/) tr+ (/ij/b/a/ ) tr+ (/ij/a/b/) tr+ (/ij/P/ L;za/ )
+
(a
b)
(PL;z a)
(a b)3

2
2
(a b) 2
tr+ (/ij/a/b/) tr+ (/ij/b/a/ )
PL;z ,
log 1
+
(4.3)
N
(a b)4
7 In our notation L corresponds to T (3) , which, however, lacks a rational term.
2
8 We drop the factor of iAtree from now on and reinstate it at the end of the calculation.
n
69
where N := N (P ) := (a b)P 2 2(P a)(P b), and we have suppressed a factor of

4(PL2 ) [28 (i j )4 ]1 on the right-hand side of (4.3), where is defined in (C.1).
We notice that (4.3) is symmetric under the simultaneous exchange of i with j and a
with b. This symmetry is manifest in the coefficient multiplying the logarithmthe last
term in (4.3); for the remaining terms, non-trivial gamma matrix identities are required.
For instance, consider the terms in the second line of (4.3). These terms are present in the
adjacent gluon case (3.3), and it is therefore natural to expect that the trace structure of this
term is separately invariant when i j and a b. Indeed this is the case, thanks to the
identity

ij/a/P/ L;z )
2(i j )
3
2 tr+ (/
32(i j ) = tr+ (/ij/P/ L;za/ )
+
(PL;z a)3
(PL;z a)2

tr+ (/ij/P/ L;za/ )

2(i j )
+ tr+ (/ij/a/P/ L;z )2
(4.4)
.
+
(PL;z a)3
(PL;z a)2
Similar identities show that the third and fourth lines of (4.3) are invariant under the simultaneous exchange i j and a b.
The next step is to perform the dispersion integral of (4.3), i.e., the integral over the vari2 )
able z. This appears in the terms involving PL;z in (4.3), and in an overall factor (PL;z
arising from the dimensionally regulated measure.
The integral over the term involving the logarithm has been evaluated in [6], with the
result

dz 2
(a b) 2
PL;z
PL;z
log 1
z
N

2
dPL;z 2
(a b) 2
P
=
log
1
P
L;z
L;z
2 P2
N
PL;z
L

(a b) 2
PL + O().
= Li2 1
(4.5)
N (P )
Notice that these terms were not present in the adjacent negative-gluon case considered in
Section 3.
Next we move on to the remaining terms in (4.3). Inspecting their z-dependence, we see
that, in complete similarity with the adjacent case of Section 3, in each term there are the
same powers of PL;z in the numerator as in the denominator. Hence, in the centre of mass
frame in which PL;z := PL;z (1, 0), one finds that PL;z cancels completely. Note that this
also immediately resolves the question of gauge invariance for these termsthis occurs
only through the dependence in PL;z = PL z. Furthermore, the box functions coming
from (4.5) are separately gauge invariant [6]. The conclusion is that our expression for the
amplitude below, built from sums over MHV diagrams of the dispersion integral of (4.3),
will be gauge invariant. Moreover, apart from (4.5), the only other dispersion integral we
will need is that computed in (3.7).
It follows from this discussion that the result of the dispersion integral of (4.3) is (suppressing a factor of 4(PL2 ) [28 (i j )4 ]1 [ csc()])
70

dz
d 42 LIPS(l2 , l1 ; PL;z ) C(a, b)
z

1 tr+ (/ij/b/a/ )
tr+ (/ij/a/P/ L )
1
2(i j )
(a
b)
tr+ (/ij/P/ La/ )2
= PL2
+

3 (a b)
(PL a)3
(PL a)2

1 tr+ (/ij/b/a/ ) tr+ (/ij/a/b/) tr+ (/ij/P/ La/ )2

+
+ (a b)
2
(a b)2
(PL a)2

tr+ (/ij/a/b/) tr+ (/ij/b/a/ ) tr+ (/ij/a/b/) tr+ (/ij/P/ La/ )
+
(a
b)
(PL a)
(a b)3

2
2
(a b) 2
tr+ (/ij/a/b/) tr+ (/ij/b/a/ )
1
.
P
+
(4.6)
Li
2
N (PL ) L
(a b)4
Now, due to the four terms in (4.1), the sum over MHV diagrams will include a signed
sum over four expressions like (4.6). Let us begin by considering the last line of (4.6). This
is a term familiar from [6,15], corresponding to one of the four dilogarithms in the novel
expression found in [6] for the finite part B of a scalar box function

(a b) 2
(a b) 2
2
2
B s, t, P , Q = Li2 1
P + Li2 1
Q
N (P )
N (P )

(a b)
(a b)
s Li2 1
t ,
Li2 1
(4.7)
N (P )
N (P )
with s := (P + a)2 , t := (P + b)2 , and P + Q + a + b = 0. By taking into account the four
terms in (4.1) and summing over MHV diagrams as specified in (2.1) and (2.2), one sees
that each of the four terms in any finite box function B appears exactly once, in complete
similarity with [6,15], so that the final contribution of this term will be9
i1
j 1

1 ij 2 2
2
2
2
bm1 m2 B qm1 ,m2 1 , qm
+1,m2 , qm1 +1,m2 1 , qm2 +1,m1 1 ,
1
2
(4.8)
m1 =j +1 m2 =i+1
where ti[k] := (pi + pi+1 + + pi+k1 )2 for k 0, and ti[k] = ti[nk] for k < 0. In writing
(4.8), we have taken into account that the dilogarithm in (4.6) is multiplied by a coefficient
ij
proportional to the square of bm1 m2 , where
ij
bm1 m2 := 2
tr+ (/ki k/j k/m1 k/m2 ) tr+ (/ki k/j k/m2 k/m1 )
.
[(ki + kj )2 ]2 [(km1 + km2 )2 ]2
(4.9)
ij
We notice that bm1 m2 is the coefficient of the box functions in the one-loop N = 1 MHV
amplitude, originally calculated by Bern, Dixon, Dunbar and Kosower in [19], and derived
in [15,16] using the MHV diagram approach for loops proposed in [6].
9 We multiply our final results by a factor of 2, which takes into account the two possible helicity assignments
for the scalars in the loop.
71
ij
Furthermore, we observe that bm1 m2 is holomorphic in the spinor variables, and as such
has simple localisation properties in twistor space. Indeed, from (4.9) it follows that
ij
bm 1 m 2 = 2
im1 im2 j m1 j m2

.
ij 2 m1 m2 2
(4.10)
Summing over the four terms for the remainder of (4.6) can be done in complete similarity
with Section 4 of [15].10 We will skip the details of this derivation, and will now present
our result.
In order to do this, we find it convenient to define the following expressions:
(ij m2 m1 )
(ij m2 + 1 m1 )
((m2 + 1) m1 ) (m2 m1 )
m2 m2 + 1
= 2[ij ]m1 im1 j
,
m2 + 1 m1 m1 m2
(ij m1 m2 + 1)(ij m2 + 1 m1 ) (ij m1 m2 )(ij m2 m1 )
ij
,
Sm1 m2 :=
((m2 + 1) m1 )2
(m2 m1 )2
(ij m1 m2 + 1)2 (ij m2 + 1 m1 ) (ij m1 m2 )2 (ij m2 m1 )
ij
,
Im1 m2 :=
((m2 + 1) m1 )3
(m2 m1 )3
ij
Am1 m2 :=
(4.11)
(4.12)
(4.13)
where for notational simplicity we set (a1 a2 a3 a4 ) := tr+ (/

a 1a/ 2a/ 3a/ 4 ) in the above. We also
note the symmetry properties
ji
ij
Am1 m2 = Am1 m2 ,
ji
ij
Sm 1 m 2 = Sm 1 m 2 .
(4.14)
The momentum flow is best described using the triangle diagram in Fig. 4, where we use
the following definitions:
P := qm2 +1,m1 1 = qm1 ,m2 ,
Q := qm1 +1,m2 .
(4.15)
The triangle in Fig. 5 also appears in the calculation, and can be converted into a triangle
as in Fig. 4but with i and j swappedif one shifts m1 1 m1 , and then swaps
m1 m2 .
We then introduce the coefficients

ij
ij
Am1 m2 := 28 (i j )4 Am1 m2 (ij m1 Q)2 (ij Qm1 ) (ij m1 Q)(ij Qm1 )2 ,
(4.16)

ij
ij
8
4
2
2
A m1 m2 := 2 (i j ) Am1 m2 (ij m1 Q) (ij Qm1 ) ,
(4.17)

ij
ij
Sm1 m2 = 28 (i j )4 Sm1 m2 (ij m1 Q)2 + (ij Qm1 )2 ,
(4.18)

ij
ji
8
4 ij
Im1 m2 := 2 (i j ) Im1 m2 (ij Qm1 ) + Im1 m2 (ij m1 Q) .
(4.19)
We will also make use of the -dependent triangle functions introduced in (3.15), whose
0 limits have been considered in (3.16)(3.18). This is in order to write a compact
10 In Section 3 we have illustrated in detail how this sum is performed for the simpler case of adjacent negativehelicity gluons.
72
expression which incorporates also the infrared-divergent terms.11

We can now present our result for the one-loop MHV amplitude (2.1)
i1
j 1

1 ij 2
bm 1 m 2
Ascalar = Atree
2
m1 =j +1 m2 =i+1

2
2
2
2
B qm
, qm
, qm
, qm
1 ,m2 1
1 +1,m2
1 +1,m2 1
2 +1,m1 1

j 1
i1
8
ij
+
Am1 m2 T (3) (m1 , P , Q)
3
m1 =j +1 m2 =i

ij
+ (i j )A m1 m2 T (2) (m1 , P , Q)
+2
j 1

ij

ij
Sm1 m2 T (2) (m1 , P , Q) + Im1 m2 T (m1 , P , Q)
i1
m1 =j +1 m2 =i

+ (i j )
(4.20)
where on the right-hand side of (4.20) a factor of 4 is understood, where is defined

in (C.1). We can also introduce the coefficient

1 (ij m2 + 1m1 )
(ij m2 m1 ) (ij m1 Q) (ij Qm1 )
ij
cm1 m2 :=
(4.21)
,
2 ((m2 + 1) m1 ) (m2 m1 )
[(i + j )2 ]2
which already appears as the coefficient multiplying the triangle function T in the N = 1
amplitude (see, e.g., Eq. (2.19) of [15]), and rewrite (4.20) as
i1
j 1

1 ij 2
bm 1 m 2
Ascalar = Atree
2
m1 =j +1 m2 =i+1
2

2
2
2
B qm
, qm
, qm
, qm
1 ,m2 1
1 +1,m2
1 +1,m2 1
2 +1,m1 1

j 1
i1
1 1 ij
cm m
+
2
3 1 2
m1 =j +1 m2 =i

(ij m1 Q)(ij Qm1 ) (3)
T (m1 , P , Q) + T (m1 , P , Q)
2(i j )2
+2
j 1

ij

ij
Sm1 m2 T (2) (m1 , P , Q) + Im1 m2 T (m1 , P , Q)
i1
m1 =j +1 m2 =i

+ (i j )
(4.22)
11 The infrared-divergent terms will be described below, and used to check that our result has the correct
infrared pole structure.
73
Several remarks are in order.

2
2
, qm
correspond to the s- and t-channel of the
(1) As usual, the variables qm
1 ,m2 1
1 +1,m2
finite part of the easy two-mass box function with massless legs m1 and m2 , and
2
2
, qm
(Fig. 3).
massive legs qm
1 +1,m2 1
2 +1,m1 1
(2) Compared to the range for m1 and m2 indicated in (2.2), we have omitted m1 = i in
the summation of the triangles, as for this value the coefficients A, S, I defined in
(4.16)(4.19) vanish. Notice also that we have i Q and j P .
(3) In the case of adjacent negative-helicity gluons, the only surviving terms are those
ij
containing the coefficient cm1 m2 , on the third and fourth lines of (4.20) or (4.22). We
will return to this point in Section 5.
Fig. 3. A box function contributing to the amplitude in the general case. The negative-helicity gluons, i and j ,
cannot be in adjacent positions, as the figure shows.
Fig. 4. One type of triangle function contributing to the amplitude in the general case, where i Q, and j P .
74
Fig. 5. Another type of triangle function contributing to the amplitude in the general case. By first shifting
m1 1 m1 , and then swapping m1 m2 , we convert this into a triangle function as in Fig. 4but with
i and j swapped. These are the triangle functions responsible for the i j swapped terms in (4.20) or (4.22).
(4) We comment that, in contrast to the adjacent case (see (3.21)), in the general case the
N = 1 chiral amplitude does not separate out naturally in the final result, as one can
see from the coefficient of the box function B in (4.20).
Next we wish to separate explicitly the infrared divergences from (4.20). We can immediately anticipate that there will be four infrared-divergent terms, corresponding to the
four possible degenerate triangles. Two of these degenerate triangles occur when either P 2
or Q2 happen to vanish. The other two originate from the i j swapped terms.
Let us consider first the terms arising from the summation with i j unswapped. When
Q2 = 0, it follows that m1 = i 1 and m2 = i (see Fig. 4). When P 2 = 0, it follows that
m1 = j + 1 and m2 = j 1 (see Fig. 5). Hence
T (r) (p, P , Q) ()r
T (r) (p, P , Q)
1
[2]
1 (ti1 )
,
(t [2] )r
i1
[2]
(tj )
,
(tj[2] )r
Q2 0,
P 2 0.
(4.23)
The infrared-divergent terms coming from Q2 = 0 are then easily extracted, and are
(ij i 1i + 1)
1 [2]
ti1 4(i j )
2
((i + 1) (i 1))

(ij i + 1i 1)(ij i 1i + 1)
8
(ij i + 1i 1)
2
,
(i j ) +
(i j ) 2
3
((i + 1) (i 1))
((i + 1) (i 1))2
(4.24)
and from
P2
=0
75
1 [2]
(ijj 1j + 1)
tj
4(i j )
2
((j + 1) (j 1))

(ijj + 1j 1)(ijj 1j + 1)
8
(ijj
+ 1j 1)
.
(i j ) +
(i j )2 2
3
((j + 1) (j 1))
((j + 1) (j 1))2
(4.25)
Likewise, from the swapped degenerate triangles we obtain the following infrareddivergent terms:
(ijj + 1j 1)
1 [2]
tj 1 4(i j )
2
((j + 1) (j 1))

(ijj 1j + 1)(ijj + 1j 1)
(ijj 1j + 1)
8
,
(i j ) +
(i j )2 2
3
((j + 1) (j 1))
((j + 1) (j 1))2
(4.26)
and
1 [2]
(ij i + 1i 1)
ti
4(i j )
2
((i + 1) (i 1))

(ij i 1i + 1)(ij i + 1i 1)
(ij
i 1i + 1)
8
.
(i j ) +
(i j )2 2
3
((i + 1) (i 1))
((i + 1) (i 1))2
(4.27)
4.1. Comments on twistor space interpretation

We would like to make some brief comments on the interpretation in twistor space of
our result (4.22).
ij
(1) As noticed earlier, the coefficient bm1 m2 appears already in the N = 1 chiral supermultiplet contribution to a one-loop MHV amplitude, where it multiplies the box function.
ij
It was noticed in Section 4 of [5] that bm1 m2 is a holomorphic function, hence it does
not affect the twistor space localisation of the finite box function.12
ij
(2) The coefficient cm1 m2 also appears in the N = 1 amplitude, as the coefficient of the
triangles (see, e.g., Eq. (2.19) of [15]). Its twistor space interpretation was considered
ij
in Section 4 of [5], where it was found that cm1 m2 has support on two lines in twistor
space. Furthermore, it was also found that the corresponding term in the amplitude has
a derivative of a delta function support on coplanar configurations.
ij
(3) The combination cm1 m2 (ij m1 Q)(ij Qm1 )/(i j )2 already appears in the case of adjacent negative-helicity gluons. The localisation properties of the corresponding term in
the amplitude were considered in Section 5.3 of [5], and found to have, similarly to the
previous case, derivative of a delta function support on coplanar configurations.
(4) On general grounds, we can argue that the remaining terms in the amplitude have a
twistor space interpretation which is similar to that of the terms already considered.
12 We thank Dave Dunbar for discussions on this point.
76
The gluons whose momenta sum to P are contained on a line; likewise, the gluons
whose momenta sum to Q localise on another line.
We observe that the rational parts of the amplitude are not generated from the MHV diagram construction presented here. Such rational terms were not present for the N = 1 and
N = 4 amplitudes derived in [6,15,16]. However, for the amplitude studied here, rational
terms are required to ensure correct factorisation properties [19].
5. Checks of the general result

In this section we present three consistency checks that we have performed for the result
(4.20) (or (4.22)) for the one-loop scalar contribution to the MHV scattering amplitude.
These checks are:
(1) For adjacent negative-helicity gluons, the general expression (4.20) should reproduce
the previously calculated form (3.21).
(2) In the case of five gluons in the configuration (1 2+ 3 4+ 5+ ), the result (4.20) should
reproduce the known amplitude given in [20].
(3) The result (4.20) should have the correct infrared-pole structure.
We next discuss these requirements in turn.
5.1. Adjacent case
The amplitude where the two negative-helicity external gluons are adjacent is given in
Section 7 of [19] and was explicitly rederived in Section 3 of this paper by combining
MHV vertices, see Eq. (3.21). It is easy to show that our general result (4.22) reproduce
correctly (3.21) as a special case.
To start with, recall that our result (4.22) is expressed in terms of box functions and
triangle functions, see Fig. 3 and Figs. 4, 5, respectively. In the adjacent case, the box
functions are not present. Indeed, in the sum (4.8) the negative-helicity gluons can never
be in adjacent positions (see Fig. 3).
Next, we focus on the triangles of Fig. 4. In terms of these triangles, requiring i and
j to be adjacent eliminates the sum over m2 , as we must have m2 = i and m2 + 1 = j .
Moreover, in this case Q = qm1 +1,i , P = qj,m1 1 and one has
1 m2
= 4(i j ),
Am
ij
Sijm1 m2 = 0,
Iijm1 m2 = 0,
(5.1)
for m2 = i, and m2 + 1 = j . Similar simplifications occur for the swapped triangle. Hence
the only surviving terms are those in the third and fourth lines of (4.20) (or (4.22)), and it is
then easy to see that they generate the same amplitude (3.8) already calculated in Section 3.
77
5.2. Five-gluon amplitude

The other special case is the non-adjacent five-gluon amplitude (1 2+ 3 4+ 5+ ), given
in Eq. (9) of [20]. This amplitude may be written as ic Atree times13
1
1
tr+ (/13/2/5/)2 tr+ (/13/5/2/)2
+ log(s34 )
B(s51 , s12 , s34 , 0)
6 6
27 (2 5)4 (1 3)4
1 tr+ (/13/2/5/)2 tr+ (/13/5/2/)2
3
24 (2 5)(1 3)4

2 log(s12 /s34 )
2 log(s34 /s51 )
tr+ (/13/5/2/)
+ tr+ (/13/2/5/)
(s12 s34 )3
(s34 s51 )3

1
1
2 log(s34 /s51 )
(/
1
3
/
4
/
2
/
)
tr
(/
1
3
/
2
/
5
/
)
tr
+
+
3 23 (1 3)3
(s34 s51 )3

tr+ (/13/2/5/)2 tr+ (/13/5/2/)2 log(s12 /s34 ) log(s34 /s51 )

+
26 (2 5)2 (1 3)4
(s12 s34 )2
(s34 s51 )2

tr+ (/13/2/5/)2 tr+ (/13/5/2/)2 log(s12 /s34 ) log(s34 /s51 )

+
(s12 s34 )
(s34 s51 )
26 (2 5)3 (1 3)4

1
1
log(s34 /s51 )
tr+ (/13/2/5/)
+
2
3 2 (1 3)
(s34 s51 )
+
+ (1, 4) (3, 5),
(5.2)
where the interchange on the last line applies to all terms above it in this equation, including
the first two terms, and the box function B is defined in (4.7). In deriving this from [20],
we have used the dilogarithm identity
Li2 (1 r) + Li2 (1 s) + log(r) log(s)

1r
1s
1s 1r
= Li2
+ Li2
Li2
.
s
r
r
s
(5.3)
We have checked explicitly that our expression for the n-gluon non-adjacent amplitude
(4.20), when specialised to the case with five gluons in the configuration (1 2+ 3 4+ 5+ ),
yields precisely the result (5.2). For the terms involving dilogarithms, this is easily done.
For the remaining terms, which contain logarithms, a more involved calculation is necessary using various spinor identities from Appendix D. A straightforward method of doing
this calculation begins with the explicit sum over MHV diagrams in this case, isolating
the coefficients of each logarithmic function such as, e.g., log(s12 ), and then checking that
these coefficients match those in (5.2). The remaining 1/ term arises from the following
discussion.
13 The derivation in [20] used string-based methods, which affects the coefficient of the pole term. In (5.2) we
have written the pole coefficient which matches the adjacent case.
78
5.3. Infrared-pole structure

The infrared-divergent terms (poles in 1/) can easily be extracted from (4.24)(4.27)
by simply replacing (tr[2] ) 1 (r = i 1, i, j 1, j ). Consider first the terms in (4.25)
and (4.26). After a little algebra, and using

(ij j + 1 j 1) + (ij j 1 i 1) = 4(i j ) (j 1) (j + 1) ,
(5.4)
one finds that these two contributions add up to
64
(i j )4 .
(5.5)
3
Similarly, the pole contribution arising from (4.24) and (4.27) gives an additional contribution of (64/3)(i j )4 . Reinstating a factor of 2 28 (i j )4 Atree , we see that the
pole part of (4.20) is simply given by
Atree
.
3
Hence our result (4.20) has the expected infrared-singular behaviour.
Ascalar | -pole =
(5.6)
Acknowledgements
It is a pleasure to thank Lance Dixon, Dave Dunbar, Michael Green, Marek Karliner,
Valya Khoze, David Kosower, Marco Matone and Sanjaye Ramgoolam for discussions.
G.T. acknowledges the support of PPARC.
Appendix A. PassarinoVeltman reduction
In Section 2 we saw that a typical term in the cut-constructible part of the YangMills
amplitude is the dispersion integral of the following phase space integral:

C(m) := dLIPS(l2 , l1 ; PL;z )
tr+ (/k1 k/2P/ L;z /l2 ) tr+ (/k1 k/2 /l2P/ L;z ) tr+ (/k1 k/2 k/m /l2 )
.
2 )2
(l2 m)(k1 k2 )3 (PL;z
(A.1)
The goal of this appendix is to perform the PassarinoVeltman reduction [35] of (A.1).
To this end, we rewrite C(m) as
tr+ (/k1 k/2P/ L;z ) tr+ (/k1 k/2 P/ L;z ) tr+ (/k1 k/2 k/m )
C(m) =
(A.2)
I (m, PL;z ),
2 )2
(k1 k2 )3 (PL;z
where14
I (m, PL ) =
dLIPS(l2 , l1 ; PL )
l2 l2 l2
.
(l2 m)
14 For the rest of this appendix we drop the subscript z in P

L;z for the sake of brevity.
(A.3)
79
On general grounds, I (m, PL ) can be decomposed as

I = m m m I1 + m m PL + m PL m + PL m m I2

+ m PL PL + PL m PL + PL PL m I3 + PL PL PL I4

+ m + m + m I5 + PL + PL + PL I6 , (A.4)
for some coefficients Ii , i = 0, . . . , 6. One can then contract with different combinations
of the independent momenta in order to solve for the Ii . Introducing the quantities
A := m m m I ,
B := m m PL I ,
C := m PL PL I ,
D := PL PL PL I ,
E := m I = 0,
F := PL I = 0,
(A.5)
the result for the PassarinoVeltman reduction of {I1 , . . . , I6 } in the basis {A, . . . , D} is

2

I2 = 5 PL2 / 2(m PL )5 , 6PL2 /(m PL )4 , 3/(m PL )3 , 0 ,

I3 = 2PL2 /(m PL )4 , 3/(m PL )3 , 0, 0 ,

I4 = 1/(m PL )3 , 0, 0, 0 ,
2

I5 = PL2 / 2(m PL )4 , 3PL2 / 2(m PL )3 , 1/(m PL )2 , 0 ,

I6 = PL2 / 2(m PL )3 , 1/(m PL )2 , 0, 0 .
(A.6)
We omit the decomposition for I1 , as the corresponding term in (A.4) drops out of all
2 = 0.
future expressions due to km
Finally, using the methods of [15] and the results of Appendix C, the integrals in (A.5)
are found to be, keeping only terms to order O( 0 ),
4
A = (m PL )2 ,
3
2
B = PL (m PL ),
2
C = PL2 ,
D=
(PL2 )3
4
,
8(m PL )
(A.7)
(A.8)
(A.9)
(A.10)
where
2
1
:=
41 ( 12 )
(A.11)
80
Appendix B. Evaluating the integral of C(a, b)

The basic expression which arises in the MHV diagram construction in this paper is
C(a, b) =
il1 j l1 2 il2 2 j l2 iaj b

.
l1 al2 b
ij 4 l1 l2 2
(B.1)
We wish to integrate this expression over the Lorentz invariant phase space. We begin by
simplifying it, using multiple applications of the Schouten identity. First note that using
this identity twice, one deduces that
il2 j l1
l1 b
l2 a
ab2 = iabj + iaaj
+ bj i b
l1 al2 b
a l1
b l2
l1 l2 ab
.
+ aj ib aj ib
al1 bl2
(B.2)
Now use this identity in C(a, b). This generates five terms, which we will label (in correspondence with the ordering arising from the order of terms in (B.2)) as Ti , i = 1, . . . , 4,
and U . The Ti have dependence on the loop momenta such that we may use the phase
space integrals of Appendix C to calculate them. The term U is more complicated; however, one may again use the identity (B.2), generating another five terms, which we will
label T5 , . . . , T8 , and V . Again, the expressions in Ti , i = 4, . . . , 8 may be calculated using
the integrals of Appendix C. Finally, the term V may be simplified, here using the identity (B.2) with i and j interchanged. This generates a further five terms, which we label
T9 , . . . , T13 . The explicit forms of these terms follow:
tr+ (/ij/b/a/ )2 tr+ (/ij/ /l1 /l2 ) tr+ (/ij/ /l2 /l1 )
,
28 (i j )4 (a b)2 (l1 l2 )2
tr+ (/ij/a/b/) tr+ (/ij/b/a/ ) tr+ (/ij/ /l1 /l2 ) tr+ (/ij/ /l2 /l1 ) tr+ (/ib//l1a/ )
,
T2 =
210 (i j )4 (a b)2 (l1 l2 )2 (i b)(a l1 )
j a/ /l2b/)
tr+ (/ij/a/b/) tr+ (/ij/b/a/ ) tr+ (/ij/ /l1 /l2 ) tr+ (/ij/ /l2 /l1 ) tr+ (/
T3 =
,
210 (i j )4 (a b)2 (l1 l2 )2 (j a)(b l2 )
tr+ (/ij/a/b/) tr+ (/ij/b/a/ ) tr+ (/ij/ /l1 /l2 ) tr+ (/ij/ /l2 /l1 )
T4 =
,
28 (i j )4 (a b)2 (l1 l2 )2
T1 =
(B.3)
(B.4)
(B.5)
(B.6)
and
T5 =
tr+ (/ij/b/a/ )2 tr+ (/ij/a/b/) tr+ (/ij/ /l2 /l1 )

,
28 (i j )4 (a b)3 (l1 l2 )
(B.7)
T6 =
tr+ (/ij/a/b/)2 tr+ (/ij/b/a/ ) tr+ (/ij/ /l2 /l1 ) tr+ (/ib//l1a/ )
,
210 (i j )4 (a b)3 (l1 l2 )(i b)(a l1 )
(B.8)
T7 =
tr+ (/ij/a/b/) tr+ (/ij/b/a/ )2 tr+ (/ij/ /l2 /l1 ) tr+ (/ia/ /l2b/)
,
210 (i j )4 (a b)3 (l1 l2 )(i a)(b l2 )
T8 =
tr+ (/ij/a/b/)2 tr+ (/ij/b/a/ ) tr+ (/ij/ /l2 /l1 )

,
28 (i j )4 (a b)3 (l1 l2 )
(B.9)
(B.10)
81
and
T9 =
tr+ (/ij/a/b/)3 tr+ (/ij/b/a/ )

,
28 (i j )4 (a b)4
(B.11)
T10 =
j b//l1a/ )
tr+ (/ij/a/b/)2 tr+ (/ij/b/a/ )2 tr+ (/
,
10
4
4
2 (i j ) (a b) (j b)(a l1 )
(B.12)
T11 =
tr+ (/ij/a/b/)2 tr+ (/ij/b/a/ )2 tr+ (/ia/ /l2b/)

,
210 (i j )4 (a b)4 (i a)(b l2 )
(B.13)
T12 =
T13 =
tr+ (/ij/a/b/)2 tr+ (/ij/b/a/ )2

,
28 (i j )4 (a b)4
tr+ (/ij/b/a/ )2 tr+ (/ij/a/b/)2 tr+ (/b/l2 /l1a/ )

.
210 (i j )4 (a b)4 (a l1 )(b l2 )
(B.14)
(B.15)
The expression C(a, b) is then the sum of the terms Ti , i = 1, . . . , 13.

Before performing the phase space integrals, it proves convenient to collect the resulting
expressions in pairs as T1 + T2 , T3 + T4 , T5 + T6 , T7 + T8 , T9 + T11 and T10 + T12 , we are
led to the following decomposition:
tr+ (/ij/ /l1 /l2 ) tr+ (/ij/ /l2 /l1 ) tr+ (/ij/ /l1a/ ) tr+ (/ij/b//l2 )
28 (i j )4 (l1 l2 )2 (l1 a)(l2 b)
1
(H1 + + H4 ),
= 8
2 (i j )4
C(a, b) =
(B.16)
where

tr+ (/ij/b/a/ ) tr+ (/ij/ /l1 /l2 ) tr+ (/ij/ /l2 /l1 ) tr+ (/ij/ /l1a/ ) tr+ (/ij/ /l2b/)
,
H1 :=
(l1 a)
(l2 b)
(l1 l2 )2 (a b)

tr+ (/ij/a/b/) tr+ (/ij/b/a/ ) tr+ (/ij/ /l2 /l1 ) tr+ (/ij/ /l1a/ ) tr+ (/ij/ /l2b/)
H2 :=
,
(l1 a)
(l2 b)
(l1 l2 )(a b)2

(tr+ (/ij/a/b/))2 tr+ (/ij/b/a/ ) tr+ (/ij/ /l1a/ ) tr+ (/ij/ /l2b/)
H3 :=
,
(l1 a)
(l2 b)
(a b)3
H4 :=
(tr+ (/ij/a/b/))2 (tr+ (/ij/b/a/ ))2 tr+ (/l1a/b//l2 )

.
4(a b)4 (l1 a)(l2 b)
(B.17)
Finally, we perform the phase space integrals of the above expressions, using the formulae in Appendix C. One finds quickly that the divergent (as 0) part of the total
expression is zero. The finite part, after further spinor manipulations, becomes the expression we have given in (4.3).
Appendix C. Phase space integrals

The basic method which we use for evaluating Lorentz-invariant phase space integrals has been outlined in our earlier papers [6,15]. Here we will just quote the results which we need. In the following we will use a shorthand notation where
82
d 42 LIPS(l2 , l1 ; PL ), and a common factor of 4(PL;z )2 is understood to multiply all expressions, where is the ubiquitous factor
2

.
41 12
1
:=
(C.1)
In the following we define = (a P ), = (b P ), N (P ) = (a b)P 2 2(a P )(b P )

and drop the L; z subscripts on PL;z for clarity.
Firstly we quote the results from Appendix B of [15] up terms of order O( 0 ):

1
1
1
1
= ,
=
,
1 = 1,
(a l1 )

(b l2 )

4
1
1
=
+L ,
(C.2)
(a l1 )(b l2 )
N (P )
where

(a b) 2
L = log 1
P .
N
From this, we can derive recursively the following integrals (up to O( 0 )):

1
1
l1 = P ,
l2 = P ,
2
2

1
1
P P P 2 ,
l1 l1 = l2 l2 =
3
4
2
l1
1 P2
P
a
+
=
P 2a ,
(a l1 )
2 2
2
l2
P
1 P2
=
P 2b ,
b
+
(b l2 )
2 2
(C.3)
and

l1 l1
P4
1
P 2 ( ) 3P 4 P 2
=
P
,
a
a
+
P
+
P a
a a
(a l1 )
2
4
4 3
2 2
4 3
l2 l2
P4
1
P2
3P 4 P 2
=
P P 2 P ( b) +
,
b b
b b +
3
(b l2 ) 4
2
4
2
4 3

l2
1
P2 P2
=
2P
a +
b
(a l1 )(b l2 ) N

2L
P
a +
b .
+
N
(a b)
(a b)
(C.4)
Finally, there are integrals involving cubic powers of loop momenta in the numerator. The
first is
83
l1 l1 l1
P2
1
P4
P4
= 3 P ( a a ) + 2 P ( P a ) +
P P P 2 ( a )
(a l1 ) 4
3
4
8
P 2 ( )
P ,
(C.5)
4
where we have suppressed terms cubic in a as they prove not to contribute when this
integral is contracted into the products of Dirac traces which appear in the expressions in
Appendix B. The second cubic integral required is

l2 l 2 l 2
P4
P2
1
P4
= 3 P ( b b) + 2 P ( P b) +
P P P 2 ( b)
(b l2 ) 4
3
4
8
P 2 ( )
P ,
4
(C.6)
again suppressing terms cubic in b which will not contribute.
Appendix D. Spinor identities

We collect here some formulae useful for the calculations presented in this paper. The
Schouten identity is
ij kl = ikj l + ilkj .
(D.1)
Other identities are:

[ij ]j i = tr+ (/ki k/j ) = 2(ki kj ),
(D.2)
[ij ]j l[lm]mi = tr+ (/ki k/j k/l k/m ),
(D.3)
[ij ]j l[lm]mn[np]pi = tr+ (/ki k/j k/l k/m k/n k/p ),
(D.4)
for momenta ki , kj , kl , km , kn , kp . We also have, for null momenta i, j, k, a, b,

tr+ (/ij/b/a/ ) tr+ (/ia/k/b/)
j a/k/b/)
tr+ (/ij/a/b/) tr+ (/
=
.
(j a)
(i a)
(D.5)
For dealing with Dirac traces, we have the following identities15

tr+ (/ki k/j k/l k/m ) = tr+ (/km k/l k/j k/i ) = tr+ (/kl k/m k/i k/j ),
(D.6)
tr+ (/ki k/j k/l k/m ) = 4(ki kj )(kl km ) tr+ (/kj k/i k/l k/m ),
(D.7)
tr+ (/ij/ /
P ) tr+ (/ij/ m)
/ = 0,
(D.8)
tr+ (/ij/ /
P ) tr+ (/ij/m)
/ = 4(i j ) tr+ (/ij/m/
/P ).
(D.9)
15 The appearance of a Greek letter such as inside a trace indicates that the relevant gamma matrix is to be
inserted at that point.
84
References
[1] E. Witten, Perturbative gauge theory as a string theory in twistor space, hep-th/0312171.
[2] F. Cachazo, P. Svrcek, E. Witten, MHV vertices and tree amplitudes in gauge theory, JHEP 0409 (2004)
006, hep-th/0403047.
[3] V.V. Khoze, Gauge theory amplitudes, scalar graphs and twistor space, hep-th/0408233, in: From Fields to
Strings: Circumnavigating Theoretical Physics, in Memory of Ian Kogan, in press.
[4] N. Berkovits, E. Witten, Conformal supergravity in twistor-string theory, JHEP 0408 (2004) 009, hepth/0406051.
[5] F. Cachazo, P. Svrcek, E. Witten, Twistor space structure of one loop amplitudes in gauge theory, JHEP 0410
(2004) 074, hep-th/0406177.
[6] A. Brandhuber, B. Spence, G. Travaglini, One-loop gauge theory amplitudes in N = 4 super-YangMills
from MHV vertices, Nucl. Phys. B 706 (2005) 150, hep-th/0407214.
[7] F. Cachazo, P. Svrcek, E. Witten, Gauge theory amplitudes in twistor space and holomorphic anomaly,
JHEP 0410 (2004) 077, hep-th/0409245.
[8] I. Bena, Z. Bern, D.A. Kosower, R. Roiban, Loops in twistor space, hep-th/0410054.
[9] R. Britto, F. Cachazo, B. Feng, Coplanarity in twistor space of N = 4 next-to-MHV one-loop amplitude
coefficients, hep-th/0411107.
[10] F. Cachazo, Holomorphic anomaly of unitarity cuts and one-loop gauge theory amplitudes, hep-th/0410077.
[11] R. Britto, F. Cachazo, B. Feng, Computing one-loop amplitudes from the holomorphic anomaly of unitary
cuts, hep-th/0410179.
[12] Z. Bern, V. Del Duca, L.J. Dixon, D.A. Kosower, All non-maximally-helicity-violating one-loop sevengluon amplitudes in N = 4 super-YangMills theory, hep-th/0410224.
[13] S.J. Bidder, N.E.J. Bjerrum-Bohr, L.J. Dixon, D.C. Dunbar, N = 1 supersymmetric one-loop amplitudes
and the holomorphic anomaly of unitarity cuts, hep-th/0410296.
[14] S.J. Bidder, N.E.J. Bjerrum-Bohr, D.C. Dunbar, W.B. Perkins, Twistor space structure of the box coefficients
of N = 1 one-loop amplitudes, hep-th/0412023.
[15] J. Bedford, A. Brandhuber, B. Spence, G. Travaglini, A twistor approach to one-loop amplitudes in N = 1
supersymmetric YangMills theory, Nucl. Phys. B 706 (2005) 100, hep-th/0410280.
[16] C. Quigley, M. Rozali, One-loop MHV amplitudes in supersymmetric gauge theories, hep-th/0410278.
[17] Z. Bern, L.J. Dixon, D.C. Dunbar, D.A. Kosower, One loop N point gauge theory amplitudes, unitarity and
collinear limits, Nucl. Phys. B 425 (1994) 217, hep-ph/9403226.
[18] L.J. Dixon, E.W.N. Glover, V.V. Khoze, MHV rules for Higgs plus multi-gluon amplitudes, JHEP 0412
(2004) 015, hep-th/0411092.
[19] Z. Bern, L.J. Dixon, D.C. Dunbar, D.A. Kosower, Fusing gauge theory tree amplitudes into loop amplitudes,
Nucl. Phys. B 435 (1995) 59, hep-ph/9409265.
[20] Z. Bern, L.J. Dixon, D.A. Kosower, One-loop corrections to five-gluon amplitudes, Phys. Rev. Lett. 70
(1993) 26772680, hep-ph/9302280.
[21] L.J. Dixon, Calculating scattering amplitudes efficiently, TASI Lectures 1995, hep-ph/9601359.
[22] R. Roiban, M. Spradlin, A. Volovich, A googly amplitude from the B-model in twistor space, JHEP 0404
(2004) 012, hep-th/0402016.
[23] N. Berkovits, An alternative string theory in twistor space for N = 4 super-YangMills, hep-th/0402045.
[24] R. Roiban, A. Volovich, All googly amplitudes from the B-model in twistor space, hep-th/0402121.
[25] R. Roiban, M. Spradlin, A. Volovich, On the tree-level S-matrix of YangMills theory, Phys. Rev. D 70
(2004) 026009, hep-th/0403190.
[26] N. Berkovits, L. Motl, Cubic twistorial string field theory, JHEP 0404 (2004) 056, hep-th/0403187.
[27] S. Gukov, L. Motl, A. Neitzke, Equivalence of twistor prescriptions for super-YangMills, hep-th/0404085.
[28] C.J. Zhu, The googly amplitudes in gauge theory, JHEP 0404 (2004) 032, hep-th/0403115.
[29] G. Georgiou, V.V. Khoze, Tree amplitudes in gauge theory as scalar MHV diagrams, JHEP 0405 (2004) 070,
hep-th/0404072.
[30] J.-B. Wu, C.-J. Zhu, MHV vertices and scattering amplitudes in gauge theory, hep-th/0406085.
[31] J.-B. Wu, C.-J. Zhu, MHV vertices and fermionic scattering amplitudes in gauge theory with quarks and
gluinos, hep-th/0406146.
85
[32] I. Bena, Z. Bern, D.A. Kosower, Twistor-space recursive formulation of gauge theory amplitudes, hepth/0406133.
[33] D. Kosower, Next-to-maximal helicity violating amplitudes in gauge theory, hep-th/0406175.
[34] G. Georgiou, E.W.N. Glover, V.V. Khoze, Non-MHV tree amplitudes in gauge theory, JHEP 0407 (2004)
048, hep-th/0407027.
[35] G. Passarino, M.J.G. Veltman, One loop corrections for e+ e annihilation into + in the Weinberg
model, Nucl. Phys. B 160 (1979) 151.
Neutralino dark matter detection in split

supersymmetry scenarios
A. Masiero a , S. Profumo b,c,d , P. Ullio c,d
a Dipartimento di Fisica G. Galilei, Universit di Padova, and Istituto Nazionale di Fisica Nucleare,
Sezione di Padova, Via Marzolo 8, I-35131 Padova, Italy

b Department of Physics, Florida State University, 505 Keen Building, FL 32306-4350, USA
c Scuola Internazionale Superiore di Studi Avanzati, Via Beirut 2-4, I-34014 Trieste, Italy
d Istituto Nazionale di Fisica Nucleare, Sezione di Trieste, I-34014 Trieste, Italy
Received 8 December 2004; received in revised form 19 January 2005; accepted 19 January 2005
Abstract
We study the phenomenology of neutralino dark matter within generic supersymmetric scenarios where the gaugino and higgsino masses are much lighter than the scalar soft breaking masses
(split supersymmetry). We consider a low-energy model-independent approach and show that the
guidelines in the definition of this general framework come from cosmology, which forces the lightest neutralino to have a mass smaller than 2.2 TeV. The testability of the framework is addressed
by discussing all viable dark matter detection techniques. Current data on cosmic rays antimatter,
gamma-rays and on the abundance of primordial 6 Li already set significant constraints on the parameter space. Complementarity among future direct detection experiments, indirect searches for
antimatter and with neutrino telescopes, and tests of the theory at future accelerators, such as the
LHC and a NLC, is highlighted. In particular, we study in detail the regimes of winohiggsino mixing and binowino transition, which have been most often neglected in the past. We emphasize that
our analysis may apply to more general supersymmetric models where scalar exchanges do not provide the dominant contribution to annihilation rates.
PACS: 12.60.-i; 95.35.+d
E-mail addresses: antonio.masiero@pd.infn.it (A. Masiero), profumo@hep.fsu.edu,

profumo@sissa.it (S. Profumo), ullio@sissa.it (P. Ullio).
doi:10.1016/j.nuclphysb.2005.01.028
A. Masiero et al. / Nuclear Physics B 712 (2005) 86114
87
1. Introduction
Soon after the first explicit formulations of supersymmetric (SUSY) versions of the
Standard Model (SM) [1], enforcing R-parity conservation to prevent sources of violation
of baryon and lepton number at tree-level, it was realized, as an interesting by-product, that
these theories could naturally embed a candidate for cold dark matter [2]. In fact, it had
been shown long before that weakly-interacting massive particles (WIMPs) have a thermal
relic abundance that is generically at a level relevant for cosmology, and the stable LSP
falls under this category if it happens to be electric and color neutral (and as it is indeed
the case in large portions of the SUSY parameter space).
Two decades later, no SUSY particle has unfortunately been found at particle accelerators and no clean indication of SUSY has emerged from tests of rare processes or precision
measurements. The initial theoretical motivations for SUSY, although still very appealing,
are being questioned, with the scale of physics beyond the SM that is slowly, but steadily,
drifting above the weak scale. On the other hand, in these twenty years, the case for nonbaryonic cold dark matter (CDM) being the building block of the Universe has become
stronger and stronger, with the determination of the CDM contribution to the Universe energy density which has already reached a 20% level of accuracy [3] and is going to improve
further in the upcoming years. It is then not surprising that in formulating an extension to
the SM, sometimes, the issue of incorporating a dark matter candidate has changed from
being a by-product of the proposed scenario into being one of the hinges of the theory itself.
This is the case for a recently proposed SUSY framework, dubbed split supersymmetry
[4,5].
Split supersymmetry indicates a generic realization of the SUSY extension to the SM
where fermionic superpartners feature a low mass spectrum (say at the TeV scale or lower),
while scalar superpartners are heavy, with a mass scale which can in principle range from
hundreds of TeV up to the GUT or the Planck scale [4]. In split SUSY one gives up on the
idea that SUSY is the tool to stabilize the weak scale, since some other mechanism should
anyway account for other fine tuning issues, such as that of the cosmological constant.
On the other hand, a number of phenomenological problems appearing in ordinary SUSY
setups, where the whole spectrum of superpartners is supposed to be light, are cured: the
heavy sfermions minimize flavor and CP violating effects, alleviate proton decay and avoid
an excessively light Higgs boson. At the same time, two major features are maintained: the
successful unification of gauge couplings, andthe central issue in our analysisthe LSP
as a viable particle candidate for CDM. Finally, it is certainly a well-motivated scenario,
as it has been shown that the occurrence of a split SUSY spectrum is indeed a generic
feature of a wide class of theories [69]. It arises, among others, in frameworks where the
mechanism of SUSY breaking preserves a R-symmetry and forbids gaugino and higgsino
mass terms, or in case of direct SUSY breaking through renormalizable interactions with
the SUSY breaking sector (direct mediation) [6,10,11].
Rather than focusing on some specific realizations, we consider here a generic split
SUSY setup: we refer to the minimal SUSY extension to the Standard Model (MSSM),
set the scalar sector at a heavy scale, but allow for the most general fermion sector. We
investigate how cosmological constraints affect the definition of such framework and discuss its testability by examining the issue of dark matter detection within this wide class
88
of models, comparing also with the perspectives to test this scenario at accelerators. As a
result, we work out here a fully general analysis of neutralino dark matter on the basis of
the lightest neutralino mass and composition, largely independent of the split supersymmetry framework; in this respect, we outline a reference guide to the sensitivity of upcoming
detection experiments which applies, especially as fas as indirect detection is concerned,
to supersymmetric models with a generic (moderatly) heavy scalar sector. Once the LSP
composition and mass is known, our results allow to get a fair estimate of the detection
prospects.
The paper is organized as follows: in Section 2 we describe our working framework
and discuss the current constraints to its parameter space. In Section 3 we introduce the
relevant dark matter detection techniques and show rates in extremal regimes of the parameter space. In Section 4 we discuss in detail the phenomenology of the split SUSY
setup along a sample slice in the parameter space, show constraints from current data and
address detectability in future searches, cross-checking the outreach of the different techniques. In Section 5 we indicate how to generalize the results of the previous section to the
full parameter space. Finally, we draw an outlook and conclude in Section 6.
2. Definition of the model and parameter space

We consider a MSSM setup, with the relevant parameters defined at the low energy
scale and without implementing any unification scheme. We restrain to the split SUSY
framework by assuming that the sfermion sector is much heavier than the fermion sector
and hence is completely decoupled; as for the Higgs sector, we assume there is only one
light SM-like Higgs, with a mass labeled by the parameter mh , while the other two neutral
Higgs and the charged Higgs are taken again to be very heavy, and hence decoupled. Finally, we will leave out of most of our discussion the gluino and the gravitino, supposing
they are (moderately) heavy, just mentioning what are possible consequences in case this
is not true.
In this scheme, the lightest neutralino, defined as the lightest mass eigenstate from the
superposition of the two neutral gaugino and the two neutral higgsino fields,
10 = N11 B + N12 W 3 + N13 H 10 + N14 H 20 ,
(1)
is automatically the LSP.1 The coefficients N1j , obtained by diagonalizing the neutralino
mass matrix, are mainly a function of the bino and the wino mass parameters M1 and M2 ,
and of the Higgs superfield parameter , while depending rather weakly on tan , the
ratio of the vacuum expectation values of the two neutral components of the SU(2) Higgs
doublets. The hierarchy between M2 and determines also whether the lightest chargino
is wino like or higgsino like, and again the role tan is minor. The phenomenology of the
1 The lightest chargino is always heavier than the lightest neutralino at tree level; this mass hierarchy is more-
over stable, in the parameter space regions under scrutiny, against the inclusion of radiative corrections in the
relevant mass matrices [12].
89
scheme we are considering is hence fully defined by only five parameters:

M1 ,
M2 ,
tan
and mh .
(2)
In this construction, constraints from accelerator data are rather weak, essentially just in
connection with the fact that charginos have not been observed so far: we implement, as a
conservative limit on the lightest chargino mass, the kinematic limit from the last phase of
LEP2, i.e., m + > 103.5 GeV [13] (notice however that in large portions of our parameter
space the lightest neutralino and lightest chargino are nearly degenerate in mass, and this
limit should be slightly relaxed; however, this does not critically enter in our discussion).
Much more powerful are constraints on the model from cosmology; to implement the
cosmological bounds, we refer to the determination of the CDM component of the Universe by WMAP [3]: CDM h2 = 0.113 0.009. LSP relic densities are computed with
the DarkSUSY package [14], which allows for high accuracy solutions of the equations
describing thermal decoupling. We study the parameter space varying freely the mass parameters M1 , M2 and (no GUT relations are assumed between M1 and M2 ; we restrict
to positive , since in this contest the case of negative is just specular), and taking the
sample value tan = 50 and mh = 115 GeV (close to the current lower limit on the mass
of a SM-like Higgs).
In Fig. 1 we consider slices, in the parameter space, along the planes (M1 , M2 ) (lefthand side panel) and (M1 , ) (right-hand side panel) and for large values of the third mass
parameter (which we set to 10 TeV). In the first case, since M1 and M2 are both diagonal
entries in the neutralino mass matrix, the transition of the LSP from being bino like (above
the diagonal in the figure) to being wino like (below it) is very sharp. The two regimes are
very different because winos annihilate very efficiently into gauge bosons, while binos can
annihilate just into fermions. The latter annihilation processes are very sharply suppressed,
as compared to ordinary SUSY setups, since diagrams with t- and u-channel sfermions
exchanges do not contribute here. In general, the relic density of the LSP scales with the
inverse of the annihilation rate: it is very large for binos regardless of the LSP mass, while
it tends to be too small for winos, unless one considers M2 in the 2 TeV range. The exception is in a narrow strip above the diagonal where the bino LSP is nearly degenerate
in mass with the next-to-lightest wino like neutralino and chargino states, and the thermal
equilibrium of the LSP in the early Universe is enforced through these slightly heavier particles rather than by just LSP pair annihilations into SM particles: this is a very well-studied
effect, usually dubbed coannihilation [15], but that has been very seldom considered in
this specific realization, since usually M1 and M2 are not taken as independent parameters.
The region in the plane where the LSP relic abundance is too large is shaded. Starting from
light LSPs (the mass of the LSP is essentially the minimum between M1 and M2 ) in the
bottom-left corner, the border of the excluded region approaches the diagonal, since increasing the mass scale the cross-sections decreases, and hence coannihilation effects have
to become larger and larger, i.e., mass splitting with coannihilating particles smaller and
smaller. We reach then the mass value at which this process is saturated, and the LSP turns
into wino like; above this scale, even for winos, the annihilation cross-section becomes too
small and CDM is overproduced. In the figure, since we are plotting results on logarithmic
scales, the iso-level curve at CDM h2 = 0.113 is essentially overlapping with the border
90
of the excluded region, while in the remaining part of the plane, CDM is underproduced.
In Figs. 1 and 2 we shade in red the regions with an excessively light chargino.
For the right-hand side panel of Fig. 1, the picture is analogous. The wino mass parameter is assumed to be heavy, hence we are left with the possibility that the LSP is either
bino like or higgsino like, depending on the relative values of M1 and . Being the parameter in off-diagonal entries in the neutralino mass matrix, the transition between the
two regimes is smoother than in the previous case, but the trend is the same, with higgsinos
annihilating efficiently into gauge bosons and binos with suppressed annihilation rates. In
Fig. 1. The parameter space region in the (M1 , M2 ) and (M1 , ) plane allowed by direct chargino searches at
LEP and by the requirement that the neutralino relic abundance does not exceed the CDM density. The third
parameter (respectively, (left) and M2 (right)) is always set to 10 TeV. The excluded regions are, respectively,
shaded in red and light blue. (For interpretation of the references to colour in all figures legends, the reader is
referred to the web version of this article.)
Fig. 2. Left: the allowed regions in the (M2 , ), with the same color code as in Fig. 1; the cosmological limit
is indicated for various values of M1 . Right: a 3D representation of the hypersurface at h2 = 0.113 in the
(M1 , M2 , ) space.
91
the computation of the relic density both the effects of mixing and coannihilations enter,
and, for instance, a wiggle in the borderline of the excluded region corresponding to the
tt threshold is clearly visible. This is the scenario which has been considered up to now
for estimates of the LSP relic abundance in split SUSY contexts [5,6,16].
The value of the higgsino parameter, hence of the higgsino mass, saturating the relic
density constraint is smaller than for winos. This is illustrated also in the left-panel of
Fig. 2, where iso-level curves for CDM h2 = 0.13 are shown in the plane (M2 , ) and
for a few values of M1 . In case M1 is very heavy, we see the transition of the LSP from
being a pure wino (upper part of the figure) to being a pure higgsino (right-hand side). The
size of the saturating parameters
differs because the coupling in the most important vertex,
namely W + 10 i+ , reads g/ 2 for higgsinos and g for winos, but also because the number
of degrees of freedom in the coannihilation process in the two cases is different: as far as
higgsinos are concerned, the coannihilating particles are two neutralinos and one chargino,
all nearly degenerate in mass, while for winos only one neutralino and one chargino enter
(and this matters, because pair annihilation rates of charginos are larger than those for
neutralinos). We find that a pure higgsino of mass equal to 1030 GeV and a pure wino
of mass 2210 GeV have a relic abundance equal to CDM h2 = 0.113, the current best fit
value from the WMAP data.
Lower values of M1 imply smaller upper limits on the neutralino mass, corresponding,
as sketched before, to bino LSPs coannihilating either with winos or higgsinos, with the
transition regimes getting progressively wider as M1 decreases. A 3D representation of
the hypersurface at h2 = 0.113 in the (M1 , M2 , ) space is given in the right-panel of
Fig. 2; the allowed low relic density region stands behind the plotted surface.
The value of tan and mh have not played much of a role in our discussion, since we
expect very minor changes in the cosmologically allowed portion of the parameter space if
these are varied (and we have indeed verified numerically that this is the case). Such choice
of parameters is actually crucial only for direct detection rates, as we will stress below.
Finally, the bounds we have derived are valid under the assumption of a standard cosmological setup, and including only thermal sources of dark matter. The bounds get tighter,
or, from another perspectives, models that give here subdominant CDM candidates may
become fully consistent with observations, if one introduces particle physics models with
non-thermal sources of CDM, such as neutralino productions from gravitinos or moduli
decays [17], or considers cosmological scenarios with faster expansion rates at the time of
decoupling: sample cases are cosmologies with a quintessence energy density term dominating at the LSP freeze-out temperature [1820], anisotropic Universes with effective
shear density terms [20,21] or in scalar-tensor theories [22]. Relaxing the upper bound is
also possible, though maybe slightly more contrived, as it would involve wiping out (totally, or in part) the thermal relic component with an entropy injection: a mechanism of
reheating at low temperature, i.e., lower than the LSP freeze-out temperature which is in
the range m /20m /25, would give this effect [23]. However, excluding this latter possibility, it is interesting to notice that, in this generic split SUSY setup, the cosmological
bound is indeed the only issue forcing one of the sectors of the theory to be light, i.e., at a
scale lower than 2.2 TeV.
92
3. Dark matter detection rates

The issue of WIMP dark matter detection has been studied at length (for a review, see,
e.g., [24,25]; see also the more recent works, e.g., Refs. [2628]). We will show here that
this is especially relevant in the split SUSY scenario by systematically discussing all WIMP
detection techniques that currently provide, or that will provide in the future, constraints
on the model. Rates are computed with the DarkSUSY package [14]; the set of underlining
assumptions will be briefly described here. In this section we will present sample rates
for dark matter candidates in the three extreme regimes described above, and focusing on
models with thermal relic densities in the WMAP preferred range [3], in case of a standard
cosmological setup. In the next two sections, these results will be generalized, including
also low relic density models, either by relaxing the requirement of thermal production
of the whole CDM content of the Universe or by referring to non-standard cosmological
scenarios; at that stage we will also put more emphasis on the comparison between different
detection techniques.
In the last decade there has been a considerable experimental effort to detect WIMPs
directly, measuring the recoil energy from WIMPs elastic scattering on nuclei [29]. Our
theoretical predictions are derived with the usual effective Lagrangian approach, except
that no contributions mediated by squarks are present, and taking a standard set of parameters [30,31] for nucleonic matrix elements (note, e.g., the strange content here is
slightly smaller than the values implemented in other analyses, see [14,32] for details).
In the left panel of Fig. 3 we plot the spin-independent (SI) neutralinoproton scattering
SI , as a function of the neutralino mass. The thick lines label models on the
cross section P
2
h = 0.113 isolevel curves in Figs. 1 and 2, left, i.e., the regimes with one fermion mass
parameter heavier than the other two. From the upper left side of figure, along a dashed
Fig. 3. Direct, spin-independent (left) and spin-dependent (right) neutralinoproton scattering cross-sections
along the lines at h2 = 0.113, obtained in the (M1 , M2 , ) space setting one of the parameters to 10 TeV
and varying the other two. The dotted black lines and the green dot-dashed lines correspond to the current and
projected experimental limits. In the spin independent scattering cross sections plot, left panel, we also shade the
region corresponding to the whole parameter space points giving the correct relic abundance, i.e., the hypersurface
of Fig. 2, right panel.
93
SI for a light LSP with significant binohiggsino mixing turning into a heavline, we plot P
ier pure higgsino state; along the dasheddotted curve the higgsino state makes a transition
into a pure wino, going through models with large higgsinowino mixing, and finally on
the dotted line the parameter is large and there is just a very sharp transition from pure
winos to pure binos when moving again to smaller masses. Since the SI scattering cross
section is mediated by the light Higgs in a t-channel, and the 10 10 h vertex scales with the
higgsinogaugino mixing, whenever this is small the cross section is small as well, as it
can be clearly seen in the figure. The gray-shaded region corresponds to the projection of
all points lying onto the hypersurface at h2 = 0.113 in the (M1 , M2 , ) space (the hypersurface plotted in Fig. 2, right). Not surprisingly, this region is close to being the zone
delimited by the curves labeling the three extreme regimes just described (and the same
holds for all other detection techniques we discuss in the present section, hence, for the
sake of clarity, we omit the corresponding shadings). The black dotted line indicates the
SI from the CDMS Collaboration [33], while the green dot-dashed line
current bound on P
stands for the projected sensitivity of next-generation (sometimes dubbed stage 3) detectors; for definiteness, we take as a benchmark experimental setup the XENON facility [34].
As it can be seen, the present sensitivity does not allow to set bounds on these models, but
future projects will test a large portion of the parameter space.
In the right panel of Fig. 3 we show the spin-dependent neutralinoproton cross sections,
along the same parameter space slices as in the left panel. We also indicate the current
[35] and projected [36] experimental sensitivities in this detection channel. As already
pointed out [28], spin-dependent direct neutralino searches in the general MSSM appear
to be largely disfavored with respect to spin-independent ones: future facilities will reach
a sensitivity which is expected to be various orders of magnitude below the theoretical
expectations for the models under consideration.
The search for neutrinos produced by the annihilation of neutralinos trapped in the core
of the gravitational wells of the Sun or of the Earth has been since long recognized as a
promising indirect detection technique. In the present framework, fluxes from the Earth
are actually very low and will not be considered further; to estimate neutrino fluxes from
the Sun we implement the standard procedure described in Refs. [25,37], except for a
more careful treatment of neutralino capture rates [14,38]. In the left panel of Fig. 4, we
present results in terms of muon fluxes, above the threshold of 1 GeV, and compare them
to the current best limits from the Super-Kamiokande Collaboration [39] and with the
future projected sensitivity of the IceCube experiment [40] (the mismatch in the energy
threshold of IceCube and the threshold considered here has been taken into account). The
color coding on the isolevel curves is the same as in the left panel, making it transparent
that the emerging pattern is perfectly specular. The neutrino flux in the Sun scales with a
capture rate which is dominated by the spin dependent (SD) neutralinoproton coupling,
and, as for the SI term, this is suppressed in models with small higgsinogaugino mixing
(and the relevant vertex here is 10 10 Z 0 ). While the sensitivity of current direct dark matter
searches is more than one order of magnitude above the largest signals we obtain in the
present framework, in the lowest mass range ( 100 GeV) neutrino telescopes already rule
out a few models. However, future direct detection experiments will probe a much wider
portion of the parameter space, as compared to what IceCube is expected to achieve.
94
The results shown in Figs. 3 and 4 are derived assuming, as in the previous section,
SI scales with m4 , while the dependence on m
that tan = 50 and mh = 115 GeV. P
h
h
of the neutrino fluxes from the Sun is negligible. tan does not enter critically in any
of the relevant couplings, however it slightly affects the higgsinogaugino mixing. We
sketch a sample effect of this kind in the right panel of Fig. 4, where the main point is
however to illustrate the reason why the rates in the cases considered so far change sharply
at m 1.51.7 TeV, in the binowino transition along the isolevel curve at large
(red dotted lines). The figure shows that going from the regime M2 < M1 to the regime
M1 < M2 , there is not only a sharp switch in the gaugino content, but also a step-like fall
in the (already small) higgsino fraction. As we show, a slight dependence on tan (which
enters in the neutralino mass matrix) is present. We will further discuss this effect in the
next section.
Other indirect detection techniques rely on the fact that there is a finite probability for
dark matter WIMPs to annihilate in pairs in galactic halos, and in particular in the halo of
the Milky Way, possibly giving rise to exotic sources of radiation and cosmic rays. In this
analysis we will consider the production of gamma-rays, antiprotons, positrons and antideuterons within our own Galaxy. The yield per annihilation in all these channels is read
out of simulations with the PYTHIA [41] 6.154 Monte Carlo code as included in the Dark where we use the prescription suggested in Ref. [42]. Since
SUSY package, except for D,
the source strength scales with the number density of neutralino pairs locally in space, i.e.,
in terms of the dark matter density profile, with 1/2( (
x )/m )2 , the predictions for these
signals are very sensitive to which (
x ) is considered for the Milky Way. In fact, the latter
is unfortunately not well known, and one is forced to make some assumptions.
In this analysis, we will mainly focus on two extreme models for the Milky Way dark
matter halo. They are both inspired to the current picture emerging from N -body sim-
Fig. 4. Left: the neutralinoannihilations-induced muon flux from the Sun along the lines at h2 = 0.113,
obtained in the (M1 , M2 , ) space setting one of the parameters to 10 TeV, and varying the other two. The dotted
black lines and the green dot-dashed lines correspond to the current and projected experimental limits. Right: the
higgsino fraction, as a function of the parameter M2 , at M1 = 500 GeV and = 2 TeV, for two values of tan ,
respectively, 5 (solid blue line) and 50 (red dashed line).
95
ulations of hierarchical clustering in CDM cosmologies, respecting, e.g., the relation

between halo mass and concentration parameter suggested therein. The two models diverge however in the way the effect of the baryon infall in the dark matter potential well
is sketched. One extreme it to suppose that this process happened with a large transfer of
angular momentum between the baryons and the dark matter. In this case the system of
cold particles, sitting at the center of dark matter halo and forming the cusp seen in the
simulations (which describe the evolution of the collision-less dark halo component only),
are significantly heated and removed from the central part of the halo. One such extreme
scenario was proposed in Ref. [43], suggesting that the resulting dark halo profile can be
modeled by the so-called Burkert profile [44]:
B (r) =
B0
.
(1 + r/a)(1 + (r/a)2 )
(3)
This profile has been shown to provide fair fits in case of a large sample of the rotation
curves for spiral galaxies [45], and, from our perspective, having a very large core radius
(the parameter a), is very conservative. We will consider a sample choice of the free parameters in Eq. (3), fixing the length scale parameter a = 11.7 kpc and the local halo density
B (r0 ) = 0.34 GeV cm3 . The second model we consider stands on the other extreme, and
it is derived by supposing that the settling in of the baryons happened with no net transfer
of angular momentum between baryons and dark matter. The baryon infall happened as
a slow and smooth process which gradually drove the potential well at the center of the
Galaxy to become deeper and deeper, accreting more and more dark matter particles in
this central region. This is the adiabatic contraction limit, which we implement according to the prescription of Blumenthal et al. [46] (circular orbit approximation). We start
from the CDM profile of the non-singular form extrapolated in Ref. [47] (which we label
as N03 profile), assume as sample virial mass and concentration parameter, respectively,
Mvir = 1.8 1012 M and cvir = 12, derive the enhancement on it induced by the stellar
bulge and the disc components, and just cut the profile in the region which is dynamically
dominated by the central black hole (i.e., we assume complete relaxation due to the black
hole formation). The resulting spherical profile has a local dark matter density equal to
N03 (r0 ) = 0.38 GeV cm3 . The choice of parameters in both scenarios is justified by testing the dark matter model against available dynamical constraints. Finally, a self-consistent
velocity distribution is derived as well, and the result for the Burkert profile has already
been exploited above when showing sensitivity curves for direct detection and computing
capture rates in the Sun (and both of these change are essentially unchanged in case the
adiabatically-contracted N03 profile is chosen). Further details on the definition of the halo
models can be found in Refs. [38,48,49].
Antimatter searches do not play a main role for the models we have considered so far,
since we find that the predicted fluxes are well below both current constraints and the sensitivity of next-generation experiments; at the same time, they look as the best chance for
detection in other regions of the parameter space, hence we postpone the relative discussion. On the other hand, data on the gamma-ray flux in the Galactic center (GC) direction
are already relevant to set combined constraints on our particle physics framework and the
dark halo profile. Gamma-ray maps of the GC have been derived both from space, by the
EGRET experiment on the Compton Gamma-Ray Observatory, below 10 GeV [50], and,
96
more recently, from ground by three different air Cherenkov telescopes (ACTs), with an
overall energy coverage between 300 GeV and 8 TeV, and apparently discrepant results
concerning spectral shapes of the source and normalization of the flux [5153]. An interpretation of one of these signals in term of WIMP pair annihilations is possible, see, e.g.,
the modeling of the EGRET source in Ref. [54], while most probably it is not possible
to fit with such a source all of them simultaneously. We take here a conservative attitude
and just assume that the GC gamma-ray flux produced in our model should not exceed the
measured one; of course, in case it were demonstrated that these fluxes were associated
to another source, possibly not even located in the GC, as it has already been argued for
the EGRET source [55], the constraints we derive here would get stronger. A clearer statement on this respect should come with the next gamma-ray mission in space, the GLAST
satellite [56].
In Fig. 5, left panel, we plot, with a black dotted horizontal line, the most stringent 2
bound from the EGRET data set [50] (which turned out to be the bin at the largest energy
4 GeV < E < 10 GeV), compared to the fluxes in this energy bin and computed for the
cuspy, adiabatically contracted N03 model (upper lines) and cored Burkert profile (lower
lines). The fluxes are averaged over the EGRET angular resolution 1.5 . Going to larger
masses the expected flux decreases both because of the 1/m2 scaling in the number density of neutralino pairs and because we are looking at a rather low energy bin; the trend
is partially smoothed, especially for winos, by the transition along the isolevel curves
into states that annihilate more and more efficiently into W bosons (for pure winos this is
maximal). Note also, even in this case, the sharp drop of the signal for pure binos, which
is now due to the fact that they can annihilate just into fermions and with very suppressed
rates (of course coannihilation effects make a compensation in the effective thermally averaged annihilation cross section setting the relic abundance in the early Universe, but not
in the zero temperature cross section for annihilation for lightest state pairs, which is the
relevant quantity here). At higher energies we consider the constraint from the recent mea-
Fig. 5. Current constraints on the points at h2 = 0.113 from gamma-rays observations at the center of
our Galaxy, as performed by EGRET (left) and by HESS (right). The upper solid lines correspond to the
N03-adiabatically contracted halo model [47], while the lower dotted lines to the Burkert profile [44].
97
surement by the HESS telescope, which has the largest statistics among recently reported
results [53]. Since the flux detected by HESS is surprisingly hard, the data point setting the
strongest 2 bound is the one corresponding to the smallest energy bin, E = 281 GeV.
The predicted fluxes at this energy are shown in Fig. 5, right panel; in this case they are averaged over a 5.8 cone around the GC, corresponding to the HESS angular resolution for
this source. Obviously the fluxes are zero for neutralino masses below the energy we are
considering; above that mass, the 1/m2 scaling in the number density of neutralino pairs is
compensated (except for the pure wino branch in the heavy case) by the fact that we are
considering a high energy bin, in which the pile up of the produced photons get efficient
just for heavy masses, and again by the fact that the maximal pair annihilation rate involves
pure winos. Note also the huge span in the predictions between the two halo models we are
considering, even larger than in the case of the EGRET data, since the angular acceptance
here is smaller.
4. A case study
We are ready to examine in more detail the role of dark matter searches in the split SUSY
framework. The approach we follow here is to choose a direction in the parameter space,
to postulate that these selected models provide the LSP as a dark matter candidate (i.e.,
relaxing the hypothesis that the thermal relic abundance of s in a standard cosmological
setup should match the WMAP preferred range) and to check their detectability.
In this section we concentrate, if not differently specified, on a case study where we take
the soft breaking mass of the hypercharge gaugino M1 = 500 GeV, tan = 50 and mh =
115 GeV. This corresponds to a foliation of the parameter space along (, M2 ) planes; we
consider a scan on these two parameters varying them in the range from 100 GeV to 10 TeV
(we remind again that no GUT relation is assumed, and that M1 , M2 and are considered
as independent parameters). This sample case is particularly illuminating, because, within
a simple and readable setting of the relevant low-energy parameters in the neutralino sector,
it allows to investigate the full set of neutralino compositions. The pattern of mixing and
composition for the LSP is schematically illustrated in Fig. 6. We give here a detailed, and,
to our knowledge novel, analysis of the physics occurring when the wino and higgsino
mixing is large (violet strip in Fig. 6), as well as an analysis of the winobino edge (green
strip, in Fig. 6).
We present our results by resorting to visibility ratios (VRs), i.e., ratios of expected
signals to projected (or current) sensitivities. A model featuring a VR > 1 in a given search
channel will therefore be within the projected reach of (or excluded by) the corresponding
experimental facility.
We start with the existing bounds from measurements of the gamma-ray flux towards
the GC. Fig. 7 shows the exclusion curves one draws from the EGRET results, under the
hypothesis of subdominant background (see Section 3). The VR is defined as:

VREGRET
(4)
.
+ 2 E [4,10] GeV
98
Here and in the analogous figures below, on the left we show 3D plots with values of VR on
the (M2 , ) plane, on the right the corresponding isolevel curves. The cuspy adiabatically
contracted N03 profile is assumed in the plots; fluxes and VRs get a factor of 104 smaller
in case of the cored Burkert halo. The sharp decrease in the low masses regions, at M2
100 GeV, is due to the gauge boson threshold, i.e., when m < mW , in which case
the gamma-rays production cross-section suddenly decreases. We recall that this region
Fig. 6. A sketch of the neutralino composition in the case study plane (M2 , ) at M1 = 500 GeV. The plane
collects a sample of all possible mixing patterns in the composition of the lightest neutralino within the MSSM.
Fig. 7. The 95% C.L. exclusion ratio from the EGRET data (namely, the gamma-ray flux from neutralino annihilation over the maximal allowed observed gamma-ray flux) on the (M2 , ) plane at M1 = 500 GeV. The
N03-adiabatically contracted halo model is assumed. In the right panel we show the corresponding iso-level
curves.
99
is, however, already ruled out by chargino searches at LEP, see Fig. 2, right. The largest
production of gamma rays occurs in the light M2 region, and it does not depend either
on M1 or . In the region of bino like neutralino, the VRs sensibly decrease, along with
the decrease in the higgsino fraction, and the expected fluxes are always well below the
measured signal. The EGRET data also constrain pure higgsinos, up to masses around
350 GeV.
The VR from the HESS measurement is defined analogously, except that now in Eq. (4)
the reference energy bin is the lowest in the HESS dataset, i.e., the one at 281 GeV; this sets
the threshold in VR at approximately 300 GeV. Once again, the largest signal is obtained
for wino like LSPs, see Fig. 8. Relaxing the M1 = 500 GeV assumption, we find that
sufficiently pure winos with masses up to around 2 TeV are excluded by HESS data, for
the cuspy halo model we have considered (but are perfectly viable in case of the cored
profile, since fluxes are suppressed, in that case, by more than six orders of magnitude).
VRs for higgsinos are smaller, but still many configurations are excluded, while again binos
give very small fluxes.
Turning to spin-independent direct detection, we take here, as a benchmark of next
generation search experiments, the future reach of the XENON-1ton facility [34]. The
SI ) is shown in the left frame of
corresponding projected sensitivity curve in the (m , P
Fig. 3; the VR is defined here as the ratio between the expected cross section and the
projected sensitivity, at the corresponding neutralino mass.
Fig. 9 highlights again the role of gauginohiggsino mixing in direct detections
SI (Z Z )2 , with Z and Z , respectively, the lightest neutralino higgsino
searches: P
h g
h
g
and gaugino fractions; their product is maximized on the ridge along the M2 line (explaining the shape of the VR = 100 isolevel curve) and has a line of local maxima around
M1 = 500 GeV. Switching from a bino like to a wino like lightest neutralino, e.g., on
Fig. 8. The 95% C.L. exclusion ratio from the HESS data on the (M2 , ) plane at M1 = 500 GeV, with the
N03-adiabatically contracted halo model. The right panel shows the corresponding iso-level curves.
100
lines at constant > 500 GeV, the scattering cross section is enhanced by one order of
magnitude or even more. This can be retraced to the sudden increase in the higgsino fraction when the neutralino becomes Wino like we have shown in the right panel of Fig. 4,
and slightly depends on tan . The region at , M2 > M1 = 500 GeV features a 500 GeV
bino like neutralino, whose scattering cross section off a proton keeps decreasing as the
higgsino fraction, with increasing . The isolevel curves indicate that, in this case study,
Fig. 9. The visibility ratio for direct spin-independent searches (expected signal over future projected sensitivity
for a XENON-1ton like experiment) on the (M2 , ) plane at M1 = 500 GeV, and the corresponding iso-level
curves.
Fig. 10. The same as in Fig. 9, but with M1 = 5000 GeV, for > 0 (left) and < 0 (right). We draw the isolevel
curves not on the surface but rather project them on the (M2 , ) plane. Cancellations occur for < 0, since the
Higgs-neutralino coupling gh (sin Z13 + cos Z14 ); while for > 0 both terms have the same sign, at
< 0 they have opposite signs, thus giving rise to the possibility of cancellations.
101
neutralinos as heavy as 700800 GeV will be within reach of future detection facilities; the
critical parameter is found to be the value of , while the M2 dependence is quite mild.
We show in Fig. 10 the VR for a larger value of M1 = 5 TeV, and for both signs of . In
this case, the reach of direct searches is clearly correlated to the degree of winohiggsino
mixing. Strikingly enough, we find that, for both signs of , neutralinos as heavy as 5 to
6 TeV will give a detectable signal at next generation direct detection searches, provided
M2 . We emphasize that this situation is not theoretically unrealistic, as it is naturally
realized along the hyperbolic-branch/focus-point (HB/FP) of mSUGRA (see Ref. [57],
where the phenomenology of binohiggsinoneutralino dark matter has also been extensively addressed); these is the slice in the parameter space close to the region where there
is no viable electroweak symmetry breaking) of models featuring a wino like LSP, as in
SI for < 0 (see
the mAMSB model [58]. Interestingly enough, we find cancellations in P
Fig. 10, right). While in the past it was realized that interferences of heavy and light Higgs
bosons exchanges could induce low values for the scattering cross sections, we point out
here that what we find is of a different nature. In split SUSY, in fact, the heavy Higgs conSI is driven to low values by the neutralinoneutralinoHiggs
tribution vanishes, and the P
coupling itself. The latter is in fact gh (sin Z13 + cos Z14 ), hence a suitable interplay of M2 and may entail, for a given value of tan , gh 0. Apart from this novel
feature, we remark that the sign of does not affect much the VR in this channel. The
same conclusion holds for all other detection channels: this is why we always restrict to
the case > 0.
To examine the prospects for indirect detection with neutrino telescopes, we consider
as a projected future sensitivity the IceCube expected limit on the muon flux from the
Sun [40], i.e., the green curve of Fig. 3, right. The intensity of the signal we find highlights
Fig. 11. The visibility ratio for the neutralinoannihilations-induced muon flux from the Sun (expected signal
over future projected sensitivity for the IceCube experiment) on the (M2 , ) plane at M1 = 500 GeV, and the
corresponding iso-level curves.
102
again the role of winohiggsino mixing: the line at M2 produces in fact the largest
muon fluxes, as it combines large capture rates in the Sun, due to large spin-dependent
neutralino-matter couplings, and large annihilation rates. As the LSP turns into a bino, the
capture-annihilation equilibrium ceases, and the VR suddenly collapses. In the wino like
regions, at M2 < M1 , , we observe the role of the higgsino fraction at fixed neutralino
mass (i.e., along a given constant M2 ), an effect which is fully accounted for by the decrease in the capture rate of neutralinos inside the Sun. The same effect is visible, though
less evident, in the higgsino like region.
Turning to antiprotons and positrons searches, we address the perspectives of discrimination of the signal from neutralino pair annihilations against the background from the
interaction of primary cosmic ray with the interstellar medium, following the approach
outlined in Ref. [48]. We define the quantity:
E
max
[s (E)]2
dE,
b (E)
(5)
Emin
where s (E) and b (E) refer, respectively, to the signal and the background flux, and the
integral extends over the energy range in which the integrand is non-negligible. This quantity stems from the continuum limit, up to an overall factor accounting for the exposure
times effective area of a given future experiment, of a 2 statistical variable defined assuming that the background is known and the signal is subdominant with respect to the
background. Antiproton and positron backgrounds are obtained with the Galprop [59]
propagation code, in a scheme with diffusion and convection tuned to give a fair estimate
of ratios of primary to secondary cosmic ray nuclei. Signals are obtained within the same
propagation framework. VRs are defined as ratios between I and the critical value in this
quantity as estimated for the PAMELA experiment [60] after three years of data taking,
3y
i.e., I = 3.2 108 cm2 sr1 s1 , see Ref. [48] for details.
In Figs. 12 and 13 we plot visibility ratios in case of the adiabatically contracted N03
halo model; contrary to gamma rays, since fluxes are dominated by nearby sources, the
suppression for the cored profile is not dramatic and discrimination pattern are comparable
as we will discuss in Section 5. Not surprisingly, we find a large dependence on the neutralino mass, and on the higgsino fraction in the bino like neutralino region at M2 , > M1 .
While antiprotons fluxes will be sensitive to winos and higgsinos up to masses as large as
500 GeV, the sensitivity in the positrons fluxes is somewhat less stringent, extending up to
higgsinos as heavy as 400 GeV. The fact that annihilating neutralinos give larger signal to
background ratios for antiprotons than for positrons, has been already pointed out [28,48],
and we give here a further confirmation to this point.
The case for antideuterons is different because here, restricting to a low energy window, the background flux is expected to be negligible [42], and even detection of 1 event
would imply discovery of an exotic component. Regarding the detection prospects, we
will consider, as the ultimate reach for an experiment in the future, that of the gaseous
antiparticle spectrometer (GAPS) [61]. This is a proposal for an instrument looking for antideuterons in the energy interval 0.10.4 GeV per nucleon, with estimated sensitivity level
of 2.6 109 m2 sr1 GeV1 s1 , to be placed either on a satellite orbiting around the
103
earth or on a probe to be sent into deep space. We compute the antideuteron flux induced
by neutralino annihilations in this energy bin, and define the VR as the ratio between this
and the mentioned expected sensitivity of GAPS [48].
Fig. 12. The visibility ratio for antiprotons searches (expected value of the I parameter over future corresponding projected sensitivity for the PAMELA experiment after 3 years of data-taking) on the (M2 , ) plane at
M1 = 500 GeV, and the corresponding iso-level curves.
Fig. 13. The visibility ratio for positrons searches (expected value of the I parameter over future corresponding projected sensitivity for the PAMELA experiment after 3 years of data-taking) on the (M2 , ) plane at
M1 = 500 GeV, and the corresponding iso-level curves.
104
Results for VRs in Fig. 14 show that a critical threshold in the signal is that of top
quarks; once again, the signal is maximized along the largest winohiggsino mixing line.
The binohiggsino mixing effect is also clearly distinguishable in the line of maxima in
VR around M1 < M2 ; the higgsino fraction is found to enhance the antideuteron production (region > M1 M2 ). As discussed in [48], binos are somewhat less disfavored
Fig. 14. The visibility ratio for antideuterons searches (expected signal over the projected sensitivity for a
GAPS-like experiment) on the (M2 , ) plane at M1 = 500 GeV, and the corresponding iso-level curves.
Fig. 15. The total production cross section for all possible neutralinoneutralino, charginoneutralino and
charginochargino reactions at the LHC (center-of-mass energy of 14 TeV), in fb, and the corresponding iso-level
curves.
105
in antideuterons rather than in antiprotons or positrons searches, due to the hadronization

The critical detection line (VR = 1) exof Binos dominant decay products, such as bb.
tends here well beyond that of future antiprotons and positrons searches, as envisaged in
Ref. [48], entering the region of bino like neutralinos with masses as large as 700 GeV.
The last process we consider is the production of neutralinos and charginos at the CERN
large hadron collider; we computed, with the Prospino2.0 package [62], the total production cross section of charginos and neutralinos at a center of mass energy of 14 TeV,
i.e.:
TOT
4

i<j =1
0 0 +
i
4,2

i,j =1
0 +
i
2

i,j =1
.
i
(6)
A thorough analysis of the expected LHC background and of the suitable cuts to optimize
searches for light neutralinos and charginos goes beyond the scope of the present work. We
outline, however, that, in two extreme cases, dedicated studies have faced this very same
problem: pure winos in mAMSB with large common scalar masses m0 , as discussed in [63]
and more recently in [64]; and higgsinos in the HB/FP region of mSUGRA, which have
been addressed in [65,66]. We find that in both cases the critical total production crosssection tentatively amounts to values around 500 fb, which we reproduce for illustrative
purposes in the iso-level curves of Fig. 15.
4.1. Summary of the case study results
With the aim of comparing different future neutralino dark matter detection strategies
in the case study under scrutiny, we collect in Figs. 16 and 17 the iso-level curves corresponding to the critical VR = 1 lines. We also include the bound stemming from the
Fig. 16. The reach of future facilities in the direct detection and muon-flux from the Sun channels, on the (M2 , )
plane at M1 = 500 GeV. We shade in yellow the region which is not consistent with the measured lithium 6
abundance.
106
primordial production of 6 Li [67] induced by the residual pair annihilations of neutralinos, after freeze-out and during the period of synthesis of light elements (the production
of 6 Li in the standard bing bang nucleosynthesis scenario is very strongly suppressed).
The limit we implemented is extrapolated from Ref. [67] for the cases of interest here
(namely, those of neutralino annihilations with gauge bosons final states); we shade the
corresponding excluded regions in Fig. 16, where we show also the reach contours of direct spin-independent detection and of the muon-flux from the Sun. Fig. 17 shows our
results concerning halo-dependent quantities: results for the cuspy adiabatically contracted
N03 profile and for the Burkert cored profile are respectively given in the left and right
panels.
Interestingly enough, we find that the region excluded by the 6 Li bound rules out a wide
portion of parameter space where direct searches are not effective, namely, that of pure
winos. We also remark that the whole region covered by the detection of a neutralino induced muon flux from the Sun is completely contained in that of direct searches: a positive
sign from IceCube would imply, within this framework, a visible signal at next generation
direct spin-independent facilities.
Fig. 17 shows the excluded regions from current data on antimatter fluxes [6871]:
remarkably, in both halo models a large portion of pure winos and higgsinos is already
ruled out by available data. Constraints from gamma rays are effective only in the case of
a cuspy profile, and exclude an even larger portion of models. Future antimatter searches
will cover essentially the whole parameter space in the large region; in the higgsinobino
mixing region, on the other hand, a clear hierarchy among antideuterons, antiprotons and
positrons is visible; with a cuspy profile, the region will be in all cases thoroughly probed
by the PAMELA experiment.
Fig. 17. The reach of future facilities for neutralino detection through antimatter searches on the (M2 , ) plane at
M1 = 500 GeV, with the N03-adiabatically contracted profile (left) and with the Burkert profile (right). The gray
and brown shadings respectively reproduce the parameter space inconsistent with current data on antiprotons and
positrons. We also indicate the EGRET exclusion bound and shade the region excluded by the HESS data (these
constraints are null for the Burkert cored profile).
107
Turning to a conservative cored profile, we find again a large antideuterons reach along
the higgsinobino large mixing region; in the pure wino region, instead, the lions share is
given by antiprotons. As a result, the potentially undetectable regions shrink to very narrow
corners of massive pure winos or higgsinos. We remark, comparing Figs. 16 and 17, a noticeable complementarity among neutralino dark matter searches: within this scenario, in
fact, the higgsino like part of the parameter space would be thoroughly searched for by direct detection experiments, while the pure wino region, already constrained by current data
on antimatter fluxes, will be largely accessible to next generation space based antimatter
search facilities.
The bottom line is therefore that, except for a thin slice at the edge of the winobino
transition line, in the case study we considered here, the whole parameter space will be
accessible to future dark matter detection experiments; this result holds quite independently of the halo model under consideration: a remarkable complementarity among direct
searches and antimatter searches ensures in fact most corners of the parameter space to be
within future reach. A naive estimate of the LHC reach, instead, indicates that these models
mostly lie beyond the reach of future super-colliders.
5. Discussion
In order to summarize the neutralino detection prospects at future dark matter search
facilities, we now go back to the parameter space slices of split SUSY introduced in Section 2. We will show on those planes our results concerning current and future reaches. We
recall that relic abundances inside the WMAP range correspond to the blue contours, and
that the regions shaded in red are ruled out by the LEP2 searches for charginos [13]. We
shade in yellow the region which is not consistent with the 6 Li abundance. In order to be as
conservative as possible, we resort here to the cored Burkert profile: 95% C.L. exclusions
limits from current positrons and antiprotons data are also indicated in the figures.
Fig. 18 details on the (M2 , ) plane at large M1 = 10 TeV. As a first remark, we point
out once again the role of winohiggsino mixing, which dramatically enters in direct detection searches: in particular, when the mixing is maximal, masses as large as 10 TeV
will be probed at future planned experiments! Antideuterons fluxes, as well as the flux
of muons from the Sun, are also sensitive to the degree of mixing between the wino and
higgsino components in the lightest neutralino: in this setup, IceCube may detect a signal
for neutralino masses as heavy as 1 TeV. As regards the case of pure neutralinos, either
wino or higgsino like, we remark that direct detection techniques are still effective when
M2 , 1 TeV; on the other hand, if the mixing is tiny, then only antimatter searches are
supposed to be effective. We notice, moreover, an interesting and non-trivial complementarity between antiprotons and antideuterons searches: while the first will be more effective
in the pure wino region, the latter are expected to do better in the pure higgsino case. In general, however, we find that the reach of antimatter searches is limited to around 500 GeV
in the wino like case and to 250 GeV in the higgsino like one.
Let us now turn to the (M1 , ) and (M1 , M2 ) planes, Fig. 19. In the well-known bino
higgsino plane, a large region of sufficiently pure higgsinos turns out to be excluded by
current antiprotons data. As expected, direct detection will probe the region at large mix-
108
ing, i.e., when M1 , even at very large masses. As the higgsino fraction drops (upper
part of the figure), all detection methods are not effective, in this plane. The case of pure
higgsinos reflects what we pointed out above: the region may be best searched for at antimatter detection experiments, up to masses around 250 GeV.
The (M1 , ) plane may be regarded, to some extent, as a low-energy blow-up of the
so-called hyperbolic branch/focus point region (HB/FP) of minimal supergravity [57]. In
Fig. 18. A collection of future exclusion limits in various direct and indirect neutralino detection techniques on
the (M2 , ) plane, at M1 = 10 TeV. A conservative cored dark halo model (Burkert profile) is assumed. The gray
and brown shadings respectively reproduce the parameter space inconsistent with current data on antiprotons and
positrons. The region shaded in yellow is not consistent with the measured lithium 6 abundance.
Fig. 19. Future exclusion limits on the (M1 , ) plane at M2 = 10 TeV (left), and on the (M1 , M2 ) plane at
= 10 TeV (right). The cored Burkert profile is again assumed for the dark matter density and velocity distributions. As in the previous figure, the gray and brown shadings respectively reproduce the parameter space
inconsistent with current data on antiprotons and positrons.
109
this respect, we can directly compare the results we show in the left panel of Fig. 19 with
those recent studies addressing the detection of neutralino dark matter in the HB/FP region
[27,38]. We remark that, as far as those quantities that weakly depend on assumptions on
the halo models are concerned, we mostly agree with those works. In particular, we confirm
the importance of the role of neutrino telescopes in the HB/FP region, already pointed out
in Baer et al., in Ref. [27]: the timeline of the future experimental facilities might give the
opportunity to IceCube to provide the first indirect evidences for neutralino dark matter in
this region of the mSUGRA parameter space; the would-be signal might later be confirmed
by stage-3 direct detection experiments, whose sensitivity extends well beyond that of the
future antartic neutrino telescope.
In the present analysis we extended previous studies of the binohiggsino mixing region
of mSUGRA with a detailed and thorough investigation of indirect neutralino detection
through antimatter searches. With a conservative dark halo profile, we find that positrons
and antiprotons searches will not be able to probe models, in the HB/FP region, featuring
a thermal neutralino relic abundance within the WMAP range. Nevertheless, we point out,
as a novel and somewhat exciting result, that antideuterons searches on a GAPS-like experiment could be extremely competitive in the HB/FP region, with a reach comparable to
that of neutrino telescopes.
The (M1 , M2 ) plane, which reproduces a conceivable scenario in which the gaugino
masses are light and the parameter is large, is essentially split in a pure wino and pure
bino region, respectively below and above the diagonal, dashed line. Interestingly enough,
in this plane direct detection and muon fluxes feature extremely low rates, well below future projected sensitivities. The bino like region, in which most of the points producing
a thermal relic abundance in the WMAP range lie, is completely off-limits for any detection technique. The wino like region, which is already constrained by current data up
to masses as large as 300 GeV, is going to be partly covered by antimatter searches,
particularly in the antiprotons channel.
So far we have assumed that, at each point in the parameter space, the lightest neutralino
has a relic density matching the CDM component, regardless of its thermal relic density in
the standard cosmological framework. We also mentioned that relaxing the overproduction
bound is possible but contrived, while there is a plethora of possibilities to enhance a low
thermal relic abundance into the CDM range. In a more conservative approach one can
assume that only a fraction of the CDM is accounted for by supersymmetric models whose
WMAP ). Under
thermal relic abundance lies below the WMAP preferred range ( < max
this assumption, the simplest possible way to compute the detection rates is to rescale the
neutralino dark matter density distribution according to the prescription

(r) CDM (r) min 1, WMAP .

(7)
min
Direct detection and capture rates in the Sun linearly scale with the local neutralino density
(r0 ); on the other hand, gamma-ray and antimatter rates involve the squared of the neutralino density distributions. We find that the only technique which could give a signal, in
this approach, and in the parameter space we analyze here, is the spin-independent direct
detection. In Fig. 20 we present the rescaled direct detection reach in the (M2 , ) plane, at
M1 = 10 TeV. Accelerator searches would naturally not be involved in the above outlined
110
rescaling: we show in the figure the iso-500 fb total production cross section at the LHC,
and the kinematic reach of a future next linear collider (NLC) with a center-of-mass energy
of 1 TeV (i.e., the iso-m + = 500 GeV line).2
For completeness, we also mention the detection prospects connected to another wellknown feature of the split SUSY scenario, i.e., the fact that gluinos have very long lifetimes. The phenomenological implications of a quasi-stable gluino have been recently studied in Refs. [11,72,73]; they mildly depend on the details of the spectrum, one of the crucial
quantities being the gluino mass. The maximal LHC reach is around Mg0 2.3 TeV [72].
In the general setup we have considered, the gluino mass is a free parameter. In order to
give a flavor of what could be the LHC reach in the long-lived gluino detection channel,
one has to resort to a particular framework to relate the gaugino masses. For instance, in
the (M2 , ) plane it is worthwhile inspecting the consequences of assuming an anomalymediated-inspired relation, where M2 /M3 67, the spread being given by RG effects
on the g2 and g3 coupling which depend on the value of the gravitino mass parameter
m3/2 [58]. In Fig. 20 we plot, with a solid violet line, the putative reach of the LHC in the
long-lived gluino detection channel assuming an anomaly mediated relation between M2
and M3 , which corresponds to M2 330 GeV. The latter value is obtained solving for the
equation:

M2 =
AMSB

M3
Mg0 ,
(M2 )
M2
Mg0 = 2.3 TeV.
(8)
Fig. 20. The isolevel curve corresponding to the 500 fb1 total neutralinochargino production cross section at
the LHC and the (kinematic) reach of a NLC, compared with direct dark matter searches on the (M2 , ) plane,
at M1 = 10 TeV. We rescale here the local abundance of neutralinos according to their thermal relic density (see
Eq. (7)).
2 Notice that for nearly degenerate lightest chargino and neutralino the actual reach of a NLC will be slightly
lower than the kinematical reach.
111
In summary, Fig. 20 shows that in case the rescaling prescription is applied, models which
could be discovered with the direct detection technique have non-negligible winohiggsino
mixing; since the relic abundance of winos is smaller than that of higgsinos, at a given
mass, we also notice that in this scheme the higgsino like region is somewhat favored. The
pure wino and pure higgsino regions could only be explored at accelerators: the complementarity between accelerator searches and the quest for dark matter therefore emerges
even in the present setup of pure electroweak production of supersymmetric particles at
colliders.
As a concluding remark, we want to stress that the results we presented here hold
for more general supersymmetric models, where the scalar sector need not necessarily
be largely split. We can conservatively state that postulating a scalar sector lying even
only one order of magnitude above the LSP mass scale would leave most of our results essentially unaffected. Even lighter scalars may, however, give rise to some possible caveats,
particularly for the neutralino relic density, for instance, as far as sfermion coannihilations,
resonant annihilations with the heavy Higgses, and t- and u-channels sfermions exchange
diagrams are concerned.
6. Conclusions
We studied in full generality the neutralino dark matter phenomenology of models in
which the scalar sector is heavy, i.e., in split supersymmetry scenarios. The relevant parameters are the entries of the neutralino mass matrix, namely the soft breaking gaugino mass
parameters M1 and M2 and the Higgs mixing mass term . Requiring that the thermal relic
abundance of neutralinos does not exceed the observed amount of cold dark matter defines
an hypersurface in the three-dimensional space of parameters: the cosmological bound is
indeed the only issue forcing one of the sectors of the theory to be light, i.e., at a scale lower
than at least 2.2 TeV; the bound can be violated only invoking mechanisms for entropy production at low energy, a rather contrived setup. Models defined by parameters lying below
the hypersurface are viable, and suitable mechanisms of relic density enhancement may
drive low relic density models to produce the required amount of cold dark matter.
We studied in detail the physics of winohiggsino mixing, and pointed out that future
experiments may be able to probe extremely large neutralino masses when the mixing is
maximal: 10 and 1 TeV neutralinos may give detectable signals respectively in next generation spin-independent searches and at IceCube. Interestingly enough, we find that even
though only one Feynman diagram contributes to neutralinoproton scattering, cancellations among the various neutralino interaction-eigenstates components may conspire and
suppress the relevant couplings.
Resorting to two extreme cases for the dark matter distribution in our Galaxy, we
showed that large parameter space portions are already ruled out by currently available
data on antiprotons and positrons fluxes, as well as by the primordial 6 Li abundance induced by neutralino residual annihilations. Depending on the structure of the dark halo in
the GC, measurements of the gamma-ray in the Galactic center direction from the EGRET
and the HESS telescopes have also been shown to set strong constraints in the parameter
space. Future prospects for antimatter searches look promising, particularly in the pure
112
Wino and pure higgsino cases; a remarkable complementarity between antideuterons and
antiprotons detection has also been outlined.
We worked out an explicit analysis of a case study, in which we fixed the bino mass
term M1 = 0.5 TeV. We showed that while most of the resulting parameter space slice
will not be within the reach of the LHC, an interplay among direct detection and antimatter
searches will allow future dark matter detection experiments to thoroughly explore the split
supersymmetry parameter space.
Restricting to a standard cosmological scenario and taking into account for thermal
components only, a rescaling procedure should be implemented for subdominant supersymmetric dark matter candidates, suppressing detection rates and leaving a chance only
for direct detection in the large winohiggsino mixing region. In this scenario, the complementarity among collider searches and direct detection experiments explicitly emerges,
although here superpartners can be produced only through electroweak processes.
As a last remark, we point out that a lighter scalar sector would in general yield, modulo cancellations, larger dark matter detection signals: in this respect our results may be
regarded as lower bounds to neutralino searches in more general supersymmetric setups.
Note added
While this manuscript was being completed, Ref. [74] appeared, where some indirect
dark matter detection signals are discussed in the framework of split supersymmetry.
Acknowledgements
The work of A.M. and P.U. was supported by the Italian INFN under the project Fisica
Astroparticellare and the MIUR PRIN Fisica Astroparticellare. The work of S.P. was
supported in part by the US Department of Energy under contract number DE-FG0297ER41022. S.P. would like to thank H. Baer, K. Matchev and A. Birkedal for useful
discussions and remarks.
References
[1] P. Fayet, Phys. Lett. B 86 (1979) 272.
[2] H. Goldberg, Phys. Rev. Lett. 50 (1983) 1419;
J.R. Ellis, J.S. Hagelin, D.V. Nanopoulos, K.A. Olive, M. Srednicki, Nucl. Phys. B 238 (1984) 453;
L.E. Ibanez, Phys. Lett. B 137 (1984) 160.
[3] D.N. Spergel, et al., WMAP Collaboration, Astrophys. J. Suppl. 148 (2003) 175, astro-ph/0302209.
[4] N. Arkani-Hamed, S. Dimopoulos, hep-th/0405159.
[5] G.F. Giudice, A. Romanino, Nucl. Phys. B 699 (2004) 65, hep-ph/0406088.
[6] N. Arkani-Hamed, S. Dimopoulos, G.F. Giudice, A. Romanino, hep-ph/0409232.
[7] I. Antoniadis, S. Dimopoulos, hep-th/0411032.
[8] B. Kors, P. Nath, hep-th/0411201.
[9] C. Kokorelis, hep-th/0406258.
[10] P. Kumar, J.D. Lykken, JHEP 0407 (2004) 001, hep-ph/0401140.
113
[11] J.L. Hewett, B. Lillie, M. Masip, T.G. Rizzo, JHEP 0409 (2004) 070, hep-ph/0408248.
[12] D. Pierce, A. Papadopoulos, Phys. Rev. D 50 (1994) 565, hep-ph/9312248;
D. Pierce, A. Papadopoulos, Nucl. Phys. B 430 (1994) 278, hep-ph/9403240;
A.B. Lahanas, K. Tamvakis, N.D. Tracas, Phys. Lett. B 324 (1994) 387, hep-ph/9312251;
M. Drees, M.M. Nojiri, D.P. Roy, Y. Yamada, Phys. Rev. D 56 (1997) 276, hep-ph/9701219;
M. Drees, M.M. Nojiri, D.P. Roy, Y. Yamada, Phys. Rev. D 64 (2001) 039901, Erratum.
[13] ALEPH, DELPHI, L3, OPAL Experiments, LEPSUSYWG/01-03.1, http://lepsusy.web.cern.ch/lepsusy/.
[14] P. Gondolo, J. Edsjo, P. Ullio, L. Bergstrom, M. Schelke, E.A. Baltz, JCAP 0407 (2004) 008, astro-ph/
0406204.
[15] P. Binetruy, G. Girardi, P. Salati, Nucl. Phys. B 237 (1984) 285;
K. Griest, D. Seckel, Phys. Rev. D 43 (1991) 3191;
J. Edsjo, P. Gondolo, Phys. Rev. D 56 (1997) 1879, hep-ph/9704361.
[16] A. Pierce, Phys. Rev. D 70 (2004) 075006, hep-ph/0406144.
[17] B. Murakami, J.D. Wells, Phys. Rev. D 64 (2001) 015001;
T. Moroi, L. Randall, Nucl. Phys. B 570 (2000) 455;
M. Fujii, K. Hamaguchi, Phys. Lett. B 525 (2002) 143;
M. Fujii, K. Hamaguchi, Phys. Rev. D 66 (2002) 083501;
R. Jeannerot, X. Zhang, R.H. Brandenberger, JHEP 9912 (1999) 003;
W.B. Lin, D.H. Huang, X. Zhang, R.H. Brandenberger, Phys. Rev. Lett. 86 (2001) 954.
[18] P. Salati, astro-ph/0207396;
F. Rosati, Phys. Lett. B 570 (2003) 5, hep-ph/0302159.
[19] S. Profumo, P. Ullio, JCAP 0311 (2003) 006, hep-ph/0309220.
[20] S. Profumo, P. Ullio, astro-ph/0404390.
[21] M. Kamionkowski, M.S. Turner, Phys. Rev. D 42 (1990) 3310.
[22] R. Catena, N. Fornengo, A. Masiero, M. Pietroni, F. Rosati, astro-ph/0403614.
[23] N. Fornengo, A. Riotto, S. Scopel, Phys. Rev. D 67 (2003) 023514, hep-ph/0208072.
[24] M. Kamionkowski, K. Griest, G. Jungman, B. Sadoulet, Phys. Rev. Lett. 74 (1995) 5174, hep-ph/9412213.
[25] L. Bergstrom, J. Edsjo, P. Gondolo, Phys. Rev. D 58 (1998) 103519, hep-ph/9806293.
[26] J.L. Feng, K.T. Matchev, F. Wilczek, Phys. Rev. D 63 (2001) 045024, astro-ph/0008115;
J.R. Ellis, J.L. Feng, A. Ferstl, K.T. Matchev, K.A. Olive, Eur. Phys. J. C 24 (2002) 311, astro-ph/0110225;
J.R. Ellis, K.A. Olive, Y. Santoso, New J. Phys. 4 (2002) 32, hep-ph/0202110;
R. Dermisek, S. Raby, L. Roszkowski, R. Ruiz De Austri, JHEP 0304 (2003) 037, hep-ph/0304101;
A. Bottino, F. Donato, N. Fornengo, S. Scopel, Phys. Rev. D 70 (2004) 015005, hep-ph/0401186;
D. Hooper, J. Silk, hep-ph/0409104.
[27] J.L. Feng, K.T. Matchev, F. Wilczek, Phys. Lett. B 482 (2000) 388, hep-ph/0004043;
U. Chattopadhyay, A. Corsetti, P. Nath, Phys. Rev. D 68 (2003) 035005, hep-ph/0303201;
H. Baer, A. Belyaev, T. Krupovnickas, J. OFarrill, JCAP 0408 (2004) 005, hep-ph/0405210.
[28] S. Profumo, C.E. Yaguna, Phys. Rev. D 70 (2004) 095004, hep-ph/0407036.
[29] M.W. Goodman, E. Witten, Phys. Rev. D 31 (1986) 3059;
I. Wasserman, Phys. Rev. D 33 (1986) 2071.
[30] J. Gasser, H. Leutwyler, M.E. Sainio, Phys. Lett. B 253 (1991) 252.
[31] D. Adams, et al., Phys. Lett. B 357 (1995) 248.
[32] L. Bergstrom, P. Gondolo, Astropart. Phys. 5 (1996) 263.
[33] D.S. Akerib, et al., CDMS Collaboration, astro-ph/0405033.
[34] E. Aprile, et al., astro-ph/0207670.
[35] R. Bernabai, et al., Phys. Lett. B 389 (1996) 35;
M. Altmann, et al., astro-ph/0106314;
See also: R. Gaitskell, V. Mandic, http://dmtools.berkeley.edu/limitplots/.
[36] N.J.C. Spooner, et al., Phys. Lett. B 473 (2000) 330.
[37] L. Bergstrm, J. Edsj, P. Gondolo, Phys. Rev. D 58 (1998) 103519.
[38] J. Edsjo, M. Schelke, P. Ullio, JCAP 0409 (2004) 004, astro-ph/0405414.
[39] A. Habig, Super-Kamiokande Collaboration, hep-ex/0106024.
[40] J. Edsj, Internal Amanda/IceCube report, 2000.
114
[41] T. Sjstrand, fixxx cpc82199474;

T. Sjstrand, PYTHIA 5.7 and JETSET 7.4, Physics and Manual, CERN-TH.7112/93, hep-ph/9508391.
[42] F. Donato, N. Fornengo, P. Salati, Phys. Rev. D 62 (2000) 043003;
F. Donato, N. Fornengo, D. Maurin, P. Salati, R. Taillet, Phys. Rev. D 69 (2004) 063501.
[43] A. El-Zant, I. Shlosman, Y. Hoffman, Astrophys. J. 560 (2001) 336.
[44] A. Burkert, Astrophys. J. 447 (1995) L25.
[45] P. Salucci, A. Burkert, Astrophys. J. 537 (2000) L9.
[46] G.R. Blumental, S.M. Faber, R. Flores, J.R. Primack, Astrophys. J. 301 (1986) 27.
[47] J.F. Navarro, et al., astro-ph/0311231, Mon. Not. R. Astron. Soc. (2004), in press.
[48] S. Profumo, P. Ullio, JCAP 0407 (2004) 006, hep-ph/0406018.
[49] P. Ullio, in: Proceedings of the Third International Conference on Frontier Science, Physics and Astrophysics
in Space, Villa Mondragone (Rome), Italy, published on Frascati Physics Series.
[50] H.A. Mayer-Hasselwander, et al., Astron. Astrophys. 335 (1998) 161.
[51] K. Kosack, et al., VERITAS Collaboration, astro-ph/0403422.
[52] K. Tsuchiya, et al., CANGAROO-II Collaboration, Astrophys. J. 606 (2004) L115.
[53] F. Aharonian, et al., HESS Collaboration, Astron. Astrophys. 425 (2004) L13, astro-ph/0408145.
[54] A. Cesarini, et al., Astropart. Phys. (2004), in press.
[55] D. Hooper, B. Dingus, astro-ph/0212509.
[56] GLAST Proposal to NASA A0-99-055-03, 1999.
[57] K.L. Chan, U. Chattopadhyay, P. Nath, Phys. Rev. D 58 (1998) 096004, hep-ph/9710473;
J.L. Feng, K.T. Matchev, T. Moroi, Phys. Rev. Lett. 84 (2000) 2322, hep-ph/9908309;
J.L. Feng, K.T. Matchev, T. Moroi, Phys. Rev. D 61 (2000) 075005, hep-ph/9909334.
[58] L. Randall, R. Sundrum, Nucl. Phys. B 557 (1999) 79, hep-th/9810155;
G.F. Giudice, M.A. Luty, H. Murayama, R. Rattazzi, JHEP 9812 (1998) 027, hep-ph/9810442;
T. Gherghetta, G.F. Giudice, J.D. Wells, Nucl. Phys. B 559 (1999) 27, hep-ph/9904378;
J.L. Feng, T. Moroi, Phys. Rev. D 61 (2000) 095004, hep-ph/9907319.
[59] Galprop numerical package, http://www.mpe.mpg.de/~aws/propagate.html.
[60] O. Adriani, et al., PAMELA Collaboration, in: Proceedings of the 26th ICRC, Salt Lake City, 1999,
OG.4.2.04;
P. Picozza, A. Morselli, J. Phys. G 29 (2003) 903.
[61] K. Mori, C.J. Hailey, E.A. Baltz, W.W. Craig, M. Kamionkowski, W.T. Serber, P. Ullio, Astrophys. J. 566
(2002) 604, astro-ph/0109463.
[62] W. Beenakker, R. Hopker, M. Spira, hep-ph/9611232.
[63] H. Baer, J.K. Mizukoshi, X. Tata, Phys. Lett. B 488 (2000) 367, hep-ph/0007073.
[64] A.J. Barr, C.G. Lester, M.A. Parker, B.C. Allanach, P. Richardson, JHEP 0303 (2003) 045, hep-ph/0208214.
[65] H. Baer, T. Krupovnickas, X. Tata, JHEP 0307 (2003) 020, hep-ph/0305325.
[66] H. Baer, C. Balazs, A. Belyaev, T. Krupovnickas, X. Tata, JHEP 0306 (2003) 054, hep-ph/0304303.
[67] K. Jedamzik, Phys. Rev. D 70 (2004) 063524;
K. Jedamzik, Phys. Rev. D 70 (2004) 083510.
[68] S. Orito, et al., Phys. Rev. Lett. 84 (2000) 1078;
Y. Asaoka, et al., Phys. Rev. Lett. 88 (2002) 05110.
[69] Boezio, et al., Astrophys. J. 561 (2001) 787.
[70] DuVernois, et al., Astrophys. J. 559 (2001) 296.
[71] Boezio, et al., Astrophys. J. 532 (2000) 653.
[72] W. Kilian, T. Plehn, P. Richardson, E. Schmidt, hep-ph/0408088.
[73] K. Cheung, W.Y. Keung, hep-ph/0408335.
[74] A. Arvanitaki, P.W. Graham, hep-ph/0411376.
A geometric look on the microstates of supertubes

Dongsu Bak a , Yoshifumi Hyakutake b , Seok Kim c ,
Nobuyoshi Ohta b
a Physics Department, University of Seoul, Seoul 130-743, South Korea
b Department of Physics, Osaka University, Toyonaka, Osaka 560-0043, Japan
c School of Physics, Seoul National University, Seoul 151-747, South Korea
Received 12 August 2004; accepted 25 January 2005
Abstract
We give a geometric interpretation of the entropy of the supertubes with fixed conserved charges
and angular momenta in two different approaches using the DBI action and the supermembrane
theory. By counting the geometrically allowed microstates, it is shown that both the methods give
consistent result on the entropy. In doing so, we make the connection to the gravity microstates clear.
PACS: 11.25.-w; 11.27.+d; 04.70.Dy
1. Introduction
Supertubes are tubular shaped bound states of D0-branes, fundamental strings (F1) and
D2-branes [131]. While having translational invariance in the axial direction along which
the F1 strings are stretched, the cross sectional shape of the supertubes may be arbitrary in
the eight transverse dimensions. As shown in Ref. [6], the cross sectional shape could be
either open and stretched to infinity or closed but here we would like to focus on the closed
cases.
E-mail addresses: dsbak@mach.uos.ac.kr (D. Bak), hyaku@het.phys.sci.osaka-u.ac.jp (Y. Hyakutake),
calaf2@snu.ac.kr (S. Kim), ohta@phys.sci.osaka-u.ac.jp (N. Ohta).
doi:10.1016/j.nuclphysb.2005.01.042
116
D. Bak et al. / Nuclear Physics B 712 (2005) 115138
Let us begin our discussion with the cases where the cross sectional curve lies in x 1 and
x 2 plane. The supertube then carries an angular momentum density J = J12 proportional to
the cross sectional area. For the fixed conserved charges, the moduli space of supertubes is
consisting of the geometric fluctuations of the cross sectional shape [28]. Since the angular
momentum is fixed, the fluctuation of the curve hasto be area preserving. The length
L of the cross sectional curve is further limited by Q0 Q1 where Q0 and Q1 denote
lineal D0 density in the axis direction and
by 2 , respectively. Thus
F1 charges divided
one has the restriction of the length by J /T2 L Q0 Q1 /T2 , where T2 is the D2brane tension. This space of arbitrary fluctuation of the curve forms an infinite-dimensional
moduli space. For given curve, the magnetic field representing the density of D0 may be
arbitrary with total number of D0-branes fixed. Moreover the shape may fluctuate into the
six more transverse directions. Hence eight arbitrary bosonic functional fluctuations are
involved as the moduli deformation. Since the supertubes involve a nonvanishing electric
field and linear momentum densities fixed by the shape of the curve, the above moduli
space is not a configuration space but a phase space.
The supertubes allow corresponding supergravity solutions [4,9] of an arbitrary cross
sectional shape and arbitrary density of D0-brane as a function of the world-volume coordinate of the curve direction. Therefore the solution involves the same number of
arbitrary functions of bosonic degrees. The geometry is nonsingular everywhere as argued
in Ref. [32] in the U -dual picture. The solution does not have a horizon either. The recently
emerging picture is that such a regular, no-horizon solution corresponds to distinguishable
gravity microstates represented by supergravity fields [33]. When all the conserved charges
and certain asymptotic conditions on the geometries are fixed, the logarithm of the number
of above microstates is the entropy of the gravity system with certain macroscopic parameters fixed. In case of supergravity supertubes, we are interested in all the supersymmetric
solutions with fixed energy, D0 and F1 charges and the angular momenta. The system
may have many components of angular momenta of SO(8) since the system involves eight
transverse dimensions. We fix here the four independent Cartan elements of SO(8).
This solution space with all the macroscopic conserved quantities fixed, forms a moduli
space of the supergravity supertubes. As we see in the case of DBI description of supertubes, this space must be a phase space instead of a configuration space. Since the phase
moduli space involves arbitrary functions, it is an infinite-dimensional space. Hence its
volume divided by (2 h)
dim is either zero or infinity. Consequently it requires at least a
regularization procedure. By quantization, the above problem may be avoided but this will
not be that simple since a direct quantization of gravity is not well defined as we know very
well.
But the two sides of the system has a striking similarity in its geometric nature within
the moduli space. Namely the bosonic sector in each side may be visualized as a geometric
shape. The cross sectional shape of the DBI description has a gravity counterpart of the
supertube shape. The moduli space of a supertube consists of the shape fluctuations and
again each has its own counterpart in the supergravity side.
In fact, there is a regime where both descriptions may have their validity. Note that the
radius squared of the circular supertubes is given by
R 2 = 2gs 2s N0 N1
s
,
Lz
(1)
117
where N0 = Lz Q0 and N1 = 2Q1 are the numbers of D0 and F1 and Lz is the size of
the compactified circle along which the axis direction of the supertube is wrapped. This
is a new length scale introduced by supertubes and this estimation of the supertube size is
valid unless J Lz /(Q0 Q1 ) 1. The cross sectional area is quantized, which is related to
the quantization of the angular momentum. Considering the case where = Lz /(2s ) is
of order one, the validity of supergravity description requires that
gs N0 N1 1,
by R s .
Since the energy of the supertubes are given by

1
N0
1
Lz =
M=
N
+
(gs N1 + N0 ),
1
gs s Lz
gs s
22s
the Schwarzschild radius RS = (Mgs2 8s /Lz )1/6 is

1/6
.
RS = s (gs N1 + N0 / )gs /2
(2)
(3)
(4)
Thus, for gs N0 N1 1, R RS , which may explain the regularity of the supertube solutions.
On the other hand, the DBI description has its validity in the decoupling limit of s
0 and gs 0. Thus the overlapping region of the validity is given by the open-string
decoupling limit
s 0,
gs 0,
(5)
while keeping the combination gs N0 N1 large. Thus we have here the gravity and the supertube field theory correspondence in the overlapping regime of the validity. The decoupled
field theory is the world-volume field theory of supertubes. In this limit, the field theory
obtained by expanding DBI theory around the circular supertube background is eventually described by a peculiar (2 + 1)-dimensional (noncommutative) YangMills theory in
the decoupling limit where s 0, which is equivalent to the matrix theory in a circular
supertube background [2,6,14].
In this note, we would like to count the entropy of the geometries using the gravity/field
theory correspondence and the structure of the phase moduli obtained in Ref. [28]. We
shall be using basically the DBI action to count the degeneracy the states. This problem
is in some sense already treated in Ref. [25], but the perspective and the emphasis on the
geometric nature are the main differences.
We would like to first make it clear that we are basically counting the geometric fluctuations in the sense that, even in the field theory, we are counting the freedom of the
shape fluctuation including other accompanying bosonic and fermionic degrees. This simpler version of the account of entropy using the shape fluctuation is presented first. The
full derivation of the entropy in the decoupling limit is done via two different methods. One is the description of DBI action. Here we identify the infinite-dimensional
fermionic moduli space and count the entropy including all the fermionic fluctuations.
In this case we use the near circular condition q = (Q0 Q1 J )/J 1 to simplify
the calculation. The others are via the M-theory description of the M2-brane. In this
118
M2-brane picture, we find that the near circular condition is not necessary for the counting.
As stated earlier, the quantization is necessary to get the correct expression and in this
sense the decoupling limit is essential. At the end of the day, the decoupled theory in the
same limit should lead to a unique theory in any paths.
The paper is organized as follows. In Sections 2 and 3, we review the phase moduli space
of supertubes and count the cross sectional shape fluctuations. In Section 4, we explain the
fermionic part of the moduli space using DBI description and count the full degeneracy in
the near circular limit. In Section 5, we count again the entropy using the M-theory. Section
6 is devoted to conclusions.
2. BPS equations and conserved charges for a closed supertube

A tubular D2-brane with electric and magnetic fluxes on the world-volume becomes a
closed supertube if it satisfies suitable BPS conditions. We first review these BPS equations
and conserved charges for a closed supertube. This also serves to establish our notations.
The tubular D2-brane is embedded in the 10-dimensional flat spacetime, and the worldvolume is parametrized by (t, , z). The pullback metric and the field strength on the D2brane is written as

2
dspb
= 1 |x |2 dt 2 + |
x |2 d 2 + 2x x dt d + dz2 ,
F = E dt dz + B dz d.
(6)
x

Here x = (x 1 , . . . , x 8 ), x x
t and x . The cross section of the D2-brane is expressed
8
as an arbitrary loop in R . The angle ( ) represents the direction along the
loop and z lies in the transverse direction. Thus we are considering only the configurations
with the translational invariance in the z direction.
The bosonic part of the D2-brane DBI action is evaluated as

S = T2 dt d dz

x |2 + (x x )2 22 EB x x ,
1 |x |2 |
x |2 + 2 B 2 2 E 2 |
(7)
where a D2-brane tension T2 and are written as T2 = (2)12 3 g and = 22s , respecs s
tively, in terms of the string length s and coupling gs . Canonical momenta pi and
conjugate to x i and Az are written as

T22 2
xi |
x | + 2 B 2 (x x )xi + 2 EBxi ,
L
T22 2
E|
=
x |2 + 2 B x x ,
L
pi =
(8)
where L is the Lagrangian density. The Hamiltonian density is given by

2
x |2 + T22 2 B 2 + |p|
2+ 2
H = T22 |

2
B 2
2B 2
+ T2 B + T2 |
=
x|
+ |p|
2
|
x|
|
x |2
B
T + T0 ,
2
119
(9)
1
1
where T = 2
2 is the tension of the fundamental string and T0 = s gs is the mass of a
s
D0-brane. It can be verified that the third term in the square root in the second line is
non-negative and vanishes when x x , which is equivalent to x = 0 by suitable reparametrization of . Thus the Hamiltonian density is bounded from below by the mass density
of fundamental strings and D0-branes, which precisely matches with the energy of the supertube. The equality in (9) is saturated when
2
ds
x |2 T2
,
x = 0,
B = T2 |
(10)
d
where ds 2 = d x d x is the line element of R8 . These are the BPS conditions which must
be satisfied by all supertubes.
The closed supertube carries two charges, which corresponds to those of fundamental
strings and D0-branes, and angular momenta. Defining
1
Q1 =
2

d ,
1
Q0 =
2

d B,
(11)
we find that 2Q1 Z is the number of fundamental strings dissolved in the D2-brane,
and Q0 Lz Z is that of D0-branes. The angular momenta are given as
1
L =
2
ij
T2

d x i p j x j p i =

dx i dx j ,
(12)
where i, j = 1, . . . , 8. With the aid of the BPS equations (10), the canonical momenta are

expressed as p i = T2 x i . The last equality in (12) is obtained by using this relation. Thus,
when we fix the angular momenta, the area made by projecting the loop of the supertube
onto (i, j )-plane should be preserved during the deformation in the flat directions.
The flat directions of supertubes of general shape with fixed fundamental string charge
Q1 , D0-brane charge Q0 and angular momentum J are of our interest. They make the
moduli space of the supertubes and are related to the number of gravity microstates. It was
shown [28] that the perimeter 2L of the supertubes is restricted by

J /T2 L Q0 Q1 /T2 .
(13)
The number of microstates allowed by this bound is the problem we are going to discuss
in this paper.
120
3. A first look on the microstates of supertubes

In order to get the idea how to count the microstates of the supertubes, let us first discuss
the bosonic fluctuations (), a() and b() around the circular background,
= x1 + ix2 = R(1 + )ei ,
= Q1 (1 + a),
B = Q0 (1 + b),
(14)
where , a and b are real. We consider the landscape of vacua having Q0 Q1 > T2 R 2 with
the area A = R 2 fixed. Introducing q defined by
Q0 Q1 = T2 R 2 (1 + q),
(15)
we focus on the case of q 1. The fluctuation can be expanded

() =
n ein ,
a() =
an ein ,
b() =
bn ein .
as1
nZ
nZ
(16)
nZ
Here we should impose reality conditions n = n , an = an and bn = bn .

The area given by
1
A=
4
2

Im d d
(17)
is evaluated as
A = R

1 + 20 +

|n |
(18)
nZ
The conservation of the angular momentum implies that

20 =
|n |2 .
(19)
nZ
The length 2L of the curve

2
2L =
2
|d | = R
(1 + )2 + ( )2
(20)
may be expanded as
2
2L = R

(2 + 2 + ( )2 ) 2
+
d 1 +
2
2

n2
|n |2 + .
= 2R 1 + 0 +
2
(21)
nZ
1 As we shall see later on, a and b are mixed with the fluctuation. But for simplicity, we ignore this complication here. See also Ref. [28] for the detailed classical description of this mixing.
The condition (19) may be used to eliminate 0 from (21) and we get

2
2
2L = 2R 1 +
(n 1)|n | + .
121
(22)
n>1
Here n for n > 0 are our phase space variables and 0 is constrained by the condition (19).
We are not interested in the translational mode 1 .
Since from (13)

R L Q0 Q1 /T2 ,
(23)
we get, using (15), a constraint

n2 1 |n |2 q/2.
(24)
n>1
To compute the number of states in the volume (24), let us find out the canonical variables in the phase space. First we define the coordinate corresponding to the radius as
r(= R(1 + )) which is real. From the action (7), we obtain the conjugate momentum
pr () = T2 r ,
(25)
where use has been made of the fact that E = 1/ for the BPS states of our concern here
and L = T2 B. The relation (25) is a second class constraint and we should make Dirac
quantization. After this procedure, one finds
i

r(), T2 r ( ) = ( )(z z ).
2
For the zero mode in the z direction, the above implies that

m , n =
1
mn ,
4T2 R 2 Lz n
(26)
(27)
where Lz is the length of the supertube in the z direction. Thus cn defined by
cn = n ,
n
with 2 = 1/(4T2 R 2 Lz ) satisfies the commutation relation

cm , cn = mn .
In terms of these variables, the constraint (24) becomes

1
q
n
|cn |2 2 .
n
2
(28)
(29)
(30)
n=2
Quantum mechanically, this condition is interpreted as

1
q
n
Nn 2 s,
n
2
n=2
(31)
122
where the number operator Nn is defined by cn cn .

Our task is now to evaluate the number of states restricted by (31). For large n, the 1/n
in the bracket may be ignored, and the problem reduces to the well-known case of counting
string states. It is given by

2 23 s
.
V=
(32)
e
4 s
To get this, let us consider the following quantity [34]
G(w) =

1
1
f (w) .
n)
(1
w
n=1
dn w nNn =
n=0
(33)
This is related to the Dedekind eta function

( ) = e
i/12

1 e2in ,
(34)
n=1
which has the modular transformation formula

(1/ ) = (i )1/2 ( ).
Applied to f (w), this gives the HardyRamanujan formula

2 1/2 1/2 1/12 2
f (w) =
w
w
f w ,
log w
(35)
(36)
where
w = e2
2 / log w
(37)
One can then deduce the asymptotic formula for w 1 (or w 0)

2
2 1/2
exp
f (w)
.
log w
6 log w
The degeneracy is obtained by

G(w) dw
.
dn =
w n+1 2i
(38)
(39)
Using the asymptotic expansion of f (w), this can be estimated for large n by a saddle point
evaluation. G(w) grows rapidly for w 1, while if n is very large, w n+1 is very small
for w < 1. There is a sharp saddle point for w near 1. The integrand (for the integration
variable log w)

2
n log w ,
exp
(40)
6 log w
is stationary for log w / 6n. Evaluating (39) around this saddle point, we get [35]
1
dn = e
4 3n
2
3n
(41)
in the large n. Integration of dn up to s gives V in (32).

Thus the entropy becomes

q
4
S = ln V =
Lz (Q0 Q1 J ).
=
2
3
3
123
(42)
We note that this is the entropy from a single fluctuating boson around the supertube. In
what follows we are going to extend this to supertubes with other modes.
4. Supertube solutions in DBI action and entropy

In this section we find exact BPS supertube solutions including fermion backgrounds
using the DBI action of D2. The BPS solutions preserve 1/4 supersymmetry. One may
identify fermionic flat directions of the classical solutions. We then give the quantization
rules for the flat modes and we count the contributions to the entropy from fermions as
well as bosons.
4.1. The solutions
We start by summarizing the supersymmetric DBI theory for the D2 in the notations
and conventions of [36]. This action has gauge invariances coming from world-volume
diffeomorphism and the local kappa symmetry. Since the full gauge invariant action is
complicated, we start from the action with gauge fixed kappa symmetry:

M

+
M ) ,
S = d 3 det g + F 2
(43)
(
where , = t, , z are the world-volume indices, M is the R 9+1 vector index, g is
M
the pullback of 10-dimensional flat metric onto the world-volume, = M X
is the
induced gamma matrix, and is the MajoranaWeyl fermion in the target space, where the
Weyl condition is imposed by the gauge choice for the local kappa symmetry. We use the
convention = (i 0 ). As in Section 2, one should also include an overall coefficient
T2 , the D2-brane tension, and also replace F by 22s F .
For later use, we summarize our gamma matrix conventions. The SO(9, 1) 32 32 M
is expressed in terms of SO(1, 1) 22 gamma matrices and SO(8) 1616 gamma matrices
as follows:

z = 1 116 , others 3 i (i = 1, 2, . . . , 8).
0 = i 2 116 ,
(44)
i s are the SO(8) spinors in a suitable representation: we use the convention that the former/latter 8 indices act on left/right chiral components, respectively. The last eight gamma
matrices in (44) will be written as = 3 , where the vector lives in R 8 . The chirality
operator is defined as
11 = 0 1 8 z = 3 9 ,
(45)
124
with the choice 9 = diag(18 , 18 ). The Weyl condition on is chosen to be 11 = +.

With our convention (44) and (45), this chiral is written as

1
( )
0
0
( ) =
,
(46)
0
0
1
( )
where , are 8-component SO(8) spinors with 9 eigenvalues 1, respectively.
Let us make the partial gauge fixing for the world-volume diffeomorphism
T = t,
Z = z,
0 < 2,
x(t, , z) R 8 ,
(47)
i.e., we leave unspecified one scalar field in R 8 tangent to direction. This is harmless as
far as we do the classical analysis.2 In this gauge, the action (43) is written as

S = d 3 det M ,
(48)
where
M

+ x) +
M ).
M = + F + x x 2(
(
(49)
This is the final supersymmetric DBI action that we need. Note that, again due to the partial
gauge fixing, tt = zz = 1 but = 0; furthermore, t = 0 and z = z but = 0.
The solution we are looking for is independent of t and z, so we take x(), F =
E() dt dz + B() dz d and () and insert them into the equations of motion.
In this process, we may set all t, z derivatives of x in the Lagrangian to zero, since terms
containing
these derivatives would not survive the equations of motion for x . We can then
rewrite det M as

M 2
x ) + 2B{E
2 },
0
|
x |2 1 E 2 + B 2 + 1 E 2
2(
(50)
where the prime denotes derivative. Variation of this quantity in , x and A yields
equations of motion. As we know that the supertube solution is obtained for E = 1, let us
set E = 1 after variation in these fields, which simplifies the resulting equations drastically.
The variation of the Lagrangian in (with since it is Majorana) gives (after setting
E = 1)
L =

B
0 z ) + (
0 z ) ,
(
0 z )
B 2 + 2B (
(51)
and the equation of motion

( 0 z ) = 0.
(52)
One can easily check that the full equations of motion are solved by E = 1 and (52) for any
functions B() and x(). This is a simple generalization of the original supertube solution.
2 However, when we consider quantization of near-circular supertubes, we should fix this extra gauge. There
we will set this scalar equal to .
125
With the representation (44), (52) becomes [ 3 116 ] = . This means that the second term in (46) should be -independent:

1
()
0
0
() =
(: any constant spinor).

(53)
0
0
1
Note that the fermionic part contains arbitrary functions of , an 8-component SO(8)
spinor () with positive chirality. The bosonic part has one from B() plus seven
gauge-invariant components from x(). Below we shall show that = 0 for the 1/4 supersymmetry. Hence the fermionic part of supertubes also involves eight arbitrary functions
of moduli fluctuation, which is expected from the number of remaining supersymmetries.
4.2. Supersymmetry
We now check whether the above solution preserves 1/4 supersymmetry. The 32 supersymmetry parameters of type-IIA string theory are split into satisfying 11 = ,
respectively. The supersymmetry transformations, combined with compensating kappa
transformation and world-volume diffeomorphism to restore the gauge, are
= + + (z) + ,

x = + (z) + x,

A = (z) + ( + x)

1
+ (z) M M + A + A ,
+
3
(54)
where = ( (z) + ) (with = t, z only in the superscript) and the 32 32

matrix (2) in our case is (using E = 1 and (52) to simplify the expression)
x ( )
11 0 .
(55)
B
Note that we have fixed the world-volume diffeomorphism only partially, i.e., T = t and
Z = z, so in the above definition we have two gauge-keeping parameters t and z but
nothing like .
The last term is absent after inserting our solution, so we have
Let us start from .

x ( ) 2
+ i
= + + 1 3
(56)
,
B
(z) = ( 0 z + 11 )
where we have used 11 = + , and for all Pauli matrices [ 116 ] is implicit. Considering the chiralities of , we can write them in the following form:

0
0
1
0
1
+
0
,
=
.
+
+ =
+
0
1
+
0
0
0
1
(57)
In order for our solution to be supersymmetric, the first term in the curly bracket should
vanish. This is true if = 0. The cancellation of the remaining second term (i 2 ) with
126
the term requires + = and + = 0. Therefore we have 1/4 of the components

preserved; + , and + are broken.
Let us now consider the remaining transformations for bosonic fields. Using the expression (55) and the conditions = 0, + = , the scalar variation x reduces
to

0
x = 2i [ 0 ]
(58)
.
One can see from the above variation that all the supersymmetry is broken if the constant
mode is nonzero. Therefore, as promised, we should set = 0 in order to have 1/4-BPS
deformation.
In order to check the last transformation, we should specify the vector potential giving
rise to the field strength F = dt dz + B() dz d. The simplest choice may be A(1) =
t dz + B()z d. For later use, let us also consider an alternative choice. To this end, we
decompose the magnetic field as

1
B0 =
(59)
d B(),
b() = B() B0 a (),
2

where a() is a periodic function, due to the fact d b() = 0. Then we can choose the
1-form potential as

A(2) = t + a() dz + zB0 d.
(60)
This form would be more convenient later, when we consider quantization. The two choices
are related by a gauge transformation.
After inserting our supertube solution with = 0, and turning on only, the supersymmetry transformation (54) for gauge field becomes

(1)
A0 = t 2it () ,
A(1)
z = 0,

(1)
A = 2it () + 2i ()B().
(61)
If the last term in the third line in (61) is absent, one can make a compensating gauge
transformation and have 1/4 supersymmetry. The subtle term proportional to ()B()
can be decomposed into the 0-mode piece plus the remainder as

1
(62)
B
d ()B(),
()B() r ()B() B.
2
The remainder piece {()B()}r can be rewritten as [ ], where we have a welldefined periodic function of inside the square bracket. This can safely be compensated
by a gauge transformation. The 0-mode piece can be written as d{2i B}, which is not
a gauge transformation in general. To ensure supersymmetry, we require B = 0, which
in turn implies that 0-modes are expressed in terms of and B nonzero modes.
127
4.3. Conserved charges

L
The electric displacement ( ) = E
is obtained from (50) and setting E = 1 after
differentiation:
|
x |2
1 M 2
x ) .
0 +
(63)
2(

B
B
One may show that all the higher-order terms in the curly bracket vanish for the supertube
solution. The electric displacement reduces to
( ) =
|
x |2
+ i ,
B
for the supertubes.
The linear momentum density conjugate to x becomes
( ) =
(64)
p = x .
(65)
Note that there is no correction from fermions. Consequently the field angular momentum
also takes the same form as in the bosonic case,

1

1
Lij =
(66)
d x i p j x j p i =
dx i dx j ,
2
proportional to the cross section area of the tube. To obtain the conserved total angular
momentum, we should add to it the spin angular momentum. The total angular momentum
is then

1
i
J ij =
(67)
dx i dx j
d B ij ,
4
where ij [i j ] is the anti-Hermitian SO(8) Lorentz generators acting on spinors,
which we understand as being reduced from 16 16 to 8 8 and acts on positive chirality
subspace.
4.4. Quantization and entropy of near-circular supertubes
In order to quantize the modes identified before, we now fix the remaining diffeomorphism. As in the previous section, let us consider a small deformation from the circular
tube with radius R0 in the 12 plane. Then x 1 + ix 2 = R()ei with |R() R0 | R0 ,
and |x i ()| R0 for all i = 3, 4, . . . , 8. To identify the quantization conditions, we have to
know the quadratic piece of the gauge-fixed Lagrangian. The full gauge fixed Lagrangian
density becomes

L = L2b + L2f ,
2

x | + R 2 + (R )2 + B 2
L2b = B 2 |x |2 + R 2 |
2

)2 2EB(x x + RR
) + 1 E 2 |
+ (x x + RR
x | + R 2 + (R )2 ,
128

2
) + (1 E) |
L2f = 2i B 2 B(x x + RR
x | + R 2 + (R )2

) ,
2i B |x |2 + R 2 + (E 1)(B + x x + RR
(68)
where vectors are in the six-dimensional x i space. With the choice of the vector potential
B = B0 a . We expand this
Az = t + a, the field strengths are given by E = 1 + a,
i
action up to quadratic order in a(, t), r(, t), x (, t), (, t). The resulting quadratic
Lagrangian density is

R0 R0
a
a + 2r + x x + r r + iB0
L2 =
B0 B0
R2
B 2 + R02 2
r + |x |2 + 0 a 2 .
+ 0
(69)
2B0
2B0
Working within the 1/4-BPS phase moduli space only, the terms of quadratic time derivative may be dropped since the BPS states are time-independent.
The mode expansion for eight bosonic/fermionic fields are given as

an (t)ein ,
Az = t + a(, t) = t +
n=0
R(, t) = R0 + r(, t) = R0 +
x i (, t) = x0i +
rn (t)ein ,
n=0
xni (t)ein ,
n=0
(, t) =
n (t)ein ,
(70)
n=0
i = (x i ) . The transverse center of mass positions x i

with an = an , rn = rn and xn
n
0
would not affect the following analysis, so we will neglect them. (, t) and n s carry
eight components. Inserting the mode expansions into (69) and integrating over and z
(for the zero mode in the z direction), we get the Lagrangian

R0
R0
xn x n + iB0 n n . (71)
inrn rn +
in an + 2rn a n + in
L2 = 2Lz
B0
B0
n=0
Introducing
R0
an ,
B0
the Lagrangian becomes
Xn rn i
(72)

L2 = 2Lz
i(n + 1)Xn+
xn x n + 2iB0 n n .
Xn+ + i(n 1)Xn
Xn + 2in
n=1
(73)
After the Dirac quantization procedure, the resulting commutation relations read

Xm , Xn
=
1
m,n ,
2Lz (n 1)

1
m,n ,
m , n =
2Lz (2B0 )
i j
x m , xn =
1
m,n i,j ,
2Lz (2n)
(74)
129
with all the other commutators vanishing (i, j = 3, 4, . . . , 8, m, n = 1, 2, . . .). Note that the
radius rn and gauge field an modes mix nontrivially in the commutation relation. Special
remark for X1 is in order: the above relation is meaningless for X1 . This is natural since
the dipole deformation of radius R() is nothing but the translation of supertube along 12
plane [28]. So we expect that there are true zero modes having the quadratic time derivative
terms only for their kinetic part.3 We are not interested in this translational zero mode.
The conserved charges are expressed as

J12 = R02 +
(75)
2|rn |2 iB0 n 12 n ,
n>0
Q0 Q1 = R02 + 2
2

2
2
rn i nan R0 + n2 |rn |2 + |
,
x
|
+
nB
|
|
n
0 n

B
(76)
n>0
where |A|2 = A A is our ordering convention. The two charges commute, as they should.
The first expression (75) determines R02 in terms of oscillators and J . Inserting this into
(76), we obtain

n(n + 1)|Xn+ |2 + n(n 1)|Xn |2 + 2n2 |
xn |2
Q0 Q1 J =
n>0
+ 2nB0 |n |2 + iB0 n 12 n .
(77)
Here we choose a basis for the spinor such that i12 is diagonal with four 1 eigenvalues
( = 1, 2, 3, 4). Furthermore, we normalize the oscillators
with corresponding modes n
in the canonical way as follows:
1
Yn ,
Xn =
2Lz (n 1)
1
=
n
n
2Lz (2B0 )
xn =
1
2Lz (2n)
yn ,
(78)
with n > 0. The new oscillators satisfy the commutation relations [Yn , Yn
] = [yni , yni ] =
, } = 1. Then we can rewrite (77) as
{n
n
2Lz (Q0 Q1 J )

1 2
1 2
n|Yn+ |2 + n|Yn |2 + n|
n+ + n
n
yn |2 + n +
=
2
2
n>0
8

4

1
1
n+
,
Nn+
Nn
=
(79)
nNnI +
+ n
2
2
n>0
I =1
=1
where the last expression contains 8 classes of bosonic number operators NnI (I =
with spins. 2L (Q Q
1, 2, . . . , 8) and 4 classes of fermionic number operators Nn
z
0 1
J ) may take half-integer eigenvalues.
3 In the gravity description, this mode is like the freedom of translating black holes or supertubes.
130
The entropy can be counted by considering the generating function tr(2N ) =

n
n=0 dn where N is the number operator (79), and obtaining the degeneracy dn with
n
=
2L
(Q
z
0 Q1 J ) being a large half-integer. One has
2

8
4
4

2N
2m
2m+1
2m1
=
1
1+
1+
.
tr
(80)
m=1
m=1
The saddle-point evaluation of the degeneracy

1
tr(2N )
dn =
d n+1 ,
2i
m=1
(81)

n
requires the behavior of the functions f (z)

n=1 (1 z ) near z 1 . Up to the prefactors that we do not need, we have

2

1
2 1
,
f (z) exp
.
f+ (z) exp
(82)
12 1 z
6 1z
We also note that, as long as we are interested in saddle point evaluation for large
2Lz (Q0 Q1 J ), we may regard 2n1 in (80) as 2n . Using the above formulae and
noting that log (1

) for 1 , one can see that (81) gets dominant contribu2
(8+4)
tions near log 12(n+1)
n , where cB = 8 and cF = 4 denotes boson/fermion
contributions, respectively. The result is

n
= exp[2 n ] (cB = 2cF = 8),
dn exp 2 (cB + cF )
(83)
12
up to the prefactor, which is a suitable power of n. Inserting n = 4Lz (Q0 Q1 J ), we get

the final expression for the supertube entropy

S = log(dn ) = 4 Lz (Q0 Q1 J ).
(84)
As expected, this is cB + cF = 12 times the entropy (42) from one boson.

One may consider more general case of the multiple circular supertubes carrying SO(8)
Cartans Ja = J2a1,2a for a = 1, 2, 3, 4. The relevant background is described by
x2a1 + ix2a = Ra ei
(85)
with
0 B0 = T2
4

Ra2 .
(86)
a=1
The corresponding angular momentum becomes Ja = T2 Ra2 . Repeating the above analysis,
one may get straightforwardly

4

S = 4 Lz Q0 Q1
(87)
|Ja | ,
a=1
in an appropriate near circular limit.
131
5. Microstates in supermembrane picture

In this section we derive the entropy of supertube with charges Q1 , Q0 and angular
momentum Ja (a = 1, . . . , 4) from the 11-dimensional M-theory point of view including
the contribution from bosons and fermions. We study equations of motion for a supermembrane with winding number and momentum along the 11th direction, which preserves 1/4
supersymmetry. This approach gives a simple derivation of BPS equations because fields
on the supermembrane are only 11 bosons X M = (t, z, x i (i = 1, . . . , 8), x ) and a Majorana fermion , which denote the embedding of the supermembrane into the superspace.
Let us investigate BPS equations for the supermembrane. These are obtained by analyzing the Killing spinor equations of [37]
M (1 + ) = 0,
X M = i M + i
= + (1 + ) = 0,
M
are 11-dimensional gamma matrices, and is defined as
1

abc a L b M c N LMN .
3! det P [G(X M , )]ab
where
(88)
(89)
Here P [G(X M , )]ab is the induced metric on the world-volume, and M are super M d. Note that satisfies 2 = 1 and tr = 0.
invariant 1-forms, M = dX M i
By using the former property, can be eliminated and the Killing spinor equations simply
become
M = 0,
(90)
(1 ) = 0.
(91)
Since we are considering the supermembrane which corresponds to the supertube, the solution should be 1/4 supersymmetric [12]:
1 + tz 1 + t
0 ,
2
2
where 0 is an arbitrary Majorana spinor.
We then find that Eq. (90) has the solution (92) iff
=
(92)
1 tz 1 + t
(93)
0 ,
2
2
where 0 is an arbitrary Majorana spinor. We see that has 8 real components. It is easy
to verify the following relations:
=
z =
i =
t =
i = t =
z = 0,
t = = z = tz .
(94)
Next we consider Eq. (91). The world-volume coordinates on the supermembrane are
identified with (, , z), and we assume that x i , x and depend only on and . Then
the super invariant 1-forms M are expressed as
132
M = M d + M d (M = t, i, ),
z = dz + z d + z d,
(95)
M and M = X M i
M . Notations X and X represent
where M = X M i
X
X
and , respectively. Then is written as

1 t i
t i tiz + i i iz + t t tz + i j ij z ,
=
X

2

X = t 2 + i 2 + 2 t 2 + i 2 + 2 + t t + i i + ,
(96)
where we have defined the volume factor X. From these equations, we find that Eq. (92)
becomes the solution of (91) when
i = k i ,
t k t = k ,
(97)
where k is a constant. Later we set k = 1. Therefore the BPS equations of the supermembrane corresponding to the supertube are given by (93) and (97).
Now that we have obtained the BPS equations which minimize the energy of the supermembrane, our next task is to derive conditions to fix two charges Q1 , Q0 and angular
momenta. The two charges are winding number and momentum along the x direction, so
the conditions are written as
1
Q1 =
2
x
,
2R11

Q0 =
d R11 p ,
1
Jij =
2

1
d x i p j x j p i S ij ,
2
(98)
where R11 is the radius of the 11th circle, and p and S are the conjugate momenta to x
and , respectively. Note that 2Q1 and Q0 Lz are integers. From now on we identify
and with t = 2Q1 R11 and x = 2Q1 R11 , respectively. With this choice, the first
equation in (98) is trivially satisfied.
To explicitly write down the other conditions, we need the conjugate momenta. These
are calculated from the supermembrane action:
SM2 = SNG + SWZ ,

SNG = T2 d d dz X,

t i x
t + i t
t .
SWZ = T2 d d dz ix
(99)
133
We have reduced the degrees of freedom of by using (93), and assumed the same ansatz
as (95). By imposing the BPS condition (97), the conjugate momenta are given by
p i = T2 x i ,
x |2
T2 |
t ,
2iT2
p =
x
t.
S = 2iT2 x
(100)
(101)
(102)
The conditions for fixing charges and angular momenta are then written as

2
Q1 Q0
1
t ,
=
d |
x | 2ix
T2
2

1
Ja
t 2a1,2a ,
=
(103)
d x 2a1 x 2a + x 2a x 2a1 ix
T2 2
where a = 1, 2, 3, 4 labels the SO(8) Cartans. These are the constraints obtained by fixing
Q1 , Q0 and angular momenta.
Now let us consider the quantization of x i and . From the BPS equations ( +
)x i = 0, x i contain only right moving modes,
|m|
p0i
i
x i eim( ) .
x i (, ) = x0i +
(104)
( ) +
2Lz T2
m m
4T2 Lz
m=0
xi
pi
Since
and
are related as (100), we need the Dirac quantization of the constrained
system. After some calculations, we obtain the commutation relations [x i (), p j ( )] =
i ij

2Lz ( ), and hence
i j
xm , xn = ij mn .
(105)
The Majorana fermion is also treated similarly. From the equations of motion obtained
by (99) and BPS equations (97), we only need the right moving modes and ( =
1, . . . , 8) are expanded as

1
m eim( ) .
=
(106)
8T2 x Lz m
Since and S are related as (102), after the calculation of Dirac brackets we obtain
{ , S } = 2Li z ( ), and hence

m , n = mn ,
(107)
where = i T C 1 t .
After the quantization, the constraints (103) reduce to
2Lz Q0 Q1 =
8

i=1 m=1
2Lz Ja = i
i i
mxm
xm +
8
mm
m ,
=1 m=1
8

2a1 2a

1 2a1,2a
2a 2a1
xm xm xm
i
xm
m .

2 m
m=1
=1 m=1
(108)
134
Note that 2Lz Q0 Q1 is an integer. The volume of the phase moduli space is obtained
exactly by counting the configurations of x i and which satisfy the above constraints.
We choose the spinor eigen basis digonalizing i 2b1,2b (b = 1, 2, 3). One has
i 2b1,2b n,s = sb n,s ,
(109)

where s = (s1 , s2 , s3 ) with sb = 1. Since 4a=1 2a1,2a n,s = n,s , the eigenvalue s4
of i 78 is given by s1 s2 s3 .
The combination of the constraints in (108) give

N = 2Lz Q0 Q1
Ja
=
m=1
a
4

(m + 1)Aam Aam + (m 1)Bam

Bam
a=1

4

1
m+
+
sa ms ms ,
2
s
(110)
a=1
where we have defined

1 2a1
1 2a1
2a
2a
Aam = xm
,
Bam = xm
,
+ ixm
ixm
2
2

Aam , Abm = ab ,
= ab .
Bam , Bbm
(111)
The second constraint in (108) is written as

1
Bam Bam Aam Aam

sa m
s ms = 2Lz Ja .
2
s
m=1
(112)
m=1
We can consider this determines Ba1 mode which is absent from (110). Thus the number
of microstates can be counted by taking only the constraint (110) into account.
We consider the case N 1. As in the previous sections, let us compute the partition
function
G(w) = tr w N =
dn w n
n=0
2(1 + w 1 )(1 w)4 (1 + w m )8

(1 + w)(1 + w 2 )
(1 w m )8
m=1
(1 w)4 (1 + w 1 )
4 (0, )
10
23 (1 + w)(1 + w 2 ) 12 ( )
4

(0, 1/ )
ln w 4 01
(1 w)4 (1 + w 1 )
,
= 3
2
2 (1 + w)(1 + w 2 )
12 (1/ )
(113)
135
where we have used the eta function defined in Eq. (34) and
10 (0, ) = 2w 1/8

2
1 wm 1 + wm ,
m=1

2
1 w m 1 w m1/2 ,
01 (0, ) =
(114)
m=1
with w = e2i , and the modular transformations of the eta (35) and
10 (0, 1/ ) = i 01 (0, ).
The result (113) gives the asymptotic formula

log w 8
2 2
,
exp
G(w) 4
2
log w
(115)
(116)
for w 1 . The saddle point approximation enables us to derive the final result for the
degeneracy for N = n as

1
G(w)
2 2n
dw
e
.
dn =
(117)
2i
w n+1
In this way we obtain the entropy

4

S = log dN 4 Lz Q0 Q1
(118)
|Ja | ,
a=1
in agreement with (87).

We would like to emphasize that we have not used the near circular condition (Q0 Q1
J )/J 1 anywhere for the evaluation of the entropy in this section. So the entropy formula (118) will be valid beyond the near circular limit. Let us confirm this for the case of
= 0 and Ja = 0 (a = 2, 3, 4) for simplicity. The expectation value of the radius squared
is given by

4

1
1
1
1
2
2
A Aam + Bam Bam .
R
(119)
d |
x| =
2
2Lz T2
m am
m
m=1 a=1
As discussed in the introduction, we need to have RS R for the validity of our counting
of microstates. The value (119) of R should be estimated under the constraints (108) or
(110) and (112). The constraint (112) tells us that we must excite certain amount of Bam
and Aam , and R is smaller if those with larger m are excited, but there is an upper limit on
possible m from the first equation in (108). We thus find that the minimum of R is attained
when we excite 2Lz J1 of B1m for m Q0 Q1 /J1 , giving

J1
J1
R
(120)
.
Q0 Q1 T2
Thus, as long as J1 /Q0 Q1 is not very small, RS R can be satisfied for large J1 and
therefore the entropy formula (118) is valid.
136
6. Conclusions
In this paper we have presented two approaches to the counting of the number of microstates for supertubes specified by the F1 and D0 charges and angular momenta, and
derived consistent entropy formula.
There are corresponding supergravity microstates and, thus, we count the degeneracy
of the geometries with the asymptotic geometry and charges fixed. The correspondence
demonstrates the existence of the quantized microstates specified by the distinguishable
supergravity fields. Thus although we do not know how to do precisely, there must be a
clear way to sum over geometries with an appropriate measure. This has been suspected in
many cases including the thermal AdS/CFT correspondence [38], where one has competing
contributions from the AdS Schwarzschild black hole and the Euclidean AdS geometry of
temporal circle size related to the inverse temperature.
Indeed the related black hole entropy may be understood from the microstates. Since
the horizon area of the supergravity supertubes are zero in any cases, the situation here
is rather confusing. However, the proposal of Sen [39] may be applied and the stretched
horizon area of the rotationally symmetric solution may be shown to agree to the entropy
[33].
The situation of D1-D5-P [27] which is related to F1-D0-D4 by a U -duality is different. [D1 (5)-D5 (56789)-P (5) where the numbers in the parenthesis represent momentum
direction or extending directions, is related to F1 (5)-D0-D4 (6789) by the successive
transformations of S, T56, S, T56789.] The rotationally symmetric black hole solution of
D1-D5-P has a nonvanishing horizon area and the corresponding entropy can be explained
by the CFT counting [41].
When we add J12 angular momentum to the F1 (5)-D0-D4 (6789)-D2 (512 ), the configuration describes the supertubes intersecting with D4-branes, which preserve four real
supersymmetries. By the same U -duality transformation, the above is related to the D1
(5)-D5 (56789)-P (5)-KK5 (678912 ) [32] where 12 represents that the KK monopole or
the D2 form a curve in the (12) plane. Similarly J34 may be added too.
Since the supertube ending on D4 in Refs. [11,15,19,40] has angular momenta in (6789)
plane only, e.g., F1 (5)-D0-D4 (6789)-D2 (567 ), the above configurations of the curve in
(1234) plane are different in their expansion directions of D2 and have not been found in
the field theory description.
Considering the supertubes suspended between
two D4-branes

of large separations, the
4
corresponding entropy is expected as S =
L(Q0 Q1 |J |), where the sum is over
2
the SO(4) Cartans in (6789) plane. The curve cannot escape to the (1234) plane because
the supertube ends on D4-branes. Thus only four arbitrary bosonic fluctuations remain.
Furthermore they preserve four real supersymmetries and the number
of arbitrary fermionic
fluctuations should be reduced to four. Hence one has the 1/ 2 factors. (The half factor
for the numbers of degrees goes inside of the square root.)
For many D4-branes, the above formula would have a straightforward generalization.
For the supertubes connecting D4-branes, one may have in principle five independent
charges; D0, F1, D4 and two Cartans of the angular momenta.
137
For these cases, one has clean examples of the gravity microstates, black hole solutions
whose horizon area reproduces the entropy, and the corresponding filed theory description
of the microstates. But the detailed and complete construction awaits more endeavors.
Finally, the formula for the cross sectional area A gs 2s N0 N1 in (1) is reminiscent of
the quantum foam in Ref. [42]. Since the counting and the partition function may be also
related to the black hole partition function [43], there seems to be some connections of
the microstates to the quantum foam in Ref. [42]. Any clue in this direction will be very
interesting.
Acknowledgements
Y.H. would like to thank H. Kajiura, N. Sasakura and S. Sugimoto. D.B. is supported in
part by 2004 UOS Academic Research Grant. The work of Y.H. was supported in part by
a Grant-in-Aid for JSPS fellows. S.K. is supported in part by BK21 project of the Ministry
of Education, Korea, and a 2003 Interdisciplinary Research Grant of Seoul National University. NO was supported in part by Grants-in-Aid for Scientific Research Nos. 12640270
and 02041.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
D. Mateos, P.K. Townsend, Supertubes, Phys. Rev. Lett. 87 (2001) 011602, hep-th/0103030.
D. Bak, K. Lee, Noncommutative supersymmetric tubes, Phys. Lett. B 509 (2001) 168, hep-th/0103148.
J.H. Cho, P. Oh, Super D-helix, Phys. Rev. D 64 (2001) 106010, hep-th/0105095.
R. Emparan, D. Mateos, P.K. Townsend, Supergravity supertubes, JHEP 0107 (2001) 011, hep-th/0106012.
D. Bak, S.W. Kim, Junctions of supersymmetric tubes, Nucl. Phys. B 622 (2002) 95, hep-th/0108207.
D. Bak, A. Karch, Supersymmetric braneantibrane configurations, Nucl. Phys. B 626 (2002) 165, hepth/0110039.
I. Bena, The polarization of F1 strings into D2 branes: Aut Caesar aut nihil, Phys. Rev. D 67 (2003) 026004,
hep-th/0111156.
D. Bak, N. Ohta, Supersymmetric D2anti-D2 strings, Phys. Lett. B 527 (2002) 131, hep-th/0112034.
D. Mateos, S. Ng, P.K. Townsend, Tachyons, supertubes and brane/anti-brane systems, JHEP 0203 (2002)
016, hep-th/0112054.
D. Mateos, S. Ng, P.K. Townsend, Supercurves, Phys. Lett. B 538 (2002) 366, hep-th/0204062.
M. Kruczenski, R.C. Myers, A.W. Peet, D.J. Winters, Aspects of supertubes, JHEP 0205 (2002) 017, hepth/0204103.
Y. Hyakutake, N. Ohta, Supertubes and supercurves from M-ribbons, Phys. Lett. B 539 (2002) 153, hepth/0204161.
J.H. Cho, P. Oh, Elliptic supertube and a BogomolnyiPrasadSommerfield D2-braneanti-D2-brane pair,
Phys. Rev. D 65 (2002) 121901.
D. Bak, N. Ohta, M.M. Sheikh-Jabbari, Supersymmetric braneanti-brane systems: matrix model description, stability and decoupling limits, JHEP 0209 (2002) 048, hep-th/0205265.
D. Bak, K.M. Lee, Supertubes connecting D4 branes, Phys. Lett. B 544 (2002) 329, hep-th/0206185.
N.E. Grandi, A.R. Lugo, Supertubes and special holonomy, Phys. Lett. B 553 (2003) 293, hep-th/0212159.
N. Drukker, B. Fiol, J. Simon, Goedels universe in a supertube shroud, Phys. Rev. Lett. 91 (2003) 231601,
hep-th/0306057.
L. Martucci, P.J. Silva, JHEP 0308 (2003) 026, hep-th/0306295.
S. Kim, K. Lee, Dyonic instanton as supertube between D4 branes, JHEP 0309 (2003) 035, hep-th/0307048.
C.J. Kim, Y.B. Kim, O.K. Kwon, P. Yi, Tachyon tube and supertube, JHEP 0309 (2003) 042, hep-th/0307184.
138
[21] H. Takayanagi, Boundary states for supertubes in flat spacetime and Goedel universe, JHEP 0312 (2003)
011, hep-th/0309135.
[22] H. Elvang, R. Emparan, Black rings, supertubes, and a stringy resolution of black hole non-uniqueness,
JHEP 0311 (2003) 035, hep-th/0310008.
[23] S.D. Mathur, A. Saxena, Y.K. Srivastava, Constructing hair for the three charge hole, Nucl. Phys. B 680
(2004) 415, hep-th/0311092.
[24] I. Bena, P. Kraus, Three charge supertubes and black hole hair, hep-th/0402144.
[25] B.C. Palmer, D. Marolf, Counting supertubes, JHEP 0406 (2004) 028, hep-th/0403025.
[26] I. Bena, Splitting hairs of the three charge black hole, hep-th/0404073.
[27] O. Lunin, Adding momentum to D1D5 system, JHEP 0404 (2004) 054, hep-th/0404006.
[28] D. Bak, Y. Hyakutake, N. Ohta, Phase moduli space of supertubes, Nucl. Phys. B 696 (2004) 251, hepth/0404104.
[29] N. Drukker, Supertube domain-walls and elimination of closed time-like curves in string theory, hepth/0404239.
[30] S. Giusto, S.D. Mathur, A. Saxena, Dual geometries for a set of 3-charge microstates, hep-th/0405017;
S. Giusto, S.D. Mathur, A. Saxena, 3-charge geometries and their CFT duals, hep-th/0406103.
[31] W.H. Huang, Tachyon tube on non-BPS D-branes, hep-th/0407081.
[32] O. Lunin, J. Maldacena, L. Maoz, Gravity solutions for the D1D5 system with angular momentum, hepth/0212210.
[33] O. Lunin, S.D. Mathur, AdS/CFT duality and the black hole information paradox, Nucl. Phys. B 623 (2002)
342, hep-th/0109154;
S.D. Mathur, Where are the states of a black hole?, hep-th/0401115.
[34] M.B. Green, J.H. Schwarz, E. Witten, Superstring Theory, Cambridge Univ. Press, Cambridge, 1987.
[35] See, for instance, (15) of P. Blanchard, S. Fortunato, H. Satz, The Hagedorn temperature and partition thermodynamics, hep-ph/0401103.
[36] M. Aganagic, C. Popescu, J.H. Schwarz, Gauge-invariant and gauge-fixed D-brane actions, Nucl. Phys.
B 495 (1997) 99, hep-th/9612080.
[37] B. de Wit, J. Hoppe, H. Nicolai, On the quantum mechanics of supermembranes, Nucl. Phys. B 305 (1988)
545.
[38] E. Witten, Anti-de Sitter space, thermal phase transition, and confinement in gauge theories, Adv. Theor.
Math. Phys. 2 (1998) 505, hep-th/9803131.
[39] A. Sen, Extremal black holes and elementary string states, Mod. Phys. Lett. A 10 (1995) 2081, hepth/9504147.
[40] For the discussions of the dyonic instanton, see N.D. Lambert, D. Tong, Phys. Lett. B 462 (1999) 89, hepth/9907014;
M. Zamaklar, Phys. Lett. B 493 (2000) 411, hep-th/0006090;
E. Eyras, P.K. Townsend, M. Zamaklar, JHEP 0105 (2001) 046, hep-th/0012016.
[41] A. Strominger, C. Vafa, Microscopic origin of the BekensteinHawking entropy, Phys. Lett. B 379 (1996)
99, hep-th/9601029.
[42] A. Iqbal, N. Nekrasov, A. Okounkov, C. Vafa, Quantum foam and topological strings, hep-th/0312022.
[43] H. Ooguri, A. Strominger, C. Vafa, Black hole attractors and the topological string, hep-th/0405146.
Dual models of gauge unification in various

dimensions
Wilfried Buchmller a , Koichi Hamaguchi a , Oleg Lebedev a ,
Michael Ratz b
a Deutsches Elektronen-Synchrotron DESY, 22603 Hamburg, Germany
b Physikalisches Institut der Universitt Bonn, Nussallee 12, 53115 Bonn, Germany
Abstract
We construct a compactification of the heterotic string on an orbifold T 6 /Z6 leading to the standard model spectrum plus vector-like matter. The standard model gauge group is obtained as an
intersection of three SO(10) subgroups of E8 . Three families of SO(10) 16-plets are localized at
three equivalent fixed points. Gauge coupling unification favours existence of an intermediate GUT
which can have any dimension between five and ten. Various GUT gauge groups occur. For example,
in six dimensions one can have E6 SU(3), SU(4) SU(4) U(1)2 or SO(8) SO(8), depending
on which of the compact dimensions are large. The different higher-dimensional GUTs are dual to
each other. They represent different points in moduli space, with the same massless spectrum and
ultraviolet completion.
PACS: 12.10.-g; 11.25.-w; 11.25.Mj
1. Embedding the standard model in E8

The symmetries and the particle content of the standard model point towards grand
unified theories (GUTs) as the next step in the unification of all forces. Left- and rightE-mail address: koichi.hamaguchi@desy.de (K. Hamaguchi).
doi:10.1016/j.nuclphysb.2005.01.038
140
W. Buchmller et al. / Nuclear Physics B 712 (2005) 139156
handed quarks and leptons can be grouped in three SU(5) multiplets [1], 10 = (qL , ucR , eRc ),
5 = (dRc , L ) and 1 = Rc . Here we have added right-handed neutrinos which are suggested
by the evidence for neutrino masses. All quarks and leptons of one generation can be
unified in a single multiplet of the GUT group SO(10) [2],
16 = 10 + 5 + 1.
(1)
The group SO(10) contains as subgroups the PatiSalam group [3], GPS = SU(4)
SU(2)SU(2), the GeorgiGlashow group SU(5), GGG = SU(5)U(1), and the flipped
SU(5) group, Gfl = SU(5) U(1) [4], where the right-handed up- and down-quarks are
interchanged, yielding another viable GUT group.
It is a remarkable property of the standard model that the matter fields form complete
SO(10) multiplets whereas the gauge and Higgs fields are split multiplets. They have to
be combined with other split multiplets, not contained in the standard model, in order to
obtain a complete unified theory. It is also well known [5] that exceptional groups play an
exceptional role in grand unification, and the embedding
SO(10) E6 E8
(2)
appears, in particular, in compactifications of the heterotic string [6] on CalabiYau manifolds [7].
As we shall see, complete SO(10) matter multiplets together with split gauge and Higgs
multiplets arise naturally in orbifold compactifications of higher-dimensional unified theories. Orbifold compactifications have first been considered in string theory [8,9] and
subsequently in effective higher-dimensional field theories [10,11]. They provide a simple
and elegant way to break GUT symmetries, while avoiding the notorious doublettriplet
splitting problem. More recently, it has been shown how orbifold GUTs can occur in orbifold string compactifications [1214].
In the following we shall first search for a scheme of ZN twists which allows to break
E8 , a common ingredient of string models, to the standard model group. A ZN twist is an
element of the gauge group G, with
P = exp (2iVN H ),
P N = 1.
(3)
Here the generators H i form the (Abelian) Cartan subalgebra of G, and VN is a real vector.
The twist P acts on the Cartan and step generators E as follows:
P H i P 1 = H i ,
P E P 1 = exp (2iVN )E ,
(4)
where is a root associated with E . Clearly, P breaks G to a subgroup containing all

step generators which commute with P , i.e., [P , E ] = 0.
The symmetry breaking is conveniently expressed in terms of the Dynkin diagrams.
This technique has been employed to classify possible symmetry breaking patterns in string
models [1517] and, more recently, in orbifold GUTs [18,19]. Starting with the extended
Dynkin diagram which contains the most negative root in addition to the simple roots,
regular subgroups of a given group are obtained by crossing out some of the roots of the
Dynkin diagram. In particular, the action of the ZN orbifold twist essentially amounts to
141
Fig. 1. SO(10) breaking patterns by a Z2 twist. The action of the PatiSalam twist is indicated by crosses, while
that of the GeorgiGlashow twist is indicated by a slash.
crossing out a root with the (Coxeter) label N , or more generally, roots whose labels sum
up to N .
As an example, consider the breaking of SO(10), displayed in Fig. 1. For each simple
root the Coxeter label is listed, which determines the order of the twist required for the corresponding symmetry breaking. Crossing out one of the nodes with label 2 breaks SO(10)
to the semi-simple subgroup GPS , while crossing out one of the roots with label 1 together
with the most negative root breaks SO(10) to GGG . The intersection of the two groups
gives the standard model with an additional U(1) factor [11],
GGG GPS = SU(3) SU(2) U(1)2 GSM ,
(5)
where means modulo U(1) factors. Under the Z2 twisting, the group generators divide into those with positive and negative parities P with respect to the twist. Combining
the two parities PGG and PPS , one can construct the third Z2 parity PGG PPS = Pfl which
breaks SO(10) to the flipped SU(5),
Pfl
SO(10) Gfl = SU(5) U(1) .
(6)
The standard model group GSM can also be obtained as an intersection of the two SU(5)
embeddings, GGG and Gfl ,
GGG Gfl = SU(3) SU(2) U(1)2 GSM .
(7)
As another example, consider now E6 breaking to the standard model group. From the
extended Dynkin diagram Fig. 2 it is clear, in analogy with the SO(10) breaking, that three
Z2 twists,
PA
E6 SO(10) U(1),
PB
E6 SU(6) SU(2) ,
PC
E6 SU(6) SU(2) ,
(8)
can break E6 to the standard model up to U(1) factors,

SO(10) U(1) SU(6) SU(2) SU(6) SU(2)
= SU(3) SU(2) U(1)2 GSM .
(9)
As in the SO(10) example, one can check that the same breaking can be obtained as an
intersection of three different SO(10) embeddings in E6 which correspond to the twists
PA PC
E6 SO(10) U(1) ,
E6
PA PB PC
SO(10) U(1) ,
(10)
142
Fig. 2. E6 breaking patterns under Z2 twisting. Three different Z2 twists are indicated by crosses, a slash and a
backslash, respectively.
Fig. 3. E8 breaking to E6 SU(3).
such that
SO(10) U(1) SO(10) U(1) SO(10) U(1)
= SU(3) SU(2) U(1)2 GSM .
(11)
Let us remark that it is not possible to distinguish the three SO(10) embeddings in E6
(as well as GGG and Gfl embeddings in SO(10)) at the level of Dynkin diagrams. The
corresponding subalgebras are related by Weyl reflections within the embedding group. To
distinguish them, an explicit analysis of the shift vectors VN is required.
Our final goal is to break E8 to the standard model gauge group. This can be achieved
by combining the above three Z2 twists with a Z3 twist which breaks E8 to E6 SU(3)
(cf. Fig. 3). The Z2 twists can then also break the SU(3) factor to SU(2) U(1). In this
way one obtains three Z6 twists which break E8 to subgroups containing SO(10),
P6
E8 SO(10) SU(3) U(1),
P6
E8 SO(10) SU(2) U(1)2 ,
P6
E8 SO(10) SU(2) U(1)2 ,
(12)
such that the intersection is the standard model group up to U(1) factors,
SO(10) SU(3) U(1) SO(10) SU(2) U(1)2
SO(10) SU(2) U(1)2 GSM .
(13)
In an orthonormal basis of E8 roots, three Z6 shift vectors which realize the described
symmetry breaking read explicitly:

1 1 1
, , , 0, 0, 0, 0, 0 ,
V6 =
3 3 3

7 7 1 1 1 3 3 3
, , , , , , , ,
V6 =
12 12 12 4 4 4 4 4

7 13 7 3 1 3 3 3
V6 =
(14)
, , , , , , , .
12 12 12 4 4 4 4 4
Note that the differences between the Z6 shift vectors are Z2 shift vectors,

1 1 1 1

W2 = V6 V6 = 0, , , , , 0, 0, 0 ,
2 2 2 2

1 1 1 1 1 3 3 3

, , , , , , , ,
W2 = V6 V6 =
4 4 4 4 4 4 4 4
143
(15)
which will play the role of Wilson lines in the next section.
To summarize, in this section we have presented a group-theoretical analysis of E8
breaking to the standard model with intermediate E6 and SO(10) GUTs, suggested by the
structure of matter multiplets.
2. Orbifold compactification
Let us now construct an orbifold compactification of the heterotic string, which realizes
the symmetry breaking described above. As is clear from the above discussion, we will
need a Z6 or a higher-order orbifold and choose the former for simplicity.
In the light cone gauge the heterotic string [6] can be described by the following bosonic
world-sheet fields: 8 string coordinates X i , i = 1, . . . , 8, 16 internal left-moving coordinates X I , I = 1, . . . , 16, and 4 right-moving fields i , i = 1, . . . , 4, which correspond to
the bosonized NeveuSchwarzRamond fermions (cf. [16,20,21]). The 16 left-moving internal coordinates are compactified on a torus. The associated quantized momenta lie on
the E8 E8 root lattice. In an orthonormal basis, vectors of the E8 root lattice are given by

1
1
n 1 + , . . . , n8 +
pE8 = (n1 , . . . , n8 ) or
(16)
,
2
2

with integer ni satisfying 8i=1 ni = 0 mod 2. The massless spectrum of this 10D string is
10D supergravity coupled to E8 E8 super-YangMills theory.
To obtain a four-dimensional theory, 6 dimensions of the 10D heterotic string are compactified on an orbifold. In our case, this is a Z6 orbifold obtained by modding a 6D torus
together with the 16D gauge torus by a Z6 twist,
O = T 6 TE8 E8 /Z6 .
(17)
On the three complex torus coordinates zi , i = 1, 2, 3, the Z6 twist acts as

i
zi e2iv6 zi .
(18)
Here 6v6 has integer components. The compact string coordinates are described by the
complex variables Z i = X 2i1 +iX 2i , i = 1, . . . , 3. The Z6 action on the string coordinates
reads, up to lattice translations (cf. [16]),
i
Z i ( = 2) = e2ikv6 Z i ( = 0),
i ( = 2) = i ( = 0) kv6i ,
k = 0, . . . , 5,
X I ( = 2) = X I ( = 0) + kV6I ,
(19)
(20)
144
where 6V I is an E8 E8 lattice vector.

The torus T 6 is spanned by basis vectors e , = 1, . . . , 6. In general, a torus allows for
the presence of Wilson lines, i.e., a translation by a lattice vector n ei can be accompanied
by a shift of the internal string coordinates,
X i ( = 2) = X i ( = 0) + 2n ei ,
X ( = 2) = X (
I
n Z,
= 0) + n WI .
(21)
Here the discrete Wilson lines W are restricted by symmetry and by modular invariance.
The basis vectors e are taken to be simple roots of a Lie algebra, whose choice is
dictated by the required symmetry of the lattice. In our case the lattice must have a Z6
symmetry and allow for the existence of 3 independent V6 shift vectors (14) (or two Wilson
lines of order 2). This leaves two possibilities for the Lie lattice [16]
G2 SU(3) SO(4)
or SU(3)[2] SU(3) SO(4).
(22)
We shall base our analysis on the first lattice, which has recently been studied in detail by
Kobayashi, Raby and Zhang [12]. These authors have obtained models with the PatiSalam
gauge group in four dimensions, which then has to be broken to the standard model by the
Higgs mechanism. The model described in the following differs from those in the choice
of Z6 twists and the pattern of symmetry breaking.
For the G2 SU(3) SO(4) lattice, the action of the Z6 twist is given by Eq. (18) with
1
v6 = (1, 2, 3).
(23)
6
z1 , z2 and z3 are the coordinates of the G2 , SU(3) and SO(4) T 2 -tori, respectively. The Z6
twist v6 has two subtwists,
1
1
Z2 : v2 = 3v6 = (1, 2, 3).
Z3 : v3 = 2v6 = (1, 2, 3),
(24)
3
2
An interesting feature of this orbifold is the occurrence of invariant planes. Clearly, the
Z3 twist leaves the SO(4)-plane invariant whereas the Z2 twist leaves the SU(3)-plane
invariant. The corresponding fixed points and invariant planes are shown in Fig. 4. Our
construction requires two Wilson lines in the SO(4) plane, W2 and W2 , such that there are
3 independent gauge shift vectors (14) acting at different fixed points in this plane.
The rules of orbifold compactifications of the heterotic string have recently been reviewed in [13,14]. We are interested in the states whose masses are small compared to the
string scale MS . These states are described by fields
r,s (x; z1 , z2 , z3 ).
Here r labels the gauge quantum numbers and is given by

p
for the untwisted sector,
r=
p + kV6 for the kth twisted sector,
(25)
(26)
where p lies on the E8 E8 root lattice (16) and we have absorbed the Wilson lines in the
definition of the local twist kV6 . Similarly, s carries information about the spin,

q
for the untwisted sector,
s=
(27)
q + kv6 for the kth twisted sector,
145
Fig. 4. Fixed points and invariant planes (hatched) under the Z6 twist and Z3 , Z2 subtwists, describing localization of different twisted sectors.
where q is an element of the SO(8) weight lattice and v64 = 0. In our convention, the
last component of q gives the 4D helicity. For example, 4D vectors correspond to q =
(0, 0, 0, 1), 4D scalars to q = (1, 0, 0, 0) with all permutations of the first 3 entries, and
fermions correspond to q = ( 12 , 12 , 12 , 12 ) with an even number of + signs.1
The physical states are invariant under the action of the orbifold symmetry group which
consists of twists and translations. In our case only translations in the SO(4) plane have
a non-trivial action on the gauge degrees of freedom, due to the presence of Wilson lines.
Then the invariance conditions read2 ( = 1, . . . , 5):

r,s (x; z1 , z2 , z3 ) = e2i(rV6 sv6 ) r,s x; e2i 6 z1 , e2i 3 z2 , e2i 2 z3 ,
r,s (x; z1 , z2 , z3 ) = e2irW2 r,s (x; z1 , z2 , z3 + 1),
r,s (x; z1 , z2 , z3 ) = e2irW2 r,s (x; z1 , z2 , z3 + i),
(28)
where we have included the Wilson lines in the local shift vectors kV6 . We note that here
two sources of symmetry breaking are present: local, due to twisting, and non-local, due
to the Wilson lines. In the first case, symmetry breaking is restricted to the fixed points
in the compact space. Indeed, since orbifold fixed points are invariant under twisting (up
1 These qs may have to be shifted by an SO(8) root vector to satisfy masslessness conditions in twisted
sectors.
2 Here we omit string oscillator states.
146
to a lattice vector), the first condition can be satisfied only for certain p, which indicates
symmetry breaking at the fixed points. These sets of ps are generally different at different
fixed points and only their intersection survives in 4D, since in this case the wave function
can be constant in the compactified dimensions leading to a massless state. In the case of
Wilson line symmetry breaking, the second and third conditions apply to all points in the
G2 and SU(3) planes and the symmetry breaking is non-local.
To define our string model, it is necessary to specify the action of the twist on the second,
hidden, E8 . We find that the desired symmetry breaking pattern and the appearance of
three 16-plets at fixed points with unbroken SO(10) lead to

1 1
1 1 1
, , , 0, 0, 0, 0, 0
, , 0, 0, 0, 0, 0, 0 ,
V6 =
(29)
3 3 3
6 6

1 1 1
1
1 1 1 1
, , , 0, 0, 0, 0,
,
W2 = 0, , , , , 0, 0, 0
(30)
2 2 2 2
2 2 2
2

1 1 1 1
1 1 1 1 1 3 3 3
, , , , , , ,
0, , , , , 0, 0, 0
W2 =
(31)
4 4 4 4 4 4 4 4
2 2 2 2
in the orthonormal E8 E8 basis. In string theory, these quantities must satisfy certain
consistency conditions (see [13] for a recent discussion). First of all, 6V6 and 2W2 , 2W2
must be elements of the E8 E8 root lattice which is required by embedding of the orbifold symmetry group (space group) in the gauge degrees of freedom. Second, modular
invariance requires

6 (mV6 + nW2 + n W2 )2 m2 v62 = 0 mod 2, m, n, n = 0, 1.
(32)
Our choice of the hidden sector components of V6 , W2 , W2 is strongly affected by these
conditions.
We note that N = 1 supersymmetry in 4D requires
3

v6i = 0 mod 1,
(33)
i=1
whereas N = 2 would require, in addition, v6i = 0 mod 1 for some i. In the former case,
there is one gravitino satisfying q v6 = 0 mod 1 whereas in the latter case there are two of
them.
Finally, massless states in 4D must satisfy the following conditions:
q 2 = 1,
p 2 = 2 2N
(34)
for the untwisted sector, and

(q + kv6 )2 = ck ,
(p + kV6 + nW2 + n W2 )2 = ck,N
(35)
for the kth twisted sector. Here N is an oscillator number and ck , ck,N are certain constants
(see, e.g., [13]). In our model, all states which transform non-trivially under SU(3)c
SU(2)L have N = 0.
In this section we have described the necessary ingredients of our orbifold model. In the
next section we compute the massless spectrum of the model and discuss localization of
various states.
147
Fig. 5. Local gauge symmetries in the SO(4)-plane.
3. Massless spectrum of the model

First let us identify the gauge group in 4D. For N = 1 vector multiplets q v6 = 0.
Hence, the surviving gauge group in 4D is given by the root vectors satisfying
p V6 , p W2 , p W2 Z,
p 2 = 2.
(36)
It is straightforward to verify that these roots together with the Cartan generators form the
Lie algebra of
SU(3) SU(2) U(1)5 ,
while the hidden sector E8 is broken to SU(4) SU(4) U(1)2 . This result can be understood by examining the enhanced gauge groups at the four orbifold fixed points in the
SO(4)-plane. These gauge groups are determined by
p (V6 + nW2 + n W2 ) Z,
(37)
where n, n = {0, 1} specify the fixed point in the SO(4) plane. Then, omitting the hidden
sector the local gauge groups are (Fig. 5)
(n = 0, n = 0): SO(10) SU(3) U(1),
(n = 1, n = 0): SU(6) SU(2) SU(2) U(1),
(n = 0, n = 1): SO(10) SU(2) U(1)2 ,
(n = 1, n = 1): SO(10) SU(2) U(1)2 .
(38)
These are precisely the groups discussed in the first section. Their intersection yields the
surviving group SU(3) SU(2) U(1)5 .
Let us now consider matter fields. These can be either in the untwisted sector U or in
one of the twisted sectors T1 . . . T5 . Below we analyze each of them separately. Before we
proceed, let us fix the chirality of the matter fields to be positive3 , i.e., q4 = +1/2 for their
fermionic components.
3 This is necessary to distinguish matter fields from their CP conjugates.
148
3.1. U sector
For N = 1 chiral multiplets in the untwisted sector we have q v6 = 1/6, 1/3, 1/2,
and therefore
p V6 = {1/6, 1/3, 1/2} mod 1,
p W2 , p W2 Z.
(39)
The states represent bulk matter of the orbifold. By choosing an appropriate right-mover,
these massless states can be made invariant under the Z6 orbifold action and thus are
present in the 4D spectrum. Apart from SU(3)c SU(2)L singlets,4 the untwisted sector
of our model contains
1) 5 (1, 2)
U = 2 (3, 1) (3,
(40)
in terms of the SU(3)c SU(2)L quantum numbers. From the field-theory perspective,
these fields correspond to the compact space components of the E8 gauge fields and their
superpartners.
3.2. T1 + T5 sector
These matter fields are located at the 12 orbifold fixed points (Fig. 4) and satisfy
25
.
(41)
18
Since Wilson lines are present only in the SO(4) plane, only SO(4)-plane projections of the
fixed points matter. The G2 and SU(3) projections do not affect the local twist. They only
lead to a multiplicity factor 3 due to the three identical SU(3) fixed points. Any massless
state in the T1 sector survives the orbifold projection, i.e., is invariant under the Z6 action,
and is therefore present in the 4D spectrum.
The twisted matter fields located at a given fixed point appear in a representation of the
local gauge group at this point. In our case, twisted matter with SU(3)c SU(2)L quantum
numbers is
(p + V6 + nW2 + n W2 )2 =
(n = 0, n = 0): 3 (16, 1),

(n = 1, n = 0): 6 (1, 2, 1),
(n = 0, n = 1): ,
(n = 1, n = 1): .
(42)
It is convenient to keep the notation (16) of SO(10) even though the unbroken group in
4D is only GSM , since it represents one complete generation of SM fermions including
right-handed neutrinos. In terms of SU(3)c SU(2)L quantum numbers we have
T1 + T5 = 3 (16) 6 (1, 2),
where again we have omitted singlets.
4 We defer the analysis of U(1) charges until a subsequent publication.
(43)
149
Fig. 6. Local gauge symmetries in the SU(3)-plane after the Z3 subtwist.
3.3. T2 + T4 sector
These states are localized at the fixed points in the G2 and SU(3) planes, while being
bulk states in the SO(4) plane (Fig. 4). If the T1 sector corresponds to the string with the
3
1
2
boundary condition twisted by = diag(e2iv6 , e2iv6 , e2iv6 ), the T2 sector corresponds
2
2
to the strings twisted by . Since has a fixed plane, T2 states are bulk states in this
plane and localized states in the other two planes.
The orbifold action on this sector is Z3 , and is given by
v3 = 2v6 ,
V3 = 2V6 .
(44)
Since there are no Wilson lines in the G2 and SU(3) planes, all fixed points are equivalent.
The massless N = 1 multiplets obey
14
.
(45)
9
Both the G2 and the SU(3) lattice have 3 fixed points under Z3 , so the multiplicity factor
is 9. The local gauge groups at the fixed points are determined by
(p + V3 )2 =
p V3 = 0.
(46)
At each Z3 fixed point, the unbroken gauge group and the twisted sector matter fields are
(cf. Fig. 6)
E6 SU(3):
(27, 1),
(47)
plus SU(3)c SU(2)L singlets.

These states are subject to further projection and not all of them survive. Indeed, by construction they are only invariant under the Z3 action, but not under the full Z6 . Furthermore,
the Z3 fixed points in the G2 -plane are only fixed under Z3 and the Z6 action transforms
them into one another. Physical states are formed out of their linear combinations which
are eigenstates of the Z6 twist.
The Z3 invariance of a physical state requires
(q + v3 ) v3 = (p + V3 ) V3 mod 1,
(48)
where q + v3 is the shifted SO(8) momentum and p + V3 is the shifted E8 E8 momentum.

This is satisfied automatically as long as the gauge embedding of the twist and the Wilson
150
Fig. 7. Local gauge symmetries in the SO(4)-plane after the Z2 subtwist.
lines obey modular invariance (yet it may require shifts by a lattice vector). A non-trivial
Z2 invariance condition is
(q + v3 ) v2 = (p + V3 ) V2 + mod 1,
p W2 , p W2 Z,
(49)
where
v2 = 3v6 ,
V2 = 3V6 .
(50)
The extra term = (0, 0, 1/2) appears due to the mixing of the fixed points [12,22].
There are three combinations of the Z3 fixed points which are eigenstates of Z6 with eigenvalues e2i .
An important note is in order. The SO(8) lattice momentum q is found via the masslessness condition for the right-movers,
5
(q + v3 )2 = .
(51)
9
Since v3 has a fixed plane, there are always two sets of solutions, with opposite chiralities.
Both of them survive the projection (48), which leads to N = 2 hypermultiplets. The conditions (49) break the symmetry between the two chiralities and one obtains N = 1 chiral
multiplets.
As a result, 9 (27)N = 2 hypermultiplets produce the following N = 1 multiplets
with SU(3)c SU(2)L quantum numbers:
1) 9 (1, 2).
T2 + T4 = 3 (3, 1) 6 (3,
(52)
3.4. T3 sector
These states are localized at the Z2 fixed points in the G2 and SO(4) planes and are bulk
states in the SU(3) plane (Fig. 4). They correspond to strings twisted by 3 . The massless
T3 states satisfy
3
(p + V2 + nW2 + n W2 )2 = ,
2
(53)
151
and the local gauge groups at the fixed points are determined by
p (V2 + nW2 + n W2 ) = 0.
(54)
The result for gauge groups and matter multiplets reads (cf. Fig. 7)
(n = 0, n = 0): SO(16),
(n = 1, n = 0): SO(16),
8 (16),
8 (16),
(n = 0, n = 1): E7 SU(2),
(n = 1, n = 1): E7 SU(2),
(55)
As usual we have omitted SU(3)c SU(2)L singlets and included a multiplicity factor 4
from the G2 -plane fixed points. These states are located at the Z2 fixed points which are
mixed by the action of the full Z6 twist. Again, one has to form linear combinations of the
states transforming covariantly under Z6 .
The matter states are, as before, subject to projection conditions. The Z2 condition for
the relevant states is
(q + v2 ) v2 = (p + V2 + nW2 ) (V2 + nW2 ) mod 1.
(56)
With a proper redefinition of W2 by a lattice vector shift, this condition is satisfied by all
states with both chiralities, i.e., both solutions q of the equation for massless right-movers,
1
(q + v2 )2 = .
(57)
2
Therefore, these states form N = 2 hypermultiplets. The further Z3 projection reads
(q + v2 ) v3 = (p + V2 + nW2 ) V3 + mod 1,
(58)
where now = (0, 0, 1/3, 1/3). The four Z2 fixed points in the G2 -plane lead to four
eigenstates under Z3 with eigenvalues e2i . The above condition projects out some of the
states. The surviving N = 1 multiplets with SU(3)c SU(2)L quantum numbers are
1) 10 (1, 2).
T3 = 7 (3, 1) 5 (3,
(59)
3.5. Summary of the massless spectrum

Combining all matter multiplets from the untwisted and the five twisted sectors we
finally obtain
M = U + T1 + T2 + T3 + T4 + T5
(60)
1) 30 (1, 2),
= 3 (16) 12 (3, 1) 12 (3,
plus SU(3)c SU(2)L singlets. Note that in addition to three SM generations contained
in the three 16-plets we have only vector-like matter. This result is partly dictated by the
requirement of anomaly cancellations. Vector-like fields can attain large masses and decouple from the low energy theory. A detailed analysis of this issue, including U(1) factors,
will be presented in a subsequent publication.
152
Fig. 8. 6D E6 SU(3) orbifold GUT for a large compactification radius of the SO(4)-plane.
4. Intermediate GUTs
So far we have made no assumption on the size of the compact dimensions. These are
usually assumed to be given by the string scale, Ri 1/MS . However, this is not necessarily the case and, furthermore, unification of the gauge couplings favours anisotropic
compactifications where some of the radii are significantly larger than the others [23,24].
In this case one encounters a higher-dimensional GUT at an intermediate energy scale. Indeed, the KaluzaKlein modes associated with a large dimension of radius R become light
and are excited at energy scales above 1/R MS . At these energy scales we obtain an
effective higher-dimensional field theory with enhanced symmetry in the bulk.
In our model there are four independent radii: two are associated with the G2 and SU(3)
planes, respectively, and the other two are associated with the two independent directions
in the SO(4)-plane. Any of these radii can in principle be large leading to a distinct GUT
model.
The bulk gauge group and the amount of supersymmetry are found via a subset of the
invariance conditions (28). Consider a subspace S of the 6D compact space with large
compactification radii. This subspace is left invariant under the action of some elements of
the orbifold space group, i.e. a subset of twists and translations G. The bulk gauge multiplet
in S is a subset of the N = 4 E8 E8 gauge multiplet which is invariant under the action
of G, i.e. a subset of conditions (28) restricted to G.
Consider first the case with two large compact dimensions, for instance those associated with the SO(4)-plane. The SO(4)-plane is invariant under the Z3 subtwist as well
as translations by a lattice vector in the G2 and SU(3) planes. The latter do not lead to
non-trivial projection conditions since there are no Wilson lines in these planes, while the
former leads to gauge symmetry and supersymmetry breaking. The light gauge states are
described by fields which are constant with respect to z1 and z2 . Invariance under Z3 requires (see Eq. (28) with = 2)
p,q (x; z3 ) = e2i(pV3 qv3 ) p,q (x; z3 ).
(61)
Gauge multiplets satisfy q v3 = 0 which has two sets of solutions for q corresponding to
N = 2 supersymmetry. Then the condition p V3 = 0 breaks E8 to E6 SU(3). At the four
fixed points of the SO(4)-plane symmetry is broken further to the four subgroups discussed
153
Fig. 9. 6D SU(4) SU(4) U(1)2 orbifold GUT for a large compactification radius of the SU(3)-plane.
in Section 3. Altogether, we obtain a 6D E6 SU(3) orbifold GUT with the distribution

of gauge symmetries in the fundamental region of the orbifold shown in Fig. 8. Similarly,
untwisted matter satisfies (61) with q v3 = 1/3. We note that all three SM generations
live at the origin in Fig. 8.
A similar analysis can be carried out for the SU(3)-plane, which is invariant under the
Z2 subtwist and lattice translations in the G2 and SO(4) planes. In this case, there are also
non-trivial projection conditions due to the Wilson lines,
p,q (x; z2 ) = e2i(pV2 qv2 ) p,q (x; z2 ),
p,q (x; z2 ) = e2ipW2 p,q (x; z2 ),
p,q (x; z2 ) = e2ipW2 p,q (x; z2 ).
(62)
This breaks N = 4 E8 to N = 2 SU(4) SU(4) U(1)2 in the bulk (Fig. 9). At the fixed
points, the symmetry is broken further by the Z3 twist leaving only the standard model
gauge group (up to U(1)s). Each of the three fixed points carries one generation of the
standard model matter.
A different picture arises when the G2 -plane compactification radius is large. The G2 plane is not invariant under any of the twists, thus there is no projection condition due to
twisting. The only non-trivial projection conditions are due to the Wilson lines,
p,q (x; z1 ) = e2ipW2 p,q (x; z1 ),
p,q (x; z1 ) = e2ipW2 p,q (x; z1 ).
(63)
Thus we have N = 4 supersymmetry and the gauge group is SO(8) SO(8). Three generations of the standard model are localized at the origin where the Z6 twist breaks the
symmetry to the standard model gauge group (Fig. 10).
In principle, there is nothing special about six dimensions, and the same analysis can
be carried out for five, seven, eight, nine and ten dimensions. The results are summarized
in Table 1. A variety of orbifold GUTs appears, with gauge groups ranging from E8 to
SU(4) SU(4) U(1)2 . These GUTs represent different points in moduli space. Values
of the corresponding T-moduli determine the compactification radii.
154
Fig. 10. 6D SO(8) SO(8) orbifold GUT for a large G2 -plane compactification radius.
It is remarkable that all these GUT models in various dimensions are consistent with
gauge coupling unification.5 This is true even though in some cases SU(3)c and SU(2)L
are contained in different simple factors, i.e. SO(8) SO(8) or SU(4) SU(4). The beta
functions for both SO(8)s or SU(4)s are the same. In the former case this is enforced by
N = 4 supersymmetry, while in the latter case the two SU(4)s have identical bulk matter
content, 2 (6, 1) + 16 (4, 1) N = 2 multiplets. In all other cases SU(3)c SU(2)L
is contained in a simple factor such that unification of the gauge couplings in the bulk is
automatic.
On the other hand, different GUTs differ in the value of the gauge coupling at the unification scale, since the power law running depends on the number of extra dimensions
and the bulk gauge group. Realization of some of the GUTs may require non-perturbative
string coupling [23,24]. Different models also lead to different Yukawa couplings which
depend on the compactification radii. These phenomenological aspects are similar to those
of orbifold GUTs [25] and will be discussed elsewhere.
5. Summary
We have presented a Z6 heterotic orbifold model leading to the standard model spectrum
and additional vector-like matter in four dimensions. Standard model generations appear
as 16-plets of SO(10). They are localized at different fixed points in the compact space
with local SO(10) SU(3) U(1) symmetry.
If some of the compactification radii are significantly larger than the others, we recover
various higher-dimensional GUTs as an intermediate step at energies below MS . These
GUTs have the same 4D massless spectrum and the same ultraviolet completion, but represent different points in moduli space. All of them are consistent with gauge coupling
unification, yet differ in other phenomenological aspects.
5 Here we only consider running of the gauge couplings in the bulk. An analysis of localized contributions
will be presented elsewhere.
155
Table 1
Survey of the various orbifold GUTs in different dimensions. The bullet indicates small compact dimensions.
U(1) factors are omitted
Dim.
Plane
Conditions
SUSY, bulk groups
N = 4, E8
p W2 Z
N = 4, SO(16)
p W2 Z
N = 4, SO(16)
N = 4, E8
N = 4, E8
p W2 , p W2 Z
N = 4, SO(8) SO(8)
p W2 Z
N = 4, SO(16)
p W2 Z
N = 4, SO(16)
p W2 Z
N = 4, SO(16)
p W2 Z
N = 4, SO(16)
p 2V6 Z
N = 2, E6 SU(3)
p 3V6 , p W2 , p W2 Z
N = 2, SU(4) SU(4)
p W2 , p W2 Z
N = 4, SO(8) SO(8)
p 2V6 , p W2 Z
N = 2, SU(6) SU(2)2
p 2V6 , p W2 Z
N = 2, SU(6) SU(2)2
p V6 , p W2 , p W2 Z
N = 1, SU(3) SU(2)
GSM
G2
10
SU(3)
SO(3)
Acknowledgements
We would like to thank S. Frste, A. Hebecker, T. Kobayashi, H.-P. Nilles, M. Trapletti,
P.K.S. Vaudrevange and A. Wingerter for discussions. One of us (M.R.) would like to
156
thank the Aspen Center for Physics for support. This work was partially supported by the
EU 6th Framework Program MRTN-CT-2004-503369 Quest for Unification and MRTNCT-2004-005104 Forces Universe.
References
[1] H. Georgi, S.L. Glashow, Phys. Rev. Lett. 32 (1974) 438.
[2] H. Georgi, in: C.E. Carlson (Ed.), Particles and Fields 1974, American Institute of Physics, New York, 1975,
p. 575;
H. Fritzsch, P. Minkowski, Ann. Phys. 93 (1975) 193.
[3] J.C. Pati, A. Salam, Phys. Rev. D 10 (1974) 275.
[4] S.M. Barr, Phys. Lett. B 112 (1982) 219.
[5] D.I. Olive, in: J. Ellis, S. Ferrara (Eds.), Unification of the Fundamental Particle Interactions, vol. II, Erice,
1981, Plenum, New York, 1983.
[6] D.J. Gross, J.A. Harvey, E.J. Martinec, R. Rohm, Phys. Rev. Lett. 54 (1985) 502;
D.J. Gross, J.A. Harvey, E.J. Martinec, R. Rohm, Nucl. Phys. B 256 (1985) 253.
[7] P. Candelas, G.T. Horowitz, A. Strominger, E. Witten, Nucl. Phys. B 258 (1985) 46.
[8] L.J. Dixon, J.A. Harvey, C. Vafa, E. Witten, Nucl. Phys. B 261 (1985) 678;
L.J. Dixon, J.A. Harvey, C. Vafa, E. Witten, Nucl. Phys. B 274 (1986) 285.
[9] L.E. Ibez, H.P. Nilles, F. Quevedo, Phys. Lett. B 187 (1987) 25;
L.E. Ibez, J.E. Kim, H.P. Nilles, F. Quevedo, Phys. Lett. B 191 (1987) 282.
[10] Y. Kawamura, Prog. Theor. Phys. 103 (2000) 613;
Y. Kawamura, Prog. Theor. Phys. 105 (2001) 999;
G. Altarelli, F. Feruglio, Phys. Lett. B 511 (2001) 257;
L.J. Hall, Y. Nomura, Phys. Rev. D 64 (2001) 055003;
A. Hebecker, J. March-Russell, Nucl. Phys. B 613 (2001) 3.
[11] T. Asaka, W. Buchmller, L. Covi, Phys. Lett. B 523 (2001) 199;
L.J. Hall, Y. Nomura, T. Okui, D.R. Smith, Phys. Rev. D 65 (2002) 035008.
[12] T. Kobayashi, S. Raby, R.-J. Zhang, Phys. Lett. B 593 (2004) 262.
[13] S. Frste, H.P. Nilles, P. Vaudrevange, A. Wingerter, Phys. Rev. D 70 (2004) 106008.
[14] T. Kobayashi, S. Raby, R.-J. Zhang, hep-ph/0409098.
[15] J.D. Breit, B.A. Ovrut, G.C. Segre, Phys. Lett. B 158 (1985) 33.
[16] Y. Katsuki, Y. Kawamura, T. Kobayashi, N. Ohtsubo, Y. Ono, K. Tanioka, Nucl. Phys. B 341 (1990) 611.
[17] K.S. Choi, K. Hwang, J.E. Kim, Nucl. Phys. B 662 (2003) 476.
[18] A. Hebecker, J. March-Russell, Nucl. Phys. B 625 (2002) 128.
[19] A. Hebecker, M. Ratz, Nucl. Phys. B 670 (2003) 3.
[20] L.E. Ibez, J. Mas, H.P. Nilles, F. Quevedo, Nucl. Phys. B 301 (1988) 157.
[21] D. Bailin, A. Love, Phys. Rep. 315 (1999) 285.
[22] T. Kobayashi, N. Ohtsubo, Phys. Lett. B 245 (1990) 441.
[23] E. Witten, Nucl. Phys. B 471 (1996) 135.
[24] A. Hebecker, M. Trapletti, hep-th/0411131.
[25] T. Asaka, W. Buchmller, L. Covi, Phys. Lett. B 563 (2003) 209.
A three-loop test of the dilatation operator

in N = 4 SYM
B. Eden a,b , C. Jarczak c , E. Sokatchev c
a Max-Planck-Institut fr Gravitationsphysik, Albert-Einstein-Institut,
Am Mhlenberg 1, D-14476 Golm, Germany

b Dipartimento di Fisica, Universit di Roma Tor Vergata, I.N.F.N., Sezione di Roma Tor Vergata,
Via della Ricerca Scientifica, 00133 Roma, Italy

c Laboratoire dAnnecy-le-Vieux de Physique Thorique LAPTH,
B.P. 110, F-74941 Annecy-le-Vieux, France 1
Abstract
We compute the three-loop anomalous dimension of the BMN operators with charges J = 0 (the
Konishi multiplet) and J = 1 in the N = 4 super-YangMills theory. We employ a method which
effectively reduces the calculation to two loops. Instead of using the superconformal primary states,
we consider the ratio of the two-point functions of suitable descendants of the corresponding multiplets. Our results unambiguously select the form of the N = 4 SYM dilatation operator which is
compatible with BMN scaling. Thus, we provide evidence for BMN scaling at three loops.
PACS: 11.15.-q; 11.25.Hf; 11.30.Pb; 11.40.-q
1. Introduction
During the last couple of years, following the important work of BMN [1], spectacular
progress has been made in the understanding of the integrable structure underlying N = 4
E-mail address: beden@aei.mpg.de (B. Eden).
1 UMR 5108 associe lUniversit de Savoie.
doi:10.1016/j.nuclphysb.2005.01.036
158
B. Eden et al. / Nuclear Physics B 712 (2005) 157195
supersymmetric YangMills (SYM) theory [24], on the one hand, and its correspondence
to string theory [57], on the other. An important part of this program is the development
of efficient tools for calculating anomalous dimensions of gauge-invariant composite operators and then comparing to string theory predictions.
One such tool is the so-called dilatation operator of N = 4 SYM [3,8]. In this approach, without doing complicated higher-loop perturbative calculations, one is able to
predict the anomalous dimensions of large classes of operators. The form of the dilatation operator can be determined by combining various symmetry arguments with some
field theory input. In [3] the notion of one-loop integrability was extended to higher loops.
Based on this and on the additional assumption of BMN scaling, the planar dilatation operator of the SU(2) sector of N = 4 SYM was elaborated up to three loops. Later on, a
comprehensive treatment of the dilatation operator for the SU(2|3) sector was given in [9].
It was shown that the two-loop dilatation operator is determined up to one free constant,
which can be fixed from the known value of the two-loop anomalous dimension of the
Konishi operator. At three loops, after taking into account the known quantum symmetries,
a two-parameter freedom remains. The higher-loop integrability conjectured in [3] was
shown to follow from conformal supersymmetry and the dynamics of the theory, keeping
the above-mentioned parameters arbitrary.
In [9] Beisert used the extra assumption of BMN scaling to fix the remaining freedom.
Alternatively, experimental data in the form of the three-loop anomalous dimensions of
any two BMN operators, obtained by a direct perturbative calculation, would unambiguously determine the three-loop dilatation operator. This would also provide an indirect
check on the hypothesis of BMN scaling at this level of perturbation theory.2
Thus, the necessity of such a three-loop perturbative experiment appears quite clear.
Rather recently, a whole series of three-loop anomalous dimensions of operators of twist
two were computed in [13,14], using sophisticated QCD methods. The simplest among
these operators is the Konishi scalar, which is also the lowest state in the BMN series. The
value of its anomalous dimension obtained in [14] is in perfect agreement with the prediction of [3,9] under the assumption of BMN scaling. However, it is not clear if the method
of [13,14] generalizes to operators of higher twist, such as the scalar BMN operators with
non-vanishing charge.3 So, obtaining the second piece of experimental data needed for
the above test is still an open problem.
In this paper we carry out a three-loop calculation of the anomalous dimension of the
first two operators in the BMN series (with two impurities), the Konishi one (i.e., twist two
or J = 0) and the twist three (or J = 1) one. In the Konishi case we reproduce the result
of [14] (unlike Ref. [14], we use standard field theory techniques). The result for the J = 1
2 At two loops BMN scaling has already been confirmed in [10], but it seems rather difficult to extend this
result to three loops. A general argument to all orders was proposed in [11], but it is based on a number of
assumptions which, although plausible, seem hard to justify. Also, an indication of BMN scaling at three loops
through the matrix model has appeared in [12].
3 Here we mean the twist and the charge of the primary state of a superconformal multiplet. Thus, the Konishi
multiplet starts with a scalar of dimension two and charge zero, so it has twist two and J = 0; the scalar BMN
operators with two impurities of charge J have dimension and twist J + 2, etc. This differs from the classification
of [1], where the dimension and the charge are shifted by two units by considering a certain descendant of the
multiplet.
159
case is new. Most remarkably, the two values are exactly those predicted by the dilatation
operator whose form is compatible with BMN scaling!
Direct three-loop calculations of anomalous dimensions are prohibitively complicated.
Here we exploit a method based on an idea of Anselmi [15] and further refined and applied
at two loops in [16]. It allows us to effectively reduce the perturbative order to two loops.4
In this section we give a brief description of the basic ideas.
In conformal field theory operators are labeled by their spin and dilatation weight. The
two-point function of a scalar operator has the functional form

O(x)O (0) =

with
c(g 2 )
(x 2 )(g
2)
(1)

g 2 = 0 + g 2 .
(g 2 ) =
(2)
g 2n
is the
Here 0 is the canonical (or tree-level) scaling dimension and
n=1 n
anomalous part of the dilatation weight. The latter is due to logarithmic divergences of
the two-point function which, after proper renormalization of the quantum operator O, are
supposed to sum up into the power behavior indicated in Eq. (1).
The idea of Anselmis trick is most easily understood on the example of a current
k (x) possessing anomalous dimension = 3 + (g 2 ). Conformal invariance fixes its
two-point function up to normalization:

x 2 2x x
.
k (x)k (0) = c g 2
2
(x 2 )4+ (g )
In the free field theory, where g = 0 and (g 2 ) = 0, this current is conserved, so

k (x) k (0) g=0 = 0,
(3)
(4)
up to contact terms which we do not consider here. In the presence of (quantum) interaction
(g 2 ) = 0 and we find

( + 2)
k (x) k (0) g=0 = 4c g 2
.
2
(x 2 )4+ (g )
(5)
Now, consider the ratio

k (x) k (0)
k (x)k (0)
=2
( + 2)
.
x2
(6)
Notice that the right-hand side of this equation is O(g 2 ), since (g 2 ) = 1 g 2 + O(g 4 ),
and so must be the left-hand side. This is indeed true, given the fact that k = O(g) as
a consequence of the field equations (i.e., conservation in the absence of interaction). In
4 Loop counting for composite operators is not the same as counting powers of the coupling g. Thus, the
tree-level two-point function of an n-linear composite operators contains n 1 loops, but no interactions, so it
is O(g 0 ). In this article, as is customary in the AdS/CFT literature, n loops means O(g 2n ). However, the actual
loop integrals to be evaluated may be of higher order, see Section 3.3.4.
160
practical terms this means that we have gained an order in g 2 : if we wished to compute, e.g.,
1 starting from Eq. (3), we would have to evaluate one-loop divergent integrals involved
in the two-point function of the current itself. However, using the ratio (6) instead, it is
sufficient to compute the tree-level two-point functions of the current and of its divergence.
It is precisely this trick that Anselmi [15] applied to the current in the Konishi multiplet in
order to obtain its one-loop anomalous dimension, without doing any loop integrals!
The generalization to higher orders in perturbation theory is complicated by the quantum anomaly phenomenon. By this we mean that the non-conservation of the current may
not be correctly explained by the classical field equations alone, but it might be modified
by an extra, purely quantum term. This is exactly what happens to the Konishi current, as
we explain below.
In the N = 4 SYM theory the Konishi current and its divergence are members of one
long (or unprotected) superconformal multiplet. The lowest component (or superconformal primary state) of the Konishi superfield K1 (x, , ) is an SU(4) singlet scalar of dimension 2. This superfield contains 15 + 1 Konishi currents, K1 = + A B (k )A B (x) +
, where A, B = 1, 2, 3, 4 are SU(4) indices. The divergence in (5) is contained in
the double spinor derivative of the superfield D A D B K1 (here D A D B = D B D A =
). Taking such derivatives and subsequently using the superfield equations
(1/4)D A
D
B
of motion is equivalent to making on-shell supersymmetry transformations. In the free
theory D A D B K1 = 0, which implies conservation of the type (4), (k )A B = 0. In
the interacting theory this equation is modified, D A D B K1 = g(K10 )AB . Its right-hand
side gives rise to a new scalar superfield of dimension 3 in the 10 of SU(4), which contains the divergence k (x). This new scalar is a superconformal descendant of the
Konishi multiplet [17,18]. Notice the presence of the gauge coupling g in the definition of K10 , indicating that this descendant does not occur in the free theory. Now, in
the N = 4 theory we can go a step further, applying two more spinor derivatives (i.e.,
on-shell supersymmetry transformations) to produce a scalar of dimension 4 in the 84,
g 2 (K84 )CD AB = D C D D D A D B K1 (traces). This time we have a factor of g 2 in the
definition of K84 , showing that we have used the field equations twice.
The supersymmetric version of Eq. (6) is the ratio of two-point functions

K10 K10
K1 K1
(7)
In principle, it contains information about the anomalous dimension common to all members of the Konishi multiplet (which include K1 and K10 ). However, the procedure is not
safe beyond the lowest perturbative level. The reason is that the naive definition of the
descendant K10 obtained by applying the classical field equations (or on-shell supersymmetry) to the quantum (renormalized) operator K1 is not correct. What this naive
procedure gives is the classical anomaly K10 = B, where B is the superpotential of the
N = 4 theory. The procedure misses, however, another term gF , where F is a composite
operator with the same quantum numbers (dimension and SU(4) representation), but this
time made out of fermions. This is the so-called quantum Konishi anomaly [19]. So, the
correct form of the descendant (schematically) is K10 = B + gF . Even knowing this, the
correct evaluation of the ratio (7) is still not obvious. The problem is to properly fix the
161
normalization of the quantum corrected descendant K10 . Recall that when going from (3),
to (5), we tacitly assumed that the normalization c(g 2 ) was the same; this is far from obvious when the divergence of the current is affected by a quantum anomaly. So, it might
seem that the Anselmi trick is of little use beyond one loop.
The way out was proposed and successfully used at order O(g 4 ) in [16]. The idea is to
evaluate another ratio,

K84 K84
K10 K10
(8)
Why is this better? We can argue that the second step in the chain of descendants above,
K10 K84 , is anomaly-free. We identify the origin of the Konishi anomaly with the
operator mixing taking place at the level of the descendant K10 . As we just mentioned, at
this level there are two possible composite operators with the same quantum numbers, B
and F , which can and do mix between themselves. The correct definition of K10 is achieved
by resolving this mixing problem (see Section 2 and [20] for a detailed discussion). On the
other hand, at the level of the descendant K84 there is only one candidate, a single-trace
operator of dimension 4 (see again Section 2).5 So, we can conclude that the absence of
mixing at the level of K84 implies the absence of the anomaly. In this case, we assume that
the step K10 K84 can be done naively, using the classical superfield equations of motion
or on-shell supersymmetry. In particular, this means that K84 inherits its normalization
from its ancestor K10 , so that it drops in the ratio (8).
Apart from the central issue of anomalies, considering the ratio (8) rather than (7) has
another, technical but very important advantage. We plan to do our Feynman graph calculations in the N = 1 superspace formalism. There K1 is a real (non-chiral) superfield,
whereas for both K10 and K84 we can find an SU(4) projection which is chiral.6 This significantly reduces the number of Feynman graphs involved. Further, evaluating the ratio
instead of the two-point functions themselves results in cancellation of many Feynman integrals. Finally, the set of graphs for the two-point function of the chiral projection has
a large overlap with those for a certain half-BPS operator. The graphs for the latter are
known to sum up to zero (protectedness), which allows us to eliminate large sets of
graphs without calculating them (see also [23,24]).
This method applies not only to the Konishi multiplet, but also to any multiplet which
lies on the unitarity bound of the continuous series of N = 4 superconformal representations [25,26]. Their common property is that they satisfy a generalized conservation
condition D 2 O = 0 in the free case. In the presence of interaction some of these multiplets still satisfy this condition, and so they have protected conformal weight (these are
5 To be more precise, there is also a double-trace version of it which could possibly mix. However, one can
show (see [21] for a one-loop and [22] and Appendix A.2 to the present paper for a two-loop proof) that this
mixing only affects another operator, a member of a protected quarter-BPS multiplet; the Konishi descendant
K84 stays pure.
6 One should not confuse chiral operators with chiral primary operators (CPO). The latter are known to be
protected from quantum correction. The chiral projections of the Konishi descendants we are considering are not
protected, precisely because they are not superconformal primary. For this reason we prefer the more adequate
term BPS operators instead of the popular term CPO.
162
the so-called semishort multiplets [18,27]). However, the generic multiplet of this type
is not conserved in the interacting theory, so it is not protected and acquires an anomalous
dimension. It is precisely this non-conservation that we wish to exploit according to the
scenario described above. In principle, the entire series of BMN multiplets falls into this
class,7 but in practice the quantum calculation are further complicated by operator mixing
already at the level of the superconformal primaries. Only the first two members of the
BMN family, the J = 0 (or twist two) multiplet (which is the Konishi multiplet) and the
J = 1 (or twist three), are realized in terms of pure composite operators (without mixing).
This is why we limit the scope of our three-loop calculations to these two cases. As explained earlier, our results are sufficient to unambiguously confirm the form of the N = 4
dilatation operator proposed in [3,9].
The paper is organized as follows. In Section 2 we recall the general properties of the
Konishi multiplet and its descendants. We explain in detail how superspace differentiation
reduces the loop order. We also give the realization of the various objects as gauge invariant
composite operators both in N = 4 and in N = 1 superspace. We proceed to determine
2 and 3 from a graph calculation: in Section 3 this is done for the Konishi case. In
Section 4 we adapt the calculation, with relatively minor changes, to the case of the BMN
multiplet with J = 1. In the conclusions we summarize the results obtained and discuss
their implications. Appendix A contains our conventions and superspace Feynman rules,
discusses the absence of mixing for the level two descendants, and comments on the use of
the MS scheme.
2. Reducing the loop order by superspace differentiation

2.1. The Konishi operator and its descendants
The Konishi operator is the simplest example of a composite gauge invariant operator
in the N = 4 super-YangMills theory. The basic object in this theory is the super-field ) = W
strength WAB (x , A , A
BA where A, B = 1, 2, 3, 4 are indices of the fundamental representation of the R symmetry group SU(4). The lowest component of this superfield
describes six real scalars due to the reality condition
1
(WAB ) = W AB =
ABCD WCD .
2
(9)
In addition, WAB is subject to the on-shell constraints

{C WA}B = 0,
(C
WA)B = 0,
(10)
7 We emphasize the point that only the Konishi multiplet contains an anomalous vector current. The highertwist BMN multiplets do not contain such currents, but nevertheless satisfy an anomalous semishortness
condition in superspace. Thus, focusing on the analogy between the anomaly of the Konishi current and the
standard axial anomaly is somewhat misleading, in our opinion.
163
where {C A} denotes the traceless part and (CA) means symmetrization with unit weight.
The gauge-covariant superspace derivatives , satisfy the algebra
A B
, = 4g
W AB ,
A

, B
(11)
= 2iBA
(we have explicitly written down the gauge coupling g). A useful corollary of Eqs. (10)
and (11) is

(A B) W CD = g W (AC , W B)D ,
(12)
where A B = (1/4) A B .
The (on-shell) Konishi operator is defined by

1
K1 = Tr W AB WAB .
(13)
4
Its lowest component (at = = 0) is a scalar of naive dimension 2. This scalar is an
SU(4) singlet, as indicated by the subscript 1. The N = 4 supersymmetric generalization
of the two-point function (1) for K1 is (see [28] for the N = 1 case)

c(g 2 )
K1 (x1 , 1 , 1 )K1 (x2 , 2 , 2 ) =
(g 2 )/2
2
2
xL1R2
xL2R1
(14)
with the chiral/antichiral supersymmetric invariant coordinate
xL1R2 = x1 x2 + i1A 1A + i2A 2A 2i1A 2A .

(15)
Clearly, (14) reduces to (1) when , = 0.
Applying spinor derivatives to the superfield K1 , we can produce all of its components
(or descendants). In this paper we exploit the following two descendants [17,18]:
g(K10 )AB = D A D B K1
,=0
(16)
,
2
CD
{C
{D
g (K84 ) AB = D D DA} D B} K1
,=0
(17)
.
In Eq. (17) we are taking the traceless part only, so the order of the derivatives D and D
does not matter. It is clear that the descendants (16) and (17) are scalars of naive dimension
3 and 4, in the 10 (Dynkin labels [200]) and 84 ([202]) of SU(4), respectively. Notice that
we write down the gauge coupling g explicitly, in order to remind ourselves that these
descendants do not occur (or rather decouple) in the free theory.
Further, for reasons which will become clear below, we find it very convenient to work
with the highest weight projections of the above SU(4) representations:8
gK10 g(K10 )11 = D 1 D 1 K1

,=0
,
8 The SU(4) representations can be read off the highest weight by applying the following rule: count the
occurrences of the lower indices 1, 2, 3 and fill in the boxes of a Young tableau (the first row with 1, etc). When a
lower index 4 occurs, it can be removed by simultaneously dropping a full set 1234; an upper index 4 is replaced
by a lower set 123. Thus, the Young tableau corresponding to K10 in (18) has labels (200) which translates into
Dynkin labels [200], i.e., the 10. For K84 we have 44
11 1111 22 33, which is the Young tableau (422) or Dynkin
labels [202], i.e., the 84.
164
g 2 K84 g(K84 )44 11 = D 4 D 4 D 1 D 1 K1

,=0
.
(18)
Their two-point functions can be found from (14) by applying the same spinor derivatives
as in the definition of the operators:

g 2 K10 (x1 )K10

(x2 ) = D 1 D 1
1 D 1 D 1
2 K1 (1)K1 (2)
,=0
,

(x2 ) = D 4 D 4 D 1 D 1
1 D 4 D 4 D 1 D 1
2 K1 (1)K1 (2)
,=0
g 4 K84 (x1 )K84
. (19)
Due to chirality, D 1 D 1 |1 D 1 D 1 |2 only acts on xL2R1 . The superspace differentiation
results into on the corresponding factor in the two-point function. The second step is
analogous: D 4 D 4 |1 D 4 D 4 |2 only sees xL1R2 , producing another . Hence

g 2 K10 (1)K10
(2) = ( 2)

g 4 K84 (1)K84
(2) = 2 ( 2)2
c(g 2 )
2 )((g 2 )+1)
(x12
c(g 2 )
2 )((g
(x12
2 )+2)
,
.
(20)
Let us consider the ratios of the two-point functions in (14) (at = = 0) and in (20):

K10 K10
K1 K1

K84 K84
K10 K10
( 2)
,
2
g 2 x12
(21)
( 2)
.
2
g 2 x12
(22)
Apart from the trivial spacetime factor, the right-hand sides contain information about the
anomalous dimension,

( 2)
= 21 + 12 + 22 g 2 + 2(1 2 + 3 )g 4 + O g 6 .
g2
(23)
We clearly see the loop reduction effect of the superspace differentiation: it is sufficient
to compute the left-hand side of either Eq. (21) or Eq. (22) to order O(g 2(n1) ) (n 1
loop) in order to determine the n-loop anomalous dimension n via a simple algebraic
equation.
2.2. The descendants as composite operators
In the free field theory (g = 0) the descendants of K1 considered above do not occur.
Instead, the free Konishi superfield satisfies the conservation (or semishortness) condition
D A D B K1
g=0 = D A D B K1
g=0 = 0,
(24)
as easily follows from the on-shell constraints (10), (11). In particular, this implies that the
15 + 1 vector components of dimension 3 in K1 are conserved vectors (the so-called Konishi currents). However, this conservation is destroyed when the interaction is switched
on. Indeed, applying once again the on-shell constraints (10), (11), but this time with g = 0,
165
after some simple algebra we find

g
D A D B K1
g=0 = Tr [WAC , WBD ]W CD 3gBAB .
(25)
2
This is the so-called classical Konishi anomaly reflecting the fact that the conservation
(24) does not correspond to any symmetry of the interacting theory.
In the quantum theory Eq. (25) is further modified by the fermionic term
FAB =

1 C
Tr WCA D WDB ,
9
(26)
so that

D A D B K1
quantum = 3g(BAB + gf10 FAB ) + O g 3 g(K10 )AB .
(27)
N
The coefficient f10 = 32
2 of the F term (the so-called quantum Konishi anomaly)
has been calculated at one loop in [19], but the operator relation (27) receives higherorder corrections indicated by the O(g 3 ) terms.9 So, the correct definition of the quantum
operatordescendant of the Konishi operator is
(K10 )AB = 3(ZB BAB + ZF FAB ),
(28)
where ZB , ZF are renormalization factors specified in Section 3.1 (see also [20] for a
more detailed discussion). We stress that Eq. (16) simply defines a component of the
superfield K1 . The difference between a true descendant and a superfield component is
that the former is obtained through the (quantum corrected) field equations and the latter
through straightforward superspace differentiation.
The next level descendant K84 is obtained by further differentiation (see (17)):
g(K84 )CD AB = D {C D {D (K10 )A}B}
(29)
and use of the on-shell constraints. This is most easily done in terms of the highest weight
projections as defined in (18). The corresponding projections of the bosonic and fermionic
composite operators B and F are

B B11 = Tr [W12 , W13 ]W14 ,

1
F F11 = Tr C WC1 D WD1 .
(30)
9
Applying the derivatives D 4 D 4 and using the constraints, we find
D 4 D 4 B = gY,
D 4 D 4 F = 4g 2 Y,
(31)
where

Y = Tr [W12 , W13 ][W12 , W13 ]
(32)
9 In the literature there are claims that the Konishi anomaly is one-loop exact, like the AdlerBellJackiw
anomaly. We believe that this statement is strongly context- and interpretation-dependent. In [20] this issue is
discussed in the context of operator mixing and it is shown that the anomaly receives corrections.
166
is a scalar operator of dimension 4 corresponding to the highest weight of the 84. We see
that both ingredients B and F of K10 give rise to the same operator Y , so we are led to the
conclusion
gK84 = D 4 D 4 K10 = g(ZB + 4gZF )Y gZ84 Y.
(33)
Here we ought to make an important comment. We obtained the quantum operator K84
from the quantum operator K10 through the classical field equations. However, at the
preceding step K1 K10 we saw that the same procedure only produced the B term (25),
it completely missed the F term in (27). Can we be sure that something similar does not
happen at the step K10 K84 ? Our argument is as follows: the reason why the quantum
K10 is different from the classical one is operator mixing. Indeed, there exist two operators
with the same quantum numbers (scalars of dimension 3 in the 84), our B and F . In the
quantum theory they can mix, therefore the correct form of K10 should be established by
resolving this mixing problem (see Section 3 and [20]). At first sight, at the level of K84
we seem to be in a similar situation. Besides the single-trace operator Y (32), there is also
a double-trace version of it,
Y = Tr(W12 W12 ) Tr(W13 W13 ) Tr(W12 W13 ) Tr(W12 W13 )
(34)
(the antisymmetrization is needed to achieve the right Young tableau structure of the 84).
The mixing of Y and Y has been studied at one loop in [21,29,30] and the conclusion is
that the Konishi descendant K84 is indeed identified with Y , while a particular mixture of
Y and Y gives rise to a different, protected (quarter-BPS) operator which is orthogonal to
K84 . For our purposes here we need to make sure that the same picture persists at the next,
O(g 4 ) level, and this is done in Appendix A.2 (see also [22]). We view this as a sufficient
reason to assume that the step K10 K84 can be done naively, so that (33) is the exact
form of the operator K84 .
2.3. The stress-tensor multiplet
We remark that the two equations (31) imply the existence of a particular combination
of B and F which has a vanishing descendant in the 84:
O10 = 3(F 4gB) D 4 D 4 O10 = 0.
(35)
In fact, the operator O10 is itself a descendant of the protected (half-BPS) operator O20 ,
the so-called stress-tensor multiplet of N = 4 SYM:
(O20 )ABCD = Tr(WAB WCD + WAD WCB ).
(36)
Note that this combination has vanishing traces of the type (O20 )ABCD
BCDE = 0, so
it indeed corresponds to the 20 ([020]) in the decomposition 6 6 = 20 + 15 + 1. The
important property of the operator (36) is that it satisfies the half-BPS constraints
D{E (O20 )A}BCD = D{E (O20 )AB}CD = 0,
(E
D (E
(O20 )A)BCD = D
(O20 )AB)CD = 0,
(37)
167
which follow from the on-shell constraints on W . This is true even in the presence of
interactions, and as a consequence, the conformal dimension of O20 is protected from
quantum corrections. Moreover, it has been shown [3133] that the two-point function of
this operator is not renormalized (up to contact terms),

O20 O20
(38)
= O20 O20 g=0 .
The descendant O10 is obtained in the following way:
(O10 )AB = D C D D (O20 )ACBD = 3(FAB 4gBAB ),
(39)
and Eq. (35) corresponds to the highest weight projection (O10 )11 . Another consequence
of the half-BPS conditions (37) is that this multiplet is short, i.e., it does not have the full
span of descendants for a generic N = 4 multiplet. This explains why the step indicated in
(35) does not produce a new descendant.
The two-point function of the descendant O10 , like the one of the primary (38), remains
unchanged by quantum corrections,

= O10 O10
.
O10 O10
(40)
g=0
For our graph calculations in Section 3 it will be essential that not only the two-point
function of O10 , but also its explicit form (39) as a mixture of the bare operators F and B
is not renormalized. This is not obvious, since it is not easy to rule out the possibility of
finite corrections in an operator mixture, even if it has no anomalous dimension. In [20] a
perturbative test of this conjecture has been carried out.
The two operators O10 and K10 belong to different superconformal multiplets, the former to a short one with protected (canonical) dimension, the latter to a long one with
anomalous dimension. Therefore, they must be orthogonal to each other,

= 0.
O10 K10
(41)
The two properties (40) and (41) are all that we need in order to resolve the mixing of
the operators B and F . We just have to prepare the right mixtures which satisfy these two
conditions (see Section 3 and [20]).
2.4. The BMN operators
The Konishi multiplet is the first member of the family of BMN operators (here we
consider the case of two impurities only). Let us give a brief description of it.
The stress-tensor multiplet can be generalized to a family of half-BPS operators [34] of
dimension J in the representation [0J 0] of SU(4). Their highest weights can be written in
the symbolic form
O[0J 0] = (W12 )J
(42)
(the color trace (single or multiple) is understood). They satisfy half-BPS shortening conditions similar to (37). After projecting the SU(4) indices these conditions read (for details
see [26])
D 4 O[0J 0] = D 3 O[0J 0] = D 1 O[0J 0] = D 2 O[0J 0] = 0.
(43)
168
Their meaning is that the superfield O[0J 0] depends on half of the Grassmann variables
(hence the name half-BPS; such superfields are also called Grassmann analytic),

O[0J 0] a , a , a = 1, 2, a = 3, 4.
(44)
The BMN family of operators with two impurities is obtained by gluing together
O[0J 0] with the Konishi operator K1 , thus constructing a new scalar operator of dimension
J + 2 in the [0J 0]. Symbolically,10
K[0J 0] = O[0J 0] K1 ,
(45)
where means merging the two color traces. Now, combining the free theory semishortness conditions (24) with the BPS conditions (43), we see that free BMN operators satisfy
the intersection of the two, i.e., the semishortness conditions (only the highest weight is
shown)
D 4 D 4 K[0J 0] = D 1 D 1 K[0J 0] = 0,
(46)
which corresponds to a superconformal representation at the unitarity bound of the continuous series. In the interacting case, just like for Konishi, these conditions give rise to
descendants (recall (18))
gK[2J 0] = D 1 D 1 K[0J 0]
,=0
,
g 2 K[2J 2] = D 4 D 4 D 1 D 1 K[0J 0]
,=0
(47)
.
The two-point functions of the operators K[0J 0] are determined by N = 4 conformal
supersymmetry as follows. First, consider the two-point function of the half-BPS operator
(42):
O[0J 0] (1)O[0J
0] (2) =
C
(x122 )J
(48)
Here C is a normalization constant (or a constant SU(4) tensor, if not restricted to the highest weight only). The supersymmetric invariant difference (cf. (15) in the chiral/antichiral
case)

x12 = x1 x2 + i 1a 1a 1a 1a + 2a 2a 2a 2a 21a 2a + 22a 1a
(49)
is the unique combination compatible with the BPS conditions (43). What remains to do is
to multiply the structure (14) for K1 with that for O[0J 0] (48), according to (45):
K[0J 0] (1)K[0J
0] (2) =
c(g 2 )
2
2
(xL1R2
xL2R1
)(g
2 )/2
(x122 )J
(50)
where = 2 + (g 2 ). Thus, the Konishi factor carries the anomalous dimension, while
the half-BPS factor is there to adjust the canonical dimension 0 = 2 + J .
10 The full picture is more complicated, since the two constituents of the Konishi operator can be positioned in
different ways under the common color trace. Also, an operator with the same quantum numbers can be realized
using fermions of the type A WAB . Here we work in the simplest cases J = 0, 1 where these issues are irrelevant.
169
It is now rather obvious that repeating the same differentiation as in (19), we obtain
results for the descendants K[2J 0] and K[2J 2] similar to (20), since the BPS factor is annihilated by these derivatives.
In this paper we are only interested in J = 0 and J = 1. We restrict to SU(N ) gauge
group so that there is in both cases only one gauge invariant composite operator with the
right quantum numbers, the Konishi operator K1 (13) and

1
(51)
Tr WAB W CD WCD ,
2
respectively. We thus avoid the complications of operator mixing on the level of the primaries.
(K6 )AB =
2.5. Realization in terms of N = 1 superfields

The quantum calculations we have in mind are most easily carried out in N = 1 superspace. To this end we need to translate the N = 4 expressions obtained above into N = 1
notation.
The on-shell N = 4 gauge multiplet contains the fields [AB] , A , A
, A . The six
scalar fields AB = WAB |,=0
obey the reality condition (9). The spinor components are
defined by A = (1/3)B WBA |,=0

.
The transition from N = 4 to N = 1 is accompanied by a breakdown of R symmetry,
SU(4) SU(3) U (1). Accordingly, we can organize the three complex scalars and
three of the four spinors into on-shell N = 1 chiral matter multiplets:
I = 1,I +1 + ,I +1 ,
I = 1, 2, 3,
(52)
where we have identified = 1 . The remaining fields A, , (where we have set = 1 )

constitute the on-shell vector multiplet of N = 1 SYM. In the full N = 1 formulation
these are completed to off-shell chiral superfields I (xL , ) and a real YangMills gauge
superfield V (x, , ) by a set of auxiliary and pure gauge fields. Replacing the gauge field
A by its field strength F , we can form another chiral superfield, the N = 1 on-shell
super-field strength:

W = + F .
(53)
2.5.1. The operator K1
Let us now give the N = 1 expressions of the various gauge invariant operators from
the preceding subsections. The Konishi superfield is

K1 = Tr egV I egV I .
(54)
The role of the exponential factors is to make the composite operator gauge invariant. The
highest weight projection B (30) of the operator in the 10 is expressed in terms of matter
superfields of the same chirality:

B = Tr 1 , 2 3 ,
(55)
170
while the operator F is given by the chiral N = 1 field strength:

F = Tr W W .
(56)
The highest weight projection Y (32) of the operator in the 84 is again chiral:

Y = Tr 1 , 2 1 , 2 .
(57)
We now clearly see the advantage of working with the highest weight projections. In the
N = 1 language they all correspond to chiral operators. In particular, there is no need for
gauge factors like in the Konishi operator (54). The quantum calculations with such operators are significantly easier than with projections of mixed chirality, e.g., Tr([ 1 , 2 ] 3 ).
A quick way for constructing descendants is to use the on-shell N = 4 supersymmetry
transformations of the component fields (we only list the relevant part):
1
D] ,
B] +
ABCD [C
AB = [A

2

A = A () F gB AC , BC + i B D AB
(58)
(here [A, B] exceptionally means antisymmetrization with weight one). In the preceding
sections we used the superspace derivatives, e.g., D 4 D 4 , which is equivalent to making
two supersymmetry transformations with parameter 4 . With the help of the rules

(4 )2 1 = 0,
4 1 = 2g4 1 , 2 ,

4 3 = 4 1 ,
(4 )2 3 = 2g42 1 , 2 ,
4 1 = 4 2 = 0,
(59)
we can immediately obtain Eqs. (31).

2.5.2. The operator K6
Finally, here are the relevant expressions for the BMN operator K6 . The primary state
of the multiplet is in the 6 of SU(4) with N = 1 (i.e., SU(3) U (1)) projection K6I =
Tr( I J J ) + Tr( I J J ). It has a descendant K45 in the 45 (DL [210]) whose highest
weight all-chiral projection is a mixture of the operators11

F = Tr 1 W W .
B = Tr 1 1 2 , 3 ,
(60)
The right mixture can be determined by orthogonalization since, in complete analogy with
the Konishi case, only the B term follows from the classical field equations whereas the
F part is a generalized anomaly. Notice the important difference with the Konishi case:
even in the free theory, where K6 satisfies a semishortness condition of the type (46), it
does not contain any conserved current.
Performing two supersymmetry transformations with parameter 4 (recall (59)) we find
4 2
4 2
D B = gY,
(61)
D F = 2g 2 Y
11 We use the same notation as in the Konishi case, e.g., B, F , etc., although the meaning is now different.
with the bare descendant

Y = 2 Tr 1 1 1 , 2 2 ,
171
(62)
which is the highest weight of a 300 (DL [212]) [30].

The same transformation applied to the mixture
O45 = F 2gB
(63)
gives zero. This combination can be identified as a descendant of the protected half-BPS
multiplet O50 (DL [030]).
3. Calculation of 3 for the Konishi multiplet

3.1. The setup
In Section 2.1 we showed that the two ratios of two-point functions (21) and (22) contain
the same information about the anomalous dimension of the Konishi operator. The presence
of an overall factor g 2 in both sides of these equations means that in order to determine,
e.g., 3 , we need to compute one of these ratios at level g 4 (two loops). We also explained
why we choose to evaluate the second ratio,
R=

K84 K84
K10 K10
(64)
Once the ratio (64) is expanded in powers of g 2 , prior to the evaluation of any Feynman
diagram, we will see that similar linear combinations of integrals occur in both two-point
functions which leads to substantial cancellations. The remaining set of integrals can be
drastically reduced by comparison with the two-point function of a certain half-BPS operator. The latter is known to be non-renormalized, so the corresponding set of Feynman
graphs must sum into contact terms.
We begin by recalling Eq. (28) which states that the renormalized operator K10 is a
mixture of the bare operators B and F with some singular renormalization factors. In the
supersymmetric dimensional reduction scheme these factors have the following form [20]

b22 b21
b11
+ g4
+ O g6 ,
+
ZB = 1 + g 2
(65)
2

f21
+ f20 + O g 5 .
ZF = gf10 + g 3
(66)
An overall finite rescaling of the operator K10 carries over to its descendant K84 and drops
in the ratio of two-point functions (64). In the above we used this freedom to remove the
finite part of ZB . However, ZF must contain a finite part, the terms f10 and f20 (f10 is
the one-loop Konishi anomaly). In [20] it is shown that f21 as well as all the bmn are
determined by the anomalous dimension of the Konishi multiplet. In our calculations here,
as shown below, the coefficients b22 , b21 and f20 drop out, so we only need b11 and f21 ,
which will be determined directly.
172
Let us combine (31) and (61) into one equation suitable for treating both cases,
D 4 D 4 B = gY,
D 4 D 4 F = ag 2 Y,
(67)
where a = 4 in the case of K1 and a = 2 for K6 . (By definition a is the coefficient in

the protected linear combination F + agB.)
As discussed earlier, we assume that the quantum operator K84 is obtained from K10 by
naive application of the field equations:
gK84 = D 4 D 4 K10 = g(ZB agZF )Y = gZ84 Y.
Explicitly,
Z84 = 1 + g 2
(68)

6
b11
b22 b21 af21
af10 + g 4
af
+
20 + O g .
2
(69)
The detailed analysis of the two-point functions below confirms that this definition of Z84
renders the operator K84 finite.
Before going on, in order to check our normalizations, let us show how the well-known
value of 1 is obtained from the ratio (64). At the lowest order we have the equation
g 0
g 2 K84 K84
K10 K10
g 0
g 2 Y Y0
2g 2 1
= 2 .
BB0
x12
(70)
Here

2N (N 2 1)
BB0 = BB g 0 =
,
2 )3
(4 2 x12

12N (N 2 1)
Y Y0 = Y Y g 0 =
2 )4
(4 2 x12
(71)
are the tree-level two-point functions of the scalar operators B and Y (see Section 3.3.1).
It follows
3N
,
4 2
as first computed in [35].
1 =
(72)
3.2. Evaluating the two-point functions

Now we need to evaluate the two-point functions of the renormalized operators K10 and
K84 up to order g 4 . In the process we encounter the various two-point functions of the bare
operators B and F . We introduce the notation
BB2 =
F F0 =
BB g 2
g 2 BB0
F F g 0
BB0
BB4 =
F F2 =
BB g 4
g 4 BB0
F F g 2
g 2 BB0
,
,
BF3 =
BF g 3
g 3 BB0
Y Y2 =
Y Y g 2
g 2 Y Y0
Y Y4 =
Y Y g 4
g 4 Y Y0
173
We remark that BF g 1 = 0 simply because no such graph exists. For the two-point functions of the renormalized operators we find

K10 K10

2
2
4 2b11
+ BB4 g 4 + BF3 g 4 2f10
= ZB + BB2 g + g
BB0

2 2
4 2f10 f21
2
+ 2f10 f20 + F F2 g 4 f10
+ F F0 g f10 + g

+ O g6
(73)
and
K84 K84

b11 f10 + f21
a 2
= ZB2 g 2 2af10 g 4 2a
+ f20 f10
Y Y0
2

2b
11
2af10 + Y Y4 g 4 + O g 6 .
+ Y Y2 g 2 + g 4
(74)
We have normalized such that in both equations the r.h.s. goes like 1 + O(g 2 ). The ratio of
(74) and (73) can therefore straightforwardly be expanded in the coupling constant:

BB0
R = 1 + g 2 (Y Y2 BB2 ) f10 a + (a + f10 F F0 )
Y Y0

+ g 4 Y Y4 BB4 BB2 (Y Y2 BB2 )
2f21
(a + f10 F F0 )
g 4 f10 (2BF3 + aY Y2 ) g 4

2b11
2b11
Y Y2 +
f10 (a + f10 F F0 )
+ g 4 2 BB2 +

2
2
g 4 2f20 f10
(a + f10 F F0 ) (a + f10 F F0 ) g 4 f10
F F2 .
(75)
Note that the g 4 part of ZB has dropped out.

Let us try to simplify further. In Section 2.3 we explained that the combination
O10 = F + agB
(76)
(likewise O45 for the J = 1 case) is short and thus protected. In particular (see (40)),

O10 O10
(77)
=0
g2
up to contact terms. Since BF g 1 = 0 we can deduce
F F2 = a 2
(78)
up to a contact term. In dimensional regularization, two-point contact terms are O(

) as
long as the points are kept apart [33]. For our purposes we can put them to zero provided
that they are not multiplied by a singular term coming, e.g., from the Z factors. In (75)
F F2 appears with a finite factor and thus it can be replaced by a 2 .
174
Next, we recall that the operators O10 and K10 must be orthogonal, see (41). To first
order in g this means

(F + agB) B + f10 gF g 1 = gBB0 (a + f10 F F0 ) = O(
),
(79)
where we have again used BF g 1 = 0. Eq. (79) implies
a + f10 F F0 = O(
).
(80)
Notice that in dimensional regularization the right-hand side of Eq. (79) is not identically
zero. The orthogonality condition becomes exact only in the limit
0. While we try to
drop the expression a + f10 F F0 wherever possible, we cannot do so when it occurs with a
singular factor. For instance, the term f
21 (a + f10 F F0 ) must be kept.
The terms in the fourth line of Eq. (75) can be dropped since the combinations
BB2 +
2b11
2b11
= O(1) = Y Y2 +
(81)
are finite. Indeed, the poles

1 in both BB2 and Y Y2 have residues equal to their common
anomalous dimension 1 . So, these poles are removed by the same one-loop renormalization factor with b11 = 1 /2, as shown in (81) (see also (96) below).
Further, the difference Y Y2 BB2 in the first line of Eq. (75) can be dropped as well.
Indeed, both BB2 and Y Y2 contain only two graphs which are denoted by A and C in
Fig. 1. In Appendix A.1 it is shown that Graph C is a contact term. The operators B and
Y have the same poles
1 (and hence the same anomalous dimension 1 ) exactly because
diagram A occurs in Y Y2 and BB2 with the same coefficient, see (96) below. Thus,
Y Y2 BB2 = O(
).
(82)
Notice, however, that the same difference cannot be neglected in the second line of Eq. (75)
because it is multiplied by the divergent two-point function BB2 .
Putting together Eqs. (78)(82), we can considerably simplify Eq. (75):

2

BB0
R = 1 af10 g 2 + af10 g 2 + g 4 Y Y4 BB4 BB2 (Y Y2 BB2 )
Y Y0

f10
2af21
1+
F F0 .
g 4 f10 (2BF3 + aY Y2 ) g 4
(83)
We can still simplify the last term in (83) in a fashion that is universal to the J = 0, 1
cases:
(J +3)
BB0 12 ,

J
F F g 0 12
(12 )2 .
(84)
(For the exact expressions see the next subsection.) The configuration space propagator in
dimensional regularization is
12 =
(1
)
.
2 )(1
)
4 2 (x12
(85)
175
Now, the requirement of orthogonality (80) implies

f10 F F g 0
f10
F F0 = 1 + O(
).
=
a
BB0
a
(86)
The
dependence in this expression comes from the expansion of the propagators (85). In
the MS scheme the fractional powers of as well as the EulerMascheroni constant from
the expansion of -functions are absorbed into the mass scale, see Appendix A.3. In this
scheme

2 (
)
(2)
f10
1
2
F F0 = (1
)2 x12
+
(87)
a
2
and finally

2
1
f10
1+
+ 2 + O(
).
F F0 = ln x12
(88)
3.3. Evaluation of the graphs

In this section we elaborate the contributions BB2 , Y Y2 , BF3 and Y Y4 BB4 which
occur in (83).
3.3.1. Level g 0
At tree level we have

3
BB0 = BB g 0 = 1 2 , 3
3 , 2 1 = 2N N 2 1 12
.
(89)
One comment on the convention: the linear part of the W field is

1
W = D 2 D V .
(90)
4
Its lowest component is the physical fermion of the N = 1 SYM multiplet. One may
verify from the definitions in Appendix A.1 that indeed

W (1)W (2) g 0 ,,=0
(91)
= 2i
1 12
,=0
in agreement with our convention for the matter fermion propagator. Hence we find

F F g 0 = 16 N 2 1 ( 12 )2 .
(92)
We can now use the orthogonalization condition (80) to fix the one-loop anomaly coefficient f10 . In O10 we have a = 4 so that
N
.
32 2
The remaining tree-level correlator is

4
.
Y Y0 = Y Y g 0 = 12N 2 N 2 1 12
f10 =
(93)
(94)
176
Fig. 1. Graphs contributing to BB2 and Y Y2 ; the free matter lines are not shown.
3.3.2. Level g 2
We now turn our attention to the O(g 2 ) contributions BB2 and Y Y2 . Here and in the
sequel we ignore the one-loop self-energy diagrams since they cancel in the N = 1 formulation of N = 4 SYM. Further, we do not display free matter lines.
We find

BB g 2 = 6g 2 N 2 N 2 1 (212 A + 12 C),

4 2
2
A + 12
C ,
Y Y g 2 = 36g 2 N 3 N 2 1 212
(95)
3
with the graphs from Fig. 1, which are calculated in Appendix A.1. Note that in our conventions the YM propagator has a minus sign relative to the matter propagator 12 . This
sign has already been taken into account in the combinatorial factors in (95). On scaling
down by the respective tree-levels, we obtain

A
C
BB2 = 3N 2 2 + 2 ,
12 12

A
4 C
Y Y2 = 3N 2 2 +
(96)
.
2
12 3 12
This confirms the claim made in (82).
3.3.3. Level g 3
Next, we turn our attention to BF g 3 . We recall that the operators B and F are highest weight projections of the 10 of SU(4). It proves more convenient to calculate the
two-point function of a different projection of the 10, namely B = ( 1 [ 2 , 3 ]) and
F = (( 1 )( 1 )). One sees immediately that BB g 0 = B B g 0 , i.e., the normalization of the two components is the same. For F and F we have checked this in the
beginning of the section. Thus we can safely appeal to SU(4) and generalize results found
for this projection of the 10 to the entire SU(4) multiplet.12
The operator F contains the YM covariant derivative

1 = D 1 g D V , 1 + O g2
(97)
and B is made gauge invariant by insertion of the gauge bridge egV between superfields
of different chirality.
12 A word of caution is due here. Part of the SU(4) R symmetry may turn out anomalous in the quantum theory.
However, the transformation leading from B to B and from F to F actually uses the SU(2) SU(4) which can
be maintained manifest in the off-shell N = 2 harmonic superspace [36] formulation, see [23].
177
Fig. 2. Graphs GI...IV ; the crosses denote spinor derivatives.
To order g 3 there are three types of diagrams: first, those that do not involve the gauge
connection at the F point. These are the topologies GI through GIII in Fig. 2 and a few
other graphs that vanish when , = 0 at the outer points. Second, we may have one
connection line coming out of F . There are six related graphs, of which the only non-zero
one is displayed in Fig. 2 as graph GIV . Third, there are also graphs with a single YM line
emanating from the gauge bridge in B . However, they do not contribute.
The combinatorics for the non-vanishing graphs yields the sum

2
1
3 2
F B g 3 = 8g N N 1
(98)
GI GII + GIII GIV ,
2
where the minus sign from the YM propagator is included. The signs in the latter formula
are valid when the propagator with D is written to the left of the one with D and when
in GIV the D is on the YM line. The Grassmann integrations in all four supergraphs are
straightforward; in the YM sector we additionally assemble double derivatives into box
operators to find
GI = GII + GIII GIV = 2J,
and hence

FB g 3 = 24g 3 N 2 N 2 1 J.
(99)
(100)
The letter J stands for the Feynman integral (the notation is explained in Appendix A)

d 42
x3,5,6 (15 53 63 16 )
J = c07
(101)
.
2 x 2 x 2 x 2 x 2 x 2 x 2 )(1
)
(x15
16 35 36 25 23 26
(The factor c0 = (1
)
/(4 2 ) comes from the normalization of the configuration
space propagator.)
The evaluation of J can be done by switching over to momentum space and using the
powerful Mincer computer algorithm [37]. J is a three-loop integral of topology BU in
the Mincer classification. The program computes

2 2 (13
)
1 1 1 14
+
+
O
q
J=
(102)
3
(4)6
2 2
178
in the MS scheme. We transform back to configuration space:

J =

2 (4
)
3
1
1
x12
.
+
1
+
O(
)
2
2
16
(4 2 x12 )3
(103)
For the definitions of the Fourier transform and the mass scales in p- and x-space see
Appendix A.3. Finally, since J is real:

2
3N 3
+ 1 + O(
) x12
BF3 = 2
.
4
(104)
Next, we turn to Y Y2 , which we need up to O(

). Graph C is of no interest in this
context as it is a contact term. Graph A is computed in detail in Appendix A.1, Eq. (A.23).
On dividing by the tree-level BB0 we obtain
Y Y2 =

2
3N 1
+
1
+
O(
)
x12 .
2
4
(105)
The one-loop anomalous dimension of the operator Y (and in the same way of B) is therefore 1 = 3N/(4 2 ), as stated earlier (see (81)).
Collecting terms:
f10 (2BF3 4Y Y2 ) =

2
1
3 N2
x12 .
1
+
O(
)
2
2
4 (4 )
(106)
3.3.4. Level g 4
It might seem that the evaluation of the genuine four-loop integrals in Y Y4 and BB4 is
an insurmountable obstaclewe are now going to show how to get away with calculating
only one difficult integral.
The O(g 4 ) diagrams that occur in Y Y4 and BB4 are listed in Fig. 3. One of the advantages of the all-chiral choice for Y and B is that the gauge bridge egV does not occur.
Therefore, all diagrams have the same number of outer lines at both ends, namely four
lines in Y Y4 and three in BB4 . Also, all the outer lines connect to matter fields. Now, any
graph has one or two connected pieces and a number of free lines. In Fig. 3 we do not
show such spectator propagators. Also, we do not display the mirror graphs obtained
by permuting points 1 and 2; if needed, they will be denoted by, e.g., G17b (see (108)).
Above we have first computed BF3 and Y Y2 and then divided by the respective treelevels. Here we divide out the tree-level before calculationin practical terms this means
first of all to divide the combinatorial coefficient of any graph in Y Y4 by 12N 2 (N 2 1)
4
and that of any graph in BB4 by 2N (N 2 1). Further, we normalize the integrals by 12
3
and 12 , respectively. In doing so we may cancel spectator lines, because the propagator is
non-singular. As a result, an integral with n outer legs at each end will occur with n inverse
powers of 12 .
179
Fig. 3. Graphs G1...19 ; the mirror graphs are not shown; the outer points are split for clarity.
Apart from the 19 graphs in the list the correlators also contain six further diagrams that
together constitute the two-loop propagator correction. The propagator blob occurs once
per matter line, so once more in Y Y4 than in BB4 .13
13 We do not argue in this paper whether the two-loop blob is just finite or really vanishing. This is presumably
a scheme-dependent issue; here we do not need the explicit result.
180
Fig. 4. Propagator correction graphs denoted by in the text.
To be more tidy, we directly give the difference Y Y4 BB4 in which graphs G6 through
G10 drop:

1
1
1
Y Y4 BB4 = 12N 2 G1 + G4 + G11 G2 + 2N 2 1 + 242 .
(107)
2
2
8
The two sums of diagrams are
1
1
1 = 2G5 + G11 + G12 G13 + G15 + G16
4
2
1
3(G17 + G17b ) + G18 + 2G19 +
[ ],
2N 2 g 4
(108)
where G17b denotes the mirror image of G17 and the blob stands for the propagator corrections from Fig. 4; further,
1
2 = G3 + G11 2G12 + G14 + G15 2G16 + 4G18 .
2
We have organized the result in this form because
1 = O(
) = 2 .
(109)
(110)
These two conditions follow from the protectedness of the half-BPS operator O105 (DL
[040]). It has highest weight chiral projection Tr( 1 )4 or, in N = 4 notation, Tr(W12 )4 . In
[32] it has been shown that Tr( 1 )4 Tr( 1 )4 g 4 contains only contact, i.e., O(
) terms.
We can assume that this is true for any other projection of O105 . For our purposes it is most
convenient to chose the projection obtained by the SU(4) lowering operator T3 2 which
replaces a lower index 2 in WAB by 3, e.g., T3 2 W12 = W13 , i.e., T3 2 1 = 2 . Acting with
T3 2 on Tr( 1 )4 twice, we convert it into 2 Tr( 1 1 2 2 ) + Tr( 1 2 1 2 ). When
evaluating the two-point function of this chiral operator, we encounter two different color
structures corresponding to the sums of graphs 1 and 2 (the phenomenon is described
in [33]). Thus, protectedness of O105 implies that 1 = O(
) and 2 = O(
) separately.
Graphs G1 , G4 and G11 are products of the one-loop diagrams from Fig. 1:
G1 = A2 ,
G4 = AC,
G11 = C 2 .
We have already encountered the same graphs in Eq. (96) above:

1 2
1
2 1
C = 12N
AC + C .
BB2 (Y Y2 BB2 ) = 3N (2A + C)3N
3
2
4
(111)
(112)
181
The term C 2 is O(
2 ) and can be neglected. Even though AC = O(1) and thus cannot be dropped, it cancels against a similar term in Y Y4 BB4 in the second line
of Eq. (83):

2
G2
2 1 A
Y Y4 BB4 BB2 (Y Y2 BB2 ) = 12N
(113)
3 + O(
).
4
2 12
12
The evaluation of the Grassmann integration for graph G2 is trivialit actually uses
twice the same trick as for diagram A: the two parallel propagators coming in from the
outer points are evaluated at outer , = 0, hence saturation of the spinor integral at the
vertices at the inner end of the double propagators produces box operators on the subsequent lines. We find

dx3,5
7
.
G2 = c0
(114)
4
2
2
2 x 4 )(1
)
(x13 x23 x35 x15
52
This is a genuine four-loop integral of the 4 type. A method for evaluation of such integrals was developed in Dubna in the late 1970s [38]. The result in the MS scheme is14

(14
)
1
1
1 + 14
+ 121 2 (2)
2 + O
3 q 2
.
(115)
8
3
(4) 4
Like in the case of the integral J we return to x-space by the inverse Fourier transform and
divide out three powers of the propagator:

2 (2
)
1
5 (2)
1
5
G2
+
+
+
O(
)
x12
=
+
.
(116)
3
2
2
2
4
4
2
12 4(4 ) 2
G2 =
Combined with the result for graph A (A.23):

2
2 (2
)
3N 2
1
3
G2
2 1 A
12N
+ + O(
) x12
3 =
.
4
2
2
2 12
4
4
(4 )
12
(117)
Finally, Eq. (83) becomes:

BB0
1 g2N
1 g4N 2
R=1+
+
Y Y0
2 4 2
4 (4 2 )2

3 2 (2
) 3g 4 N 2 1
1 2
3g 4 N 2 1
+
x
x12
+
12
(4 2 )2 4
4
(4 2 )2 4
4
2

+ 2 + O(
).
+ g 4 8f21 ln x12
(118)
We see that the pole itself cancels between the second and the third line, whereas a simple
logarithm persists in the
-expansion. On the other hand, conformal behavior dictates the
absence of any logarithm from the ratio R, because K10 and K84 have the same anomalous
dimension. We must therefore choose
f21 =
3N 2
.
32(4 2 )2
14 We are indebted to D. Kazakov for help on this point.
(119)
182
The value is fully consistent with the abstract analysis in [20]: one must find f21 = 14 1 f10 .
Our last version of Eq. (83) is

BB0
1 g2N
5 g4N 2
R=1+
+ O g6 ,
Y Y0
2 4 2
4 (4 2 )2
(120)
which we match with the superspace prediction (22): the normalizations of the two-point
functions ought to differ by

( 2)
2
3
2 1
4
(121)
+ g 2 +
+ O g6 .
+
=1+g
2
1
1
2g 2 1
(The equation was scaled down by 21 in order to set the lowest order equal to one.) We
put in 1 = 3N/(4 2 ) and extract
2 = 3
N2
,
(4 2 )2
3 =
21 N 3
.
4 (4 2 )3
(122)
The value of 2 has first been obtained in [22]; the value of 3 is in agreement with [14].
4. The BMN operator K6
In this section we repeat the calculation, with relatively minor changes, for the next
higher BMN multiplet K6 with J = 1. For the definition and the relevant N = 1 projections
we refer the reader to Section 2.5.2.
The tree-level correlators are

4
BB0 = 2 N 2 1 N 2 4 12
,

8
F F0 = N 2 1 N 2 4 ( 12 )2 12 ,
N

5
.
Y Y0 = 8N N 2 1 N 2 4 12
(123)
Consequently (recall a = 2),
f10 =
N
.
32 2
(124)
At O(g 2 ) we find
BB2 = 2N (2A + 2C),

5
Y Y2 = 2N 2A + C ,
2
(125)
(126)
and therefore 1 = 2N/(4 2 ), in agreement with [3].

For the calculation of BF g 3 we swap once again SU(4) projections: the choice

B = ( 1 1 [ 2 , 3 ]) and F = ( 1 ( 1 )( 1 )) is more convenient. The graphs that
contribute are the same as before,15 just multiplied by an additional free line.

F B g 3 = 4g 3 N N 2 1 N 2 4 (GII + GIII GIV ).
183
(127)
Curiously, the matter sector graph GI does not occur. The YM sector diagrams sum into
2J as before, so that
BF3 = 4N J.
In (83) we have the term
f10 (2BF3 + aY Y2 ) = 8Nf10
(128)

J
3
12
A
2
12

+ O(
),
(129)
where we have dropped graph C from Y Y2 . Remarkably, this is the same combination of
J with A as above for the Konishi multiplet. Since the one-loop anomaly coefficient f10
is identical (by accident?), too, we can simply scale down Eq. (106) by a factor 3 without
doing any further calculation:

2
1
N2
1
+
O(
)
x12 .
f10 (2BF3 2Y Y2 ) =
(130)
2
2
4(4 )
In the genuine O(g 4 ) part of the calculation there are no new graphs. We find

1
2 1
G1 + G4 + G11 G2
Y Y4 BB4 = 4N
2
2

1
2
+ 2N 1 + G11 + 362 .
4
(131)
The sums 1 , 2 are as defined in the last section. We have written the result in this form,
because (1 + 1/4G11 ) = O(
) = 2 follows from the protectedness of the next higher
half-BPS multiplet O196 . It is interesting to note that the protectedness of this operator is
equivalent to that of O105 since graph G11 is a contact term. The phenomenon has been
observed in a variety of cases [39].
Last,

1
C = 4N 2 AC + C 2 ,
BB2 (Y Y2 BB2 ) = 2N (2A + 2C)2N
(132)
2
and by neglecting graph G11 = C 2 ,
Y Y4 BB4 BB2 (Y Y2 BB2 ) = 4N 2
1 A2
G2
3
4
2 12
12

+ O(
).
(133)
Hence also this term can be inferred from the results of the last section just by scaling down
equation (117) by a factor 3:

2 (2
)
1
N2
3
Y Y4 BB4 BB2 (Y Y2 BB2 ) =
(134)
.
+ + O(
) x12
2
2
(4 ) 4
4
15 There are a few new vanishing diagrams.
184
We substitute into (83):

1 g2N
1 g4N 2
a 2 BB0
R
=
1
+
+
4 4 2
16 (4 2 )2
g 2 Y Y0

4
2
1
3 2 (2
)
1 2
g N
g4N 2 1
+
x
x12
+
12
(4 2 )2 4
4
(4 2 )2 4
4
2

+ 2 + O(
).
+ g 4 4f21 ln x12
(135)
The logarithm is eliminated by setting

f21 =
N2
,
16(4 2 )2
(136)
which conforms with the requirement f21 = 14 1 f10 .

Finally, (83) becomes

BB0
1 g2N
7 g4N 2
R=1+
+ O g6 .
2
2
2
Y Y0
4 4
16 (4 )
(137)
We equate with (121) and use 1 = 2N/(4 2 ) to solve:

2 =
3 N2
,
2 (4 2 )2
3 =
17 N 3
.
8 (4 2 )3
(138)
(The result for 2 was originally stated in [3].)
5. Conclusions
In this paper we presented a three-loop calculation of the anomalous dimensions of
the first two operators of the BMN family corresponding to charges J = 0, 1. Here we
summarize the results, including the lower-order ones:
N2
,
(4 2 )2
3 =
21 N 3
,
4 (4 2 )3
3 N2
,
2 (4 2 )2
3 =
17 N 3
.
8 (4 2 )3
J = 0:
1 =
3N
,
4 2
2 = 3
J = 1:
1 =
2N
,
4 2
2 =
(139)
Our result for 3 , J = 0 (Konishi) agrees with that of [14]. The result for 3 , J = 1 is new.
As explained in the introduction, these two values of 3 are sufficient to test the form of
the three-loop dilatation operator proposed in [3,9]. We recall that in both of these references the requirement of BMN scaling was needed to fix the operator (in [3] the additional
assumption of integrability was made, but later in [9] it was shown to follow from the
larger symmetry taken into account). Thus, we can say that our result not only confirms the
validity of the dilatation operator approach, but it also provides indirect evidence for BMN
scaling at three loops.
We point out that our values (139) are exact, i.e., we have not made use of the large N
(planar) approximation. Yet, they coincide with the predictions of [3,9] which are supposed
185
to apply to the planar sector only. This absence of subleading 1/N corrections may be
explained by the relative shortness (small J ) of our operators, for which the non-planar
effects may not be important.
Let us finish by a few remarks about the possible further developments. Two generalizations of our calculation could be of interest. One of them would be to go to four loops.
An obvious difficulty is the exponential growth of the combinatorics. However, our present
experience shows that in the end these massive calculations always produce rather simple
results, so one might hope this is not just a low loop effect. Another complication would
be the occurrence of new, five- or maybe six-loop scalar integrals. In this paper we had to
deal with a single difficult four-loop integral. Hopefully, the method of [38] is applicable to
higher loops as well. Finally, at four loops we may reach the limitations of the supersymmetric dimensional reduction scheme (see [40]). This scheme, although not indispensable,
is highly efficient due to the manifest N = 1 supersymmetry.
The other generalization would be to look at the next value J = 2. This would be of great
interest, since our results so far helped to unambiguously fix the form of the three-loop
dilatation operator, but new results for J = 2 would provide a check on a true prediction.
The obvious major complication in this direction is the possibility of mixing among the
various realization of the J = 2 BMN operator. Still, we believe that trying is worthwhile.
Finally, we should mention another interesting line of research. Instead of going in the
direction of higher values of J (higher twist), one might stay with twist two operators,
but increase their spin (see, e.g., [41] for a recent discussion). Three-loop results of this
type are already available [14], but our method would provide an alternative check using
more conventional field theory methods. Yet another approach to higher spins consists in
calculating the four-point function of the stress-tensor multiplet and extracting anomalous
dimensions by OPE techniques. We believe that a generalization of the harmonic superspace approach to four-point functions developed in [42] could still be applicable at order
O(g 6 ).
Acknowledgements
Most of this work was done while B.E. was a member of the theory group of the Dipartimento di Fisica, Universit di Roma Tor Vergata, where he held a DFG postdoctoral
fellowship. E.S. is grateful to this group and especially to Augusto Sagnotti for extending
to him their warm hospitality at Tor Vergata. We are deeply indebted to Yassen Stanev
who participated at the initial stages of this project and has continuously shared his experience with us. We also profited from numerous discussions with G. Arutyunov, N. Beisert,
M. Bianchi, T. Binoth, D. Kazakov, G.-C. Rossi, Ch. Schubert, M. Staudacher, F. Tkachov.
This work was supported in part by the MUIR-COFIN contract 2003-023852 and by the
INTAS grant No. 00-00254.
186
Appendix A
A.1. Feynman rules and conventions
The classical N = 4 SYM action formulated in terms of N = 1 superfields has the form

1

SN =4 = d 4 x d 2 d 2 Tr egV I egV I +
d 4 xL d 2 Tr W W
4

g
4
2
I
J
K
+
(A.1)
+ c.c. .
d xL d
I J K Tr ,
3!
Here
xL = x + i ,
xR = x i
(A.2)
are the spacetime coordinate in the (left-handed) chiral and (right-handed) antichiral bases
of superspace. All the superfields are in the adjoint representation of the gauge group
SU(N ), and the generators and the structure constants satisfy the relations

Tr T a T b = ab ,
(A.3)
f abc f abd = 2N cd .
To this classical action one should add (Feynman) gauge fixing and ghost terms. For our
purposes we only need the cubic vertices (except the V V V one). The relevant part of the
gauge-fixed action is

1
g.f.
SN =4 = d 4 x d 2 d 2 Tr I I V V + g[V , I ] I
2

+ g d 4 xL d 2 Tr 1 2 , 3

g d 4 xR d 2 Tr 1 2 , 3 + .
(A.4)
Our spinor conventions are [43]:
=
,
=
,
12 = 1,

= ,
(A.5)
and exactly the same with dotted indices. Complex conjugation replaces an undotted by a
dotted index and vice versa; however, it does not exchange lower and upper position.
The 2 2 sigma matrices are Hermitian. The matrix is obtained from a by raising
of both indices as defined by the last equation. They satisfy the following relations:
= i ,
= i ,

( ) = 2 ,
( )
= 2 .
(A.6)
In N = 4 supersymmetry spinors carry SU(4) indices with canonical positions A , A .

Complex conjugation makes a lower internal index go up and an upper go down, while the
sequence of indices of a tensor is not changed. Last,
1234 =
1234 = 1.
The superspace covariant derivatives are

D = + i ,
(A.7)
D = i ,
= .
187
The derivatives with upper indices are obtained by raising in the usual fashion. A useful
shorthand is
2 = ,
2 = ,
(A.8)
and similarly for the product of two different spinors. Consequently,

1
1
D D 2 D 2 2 = 1,
D D 2 D 2 2 = 1.
4
4
Under the x-integral we may thus identify
d 2 = D D.
d 2 = D D,
(A.9)
(A.10)
We are going to perform our graph calculations in x space,16 since the supersymmetry
algebra is easier there. The propagator of the chiral superfields is

I (1) J (2) = JI 12 ,
12 =
1
2
4 2 x12
(A.11)
(We find it more convenient not to write the gauge group trace ab it is there, of course.)
In the supersymmetric dimensional reduction (SSDR) scheme with D = 4 2
, the last
formula becomes
12 =
(1
)
.
2 )(1
)
4 2 (x12
(A.12)
In these expressions we use the coordinate difference

x12 = x1L x2R 2i1 2 ,
(A.13)
which is manifestly chiral/antichiral and invariant under the supersymmetry transformations

x = i
+ i
,
=
,
=
.
(A.14)
The various nilpotent shifts in the propagator (A.11) can conveniently be assembled into
an exponential factor:
1

1
1
1 + 2 2 21 2 ]
=
exp
i[
e12 2 .
1
1
2
2
x12
x12
x12
(A.15)
Our technique for evaluating supergraphs is to expand these exponential shifts in , and to
integrate out the Grassmann variables. What is left is a multiple integral of some differential
operator acting on a product of x-space propagators. Simplifications within a sum of graphs
can usually be demonstrated just by partial integration in configuration space. The matter
propagator satisfies the Greens function property
12 = D (x12 ),
(A.16)
16 Occasionally, when we need to evaluate some complicated integrals, we switch over to momentum space
where we can use the powerful Mincer algorithm.
188
which can often be used to remove an integration.

Correlation functions in Minkowski space have the weight factor eiS under the path
integral. This would usually mean to put a factor i on every vertex in the action (A.4). On
the other hand, the i is clearly absorbed into the integration measure when Wick-rotating
to Euclidean space. It is therefore a consistent prescription to take the vertices from (A.4)
without introducing extra is and to simply evaluate all configuration space integrals in
Euclidean signature. Note that the Greens function property (A.16) is metric-independent.
The propagator for the gauge superfield in Feynman gauge is

2 2
V (1)V (2) = + 12 122 .
4 2 x12
(A.17)
(We omitted ab once again.) Due to the Grassmann delta functions in the numerator, we
could also use (x1L x2L )2 or (x1R x2R )2 in the denominator.
The regularized version of (A.17) picks up a factor (1
)
, and 1/x 2 is modified
to 1/(x 2 )(1
) . The numerator is not changedthe very essence of the SSDR scheme is to
handle the spinors and the Dirac algebra like in four dimensions.
The graphs GI and GII from Section 3.3.3 contain the structure

(i j k l ) Tr i j k l

= 2 (i j )(k l ) (i k )(j l ) + (i l )(j k )
2i
i j k l ,
(A.18)
with (i j ) = i j . The four-index antisymmetric tensor drops in the integral J since it
is a pseudo-tensor: the final result for this two-point integral can only depend on x12 (or the
outer momentum, respectively) and out of one vector one cannot construct a pseudo-scalar.
The other graphs we actually have to evaluate do not involve any complicated trace structures. SSDR may become inconsistent when a contraction of two antisymmetric tensors
occurs [40]. Our calculation stays clear of this problem.
To give an impression of the formalism we evaluate the one-loop diagrams A and C
from Fig. 1. Graph A is particularly simple:

2
2
43 42
.
A = d D xR3 d 2 3 d D xL4 d 2 4 13
(A.19)
We want the 1 , 1 , 2 , 2 = 0 component of the supergraph, i.e., an exchange between
the physical scalars at the outer points. Let us inspect the nilpotent shifts of the various
propagators. We find
x13 = x1 x3R ,
x42 = x4L x2
(A.20)
exactly because the outer thetas are zero. Hence, we do not have to expand these propagators. The middle line has the shift
x43 = x4L x3R 2i4 3 ,
(A.21)
which can be written in the exponential form
e2i(4 4 3 )
42 32 D
1
1
2 2

=
(x4L3R ).
4
4
3
2
2
c0
(x4L3R
)(1
)
(x4L3R
)(1
)
(A.22)
189
Here c0 = (1
)
/(4 2 ) comes from the standard scalar propagator. In the x-space
MS scheme (A.33) we obtain:

d D x3
A = c04
2 x 2 )2(1
)
(x13
23

2
2 (3
)
3 (2)
1
1
1
+1+
+O
x12
=
.
(A.23)
2
2
2
2
2(4 2 )
(4 x12 )
Next, we consider graph C:

d D x5,6 d 2 5,6 d 2 5,6 15 52 16 62 V (5)V (6) .
(A.24)
We simplify the calculation by a trick: let us shift the integration variables x5 , x6 to

x5R , x6R . The Jacobian for such a nilpotent shift is 1. Since once again 1 , 1 , 2 , 2 = 0,
the propagators connecting to point 1 will no longer be expanded; there is also no derivative
acting on the YM line. For the other two lines we find

2i(5 2 5 ) 1
2i(6 2 6 ) 1
e
.
e
(A.25)
2
2
x25
x26
The numerator of the YM propagator behaves like a delta-function. We can therefore
replace 6 5 , 6 5 in the last equation. The thetas square up and yield a trace on the
derivatives which is a perfect square. It follows

d D x5 d D x6
C = +c05 2
(A.26)
,
2 x 2 x 2 x 2 x 2 )(1
)
(x15
25 56 16 26
where one minus comes from the sign of the YM propagator (we have often absorbed this
into the combinatorics) and the second one from the i 2 . The integral itself is in fact finite
2 )(13
) . We
(O(
0 ) and higher). In dimensional regularization its functional form is 1/(x12
conclude
C = O(
)
1
2 )(23
)
(x12
(A.27)
A.2. Decoupling of the BPS operators D84 and D300

We first discuss the J = 0 case. The bare second level descendant of the Konishi operator is (32)

Y = Tr 1 , 2 1 , 2 ,
which is tree-level and one-loop orthogonal to

1

D84 = Tr 1 1 Tr 2 2 Tr 1 2 Tr 1 2 Y,
(A.28)
N
cf. [21]. The operator D84 is one-loop protected since in its two-point function graph A
drops out and only graph C survives.
190
We have to show that at O(g 4 ), too, D84 is a pure state and that it stays orthogonal to
Y; implying that Y remains an eigenvector up to O(g 4 ) without receiving any admixtures
of D84 .
Calculation of the combinatorics yields

D84 Yg 4 = 72N N 2 1 N 2 4 2 ,

D84 D84 g 4 = 24N 2 N 2 1 N 2 4 2N 2 1 + N 2 3 2 ,
(A.29)
with the sums of graphs 1 , 2 from Section 3.3.4. Thereby, both correlators vanish up to
contact terms.
For our other example with J = 1 the situation is strictly analogous: there is the second
level descendant (62)

Y = 2 Tr 1 1 1 , 2 2
of the operator L6 and the 1/4 BPS operator

D300 = Tr 1 1 1 Tr 2 2 2 Tr 1 1 2 Tr 1 2

3

+ Tr 1 2 2 Tr 1 1 + Y,
N
which is tree- and one-loop orthogonal to Y [30].
At two-loop level we have

D300 Yg 4 = 96 N 2 1 N 2 4 N 2 9 2 ,
D300 D300 g 4 =
12(N 2
1)(N 2
4)(N 2
(A.30)
9)
N

1
2
10N 1 + G11 + 2N 2 24 2
4
thus proving our assertions. Remarkably, the sum 1 is shifted by 1/4G11 = 1/4C 2 =
O(
2 ) exactly as in the two-loop two-point function of the 1/2 BPS single trace state
O196 , see Section 4.
A.3. The mass scale and the MS scheme
We work in D = 4 2
dimensions. Recall the configuration space propagator (85)
12 =
(1
)
(2 )
2 )(1
)
4 2 (x12
into which we have introduced the mass scale in order to keep the canonical dimension
of the chiral superfield = 1. Note that

2 (2)
3 (3)
1+
+ ,
(1
) = e
(A.31)
2
3
191
so that we can write

12 =

3
1 2 2
2 (2)
1
+
+
O
,
12
2
2
4 2 x12
(A.32)
which defines the MS scheme mass scale

2 = 2 e .
(A.33)
In this scheme the fractional powers of and the EulerMascheroni constant are made
to disappear from the results for Feynman integrals. We use this definition throughout
the paper. Nevertheless, one could do slightly better: the form of the propagator actually
suggests to define
2
2
= (1
).
(A.34)
This redefinition of the mass scale additionally makes the (2) = 2 /6 disappear from our
formulas (87), (116), (A.23).
The MS scheme defined in (A.33) is particular to configuration space; it is inequivalent
to the usual definition in momentum space, where
2 = 2 4e
(A.35)
helps to eliminate the EulerMascheroni constant as well as fractional powers of 4 and

from the results. It was pointed out to us by D. Kazakov that the choice
2
(2 4)
=
(1
)
(A.36)
also hides all (2), which inspired our second x-space prescription above.
As we have kept the dimension of the fields at their canonical values, every vertex
carries a factor (2 )
in x-space (or its inverse in momentum space). In order to simplify
the notation we have suppressed the mass scale throughout the paper. It can simply be
restored after integration by completing any fractional power like
2
2
2 2
2
q
x
(A.37)
x ,
q
.
2
Let us illustrate the change of schemes on the example of the integral from graph A:

d D x3
h(x12 ) = c04
(A.38)
.
2 x 2 )2(1
)
(x13
23
The integral can be evaluated directly in x-space, e.g., by the standard Feynman parameter
trick. We rather choose to mimic the steps needed to use Mincer; therefore, we go through
momentum space and then transform back. The forward Fourier-transform is
D

D
1
D
iqx 1
D2 2
2
d xe
(A.39)
=4
.
() (q 2 )( D2 )
(x 2 )
The backward Fourier-transform has the same form with the roles of x and q exchanged,
but it comes with a factor 1/(2)D . In our integral h we replace every x-space propagator
192
by its back transform from momentum space and then integrate out x3 :

h(x12 ) =
(2 )
d D q iqx12
e
H (q),
(2)D

H (q) =
(2 )
d D p
(2)D p 2 (p q)2
2
. (A.40)
We have reinstated the mass scale on the measure since we intend to demonstrate explicitly
how redefinitions of 2 help to cast the result into a tidier form. The subintegral in H (q)
is again elementary. We find:
2 (2
)
1
q
(
)2 (1
)4
4
2
(4) 4
(2 2
)2
2 (2
)

q
1
1
4 2
2
=
+
(2)
+
O(
)
+
12
8
+
2
(4)4 42
2
2 (2
)

q
1
4
1
+
12
(2)
+
O(
)
+
=
(4)4 2
2 (2
)

q
1
4
1
+
12
+
O(
)
.
+
=
(A.41)
(4)4 2
H (q) =
In the second to the last line the momentum space MS mass scale is used, in the last line
the further modified version of it.
The MS result could also be taken from Mincer: the integral can e.g. be obtained by
topology T1 with p52 in the numerator. In general, the programs results are to be multiplied
with a factor 1/(4)2l , where l is the loop order as seen in momentum space. Next, we undo
the p-space MS scheme. We should remark, however, that the reconstruction of the (2)
terms is not obvious. Luckily, these are O(
2 ) subleading, and thus the part of the integral
J needed in our calculations is insensitive to the issue.
Finally, we transform back to configuration space, where we change to the new MS
scheme (A.33) or, if desired, to its modified version with 2 .
2 )(3
)
(2 3
)(
2 )(1
)4
1 (2 x12
2 )2
(4 2 )3
4(2
)(2 2
)2
(x12

2 )(3
)
2
1
1 + 3
12 + 9 2 + 3 (2)
1 (2 x12
+
+
+
O
=
2 )2
2
2
4
(4 2 )3
(x12

2 )(3
)
2
1
1 3 (2)
1 ( 2 x12
+
+
+
O
=
2 )2
2
2
4
(4 2 )3 (x12

2 )(3
)
2
1
1 ( 2 x12
1
=
(A.42)
.
+ + 0
+ O
2 )2
2
2
(4 2 )3 (x12
h(x12 ) =
Last, we remark that these rescalings of 2 cannot influence our results for the anomalous
dimensions, because the logarithmsand with it the mass scaledrop from the defining
Eq. (83). In fact, in N = 4 the anomalous dimensions should be fully scheme-independent.
193
References
[1] D. Berenstein, J.M. Maldacena, H. Nastase, Strings in flat space and pp waves from N = 4 super-Yang
Mills, JHEP 0204 (2002) 013, hep-th/0202021.
[2] J.A. Minahan, K. Zarembo, The Bethe-ansatz for N = 4 super-YangMills, JHEP 0303 (2003) 013, hepth/0212208;
A.V. Belitsky, S.E. Derkachov, G.P. Korchemsky, A.N. Manashov, Superconformal operators in N = 4
super-YangMills theory, hep-th/0311104.
[3] N. Beisert, C. Kristjansen, M. Staudacher, The dilatation operator of N = 4 super-YangMills theory, Nucl.
Phys. B 664 (2003) 131, hep-th/0303060.
[4] N. Beisert, The complete one-loop dilatation operator of N = 4 super-YangMills theory, Nucl. Phys. B 676
(2004) 3, hep-th/0307015;
N. Beisert, M. Staudacher, The N = 4 SYM integrable super spin chain, Nucl. Phys. B 670 (2003) 439,
hep-th/0307042.
[5] S. Frolov, A.A. Tseytlin, Multi-spin string solutions in AdS(5) S 5 , Nucl. Phys. B 668 (2003) 77, hepth/0304255;
G. Arutyunov, S. Frolov, J. Russo, A.A. Tseytlin, Spinning strings in AdS(5) S 5 and integrable systems,
Nucl. Phys. B 671 (2003) 3, hep-th/0307191;
G. Arutyunov, J. Russo, A.A. Tseytlin, Spinning strings in AdS(5) S 5 : new integrable system relations,
Phys. Rev. D 69 (2004) 086009, hep-th/0311004.
[6] N. Beisert, J.A. Minahan, M. Staudacher, K. Zarembo, Stringing spins and spinning strings, JHEP 0309
(2003) 010, hep-th/0306139.
[7] G. Arutyunov, M. Staudacher, Matching higher conserved charges for strings and spins, JHEP 0403 (2004)
004, hep-th/0310182;
V.A. Kazakov, A. Marshakov, J.A. Minahan, K. Zarembo, Classical/quantum integrability in AdS/CFT,
JHEP 0405 (2004) 024, hep-th/0402207;
G. Arutyunov, M. Staudacher, Two-loop commuting charges and the string/gauge duality, hep-th/0403077;
G. Arutyunov, S. Frolov, M. Staudacher, Bethe ansatz for quantum strings, hep-th/0406256.
[8] N. Beisert, C. Kristjansen, J. Plefka, M. Staudacher, BMN gauge theory as a quantum mechanical system,
Phys. Lett. B 558 (2003) 229, hep-th/0212269.
[9] N. Beisert, The su(2|3) dynamic spin chain, Nucl. Phys. B 682 (2004) 487, hep-th/0310252.
[10] D.J. Gross, A. Mikhailov, R. Roiban, Operators with large R charge in N = 4 YangMills theory, Ann.
Phys. 301 (2002) 31, hep-th/0205066.
[11] A. Santambrogio, D. Zanon, Exact anomalous dimensions of N = 4 YangMills operators with large R
charge, Phys. Lett. B 545 (2002) 425, hep-th/0206079.
[12] T. Klose, J. Plefka, On the integrability of large N plane-wave matrix theory, Nucl. Phys. B 679 (2004) 127,
hep-th/0310232.
[13] A. Vogt, S. Moch, J.A.M. Vermaseren, The three-loop splitting functions in QCD: the non-singlet case, Nucl.
Phys. B 688 (2004) 101, hep-ph/0403192.
[14] A.V. Kotikov, L.N. Lipatov, A.I. Onishchenko, V.N. Velizhanin, Three-loop universal anomalous dimension
of the Wilson operators in N = 4 SUSY YangMills model, Phys. Lett. B 595 (2004) 521, hep-th/0404092.
[15] D. Anselmi, The N = 4 quantum conformal algebra, Nucl. Phys. B 541 (1999) 369, hep-th/9809192.
[16] B. Eden, On two fermion BMN operators, Nucl. Phys. B 681 (2004) 195, hep-th/0307081.
[17] K.A. Intriligator, W. Skiba, Bonus symmetry and the operator product expansion of N = 4 super-Yang
Mills, Nucl. Phys. B 559 (1999) 165, hep-th/9905020.
[18] M. Bianchi, S. Kovacs, G. Rossi, Y.S. Stanev, Properties of the Konishi multiplet in N = 4 SYM theory,
JHEP 0105 (2001) 042, hep-th/0104016.
[19] T.E. Clark, O. Piguet, K. Sibold, Supercurrents, renormalization and anomalies, Nucl. Phys. B 143 (1978)
445;
K. Konishi, Anomalous supersymmetry transformation of some composite operators in SQCD, Phys. Lett.
B 135 (1984) 439.
[20] B. Eden, C. Jarczak, E. Sokatchev, Y. Stanev, Operator mixing at level g 4 in N = 4 SYM: the Konishi
anomaly revisited, hep-th/0501077.
194
[21] M. Bianchi, S. Kovacs, G. Rossi, Y.S. Stanev, On the logarithmic behavior in N = 4 SYM theory, JHEP 9908
(1999) 020, hep-th/9906188.
[22] M. Bianchi, S. Kovacs, G. Rossi, Y.S. Stanev, Anomalous dimensions in N = 4 SYM theory at order g 4 ,
Nucl. Phys. B 584 (2000) 216, hep-th/0003203.
[23] G. Arutyunov, E. Sokatchev, A note on the perturbative properties of BPS operators, Class. Quantum
Grav. 20 (2003) L123, hep-th/0209103.
[24] M. Bianchi, G. Rossi, Y.S. Stanev, Surprises from the resolution of operator mixing in N = 4 SYM, Nucl.
Phys. B 685 (2004) 65, hep-th/0312228.
[25] V.K. Dobrev, V.B. Petkova, All positive energy unitary irreducible representations of extended conformal
supersymmetry, Phys. Lett. B 162 (1985) 127;
L. Andrianopoli, S. Ferrara, E. Sokatchev, B. Zupnik, Shortening of primary operators in N -extended
SCFT(4) and harmonic-superspace analyticity, Adv. Theor. Math. Phys. 3 (1999) 1149, hep-th/9912007;
P. Heslop, P.S. Howe, On harmonic superspaces and superconformal fields in four dimensions, Class. Quantum Grav. 17 (2000) 3743, hep-th/0005135.
[26] S. Ferrara, E. Sokatchev, Superconformal interpretation of BPS states in AdS geometries, Int. J. Theor.
Phys. 40 (2001) 935, hep-th/0005151.
[27] G. Arutyunov, S. Frolov, A.C. Petkou, Operator product expansion of the lowest weight CPOs in N = 4
SYM(4) at strong coupling, Nucl. Phys. B 586 (2000) 547, hep-th/0005182;
G. Arutyunov, S. Frolov, A.C. Petkou, Operator product expansion of the lowest weight CPOs in N = 4
SYM(4) at strong coupling, Nucl. Phys. B 609 (2001) 539, Erratum;
G. Arutyunov, B. Eden, E. Sokatchev, On non-renormalization and OPE in superconformal field theories,
Nucl. Phys. B 619 (2001) 359, hep-th/0105254;
B. Eden, E. Sokatchev, On the OPE of 1/2 BPS short operators in N = 4 SCFT(4), Nucl. Phys. B 618 (2001)
259, hep-th/0106249;
P.J. Heslop, P.S. Howe, A note on composite operators in N = 4 SYM, Phys. Lett. B 516 (2001) 367, hepth/0106238;
S. Penati, A. Santambrogio, Superspace approach to anomalous dimensions in N = 4 SYM, Nucl. Phys.
B 614 (2001) 367, hep-th/0107071;
F.A. Dolan, H. Osborn, On short and semi-short representations for four-dimensional superconformal symmetry, Ann. Phys. 307 (2003) 41, hep-th/0209056.
[28] J.H. Park, N = 1 superconformal symmetry in 4 dimensions, Int. J. Mod. Phys. A 13 (1998) 1743, hepth/9703191;
H. Osborn, N = 1 superconformal symmetry in four-dimensional quantum field theory, Ann. Phys. 272
(1999) 243, hep-th/9808041.
[29] M. Bianchi, B. Eden, G. Rossi, Y.S. Stanev, On operator mixing in N = 4 SYM, Nucl. Phys. B 646 (2002)
69, hep-th/0205321.
[30] A.V. Ryzhov, Quarter BPS operators in N = 4 SYM, JHEP 0111 (2001) 046, hep-th/0109064.
[31] B. Eden, P.S. Howe, P.C. West, Nilpotent invariants in N = 4 SYM, Phys. Lett. B 463 (1999) 19, hepth/9905085.
[32] S. Penati, A. Santambrogio, D. Zanon, Two-point functions of chiral operators in N = 4 SYM at order g 4 ,
JHEP 9912 (1999) 006, hep-th/9910197.
[33] S. Penati, A. Santambrogio, D. Zanon, More on correlators and contact terms in N = 4 SYM at order g 4 ,
Nucl. Phys. B 593 (2001) 651, hep-th/0005223.
[34] P.S. Howe, P.C. West, Superconformal invariants and extended supersymmetry, Phys. Lett. B 400 (1997)
307, hep-th/9611075.
[35] D. Anselmi, M.T. Grisaru, A. Johansen, A critical behaviour of anomalous currents, electricmagnetic universality and CFT4 , Nucl. Phys. B 491 (1997) 221, hep-th/9601023.
[36] A. Galperin, E. Ivanov, S. Kalitsyn, V. Ogievetsky, E. Sokatchev, Unconstrained N = 2 matter, YangMills
and supergravity theories in harmonic superspace, Class. Quantum Grav. 1 (1984) 469.
[37] K.G. Chetyrkin, A.L. Kataev, F.V. Tkachov, New approach to evaluation of multiloop Feynman integrals:
the Gegenbauer polynomial X space technique, Nucl. Phys. B 174 (1980) 345;
K.G. Chetyrkin, F.V. Tkachov, Integration by parts: the algorithm to calculate beta functions in 4 loops,
Nucl. Phys. B 192 (1981) 159;
[38]
[39]
[40]
[41]
[42]
[43]
195
D.I. Kazakov, The method of uniqueness, a new powerful technique for multiloop calculations, Phys. Lett.
B 133 (1983) 406;
S.A. Larin, F.V. Tkachov, J.A.M. Vermaseren, The FORM version of MINCER, NIKHEF-H-91-18.
A.A. Vladimirov, Method for computing renormalization group functions in dimensional renormalization
scheme, Theor. Math. Phys. 43 (1980) 417, Teor. Mat. Fiz. 43 (1980) 210 (in Russian);
D.I. Kazakov, O.V. Tarasov, A.A. Vladimirov, Calculation of critical exponents by quantum field theory
methods, Sov. Phys. JETP 50 (1979) 521, Zh. Eksp. Teor. Fiz. 77 (1979) 1035 (in Russian);
O.V. Tarasov, A.A. Vladimirov, Three loop calculations in non-Abelian gauge theories, JINR-E2-80-483.
Y.S. Stanev, private communication.
W. Siegel, Inconsistency of supersymmetric dimensional regularization, Phys. Lett. B 94 (1980) 37.
N. Beisert, M. Bianchi, J.F. Morales, H. Samtleben, Higher spin symmetry and N = 4 SYM, JHEP 0407
(2004) 058, hep-th/0405057.
B. Eden, C. Schubert, E. Sokatchev, Three-loop four-point correlator in N = 4 SYM, Phys. Lett. B 482
(2000) 309, hep-th/0003096;
B. Eden, C. Schubert, E. Sokatchev, Four-point functions of chiral primary operators in N = 4 SYM, hepth/0010005.
A. Galperin, E. Ivanov, V. Ogievetsky, E. Sokatchev, Harmonic Superspace, Cambridge Univ. Press, Cambridge, 2001.
Wightman function and Casimir densities

on AdS bulk with application
to the RandallSundrum braneworld
Aram A. Saharian
Department of Physics, Yerevan State University, 1 Alex Manoogian Str., 375049 Yerevan, Armenia
Received 10 December 2003; received in revised form 15 October 2004; accepted 19 January 2005
Abstract
Positive frequency Wightman function and vacuum expectation value of the energymomentum
tensor are computed for a massive scalar field with general curvature coupling parameter subject to
Robin boundary conditions on two parallel plates located on (D + 1)-dimensional AdS background.
The general case of different Robin coefficients on separate plates is considered. The mode summation method is used with a combination of a variant of the generalized AbelPlana formula for the
series over zeros of combinations of cylinder functions. This allows us to extract manifestly the parts
due to the AdS spacetime without boundaries and boundary induced parts. The asymptotic behavior
of the vacuum densities near the plates and at large distances is investigated. The vacuum forces
acting on the boundaries are presented as a sum of the self-action and interaction forces. The first
one contains well-known surface divergences and needs further regularization. The interaction forces
between the plates are attractive for Dirichlet scalar. We show that there is a region in the space of
parameters defining the boundary conditions in which the interaction forces are repulsive for small
distances and attractive for large distances. An application to the RandallSundrum braneworld with
arbitrary mass terms on the branes is discussed.
PACS: 04.62.+v; 11.10.Kk
E-mail address: saharyan@server.physdep.r.am (A.A. Saharian).

doi:10.1016/j.nuclphysb.2005.01.033
A.A. Saharian / Nuclear Physics B 712 (2005) 196228
197
1. Introduction
Anti-de Sitter (AdS) spacetime is one of the simplest and most interesting spacetimes
allowed by general relativity. Quantum field theory in this background has been discussed
by several authors (see, for instance, Refs. [117]). Much of early interest to AdS spacetime
was motivated by the questions of principle related to the quantization of fields propagating on curved backgrounds. The importance of this theoretical work increased when it was
realized that AdS spacetime emerges as a stable ground state solution in extended supergravity and KaluzaKlein models and in string theories. The appearance of the AdS/CFT
correspondence and braneworld models of RandallSundrum type has revived interest in
this subject considerably. The AdS/CFT correspondence (for a review see [18]) represents
a realization of the holographic principle and relates string theories or supergravity in the
bulk of AdS with a conformal field theory living on its boundary. It has many interesting
formal and physical facets and provides a powerful tool to investigate gauge field theories,
in particular QCD. Recently it has been suggested that the introduction of compactified
extra spatial dimensions may provide a solution to the hierarchy problem between the gravitational and electroweak mass scales [1921]. The main idea to resolve the large hierarchy
is that the small coupling of four-dimensional gravity is generated by the large physical
volume of extra dimensions. These theories provide a novel setting for discussing phenomenological and cosmological issues related to extra dimensions. The model introduced
by Randall and Sundrum is particularly attractive. Their background solution consists of
two parallel flat branes, one with positive tension and another with negative tension embedded in a five-dimensional AdS bulk [20]. The fifth coordinate is compactified on S 1 /Z2 ,
and the branes are on the two fixed points. It is assumed that all matter fields are confined
on the branes and only the gravity propagates freely in the five-dimensional bulk. In this
model, the hierarchy problem is solved if the distance between the branes is about 40 times
the AdS radius and we live on the negative tension brane. More recently, scenarios with
additional bulk fields have been considered [2226].
In the scenario presented in [20] the distance between the branes is associated with the
vacuum expectation value of a massless scalar field, called the radion. This modulus field
has zero potential and consequently the distance is not determined by the dynamics of the
model. For this scenario to be relevant, it is necessary to find a mechanism for generating a potential to stabilize the distance between the branes. Classical stabilization forces
due to the nontrivial background configurations of a scalar field along an extra dimension
were first discussed by Gell-Mann and Zwiebach [27]. With the revived interest in extra
dimensions and braneworlds, as modified version of this mechanism, which exploits a classical force due to a bulk scalar field with different interactions with the branes, received
significant attention [28] (and references therein). Another possibility for the stabilization
mechanism arises due to the vacuum force generated by the quantum fluctuations about
a constant background of a bulk field. The braneworld corresponds to a manifold with
dynamical boundaries and all fields which propagate in the bulk will give Casimir-type contributions to the vacuum energy (for the Casimir effect see Refs. [2934]), and as a result
to the vacuum forces acting on the branes. In dependence of the type of a field and boundary conditions imposed, these forces can either stabilize or destabilize the braneworld. In
addition, the Casimir energy gives a contribution to both the brane and bulk cosmological
198
constants and, hence, has to be taken into account in the self-consistent formulation of the
braneworld dynamics. Motivated by these, the role of quantum effects in braneworld scenarios has received some recent attention. For a conformally coupled scalar this effect was
initially studied in Ref. [35] in the context of M-theory, and subsequently in Refs. [36
49] for a background RandallSundrum geometry (for the related heat kernel expansions
see Refs. [5052]). The models with dS branes are considered as well [48,5355]. For a
conformally coupled bulk scalar the cosmological backreaction of the Casimir energy is
investigated in Refs. [35,48,53,5658].
Investigation of local physical characteristics in the Casimir effect, such as expectation
value of the energymomentum tensor, is of considerable interest. In addition to describing the physical structure of the quantum field at a given point, the energymomentum
tensor acts as the source in the Einstein equations and therefore plays an important role in
modelling a self-consistent dynamics involving the gravitational field. In this paper we will
study the vacuum expectation value of the energymomentum tensor of a scalar field with
arbitrary curvature coupling parameter obeying Robin boundary conditions on two parallel plates in (D + 1)-dimensional AdS spacetime. The general case is considered when
the constants in the Robin boundary conditions are different for separate plates. Robin type
conditions are an extension of Dirichlet and Neumann boundary conditions and appear in a
variety of situations, including the considerations of vacuum effects for a confined charged
scalar field in external fields [59], spinor and gauge field theories, quantum gravity and
supergravity [60,61]. Robin conditions can be made conformally invariant, while purelyNeumann conditions cannot. Thus, Robin-type conditions are needed when one deals with
conformally invariant theories in the presence of boundaries and wishes to preserve this invariance. It is interesting to note that the quantum scalar field satisfying the Robin condition
on the boundary of the cavity violates the Bekensteins entropy-to-energy bound near certain points in the space of the parameter defining the boundary condition [62]. The Robin
boundary conditions are an extension of those imposed on perfectly conducting boundaries and may, in some geometries, be useful for depicting the finite penetration of the field
into the boundary with the skin-depth parameter related to the Robin coefficient. Mixed
boundary conditions naturally arise for scalar and fermion bulk fields in the Randall
Sundrum model [22,43]. To obtain the expectation values of the energymomentum tensor,
we first construct the positive frequency Wightman function. The application of the generalized AbelPlana formula to the corresponding mode sum allows us to extract manifestly
the boundary-free AdS part. The expressions for the boundary induced vacuum expectation
values of the energymomentum tensor are obtained by applying on the subtracted part a
certain second order differential operator and taking the coincidence limit. Note that the
Wightman function is also important in consideration of the response of particle detectors
at a given state of motion (see, for instance, [63]).
We have organized the paper as follows. In Section 2 we consider the vacuum in the
region between two parallel plates. The corresponding positive frequency Wightman function is evaluated by using the generalized AbelPlana summation formula for the series
over zeros of a combination of cylinder functions. This allows us to present the boundary induced part in terms of integrals with exponential convergence for the points away
the boundaries. In Section 3 we consider the vacuum expectation value of the energy
momentum tensor for the case of a single plate. Both regions on the right and on the left
199
from the plate are investigated. Various limiting cases are discussed. Section 4 is devoted
to the vacuum energymomentum tensor for the geometry of two parallel plates. The corresponding vacuum expectation values are presented in the form of the sum of single plates
and interference parts. The latter is finite everywhere including the points on the boundaries. The interaction forces between the plates are discussed as well. In Section 5 we show
that the vacuum expectation value for the energymomentum tensor of a bulk scalar in the
RandallSundrum braneworld scenario is obtained from our results as a special case. The
last section contains a summary of the work.
2. Wightman function
Consider a scalar field (x) on background of a (D + 1)-dimensional plane-symmetric
spacetime with the line element
ds 2 = gik dx i dx k = e2 (y) dx dx dy 2 ,
(2.1)
and with = diag(1, 1, . . . , 1) being the metric for the D-dimensional Minkowski
spacetime. Here and below i, k = 0, 1, . . . , D, and , = 0, 1, . . . , D 1. The corresponding field equation has the form

ik
g i k + m2 + R (x) = 0,
(2.2)
where the symbol i is the operator for the covariant derivative associated with the metric gik , R is the corresponding Ricci scalar, and is the curvature coupling parameter.
For minimally and conformally coupled scalars one has = 0 and = c = (D 1)/4D
correspondingly. Note that by making a coordinate transformation

z = e (y) dy,
(2.3)
metric (2.1) is written in a conformally-flat form ds 2 = e2 ik dx i dx k .
Below we will study quantum vacuum effects brought about by the presence of parallel infinite plane boundaries, located at y = a and y = b, a < b, with mixed boundary
conditions
(A y + B y y )(x) = 0,
y = a, b,
(2.4)
and constant coefficients A y , B y . The presence of boundaries modifies the spectrum for the
zero-point fluctuations of the scalar field under consideration. This leads to the modification of the vacuum expectation values (VEVs) of physical quantities to compared with the
case without boundaries. In particular, vacuum forces arise acting on the boundaries. This
is well-known Casimir effect. As a first stage, in this section we will consider the positive
frequency Wightman function defined as the expectation value
G+ (x, x ) = 0|(x)(x )|0,
(2.5)
where |0 is the amplitude for the vacuum state. In the next section we use this function
to evaluate the VEVs for the energymomentum tensor. Note that the Wightman function
200
also determines the response of particle detectors in a given state of motion. By expanding
the field operator over eigenfunctions and using the commutation relations one can see that

G+ (x, x ) =
(2.6)
(x) (x ),
where denotes a set of quantum numbers, and { (x)} is a complete set of solutions to
the field equation (2.2) satisfying boundary conditions (2.4).
On the base of the plane symmetry of the problem under consideration the corresponding eigenfunctions can be presented in the form

x i = k x fn (y),
(2.7)
where k (x ) are the standard Minkowskian modes on the hyperplane parallel to the plates

ei k x
k x =
(2.8)
, k = (, k), = k 2 + m2n , k = |k|.
2(2)D1
Here the separation constants mn are determined by boundary conditions (2.4) and will
be given below. Substituting eigenfunctions (2.7) into the field equation (2.2) for the function fn (y) one obtains the following equation

D d
D dfn
e
+ m2 + R fn = m2n e2 fn .
e
(2.9)
dy
dy
2 , where
For the AdS geometry one has (y) = kD y, z = e (y) /kD , and R = D(D + 1)kD
the AdS curvature radius is given by 1/kD . In this case the solution to Eq. (2.9) for the
region a < y < b is

fn (y) = cn eD/2 J (mn z) + b Y (mn z) ,

(2.10)
where J (x), Y (x) are the Bessel and Neumann functions, and

2.
= (D/2)2 D(D + 1) + m2 /kD
(2.11)
Here we will assume values of the curvature coupling parameter for which is real. For
imaginary the ground state becomes unstable [4]. Note that for a conformally coupled
massless scalar one has = 1/2 and the cylinder functions in Eq. (2.10) are expressed via
the elementary functions. From the boundary condition on y = a one finds
b =
(a)
J (mn za )
,
(a)
Y (mn za )
zj = e (j ) /kD , j = a, b,
(2.12)
where we use the notation

F (j ) (x) = Aj F (x) + Bj xF (x),
Aj = A j + B j kD D/2, Bj = B j kD , j = a, b
(2.13)
for a given function F (x). Note that for Neumann scalar one has Aj = Bj D/2 and

F (j ) (x) = F (N) (x) x 1D/2 x D/2 F (x) ,
(2.14)
A j = 0.
201
From the boundary condition on the plane y = b we receive that the eigenvalues mn have
to be solutions to the equation
Cab (zb /za , mn za ) J(a) (mn za )Y(b) (mn zb ) Y(a) (mn za )J(b) (mn zb ) = 0.
(2.15)
We denote by z = ,n , n = 1, 2, . . . , the zeros of the function Cab (, z) in the right

half-plane of the complex variable z, arranged in the ascending order, ,n < ,n+1 . The
eigenvalues for mn are related to these zeros as
mn = kD ,n e (a) = ,n /za .
(2.16)
The coefficient cn in Eq. (2.10) is determined from the orthonormality condition

b
dy e(2D) fn (y)fn (y) = nn
(2.17)
and is equal to
cn2 =
2 u (a)2
(u)Tab (, u),
Y
2kD za2
u = ,n , = zb /za ,
where we have introduced the notation

(a)2

1
J (u) 2
Tab (, u) = u (b)2
Ab + Bb2 2 u2 2 A2a + Ba2 u2 2
.
J (u)
(2.18)
(2.19)
Note that, as we consider the quantization in the region between the branes, za z zb ,
the modes defined by (2.10) are normalizable for all real values of from Eq. (2.11).
Substituting the eigenfunctions (2.7) into the mode sum (2.6), for the expectation value
of the field product one finds
0|(x)(x )|0 =
D1
kD
(zz )D/2
D+1
2
D3 za2
dk eik(xx )
h (,n )Tab (, ,n ),
(2.20)
n=1
where x = (x 1 , x 2 , . . . , x D1 ) represents the coordinates in (D 1)-hyperplane parallel to

the plates and
2 2 2
uei u /za +k (t t)
h (u) =
(2.21)
g (u, uz/za )g (u, uz /za ).
u2 /za2 + k 2
Here and below we use the notation
g (u, v) = J (v)Y(a) (u) J(a) (u)Y (v).
(2.22)
To sum over n we will use the summation formula derived in Refs. [64,65] by making use
of the generalized AbelPlana formula [64,66]. For a function h(u) analytic in the right
half-plane Re u > 0 this formula has the form
202
2
h(,n )T (, ,n )
2
n=1
=
0

(1b)
h(u)H (u)
h(x) dx
Res
2 u=0 Cab (, u)H (1a) (u)

J(a)2 (x) + Y(a)2 (x)
(b)
[h(xei/2 ) + h(xei/2 )] dx
K (x)
,
(a)
(a)
(b)
(b)
(a)
K (x) K (x)I (x) K (x)I (x)
(2.23)
where I (u) and K (u) are the Bessel modified functions. Formula (2.23) is valid for
functions h(u) satisfying the conditions
|h(u)| < 1 (x)ec1 |y| ,
|u| , u = x + iy,
(2.24)
and

h uei = h(u) + o u1 ,
u 0,
(2.25)
where c1 < 2( 1), x 2Ba 0 1 1 (x) 0 for x +. Using the asymptotic formulae for
the Bessel functions for large arguments when is fixed (see, e.g., [67]), we can see that for
the function h (u) from Eq. (2.21) the condition (2.24) is satisfied if z + z + |t t | < 2zb .
In particular, this is the case in the coincidence limit t = t for the region under consideration, za < z, z < zb . As for |u| < k one has h (uei ) = h (u), the condition (2.25) is
also satisfied for the function h (u). Note that h (u) u1k0 for u 0 and the residue
term on the right of formula (2.23) vanishes. Applying to the sum over n in Eq. (2.20)
formula (2.23), one obtains
0|(x)(x )|0
k D1 (zz )D/2
= D D D1
2

dk e
ik(xx )
1
za2

0
h (u) du
(a)2
(a)2
J (u) + Y (u)
a (uza , uzb ) (a)

G (uza , uz)G(a)
(uza , uz )
u2 k 2
k

2
2
cosh u k (t t ) ,
du u
(2.26)
where we have introduced notations

)
(j )
(j )
G(j
(u, v) = I (v)K (u) I (u)K (v),
(b)
(a)
K (v)/K (u)
a (u, v) = (a)
(b)
(b)
(a)
K (u)I (v) K (v)I (u)
j = a, b,
(2.27)
(2.28)
(the function with j = b will be used below). Note that we have assumed values of the
coefficients A and B for which all zeros for Eq. (2.15) are real and have omitted the
203
residue terms in the original formula in Refs. [64,65]. In the following we will consider
this case only.
To simplify the first term in the figure braces in Eq. (2.26), let us use the relation
g (u, uz/za )g (u, uz /za )
(a)2
(a)2
J (u) + Y (u)
= J (uz/za )J (uz /za )
2
1 J(a) (u) ()
H (uz/za )H() (uz /za )
(a)
2
H (u)
(2.29)
=1
()
with H (z), = 1, 2, being the Hankel functions. Substituting this into the first integral
in the figure braces of Eq. (2.26) we rotate the integration contour over u by the angle /2
for = 1 and by the angle /2 for = 2. Under the condition z + z |t t | > 2za , the
integrals over the arcs of the circle with large radius vanish. The integrals over (0, ikza )
and (0, ikza ) cancel out and after introducing the Bessel modified functions one obtains

0
h (u) du
(a)2
(a)2
J (u) + Y (u)

i u2 +k 2 (t t)
e
= za2 du u
J (uz)J (uz )
x 2 + k2
0
2z2
a

du u
k
(a)

I (uza ) K (uz)K (uz )

cosh u2 k 2 (t t ) .
(a)
u2 k 2
K (uza )
(2.30)
Substituting this into formula (2.26), the Wightman function can be presented in the form
0|(x)(x )|0

(a)
= 0S |(x)(x )|0S + (x)(x )
k D1 (zz )D/2
D D1 D
2

dk e
ik(xx )

du u
k
(a)

G(a)
(uza , uz)G (uza , uz ) cosh
a (uza , uzb )
u2 k 2
u2 k 2 (t t ) .
(2.31)
Here the term

0S |(x)(x )|0S
k D1 (zz )D/2
= D D D1
2

dk e
ik(xx )

0
ei
u2 +k 2 (t t)
du u
u2 + k 2
J (uz)J (uz )
(2.32)
does not depend on the boundary conditions and is the Wightman function for the AdS
space without boundaries. The second term on the right of Eq. (2.31),
204
(a)
(x)(x )

(a)
D1
kD
(zz )D/2
I (uza ) K (uz)K (uz )
ik(xx )
=
du
u
dk
e
(a)
2D1 D
u2 k 2
K (uza )
k

cosh u2 k 2 (t t ) ,
(2.33)
does not depend on the parameters of the boundary at z = zb and is induced in the region
z > za by a single boundary at z = za when the boundary z = zb is absent.
Note that expression (2.32) for the boundary-free Wightman function can also be written
in the form
0S |(x)(x
D1
)|0S = kD
(zz )D/2

dm mG+
MD (x , x ; m)J (mz)J (mz ),
(2.34)

where G+
MD (x , x ; m) is the Wightman function for a scalar field with mass m in the Ddimensional Minkowski spacetime MD . The right-hand side in Eq. (2.34) can be further
simplified by using the expression

mD/21 KD/21 [m (x x )2 (t t i)2 ]
+
GMD x , x ; m =
(2.35)
,
(2)D/2 [(x x )2 (t t i)2 ](D2)/4
with > 0. Substituting this into formula (2.34) and making use of the integration formula
from [68], one obtains
0S |(x)(x )|0S
D1

(1D)/4 (D1)/2
kD
e(D1)i/2 v 2 1
Q1/2 (v)
(D+1)/2
(2)

k D1 ( + D/2)v D/2
D + 2 + 2 D + 2
1
,
,
;
+
1;
= DD/2++1 D/2
F
2 1
4
4
2
( + 1)
v2
(2.36)
where Q (v) is the associated Legendre function of the second kind, 2 F1 (a, b; c; u) is the
hypergeometric function (see, for instance, [67]), and
v=1+
(z z )2 + (x x )2 (t t i)2
.
2zz
(2.37)
It can be checked that in the limit kD 0 from (2.32) the expression for the
(D + 1)-dimensional Minkowski Wightman function is obtained. To see this note that
in this limit m/kD , kD z 1 + kD y and, hence, as it follows from (2.37), one
2 w 2 /2, where w 2 = (y y )2 + (x x )2 (t t i)2 . Using the intehas v 1 + kD
gral representations for the functions Q (v) and K (v) (formulae (8.8.2) and (9.6.23) in
Ref. [67]), it can be seen that
2 i 2
m
e
(v 1)/2 Q1/2 (v) = K (mw).
lim kD
kD 0
w
(2.38)
205
For = (D 1)/2 the expression on the right of this formula coincides with the
(D + 1)-dimensional Minkowskian Wightman function up to the coefficient (2)(D+1)/2 .
Hence, in combination with the first formula in (2.36), relation (2.38) proves our statement.
Alternatively, we can use formula (2.32), replacing the Bessel functions by their asymptotic
expressions for large values of the order.
In the coincidence limit x = x expression (2.36) is finite for D < 1 and for the VEV of
the field square one obtains
0S | 2 (x)|0S =
D1
( 1D
kD
2 )(D/2 + )
.
(D+1)/2
(1 + D/2)
(4)
(2.39)
This quantity is independent on the spacetime point, which is a direct consequence of the
maximal symmetry of the AdS bulk. Formula (2.39) coincides with the result of Refs. [10,
14] (a typo of [10] is corrected in [14]) obtained from the Feynman propagator in the coincidence limit (for the zeta function based calculations see [12,16]). The expression on
the right of Eq. (2.39) is analytic in the complex D-plane apart from simple poles coming from the gamma function in the nominator. Hence, it can be extended throughout the
whole complex plane. For even D this expression is finite and according to the dimensional regularization procedure [63] can be taken as the regularized value for the field
square. Formula (2.39) can be also obtained directly from (2.32) firstly integrating over k
and by making use of the integration formula

I (D)
0
( 1D
2 )(D/2 + )
dx x D1 J2 (x) =
.
2 (1 D/2)(1 + D/2)
(2.40)
By using the identity

(a)
I (uza )
(a)

K (uz)K (uz ) + a (uza , uzb )G(a)
(uza , uz)G (uza , uz )
(a)
K (uza )
K (b) (uzb )
(b)

I (uz)I (uz ) + b (uza , uzb )G(b)
(uzb , uz)G (uzb , uz ), (2.41)
(b)
I (uzb )
with
b (u, v) =
(a)
(b)
I (u)/I (v)
,
K (a) (u)I(b) (v) K (b) (v)I(a) (u)
(2.42)
it can be seen that the Wightman function in the region za z zb can also be presented
in the equivalent form
0|(x)(x )|0

(b)
= 0S |(x)(x )|0S + (x)(x )

D1
(zz )D/2
kD
b (uza , uzb )
ik(xx )
du u
dk e
D1
D
2
u2 k 2
k

(b)

G(b)
u2 k 2 (t t ) ,
(uzb , uz)G (uzb , uz ) cosh
(2.43)
206
where
(b)

(x)(x )

(b)
D1
kD
(zz )D/2
K (uzb ) I (uz)I (uz )
ik(xx )
=
du
u
dk
e
(b)
2D1 D
u2 k 2
I (uzb )
k

cosh u2 k 2 (t t )
(2.44)
is the boundary part induced in the region z < zb by a single plate at z = zb when the plate
z = za is absent. Note that in the formulae given above the integration over angular part
can be done by using the formula

dk e
ik(xx )

F (k) = (2)
(D1)/2
dk k D2 F (k)
0
J(D3)/2 (k|x x |)
,
(k|x x |)(D3)/2
(2.45)
for a given function F (k). Combining two forms, formulae (2.31) and (2.43), we see that
the expressions for the Wightman function in the region za z zb is symmetric under
the interchange a b and I K . Note that the expression for the Wightman function
is not symmetric with respect to the interchange of the plate indices. The reason for this is
that, though the background AdS spacetime is homogeneous, the boundaries have nonzero
extrinsic curvature tensors and two sides of the boundaries are not equivalent. In particular,
for the geometry of a single brane the VEVs differ for the regions on the left and on the
right of the brane. Here the situation is similar to that for the case of a spherical shell on
background of the Minkowski spacetime.
3. Casimir densities for a single plate

In this section we will consider the VEV of the energymomentum tensor for a scalar
field in the case of a single plate located at z = za . As it has been shown in the previous
section the Wightman function for this geometry is presented in the form

(a)
0|(x)(x )|0 = 0S |(x)(x )|0S + (x)(x ) ,
(3.1)
where |0 is the amplitude for the corresponding vacuum state. The boundary induced part
(x)(x )(a) is given by formula (2.33) in the region z > za and by formula (2.44) with
replacement zb za in the region z < za . For points away from the plate this part is finite
in the coincidence limit and in the corresponding formulae for theWightman function we
can directly put x = x . Introducing a new integration variable v = u2 k 2 , transforming
to the polar coordinates in the plane (v, k) and integrating over angular part, the following
formula can be derived

( D1
uf (u)
D2
2 )
dk k
du
=
du uD1 f (u).
(3.2)
2(D/2)
u2 k 2
0
207
By using this formula and Eq. (2.33), the boundary induced VEV for the field square in the
region z > za is presented in the form

(a)
(x)
=
2
D1 D
z
kD
D1
D/2
2
(D/2)

du uD1
0
I(a) (uza ) 2
K (uz),
(a)
K (uza )
z > za .
(3.3)
The corresponding formula in the region z < za is obtained from Eq. (2.44) by a similar
way and differs from Eq. (3.3) by replacements I K .
By using the field equation, the expression for the metric energymomentum tensor of
a scalar field can be presented in the form

1
l
Tik = i k +
(3.4)
gik l i k Rik 2 .
4
By virtue of this, for the VEV of the energymomentum tensor we have
i k 0|(x)(x )|0
0|Tik (x)|0 = lim
x x

1
gik l l i k Rik 0| 2 (x)|0.
+
4
(3.5)
This corresponds to the point-splitting regularization technique. VEV (3.5) can be evaluated by substituting expression (3.1) for the positive frequency Wightman function and
VEV of the field square into Eq. (3.5). First of all we will consider the region z > za . The
vacuum energymomentum tensor is diagonal and can be presented in the form
(a)
0|Tik |0 = 0S |Tik |0S + Tik ,
(3.6)
where
0S |Tik |0S =
D+1 D k
kD
z i
2D1 (D1)/2 ( D1
2 )

dk k
D2
f(i) [J (uz)]
du u
u2 + k 2
(3.7)
is the VEV for the energymomentum tensor in the AdS background without boundaries,
and the term
k (a)
=
Ti
D+1 D k
z i
kD
2D2 (D+1)/2 ( D1
2 )
du u
k

dk k D2
0
(a)
I (uza ) F (i) [K (uz)]
u2 k 2
K (a) (uza )
(3.8)
is induced by a single boundary at z = za . For a given function g(v) the functions

F (i) [g(v)] in formula (3.8) are defined as
208
F (0) g(v) =

1
4
2 v 2 g 2 (v) + D +
vg(v)g (v)
2
4 1

+ v 2 + 2 g 2 (v) + z2 k 2 v 2 g 2 (v)

(3.9)

Dz2 k 2 2
(i)
(0)
2
F g(v) = F
(3.10)
g(v) + v
g (v), i = 1, . . . , D 1,
D1

v2
D
F (D) g(v) = g 2 (v) + (4 1)vg(v)g (v)
2
2
1 2
2
+ v + + 2 D(D + 1) D 2 /2 g 2 (v),
(3.11)
2
and the expressions for the functions f(i) [g(v)] are obtained from those for F (i) [g(v)] by
the replacement v iv. Note that the boundary-induced part (3.8) is finite for the points
away the brane and, hence, the renormalization procedure is needed for the boundary-free
part only. The latter is well-investigated in literature.
Using formula (3.2), it can be seen that the contribution of the second term on the right
of Eq. (3.10) to the boundary part of the VEVs vanishes. From Eq. (3.8) now one obtains

(a)
Tik
k D+1 zD k
= D1D D/2 i
2
(D/2)

du uD1
0
(a)
I (uza ) (i)
F K (uz) ,
(a)
K (uza )
z > za ,
(3.12)
where the expressions for the functions F (i) [g(v)] directly follow from Eqs. (3.9)(3.11)
after the integration using formula (3.2)

1
4
(i)
2 2
2 v g (v) + D +
vg(v)g (v)
F g(v) =
2
4 1

2v 2
2
2
2
+ +v +
(3.13)
g (v) , i = 0, 1, . . . , D 1,
D(4 1)

F (D) g(v) = F (D) g(v) .

(3.14)
By a similar way for the VEVs induced by a single brane in the region z < za , by making
use of expression (2.44) (with replacement zb za ), one obtains

(a)
Tik
k D+1 zD k
= D1D D/2 i
2
(D/2)

du uD1
0
(a)
K (uza ) (i)
F I (uz) ,
(a)
I (uza )
z < za .
(3.15)
Note that VEVs (3.12), (3.15) depend only on the ratio z/za which is related to the proper
distance from the plate by equation
z/za = ekD (ya) .
(3.16)
As we see, for the part of the energymomentum tensor corresponding to the coordinates
in the hyperplane parallel to the plates one has T (a) . Of course, we could expect
209
this result from the problem symmetry. It can be seen that the VEVs obtained above obey
k = 0 which for the AdS metric takes the form
the continuity equation Ti;k
D D
z TD + DT00 = 0.
(3.17)
z
In the AdS part without boundary the integration over k can be done using the formula
zD+1

0

D
k D2 dk
uD2
D1
1
.
=
2
2
2
u2 + k 2
(3.18)
As for the boundary terms, the contribution corresponding to the second term on the right
of Eq. (3.10) vanishes and one obtains
0S |Tik |0S =

D+1 D k

z i
kD
D
1
du uD1 f (i) J (uz) ,
D
D/2
2
2
(3.19)
f (i) [g(v)]
where the expressions for the functions

are obtained from those for F (i) [g(v)],
Eqs. (3.13), (3.14), by the replacement v iv. Now from Eq. (3.13) we have the following
relation between the vacuum energy density and pressures in the parallel directions (no
summation over i)
0|T00 |0 = 0|Tii |0,
i = 1, . . . , D 1.
To evaluate the u-integral in (3.19) we use formula (2.40) and the relations

D 2 DI (D)
,
I (D + 2) = 2
4
D+1

D DI (D)
D
I1 (D + 2) = 1
,
2
2
D+1

D
I (D).
dx x D J (x)J1 (x) =
2
(3.20)
(3.21)
(3.22)
This yields

0
m2 I (D)
,
dx x D1 f (i) J (x) = 2
kD D + 1
i = 0, 1, . . . , D.
(3.23)
As a result for the boundary-free AdS VEVs one receives

0S |Tik |0S = ik
D1 2
( 1D
kD
m
2 )(D/2 + )
.
(D+1)/2
(D + 1)(1 + D/2)
(4)
(3.24)
Hence, the AdS VEVs are proportional to the corresponding metric tensor. Again, we could
expect this result due to the maximally symmetry of the AdS background. Formula (3.24)
210
can be obtained also directly from (2.39), by using the standard relation between the field
square and the trace of the energymomentum tensor.
For a conformally coupled massless scalar = 1/2, and by making use of the expressions for the modified Bessel functions, it can be seen that Tik (a) = 0 in the region z > za
and
(a)
TDD
= D
(a)
T00
(kD z/za )D+1

=
(4)D/2 (D/2)

0
t D dt
Ba (t1)+2Aa t
Ba (t+1)2Aa e
+1
(3.25)
in the region z < za . Note that the corresponding energymomentum tensor for a single
Robin plate in the Minkowski bulk vanishes [69] and the result for the region z > za is
obtained by a simple conformal transformation from that for the Minkowski case. In the
region z < za this procedure does not work as in the AdS problem one has 0 < z < za
instead of < z < za in the Minkowski problem and, hence, the part of AdS under
consideration is not a conformal image of the corresponding manifold in the Minkowski
spacetime (for a general discussion of this question in conformally related problems see
Ref. [70]).
The boundary induced VEVs given by Eqs. (3.12) and (3.15), in general, cannot be further simplified and need numerical calculations. Relatively simple analytic formulae for
the brane-induced parts can be obtained in limiting cases. First of all, as a partial check,
in the limit kD 0 the corresponding formulae for a single plate on the Minkowski bulk
are obtained (see Ref. [69] for the massless case and Ref. [71] for the massive case). This
can be seen noting that in this limit m/kD is large and by introducing the new integration variable in accordance with u = y, we can replace the Bessel modified functions by
their uniform asymptotic expansions for large values of the order. The Minkowski result is
obtained in the leading order.
In the limit z za for a fixed kD expressions (3.12) and (3.15) diverge. In accordance
with (3.16), this corresponds to small proper distances from the brane. The surface divergences in the VEVs of the energymomentum tensor are well-known in quantum field
theory with boundaries and are investigated for various types of boundary geometries and
boundary conditions (see, for instance, Refs. [72,73]). Near the brane the main contribution into the integral over u in Eqs. (3.12), (3.15) comes from the large values of u and we
can replace the Bessel modified functions by their asymptotic expressions for large values
of the argument when the order is fixed (see, for instance, [67]). To the leading order this
yields

D+1
0 (a)
( c )(Ba )
DkD
D+1
T0
(3.26)

,
D
(D+1)/2
2
2
|1 za /z|D+1

D (a) 0 (a)
za
1
,
T0
(3.27)
TD
z
where we use the notation

1
if x = 0,
(x) =
1 if x = 0.
(3.28)
211
Note that the leading terms for the components with i = 0, 1, . . . , D 1 are symmetric
with respect to the plate, and the D
D -component has opposite signs for the different sides
of the plate. Near the brane the vacuum energy densities have opposite signs for the cases
of Dirichlet (Ba = 0) and non-Dirichlet (Ba = 0) boundary conditions. Recall that for a
conformally coupled massless scalar the vacuum energymomentum tensor vanishes in
the region z > za and is given by expression (3.25) in the region z < za . The latter is finite
everywhere including the points on the plate.
For the points with the proper distances from the plate much larger compared with
the AdS curvature radius one has z za . This limit is important from the point of view
of the application to the RandallSundrum braneworld. Introducing in Eq. (3.12) a new
integration variable y = zu, by making use of formulae for the Bessel modified functions
for small values of the argument, and assuming Aa Ba = 0, to the leading order we
receive

k (a)
k D+1 ik 22D D/2 Aa + Ba za 2
D
Ti
(D/2)()( + 1) Aa Ba 2z

dy y D+21 F (i) K (y) .

(3.29)
0
The integral in this formula can be evaluated on the base of formulae

dx x
K2 (x) =
dx x
K 2 (x) =

23

+
2
,
()
2
2
2
(3.30)

0

2
2
2

dx x 1 K2 (x).
2
+1 4
(3.31)
This yields

T00
(a)

D+1
kD
D + 2
Aa + Ba za 2
4
D + 2 + 1
(4)(D1)/2 Aa Ba 2z
(2 1)(D/2 + + 1)(D/2 + 2)
,
( + 1)(D/2 + 1/2 + )
(3.32)
D 0 (a)
T
(3.33)
, z za ,
D + 2 0
and the boundary-induced VEVs are exponentially suppressed by the factor exp[2kD
(a y)]. In the limit under consideration the ratio of the energy density and the perpendicular pressure is a negative constant. We see that in the case c , at large distances from the
plate the energy density is positive for |Ba /Aa | > 1/ and is negative for |Ba /Aa | < 1/.
Combining this with the asymptotic behavior near the plate, Eq. (3.26), we conclude that
for Ba = 0 and |Ba /Aa | < 1/ the energy density is positive near the plate and is negative
at large distances approaching to zero and, hence, for some value of z it has a minimum
with the negative energy density.

TDD
(a)
212
Now we turn to the limit z za for a fixed kD . This corresponds to the points near the
AdS boundary presented by the hyperplane z = 0 with the proper distances from the brane
much larger compared with the AdS curvature radius. Introducing in Eq. (3.15) a new
integration variable y = uza and by making use of the formulae for the Bessel modified
functions for small values of the argument, to the leading order one receives

0 (a)
k D+1
z D+2 D + 2 4 (D + 2 + 1)
D/2D
T0
()( + 1)
(D/2) 2za

(a)
K (x)
dx x D+21 (a)
(3.34)
,
I (x)
0
D 0 (a)
T
, z za .
(3.35)
2 0
In this limit the exponential suppression takes place by the factor exp[(D + 2)kD (y a)].
The ratio of the energy density and the perpendicular pressure is a positive constant. Due
to the well-known properties of the Bessel modified functions, the integral in Eq. (3.34)
is positive for small values of the ratio Ba /Aa and is negative for large values and, hence,
as in the previous case there is a possibility for the change of the sign for the energy
density as a function on the distance from the boundary. Now by taking into account relation (3.16), from Eqs. (3.32), (3.34) we conclude that for large proper distances from the
plate to compared with the AdS curvature radius the boundary induced parts are exponentially suppressed.
In the large mass limit, m kD , from (2.11) one has m/kD 1. To find the leading terms of the corresponding asymptotic expansions for the vacuum energymomentum
tensor components we replace the integration variable in Eqs. (3.12) and (3.15), x = t,
and use the uniform asymptotic expansions for the Bessel modified functions for large values of the order. The main contribution to the integrals over t comes from small values of t.
For a fixed z to the leading order one receives

0 (a)
k D+1 m D/2+1 e2| ln(za /z)|m/kD
(4 1)(Ba ) DD/2
,
T0
(3.36)
2kD
|1 za2 /z2 |D/2

D (a) DkD 0 (a)
TD
(3.37)
T
, m kD ,
2m 0
and we have an exponential suppression of the vacuum expectation values for large values
of the mass.
For large values of the parameter kD and fixed proper distances from the plate,
kD |y a| 1, from relation (3.16) one has za z or za z. Hence, in this limit, which
corresponds to strong gravitational fields, we have an asymptotic behavior described by
Eqs. (3.32), (3.33) in the region z > za and by Eqs. (3.34), (3.35) in the region z < za with
the exponential suppression of the boundary induced parts by the factor exp[2kD (a y)]
in the first case and by the factor exp[(2 + D)kD (y a)] in the second case.
In Figs. 1 and 2 we have plotted the vacuum energy density [curves (a)] and D
D -stress
[curves (b)] induced by a single plate, as functions on kD (y a) for a minimally coupled
massless scalar field in D = 3. In Fig. 1 the cases of Dirichlet (top panel) and Neumann

D (a)
TD
213
D+1
D+1
Fig. 1. Boundary induced vacuum densities T00 (a) /kD
and TDD (a) /kD
as functions on kD (y a) for
a minimally coupled massless scalar in D = 3. The curves a correspond to the energy density and the curves b
correspond to the D
D -component of the vacuum stress. The top panel is for Dirichlet scalar and the bottom one is
for Neumann scalar.
214
Fig. 2. The same as in Fig. 1 for Robin scalar with Ba /Aa = 1/4.
(bottom panel) boundary conditions are presented. The energy density is negative everywhere for Dirichlet scalar and is positive for Neumann scalar. It can be seen that for both
Dirichlet and Neumann cases the boundary-induced energymomentum tensors violate the
strong energy condition. The weak energy condition is violated by Dirichlet scalar and is
satisfied by Neumann scalar. The energy conditions play a key role in the singularity theorems of general relativity and these properties may have important consequences in the
consideration of the gravitational back reaction of vacuum quantum effects. Fig. 2 corresponds to the Robin boundary condition with Ba /Aa = 1/4 and illustrates the possibility
for the change of the sign for the energy density as a function on distance from the plate: the
energy density is positive near the plate and is negative for large distances from the plate.
Note that the ratio TDD (a) /T00 (a) is a coordinate dependent function. In accordance with
the asymptotic estimates given above, this ratio tends to zero for the points near the brane
and to the constant values D/(D + 2) and D/(2) in the limits z za and z za ,
respectively.
4. Two-plates geometry
4.1. Vacuum densities
In this section we will investigate the VEVs for the energymomentum tensor in the region between two parallel plates. Substituting the corresponding Wightman function from
Eq. (2.31) into the mode sum formula (3.5), we obtain
0|Tik |0 = 0S |Tik |0S +

du u
k
(a)
Tik
D+1 D k
kD
z i
2D2 (D+1)/2 ( D1
2 )
a (uza , uzb ) (i) (a)

F G (uza , uz) ,
2
2
u k
215

dk k D2
0
za < z < zb ,
(4.1)
where for a given function g(v) the functions F (i) [g(v)] are defined in accordance with
Eqs. (3.9)(3.11). By using formula (3.2), we see that the contribution of the second term
on the right of Eq. (3.10) vanishes and from Eq. (4.1) one obtains
(a)
0|Tik |0 = 0S |Tik |0S + Tik

D+1 D k

kD
z i
du uD1 a (uza , uzb )F (i) G(a)

(uza , uz) , (4.2)
D
2D1 2 ( D2 )
0
with the functions F (i) [g(v)] from Eqs. (3.13), (3.14).

By making use of the Wightman function in form (2.43), we obtain an alternative representation of the VEVs:
(b)
0|Tik |0 = 0S |Tik |0S + Tik

D+1 D k

z i
kD
du uD1 b (uza , uzb )F (i) G(b)

(uzb , uz) . (4.3)
D
2D1 2 ( D2 )
0
This form is obtained from (4.2) by the replacements a b, I K . The vacuum

energymomentum tensor in the region between the branes is not symmetric under the
interchange of indices of the branes. As it has been mentioned above, the reason for this
is that, though the background spacetime is homogeneous, due to the nonzero extrinsic
curvature tensors for the branes, the regions on the left and on the right of the brane are not
equivalent. By the same way as for the case of a single plate, it can be seen that in the limit
kD 0 from formulae (4.2) and (4.3) the corresponding results for two plates geometry
in the Minkowski bulk (see Refs. [69,71]) are obtained.
On the base of formula (4.2), the VEVs in the region za < z < zb can be written in the
form
(a) k (b)

0|Tik |0 = 0S |Tik |0S + Tik
(4.4)
+ Ti
+ Tik ,
with the interference term

Tik =
D+1 D k
kD
z i
2D1 D/2 (D/2)

du uD1
0
K (b) (uzb ) (i)
F
I
a (uza , uzb )F (i) G(a)
(uz
,
uz)
(uz)
. (4.5)
a
I(b) (uzb )
216
Another form we can obtain by using formulae (3.12) and (4.3). Note that the surface
divergences are contained in the second and third terms on the right of Eq. (4.4) and
expression (4.5) is finite for all values za z zb (for large values u the subintegrand
behaves as uD e2u(za zb ) ). For the proper distances between the plates much less than the
AdS curvature radius, kD (b a) 1, one has (1 za /zb ) 1 and the main contribution
into the integral in Eq. (4.5) comes from large values of u. Replacing the Bessel modified
functions by their asymptotic expansions for large values of the argument, for the VEVs
from Eq. (4.5) to the leading order one receives
D1 [1 (B )(B )]

D( D+1
a
b
2 )R (D + 1) 1 2
,
TDD
(4.6)
(Ba )(Bb )
(4)(D+1)/2 (b a)D+1

0
t D dt
1 D
( c )
T0 TD 2D1 D/2
D+1
D
(Ba )(Bb )et 1
2
(D/2)(b a)
0

za z
z zb
+ (Bb ) exp t
,
(Ba ) exp t
(4.7)
zb za
zb za
where R (z) is the Riemann zeta function and the function (x) is defined by Eq. (3.28).
The leading terms given by formulae (4.6) and (4.7) are the same as for the corresponding
quantities on Minkowski spacetime background. In particular, the interference part of the
D -component does not depend on the curvature coupling parameter. In the limit of large
D
distances between the plates, when kD (b a) 1, introducing a new integration variable
v = uzb and expanding over za /zb , to the leading order from Eq. (4.5) one obtains
2

za
z
,
g (i)
Tik ik
(4.8)
zb
zb
where the form of the function g (i) (v) can be found from Eq. (4.5). For v 1 one has
g(v) v D and from (4.8) it follows that the interference part of the vacuum energy
momentum tensor is exponentially suppressed for large interbrane distances.
4.2. Interaction forces
Now we turn to the interaction forces between the plates. The vacuum force acting per
unit surface of the plate at z = zj is determined by the D
D -component of the vacuum energy
momentum tensor at this point. The corresponding effective pressures can be presented as
a sum of two terms:
(j )
(j )
p (j ) = p1 + p(int) ,
j = a, b.
(4.9)
The first term on the right is the pressure for a single plate at z = zj when the second plate
is absent. This term is divergent due to the surface divergences in the vacuum expectation
values. The second term on the right of Eq. (4.9),
(j )

(j )
p(int) = TDD 1 TDD , z = zj , j, j1 = a, b, j1 = j,
(4.10)
is the pressure induced by the presence of the second plate, and can be termed as an interaction force. It is determined by the last terms on the right of formulae (4.2) and (4.3) for
217
the plate at z = za and z = zb , respectively. Substituting into these terms z = zj and using
relations
)
G(j
(u, u) = Bj ,
)
G(j
(u, u) = Aj /u,
(4.11)
one has
(j )
p(int)
D+1
kD
2D D/2 ( D2 )

dx x D1 j (xza /zj , xzb /zj )
0
2

2
Bj D(4 1)Aj Bj A2j .
x 2 2 + 2m2 /kD
(4.12)
Note that due to the asymmetry in the VEV of the energymomentum tensor, the interaction forces acting on the branes are not symmetric under the interchange of the brane
indices. By taking into account that K (u)I (v) K (v)I (u) > 0 and K (u)I (v)
K (v)I (u) < 0 for u < v, it can be easily seen that the vacuum effective pressures are negative for Dirichlet scalar and for a scalar with Aa = Ab = 0 and, hence, the corresponding
interaction forces are attractive for all values of the interplate distance. In Fig. 3 these forces
are plotted as functions on za /zb for massless minimally coupled Dirichlet and Neumann
scalars. As it has been shown above, in the case of a conformally coupled massless scalar
(a)
(b)
field for the first term on the right of Eq. (4.9) one has p1 = 0 and p1 is determined from
Eq. (3.25) with the replacement a b (recall that we consider the region za z zb ). It
can be seen that the corresponding formulae for p (j ) obtained from Eqs. (4.9), (4.12) coincide with those given in Ref. [47]. Note that in this case p (a) /zaD+1 = p (b) /zbD+1 . Using
the Wronskian for the Bessel modified functions, it can be seen that

2 2 2
ln1
Bj x zj + 2 A2j j (xza , xzb ) = nj zj
zj

(a)
(b)
I (xza )K (xzb )
,
(b)
(a)
I (xzb )K (xza )
j = a, b,
(4.13)
where na = 1, nb = 1. This allows us to write the expressions (4.12) for the interaction
forces per unit surface in another equivalent form
(j )
p(int)
D+1 D+1
zj
n j kD
2D D/2 ( D2 )
dx x
D1

2 Bj + (1 4 )A j
1 + DBj 2
Bj (x 2 zj2 + 2 ) A2j

(a)
(b)

I (xza )K (xzb )
ln 1 (b)
.
zj
I (xzb )K (a) (xza )
(4.14)
Note that these forces in general are different for j = a and j = b. For Dirichlet scalar the
second term in the square brackets is zero.
Let us consider the limiting cases for the interaction forces described by Eq. (4.12). For
small distances to compared with the AdS curvature radius, kD (b a) 1, the leading
218
(j )
D+1
Fig. 3. The vacuum effective pressures (zb /zj )D p(int) /kD
, j = a, b, as functions on za /zb for a massless
minimally coupled scalar field in D = 3 determining the interaction forces between Dirichlet (top panel) and
Neumann (bottom panel) plates. Curves a and b correspond to the pressures on the plates z = zj , j = a, b,
respectively.
219
terms are the same as for the plates in the Minkowski bulk
p(int) = D(1 2D )
(j )
( D+1
2 )R (D + 1)
,
(D+1)/2
(4)
(b a)D+1
(4.15)
in the case of Dirichlet boundary condition on one plate and non-Dirichlet boundary condition on the another, and
(j )
p(int) = D
( D+1
2 )R (D + 1)
,
(D+1)/2
(4)
(b a)D+1
(4.16)
for all other cases. Note that in the first case the interaction forces are repulsive and are
attractive for the second case. For large distances between the plates, kD (b a) 1 (this
limit is realized in the RandallSundrum model), by using the expressions for the modified
Bessel functions for small values of the argument, one finds

D+1
2 2 )B 2 D(4 1)A B A2 ]
[(2m2 /kD
za D+2 4kD
a a
a
a
2zb
D/2 (D/2) 2 ()(Ba Aa )2

(b)
K (x)
,
dx x 2+D1 (b)
I (x)

(a)
p(int)
(4.17)

(b)
p(int)
za
2zb
2
D+1
kD
Aa + Ba
Aa Ba 2D1 D/2 (D/2) 2 ()

dx
0
x 2+D1
(b)2
I (x)
2

2
Bb D(4 1)Ab Bb A2b .
x 2 2 + 2m2 /kD
(4.18)
In dependence of the values for the coefficients in the boundary conditions, these effective
pressures can be either positive or negative, leading to repulsive or attractive forces at
large distances. To illustrate these possibilities, in Fig. 4 we have plotted the effective
pressures as functions on za /zb for Robin boundary conditions. In the case of the top
panel, the vacuum interaction forces are attractive for small distances and repulsive for
large distances. For the bottom panel we have an opposite situation, the interaction forces
are attractive for large distances and repulsive for small distances. For the latter case the
vacuum interaction forces provide a possibility for a stabilization of the interplate distance.
The dependence of the vacuum effective pressures on the ratio za /zb for various values of
the Robin coefficient Bb is given in Figs. 5 and 6.
5. Application to the RandallSundrum braneworld

In this section we will consider the application of the results obtained in the previous sections to the RandallSundrum braneworld model [20] based on the AdS geometry
with one extra dimension. The fifth dimension y is compactified on an orbifold, S 1 /Z2 of
length L, with L y L. The orbifold fixed points at y = 0 and y = L are the locations
220
Fig. 4. The same as in Fig. 3 for mixed boundary conditions. Top panel corresponds to A a = 0, Ba = 1, A b = 1,
Bb = 0.2 and bottom panel corresponds to A a = 1, Ba = 0, A b = 1, Bb = 0.2.
of two 3-branes. Below we will allow these submanifolds to have an arbitrary dimension D.
The metric in the RandallSundrum model has the form (2.1) with
(y) = kD |y|.
(5.1)
(j )
221
D+1
Fig. 5. The D = 3 vacuum effective pressures (zb /zj )D p(int) /kD
, j = a, b, for a massless minimally coupled
scalar field as functions on za /zb and Bb . The values for the other Robin coefficients are A a = 1, Ba = 0, and
A b = 1. Top panel corresponds to j = a and bottom panel corresponds to j = b.
222
Fig. 6. The same as in Fig. 5 for the values of Robin coefficients A a = 0, Ba = 1, and A b = 1.
For the corresponding Ricci scalar one has

2
R = 4DkD (y) (y L) D(D + 1)kD
.
(5.2)
223
Let us consider a bulk scalar with the curvature coupling parameter and with the action
functional

1
S=
d D x dy |g| g ik i k m2 + c1 (y) + c2 (y L) + R 2 , (5.3)
2
where m is the bulk mass, c1 and c2 are the brane mass terms on the branes y = 0 and
y = L, respectively. The field equation obtained from (5.3) is

i i + m2 + c1 (y) + c2 (y L) + R = 0.
(5.4)
The corresponding eigenfunctions can be written in form (2.7) and (2.8), where fn (y) is a
solution to the equation

D d
D dfn
e
+ m2 + (c1 + 4D kD )(y) + (c2 4D kD )(y L)
e
dy
dy
2
D(D + 1) kD
(5.5)
fn (y) = m2n e2 fn (y).
To obtain the boundary conditions for the function fn (y) we integrate (5.5) first about
y = 0 and then about y = L. Assuming that the function fn (y) is continuous at these
points, we receive

j

dfn (y) y=yj (1)

lim
= cj (1)j 4D kD fn (yj ),

0 dy
y=yj +(1)j
j = 1, 2, y1 = 0, y2 = L,
(5.6)
with > 0. First consider the case of the untwisted scalar field for which fn (y) = fn (y),
and the solution to Eq. (5.5) for y = 0, L is given by expression (2.10) with z = ekD |y| .
The boundary conditions which follow from relations (5.6) are in form (2.4) with (see also
Ref. [43] for the case c1 = c2 = 0)
A a
1
= (c1 + 4D kD ),
2
B a
A b
1
= (c2 + 4D kD ),
2
B b
(5.7)
and respectively
Aa
1
= D(1 4 ) c1 /kD ,
Ba
2
Ab
1
= D(1 4 ) + c2 /kD .
Bb
2
(5.8)
For the twisted scalar fn (y) = fn (y) and from the Z2 identification on S 1 one has
fn (0) = fn (L) = 0 and, hence, B a = B b = 0, which correspond to Dirichlet boundary
conditions on both branes. The normalization condition for the eigenfunctions fn (y) now
has the form
L
dy e(2D) (y) fn (y)fn (y) = nn .
(5.9)
As a result the normalization coefficient cn differs from (2.15) by additional factor 1/2.
Hence, the Wightman function in the RandallSundrum braneworld is given by formulae (2.31) or equivalently by (2.43) with an additional factor 1/2 and with Robin coeffi-
224
cients given by Eq. (5.7). Similarly, the VEV of the energymomentum tensor is defined
by formulae (4.2) or (4.3) by an additional factor 1/2. Note that global quantities, such as
the total Casimir energy, are the same. Motivated by the possibility for the stabilization of
extra dimensions by quantum effects, the one-loop Casimir energy of various braneworld
compactifications has been investigated by several authors for both scalar and fermion
fields (see references cited in Introduction). It has been shown that for the simultaneous
solution of the stabilization and hierarchy problems a fine tuning of the model parameters
is essential. From the point of view of backreaction of quantum effects the investigation
of local densities is also of considerable interest. For a conformally coupled massless
scalar field the corresponding results can be obtained from the results in the Minkowski
spacetime by using the standard formula for conformally related problems. For the case
of plane-symmetric line element (2.1) with an arbitrary function (y) this has been done
in Ref. [47] by using the results from [69]. Recently the energymomentum tensor in the
RandallSundrum braneworld for a bulk scalar with zero brane mass terms c1 and c2 is
considered in Ref. [49]. This paper appeared when our calculations were in progress. Note
that in [49] only a general formula is given for the unrenormalized VEV in terms of the
differential operator acting on the Green function. In our approach the application of the
generalized AbelPlana formula allowed us to extract manifestly the part due to the AdS
bulk without boundaries and for the points away from the boundaries the renormalization
procedure is the same as for the boundary-free parts. The latter is well-investigated in literature. In addition, the boundary induced parts are presented in terms of exponentially
convergent integrals convenient for numerical calculations.
6. Conclusion
The natural appearance of AdS in a variety of situations has stimulated considerable
interest in the behavior of quantum fields propagating in this background. In the present
paper we have investigated the Wightman function and the vacuum expectation value of
the energymomentum tensor for a scalar field with an arbitrary curvature coupling parameter satisfying Robin boundary conditions on two parallel plates in AdS spacetime. The
application of the generalized AbelPlana formula to the mode sum over zeros of the corresponding combinations of the cylinder functions allowed us to extract the boundary-free
AdS part and to present the boundary induced parts in terms of exponentially convergent
integrals. In the region between two plates the Wightman function is presented in two
equivalent forms given by Eqs. (2.31) and (2.43). The first terms on the right of these formulae are the Wightman function for AdS bulk without boundaries. The second ones are
induced by a single plate and the third terms are due to the presence of the second plate.
The expectation values for the energymomentum tensor are obtained by applying on the
Wightman function a certain second order differential operator and taking the coincidence
limit. For the case of a single plate geometry this leads to formula (3.12) for the region
z > za and to formula (3.15) for the region z < za . As we could expect from the problem symmetry the part of this tensor corresponding to the components on the hyperplane
parallel to the plate is proportional to the corresponding metric tensor. On the boundary
the vacuum energymomentum tensor diverges, except the case of a conformally coupled
225
massless scalar. The leading terms of the corresponding asymptotic expansion near the
boundary are given by expressions (3.26) and (3.27). These terms are the same as for a
plate in the Minkowski bulk. They do not depend on the Robin coefficient, have different signs for Dirichlet and Neumann scalars and vanish for a conformally coupled scalar.
The coefficients for the subleading asymptotic terms will depend on the AdS curvature
radius, Robin coefficient and on the mass of the field. For large proper distances from
the plate to compared with the AdS curvature radius, kD |y a| 1, the boundary induced energymomentum tensor vanishes as exp[2kD (a y)] in the region y > a and as
exp[kD (2 + D)(y a)] in the region y < a. The same behavior takes place for a fixed
y a and large values of the parameter kD . In the large mass limit, m kD , the boundary
parts are exponentially suppressed. Note that here we consider the bulk energymomentum
tensor. For a scalar field on manifolds with boundaries in addition to this part, the energy
momentum tensor contains a contribution located on the boundary (for the expression of
the surface energymomentum tensor in the case of arbitrary bulk and boundary geometries see Ref. [74]). As it has been discussed in Refs. [65,69,7376], the surface part of the
energymomentum tensor is essential in considerations of the relation between local and
global characteristics in the Casimir effect. The vacuum expectation value of the surface
energymomentum tensor for the geometry of two parallel branes in AdS bulk is evaluated
in Ref. [77]. It is shown that for large distances between the branes the induced surface
densities give rise to an exponentially suppressed cosmological constant on the brane. In
particular, in the RandallSundrum model the cosmological constant generated on the visible brane is of the right order of magnitude with the value suggested by the cosmological
observations.
For the case of two-plates geometry the vacuum expectation value of the bulk energy
momentum tensor is presented as a sum of purely AdS, single plates, and interference
parts. The latter is given by Eq. (4.5) and is finite for all values za z zb . In the limit
za zb the standard result for two parallel plates in the Minkowski bulk is obtained. For
two-plates case the vacuum forces acting on boundaries contain two terms. The first ones
are the forces acting on a single boundary when the second boundary is absent. Due to
the well-known surface divergencies in the expectation values of the energymomentum
tensor these forces are infinite and need an additional regularization. The another terms
in the vacuum forces are finite and are induced by the presence of the second boundary
and correspond to the interaction forces between the plates. These forces are given by formula (4.12) with j = a, b for the plate at z = za and z = zb , respectively. For Dirichlet
scalar they are always attractive. In the case of mixed boundary conditions the interaction
forces can be either repulsive or attractive. Moreover, there is a region in the space of Robin
parameters in which the interaction forces are repulsive for small distances and are attractive for large distances. This provides a possibility to stabilize interplate distance by using
the vacuum forces. For large distances between the plates, the vacuum interaction forces
per unit surface are exponentially suppressed by the factor exp[2kD (a b)] for the plate
at y = a and by the factor exp[kD (2 + D)(a b)] for the plate at y = b. In Section 5 we
give an application of our results for two plates case to the RandallSundrum braneworld
with arbitrary mass terms on the branes. For the untwisted scalar the Robin coefficients are
expressed through these mass terms and the curvature coupling parameter. For the twisted
scalar Dirichlet boundary conditions are obtained on both branes. Note that in this paper
226
we have considered boundary induced vacuum densities which are finite away from the
boundaries. As it has been mentioned in Ref. [78], the same results will be obtained in the
model where instead of externally imposed boundary condition the fluctuating field is coupled to a smooth background potential that implements the boundary condition in a certain
limit [79].
Acknowledgements
I am grateful to Michael Bordag and Dmitri Vassilevich for valuable discussions and
suggestions. The work was supported by the DAAD grant. I acknowledge the hospitality
of the Institute for Theoretical Physics, University of Leipzig.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
C. Fronsdal, Phys. Rev. D 10 (1974) 589.

C. Fronsdal, R.B. Haugen, Phys. Rev. D 12 (1975) 3810.
S.J. Avis, C.J. Isham, D. Storey, Phys. Rev. D 18 (1978) 3565.
P. Breitenlohner, D.Z. Freedman, Ann. Phys. (N.Y.) 144 (1982) 249.
W. Heidenreich, Phys. Lett. B 110 (1982) 461.
B. Binegar, C. Fronsdal, W. Heidenreich, Ann. Phys. (N.Y.) 149 (1983) 254;
B. Binegar, C. Fronsdal, W. Heidenreich, J. Math. Phys. 24 (1983) 2828.
E.S. Fradkin, A.A. Tseytlin, Nucl. Phys. B 234 (1984) 472.
N. Sakai, Y. Tanii, Phys. Lett. B 146 (1984) 38.
C. Dullemond, E. van Beveren, J. Math. Phys. 26 (1985) 2050.
C.P. Burgess, C.A. Ltken, Phys. Lett. B 153 (1985) 137.
B. Allen, T. Jacobson, Commun. Math. Phys. 103 (1986) 669.
R. Camporesi, Phys. Rev. D 43 (1991) 3958.
R. Camporesi, A. Higuchi, Phys. Rev. D 45 (1992) 3951.
M. Kamela, C.P. Burgess, Can. J. Phys. 77 (1999) 85.
L.-X. Li, Phys. Rev. D 59 (1999) 084016.
M.M. Caldarelli, Nucl. Phys. B 549 (1999) 499.
W.D. Goldberger, I.Z. Rothstein, Phys. Rev. Lett. 89 (2002) 131601;
W.D. Goldberger, I.Z. Rothstein, hep-th/0208060.
O. Aharony, S.S. Gubser, J. Maldacena, H. Ooguri, Y. Oz, Phys. Rep. 323 (2000) 183.
N. Arkani-Hamed, S. Dimopoulos, G. Dvali, Phys. Lett. B 429 (1998) 263;
N. Arkani-Hamed, S. Dimopoulos, G. Dvali, Phys. Rev. D 59 (1999) 086004;
I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos, G. Dvali, Phys. Lett. B 436 (1998) 257.
L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370.
L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 4690.
T. Gherghetta, A. Pomarol, Nucl. Phys. B 586 (2000) 141.
A. Pomarol, Phys. Lett. B 486 (2000) 153.
H. Davoudiasl, J.L. Hewett, T.G. Rizzo, Phys. Lett. B 473 (2000) 43.
R. Altendorfer, J. Bagger, D. Nemeschansky, Phys. Rev. D 63 (2001) 125025.
S.R. Huber, Q. Shafi, Phys. Rev. D 63 (2001) 045010.
M. Gell-Mann, B. Zwiebach, Phys. Lett. B 141 (1984) 333;
M. Gell-Mann, B. Zwiebach, Nucl. Phys. B 260 (1985) 569.
W.D. Goldberger, M.B. Wise, Phys. Rev. Lett. 83 (1999) 4922;
W.D. Goldberger, M.B. Wise, Phys. Rev. D 60 (1999) 107505;
C. Csaki, J. Erlich, T. Hollowood, Y. Shirman, Nucl. Phys. B 581 (2000) 309;
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
227
K. Maeda, D. Wands, Phys. Rev. D 62 (2000) 124009;

C. Barcelo, M. Visser, Phys. Rev. D 63 (2001) 024004.
V.M. Mostepanenko, N.N. Trunov, The Casimir Effect and Its Applications, Clarendon, Oxford, 1997.
G. Plunien, B. Muller, W. Greiner, Phys. Rep. 134 (1986) 87.
M. Bordag (Ed.), The Casimir Effect. 50 Years Later, World Scientific, Singapore, 1999.
M. Bordag, U. Mohideen, V.M. Mostepanenko, Phys. Rep. 353 (2001) 1.
M. Bordag (Ed.), Proceedings of the Fifth Workshop on Quantum Field Theory under the Influence of
External Conditions, Int. J. Mod. Phys. A 17 (2002) 6, 7.
K.A. Milton, The Casimir Effect: Physical Manifestation of Zero-Point Energy, World Scientific, Singapore,
2002.
M. Fabinger, P. Horava, Nucl. Phys. B 580 (2000) 243.
S. Nojiri, S. Odintsov, S. Zerbini, Phys. Rev. D 62 (2000) 064006.
S. Nojiri, S. Odintsov, Phys. Lett. B 484 (2000) 149.
D.J. Toms, Phys. Lett. B 484 (2000) 119.
S. Nojiri, O. Obregon, S. Odintsov, Phys. Rev. D 62 (2000) 104003.
W. Goldberger, I. Rothstein, Phys. Lett. B 491 (2000) 339.
S. Nojiri, S. Odintsov, JHEP 0007 (2000) 049.
A. Flachi, D.J. Toms, Nucl. Phys. B 599 (2001) 305.
A. Flachi, D.J. Toms, Nucl. Phys. B 610 (2001) 144.
I. Brevik, K.A. Milton, S. Nojiri, S.D. Odintsov, Nucl. Phys. B 599 (2001) 305.
J. Garriga, O. Pujolas, T. Tanaka, Nucl. Phys. B 605 (2001) 192.
A. Flachi, I.G. Moss, D.J. Toms, Phys. Lett. B 518 (2001) 153;
A. Flachi, I.G. Moss, D.J. Toms, Phys. Rev. D 64 (2001) 105029.
A.A. Saharian, M.R. Setare, Phys. Lett. B 552 (2003) 119.
E. Elizalde, S. Nojiri, S.D. Odintsov, S. Ogushi, Phys. Rev. D 67 (2003) 063515.
A. Knapman, D.J. Toms, Phys. Rev. D 69 (2004) 044023.
M. Bordag, D.V. Vassilevich, J. Phys. A 32 (1999) 8247.
I.G. Moss, Phys. Lett. B 491 (2000) 203.
P.B. Gilkey, K. Kirsten, D.V. Vassilevich, Nucl. Phys. B 601 (2001) 125.
S. Nojiri, S. Odintsov, S. Zerbini, Class. Quantum Grav. 17 (2000) 4855.
W. Naylor, M. Sasaki, Phys. Lett. B 542 (2002) 289.
I.G. Moss, W. Naylor, W. Santiago-Germn, M. Sasaki, Phys. Rev. D 67 (2003) 125010.
S. Mukohyama, Phys. Rev. D 63 (2001) 044008.
R. Hofmann, P. Kanti, M. Pospelov, Phys. Rev. D 63 (2001) 124020.
A.H. Yeranyan, A.A. Saharian, Astrophysics 46 (2003) 386.
J. Ambjorn, S. Wolfram, Ann. Phys. (N.Y.) 147 (1983) 33.
H. Luckock, J. Math. Phys. 32 (1991) 1755.
G. Esposito, A.Yu. Kamenshchik, G. Polifrone, Euclidean Quantum Gravity on Manifolds with Boundary,
Kluwer, Dordrecht, 1997.
S.N. Solodukhin, Phys. Rev. D 63 (2001) 044002.
N.D. Birrell, P.C.W. Davis, Quantum Fields in Curved Space, Cambridge Univ. Press, Cambridge, 1982.
A.A. Saharian, The generalized AbelPlana formula. Applications to Bessel functions and Casimir effect,
Report No. IC/2000/14, hep-th/0002239.
A.A. Saharian, Phys. Rev. D 63 (2001) 125007.
A.A. Saharian, Izv. AN Arm. SSR. Matematika 22 (1987) 166, Sov. J. Contemp. Math. Anal. 22 (1987) 70;
A.A. Saharian, PhD thesis, Yerevan, 1987 (in Russian).
M. Abramowitz, I.A. Stegun (Eds.), Handbook of Mathematical Functions, Dover, New York, 1972.
A.P. Prudnikov, Yu.A. Brychkov, O.I. Marichev, Integrals and Series, vol. 2, Gordon & Breach, New York,
1986.
A. Romeo, A.A. Saharian, J. Phys. A 35 (2002) 1297.
P. Candelas, J.S. Dowker, Phys. Rev. D 19 (1979) 2902.
H.H. Matevosyan, A.A. Saharian, in preparation.
D. Deutsch, P. Candelas, Phys. Rev. D 20 (1979) 3063.
G. Kennedy, R. Critchley, J.S. Dowker, Ann. Phys. (N.Y.) 125 (1980) 346.
228
[74]
[75]
[76]
[77]
[78]
[79]

A. Romeo, A.A. Saharian, Phys. Rev. D 63 (2001) 105019.
S.A. Fulling, J. Phys. A 36 (2003) 6857.
N. Graham, R.L. Jaffe, V. Khemani, M. Quandt, M. Scandurra, H. Weigel, Phys. Lett. B 572 (2003) 196.
N. Graham, R.L. Jaffe, V. Khemani, M. Quandt, M. Scandurra, H. Weigel, Nucl. Phys. B 645 (2002) 49;
N. Graham, R.L. Jaffe, H. Weigel, Int. J. Mod. Phys. A 17 (2002) 846.
Two-loop QCD corrections to the heavy quark form

factors: Axial vector contributions
W. Bernreuther a , R. Bonciani b , T. Gehrmann c , R. Heinesch a ,
T. Leineweber a , P. Mastrolia d , E. Remiddi e
a Institut fr Theoretische Physik, RWTH Aachen, D-52056 Aachen, Germany
b Fakultt fr Mathematik und Physik, Albert-Ludwigs-Universitt Freiburg, D-79104 Freiburg, Germany
c Institut fr Theoretische Physik, Universitt Zrich, CH-8057 Zrich, Switzerland
d Department of Physics and Astronomy, UCLA, Los Angeles, CA 90095-1547, USA
e Dipartimento di Fisica dellUniversit di Bologna, and INFN, Sezione di Bologna, I-40126 Bologna, Italy
Abstract
vertex to order 2 in the QCD coupling for an on-shell massive quark
We consider the Z QQ
S
antiquark pair and for arbitrary momentum transfer of the Z boson. We present closed analytic
expressions for the two parity-violating form factors of that vertex at the two-loop level in QCD,
excluding the contributions from triangle diagrams. These form factors are expressed in terms of
1-dimensional harmonic polylogarithms of maximum weight 4.
PACS: 11.15.Bt; 12.38.Bx; 14.65.Fy; 14.65.Ha; 13.88.+e
Keywords: Feynman diagrams; Multi-loop calculations; Vertex diagrams; Heavy quarks
E-mail addresses: breuther@physik.rwth-aachen.de (W. Bernreuther),

roberto.bonciani@physik.uni-freiburg.de (R. Bonciani), gehrt@physik.unizh.ch (T. Gehrmann),
heinesch@physik.rwth-aachen.de (R. Heinesch), leineweber@physik.rwth-aachen.de (T. Leineweber),
mastrolia@physics.ucla.edu (P. Mastrolia), ettore.remiddi@bo.infn.it (E. Remiddi).
doi:10.1016/j.nuclphysb.2005.01.035
230
W. Bernreuther et al. / Nuclear Physics B 712 (2005) 229286
1. Introduction
This paper is the second of a series that is devoted to the computation of the electromagnetic and neutral current form factors of heavy quarks Q at the two-loop level in QCD [1].
These form factors are relevant for a number of applications. For instance, they are part of
the order S2 QCD corrections to the differential electroproduction cross sections of heavy
and, in particular, part of the second order QCD correcquarks e+ e , Z QQX
tions to the forwardbackward asymmetry AQ
fb . As far as b quarks are concerned, an order
2
S calculation with a non-zero b quark mass is of interest in view of the discrepancy between experimental measurement and theoretical expectationssee [2] and the references
given in [1] concerning the state of the theoretical predictions. A review of the present
status of electroweak corrections to the forwardbackward asymmetry is given in [3]. At
a future linear collider, forwardbackward asymmetries will play a prominent role in very
precise measurements of the neutral current couplings of bottom and of top quarks [4].
Clearly, predictions will be required taking the mass of the heavy quark fully into account.
In [1] we presented closed analytic expressions to order S2 of the heavy-quark electromagnetic vertex form factors for arbitrary momentum transfer. Up to an overall coupling
factor these are identical to the corresponding vector, i.e., parity-conserving form factors
that appear in the amplitude of the decay of a virtual Z boson into a heavy quarkantiquark
pair. In this paper, we compute the axial vector form factors G1 and G2 , excluding the
anomalous triangle graph contributions, Fig. 1(a) and (b), which contribute only through
the mass splitting of a quark isospin doublet in the triangle loop. These terms deserve a
separate discussion and will be given elsewhere [5].
The paper is organized as follows. In Section 2 we fix our conventions and describe how
the form factors can be obtained from the vertex amplitude by appropriate projections. In
Section 3 the renormalization constants in the scheme that we usewhich is the same as
the one chosen in [1]are collected for the convenience of the reader. Sections 4 and 5
contain, for spacelike momentum transfer, the unsubtracted and renormalized axial vector
form factors at one-loop and two-loop order, respectively, for the cases of the renormalization scale being both equal and different from the heavy quark mass. In Section 6 the form
threshold, and we give there also their
factors are analytically continued above the QQ
threshold and asymptotic expansions. We conclude in Section 7.
2. The axial vector form factors
We consider the amplitude Vc1 c2 (p1 , p2 ) for the decay of a virtual Z boson of fourmomentum q = p1 + p2 into a massive quarkantiquark pair of momenta p1 , p2 and
are on their mass shell, p 2 = p 2 = m2 , where m denotes
colors c1 , c2 . The quarks Q, Q
1
2
the pole mass of Q. The squared center-of-mass energy is S = (p1 + p2 )2 .
A general decomposition of the vertex function V (p1 , p2 ) involves 6 form factors, two
of which odd under a CP transformation. As we consider here, besides QCD interactions,
Standard Model (SM) neutral current interactions to lowest order, CP invariance holds.
231
This implies that V depends only on 4 form factors, and we use the decomposition
Vc1 c2 (p1 , p2 ) = u c1 (p1 )c1 c2 (q)vc2 (p2 ),

1
F2 (s)i q
c1 c2 (q) = (i) vQ F1 (s) + vQ
2m

1
G2 (s)5 q c1 c2 ,
+ aQ G1 (s) 5 + aQ
2m
(1)
(2)
where s = S/m2 , = 2i [ , ], u c1 (p1 ), vc2 (p2 ) are Dirac spinors and vQ , aQ are the
SM vector and axial vector couplings of the massive quark Q to the Z boson:
Q

Q
T3
e
e T3
2
aQ =
sw QQ ,
,
vQ =
(3)
sw cw 2
sw cw 2
where sw (cw ) is the sine (cosine) of the weak mixing angle, T3Q the third component of
the weak isospin and QQ is the charge of the heavy quark in units of the positron charge
e > 0.
The dimensionless form factors Fi , Gi can be obtained from Vc1 c2 (p1 , p2 ) by appropriate projections. We consider the spinor traces

P1 = Tr (/
(4)
p 1 + m) (q)(/
p 2 m) ,

p 1 + m) (q)(/
p 2 m) ,
P2 = Tr t (/
(5)

p 1 + m) (q)(/
p 2 m) ,
P3 = Tr 5 (/
(6)

p 1 + m) (q)(/
p 2 m) ,
P4 = Tr 5 q (/
(7)
where t = p2 p1 . Since we are working in D = 4 2 dimensions we calculate these

traces in D dimensions as well. Inserting Eq. (2) into Eqs. (4)(7) and performing the
traces one obtains
i 2P2 + m(4 s)P1
,
F1 + F 2 =
(8)
vQ 4sm3 (1 )(4 s)
i [2 + (1 )s]P2 + m(4 s)P1
F2 =
(9)
,
vQ
sm3 (1 )(4 s)2
smP3 + 2P4
i
,
G1 =
(10)
aQ 4sm3 (1 )(4 s)
i msP3 + (2(3 2) (1 )s)P4
.
G2 =
(11)
aQ
s 2 m3 (1 )(4 s)
The trace of the unit matrix is kept equal to four also in D dimensions. In calculating the
diagrams considered in this paper we use an anticommuting 5 in D dimensions. (This
prescription is also used in the derivation of the formulae (10), (11).) This prescription is
appropriate as the diagrams below correspond to the order S2 non-singlet contributions
to the matrix element of the axial vector current, and it is well known that for these contributions a canonical, i.e., non-anomalous Ward identity must hold to this order. Within
dimensional regularization this is most conveniently implemented with an anticommuting 5 . In a subsequent paper [5], the contributions involving closed triangle loops, Fig. 1,
232
Fig. 1. Triangle diagram contributions to V . Crossed diagrams are not drawn. The external dashed line refers
to an incoming Z boson, the curly lines to gluons, the double straight lines to the massive quark and the simple
straight lines to massless quarks.
will be computed for a mass-split doublet of fermions in the triangle loop. In that context, the naive anticommuting 5 used here is problematic in D = 4. However, it will be
shown there that using a different 5 prescription [6,7] does not affect the non-anomalous
contributions presented here.
The formulae Eqs. (8)(11) show that with the above projections the computation of
the vector form factors F1,2 and the axial vector (i.e., parity-violating) form factors G1,2
decouple from each other. The vector form factors in Eqs. (8), (9) were computed in [1].
Here we determine the form factors in Eqs. (10), (11) to the second order in the strong
coupling constant S , excluding the contributions from the triangle diagrams shown in
Fig. 1. Expanding the renormalized form factors to the second order in S , we have

2

2
S
2
S
2
(1l)
(2l)
G1,R s, , 2 +
G1 s, , 2 = 1 +
G1,R s, , 2
2
2
m
m
m
3
S
+O
(12)
,
2

2

2
S
2
S
2
(1l)
(2l)
G2 s, , 2 =
G2,R s, , 2 +
G2,R s, , 2
2
2
m
m
m
3
S
+O
(13)
,
2
where the superscripts (1l) and (2l) denote the one- and two-loop contributions. The subscript R labels the renormalization scheme specified in the next section. After having
performed the renormalization, the form factors still depend on the parameter , which
regularizes the remaining infrared divergencies. We keep S dimensionless also in D = 4
dimensions.
The form factors are represented as series in and expressed in terms of 1-dimensional
harmonic polylogarithms H (
a ; x) up to weight 4 [8,9], which are functions of the dimensionless variable x defined by
S + 4m2 S
s + 4 s
=
x=
(14)
.
s + 4 + s
S + 4m2 + S
We give our results firstly in the kinematical region in which s is spacelike (0 x 1),
where the form factors are real. In Section 6 we shall perform the analytical continuation
233
to the physical region above threshold, s > 4, 1 < x 0, and explicitly decompose the
form factors into real and imaginary parts.
In what follows Nf denotes the number of light quarks (which we take to be massless)
running in the loops Fig. 4(g), and CF = (Nc2 1)/(2Nc ), CA = Nc , TR = 1/2, where Nc
is the number of colors.
3. Renormalization scheme
As in our previous paper [1], we use renormalized perturbation theory with S =
gs2 /(4) being defined as the standard MS coupling in QCD with Nf massless and one
massive quark, while we define the mass m and the wave-function of the heavy quark Q in
the on-shell (OS) scheme.
For the renormalization procedure we need the coupling renormalization and the gluon
wave function to one-loop

2
S 1 11
4
S
,
CA TR (Nf + 1) + O
ZgMS = 1 C()
(15)
2 4 3
3
2
where
C() = (4) (1 + )
(16)
and, in the Feynman gauge,

Z3MS = 1 + C()

2
S 1 5
4
S
.
CA TR (Nf + 1) + O
2 2 3
3
2
(17)
The renormalization constants concerning the heavy quark are defined in the on-shell
scheme. Here we need

2
1 2 3 2
S
S
OS
Zm
(18)
,
C()CF
=1+
+
O
2
2 1 2 m2
2
to one-loop order. Here and in the following denotes the mass scale of dimensional
regularization/renormalization. The wave function Z2OS is needed to two-loop order. The
latter was computed in [10,11]. Using the result of [11] and expressing it in terms of the
renormalized MS coupling S we have

2
2
1 2 3 2
S
S
S
(2)
OS
2
Z2 = 1 +
+ C ()
Z2 + O
C()CF
2
2
2 1 2 m
2
2
(19)
(2)
where Z2 is
(2)
Z2
2

9
51
433 39
=
+
+
(2) + 24 (2) ln(2) 6 (3)
32
2
8 2 16
2

11 3
4
+
8
+ CF C A
+
12 2
m2
CF2
2
m2
234
2

11
803
101 15
2
+ (2)
12 (2) ln(2) + 3 (3)
16
2
32
8
2

1 3
4
+ +8
+ CF TR Nf
3 2
m2

2 2
1
59
9
+ 2 (2) +
+
+
8
m2
2 2 4
2

4
1 3
+
+ CF T R
+
8
3 2
m2

2 2
1
1139
19
+
8
(2)
.
+
(20)
+
72
m2
2 12
2
+
m2
Further we need the renormalization constant Z1F for the QQ gluon vertex to one loop.
Using a SlavnovTaylor identity and Eqs. (15), (17), (19) we get

Z1F = ZgMS Z2OS
= 1 C()
Z3MS
2

S
2 3 2
S 1
+
O
.
CA CF
2 2
1 2 m2
2
(21)
For the counterterm contributions to the renormalized form factors it is convenient to define
(here our notation differs slightly from that in [1])
1F = Z1F 1,
(22)
(1l)
2 = Z2OS 1 = 2
(2l)
+ 2 ,
(23)
3 = Z3MS 1,

OS
1 m,
m = Z2OS Zm
(1l)
(24)
(25)
(2l)
where 2 , 2 can be read off from Eq. (19).

In this renormalization scheme the renormalized vertex function to order S2 is given
by the contributions from the 1-particle irreducible diagrams, Figs. 2 and 4, which we call
the unsubtracted contributions, and the counterterms given below.
Fig. 2. One-loop QCD contribution to Vc1 c2 (q).
235
4. One-loop unsubtracted and UV-renormalized form factors

(1l)
(1l)
In this section we give the results for G1,R , G2,R . First we write down the contribution
from the diagram Fig. 2 including the terms of order . They are needed for computing the
order S2 counterterms shown in Fig. 5(c), (d) and (f) of Section 5.2. We obtain

2

2
S
1
(Fig. 2)
s, , 2 =
CF C()
a1 + a2 + a3 + O 2 ,
G1
(26)
2
2

m
m

2
2
2
S
1
(Fig. 2)
G2
(27)
s, , 2 =
CF C()
b
+
b
+
b
1
2
3 +O ,
2
2

m
m
where C() is defined in Eq. (16).
The coefficients ai and bi (i = 1, . . . , 3) are

a1 = x 2 + 1 / (x 1)(x + 1) H (0; x) + 1/2,

a2 = 1/2 3x 2 2x + 3 / (x 1)(x + 1) H (0; x)

2 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

+ x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

(2) x 2 + 1 / (x 1)(x + 1) ,

a3 = 2 (2) x 2 + 1 / (x 1)(x + 1) H (1; x)

x 2 + 1 4 + (2) / (x 1)(x + 1) H (0; x)

3x 2 2x + 3 / (x 1)(x + 1) H (1, 0; x)

+ 4 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

2 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

+ 1/2 3x 2 2x + 3 / (x 1)(x + 1) H (0, 0; x)

2 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

+ x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)
1/2 3 (2)x 2 + 4 (3)x 2 2 (2)x

+ 3 (2) + 4 (3) / (x 1)(x + 1) ,
(28)
(29)
(30)
respectively,
b1 = 0,

b2 = 2 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x) 4/(x 1)2 x,

b3 = 4 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x)

4 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

+ 2 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)

2 3 (2)x 2 + 4x 2 2 (2)x 4 + 3 (2) / (x + 1)(x 1)3 x.
(31)
(32)
(33)
236
Fig. 3. Subtraction term for the one-loop renormalization.
The counterterm of Fig. 3 contributes to G1 :

2
(Fig. 3)
(1l)
s, , 2 = 2 .
G1
m
From Eqs. (26), (27) and (34) we obtain the renormalized one-loop form factors

2
2
2
1
(1l)
G1,R s, , 2 = CF C()
a
+
a
+

a
1
2
3 +O ,
2

m
m

2
2
(1l)
G2,R s, , 2 = CF C()
b 1 + b2 + b3 + O 2 ,

m
m2
(34)
(35)
(36)
with
a 1 = a1 3/2,
(37)
a 2 = a2 2,
(38)
a 3 = a3 4
(39)
and bi = bi for i = 1, . . . , 3.
5. Two-loop unsubtracted and renormalized form factors

5.1. Two-loop unsubtracted form factors
First we compute the two-loop unsubtracted contributions represented by the diagrams
of Fig. 4(a)(i) and the mirror diagrams to Fig. 4(a)(c) to the form factors G1 , G2 .
Performing the algebra we obtain, as explained in [1214], the form factors in
Eqs. (10), (11) expressed in terms of hundreds of different scalar integrals. These integrals are reduced to a small set of master integrals by means of the so-called Laporta
algorithm [15] with the help of integration-by-parts identities [16] and Lorentz-invariance
identities [17]. Symmetry relations which one can establish between different integrals are
also used during the reduction. The master integrals themselves were evaluated with the
method of differential equations [1720] in [12,14]. The master integrals, and thus the form
factors are represented as series in the regularization parameter and expressed in terms
of 1-dimensional harmonic polylogarithms up to weight 4 [8,9], which are functions of
the dimensionless variable x defined in Eq. (14). As in the one-loop case we take s to be
spacelike.
237
Fig. 4. Two-loop QCD contributions to Vc1 c2 (p1 , p2 ). The inner dashed line refers to ghosts. Straight lines
represent massless quarks and the double straight line the massive quark (cf. Fig. 1).
The two-loop unsubtracted form factors are found to be

2
(Fig. 4)
s, , 2
G1
m
2
2 2
S
2
=
C ()
2
m2

1
2 CF TR Nf c1 + CF TR c2 + CF CA c3 + CF2 c4

1
+ CF TR Nf c5 + CF TR c6 + CF CA c7 + CF2 c8


(Fig. 4)
G2

=
s, ,
S
2
2
2
m2
2 2
C 2 ()
m2

+ O() ,
(40)
238
CF TR Nf d1 + CF TR d2 + CF CA d3 + CF2 d4
2

1
+ CF TR Nf d5 + CF TR d6 + CF CA d7 + CF2 d8

+ CF TR Nf d9 + CF TR d10 + CF CA d11 + CF2 d12 + O() ,
(41)
where C() is defined in Eq. (16) and Nf , CF , CA , and TR are given at the end of Section 2.
One finds for the ci (i = 1, . . . , 12)
1
4
c1 = c 2 = c3
2
11

= 1/3 x 2 + 1 / (x 1)(x + 1) H (0; x) 1/6,

c4 = 1/2 x 4 + 20x 3 + 14x 2 + 20x + 1 / (x + 1)3 (x 1) H (0; x)
2

+ x 2 + 1 / (x 1)2 (x + 1)2 H (0, 0; x)

+ 1/8 25x 2 + 2x + 25 /(x + 1)2 ,

c5 = 2/9 7x 2 3x + 7 / (x 1)(x + 1) H (0; x)

+ 4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

2/3 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

+ 1/36 13x 2 + 24 (2)x 2 + 13 + 24 (2) / (x 1)(x + 1) ,

c6 = 1/3 3x 2 2x + 3 / (x 1)(x + 1) H (0; x)

+ 4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

2/3 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

+ 1/12 3x 2 + 8 (2)x 2 + 3 + 8 (2) / (x 1)(x + 1) ,

c7 = 1/18 27 (2)x 4 83x 4 + 33x 3 + 18 (2)x 2 33x 9 (2) + 83

/ (x + 1)2 (x 1)2 H (0; x)

14/3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)
2

+ x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 0; x)

+ 1/6 23x 2 + 11 / (x 1)(x + 1) H (0, 0; x)

2 x 2 + 1 / (x + 1)2 (x 1)2 x 2 H (0, 0, 0; x)

+ x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

2
x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 0; x)
1/144 48 (2)x 4 185x 4 + 72 (3)x 4 + 370x 2 + 288 (2)x 2

+ 144 (3)x 2 185 336 (2) + 72 (3) / (x + 1)2 (x 1)2 ,
c8 = 1/4 4 (2)x 5 3x 5 + 4 (2)x 4 49x 4 + 18x 3

+ 8 (2)x 3 + 8 (2)x 2 18x 2 + 4 (2)x
(42)
(43)
(44)
(45)
(46)
239

+ 49x + 3 + 4 (2) / (x + 1)3 (x 1)2 H (0; x)

x 4 + 20x 3 + 14x 2 + 20x + 1 / (x + 1)3 (x 1) H (1, 0; x)

2
2 x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 0; x)

+ 1/2 7x 5 + 21x 4 + 2x 3 + 14x 2 17x + 5 / (x + 1)3 (x 1)2 H (0, 0; x)

2
4 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 0; x)

2
+ 3 x 2 + 1 / (x 1)2 (x + 1)2 H (0, 0, 0; x)
1/16 35x 4 + 8 (2)x 4 + 6x 3 + 160 (2)x 3 + 112 (2)x 2 6x

+ 160 (2)x 35 + 8 (2) / (x + 1)3 (x 1) ,
(47)
2

c9 = 8/3 (2) x + 1 / (x 1)(x + 1) H (1; x)

1/54 353x 2 78x + 353 / (x 1)(x + 1) H (0; x)

+ 8/9 7x 2 3x + 7 / (x 1)(x + 1) H (1, 0; x)

16/3 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

+ 8/3 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

4/9 7x 2 3x + 7 / (x 1)(x + 1) H (0, 0; x)

+ 8/3 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

4/3 x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)
+ 1/216 169x 2 + 576 (3)x 2 + 528 (2)x 2 288 (2)x

+ 576 (3) + 169 + 816 (2) / (x 1)(x + 1) ,
(48)
2

c10 = 4/3 (2) x + 1 / (x 1)(x + 1) H (1; x)

1/54 409x 4 + 80x 3 + 302x 2 216 (2)x 2 + 80x + 409

/ (x 1)(x + 1)3 H (0; x)

+ 2/3 3x 2 2x + 3 / (x 1)(x + 1) H (1, 0; x)

8/3 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

+ 4/3 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

+ 2/9 5x 5 27x 4 + 5x 3 23x 2 + 6x 14 / (x + 1)4 (x 1) H (0, 0; x)

+ 4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

4/3 x 2 x + 1 x 2 + 3x + 1 / (x 1)(x + 1)3 H (0, 0, 0; x)
+ 1/216 288 (3)x 5 + 223x 5 + 1728 (2)x 5 291x 4 2304 (2)x 4

+ 864 (3)x 4 + 1152 (3)x 3 + 1872 (2)x 3 514x 3
+ 1152 (3)x 2 + 514x 2 1008 (2)x 2 + 291x + 864 (3)x

+ 3312 (2)x 1296 (2) + 288 (3) 223 / (x + 1)4 (x 1) ,
(49)
240
c11 = 1/3 (2)x 4 + 6 (3)x 4 + 81 (2)x 3 + 12 (3)x 2 108 (2)x 2

+ 81 (2)x + 6 (3) 37 (2) / (x + 1)2 (x 1)2 H (1; x)

+ (2) 7x 6 + 6x 5 29x 4 + 24x 3 31x 2 + 6x + 5

/ (x + 1)3 (x 1)3 H (0, 1; x)
+ 1/216 648 (2)x 6 + 4129x 6 1620 (3)x 6 + 3024 (3)x 5

12528 (2)x 5 570x 5 4129x 4 + 18792 (2)x 4
13716 (3)x 4 + 25920 (3)x 3 12744 (2)x 3 + 1140x 3
13500 (3)x 2 + 3456 (2)x 2 4129x 2 2808 (2)x

570x + 3024 (3)x 1404 (3) + 4129 / (x + 1)3 (x 1)3 H (0; x)

1/18 161x 4 276x 3 72 (2)x 2 + 276x 161 72 (2)

/ (x 1)2 (x + 1)2 H (1, 0; x)

2
10 x 2 + 1 / (x 1)2 (x + 1)2 H (0, 1, 1, 0; x)

1/3 79x 4 + 6x 3 66x 2 + 6x 13 / (x 1)2 (x + 1)2 H (0, 1, 0; x)
2
4 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 1, 0; x)

+ 2 11x 6 + 5x 5 15x 4 + 26x 3 27x 2 + 5x 1

/ (x + 1)3 (x 1)3 H (0, 0, 1, 0; x)

2
+ 4 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 1, 0; x)

6 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

2
+ 6 x 2 + 1 / (x 1)2 (x + 1)2 H (0, 1, 1, 0; x)
+ 1/18 18 (2)x 6 59x 6 72 (2)x 5 174x 5 + 176x 4

+ 306 (2)x 4 252 (2)x 3 + 240x 3 + 270 (2)x 2

283x 2 72 (2)x 66x + 166 18 (2) / (x + 1)3 (x 1)3 H (0, 0; x)

1/3 64x 4 141x 3 + 228x 2 141x 28

/ (x 1)2 (x + 1)2 H (1, 0, 0; x)

+ 2 7x 6 7x 5 + 33x 4 56x 3 + 25x 2 7x 1

/ (x + 1)3 (x 1)3 H (0, 1, 0, 0; x)

+ 1/3 50x 6 201x 5 + 244x 4 177x 3 + 13x 2 12x + 11

/ (x + 1)3 (x 1)3 H (0, 0, 0; x)

+ 74/3 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

+ 2 x 2 + 3 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 0, 0; x)

+ 2 x 5 5x 4 + 18x 3 26x 2 + x 5 / (x 1)2 (x + 1)3 H (1, 0, 0, 0; x)

2 2x 4 + 5x 3 14x 2 + 5x + 6 / (x 1)2 (x + 1)2 H (1, 0, 0; x)
241

12x 5 x 4 + 10x 3 28x 2 2x 1 / (x + 1)3 (x 1)3 xH (0, 0, 0, 0; x)

2 x 6 5x 5 + 21x 4 34x 3 + 17x 2 5x 3

/ (x + 1)3 (x 1)3 H (0, 1, 0, 0; x)
+ 3x 5 + 4 (2)x 5 x 4 8 (2)x 4 + 40 (2)x 3 + 2x 3

48 (2)x 2 2x 2 + x + 4 (2)x + 3 8 (2) / (x 1)2 (x + 1)3 H (1, 0; x)

6 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

2
+ 6 x 2 + 1 / (x 1)2 (x + 1)2 H (0, 1, 1, 0; x)

2
+ 4 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 1, 0; x)

2 7x 5 + 7x 4 + 12x 3 + 4x 2 + x + 1 / (x 1)2 (x + 1)3 H (0, 0, 1, 0; x)

2
4 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 1, 0; x)

+ 2 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

2
2 x 2 + 1 / (x 1)2 (x + 1)2 H (0, 1, 1, 0; x)

+ 8 x 2 + 1 / (x 1)(x + 1)2 xH (0, 1, 0; x)

x 2 + 1 (2)x 2 + 2 (3)x 2 (2) + 2 (3) / (x 1)2 (x + 1)2 H (1; x)

2
+ (2) x 2 + 1 / (x 1)2 (x + 1)2 H (0, 1; x)
1/4320 85920 (2)x 6 77760 (2) ln(2)x 6 + 25200 (3)x 6

12925x 6 + 15336 (2)2 x 6 + 155520 (2) ln(2)x 5 + 28512 (2)2 x 5
+ 181440 (3)x 5 279360 (2)x 5 + 38775x 4 99576 (2)2 x 4
344880 (3)x 4 + 230160 (2)x 4 77760 (2) ln(2)x 4
+ 83520 (2)x 3 + 109728 (2)2 x 3 + 268560 (3)x 2 38775x 2
294720 (2)x 2 + 77760 (2) ln(2)x 2 115560 (2)2 x 2
+ 28512 (2)2 x 181440 (3)x + 195840 (2)x
155520 (2) ln(2)x 648 (2)2 + 77760 (2) ln(2) + 12925

21360 (2) + 51120 (3) / (x + 1)3 (x 1)3 ,
(50)

5
4
3
2
3
2
c12 = 13x 23x + 12x + 24x 61x + 11 (2)/ (x + 1) (x 1) H (1; x)

2 7x 6 + 10x 5 65x 4 + 104x 3 75x 2 + 10x 3

/ (x + 1)3 (x 1)3 H (0, 1, 0, 0; x)
1/2 17x 6 58x 5 213x 4 + 224x 3 37x 2

46x + 17 / (x + 1)3 (x 1)3 H (0, 0, 0; x)

2
12 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 0, 0; x)

+ 27x 6 2x 5 + 27x 4 56x 3 7x 2 2x 7

/ (x + 1)3 (x 1)3 H (0, 0, 0, 0; x)
242

+ 8 x 5 2x 4 + 18x 3 14x 2 + 4x + 1 / (x + 1)3 (x 1)2 H (1, 0, 0, 0; x)

2 (2) 5x 6 + 6x 5 31x 4 + 24x 3 29x 2 + 6x + 7

/ (x + 1)3 (x 1)3 H (0, 1; x)

+ 2 8x 4 19x 3 + 46x 2 19x + 8 / (x 1)2 (x + 1)2 H (1, 0, 0; x)

4 x 6 x 5 + 14x 4 20x 3 + 14x 2 x + 1

/ (x + 1)3 (x 1)3 H (0, 1, 0, 0; x)
+ 4 4x 5 + (2)x 5 5 (2)x 4 + 2x 4 + 34 (2)x 3 2x 3 + 2x 2

30 (2)x 2 + 7 (2)x 2x 4 + (2) / (x + 1)3 (x 1)2 H (1, 0; x)

2 5x 4 10x 3 + 18x 2 10x + 5 / (x 1)2 (x + 1)2 H (0, 1, 0; x)

+ 4 3x 5 + 3x 4 + 6x 3 2x 2 x 1 / (x + 1)3 (x 1)2 H (0, 0, 1, 0; x)

2
+ 4 (3) x 2 + 1 / (x 1)2 (x + 1)2 H (1; x)
+ 1/160 1915x 6 + 4640 (2)x 6 + 2896 (2)2 x 6 640 (3)x 6

5760 (2) ln(2)x 6 + 11520 (2) ln(2)x 5 7680 (3)x 5
3520 (2)x 5 1152 (2)2 x 5 3840x 5 + 1935x 4 + 9584 (2)2 x 4
8160 (2)x 4 + 25280 (3)x 4 5760 (2) ln(2)x 4
21248 (2)2 x 3 1920 (3)x 3 5760 (2)x 3 17280 (3)x 2
+ 5760 (2) ln(2)x 2 + 13920 (2)x 2 + 7632 (2)2 x 2 1935x 2
1152 (2)2 x + 1920 (3)x + 4160 (2)x 11520 (2) ln(2)x
+ 3840x + 5760 (2) ln(2) 1915 5280 (2) + 320 (3)

+ 944 (2)2 / (x + 1)3 (x 1)3

2
+ 4 x 2 + 1 / (x 1)2 (x + 1)2 H (0, 1, 1, 0; x)

2
+ 8 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 1, 0; x)
2
8 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 1, 0; x)

2
+ 8 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 0, 1, 0; x)

+ 13x 5 59x 4 + 72x 3 + 60x 2 21x + 15

/ (x + 1)3 (x 1)2 H (0, 1, 0; x)

2 11x 6 2x 5 + 33x 4 56x 3 + 23x 2 2x + 1

/ (x + 1)3 (x 1)3 H (0, 0, 1, 0; x)
1/8 148 (2)x 6 64 (3)x 6 19x 6 160 (3)x 5 224 (2)x 5

326x 5 + 1056 (3)x 4 676 (2)x 4 + 275x 4 1536 (3)x 3
+ 992 (2)x 3 + 275x 2 724 (2)x 2 + 1056 (3)x 2 160 (3)x

+ 140x 3 326x + 96 (2)x 19 + 4 (2) 64 (3)
243

/ (x + 1)3 (x 1)3 H (0; x)
+ 4 9x 5 + (2)x 5 4x 4 + (2)x 4 + 2 (2)x 3 + 13x 3 13x 2

+ 2 (2)x 2 + 4x + (2)x + 9 + (2) / (x + 1)3 (x 1)2 H (1, 0; x)
+ 1/2 26 (2)x 6 + 123x 6 8 (2)x 5 112x 5 33x 4 + 62 (2)x 4

+ 120x 3 160 (2)x 3 + 42 (2)x 2 135x 2 8 (2)x + 56x 19 + 6 (2)

/ (x + 1)3 (x 1)3 H (0, 0; x)

+ 2 x 4 + 20x 3 + 14x 2 + 20x + 1 / (x + 1)3 (x 1) H (1, 1, 0; x)

9x 5 3x 4 + 116x 3 + 128x 2 41x + 7

/ (x + 1)3 (x 1)2 H (1, 0, 0; x)

2
+ 16 x 2 + 1 / (x 1)2 (x + 1)2 H (1, 1, 0, 0; x).

(51)
The di (i = 1, . . . , 12) are
d1 = d2 = d3 = 0,

d4 = 6 x 2 + 1 / (x + 1)3 (x 1) xH (0; x) + 6/(x + 1)2 x,

4
d5 = d6 = d7 = 4/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x)
11
+ 8/3/(x 1)2 x,

d8 = 9x 4 60x 3 + 22x 2 60x + 9 / (x + 1)3 (x 1)3 xH (0; x)

+ 12 x 2 + 1 / (x + 1)3 (x 1) xH (1, 0; x)

+ 2 x 2 + 1 3x 3 + 11x 2 7x + 9 / (x + 1)3 (x 1)4 xH (0, 0; x)
+ 2 3x 3 + 3 (2)x 3 25x 2 3 (2)x 2

+ 3 (2)x 25x 3 3 (2) / (x + 1)3 (x 1)2 x,

d9 = 2/9 87x 2 50x + 87 / (x + 1)(x 1)3 xH (0; x)

+ 16/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

8/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)
+ 4/9 31x 2 + 18 (2)x 2 12 (2)x

31 + 18 (2) / (x + 1)(x 1)3 x,
d10 = 2/9 87x 4 128x 3 46x 2 72 (2)x 2

128x + 87 / (x + 1)3 (x 1)3 xH (0; x)

+ 8/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

8/3 12x 4 3x 3 + 9x 2 5x + 3 / (x + 1)4 (x 1)3 xH (0, 0; x)

+ 16/ (x + 1)3 (x 1)3 x 3 H (0, 0, 0; x)
+ 8/9 2x 5 + 27 (2)x 5 18x 4 57 (2)x 4 20x 3 + 63 (2)x 3
(52)
(53)
(54)
(55)
(56)
244

+ 20x 2 45 (2)x 2 + 18x + 78 (2)x 2 18 (2)

/ (x + 1)4 (x 1)3 x,

d11 = 18 (2) x 2 + 1 x 2 4x + 1 / (x + 1)2 (x 1)4 xH (1; x)

+ 24 (2) x 2 + 1 x 2 5x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 1; x)
+ 1/18 1083x 6 + 324 (2)x 6 238x 5 + 1008 (3)x 5 3384 (2)x 5
(57)
+ 3960 (2)x 4 1083x 4 4752 (3)x 4 + 476x 3 + 9216 (3)x 3

7056 (2)x 3 + 1620 (2)x 2 1083x 2 4752 (3)x 2 238x

+ 1008 (3)x 648 (2)x + 1083 / (x + 1)3 (x 1)5 xH (0; x)

4/3 18x 2 55x + 18 / (x + 1)(x 1)3 xH (1, 0; x)

48 x 2 x + 1 / (x + 1)2 (x 1)4 x 2 H (0, 1, 0; x)

+ 8 5x 4 14x 3 + 30x 2 14x + 5 / (x + 1)3 (x 1)5 x 2 H (0, 0, 1, 0; x)
2/3 3x 6 + 24 (2)x 5 + 103x 5 126 (2)x 4 27x 4 98x 3

+ 24 (2)x 3 + 57x 2 126 (2)x 2 + 24 (2)x 5x 33

/ (x + 1)3 (x 1)5 xH (0, 0; x)

2 9x 4 92x 3 + 130x 2 92x + 9 / (x + 1)2 (x 1)4 xH (1, 0, 0; x)

8 7x 4 29x 3 + 62x 2 29x + 7 / (x + 1)3 (x 1)5 x 2 H (0, 1, 0, 0; x)

+ 2 9x 4 112x 3 + 146x 2 196x + 9 / (x + 1)3 (x 1)5 x 3 H (0, 0, 0; x)

+ 4 x 4 5x 3 + 38x 2 5x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 0, 0, 0; x)

16 3x 2 8x + 3 / (x + 1)3 (x 1)3 x 2 H (1, 0, 0, 0; x)

24 3x 2 2x + 3 / (x + 1)2 (x 1)4 x 2 H (1, 0, 0; x)

+ 8 5x 4 14x 3 + 42x 2 14x + 5 / (x + 1)3 (x 1)5 x 2 H (0, 1, 0, 0; x)

4 x 4 + 12 (2)x 3 2x 2 32 (2)x 2 + 12 (2)x + 1

/ (x + 1)3 (x 1)3 xH (1, 0; x)

16/ (x + 1)2 (x 1)2 x 2 H (0, 1, 0; x)

32/ (x + 1)3 (x 1)3 x 3 H (0, 0, 1, 0; x)
+ 2/45 270 (3)x 6 945 (2)x 6 935x 6 + 540 (2) ln(2)x 6

3690 (3)x 5 594 (2)2 x 5 2160 (2) ln(2)x 5 + 5415 (2)x 5
+ 2805x 4 + 2439 (2)2 x 4 + 2700 (2) ln(2)x 4 + 5490 (3)x 4
3465 (2)x 4 2010 (2)x 3 1584 (2)2 x 3 2700 (2) ln(2)x 2
+ 2439 (2)2 x 2 2805x 2 5490 (3)x 2 + 4185 (2)x 2
594 (2)2 x + 3690 (3)x 3405 (2)x + 2160 (2) ln(2)x + 935

540 (2) ln(2) + 225 (2) 270 (3) / (x + 1)3 (x 1)5 x,
(58)
245

d12 = 24 x 2 + 1 x 3 3x 2 6x + 2 (2)/ (x + 1)3 (x 1)4 xH (1; x)

48 (2) x 2 + 1 x 2 5x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 1; x)
1/2 84 (2)x 6 + 45x 6 310x 5 32 (3)x 5 168 (2)x 5

+ 467x 4 1116 (2)x 4 + 1056 (3)x 4 1664 (3)x 3 404x 3
+ 544 (2)x 3 + 467x 2 564 (2)x 2 + 1056 (3)x 2 + 104 (2)x

32 (3)x 310x + 45 36 (2) / (x + 1)3 (x 1)5 xH (0; x)

4 13x 2 + 46x + 13 / (x + 1)3 (x 1) xH (1, 0; x)

24 x 2 + 1 / (x + 1)3 (x 1) xH (1, 1, 0; x)

+ 4 9x 5 29x 4 + 114x 3 + 90x 2 11x + 3

/ (x + 1)3 (x 1)4 xH (0, 1, 0; x)

16 x 4 + 19x 3 28x 2 + 19x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 0, 1, 0; x)
+ 2 63x 6 78x 5 + 48 (2)x 4 + 13x 4 216 (2)x 3 + 188x 3

+ 48 (2)x 2 215x 2 + 50x 21 / (x + 1)3 (x 1)5 xH (0, 0; x)

8 6x 5 + 4x 4 + 85x 3 + 97x 2 5x + 9

/ (x + 1)3 (x 1)4 xH (1, 0, 0; x)

16 x 4 37x 3 + 54x 2 37x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 1, 0, 0; x)

2 3x 6 40x 5 329x 4 + 88x 3 7x 2 24x + 21

/ (x + 1)3 (x 1)5 xH (0, 0, 0; x)

8 x 4 5x 3 + 38x 2 5x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 0, 0, 0; x)

32 x 2 17x + 1 / (x + 1)3 (x 1)3 x 2 H (1, 0, 0, 0; x)

+ 8 3x 4 7x 3 + 64x 2 7x + 3 / (x + 1)2 (x 1)4 xH (1, 0, 0; x)

16 x 4 + 19x 3 16x 2 + 19x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 1, 0, 0; x)

+ 32 x 4 (2)x 3 + x 3 + 17 (2)x 2 (2)x + x + 1

/ (x + 1)3 (x 1)3 xH (1, 0; x)

8 3x 4 6x 3 + 14x 2 6x + 3 / (x + 1)2 (x 1)4 xH (0, 1, 0; x)

+ 64/ (x + 1)3 (x 1)3 x 3 H (0, 0, 1, 0; x)
1/5 75x 6 + 240 (2) ln(2)x 6 230 (2)x 6 120 (3)x 6

+ 560 (3)x 5 960 (2) ln(2)x 5 + 320x 5 + 8 (2)2 x 5 + 340 (2)x 5
865x 4 3760 (3)x 4 + 1200 (2) ln(2)x 4 1160 (2)2 x 4 + 790 (2)x 4
+ 480 (3)x 3 + 1400 (2)x 3 + 3240 (2)2 x 3
1200 (2) ln(2)x 2 + 2920 (3)x 2 + 865x 2
1160 (2)2 x 2 2330 (2)x 2 140 (2)x 320x + 8 (2)2 x
246
80 (3)x + 960 (2) ln(2)x 75 + 170 (2)

240 (2) ln(2) / (x + 1)3 (x 1)5 x,
(59)
with (n) being Riemanns -function.

5.2. Contributions from the counterterms
The terms in Fig. 5(a)(f) are defined as one-loop vertex amplitudes in D dimensions
including the corresponding one-loop renormalization insertions. They read explicitly:
Fig. 5(a) and (b):
=
def.
2
S
CF C()
2
m2

N1
DD k
,
2
2
2
[(p1 + k) m ] [(p2 k)2 m2 ]k 2
(60)
Fig. 5. Subtraction terms for the two-loop renormalization. Note that the diagrams (a)(b) and (c)(d) yield the
same contribution, respectively.
247
with
(1l)

N1 = (/k + p
/ 1 + m) 2 (/k + p
/ 1 ) m (/k + p
/ 1 + m)

Q
/ 2 + m)
v + a Q 5 (/k p
(61)
and
DD k =
2
m
dDk
1
.
2
C()
(2)D2
(62)
The contributions from these two diagrams to the axial vector form factors are

2
(a,b)
s, , 2
G1
m
2 2
2
S
CF2 C 2 ()
=
2
m2

1
2 3/2(x 1) x 2 + x + 1 /(x + 1)3 H (0; x)

3/4(x 1)2 /(x + 1)2

1
+ 1/4 17x 4 + 4x 3 + 6x 2 + 4x + 17 / (x 1)(x + 1)3 H (0; x)

3(x 1) x 2 + x + 1 /(x + 1)3 H (1, 0; x)

+ 3/2(x 1) x 2 + x + 1 /(x + 1)3 H (0, 0; x)

1/2 x 2 + x + 1 3 (2)x 4x 3 (2) 4 /(x + 1)3

+ 3(x 1) x 2 + x + 1 (2)/(x + 1)3 H (1; x)
1/2 26x 4 + 3 (2)x 4 6x 3 3 (2)x 3 + 8x 2 3 (2)x

6x + 3 (2) 26 / (x 1)(x + 1)3 H (0; x)

1/2 17x 4 + 4x 3 + 6x 2 + 4x + 17 / (x 1)(x + 1)3 H (1, 0; x)

+ 6(x 1) x 2 + x + 1 /(x + 1)3 H (1, 1, 0; x)

3(x 1) x 2 + x + 1 /(x + 1)3 H (0, 1, 0; x)

+ 1/4 17x 4 + 4x 3 + 6x 2 + 4x + 17 / (x 1)(x + 1)3 H (0, 0; x)

3(x 1) x 2 + x + 1 /(x + 1)3 H (1, 0, 0; x)

+ 3/2(x 1) x 2 + x + 1 /(x + 1)3 H (0, 0, 0; x)
1/4 17 (2)x 4 + 12 (3)x 4 + 16x 4 16x 3 12 (3)x 3

+ 4 (2)x 3 + 6 (2)x 2 + 4 (2)x + 16x 12 (3)x + 17 (2)

+ 12 (3) 16 / (x 1)(x + 1)3 ,
(63)
248

(a,b)
G2
s, ,
2
m2

2 2
=
m2

1
2 3 x 2 + 1 / (x 1)(x + 1)3 xH (0; x) 3/(x + 1)2 x

1
13x 2 + 6x + 13 / (x 1)(x + 1)3 xH (0; x)

+

6 x 2 + 1 / (x 1)(x + 1)3 xH (1, 0; x)

+ 3 x 2 + 1 / (x 1)(x + 1)3 xH (0, 0; x)

3 (2)x 2 + 4x 2 4 + 3 (2) / (x 1)(x + 1)3 x

+ 6 x 2 + 1 (2)/ (x 1)(x + 1)3 xH (1; x)
41x 4 + 3 (2)x 4 + 20x 3 6 (2)x 3 + 6 (2)x 2 6x 2

6 (2)x + 20x 41 + 3 (2) / (x 1)3 (x + 1)3 xH (0; x)

2 13x 2 + 6x + 13 / (x 1)(x + 1)3 xH (1, 0; x)

+ 12 x 2 + 1 / (x 1)(x + 1)3 xH (1, 1, 0; x)

6 x 2 + 1 / (x 1)(x + 1)3 xH (0, 1, 0; x)

+ 13x 2 + 6x + 13 / (x 1)(x + 1)3 xH (0, 0; x)

6 x 2 + 1 / (x 1)(x + 1)3 xH (1, 0, 0; x)

+ 3 x 2 + 1 / (x 1)(x + 1)3 xH (0, 0, 0; x)
13 (2)x 3 + 6 (3)x 3 + 14x 3 7 (2)x 2 6 (3)x 2

+ 10x 2 + 6 (3)x + 7 (2)x + 10x 13 (2) + 14 6 (3)

/ (x + 1)3 (x 1)2 x .
S
2

2
CF2 C 2 ()
(64)
Fig. 5(c) and (d):

=
def.
= 1F
2
S
CF C()
= 1F
2
m2

N2
DD k
,
2
2
[(p1 + k) m ][(p2 k)2 m2 ]k 2
(65)
with

N2 = (/k + p
/ 2 + m) .
/ 1 + m) v Q + a Q 5 (/k p
(66)
These diagrams yield

2
(c,d)
s, , 2
G1
m
2
2
S
=
CF CA C 2 ()
2
m2

1
2 1/2 x 2 + 1 / (x 1)(x + 1) H (0; x) 1/4

1
+ 1/4 3x 2 2x + 3 / (x 1)(x + 1) H (0; x)

+ x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

1/2 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

+ 1/2 (2) x 2 + 1 / (x 1)(x + 1)

(2) x 2 + 1 / (x 1)(x + 1) H (1; x)

+ 1/2 x 2 + 1 (2) 4 / (x 1)(x + 1) H (0; x)

+ 1/2 3x 2 2x + 3 / (x 1)(x + 1) H (1, 0; x)

2 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

+ x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

1/4 3x 2 2x + 3 / (x 1)(x + 1) H (0, 0; x)

+ x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

1/2 x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)

2
2
+ 1/4 3 (2)x + 4 (3)x 2 (2)x + 4 (3) + 3 (2) / (x 1)(x + 1)

2 2
S 2 2 2
CF C ()
2
m2

1
2 3/2 x 2 + 1 / (x 1)(x + 1) H (0; x) 3/4

1
+ 1/4 17x 2 6x + 17 / (x 1)(x + 1) H (0; x)

+ 3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

3/2 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

+ 1/2 2x 2 + 3 (2)x 2 + 2 + 3 (2) / (x 1)(x + 1)

3 (2) x 2 + 1 / (x 1)(x + 1) H (1; x)

+ 1/2 3 (2)x 2 26x 2 + 4x + 3 (2) 26 / (x 1)(x + 1) H (0; x)

+ 1/2 17x 2 6x + 17 / (x 1)(x + 1) H (1, 0; x)

6 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)
249
250

+ 3 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

1/4 17x 2 6x + 17 / (x 1)(x + 1) H (0, 0; x)

+ 3 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

3/2 x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)
+ 1/4 17 (2)x 2 8x 2 + 12 (3)x 2 6 (2)x + 8

+ 17 (2) + 12 (3) / (x 1)(x + 1) ,

(c,d)
G2
s, ,

2
S 2
CF CA C 2 ()
2
m2

1
2
3x 2x + 3 / (x + 1)(x 1)3 xH (0; x)

+ 2/(x 1)2 x

2 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x)

+ 2 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)

+ 3 (2)x 2 + 4x 2 2 (2)x + 3 (2) 4 / (x + 1)(x 1)3 x

=
2
m2
(67)

2 2
S 2 2 2
CF C ()
2
m2

1
2
3 3x 2x + 3 / (x + 1)(x 1)3 xH (0; x) + 6/(x 1)2 x

10 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x)

+ 6 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)

2
2
3
+ 20x + 9 (2)x 6 (2)x 20 + 9 (2) / (x + 1)(x 1) x .
(68)
Fig. 5(e):
def.
= 3
2
S
CF C()
2
m2
DD k
[(p1
+ k)2
N3
2
m ][(p2
k)2 m2 ](k 2 )2
(69)
251
with

N3 = (/k + p
/ 2 + m)
/ 1 + m) v Q + a Q 5 (/k p

2
k g k k .
This leads to the contributions

2
(e)
G1 s, , 2
m
2
2
S
2
=
CF TR Nf C ()
2
m2

1
2 2/3 x 2 + 1 / (x 1)(x + 1) H (0; x) + 1/3

1
2
+ 1/3 3x 2x + 3 / (x 1)(x + 1) H (0; x)

4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

+ 2/3 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

2/3 (2) x 2 + 1 / (x 1)(x + 1)

+ 4/3 (2) x 2 + 1 / (x 1)(x + 1) H (1; x)

2/3 x 2 + 1 4 + (2) / (x 1)(x + 1) H (0; x)

2/3 3x 2 2x + 3 / (x 1)(x + 1) H (1, 0; x)

+ 8/3 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

4/3 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

+ 1/3 3x 2 2x + 3 / (x 1)(x + 1) H (0, 0; x)

4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

+ 2/3 x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)

2
2
1/3 3 (2)x + 4 (3)x 2 (2)x + 4 (3) + 3 (2) / (x 1)(x + 1)

2
S 2
CF TR C 2 ()
2
m2

1
2 2/3 x 2 + 1 / (x 1)(x + 1) H (0; x) + 1/3

1
2
+ 1/3 3x 2x + 3 / (x 1)(x + 1) H (0; x)

4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

+ 2/3 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

2/3 (2) x 2 + 1 / (x 1)(x + 1)

(70)
252

+ 4/3 (2) x 2 + 1 / (x 1)(x + 1) H (1; x)

2/3 x 2 + 1 4 + (2) / (x 1)(x + 1) H (0; x)

2/3 3x 2 2x + 3 / (x 1)(x + 1) H (1, 0; x)

+ 8/3 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

4/3 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

+ 1/3 3x 2 2x + 3 / (x 1)(x + 1) H (0, 0; x)

4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

+ 2/3 x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)

1/3 3 (2)x 2 + 4 (3)x 2 2 (2)x + 4 (3) + 3 (2) / (x 1)(x + 1)

2
S 2
2
CF CA C ()
+
2
m2

1
2 5/6 x 2 + 1 / (x 1)(x + 1) H (0; x) 5/12

1
+ 5/12 3x 2 2x + 3 / (x 1)(x + 1) H (0; x)

+ 5/3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

5/6 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

+ 5/6 (2) x 2 + 1 / (x 1)(x + 1)

5/3 (2) x 2 + 1 / (x 1)(x + 1) H (1; x)

+ 5/6 x 2 + 1 4 + (2) / (x 1)(x + 1) H (0; x)

+ 5/6 3x 2 2x + 3 / (x 1)(x + 1) H (1, 0; x)

10/3 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

+ 5/3 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

5/12 3x 2 2x + 3 / (x 1)(x + 1) H (0, 0; x)

+ 5/3 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

5/6 x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)

+ 5/12 3 (2)x 2 + 4 (3)x 2 2 (2)x + 4 (3) + 3 (2)

/ (x 1)(x + 1) ,

(e)
G2

2
s, , 2
m
2
2
S
=
CF TR Nf C 2 ()
2
m2
(71)
253

1
2
4/3 3x 2x + 3 / (x + 1)(x 1)3 xH (0; x) 8/3/(x 1)2 x

+ 8/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x)

8/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

+ 4/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)

2

2
3
4/3 4x + 3 (2)x 2 (2)x 4 + 3 (2) / (x + 1)(x 1) x

2
S 2
2
+
CF TR C ()
2
m2

1
2
4/3 3x 2x + 3 / (x + 1)(x 1)3 xH (0; x) 8/3/(x 1)2 x

+ 8/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x)

8/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

+ 4/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)

4/3 4x 2 + 3 (2)x 2 2 (2)x 4 + 3 (2) / (x + 1)(x 1)3 x
2
CF CA C 2 ()
m2

1
5/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x) + 10/3/(x 1)2 x

10/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x)

+ 10/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

5/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)

+ 5/3 4x 2 + 3 (2)x 2 2 (2)x 4 + 3 (2) / (x + 1)(x 1)3 x .
(72)
S
2
2
Fig. 5(f):
= 2
(1l)
def. (1l)
= 2
2
S
CF C()
2
m2
DD k
[(p1
+ k)2
N2
2
m ][(p
k)2 m2 ]k 2
(73)
254
with N2 as defined in Eq. (66). They contribute

2
(f )
G1 s, , 2
m
2
2 2
S
2 2
=
CF C ()
2
m2

1
2 3/2(x 2 + 1)/ (x 1)(x + 1) H (0; x) 3/4

1
+ 1/4 17x 2 6x + 17 / (x 1)(x + 1) H (0; x)

+ 3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

3/2 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)

+ 1/2 2x 2 + 3 (2)x 2 + 2 + 3 (2) / (x 1)(x + 1)

3 (2) x 2 + 1 / (x 1)(x + 1) H (1; x)

+ 1/2 26x 2 + 3 (2)x 2 + 4x + 3 (2) 26 / (x 1)(x + 1) H (0; x)

+ 1/2 17x 2 6x + 17 / (x 1)(x + 1) H (1, 0; x)

6 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

+ 3 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

1/4 17x 2 6x + 17 / (x 1)(x + 1) H (0, 0; x)

+ 3 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

3/2 x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)
+ 1/4 17 (2)x 2 8x 2 + 12 (3)x 2 6 (2)x

+ 12 (3) + 8 + 17 (2) / (x 1)(x + 1) ,

2
s, , 2
m
2
2 2
S
=
CF2 C 2 ()
2
m2

1
2
3 3x 2x + 3 / (x + 1)(x 1)3 xH (0; x) + 6/(x 1)2 x

10 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0; x)

+ 6 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)

2
2
3
+ 9 (2)x + 20x 6 (2)x 20 + 9 (2) / (x + 1)(x 1) x .
(74)
(f )
G2
(75)

(2l)
Fig. 5(g) is defined as the two-loop renormalization constant 2

amplitude and reads:
def. (2l)
= 2
(2l)
= 2
255
times the Born level
(i) v Q + a Q 5 .
Its contribution to G1 is

2
(g)
(2l)
G1 s, , 2 = 2 .
m
(76)
(77)
5.3. UV-renormalized two-loop form factors

Adding the terms given in Sections 5.1 and 5.2 we obtain the UV-renormalized axial
vector form factors to second order in QCD. They still contain terms proportional to 2
and 1 due to infrared and collinear singularities in the loops. In this subsection we put
the renormalization scale equal to the on-shell mass m. (The logarithms that are present
if = m are given in Section 5.4.) Then we get

1
(2l)
G1,R (s, ) = C 2 () 2 CF TR Nf c1 + CF TR c2 + CF CA c3 + CF2 c4

1

2
+ CF TR Nf c9 + CF TR c10 + CF CA c11 + CF c12 + O() ,
(78)

1
(2l)
G2,R (s, ) = C 2 () 2 CF TR Nf d1 + CF TR d2 + CF CA d3 + CF2 d4

1
+ CF TR Nf d5 + CF TR d6 + CF CA d7 + CF2 d8

+ CF TR Nf d9 + CF TR d10 + CF CA d11 + CF d12 + O() ,

(79)
where:

4
3
c3 = c5 = 1/3 x 2 + 1 / (x 1)(x + 1) H (0; x) 1/3,
11
5
c2 = c6 = 0,

c4 = x 2 + 1 / (x 1)(x + 1) H (0; x)

2
+ x 2 + 1 / (x + 1)2 (x 1)2 H (0, 0; x) + 1/2,

c7 = 1/36 x 2 + 1 67x 2 + 54 (2)x 2 + 67 18 (2)

/ (x + 1)2 (x 1)2 H (0; x)

x 2 + 1 / (x 1)(x + 1) H (1, 0; x)
c1 =
(80)
(81)
(82)
256
2
+ x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 0; x)

+ 2/ (x 1)(x + 1) x 2 H (0, 0; x)

2 x 2 + 1 / (x + 1)2 (x 1)2 x 2 H (0, 0, 0; x)

+ x 2 + 1 / (x 1)(x + 1) H (1, 0; x)
2
x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 0; x)
+ 1/36 54 (2)x 4 49x 4 18 (3)x 4 36 (3)x 2 + 98x 2

72 (2)x 2 49 + 18 (2) 18 (3) / (x + 1)2 (x 1)2 ,
(83)

4
4
3
2
c8 = 1/2 2 (2)x + 7x 2x + 4 (2)x + 2x + 2 (2) 7

/ (x + 1)2 (x 1)2 H (0; x)

+ 2 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

2
2 x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 0; x)

+ 2 x 2 + 1 x 2 x + 2 / (x + 1)2 (x 1)2 H (0, 0; x)
2
4 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 0; x)

2
+ 3 x 2 + 1 / (x + 1)2 (x 1)2 H (0, 0, 0; x)

+ 2x 2 + (2)x 2 2 + (2) / (x 1)(x + 1) ,
(84)
2

c9 = 4/3 (2) x + 1 / (x 1)(x + 1) H (1; x)

1/54 36 (2)x 2 + 209x 2 78x + 36 (2) + 209 / (x 1)(x + 1) H (0; x)

+ 2/9 19x 2 6x + 19 / (x 1)(x + 1) H (1, 0; x)

8/3 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

+ 4/3 x 2 + 1 / (x 1)(x + 1) H (0, 1, 0; x)

1/9 19x 2 6x + 19 / (x 1)(x + 1) H (0, 0; x)

+ 4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0, 0; x)

2/3 x 2 + 1 / (x 1)(x + 1) H (0, 0, 0; x)
+ 1/27 93 (2)x 2 + 106x 2 + 36 (3)x 2 18 (2)x

+ 36 (3) 106 + 21 (2) / (x 1)(x + 1) ,
(85)
c10 = 1/54 36 (2)x 4 + 265x 4 + 72 (2)x 3 208x 3 144 (2)x 2

+ 14x 2 + 72 (2)x 208x + 36 (2) + 265 / (x + 1)3 (x 1) H (0; x)

+ 1/9 19x 4 14x 3 + 14x 2 14x + 19 /(x + 1)4 H (0, 0; x)

2/3 x 4 + 2x 3 4x 2 + 2x + 1 / (x + 1)3 (x 1) H (0, 0, 0; x)
1/27 383x 4 + 27 (2)x 4 + 1026 (2)x 3 1412x 3 2058x 2

+ 1278 (2)x 2 + 1026 (2)x 1412x 383 + 27 (2) /(x + 1)4 ,
(86)
257

c11 = 2/3 34x 4 + 3x 3 33x 2 + 3x 1 / (x + 1)2 (x 1)2 H (0, 1, 0; x)
2

4 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 1, 0; x)

+ 2 11x 6 + 5x 5 15x 4 + 26x 3 27x 2 + 5x 1

/ (x + 1)3 (x 1)3 H (0, 0, 1, 0; x)

2
+ 4 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 1, 0; x)

6 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

2
+ 6 x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 1, 0; x)
+ 1/36 36 (2)x 6 217x 6 144 (2)x 5 282x 5 + 612 (2)x 4

+ 451x 4 504 (2)x 3 + 348x 3 467x 2 + 540 (2)x 2 66x

144 (2)x 36 (2) + 233 / (x + 1)3 (x 1)3 H (0, 0; x)

1/3 53x 4 141x 3 + 228x 2 141x 17

/ (x + 1)2 (x 1)2 H (1, 0, 0; x)

+ 2 7x 6 7x 5 + 33x 4 56x 3 + 25x 2 7x 1

/ (x + 1)3 (x 1)3 H (0, 1, 0, 0; x)

+ 1/6 89x 6 402x 5 + 499x 4 354x 3 + 37x 2 24x + 11

/ (x + 1)3 (x 1)3 H (0, 0, 0; x)

+ 2 x 2 + 3 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 0, 0; x)

+ 2 x 5 5x 4 + 18x 3 26x 2 + x 5 / (x + 1)3 (x 1)2 H (1, 0, 0, 0; x)

2 2x 4 + 5x 3 14x 2 + 5x + 6 / (x + 1)2 (x 1)2 H (1, 0, 0; x)

12x 5 x 4 + 10x 3 28x 2 2x 1 / (x + 1)3 (x 1)3 xH (0, 0, 0, 0; x)

2 x 6 5x 5 + 21x 4 34x 3 + 17x 2 5x 3

/ (x + 1)3 (x 1)3 H (0, 1, 0, 0; x)
+ 3x 5 + 4 (2)x 5 x 4 8 (2)x 4 + 40 (2)x 3 + 2x 3 48 (2)x 2

2x 2 + x + 4 (2)x + 3 8 (2) / (x + 1)3 (x 1)2 H (1, 0; x)

6 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

2
+ 6 x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 1, 0; x)

2
+ 4 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 1, 0; x)

2 7x 5 + 7x 4 + 12x 3 + 4x 2 + x + 1 / (x + 1)3 (x 1)2 H (0, 0, 1, 0; x)

2
4 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 1, 0; x)

+ 2 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)

2
2 x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 1, 0; x)
258

+ 8 x 2 + 1 / (x + 1)2 (x 1) xH (0, 1, 0; x)

x 2 + 1 (2)x 2 + 2 (3)x 2 (2) + 2 (3) / (x + 1)2 (x 1)2 H (1; x)
+ 1/540 1917 (2)2 x 6 + 3240 (2) ln(2)x 6 7975x 6 5205 (2)x 6

+ 450 (3)x 6 3564 (2)2 x 5 22680 (3)x 5 + 33930 (2)x 5
19440 (2) ln(2)x 5 + 23925x 4 + 29160 (2) ln(2)x 4 + 36270 (3)x 4
+ 12447 (2)2 x 4 42405 (2)x 4 13716 (2)2 x 3 8460 (2)x 3
+ 14445 (2)2 x 2 29160 (2) ln(2)x 2 + 47505 (2)x 2 30690 (3)x 2
23925x 2 25470 (2)x + 22680 (3)x + 19440 (2) ln(2)x 3564 (2)2 x

6030 (3) 3240 (2) ln(2) + 7975 + 81 (2)2 + 105 (2)

/ (x + 1)3 (x 1)3
2
+ (2) x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1; x)

+ 52/3 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)
1/3 10 (2)x 4 6 (3)x 4 81 (2)x 3 12 (3)x 2 + 108 (2)x 2

81 (2)x + 26 (2) 6 (3) / (x + 1)2 (x 1)2 H (1; x)
+ 1/216 1044 (2)x 6 1620 (3)x 6 + 2545x 6 + 3024 (3)x 5 570x 5

12528 (2)x 5 13716 (3)x 4 2545x 4 + 18396 (2)x 4 12744 (2)x 3
+ 1140x 3 + 25920 (3)x 3 + 3060 (2)x 2 2545x 2 13500 (3)x 2 570x
2808 (2)x + 3024 (3)x 1404 (3)

+ 396 (2) + 2545 / (x + 1)3 (x 1)3 H (0; x)

+ (2) 7x 6 + 6x 5 29x 4 + 24x 3 31x 2 + 6x + 5

/ (x + 1)3 (x 1)3 H (0, 1; x)

1/9 31x 4 105x 3 36 (2)x 2 + 105x 36 (2) 31

/ (x + 1)2 (x 1)2 H (1, 0; x)
2
10 x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 1, 0; x),
(87)
2
2
c12 = 8 x + 1 / (x + 1)2 (x 1)2 H (1, 0, 1, 0; x)

2
8 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 1, 0; x)

2
+ 8 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 1, 0; x)

+ 2 (2) 5x 4 27x 3 + 36x 2 27x + 7 / (x + 1)2 (x 1)2 H (1; x)

2 (2) 5x 6 + 6x 5 31x 4 + 24x 3 29x 2 + 6x + 7

/ (x + 1)3 (x 1)3 H (0, 1; x)
1/8 85x 6 64 (3)x 6 + 136 (2)x 6 160 (3)x 5 6x 5 296 (2)x 5

+ 1056 (3)x 4 85x 4 568 (2)x 4 + 944 (2)x 3 + 12x 3 1536 (3)x 3
259
85x 2 616 (2)x 2 + 1056 (3)x 2 + 24 (2)x 160 (3)x 6x 64 (3)

+ 85 8 (2) / (x + 1)3 (x 1)3 H (0; x)

+ 1/2 55x 4 + 8 (2)x 4 + 82x 3 + 16 (2)x 2 82x + 8 (2) + 55

/ (x + 1)2 (x 1)2 H (1, 0; x)

+ 2 8x 4 27x 3 + 60x 2 27x + 6 / (x + 1)2 (x 1)2 H (0, 1, 0; x)

2 11x 6 2x 5 + 33x 4 56x 3 + 23x 2 2x + 1

/ (x + 1)3 (x 1)3 H (0, 0, 1, 0; x)

4 x 2 + 1 / (x 1)(x + 1) H (1, 1, 0; x)
+ 1/4 229x 6 + 52 (2)x 6 266x 5 16 (2)x 5 + 124 (2)x 4

+ 15x 4 320 (2)x 3 + 196x 3 + 84 (2)x 2 189x 2 + 70x

16 (2)x 55 + 12 (2) / (x + 1)3 (x 1)3 H (0, 0; x)

2 3x 4 15x 3 + 76x 2 15x + 5 / (x + 1)2 (x 1)2 H (1, 0, 0; x)

2
+ 16 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 1, 0, 0; x)

2 7x 6 + 10x 5 65x 4 + 104x 3 75x 2 + 10x 3

/ (x + 1)3 (x 1)3 H (0, 1, 0, 0; x)

2 5x 6 10x 5 60x 4 + 59x 3 16x 2 7x + 5

/ (x + 1)3 (x 1)3 H (0, 0, 0; x)

2
12 x 2 + 1 / (x + 1)2 (x 1)2 H (1, 0, 0, 0; x)

+ 27x 6 2x 5 + 27x 4 56x 3 7x 2 2x 7

/ (x + 1)3 (x 1)3 H (0, 0, 0, 0; x)

+ 8 x 5 2x 4 + 18x 3 14x 2 + 4x + 1 / (x + 1)3 (x 1)2 H (1, 0, 0, 0; x)

+ 2 8x 4 19x 3 + 46x 2 19x + 8 / (x + 1)2 (x 1)2 H (1, 0, 0; x)

4 x 6 x 5 + 14x 4 20x 3 + 14x 2 x + 1

/ (x + 1)3 (x 1)3 H (0, 1, 0, 0; x)
+ 4 (2)x 5 + 4x 5 + 2x 4 5 (2)x 4 + 34 (2)x 3 2x 3 + 2x 2

30 (2)x 2 2x + 7 (2)x 4 + (2) / (x + 1)3 (x 1)2 H (1, 0; x)

2 5x 4 10x 3 + 18x 2 10x + 5 / (x + 1)2 (x 1)2 H (0, 1, 0; x)

+ 4 3x 5 + 3x 4 + 6x 3 2x 2 x 1 / (x + 1)3 (x 1)2 H (0, 0, 1, 0; x)

2
+ 4 (3) x 2 + 1 / (x + 1)2 (x 1)2 H (1; x)
1/20 362 (2)2 x 6 + 240 (2) ln(2)x 6 230x 6 275 (2)x 6 + 140 (3)x 6
+ 144 (2)2 x 5 + 600 (3)x 5 + 230 (2)x 5 1440 (2) ln(2)x 5 + 690x 4
+ 2160 (2) ln(2)x 4 2980 (3)x 4 1198 (2)2 x 4 + 255 (2)x 4
260
+ 2656 (2)2 x 3 + 500 (2)x 3 954 (2)2 x 2 2160 (2) ln(2)x 2

165 (2)x 2 + 3060 (3)x 2 690x 2 730 (2)x 600 (3)x
+ 1440 (2) ln(2)x + 144 (2)2 x 220 (3) 240 (2) ln(2) + 230

118 (2)2 + 185 (2) / (x + 1)3 (x 1)3

2
+ 4 x 2 + 1 / (x + 1)2 (x 1)2 H (0, 1, 1, 0; x),
(88)
dj = 0 for j = 1, . . . , 7,

d8 = 2 5x 2 2x + 5 / (x 1)3 (x + 1) xH (0; x)

+ 4 3x 2 2x + 3 x 2 + 1 / (x + 1)2 (x 1)4 xH (0, 0; x)
(89)
+ 4/(x 1)2 x,

d9 = 2/9 51x 2 26x + 51 / (x + 1)(x 1)3 xH (0; x)

+ 8/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

4/3 3x 2 2x + 3 / (x + 1)(x 1)3 xH (0, 0; x)

+ 4/9 19x 2 + 9 (2)x 2 6 (2)x 19 + 9 (2)

/ (x + 1)(x 1)3 x,

d10 = 2/9 51x 4 176x 3 70x 2 72 (2)x 2 176x + 51

/ (x + 1)3 (x 1)3 xH (0; x)

+ 4/3 3x 4 14x 3 2x 2 14x + 3 / (x + 1)4 (x 1)2 xH (0, 0; x)

+ 16/ (x + 1)3 (x 1)3 x 3 H (0, 0, 0; x)
+ 4/9 8x 4 + 45 (2)x 4 90 (2)x 3 80x 3 144x 2 + 18 (2)x 2

90 (2)x 80x 8 + 45 (2) / (x + 1)4 (x 1)2 x,

d11 = 18 (2) x 2 4x + 1 x 2 + 1 / (x + 1)2 (x 1)4 xH (1; x)

+ 24 (2) x 2 + 1 x 2 5x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 1; x)
+ 1/18 324 (2)x 6 + 687x 6 + 1008 (3)x 5 3384 (2)x 5 + 26x 5
(90)
(91)
(92)
4752 (3)x 4 + 3960 (2)x 4 687x 4 7056 (2)x 3

+ 9216 (3)x 3 52x 3 687x 2 + 1620 (2)x 2

4752 (3)x 2 + 1008 (3)x + 26x 648 (2)x + 687

/ (x + 1)3 (x 1)5 xH (0; x)

2/3 3x 2 88x + 3 / (x + 1)(x 1)3 xH (1, 0; x)

48 x 2 x + 1 / (x + 1)2 (x 1)4 x 2 H (0, 1, 0; x)

+ 8 5x 4 14x 3 + 30x 2 14x + 5 / (x + 1)3 (x 1)5 x 2 H (0, 0, 1, 0; x)
1/3 39x 6 + 48 (2)x 5 + 184x 5 252 (2)x 4 87x 4 + 48 (2)x 3

152x 3 + 81x 2 252 (2)x 2 32x + 48 (2)x 33
261

/ (x + 1)3 (x 1)5 xH (0, 0; x)

2 9x 4 92x 3 + 130x 2 92x + 9 / (x + 1)2 (x 1)4 xH (1, 0, 0; x)

8 7x 4 29x 3 + 62x 2 29x + 7 / (x + 1)3 (x 1)5 x 2 H (0, 1, 0, 0; x)

+ 2 9x 4 112x 3 + 146x 2 196x + 9 / (x + 1)3 (x 1)5 x 3 H (0, 0, 0; x)

+ 4 x 4 5x 3 + 38x 2 5x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 0, 0, 0; x)

16 3x 2 8x + 3 / (x + 1)3 (x 1)3 x 2 H (1, 0, 0, 0; x)

24 3x 2 2x + 3 / (x + 1)2 (x 1)4 x 2 H (1, 0, 0; x)

+ 8 5x 4 14x 3 + 42x 2 14x + 5 / (x + 1)3 (x 1)5 x 2 H (0, 1, 0, 0; x)

4 x 4 + 12 (2)x 3 32 (2)x 2 2x 2 + 12 (2)x + 1

/ (x + 1)3 (x 1)3 xH (1, 0; x)

16/ (x + 1)2 (x 1)2 x 2 H (0, 1, 0; x)

32/ (x + 1)3 (x 1)3 x 3 H (0, 0, 1, 0; x)
+ 1/45 1080 (2) ln(2)x 6 + 540 (3)x 6 1395 (2)x 6 1210x 6

7380 (3)x 5 4320 (2) ln(2)x 5 1188 (2)2 x 5 + 10500 (2)x 5
+ 3630x 4 + 10980 (3)x 4 + 4878 (2)2 x 4 7425 (2)x 4
+ 5400 (2) ln(2)x 4 3360 (2)x 3 3168 (2)2 x 3 + 4878 (2)2 x 2
10980 (3)x 2 + 7875 (2)x 2 3630x 2 5400 (2) ln(2)x 2
+ 4320 (2) ln(2)x 1188 (2)2 x 7140 (2)x + 7380 (3)x 540 (3)

1080 (2) ln(2) + 1210 + 945 (2) / (x + 1)3 (x 1)5 x,
(93)
2

2

2
4
d12 = 36 (2) x 4x + 1 x + 1 / (x + 1) (x 1) xH (1; x)

48 (2) x 2 + 1 x 2 5x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 1; x)
1/2 61x 6 + 96 (2)x 6 22x 5 216 (2)x 5 32 (3)x 5

1032 (2)x 4 61x 4 + 1056 (3)x 4 + 448 (2)x 3 + 44x 3
1664 (3)x 3 + 1056 (3)x 2 480 (2)x 2 61x 2

32 (3)x 22x + 56 (2)x + 61 24 (2) / (x + 1)3 (x 1)5 xH (0; x)

2 25x 2 86x + 25 / (x 1)3 (x + 1) xH (1, 0; x)

+ 8 3x 4 13x 3 + 64x 2 13x + 3 / (x + 1)2 (x 1)4 xH (0, 1, 0; x)

16 x 4 + 19x 3 28x 2 + 19x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 0, 1, 0; x)
+ 125x 6 230x 5 + 187x 4 + 96 (2)x 4 432 (2)x 3 + 204x 3

+ 96 (2)x 2 269x 2 + 26x 43 / (x + 1)3 (x 1)5 xH (0, 0; x)

4 15x 4 16x 3 + 198x 2 16x + 15 / (x + 1)2 (x 1)4 xH (1, 0, 0; x)

16 x 4 37x 3 + 54x 2 37x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 1, 0, 0; x)
262

+ 4 14x 5 + 175x 4 56x 3 + 14x 2 + 6x 9

/ (x + 1)3 (x 1)5 xH (0, 0, 0; x)

8 x 4 5x 3 + 38x 2 5x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 0, 0, 0; x)

32 x 2 17x + 1 / (x + 1)3 (x 1)3 x 2 H (1, 0, 0, 0; x)

+ 8 3x 4 7x 3 + 64x 2 7x + 3 / (x + 1)2 (x 1)4 xH (1, 0, 0; x)

16 x 4 + 19x 3 16x 2 + 19x + 1 / (x + 1)3 (x 1)5 x 2 H (0, 1, 0, 0; x)

+ 32 x 4 (2)x 3 + x 3 + 17 (2)x 2 (2)x + x + 1

/ (x + 1)3 (x 1)3 xH (1, 0; x)

8 3x 4 6x 3 + 14x 2 6x + 3 / (x + 1)2 (x 1)4 xH (0, 1, 0; x)

+ 64/ (x + 1)3 (x 1)3 x 3 H (0, 0, 1, 0; x)
1/5 240 (2) ln(2)x 6 60 (3)x 6 235 (2)x 6 85x 6 + 320 (3)x 5
960 (2) ln(2)x 5 + 8 (2)2 x 5 30 (2)x 5 + 255x 4
3340 (3)x 4 1160 (2)2 x 4 + 1595 (2)x 4 + 1200 (2) ln(2)x 4
+ 540 (2)x 3 + 3240 (2)2 x 3 1160 (2)2 x 2 + 3340 (3)x 2 1525 (2)x 2
255x 2 1200 (2) ln(2)x 2 + 960 (2) ln(2)x + 8 (2)2 x
510 (2)x 320 (3)x + 60 (3) 240 (2) ln(2)

+ 85 + 165 (2) / (x + 1)3 (x 1)5 x.
(94)
5.4. Form factors for = m

In this section we give the expressions for the renormalized axial vector two-loop form
factors for the case of = m.
At the one-loop level we do not have an explicit dependence on the logarithm of the
ratio of the renormalization scale and the mass of the heavy quark, because an overall
factor (2 /m2 ) can be taken out, see Eqs. (35), (36).
At the two-loop level, such a dependence results from the coupling constant renormalization, first appearing at this level. Factoring an overall (2 /m2 )2 , we have

2
, s, 2
m
2
2 2
2
(2l)
(2l)
(2l)
2
2
Gi,R (, s) + Oi (, s) ln
+ Pi (s) ln
, (95)
= C ()
m2
m2
m2
(2l)
Gi,R
(2l)
(2l)
where the functions Gi,R (, s) are given in Eqs. (78), (79) and the functions Oi
(, s) and
(2l)
Pi (s) can be obtained either from the renormalization group [1] or by explicit expansion.
263
We find

(2l)
O1 (, s) = CF TR (Nf + 1) 11/4Nc

2
2
x + 1 / (x 1)(x + 1) H (0; x) 1
3

+ 1/3 3x 2 2x + 3 / (x 1)(x + 1) H (0; x)

4/3 x 2 + 1 / (x 1)(x + 1) H (1, 0; x)

+ 2/3 x 2 + 1 / (x 1)(x + 1) H (0, 0; x)
2

2
2/3 x (2) + 2x + (2) 2 / (x 1)(x + 1) ,

(2l)
P1 (s) = CF TR (Nf + 1) 11/4Nc

1/3 x 2 + 1 / (x 1)(x + 1) H (0; x) 1 ,

(2l)
O2 (, s) = CF TR (Nf + 1) 11/4Nc

4/3 3x 2 2x + 3 / (x 1)3 (x + 1) xH (0; x)

2(x 1)2 x ,
(2l)
P2 (s) = 0.
(96)
(97)
(98)
(99)
6. Analytical continuation above threshold

The results for the renormalized form factors can be analytically continued into the
timelike region s > 0 and in particular above the physical threshold s > 4 by using the
substitution [1]
x y + i,
(100)
s s 4
y=
.
s + s 4
(101)
with
For s > 4 the form factors become complex due to absorptive parts. We write:
G1 (s + i, ) = G1 (s, ) + iG1 (s, ),
(102)
G1 (s + i, ) = G1 (s, ) + iG1 (s, ).
(103)
These imaginary parts arise from the analytical continuation of the harmonic polylogarithms with rightmost index 0. In the following two subsections we shall give the real and
imaginary parts of G1,2 at one and two loops, putting = m.
264
6.1. One-loop form factors above threshold

The analytical continuation of the coefficients a i and bi in Eqs. (35), (36) is

a 1 = y 2 + 1 / (y 1)(y + 1) H (0; y) 1,

a 2 = 1/2 3y 2 + 2y + 3 / (y 1)(y + 1) H (0; y)

+ y 2 + 1 / (y 1)(y + 1) H (0, 0; y)

+ 2 y 2 + 1 / (y 1)(y + 1) H (1, 0; y)

2 y 2 + 2 (2)y 2 1 + 2 (2) / (y 1)(y + 1) ,

a 3 = 4 y 2 + 1 1 + (2) / (y 1)(y + 1) H (0; y)

+ 1/2 3y 2 + 2y + 3 / (y 1)(y + 1) H (0, 0; y)

+ y 2 + 1 / (y 1)(y + 1) H (0, 0, 0; y)

+ 2 y 2 + 1 / (y 1)(y + 1) H (1, 0, 0; y)

+ 3y 2 + 2y + 3 / (y 1)(y + 1) H (1, 0; y)

+ 2 y 2 + 1 / (y 1)(y + 1) H (0, 1, 0; y)

+ 4 y 2 + 1 / (y 1)(y + 1) H (1, 1, 0; y)

8 (2) y 2 + 1 / (y 1)(y + 1) H (1; y)

2 (3)y 2 + 3 (2)y 2 + 2y 2 + 2 (2)y 2 + 3 (2) + (3)

/ (y 1)(y + 1) ,

a 1 = y 2 + 1 / (y 1)(y + 1) ,

a 2 = y 2 + 1 / (y 1)(y + 1) H (0; y)

+ 2 y 2 + 1 / (y 1)(y + 1) H (1; y)

+ 1/2 3y 2 + 2y + 3 / (y 1)(y + 1) ,

a 3 = 1/2 3y 2 + 2y + 3 / (y 1)(y + 1) H (0; y)

+ y 2 + 1 / (y 1)(y + 1) H (0, 0; y)

+ 2 y 2 + 1 / (y 1)(y + 1) H (1, 0; y)

+ 3y 2 + 2y + 3 / (y 1)(y + 1) H (1; y)

+ 2 y 2 + 1 / (y 1)(y + 1) H (0, 1; y)

+ 4 y 2 + 1 / (y 1)(y + 1) H (1, 1; y)

2 y 2 + 1 (2) 2 / (y 1)(y + 1) ,

b2 = 2 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (0; y) + 4/(y + 1)2 y,

b3 = 4 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (0; y)

2 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (0, 0; y)
(104)
(105)
(106)
(107)
(108)
(109)
(110)

4 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (1, 0; y)

+ 8 y 2 + 3 (2)y 2 + 2 (2)y + 3 (2) 1 / (y 1)(y + 1)3 y,

b2 = 2 3y 2 + 2y + 3 / (y 1)(y + 1)3 y,

b3 = 2 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (0; y)

4 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (1; y)

4 3y 2 + 2y + 3 / (y 1)(y + 1)3 y.
265
(111)
(112)
(113)
The above expressions for the singular term and the term of order 0 of G1 agree with
those of [21].
6.2. Two-loop form factors above threshold
The real and imaginary parts of the coefficients ci and di in Eq. (78) are

4
3
c3 = c5 = 1/3 y 2 + 1 / (y 1)(y + 1) H (0; y) 1/3,
11
5
c2 = c6 = 0,

c4 = y 2 + 1 / (y 1)(y + 1) H (0; y)
2
+ y 2 + 1 / (y 1)2 (y + 1)2 H (0, 0; y)

1/2 y 4 + 6 (2)y 4 + 2y 2 + 12 (2)y 2 1 + 6 (2)

/ (y 1)2 (y + 1)2 ,

c7 = 1/36 y 2 + 1 67y 2 + 162 (2)y 2 67 + 18 (2)

/ (y 1)2 (y + 1)2 H (0; y)

y 2 + 1 / (y + 1)(y 1) H (1, 0; y)
2
+ y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 0; y)

+ 2/ (y + 1)(y 1) y 2 H (0, 0; y)

2 y 2 + 1 / (y 1)2 (y + 1)2 y 2 H (0, 0, 0; y)

+ y 2 + 1 / (y + 1)(y 1) H (1, 0; y)

2
y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 0; y)
1/36 49y 4 + 18 (3)y 4 + 162 (2)y 4 144 (2)y 2 + 36 (3)y 2

98y 2 18 (2) + 18 (3) + 49 / (y 1)2 (y + 1)2 ,

c8 = 1/2 20 (2)y 4 + 7y 4 + 2y 3 + 40 (2)y 2 2y 7 + 20 (2)

/ (y + 1)2 (y 1)2 H (0; y)

+ 2 y 2 + 1 y 2 + y + 2 / (y + 1)2 (y 1)2 H (0, 0; y)

2
+ 3 y 2 + 1 / (y + 1)2 (y 1)2 H (0, 0, 0; y)
c1 =
(114)
(115)
(116)
(117)
266
2
+ 4 y 2 + 1 / (y + 1)2 (y 1)2 H (1, 0, 0; y)

2 y 2 + 1 / (y + 1)(y 1) H (1, 0; y)

2
+ 2 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 0; y)

2
12 (2) y 2 + 1 / (y + 1)2 (y 1)2 H (1; y)
5 (2)y 4 2y 4 + 6 (2)y 3 + 18 (2)y 2 + 4y 2 + 6 (2)y

2 + 13 (2) / (y + 1)2 (y 1)2 ,
(118)

c9 = 1/54 72 (2)y 2 209y 2 78y + 72 (2) 209

/ (y 1)(y + 1) H (0; y)

1/9 19y 2 + 6y + 19 / (y 1)(y + 1) H (0, 0; y)

2/3 y 2 + 1 / (y 1)(y + 1) H (0, 0, 0; y)

4/3 y 2 + 1 / (y 1)(y + 1) H (1, 0, 0; y)

2/9 19y 2 + 6y + 19 / (y 1)(y + 1) H (1, 0; y)

4/3 y 2 + 1 / (y 1)(y + 1) H (0, 1, 0; y)

8/3 y 2 + 1 / (y 1)(y + 1) H (1, 1, 0; y)

+ 16/3 (2) y 2 + 1 / (y 1)(y + 1) H (1; y)
+ 2/27 18 (3)y 2 + 53y 2 + 132 (2)y 2 + 36 (2)y 53

+ 96 (2) + 18 (3) / (y 1)(y + 1) ,
(119)
c10 = 1/54 265y 4 + 72 (2)y 4 208y 3 144 (2)y 3 14y 2 288 (2)y 2

208y 144 (2)y 265 + 72 (2) / (y 1)3 (y + 1) H (0; y)

+ 1/9 19y 4 + 14y 3 + 14y 2 + 14y + 19 /(y 1)4 H (0, 0; y)

2/3 y 4 2y 3 4y 2 2y + 1 / (y 1)3 (y + 1) H (0, 0, 0; y)
1/27 383y 2 + 198 (2)y 2 504 (2)y + 646y

383 + 198 (2) /(y 1)2 ,
(120)
c11 = 11 (2)y 4 + 2 (3)y 4 + 30 (2)y 3 + 84 (2)y 2 + 4 (3)y 2

+ 30 (2)y 37 (2) + 2 (3) / (y 1)2 (y + 1)2 H (1; y)

(2) 7y 6 + 30y 5 + 127y 4 + 204y 3 + 101y 2 + 30y 19

/ (y 1)3 (y + 1)3 H (0, 1; y)
1/216 8568 (2)y 6 + 1620 (3)y 6 2545y 6 570y 5 + 3024 (3)y 5

+ 30888 (2)y 5 + 35496 (2)y 4 + 2545y 4 + 13716 (3)y 4 + 1140y 3
+ 25920 (3)y 3 + 25488 (2)y 3 + 2545y 2 + 936 (2)y 2 + 13500 (3)y 2

216 (2)y 570y + 3024 (3)y + 792 (2) 2545 + 1404 (3)

/ (y 1)3 (y + 1)3 H (0; y)
267
+ 3y 5 + 2 (2)y 5 + 22 (2)y 4 y 4 2y 3 + 68 (2)y 3 2y 2 + 108 (2)y 2

+ 2 (2)y y + 22 (2) + 3 / (y 1)3 (y + 1)2 H (1, 0; y)

+ 2 y 2 + 1 / (y + 1)(y 1) H (1, 1, 0; y)

2
2 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1, 0; y)

2
4 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1, 0; y)

+ 2 7y 5 7y 4 + 12y 3 4y 2 + y 1

/ (y 1)3 (y + 1)2 H (0, 0, 1, 0; y)

2
+ 4 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1, 0; y)

6 y 2 + 1 / (y + 1)(y 1) H (1, 1, 0; y)

2
+ 6 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1, 0; y)
+ 1/36 217y 6 + 1332 (2)y 6 + 282y 5 + 252 (2)y 5 + 1692 (2)y 4

+ 451y 4 348y 3 + 3528 (2)y 3 467y 2 + 324 (2)y 2 + 66y

+ 252 (2)y 36 (2) + 233 / (y 1)3 (y + 1)3 H (0, 0; y)

8 y 2 + 1 / (y 1)2 (y + 1) yH (0, 1, 0; y)

+ 2 2y 4 5y 3 14y 2 5y + 6 / (y 1)2 (y + 1)2 H (1, 0, 0; y)

+ 2 y 6 + 5y 5 + 21y 4 + 34y 3 + 17y 2 + 5y 3

/ (y 1)3 (y + 1)3 H (0, 1, 0, 0; y)

+ 1/6 89y 6 + 402y 5 + 499y 4 + 354y 3 + 37y 2 + 24y + 11

/ (y 1)3 (y + 1)3 H (0, 0, 0; y)

2 y 5 + 5y 4 + 18y 3 + 26y 2 + y + 5

/ (y 1)3 (y + 1)2 H (1, 0, 0, 0; y)

12y 5 + y 4 + 10y 3 + 28y 2 2y + 1

/ (y 1)3 (y + 1)3 yH (0, 0, 0, 0; y)

2 y 2 + 3 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 0, 0; y)

+ 1/3 53y 4 + 141y 3 + 228y 2 + 141y 17

/ (y 1)2 (y + 1)2 H (1, 0, 0; y)

2 7y 6 + 7y 5 + 33y 4 + 56y 3 + 25y 2 + 7y 1

/ (y 1)3 (y + 1)3 H (0, 1, 0, 0; y)
+ 1/9 54 (2)y 4 + 31y 4 + 105y 3 + 180 (2)y 2 105y

31 + 126 (2) / (y 1)2 (y + 1)2 H (1, 0; y)

6 y 2 + 1 / (y + 1)(y 1) H (1, 1, 0; y)

2
+ 6 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1, 0; y)
268

+ 2/3 34y 4 3y 3 33y 2 3y 1 / (y 1)2 (y + 1)2 H (0, 1, 0; y)

2
+ 4 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1, 0; y)

2 11y 6 5y 5 15y 4 26y 3 27y 2 5y 1

/ (y 1)3 (y + 1)3 H (0, 0, 1, 0; y)

2
4 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1, 0; y)

2
10 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1, 0; y)
1/3 6 (3)y 4 + 149 (2)y 4 + 342 (2)y 3 + 12 (3)y 2 + 576 (2)y 2

+ 342 (2)y 77 (2) + 6 (3) / (y 1)2 (y + 1)2 H (1; y)

+ (2) 35y 6 + 48y 5 + 227y 4 + 360y 3 + 181y 2 + 48y 11

/ (y 1)3 (y + 1)3 H (0, 1; y)
+ 1/540 450 (3)y 6 + 3240 (2) ln(2)y 6 7975y 6 13257 (2)2 y 6

+ 4560 (2)y 6 3726 (2)2 y 5 + 19440 (2) ln(2)y 5 46620 (2)y 5
+ 22680 (3)y 5 62700 (2)y 4 + 36270 (3)y 4 + 29160 (2) ln(2)y 4
23193 (2)2 y 4 + 23925y 4 + 24120 (2)y 3 31644 (2)2 y 3
29160 (2) ln(2)y 2 + 68520 (2)y 2 30690 (3)y 2 23925y 2
8235 (2)2 y 2 19440 (2) ln(2)y 3726 (2)2 y 22680 (3)y
+ 22500 (2)y + 1701 (2)2 6030 (3) + 7975 3240 (2) ln(2)

10380 (2) / (y 1)3 (y + 1)3

+ 52/3 y 2 + 1 / (y 1)(y + 1) H (1, 1, 0; y),

c12 = 2 5y 4 + 10y 3 + 18y 2 + 10y + 5 / (y + 1)2 (y 1)2 H (0, 1, 0; y)

4 3y 5 3y 4 + 6y 3 + 2y 2 y + 1

/ (y 1)3 (y + 1)2 H (0, 0, 1, 0; y)
1/4 272 (2)y 6 229y 6 + 8 (2)y 5 266y 5 15y 4

+ 200 (2)y 4 + 196y 3 + 352 (2)y 3 168 (2)y 2 + 189y 2

+ 8 (2)y + 70y 96 (2) + 55 / (y 1)3 (y + 1)3 H (0, 0; y)

2 8y 4 + 19y 3 + 46y 2 + 19y + 8 / (y + 1)2 (y 1)2 H (1, 0, 0; y)

+ 4 y 6 + y 5 + 14y 4 + 20y 3 + 14y 2 + y + 1

/ (y 1)3 (y + 1)3 H (0, 1, 0, 0; y)

2 5y 6 + 10y 5 60y 4 59y 3 16y 2 + 7y + 5

/ (y 1)3 (y + 1)3 H (0, 0, 0; y)

8 y 5 + 2y 4 + 18y 3 + 14y 2 + 4y 1

/ (y 1)3 (y + 1)2 H (1, 0, 0, 0; y)
269

+ 27y 6 + 2y 5 + 27y 4 + 56y 3 7y 2 + 2y 7

/ (y 1)3 (y + 1)3 H (0, 0, 0, 0; y)

2
+ 12 y 2 + 1 / (y + 1)2 (y 1)2 H (1, 0, 0, 0; y)

+ 2 3y 4 + 15y 3 + 76y 2 + 15y + 5 / (y + 1)2 (y 1)2 H (1, 0, 0; y)

+ 2 7y 6 10y 5 65y 4 104y 3 75y 2 10y 3

/ (y 1)3 (y + 1)3 H (0, 1, 0, 0; y)

2
+ 16 y 2 + 1 / (y + 1)2 (y 1)2 H (1, 1, 0, 0; y)
1/2 55y 4 + 80 (2)y 4 82y 3 + 160 (2)y 2 + 82y

+ 55 + 80 (2) / (y + 1)2 (y 1)2 H (1, 0; y)

2 8y 4 + 27y 3 + 60y 2 + 27y + 6 / (y + 1)2 (y 1)2 H (0, 1, 0; y)

+ 2 11y 6 + 2y 5 + 33y 4 + 56y 3 + 23y 2 + 2y + 1

/ (y 1)3 (y + 1)3 H (0, 0, 1, 0; y)

4 (2) 7y 4 + 36y 3 + 132y 2 + 36y + 11 / (y + 1)2 (y 1)2 H (1; y)

8 (2) 4y 6 6y 5 41y 4 72y 3 49y 2 6y 4

/ (y 1)3 (y + 1)3 H (0, 1; y)
2

48 (2) y 2 + 1 / (y + 1)2 (y 1)2 H (1, 1; y)
1/10 70 (3)y 6 + 120 (2) ln(2)y 6 115y 6 196 (2)2 y 6

+ 1580 (2)y 6 + 18 (2)2 y 5 + 720 (2) ln(2)y 5 + 1880 (2)y 5
300 (3)y 5 + 240 (2)y 4 1490 (3)y 4 + 1080 (2) ln(2)y 4 74 (2)2 y 4
+ 345y 4 1720 (2)y 3 + 232 (2)2 y 3 1080 (2) ln(2)y 2 1500 (2)y 2
+ 1530 (3)y 2 345y 2 + 258 (2)2 y 2 720 (2) ln(2)y + 18 (2)2 y
+ 300 (3)y 160 (2)y + 136 (2)2 110 (3) + 115 120 (2) ln(2)

320 (2) / (y 1)3 (y + 1)3

2
+ 8 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1, 0; y)
2 24 (2)y 4 + 2 (3)y 4 57 (2)y 3 + 4 (3)y 2

138 (2)y 2 57 (2)y 24 (2) + 2 (3) / (y + 1)2 (y 1)2 H (1; y)

2
8 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1, 0; y)

2
+ 8 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1, 0; y)

2
+ 4 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1, 0; y)
+ 1/8 85y 6 + 104 (2)y 6 + 64 (3)y 6 6y 5 160 (3)y 5

+ 184 (2)y 5 1056 (3)y 4 2312 (2)y 4 + 85y 4 + 12y 3
1888 (2)y 3 1536 (3)y 3 + 85y 2 1056 (3)y 2
270
152 (2)y 2 6y + 360 (2)y 160 (3)y + 64 (3)

+ 248 (2) 85 / (y 1)3 (y + 1)3 H (0; y)

12 (2) y 6 + y 5 + 14y 4 + 20y 3 + 14y 2 + y + 1

/ (y 1)3 (y + 1)3 H (0, 1; y)

4 y 2 + 1 / (y 1)(y + 1) H (1, 1, 0; y)
+ 4 4y 5 + 5 (2)y 5 + 2y 4 + 7 (2)y 4 + 2y 3 + 74 (2)y 3 + 2y 2

+ 54 (2)y 2 + 17 (2)y + 2y 4 5 (2)

/ (y 1)3 (y + 1)2 H (1, 0; y),
(121)
and

4
3
c3 = c5 = 1/3 y 2 + 1 / (y 1)(y + 1) ,
11
5
c2 = c6 = 0,
2

c4 = y 2 + 1 / (y 1)2 (y + 1)2 H (0; y) y 2 + 1 / (y 1)(y + 1) ,

c7 = y 2 + 1 / (y 1)(y + 1) H (1; y)

2
+ y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1; y)

+ 2/ (y 1)(y + 1) y 2 H (0; y)

2 y 2 + 1 / (y 1)2 (y + 1)2 y 2 H (0, 0; y)

+ y 2 + 1 / (y 1)(y + 1) H (1; y)

2
y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1; y)

+ 1/36 y 2 + 1 67y 2 + 18 (2)y 2 + 18 (2) 67

/ (y 1)2 (y + 1)2 ,

c8 = 2 y 2 + y + 2 y 2 + 1 / (y + 1)2 (y 1)2 H (0; y)

2
+ 3 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 0; y)

2
+ 4 y 2 + 1 / (y + 1)2 (y 1)2 H (1, 0; y)

2 y 2 + 1 / (y 1)(y + 1) H (1; y)

2
+ 2 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1; y)

1/2 7y 4 + 8 (2)y 4 + 2y 3 + 16 (2)y 2 2y 7 + 8 (2)

/ (y + 1)2 (y 1)2 ,

c9 = 1/9 19y 2 + 6y + 19 / (y 1)(y + 1) H (0; y)

2/3 y 2 + 1 / (y 1)(y + 1) H (0, 0; y)

4/3 y 2 + 1 / (y + 1)(y 1) H (1, 0; y)

2/9 19y 2 + 6y + 19 / (y 1)(y + 1) H (1; y)

c1 =
(122)
(123)
(124)
(125)
(126)
271

4/3 y 2 + 1 / (y 1)(y + 1) H (0, 1; y)

8/3 y 2 + 1 / (y 1)(y + 1) H (1, 1; y)

1/54 209y 2 + 78y + 209 / (y 1)(y + 1) ,

(127)

c10 = 1/9 19y 4 + 14y 3 + 14y 2 + 14y + 19 /(y 1)4 H (0; y)

2/3 y 4 2y 3 4y 2 2y + 1 / (y 1)3 (y + 1) H (0, 0; y)

1/54 265y 4 + 208y 3 + 14y 2 + 208y + 265

/ (y 1)3 (y + 1) ,
(128)
2

c11 = 52/3 y + 1 / (y 1)(y + 1) H (1, 1; y)

+ 2 y 2 + 1 / (y 1)(y + 1) H (1, 1; y)

2
2 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1; y)

2 (2)y 4 3y 4 2y 3 + 4 (2)y 2 + 2y + 2 (2) + 3

/ (y 1)2 (y + 1)2 H (1; y)

8 y 2 + 1 / (y 1)2 (y + 1) yH (0, 1; y)

2
4 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1; y)

+ 2 7y 5 7y 4 + 12y 3 4y 2 + y 1 / (y 1)3 (y + 1)2 H (0, 0, 1; y)

2
+ 4 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1; y)

6 y 2 + 1 / (y 1)(y + 1) H (1, 1; y)

2
+ 6 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1; y)

+ 2 y 6 + 5y 5 + 21y 4 + 34y 3 + 17y 2 + 5y 3

/ (y 1)3 (y + 1)3 H (0, 1, 0; y)

+ 1/6 89y 6 + 402y 5 + 499y 4 + 354y 3 + 37y 2 + 24y + 11

/ (y 1)3 (y + 1)3 H (0, 0; y)

2 y 5 + 5y 4 + 18y 3 + 26y 2 + y + 5 / (y 1)3 (y + 1)2 H (1, 0, 0; y)
+ 1/36 468 (2)y 5 217y 5 288 (2)y 4 + 499y 4 48y 3

+ 1260 (2)y 3 300y 2 + 252 (2)y 2 + 216 (2)y 167y

36 (2) + 233 / (y 1)3 (y + 1)2 H (0; y)

+ 2 2y 4 5y 3 14y 2 5y + 6 / (y 1)2 (y + 1)2 H (1, 0; y)

12y 5 + y 4 + 10y 3 + 28y 2 2y + 1 / (y 1)3 (y + 1)3 yH (0, 0, 0; y)

+ 1/3 53y 4 + 141y 3 + 228y 2 + 141y 17

/ (y 1)2 (y + 1)2 H (1, 0; y)

2 7y 6 + 7y 5 + 33y 4 + 56y 3 + 25y 2 + 7y 1

/ (y 1)3 (y + 1)3 H (0, 1, 0; y)
272
+ 1/9 18 (2)y 4 + 31y 4 + 105y 3 + 36 (2)y 2 105y

+ 18 (2) 31 / (y 1)2 (y + 1)2 H (1; y)

6 y 2 + 1 / (y 1)(y + 1) H (1, 1; y)

2
+ 6 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1; y)

+ 2/3 34y 4 3y 3 33y 2 3y 1 / (y 1)2 (y + 1)2 H (0, 1; y)

2
+ 4 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1; y)

2 y 2 + 3 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 0; y)

2 11y 6 5y 5 15y 4 26y 3 27y 2 5y 1

/ (y 1)3 (y + 1)3 H (0, 0, 1; y)

2
4 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1; y)
2

10 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1; y)
1/216 2545y 6 + 1620 (3)y 6 + 2160 (2)y 6 + 3024 (3)y 5

570y 5 + 1944 (2)y 5 432 (2)y 4 + 13716 (3)y 4
+ 2545y 4 + 1140y 3 + 25920 (3)y 3 + 13500 (3)y 2
+ 2545y 2 1728 (2)y 2 1944 (2)y + 3024 (3)y 570y

+ 1404 (3) 2545 / (y 1)3 (y + 1)3 ,
(129)

4
4
3
2
c12 = 4 4y + (2)y 2y + 2 (2)y + 2y + (2) + 4

/ (y + 1)2 (y 1)2 H (1; y)

+ 2 5y 4 + 10y 3 + 18y 2 + 10y + 5 / (y + 1)2 (y 1)2 H (0, 1; y)

2
+ 8 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1; y)

4 3y 5 3y 4 + 6y 3 + 2y 2 y + 1

/ (y 1)3 (y + 1)2 H (0, 0, 1; y)
1/4 229y 5 + 56 (2)y 5 64 (2)y 4 37y 4 + 48 (2)y 3

+ 22y 3 + 174y 2 144 (2)y 2 + 15y + 32 (2)y

40 (2) + 55 / (y 1)3 (y + 1)2 H (0; y)

2 8y 4 + 19y 3 + 46y 2 + 19y + 8 / (y + 1)2 (y 1)2 H (1, 0; y)

+ 4 y 6 + y 5 + 14y 4 + 20y 3 + 14y 2 + y + 1

/ (y 1)3 (y + 1)3 H (0, 1, 0; y)

2 5y 6 + 10y 5 60y 4 59y 3 16y 2 + 7y + 5

/ (y 1)3 (y + 1)3 H (0, 0; y)

8 y 5 + 2y 4 + 18y 3 + 14y 2 + 4y 1

/ (y 1)3 (y + 1)2 H (1, 0, 0; y)
273

+ 27y 6 + 2y 5 + 27y 4 + 56y 3 7y 2 + 2y 7

/ (y 1)3 (y + 1)3 H (0, 0, 0; y)
2

+ 12 y 2 + 1 / (y + 1)2 (y 1)2 H (1, 0, 0; y)

+ 2 3y 4 + 15y 3 + 76y 2 + 15y + 5 / (y + 1)2 (y 1)2 H (1, 0; y)

+ 2 7y 6 10y 5 65y 4 104y 3 75y 2 10y 3

/ (y 1)3 (y + 1)3 H (0, 1, 0; y)

2
+ 16 y 2 + 1 / (y + 1)2 (y 1)2 H (1, 1, 0; y)
1/2 32 (2)y 4 55y 4 82y 3 + 64 (2)y 2 + 82y

+ 32 (2) + 55 / (y + 1)2 (y 1)2 H (1; y)

2 8y 4 + 27y 3 + 60y 2 + 27y + 6 / (y + 1)2 (y 1)2 H (0, 1; y)

2
8 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1; y)

+ 2 11y 6 + 2y 5 + 33y 4 + 56y 3 + 23y 2 + 2y + 1

/ (y 1)3 (y + 1)3 H (0, 0, 1; y)
2

+ 8 y 2 + 1 / (y 1)2 (y + 1)2 H (1, 0, 1; y)

4 y 2 + 1 / (y 1)(y + 1) H (1, 1; y)

2
+ 4 y 2 + 1 / (y 1)2 (y + 1)2 H (0, 1, 1; y)
+ 1/8 56 (2)y 6 + 64 (3)y 6 85y 6 6y 5 160 (3)y 5 136 (2)y 5

+ 85y 4 392 (2)y 4 1056 (3)y 4 1536 (3)y 3 + 12y 3 1056 (3)y 2
+ 360 (2)y 2 + 85y 2 + 136 (2)y 6y 160 (3)y + 64 (3) 85

+ 88 (2) / (y 1)3 (y + 1)3 ,
(130)
respectively,
dj = 0 for j = 1, . . . , 7,

d8 = 2 5y 2 + 2y + 5 / (y 1)(y + 1)3 yH (0; y)

4 3y 2 + 2y + 3 y 2 + 1 / (y 1)2 (y + 1)4 yH (0, 0; y)
+ 4 y 4 + 9 (2)y 4 + 6 (2)y 3 + 18 (2)y 2 + 2y 2 + 6 (2)y

1 + 9 (2) / (y 1)2 (y + 1)4 y,

d9 = 2/9 51y 2 + 26y + 51 / (y 1)(y + 1)3 yH (0; y)

+ 4/3 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (0, 0; y)

+ 8/3 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (1, 0; y)

4/9 36 (2)y 2 + 19y 2 + 24 (2)y 19 + 36 (2)

/ (y 1)(y + 1)3 y,
(131)
(132)
(133)
274

d10 = 2/9 51y 4 + 176y 3 + 144 (2)y 2 70y 2 + 176y + 51

/ (y 1)3 (y + 1)3 yH (0; y)

4/3 3y 4 + 14y 3 2y 2 + 14y + 3 / (y 1)4 (y + 1)2 yH (0, 0; y)

16/ (y 1)3 (y + 1)3 y 3 H (0, 0, 0; y)

8/9 4y 2 + 9 (2)y 2 + 32y 4 + 9 (2)

/ (y 1)2 (y + 1)2 y,
(134)
2

2
2
4
d11 = 72 (2) 3y + 2y + 3 / (y 1) (y + 1) y H (1; y)

+ 24 (2) 5y 4 + 14y 3 + 42y 2 + 14y + 5

/ (y 1)3 (y + 1)5 y 2 H (0, 1; y)
+ 1/18 648 (2)y 6 687y 6 + 26y 5 + 1008 (3)y 5 + 8712 (2)y 5

+ 11808 (2)y 4 + 4752 (3)y 4 + 687y 4 + 9216 (3)y 3 52y 3
+ 14112 (2)y 3 + 687y 2 + 4752 (3)y 2 648 (2)y 2 + 1008 (3)y + 26y

648 (2)y 687 / (y 1)3 (y + 1)5 yH (0; y)

4 y 4 + 24 (2)y 3 + 64 (2)y 2 2y 2 + 24 (2)y + 1

/ (y 1)3 (y + 1)3 yH (1, 0; y)

+ 16/ (y 1)2 (y + 1)2 y 2 H (0, 1, 0; y)

32/ (y 1)3 (y + 1)3 y 3 H (0, 0, 1, 0; y)
1/3 39y 6 + 184y 5 + 84 (2)y 5 + 87y 4 + 432 (2)y 4

+ 1416 (2)y 3 152y 3 + 432 (2)y 2 81y 2 32y

+ 84 (2)y + 33 / (y 1)3 (y + 1)5 yH (0, 0; y)

+ 24 3y 2 + 2y + 3 / (y 1)2 (y + 1)4 y 2 H (1, 0, 0; y)

8 5y 4 + 14y 3 + 42y 2 + 14y + 5

/ (y 1)3 (y + 1)5 y 2 H (0, 1, 0, 0; y)

2 9y 4 + 112y 3 + 146y 2 + 196y + 9

/ (y 1)3 (y + 1)5 y 3 H (0, 0, 0; y)

+ 16 3y 2 + 8y + 3 / (y 1)3 (y + 1)3 y 2 H (1, 0, 0, 0; y)

+ 4 y 4 + 5y 3 + 38y 2 + 5y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 0, 0, 0; y)

2 9y 4 + 92y 3 + 130y 2 + 92y + 9 / (y 1)2 (y + 1)4 yH (1, 0, 0; y)

+ 8 7y 4 + 29y 3 + 62y 2 + 29y + 7 / (y 1)3 (y + 1)5 y 2 H (0, 1, 0, 0; y)

2/3 3y 2 + 88y + 3 / (y 1)(y + 1)3 yH (1, 0; y)

+ 48 y 2 + y + 1 / (y 1)2 (y + 1)4 y 2 H (0, 1, 0; y)

8 5y 4 + 14y 3 + 30y 2 + 14y + 5 / (y 1)3 (y + 1)5 y 2 H (0, 0, 1, 0; y)
275

+ 12 (2) 3y 4 + 40y 3 + 62y 2 + 40y + 3 / (y 1)2 (y + 1)4 yH (1; y)

48 (2) 4y 4 + 17y 3 + 32y 2 + 17y + 4

/ (y 1)3 (y + 1)5 y 2 H (0, 1; y)
+ 2/45 605y 6 540 (2) ln(2)y 6 270 (3)y 6 180 (2)y 6 + 621 (2)2 y 5
3690 (3)y 5 2160 (2) ln(2)y 5 + 9390 (2)y 5 1815y 4 + 5670 (2)y 4
+ 3906 (2)2 y 4 2700 (2) ln(2)y 4 5490 (3)y 4 5100 (2)y 3
+ 4626 (2)2 y 3 5760 (2)y 2 + 2700 (2) ln(2)y 2 + 5490 (3)y 2
+ 3906 (2)2 y 2 + 1815y 2 + 621 (2)2 y + 2160 (2) ln(2)y 4290 (2)y

+ 3690 (3)y + 540 (2) ln(2) + 270 (2) + 270 (3) 605

/ (y 1)3 (y + 1)5 y,
(135)
4

d12 = 24 (2) 3y + 7y 3 + 64y 2 + 7y + 3 / (y 1)2 (y + 1)4 yH (1; y)

48 (2) y 4 19y 3 16y 2 19y + 1

/ (y 1)3 (y + 1)5 y 2 H (0, 1; y)
+ 1/2 61y 6 + 96 (2)y 6 120 (2)y 5 + 22y 5 + 32 (3)y 5

61y 4 + 1056 (3)y 4 + 3168 (2)y 4 44y 3 + 1664 (3)y 3
+ 896 (2)y 3 144 (2)y 2 + 1056 (3)y 2 61y 2 + 32 (3)y + 22y

200 (2)y + 61 240 (2) / (y 1)3 (y + 1)5 yH (0; y)

32 y 4 + 2 (2)y 3 + y 3 + 34 (2)y 2 + 2 (2)y + y 1

/ (y 1)3 (y + 1)3 yH (1, 0; y)

8 3y 4 + 6y 3 + 14y 2 + 6y + 3 / (y 1)2 (y + 1)4 yH (0, 1, 0; y)

+ 64/ (y 1)3 (y + 1)3 y 3 H (0, 0, 1, 0; y)
+ 125y 6 + 24 (2)y 5 230y 5 + 24 (2)y 4 187y 4

+ 480 (2)y 3 + 204y 3 + 24 (2)y 2 + 269y 2 + 26y

+ 24 (2)y + 43 / (y 1)3 (y + 1)5 yH (0, 0; y)

+ 8 3y 4 + 7y 3 + 64y 2 + 7y + 3 / (y 1)2 (y + 1)4 yH (1, 0, 0; y)

+ 16 y 4 19y 3 16y 2 19y + 1

/ (y 1)3 (y + 1)5 y 2 H (0, 1, 0, 0; y)

+ 4 14y 5 175y 4 56y 3 14y 2 + 6y + 9

/ (y 1)3 (y + 1)5 yH (0, 0, 0; y)

+ 32 y 2 + 17y + 1 / (y 1)3 (y + 1)3 y 2 H (1, 0, 0, 0; y)

8 y 4 + 5y 3 + 38y 2 + 5y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 0, 0, 0; y)

4 15y 4 + 16y 3 + 198y 2 + 16y + 15 / (y 1)2 (y + 1)4 yH (1, 0, 0; y)
276

+ 16 y 4 + 37y 3 + 54y 2 + 37y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 1, 0, 0; y)

2 25y 2 + 86y + 25 / (y + 1)3 (y 1) yH (1, 0; y)

+ 8 3y 4 + 13y 3 + 64y 2 + 13y + 3 / (y 1)2 (y + 1)4 yH (0, 1, 0; y)

+ 16 y 4 19y 3 28y 2 19y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 0, 1, 0; y)

+ 24 (2) 9y 4 + 14y 3 + 102y 2 + 14y + 9 / (y 1)2 (y + 1)4 yH (1; y)

192 (2) 8y 2 + 13y + 8 / (y 1)3 (y + 1)5 y 3 H (0, 1; y)
1/5 85y 6 240 (2) ln(2)y 6 + 60 (3)y 6 1640 (2)y 6 + 68 (2)2 y 5

+ 320 (3)y 5 960 (2) ln(2)y 5 3480 (2)y 5 255y 4 4400 (2)y 4
+ 20 (2)2 y 4 1200 (2) ln(2)y 4 + 3340 (3)y 4 + 3600 (2)y 3
960 (2)2 y 3 + 5560 (2)y 2 + 1200 (2) ln(2)y 2 3340 (3)y 2
+ 20 (2)2 y 2 + 255y 2 + 68 (2)2 y + 960 (2) ln(2)y 120 (2)y
320 (3)y + 240 (2) ln(2) + 480 (2)

60 (3) 85 / (y 1)3 (y + 1)5 y,
(136)
and
dj = 0 for j = 1, . . . , 7,
(137)
2

2

2
4
d8 = 4 3y + 2y + 3 y + 1 / (y 1) (y + 1) yH (0; y)

+ 2 5y 2 + 2y + 5 / (y 1)(y + 1)3 y,
(138)
2

3
d9 = 4/3 3y + 2y + 3 / (y 1)(y + 1) yH (0; y)

+ 8/3 3y 2 + 2y + 3 / (y 1)(y + 1)3 yH (1; y)

+ 2/9 51y 2 + 26y + 51 / (y 1)(y + 1)3 y,
(139)
4

d10 = 4/3 3y + 14y 3 2y 2 + 14y + 3 / (y 1)4 (y + 1)2 yH (0; y)

16/ (y 1)3 (y + 1)3 y 3 H (0, 0; y)

+ 2/9 51y 4 + 176y 3 70y 2 + 176y + 51 / (y 1)3 (y + 1)3 y,
(140)

d11 = 4/ (y 1)(y + 1) yH (1; y)

+ 16/ (y 1)2 (y + 1)2 y 2 H (0, 1; y)

32/ (y 1)3 (y + 1)3 y 3 H (0, 0, 1; y)
+ 1/3 39y 5 223y 4 60 (2)y 4 252 (2)y 3 + 136y 3 252 (2)y 2

+ 16y 2 60 (2)y + 65y 33 / (y 1)3 (y + 1)4 yH (0; y)

+ 24 3y 2 + 2y + 3 / (y 1)2 (y + 1)4 y 2 H (1, 0; y)

8 5y 4 + 14y 3 + 42y 2 + 14y + 5 / (y 1)3 (y + 1)5 y 2 H (0, 1, 0; y)

2 9y 4 + 112y 3 + 146y 2 + 196y + 9 / (y 1)3 (y + 1)5 y 3 H (0, 0; y)

+ 16 3y 2 + 8y + 3 / (y 1)3 (y + 1)3 y 2 H (1, 0, 0; y)
277

+ 4 y 4 + 5y 3 + 38y 2 + 5y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 0, 0; y)

2 9y 4 + 92y 3 + 130y 2 + 92y + 9 / (y 1)2 (y + 1)4 yH (1, 0; y)

+ 8 7y 4 + 29y 3 + 62y 2 + 29y + 7 / (y 1)3 (y + 1)5 y 2 H (0, 1, 0; y)

2/3 3y 2 + 88y + 3 / (y 1)(y + 1)3 yH (1; y)

+ 48 y 2 + y + 1 / (y 1)2 (y + 1)4 y 2 H (0, 1; y)

8 5y 4 + 14y 3 + 30y 2 + 14y + 5 / (y 1)3 (y + 1)5 y 2 H (0, 0, 1; y)
1/18 687y 6 26y 5 648 (2)y 5 1008 (3)y 5 4752 (3)y 4

1296 (2)y 4 687y 4 9216 (3)y 3 + 52y 3
+ 1296 (2)y 2 4752 (3)y 2 687y 2 + 648 (2)y

1008 (3)y 26y + 687 / (y 1)3 (y + 1)5 y,
(141)
2

3
d12 = 32 y + y + 1 / (y + 1) (y 1) yH (1; y)

8 3y 4 + 6y 3 + 14y 2 + 6y + 3 / (y 1)2 (y + 1)4 yH (0, 1; y)

+ 64/ (y 1)3 (y + 1)3 y 3 H (0, 0, 1; y)
125y 5 8 (2)y 4 + 105y 4 + 64 (2)y 3 + 82y 3 286y 2

+ 64 (2)y 2 + 17y 8 (2)y 43 / (y 1)3 (y + 1)4 yH (0; y)

+ 8 3y 4 + 7y 3 + 64y 2 + 7y + 3 / (y 1)2 (y + 1)4 yH (1, 0; y)

+ 16 y 4 19y 3 16y 2 19y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 1, 0; y)

+ 4 14y 5 175y 4 56y 3 14y 2 + 6y + 9

/ (y 1)3 (y + 1)5 yH (0, 0; y)

+ 32 y 2 + 17y + 1 / (y 1)3 (y + 1)3 y 2 H (1, 0, 0; y)

8 y 4 + 5y 3 + 38y 2 + 5y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 0, 0; y)

4 15y 4 + 16y 3 + 198y 2 + 16y + 15 / (y 1)2 (y + 1)4 yH (1, 0; y)

+ 16 y 4 + 37y 3 + 54y 2 + 37y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 1, 0; y)

2 25y 2 + 86y + 25 / (y + 1)3 (y 1) yH (1; y)

+ 8 3y 4 + 13y 3 + 64y 2 + 13y + 3 / (y 1)2 (y + 1)4 yH (0, 1; y)

+ 16 y 4 19y 3 28y 2 19y + 1 / (y 1)3 (y + 1)5 y 2 H (0, 0, 1; y)
+ 1/2 61y 6 + 96 (2)y 6 + 104 (2)y 5 + 32 (3)y 5 + 22y 5

+ 368 (2)y 4 61y 4 + 1056 (3)y 4 + 1664 (3)y 3 44y 3
368 (2)y 2 + 1056 (3)y 2 61y 2 104 (2)y + 22y

+ 32 (3)y 96 (2) + 61 / (y 1)3 (y + 1)5 y.
(142)
278
6.3. Threshold expansions

In this section we expand the one and two loop form factors near threshold S 4m2 ,
i.e., y 1 in powers of

=
4m2
.
S
(143)
Up to and including terms of order 0 the non-vanishing real and imaginary parts of the
coefficients a i , bi , cj and dj (i = 1, . . . , 3, j = 1, . . . , 12) are
3 (2)
2,

6 (2)
a 3 =
1 + ln() + ln(2) ,
1
a 1 = ,
2

1
a 2 = 1 + ln() + ln(2) ,
a 3 = ln2 () 2 + 2 ln() ln2 (2) 2 ln(2) ln() + 2 ln(2) + (2) ,
b2 = 1,
a 2 =
b3 =
6 (2)
+ 2,
1
,
b3 = 2 2 ln() 2 ln(2) ,
3 (2) 3
c4 =
(2),
2
4 2

3 (2)
c8 =
1 + ln() + ln(2) + 3 (2) 2 ln() 1 + 2 ln(2) ,
2
14
2 (2)
c9 =
11 + 6 ln() + 6 ln(2) + ,
3
9
80 16
(2),
c10 =
9
3
9
101
(2)
66 ln() 97 + 66 ln(2) (3)

c11 =
6
2
18
89
18 (2) ln(2) + (2) 4 (2) ln(),
6
b2 =
(144)
(145)
(146)
(147)
(148)
(149)
(150)
(151)
(152)
(153)
(154)
(155)
(156)
(157)
279
3 (2)
4 ln2 (2) + 3 (2) + 8 ln(2) 8 ln(2) ln()

2 2
6 (2)
12 (2) ln2 (2)
4 + 8 ln() 4 ln2 ()
23
24 (2) ln(2) ln() 12 (2) ln2 () +
6
27
+ 9 2 (2) 8 (2) (3) + 2 (2) ln() + 21 (2) ln(2),
(158)
4
1
c1 = ,
(159)
6
11
c3 =
(160)
,
24
5
c5 =
(161)
,
18
31
c7 =
(162)
,
72
3 (2) 1
c8 =
(163)
+ 3 (2),
2 2

1 62 22
22
4
2
2
c9 =
ln(2) + ln2 (2)

ln() + ln2 () + ln(2) ln() ,
27
9
3
9
3
3
(164)

239 11 2
1 97
11 2
ln()
ln ()
ln (2)
c11 =
18
54
6
6

11
97
ln(2)
ln(2) ln() + 2 (2),
(165)
+
18
3

1
6 (2)
c12 =
1 + ln() + ln(2) + 2 2 ln() 2 ln(2)
2

(166)
+ (2) 12 ln(2) + 12 ln() 1 ,
c12 =
3 (2) 3
(167)
+ (2),
2
2
4 (2) 11
d9 =
(168)
,
9
49
8
d10 = (2) + ,
(169)
3
9
15
151
11 (2) 26
+ (2) 18 (2) ln(2) 8 (2) ln() (3) +
, (170)
d11 =
3
2
36
3 (2)
6 (2)
d12 = 2 2 ln() 1 + 2 ln(2)

15 (2) ln(2)
45
41
9
(171)
(2) (3) 20 (2) ln() + ,
2
4
12
d8 =
280
1
,
2

4
16
1 4
ln(2) + ln()
,
d9 =
3
3
9

11
1
32 11
d11 =
ln(2) +
ln() + 4 (2),
3
9
3

6 (2) 1
+ 3 ln(2) ln() + 10 (2).

d12 =
2
d8 =
(172)
(173)
(174)
(175)
These expansions are relevant, for instance, for studies of QQ production near threshold.
For a study using the one-loop axial form factors see [22].
6.4. Asymptotic expansions
In this section, we write down the expansions of the form factors in the limit S m2 ,
i.e., y 0. Putting r = mS2 and L = ln r, and keeping terms up to r 2 we find for the
coefficients that are non-vanishing to that order
1
2
(3 + 2L) 1 + L,
2
r
r
1
1
a 2 = 2 8 + 13L + 8 (2) L2 + (1 + 5L)

r
r
3
1 2
2 + L L + 4 (2),
2
2

1
1
13
a 3 = 2 4 (3) + 52 (2) + L3 L2 5 + 32L 8L (2)
3
2
r

5 2
1
20 (2) L 3 + 6L
+
r
2
3
1
+ 6 (2) 4 + 2 (3) L2 + 4L + L3 4L (2),
4
6
2
a 1 = 2 1,
r
1
5 3
a 2 = 2 (13 + 2L) + L,
r
2
r
1
1
a 3 = 2 13L 32 L2 + 4 (2) + (6 + 5L)

r
r
1 2
3
+ 2 (2) L 4 + L,
2
2
1
1
b2 = 2 (12 4L) + (4 6L),
r
r

1
b3 = 2 12 32L + 2L2 16 (2) + 8 12L + 3L2 24 (2) ,
r
r
4
6
b2 = 2 + ,
r
r
a 1 =
(176)
(177)
(178)
(179)
(180)
(181)
(182)
(183)
(184)
281
b3 =
1
1
(32 4L) + (6L + 12),
2
r
r
(185)
c1 =

1
2
2
1 1
1
+
L
+ L,
2
3
3r 3 3
r
(186)

11 11 11
1 11 11
L +
+
L,
c3 = 2
4
6
6r
12 12
r
c4 =
1
1
5 + 2L2 12 (2) 5L + (2L + 2)

2
r
r
1 2
1
+ L 3 (2) + L,
2
2

10 5 5
1 5 10
L +
+ L,
c5 = 2
9
9r
9 9
r 3
c7 =

55
1 13
1 3 47
2
(2)
2
(3)
L
L
L
6L
(2)
+
+
3
9
12
r2 2

67
1
49 1
67
1
1
+ (2) + (2)
L (2) + L (3),
+
r
18
2
36 2
36
2

1
37 2
79
3
c8 = 2 2L + 16 116 (2) + L + 40L (2) L
2
2
r

1
+ 38 (2) + 6L2 9L + 5
r
1
7
+ 2 L3 L + 2L2 13 (2) + 10L (2),
2
2
c9 =

356
8
2 3 8
49 2 23 959
1
(2)
(3)
L
L
(2)
+
L
L
+
+
9
3
9
3
9
2
27
r2

1
89
5 2 59 32
+
L+ L +
(2)
r
9
3
27
3
1 3 19 2 209
64
4
106 4
L + L
L (2) (3) +
+ L (2),
9
18
54
9
3
27
3

16
1153 229 2
1
4 3 2230
c10 = 2 L (2) +
+
L + 12 (2) + L
L
3
18
9
9
27
r

17
1 385 161
4
L + L2 + (2)
+
r 27
9
3
3
19 2 22
265
383 1 3 4
+ L (2)
L+
L + L (2),
18
3
54
27
9
3
(187)
(188)
(189)
(190)
(191)
(192)
(193)
282

173 2 17 3
1 241 98 2
59
+ (2) + 144 (2) ln(2) + L (2) +
L + L
2
4
5
3
18
18
r

1319
4111
15619
(2) 110L (3) +
(3) 10L2 (2) +
L
18
6
108

5 2 137
1 1 4 69 2
1189
L + (2) 14L (3) +
L+ L
(2)
+
r 24
10
36
12
3

7
157
+ 55 (3) + 36 (2) ln(2) + 7L (2) L2 (2)
2
108
1595 11
233 2 11 3 67
63
L (2)
L + L + (3) 2 (2)
108
3
72
36
6
20
1 2
13
2545
173
+ L (2) +
(2) L (3) + 6 (2) ln(2) +
L,
2
9
2
216

1 841 351 2
+
(2) 288 (2) ln(2) + L4 + 350L (2) 625 (2)
c12 = 2
4
5
r

85 3
37
2
2
148L (3) + 95L L + 190 (3) 55L (2) + L
6
2

1
1 4 9 2
91
16 3
+
L 74 (2) L + (2) 20L (3) L
r
3
12
5
4

59 2
107
2
+ L + L (2) + 10 (3) 72 (2) ln(2) + 113L (2) +
4
4
23 5 3
55 2
68 2
+
L + 31L (2) + L 32 (2) 11 (3) + (2)
2
3
8
5
7
85
12L2 (2) + L4 + 8L (3) 12 (2) ln(2) L,
24
8
c11 =
(194)
(195)
2
1
,
3r 2 3
(196)
c3 =
11
11
+ ,
2
12
6r
(197)
c4 =
1
2
(5 4L) + + 1 L,
r
r2
(198)
c1 =
10
5
+ ,
2
9
9r

47
1
67
1
2
c7 = 2 2 (2)
+ 2L L + (2) ,
9
2
36
r

1
1
79
2
c8 = 2 16 (2) + 6L +
37L + (9 12L)
2
r
r
7
3 2
+ 4L 4 (2) + L ,
2
2
c5 =
(199)
(200)
(201)

1
10
89
1 959 2 2 98
+
L
L
+
L
+
3
9
r
3
9
r 2 27
19
209 1 2
L+
+ L ,
9
54
3

4
2230
1 161 34
1
458
c10 = 2 L2
L+
+
L
3
9
27
r
9
3
r
1 2 19
265
+ L L+
,
3
9
54

15619 173
17
1
L + 20L (2) 31 (2) L2

c11 = 2 110 (3)
108
9
6
r

1
5
1
1189
L3 7 (2) + 5L (2)
+ 14 (3) L
+
r
6
36
6
11 2 233
13
2545
L +
L + (3) L (2)
,
12
36
2
216

85
1
37
c12 = 2 62L (2) + L2
+ 148 (3) 4L3 180 (2) 190L
2
2
r

1
91 59
1
20 (3) 49 (2) + 16L2 + L3 + 2L (2) +
L
+
r
3
4
2
7
85
55
8 (3) + 5L2 L L3 + 10L (2) 11 (2),
+
8
4
6
283
c9 =
20 4L2 + 24 (2) + 28L + 6L2 + 10L 4 + 36 (2) ,

2
r
r

32
44 4 2
1 196
L + (2)
L
d9 = 2
9
3
3
3
r

76
1
34
2
+ 16 (2) 2L + L ,
+
r
9
3

52
1
628
d10 = 2 44 L2 16 (2) +
L
3
9
r

32
1
34
8 (2) +
2L2 + L ,
+
r
9
3

157 43 2 401
138 2
1
d11 = 2 96 (2) ln(2) +
L
L 36L (2)
(2)
3
3
9
5
r

680
1
(2) 164 (3)
L4 + 56L (3) + 14L2 (2) +
6
3

229
11
1
242
12 (3)
L 12 (2) 24 (2) ln(2) + L2 +
,
+
r
6
2
9
d8 =
(202)
(203)
(204)
(205)
(206)
(207)
(208)
(209)
284

1
115 + 43L + 360 (2) 100L (2) + 16L (3) 12L2 (2)
r2

1
68
+ 192 (2) ln(2) + 2 (2) + L4 64 (3) + 4L3 67L2
5
3

61
1
96 (2) + 48 (2) ln(2) + L + 6L3 17
+
r
2

43 2
12 (3) L 120L (2) ,
2
1
1
d8 = 2 (28 + 8L) + (10 + 12L),
r
r

196
8
1
34
1
+ L +
+ 4L ,
d9 = 2
9
3
r
3
r

628 104
1
34
1
d10 = 2
+
L +
+ 4L ,
9
3
r
3
r

401
2
1
86
d11 = 2
+ L3 + L 20L (2) + 36 (2) 56 (3)
9
3
3
r

1 229
11L ,
+
r
6

4
1
d12 = 2 43 16 (3) + 52 (2) L3 + 8L (2) 12L2 + 134L
3
r

1
61
+
43L
+ 48 (2) 18L2 .
r
2
d12 =
(210)
(211)
(212)
(213)
(214)
(215)
The limit r corresponds to the massless limit m 0. Therefore the chirality-flipping

form factors F2 and G2 must be of order 1/r, and those terms in the chirality-conserving
form factor F1 that survive the limit r must be equal to the corresponding terms
in G1 . The asymptotic expansions given above and in [1] satisfy these constraints.
All the results of this section can be obtained in an electronic form by downloading the
source of this manuscript from http://www.arxiv.org.
7. Summary
In this paper we calculated the axial vector form factors G1 and G2 to second order
in the QCD coupling, excluding the contributions from Fig. 1(a) and (b). The results for
the two form factors were obtained keeping the full dependence on the mass of the heavy
quark, as well as on the arbitrary momentum transfer.
The renormalized form factors are expressed in terms of the MS coupling S for
(Nf + 1) quark flavors and of the on-shell mass of the heavy quark. The expressions for the
unsubtracted as well as the UV-renormalized form factors are given in closed analytic form
as a Laurent expansion in = (4 D)/2. The coefficients of this expansion have a suitable representation in terms of 1-dimensional harmonic polylogarithms. Poles in , which
285
correspond to soft and collinear divergences, are still present in G1,2,R , like in F1,2,R . Depending on the observable considered, these divergences must cancel either among each
other or against the divergences arising from the real radiation, which in this paper was not
taken into account.
As already discussed in the introduction and in our preceding paper [1] our results for
the vector and axial vector form factors are part of the order S2 QCD corrections to the
differential electroproduction cross sections of heavy quarks and, moreover, have a number
of immediate applications. These include studies of the extrapolation of the continuum
production threshold and studies concerning the generic singularity
amplitudes to the QQ
structure of QCD amplitudes with massive quarks. Of general interest is the computation
of the NNLO QCD corrections to the forwardbackward asymmetry of heavy quarks. We
shall report on this in a future publication.
Acknowledgements
We are grateful to J. Vermaseren for his kind assistance in the use of the algebra manipulating program FORM [23], by which many of our calculations were carried out. This work
was partially supported by the European Union under contract HPRN-CT-2000-00149, by
Deutsche Forschungsgemeinschaft (DFG), SFB/TR9, by DFG-Graduiertenkolleg RWTH
Aachen, by the Swiss National Funds (SNF) under contract 200021-101874, and by the
USA DoE under the grant DE-FG03-91ER40662, Task J.
References
[1] W. Bernreuther, R. Bonciani, T. Gehrmann, R. Heinesch, T. Leineweber, P. Mastrolia, E. Remiddi, Nucl.
Phys. B 706 (2005) 245, hep-ph/0406046.
[2] The LEP Collaborations ALEPH, DELPHI, L3, OPAL, LEP Electroweak Working Group, SLD Heavy
Flavour Group, A Combination of Preliminary Electroweak Measurements and Constraints on the Standard
Model, report LEPEWWG/2003-01, April 2003.
[3] A. Freitas, K. Moenig, hep-ph/0411304.
[4] ECFA/DESY LC Physics Working Group Collaboration, J.A. Aguilar-Saavedra, et al., TESLA technical
design report part III: Physics at an e+ e linear collider, DESY-report 2001-011, hep-ph/0106315.
[5] W. Bernreuther, R. Bonciani, T. Gehrmann, R. Heinesch, T. Leineweber, E. Remiddi, in preparation.
[6] G. t Hooft, M.J.G. Veltman, Nucl. Phys. B 44 (1972) 189.
[7] S.A. Larin, Phys. Lett. B 303 (1993) 113, hep-ph/9302240.
[8] E. Remiddi, J.A.M. Vermaseren, Int. J. Mod. Phys. A 15 (2000) 725, hep-ph/9905237.
[9] T. Gehrmann, E. Remiddi, Comput. Phys. Commun. 141 (2001) 296, hep-ph/0107173.
[10] D.J. Broadhurst, N. Gray, K. Schilcher, Z. Phys. C 52 (1991) 111.
[11] K. Melnikov, T. van Ritbergen, Nucl. Phys. B 591 (2000) 515, hep-ph/0005131.
[12] R. Bonciani, P. Mastrolia, E. Remiddi, Nucl. Phys. B 661 (2003) 289, hep-ph/0301170;
R. Bonciani, P. Mastrolia, E. Remiddi, Nucl. Phys. B 702 (2004) 259, Erratum.
[13] R. Bonciani, P. Mastrolia, E. Remiddi, Nucl. Phys. B 676 (2004) 399, hep-ph/0307295.
[14] R. Bonciani, P. Mastrolia, E. Remiddi, Nucl. Phys. B 690 (2004) 138, hep-ph/0311145.
[15] S. Laporta, E. Remiddi, Phys. Lett. B 379 (1996) 283, hep-ph/9602417;
S. Laporta, Int. J. Mod. Phys. A 15 (2000) 5087, hep-ph/0102033.
[16] F.V. Tkachov, Phys. Lett. B 100 (1981) 65;
K.G. Chetyrkin, F.V. Tkachov, Nucl. Phys. B 192 (1981) 159.
286
[17]
[18]
[19]
[20]
T. Gehrmann, E. Remiddi, Nucl. Phys. B 580 (2000) 485, hep-ph/9912329.

A.V. Kotikov, Phys. Lett. B 254 (1991) 158.
E. Remiddi, Nuovo Cimento A 110 (1997) 1435, hep-th/9711188.
M. Caffo, H. Czyz, S. Laporta, E. Remiddi, Acta Phys. Pol. B 29 (1998) 2627, hep-ph/9807119;
M. Caffo, H. Czyz, S. Laporta, E. Remiddi, Nuovo Cimento A 111 (1998) 365, hep-ph/9805118.
[21] J. Jersak, E. Laermann, P.M. Zerwas, Phys. Rev. D 25 (1982) 1218;
J. Jersak, E. Laermann, P.M. Zerwas, Phys. Rev. D 36 (1987) 310, Erratum.
[22] J.H. Khn, T. Teubner, Eur. Phys. J. C 9 (1999) 221, hep-ph/9903322.
[23] J.A.M. Vermaseren, Symbolic Manipulation with FORM, Version 2, CAN, Amsterdam, 1991;
J.A.M. Vermaseren, New features of FORM, math-ph/0010025.
Unquenched QCD Dirac operator spectra

at nonzero baryon chemical potential
G. Akemann a , J.C. Osborn b , K. Splittorff c , J.J.M. Verbaarschot d
a Department of Mathematical Sciences, Brunel University West London, Uxbridge UB8 3PH, United Kingdom
b Physics Department, Boston University, Boston, MA 02215, USA
c Nordita, Blegdamsvej 17, DK-2100, Copenhagen , Denmark
d Department of Physics and Astronomy, SUNY, Stony Brook, NY 11794, USA
Abstract
The microscopic spectral density of the QCD Dirac operator at nonzero baryon chemical potential
for an arbitrary number of quark flavors was derived recently from a random matrix model with the
global symmetries of QCD. In this paper we show that these results and extensions thereof can be
obtained from the replica limit of a Toda lattice equation. This naturally leads to a factorized form
into bosonic and fermionic QCD-like partition functions. In the microscopic limit these partition
functions are given by the static limit of a chiral Lagrangian that follows from the symmetry breaking
pattern. In particular, we elucidate the role of the singularity of the bosonic partition function in the
orthogonal polynomials approach. A detailed discussion of the spectral density for one and two
flavors is given.
PACS: 11.15.Ha; 05.50.+q; 12.38.Lg
1. Introduction
One of the features that makes QCD at nonzero baryon chemical potential both elusive
and interesting is that the Euclidean Dirac operator does not have any Hermiticity properE-mail address: split@alf.nbi.dk (K. Splittorff).
doi:10.1016/j.nuclphysb.2005.01.018
288
G. Akemann et al. / Nuclear Physics B 712 (2005) 287324
ties. The non-Hermiticity occurs in an essential way because a finite baryon number density
is obtained by promoting the propagation of quarks in the forward time direction and inhibiting the propagation in the backward time direction. Because of the non-Hermiticity
the quark determinant attains a complex phase, which prevents the analysis of the QCD
partition function by means of probabilistic methods. An approximation that is often used
at zero chemical potential is to ignore the fermion determinant altogether. However, this approximation fails dramatically at nonzero chemical potential [1]. In particular, the critical
chemical potential at zero temperature is determined by the pion mass instead of the nucleon mass. Therefore, the main source of information for QCD at nonzero baryon density
is based on simplified models or on perturbative expansions at very high densities which
may never be accessed experimentally. Lattice QCD simulations are feasible in a region
where the ratio of the chemical potential and the temperature is sufficiently small so that
extrapolation from an imaginary chemical potential [2,3] or from zero chemical potential
[4] becomes reliable (for a critical review of the field and additional references, we refer
to [5]).
In this paper, we will investigate an observable for which nonperturbative results can be
obtained for unquenched QCD at nonzero baryon density. This observable is the spectral
density of the QCD Dirac operator. At nonzero chemical potential its support is a twodimensional domain. The naive argument that observables do not depend on the chemical
potential at zero temperature does not apply to the spectral density of the Euclidean Dirac
operator. This puzzling fact can be understood as follows. To define an eigenvalue density
we need both the eigenvalues and the complex conjugate eigenvalues or the Dirac operator and its complex conjugate. However, complex conjugation is equivalent to changing
the sign of the chemical potential. Therefore, the generating function for the QCD Dirac
spectrum has to contain both quarks and conjugate quarks in the valence sector that have
an opposite baryon charge. This opens the possibility that the low energy limit of this
theory contains Goldstone bosons made out of quarks and conjugate anti-quarks which
have a nonzero baryon number [6,7]. More formally, the chemical potential in the generating function breaks the flavor symmetry resulting in a chiral Lagrangian that becomes
dependent on the chemical potential. The region of the Dirac spectrum we wish to analyze
determines the mass of the valence quarks [8,9]. In order to determine the eigenvalue density of the Dirac operator at small eigenvalues the masses of the valence quarks are equally
small. Consequently, the mass of the charged Goldstone modes is small and the generating
function depends on the chemical potential even at small values of .
The low energy effective theory for the generating function of the Dirac spectrum is
a theory of Goldstone modes determined by spontaneous breaking of chiral symmetry.
This effective theory is a version of a chiral Lagrangian [10] which takes into account the
presence of the conjugate quarks. The allowed terms in the effective theory are determined
by the flavor symmetries and the way they are broken by the quark masses and the chemical
potential: the effective theory must break the flavor symmetries in precisely the same way
as in the generating function we started from. To be specific, the Goldstone field, U , takes
values in the coset U SU(Nf + 2n), where n is the number of additional quarkconjugate
quark pairs. The leading order chiral Lagrangian in the standard counting scheme is the
usual nonlinear sigma model [11]. The dependence of the chemical potential is completely
fixed by the flavor symmetries. It enters as the zeroth component of an external vector field
289
through a shift of the Euclidean time derivative (see, e.g., [10,1215])

0 U 0 U i[B, U ],
(1)
where is the chemical potential and B is the charge matrix. Since the baryon charge of
conjugate quarks is opposite to the standard quarks [6] the charge matrix B is not proportional to the identity and the commutator is nonvanishing.
Although the eigenvalue spectrum of the Dirac operator is not a direct physical observable, we hope that detailed knowledge of this observable in a nonperturbative domain
where no other information is available, will ultimately lead to a better understanding of
the problems that hinder numerical simulations at nonzero baryon chemical potential. At
zero temperature these problems are manifest unless 2 F2 V 1, where F is the pion
decay constant. In this paper, we will focus on a scaling regime where the product 2 F2 V
is fixed as the volume, V , is taken to infinity. The fixed combination 2 F2 V can take any
value and our results will show the effect of the fermion determinant on the Dirac spectrum. The quark masses will be taken such that m scales like with m2 F2 V kept fixed.
To be specific, we consider the range [8,16]
m ,
1

L
(2)
as the volume L4 = V is taken to infinity, and is the scale of the lightest non-Goldstone
particle. This regime is sometimes known as the ergodic domain or as the epsilon regime1
of chiral perturbation theory [17]. In this domain the zero modes of the Goldstone fields
dominate the partition function which reduces to a static integral over the Goldstone manifold [14,17]. The static integral is completely determined by the flavor symmetries of
the QCD partition function. This implies that any theory with the same flavor symmetries
and flavor symmetry breaking is described by the same static integral. The only memory
of the underlying theory is two coupling constants, namely the chiral condensate and
the pion decay constant F . Hence the partition function in this limit is universal in the
sense that all theories with a given symmetry breaking pattern and a mass gap will have a
partition function given by the same static integral over the Goldstone manifold provided
that m , 1/L [9,1820]. The simplest theories in this class are invariant random
matrix theories in the limit of large matrices. Because of the large invariance group it is
sometimes simpler to analyze the random matrix theory rather than to evaluate the integral
over the Goldstone manifold directly.
In order to derive the microscopic eigenvalue density of the Dirac operator, which is
defined by rescaling its eigenvalues as zk V , we start from the microscopic partition
function with Nf + 2n flavors. Then we take derivatives with respect to the mass of the
additional n quarks and n conjugate quarks, and finally we remove the additional flavors by
taking the limit n 0. This procedure is known as the replica trick [21]. Since the microscopic partition functions are only known for integer n there is no guarantee that a correct
nonperturbative answer is obtained this way. After a two decade long discussion [22] two
1 The epsilon regime is the regime where m O( 2 ) and 1/L O(). These conditions are more strict than
the inequality (2). For example, we are still in the ergodic domain if 1/L O( 3/2 ).
290
closely related methods have emerged that result in the correct nonperturbative results: the
replica limit of the Painlev equation [23,24] and the replica limit of the Toda lattice equation [25,26]. Both the Painlev equation and the Toda lattice equation are well known in
the theory of exactly solvable systems (see, for example, [27,28]). The Painlev equation
is a complicated nonlinear differential equation, whereas the Toda lattice equation is a simple two step recursion relation. For that reason, it is much simpler to work with the Toda
lattice equation. The advantage of the Painlev equation is that nonperturbative results can
be obtained from fermionic partition functions only [23]. Recently, the consistency of the
replica limit of the Toda lattice equation [29] and the supersymmetric method [30] was
established. A necessary ingredient for the applicability of the Toda lattice method is that
the QCD partition function satisfies the Toda lattice equation. This is the case if the QCD
partition function is a -function [28]. For the low-energy limit of QCD at = 0 this was
shown for fermionic partition functions in [31,32]. These results were extended to bosonic
and supersymmetric partition functions as well as to QCD at nonzero chemical potential in
[25,26,29].
In the replica limit of the Toda lattice equation, generating functions with bosonic
quarks appear in addition to generating functions with fermionic quarks. While the fermionic generating functions for the spectral density at nonzero baryon chemical potential
are known from the integral over the Goldstone manifold [26,33], only the simplest case
with bosonic quarks is known [26]. The relevant supersymmetric generating functions, for
which the Goldstone manifold is a supermanifold, have proved to be quite challenging
to compute, and no explicit expressions have been obtained so far. In the present paper
we derive these generating functions from the random matrix model introduced in [34].
The flavor symmetries and their spontaneous and explicit breaking are exactly the same in
the large N limit of this random matrix model and in full QCD. Since both theories have
a mass gap, in the microscopic limit, they are both given by the same integral over the
Goldstone manifold.
In the framework of random matrix models, spectral correlation functions can also be
computed by means of (bi-)orthogonal polynomials in the complex plane. The complex orthogonal polynomial approach was developed in [3540] and was applied to unquenched
[34] QCD Dirac spectra at nonzero chemical potential. The outcome of our analysis here
is that both the Toda lattice and the orthogonal polynomials approach produce equivalent
results. The divergence of the bosonic generating functions also sheds some light on universality of non-Hermitian random matrix models in general.
Spectra of the non-Hermitian Dirac operator at nonzero baryon density have been obtained from lattice QCD in the quenched case [1,4143] and for QCD with two colors
[44,45] and have been compared successfully to random matrix theory [4143,45]. The
microscopic spectral density of quenched lattice QCD at nonzero baryon density was first
analyzed in [43] where quantitative agreement with analytical predictions [26,37] was
found in an asymptotic domain where the results derived from the chiral Lagrangian [26]
agree with the expression obtained in [37]. The two flavor phase quenched partition function which is calculated in this paper does not suffer from a sign problem either and could
be simulated on a lattice (see [46,47] for recent results). A first analytical prediction for
the unquenched QCD Dirac spectrum at nonzero baryon chemical potential in the nonperturbative regime was obtained in [34]. These results and the work presented in this paper
291
could in principle be compared to numerical simulations in the region where the sign problem sets in. In particular, any proposal for a solution of the sign problem on the lattice can
be tested against these predictions.
For the spectral density with one dynamical flavor we show both analytically and numerically that large fluctuations occur for increasing value of the chemical potential. These
findings further illustrate the difficulties encountered when one wishes to reproduce these
results in lattice QCD simulations.
In Section 2 we explain the derivation of the spectral density of the QCD Dirac operator
within the replica framework. The required generating functions are all given in terms of
group integrals that represent low-energy effective QCD-like partition functions. The bulk
of this paper is devoted to the calculation of the group integrals corresponding to bosonic
quarks using the large N limit of a random matrix model. After introducing the relevant
random matrix model for QCD with chemical potential in Section 3, we compute the necessary generating functions by means of complex orthogonal polynomials in Section 4. In
Section 5 we show that these partition functions satisfy the Toda lattice equation. This allows us to calculate the spectral density from the replica limit of the Toda lattice equation.
We show in general terms that the results agree with a direct computation using orthogonal
polynomials [34]. In Section 6 we present our results for one, two and any number of dynamical flavors and compare them to the quenched and phase quenched results. It also contains a discussion of the thermodynamic limit of our results. This section is self contained
and can be read independently of the previous sections. Our main findings are summarized
in the conclusions, and several technical details are worked out in two appendices.
2. The spectral density from generating functions

At nonzero baryon chemical potential the Dirac operator is non-Hermitian so that its
spectrum has a two-dimensional support in the complex plane. In this section we start with
the definition of the spectral density for complex eigenvalues. Then we show how it can be
obtained from a generating function with additional flavors.
We consider the eigenvalues of the Dirac operator given by the eigenvalue equation
(D + 0 )j = zj j .
(3)
Since (D = D and (0 = 0 the Dirac operator is neither Hermitian nor

anti-Hermitian and the eigenvalues are complex. Because of the axial symmetry {D +
0 , 5 } = 0 the nonzero eigenvalues come in pairs with opposite sign, zj . The density
of eigenvalues in the presence of Nf flavors is defined as the vacuum expectation value
over a sum of delta functions in the complex plane at the positions of the eigenvalues,
vanishing and nonvanishing,

Nf
(2)
2
z, z , {mf }; + (z)
(4)
(z zj )
,
)
QCD,
where is the chemical potential, Nf is the number of flavors and {mf } = m1 , . . . , mNf
are the quark masses. For later convenience we have not included the zero eigenvalues in
292
our definition of the spectral density. The average over gauge fields in a fixed topological
sector, , is defined by

Nf
[dA] O f =1
det(D + 0 + mf )eSYM (A)
.
OQCD,
(5)
Nf
[dA] f =1
det(D + 0 + mf )eSYM (A)
Since the phase of the product of the fermion determinants is nonvanishing, such expectation values of operators will in general be complex. In particular we will find that the
unquenched spectral density is complex rather than real nonnegative.
The two-dimensional -functions appearing in the definition of the eigenvalue density
can be obtained from the derivative

1
N
2 (z zj )
= z G f z, z , {mf }; ,
(6)
QCD,
j
with the resolvent defined by the average

1
Nf
G z, z , {mf }; =
.
z + zj QCD,
(7)
One way to calculate the resolvent is by means of the fermionic replica trick. In the usual
replica trick the resolvent is calculated from the identity
1
1
(8)
= lim z (z + zj )n .
z + zj n0 n
j
The problem is that, in most cases, the average of the right-hand side can only be obtained
for nonzero integer values of n, and the limit n 0 can only be taken after a proper
analytical continuation in n. For a non-Hermitian operator this procedure does not work.
Instead, a modified identity has been suggested [48]
1
2n
1
(9)
= lim z (z + zj ) .
z + zj n0 n
j
At the perturbative level the correct resolvent can be derived this way [7,48]. However,
for nonperturbative calculations this relation only leads to correct results if it is used in
combination with the Toda lattice equation [25].
The generating function for the resolvent is thus given by

N ,n
Z f {mf }, z, z ;
Nf

2n
det(D + 0 + mf ) det(D + 0 + z) eSYM (A) . (10)
= [dA]
f =1
The resolvent is obtained by

1
N
N ,n
G f z, z , {mf }; = lim z log Z f {mf }, z, z ; ,
n0 n
(11)
293
and the spectral density is given by

1
N
N ,n
f z, z , {mf }; + (2) (z) = lim
z z log Z f {mf }, z, z ; .
n0 n
(12)
This construction has been used previously for perturbative [7,48] and nonperturbative [24,
26] studies of the eigenvalue density in quenched theories (i.e., for Nf = 0).
The interpretation of the partition function (10) is [6,7] that in addition to the usual
quarks we have n quarks with mass z and n conjugate quarks with mass z . Since

det(D + 0 + z) = det 5 (D + 0 + z )5 = det(D 0 + z ),
(13)
conjugate quarks carry the opposite baryon number of quarks. The baryon charge matrix
in the partition function (10) is therefore not proportional to the identity and the flavor
symmetry is broken by the chemical potential. The specific way in which the flavor symmetries are broken determines the low-energy effective theory. In particular, the effective
theory will depend on for energies where the pions dominate the free energy. Physically
this is clear because mesons made out of quarks and conjugate anti-quarks carry a nonzero
baryon number. The general prescription for constructing a chiral Lagrangian is to impose
the transformation properties of the microscopic theory on the low-energy effective theory
[10]. For an external vector field the microscopic partition function is invariant under local
gauge transformations of this field. Therefore, the derivative in the chiral Lagrangian has
to be replaced by the corresponding covariant derivative [10,12].
In this paper we calculate the eigenvalue density in the microscopic limit for any number
of flavors Nf , where mf V , zV , and 2 F2 V are held fixed as V . In this limit the
kinetic terms factorize from the partition function and the low energy limit of the generating
function (10) is given by [14]
N ,n
Z f

{mf }, z, z ;
V 2 2
1
1
1
dU det(U ) e 4 F Tr[U,B][U ,B]+ 2 V Tr M(U +U ) .
(14)
U U (Nf +2n)
The quark mass matrix is given by M = diag(m1 , . . . , mNf , {z}n , {z }n ) and the charge
matrix B is a diagonal matrix with 1 appearing Nf + n times along the diagonal and 1
appearing n times. An explicit expression for this integral was derived in [26,33]. In this
paper we will show that the partition functions (14) satisfy the Toda lattice equation

1
N ,n
z z log Z f {mf }, z, z ;
n
Nf ,n+1
Z
1
= (zz )2
2
N ,n1
({mf }, z, z ; )Z f
N ,n
({mf }, z, z ; )
[Z f ({mf }, z, z ; )]2
(15)
294
where z = z d/dz and z = z d/dz . Since the replica limit of the right-hand side of this
equation is the spectral density2 we find the remarkably simple expression [25,26,29]
N ,n=1
zz Z f
N
f z, z , {mf }; =
2
N ,n=1
({mf }, z, z ; )Z f
N ,n=0
[Z f
({mf }|z, z ; )
({mf }; )]2
(16)
In (16) and elsewhere in this paper it is our notation to separate the fermionic and bosonic
quark masses in the argument of the partition function by a vertical line. In addition to
N ,n
the fermionic partition functions, we will need to evaluate Z f for n = 1, i.e., for
one bosonic quark and one bosonic conjugate quark. In the quenched case (Nf = 0), this
partition function is given by the integral [26,49]
Zn=1 (z, z ; )
V 2 2
i
dU
1
T
1
= lim C
(U )e 4 F Tr[U,B][U ,B]+ 2 V Tr M (U I U I ) ,
2||
0
det
U
(17)
where dU (U )/ det2 U is the integration measure on positive definite 2 2 Hermitian

matrices and

0 1
1 0
z
and I =
.
B=
,
M=
(18)
1 0
0 1
z
It was found that in order to obtain a finite limit 0, the normalization constant has
to be chosen C 1/ log . The appearance of this logarithmically diverging term is not
surprising since the inverse fermion determinants are regularized as
det1 (D + 0 + z) det1 (D + 0 + z )

D + 0 + z
= lim det1
.

D + 0 + z
0
(19)
Zero eigenvalues of D do not lead to zero eigenvalues of the operator in the r.h.s. of
this equation. Therefore, generically its eigenvalues k are nonzero real and occur in pairs
k , and we expect that the -dependence enters in a similar way to the mass dependence
of the QCD partition function with one bosonic quark at zero topological charge. In the
microscopic limit this is given by K0 (m) which diverges logarithmically in m.
To derive the partition functions (14) and (17) we have only used the global symmetries
of the underlying QCD partition functions. In particular, this implies that any microscopic
theory with the same global symmetries and spontaneous breaking thereof will have the
same zero momentum effective theory. The simplest theory in this class is chiral random
matrix theory at nonzero chemical potential which is obtained from the QCD partition
function by replacing the matrix elements Dirac operator by (Gaussian) random numbers.
The partition function (17) was also derived explicitly starting from a chiral random matrix model instead of only using symmetry arguments [26]. Another advantage of using
2 Notice that both in the l.h.s. and the r.h.s. the overall factors z and z are canceled so that the l.h.s. of (15)
gives the spectral density without the zero eigenvalues.
295
a random matrix model is that one can easily perform numerical simulations. For example, the quenched spectral density was calculated numerically [26] and was found to be in
agreement with (16).
Below we will show that the replica limit of the Toda lattice equation (16) can also
be used to obtain the unquenched spectral density using a similar regularization of the
N ,n=1
with Nf
= 0. This
partition function with bosonic quarks. In this case we need Z f
f |2). This integral is not known explicitly for
involves an integral over the supergroup Gl(N
N ,n=1
Nf
= 0. For that reason we will derive Z f
directly from the corresponding random
f |2).
matrix model instead of the low-energy effective partition function based on Gl(N
3. Random matrix model for QCD at nonzero chemical potential

In this section we define the random matrix model which we will use to calculate the
generating functions. Random matrix models for QCD originally [8] focused on the quark
mass dependence at zero chemical potential. Later [7], a random matrix model including
the chemical potential successfully explained why quenched lattice QCD at zero temperature has a phase transition at a chemical potential of half the pion mass. However,
a disadvantage of this model is that no eigenvalue representation is known which is required for the use of orthogonal polynomial methods. The random matrix model that will
be used in the present paper was introduced in [34]. It differs from the model in [7] by a
different form of the chemical potential term. Its partition function can be reduced to an
eigenvalue representation and allows for an explicit solution in terms of orthogonal polynomials [34]. Because this model captures the correct global symmetries of the QCD Dirac
operator, its low energy limit is also given by (14). A related model defined in terms of a
joint eigenvalue distribution was introduced in [37]. Although this model is not in the universality class of QCD partition functions, the complex orthogonal polynomial methods
that were developed for the derivation of the spectral density are applicable to the random
matrix model introduced in [34]. In Section 5 we will comment on the relations between
the different models.
The random matrix partition function with Nf quark flavors of mass mf and n replica
pairs of regular and conjugate quarks with masses y and z each is defined by

{mf }, y, z ;
N ,n
ZN f
d d wG ()wG ( )
Nf

det D() + mf
f =1

det D() + y detn D () + z ,
n
where the non-Hermitian Dirac operator is given by

0
i +
.
D() =
0
i +
(20)
(21)
Negative numbers of flavors Nf , n < 0 will denote insertions of bosons given by inverse
powers of determinants. To allow for the presence of both bosonic and fermionic flavors
we will distinguish between Nb and Nf in later formulas. Here and are complex
296
(N + ) N matrices with the same Gaussian weight function

wG (X) = exp N Tr X X .
(22)
The matrix D() replaces the QCD Dirac operator plus chemical potential in (3). It has
exactly zero modes which identifies as the absolute value of the topological charge.
The model (20) has the same flavor symmetries as QCD which are broken by the quark
masses and the chemical potential in exactly the same way as in QCD. The only difference
with the model [7] is that the chemical potential is multiplied by a complex matrix
with Gaussian distributed matrix elements. In the microscopic limit, where 2 N , mf N
and zN are fixed as N this partition function will match the partition function (14)
after fixing the scale of the parameters appropriately. The latter is uniquely determined in
terms of the flavor symmetries and their breaking, and thus universal. As was already done
in the previous section, the microscopic limit of the partition function will be denoted by
Z instead of the notation Z which we use both for the QCD partition function and the
random matrix theory partition function. Generally, the overall normalization of Z and the
microscopic limit of Z is different.
The form of the Gaussian weight wG () respects the flavor symmetries, but these symmetries do not exclude traces of higher powers of in the exponent as well as other
noninvariant terms. At zero chemical potential it was shown [5057] that partition functions and eigenvalue correlations in the microscopic limit are independent of the form of
this weight, which, in random matrix theory, is known as universality. The only condition
for the probability density is that the spectral density near the origin is nonvanishing in the
thermodynamic limit. In the derivation of the joint eigenvalue density of (20) it is essential
that the two weight functions are Gaussian. However, we believe that this is only a technical requirement and higher order invariant terms will not alter the microscopic limit of the
joint probability distribution (for more discussion see Section 5).
In [34] it was shown that the partition function (20) has an eigenvalue representation
after choosing an appropriate representation for the matrices and yielding
Nf

N

N ,n
{mf }, y, z ; (yz )n
mf
d 2 zk P f {zi }, zi ; a ,
N ,n
ZN f
f =1
(23)
C k=1
where the integration extends over the full complex plane and the joint probability distribution of the eigenvalues is given by
N

1 2 Nf ,n
{zi }, zi ; a = 2N N zl2
zk , zk ; a .
w
N ,n
P f
(24)
k=1
The Vandermonde determinant is defined as

N

2
2

zi zj2 ,
N zl
i>j =1
and the weight function reads
(25)

N ,n
zk , zk ; a
w f

=
297

Nf

2
2
n
n
N (1 + 2 )
2
2
mf zk
y zk2 z2 zk2 |zk |2+2 K
|z
|
k
22
f =1

N (1 2 ) 2
2
exp
z
,
+
z
k
k
42
(26)
where K is a modified Bessel function. The parameters of the weight function, {mf }, y, z
and are collectively denoted by a. In the quenched case, Nf = n = 0, the weight function will be denoted by w(zk , zk ; ). We have included a normalization factor 1/2N in
Eq. (24). This way it reduces to the joint eigenvalue density of the chGUE times a product
of delta-functions in the limit 0. We expect that the terms that are nonvanishing in
the microscopic limit are universal and are required for the derivation of the microscopic
spectral density.
In the asymptotic limit, N|zk |2 /2 1, the modified Bessel function in the weight
is well approximated by its leading order asymptotic expansion. The corresponding joint
eigenvalue distribution was introduced in [37] as a model for QCD at nonzero chemical
potential. Let us look closer at the condition N |zk |2 /2 1. In terms of the half-width,
xmax 2 , of the cloud of eigenvalues and the spacing, 1/N , of the imaginary part of
the eigenvalues it can be rewritten as
|zk |2 /(xmax ) 1.
(27)
For xmax > this implies that |zk | where we expect to find the weight function of
the Gaussian Unitary Ensemble perturbed by an anti-Hermitian matrix which indeed has a
Gaussian weight function (see [36]). In other words, the appearance of the modified Bessel
function K is a direct consequence of the chiral symmetry of the problem.
4. Generating functions from random matrix models

The aim of this section is to derive explicit expressions for the generating functions introduced in Section 2 using the random matrix model introduced in the previous section. In
particular, we will present new expressions for the corresponding partition functions with
any given number of fermions and one pair of conjugate bosonic quarks, as they are needed
for the replica limit of the Toda lattice equation. We will find that the partition functions
with bosonic replicas have to be regularized. This divergence will lead to a factorization
into a partition function containing only fermions and a purely bosonic one. After taking
the microscopic limit, they can be expressed entirely in terms of known partition functions
(14) and (17). In the following section we consider partition functions with Nf flavors
and n fermionic replicas to show that the random matrix model partition functions satisfy
the Toda lattice equation (15).
298
4.1. Orthogonal polynomials and Cauchy transforms

As is the case for Hermitian random matrix models, the partition functions (23) can
be expressed in terms of orthogonal polynomials, and the integrals to obtain the spectral
correlation function can be performed by means of the orthogonality relations. In this section we present the polynomials corresponding to the weight function (26) for Nf = n = 0
(which is denoted by w(z, z ; )).
The orthogonal polynomials are defined as solutions to the following orthogonality relation
(28)
d 2 z w(z, z ; )pk (z)pl (z) = rk kl
C
where the integral is over the full complex plane with measure d 2 z = d Re z d Im z, and
rk denotes the squared norms of the polynomials. Since the quenched weight w(z, z ; )
appearing in this relation is real positive the polynomials form a complete set with positive squared norms rk . The polynomials pk (z) depend only on the variable z and not its
complex conjugate, as is indicated by the notation. They are also functions of the chemical
potential , the size N of the random matrix, as well as of the topological charge . We
suppress the dependence on , N and .
In monic normalization, pk (z) = z2k + , the solution of (28) is given by [34]

1 2 k
N z2
()
pk (z) =
(29)
k! Lk
.
N
1 2
Since the Vandermonde determinant only contains squared variables only polynomials in
those squared variables will appear. The same happens for the polynomials on the real line
when mapping the Laguerre ensemble from the positive real numbers to the full real line.
For the weight (26) the norms are given by
rk =
2 (1 + 2 )2k+ k! (k + )!
.
N 2k++2
(30)
Because the Laguerre polynomials Lk (z) have real coefficients it holds that pk (z) =
pk (z ).
The Cauchy transform of the polynomial pk (z) is defined by
1
.
hk (y) d 2 z w(z, z ; )pk (z) 2
(31)
z y2
()
We note that h0 (y) has a nontrivial y-dependence, unlike the polynomial p0 (y). In contrast
to random matrix models with eigenvalues on a curve in the complex plane, the integral
over the pole in the complex plane is always well-defined. However, if we take the limit
0 of an anti-Hermitian Dirac operator we have to make sure that the argument y lies
outside the support of eigenvalues.
The leading order asymptotic large-y behavior of hk (y) is obtained by expanding
1/(y 2 z2 ) in a geometric series. Because of orthogonality all coefficients up to order
1/y 2k vanish resulting in the asymptotic behavior

rk
1
hk (y) = 2k+2 + O 2k+4 .
y
y
299
(32)
This result is useful for checking the identities derived below.

4.2. Partition functions with one flavor
The polynomials and their Cauchy transform enjoy the following relations to one flavor
partition functions. The former is related to a single fermionic flavor partition function
through the so-called Heine formula,

N
N =1

ZN f (z; )
z2 zj2
z
= z pN (z).
(33)
N =0
ZN f
j =1
Nf =n=0
Here we have written the fermion determinant as an expectation value with respect to
the quenched partition function at Nf = n = 0. (Vanishing replica (or flavor) indices will
be commonly suppressed throughout the following.) For < 1 the partition function is
manifestly real and positive for real masses z, as the orthogonal polynomials are Laguerre
polynomials of argument N z2 /(1 2 ) with zeros on the imaginary axis. For > 1 the
zeros are located on the real axis. Remarkably, the YangLee zeros [58] of the model in [7]
behave quite different from the model (20) discussed in this paper. The polynomials can
also be interpreted as characteristic polynomials.
The relation (33) can also be written down for pl
=N (z), expressing it as an average over
l variables. However, in order to be precise we would have to chose the weight function
in (26) to be N -independent as in [3840]. Since we are mainly interested in taking the
large-N limit we prefer to explicitly keep the N or volume dependence inside the weight.
For l N we can safely ignore this technical subtlety.
Turning to the Cauchy transform of pN (z), it is given by the partition function with one
bosonic quark flavor [39,40],

N =1
N

(z; )
ZN f
1
1
z
=
z hN 1 (z).
(34)
Nf =0
2 z2 )
r
(z
N 1
Z
j
N
j =1
Nf =n=0
The same remarks about positivity and the N -dependence of the weight function made
before also apply here.
Our program is to express spectral correlation functions in terms of partition functions.
This has the advantage that the microscopic limit is given by a group integral based on the
symmetries of the partition functions. To achieve our goal we also need all possible two
flavor partition functions. At zero chemical potential the program of expressing spectral
correlation functions in terms of multiflavor partition functions was carried out in [59,60].
4.3. Partition function with two flavors
It is known from Hermitian random matrix models [61] that the expectation value of two
characteristic polynomials is proportional to the kernel of the orthogonal polynomials (28).
300
As was shown in [38], for an arbitrary weight function, similar results can be derived
for non-Hermitian random matrix models. Since we are now dealing with non-Hermitian
matrices we have to distinguish between a characteristic polynomial (or mass term) and
its Hermitian conjugate. For a single pair of one fermionic quark with mass z and one
conjugate fermionic quark (with complex conjugated eigenvalues) with mass u we have
the relation [38]

N

2
2

ZNn=1 (z, u ; )

2
2
z zj u zj
(zu )
N =0
ZN f
j =1
Nf =n=0
= rN (zu ) KN +1 (z, u ),
KN +1 (z, u )
N

j =0
1
pj (z)pj (u ).
rj
(35)
(36)
Here we have defined the bare kernel KN (z, u ) which differs from the full kernel by the
square root of the weight function at each argument. The full kernel appears in the computation of correlation functions from orthogonal polynomials (see Section 5). In particular,
for z = u we obtain the result for one pair of fermionic replicas n = 1

N

2
2
ZNn=1 (z, z ; )
2
2
z z
= |z|
= rN |z|2 KN +1 (z, z ).
(37)
j
Nf =0
ZN
j =1
Nf =n=0
This quantity is manifestly real and positive. On the other hand, if we compute the partition
function with two flavors with the same eigenvalues, we obtain [38]

N =2
N

ZN f (x1 , x2 ; )
x12 zj2 x22 zj2
(x1 x2 )
N =0
ZN f
j =1
Nf =n=0

(x1 x2 ) pN (x1 ) pN +1 (x1 )
.
= 2
(38)
(x x 2 ) pN (x2 ) pN +1 (x2 )
2
For Hermitian random matrix models the two results (35) and (38) are related through the
ChristoffelDarboux formula, which in general does not hold in the complex plane (see
also Appendix A).
We now turn to the most important two-flavor partition function, containing a pair of
a bosonic quark and its conjugate. In the replica limit of the Toda lattice equation this
partition function appears when computing the quenched density (see (16)). In the microscopic limit this partition function was calculated in [26], both by starting from a random
matrix model, and by integrating explicitly over the Goldstone manifold in the chiral Lagrangian (17). It turned out that this partition function is singular and has to be regularized
according to the prescription given in (19). The partition function is obtained in the limit of
vanishing regulator, 0, where it diverges as log(). This diverging constant can be
absorbed in the normalization of the partition function. Below we will establish the nature
of this singularity in the orthogonal polynomial approach. In fact, it does not enter in the
calculation of the spectral density for the random matrix model (20) by means of complex
orthogonal polynomials [34]. However, we will show that it occurs even at finite N in the
301
random matrix model partition function with one bosonic quark and one conjugate bosonic
quark of the same mass.
Generalizing the results in [39] to the chiral case we find that the following relation
holds

N

ZNn=1 (y, x ; )
1

(yx )
N =0
(y 2 zj2 )(x 2 zj2 ) N =n=0
Z f
j =1
N
(yx )
=
AN 2 (x , y),
rN 1
AN 2 (x , y) Q(x , y; ) +
N
2

j =0
where
Q(x , y; )
d 2 z w(z, z ; )
1
hj (x )hj (y),
rj
1
.
(z2 x 2 )(z2 y 2 )
(39)
(40)
(41)
Eq. (40) defines a bare kernel containing only Cauchy transforms. For different masses
x
= y both the sum over Cauchy transforms and the function Q(x , y; ) are finite. At
equal arguments, x = y, the sum remains finite but the integral in (41) is logarithmically
divergent, just like the divergence encountered in (17). In order to regularize this divergence
we cut out a circle C(x, ) of radius around the pole at z = x and one around the pole
at z = x. To evaluate the singular part of the regularized Q (x , x; ) we expand the
weight function around the singular point and keep only the leading part. For |x| we
find for the pole at z = x
1
dz dz
Q (x, x ; ) w(x, x ; )
2
2
2i (z x )(z2 x 2 )
CC(x,)
= w(x, x ; )
CC(x,)
= w(x, x ; )
C(x,)

log((z x )/(z + x ))
dz dz 1
z
2i 2x
z2 x 2
dz 1 log((z x )/(z + x ))
2i 2x
z2 x 2

w(x, x ; )
=
log + O 0 .
(42)
2xx
We find the same contribution from the pole at z = x. Thus we have obtained the following interesting relation for the singular part of the bosonic partition function
ZNn=1 (x, x ; )
N =0
ZN f ()

w(x, x ; )
log + O 0 .
2+2
2|x|
rN 1
(43)
We are not aware that such a relation between the weight function (26) of complex orthogonal polynomials and the bosonic partition function at a degenerate mass pair was
302
previously known. It translates easily to nonchiral non-Hermitian random matrix models.

Relation (43) shows that the singularity that was found in [26] for the microscopic limit
of ZNn=1 (x, x ; ) even persists for finite size random matrices and can be expressed
simply in terms of the weight function. Only weak assumptions have been made in the
derivation of (43), and its validity extends to a wide class of weight functions. For x O()
the above derivation does not apply. Instead we find that limx0 Q (x, x ; ) 1/ 2 . As
we will explain in the next section, the logarithmic singularity we have found for the twopoint function (43) will also occur in more general partition functions with at least one pair
of conjugate bosonic quarks.
The last two-point function we need below is the partition function with one fermion
and one boson. This partition function reads [39]
N =1,Nb =1
ZN f
(x|y; )
N =0
ZN f
N

2
2
x (x zj )
y
(y 2 zj2 ) N =0
j =1
f

x

= y2 x2
NN 1 (x, y),
y
(44)
N 1
NN 1 (x, y) =
1
1
+
pj (x)hj (y).
2
2
rj
(y x )
(45)
j =0
Here, we have defined the bare kernel NN 1 (x, y) consisting of orthogonal polynomials
and Cauchy transforms. This partition function is nonsingular for any values of the arguments and correctly normalizes to unity at equal arguments y = x.
The remaining two flavor partition functions combining (conjugate) fermionic quarks
and (conjugate) bosonic quarks are given in Appendix A.
4.4. Partition functions with an arbitrary number of flavors and one conjugate flavor
General expressions for the expectation value of ratios of characteristic polynomials
valid for an arbitrary complex weight function were given in [3840]. We will apply these
results to partition functions with Nf flavors plus one pair of regular and conjugate fermionic (n = +1) or bosonic (n = 1) flavors which both enter in the expression for the
spectral density (16) after taking the replica limit n 0.
The partition function with Nf fermionic quarks with masses mf is given by [38]

ZN f ({mf }; )
N =0
ZN f
Nf

f =1
mf

N

2

2
mf zj
j =1
Nf =n=0
Nf

f =1 mf
det
pN +k1 (ml ) .
2
Nf ({mf }) 1k,lNf
(46)
303
Next we add a single pair of a fermion and its conjugate. This partition function is given
by [38]
N ,n=+1
ZN f
({mf }, z, z ; )
N =0

|z|
ZN f
Nf

mf
f =1

Nf
N

2
2
2

2
2
z z
mf zj
j
j =1
f =1
Nf
mf
(1)Nf rN |z|2 f =1
Nf +1 ({m2f }, z2 )
p

N +1 (m1 )

..

.

pN +Nf (m1 )

KN +1 (m1 , z )
...
..
.
...
...
pN +1 (mNf )
..
.
pN +Nf (mNf )
KN +1 (mNf , z )
Nf =n=0
pN +1 (z)

..

,
.

pN +Nf (z)

KN +1 (z, z )
(47)
where KN +1 (x, y) is the bare kernel introduced in (36). In the replica limit of the Toda lattice equation, the partition function with one pair of conjugate bosons instead of conjugate
fermions enters as well. It is given by [39] (see also Appendix A)
N ,n=1
ZN f
({mf }|y, x ; )
N =0
ZN f
Nf
f =1 mf
(yx )
Nf
N

j =1

Nf

2

1
2
mf zj
(zj2 y 2 )(zj2 x 2 ) f =1
N
2
2
f =1 mf (mf y )
rN 1 Nf ({m2f })(yx )

pN 1 (m1 )

..

.

pN +Nf 2 (m1 )

NN +Nf 2 (m1 , y)
...
..
.
...
...
pN 1 (mNf )
..
.
pN +Nf 2 (mNf )
NN +Nf 2 (mNf , y)
f =0

hN 1 (x )

..

.
.

hN +Nf 2 (x )

AN +Nf 2 (x , y)
(48)
If we take the limit of degenerate bosonic masses, y x, the function Q(x , y; ) inside
the matrix element AN +Nf 2 (x , y) becomes singular, while all other matrix elements
remain finite. Expanding the determinant with respect to the last row this singular part
Q (x , x; ) gives the dominant contribution,
N ,n=1
ZN f
({mf }|x, x ; )
N =0
ZN f
Nf
0
2
2

f =1 mf (mf x )
pN +k2 (ml ) AN +Nf 2 (x , x).
det
2
rN 1 Nf ({mf })|x|2 1k,lNf
(49)
304
Using the relations (40) and (46) we thus arrive at the following factorized result,
N ,n=1
lim
0
ZN f
({mf }|x, x ; )
N =0
ZN f
N
Nf

2
Z f ({mf }; )
Z n=1 (x, x ; )
mf x 2 N 1 N =0
lim N N =0
.
f
0
ZN 1
ZN f
f =1
(50)
Both sides are regularized by replacing Q(x, x ; ) by Q (x, x ; ) as discussed

above (42). This is the main result of this section. Let us stress that the factorization we obtained is not a consequence of the suppression of nonplanar diagrams in the large N limit.
In (50) it occurs at finite-N and is strictly due to the singularity for coinciding arguments.
This completes the derivation of all generating functions required for the derivation of the
spectral density from the replica limit of the Toda lattice equation (16).
If we were to compute not only the spectral density but also higher order correlation
functions from the Toda lattice approach (see, e.g., in [26]), we would have to add more
pairs of replicated bosons, each with different masses yj and xj . Taking the degenerate
n =1
limit yj xj would lead to the same type of factorization into j ZNj
(xj , xj ; )
and a remaining fermionic partition function.
5. Toda lattice equation from orthogonal polynomials

The purpose of this section is threefold. First we show that the generating functions with
n pairs of fermionic replicas satisfy the Toda lattice equation. Since the microscopic limit
N ,n
of the random matrix partition function, ZN f , is given by the effective partition function
(14) [26,33], this result also shows that the Toda lattice equation (15) for Nf = 0 [26] can
be extended to arbitrary Nf . Second, we show that the spectral density obtained from the
replica limit of the Toda Lattice equation agrees with the result computed from orthogonal
polynomials [34]. This extends the agreement between the two approaches to cases where
the weight in the partition function is no longer positive definite. We close this section with
some remarks about universality.
5.1. The spectral density from the Toda lattice equation
We start by showing that the Toda lattice equation (15) can be modified to hold for
random matrix model partition functions at finite N . The starting point of the evaluation
is the expression for the partition function with Nf + n ordinary flavors and n conjugate
fermionic flavors generalizing (47). While this result is given in [38] for nondegenerate
masses we have to take the limit where the replicated masses become degenerate. This
leads to successive derivatives in these variables of the polynomials and of the kernels (36).
For simplicity we only display the result for Nf = 1,

N =1,n
ZN f
(m, z, z ; )
N =0
ZN f
305
m |z|2n
N +n1
rj
j =N
2
2
n
n(n1)
(m z ) |z|

pN +n (z)
pN +n (m)

KN +n (m, z )
KN +n (z, z )

z KN +n (m, z ) z KN +n (z, z )
.
.

.
.

.
.
n1
KN +n (m, z ) n1
z
z KN +n (z, z )
z pN +n (z)
z KN +n (z, z )
z z KN +n (z, z )
.
.
.
n1
z z KN +n (z, z )
..
.

zn1 pN +n (z)

zn1 KN +n (z, z )
z zn1 KN +n (z, z )

.

.

.

n1 K

zn1
z
N +n (z, z )
(51)
omitting constant factors. Here the derivatives are defined as
d
d
,
z = z .
(52)
dz
dz
We observe that this expression depends on the matrix size explicitly through the combination N + n and implicitly through factors that originate from the N -dependence in the
exponent of the weight function. Using the Sylvester identity as in [26] while keeping only
the N -dependence that enters as N + n we easily derive the Toda lattice equation,
z = z
N =1,n
z z log ZN f
(m, z, z ; )
N =1,n+1
n(zz )2
f
ZN 1
N =1,n1
f
(m, z, z ; )ZN +1
N =1,n
[ZN f
(m, z, z ; )
(m, z, z ; )]2
(53)
again ignoring constant factors. This equation is valid at finite N provided that the factor N
that appears explicitly in the exponent of the weight function is not changed. The subscript
of the partition function thus denotes the total number of complex integration variables.
For large N this distinction can be ignored. It is not difficult to generalize the Toda lattice
equation to arbitrary Nf . The result is

N ,n
z z log ZN f {mf }, z, z ;
N ,n+1
n(zz )
f
ZN 1
N ,n1
f
({mf }, z, z ; )ZN +1
N ,n
({mf }, z, z ; )
[ZN f ({mf }, z, z ; )]2
(54)
We still have to fix the overall normalization constant of the partition functions which in
principle depends on N, Nf , as well as on the replica index n. Since the l.h.s. of (54)
is linear in n we have to choose the normalization constants in the r.h.s. to retain this
n-dependence. With the appropriately adjusted constants the random matrix model generating functions satisfy the equality (15).
N ,n
Since the microscopic limit of the random matrix partition functions ZN f ({mf }, z,
z ; ) is given by the group integrals (14), we conclude that (15) is satisfied for any number
of flavors. Explicit expressions for these partition functions in terms of Bessel functions
will be given in Section 6 where we also give results for the spectral densities.
306
Before doing so let us step back to discuss what we have achieved for the spectral density using the replica limit of the Toda lattice equation (54). Combining Eqs. (15) and (12)
with the factorization (50) and canceling one of the partition functions in the denominator
of the Toda lattice equation, we arrive at the following result for the spectral density,

N
f z, z , {mf };
Nf ,n=1
Nf
n=1
({mf }, z, z ; )
ZN 1
2 zz 2
2
=
,
mf z ZN +1 (z, z ; )
N
log 2
Z f ({mf }; )
f =1
(55)
where we have explicitly displayed the divergent normalization factor (/2) log which
cancels the divergent factor that was obtained in (43). The -dependence of the spectral
density is contained implicitly in the partition functions. This is the central result of the
Toda lattice approach applied to the random matrix model (20). The spectral density factorizes into QCD like partition functions which, at low energy, can be expressed as group
integrals of the form (14) and (17).
Let us discuss some properties of the spectral density. First, it is symmetric under
N ,n=1
z z (together with z z ), because the partition function ZN f
contains only
n=1
is even as well. This is a consequence of the axial symmetry
squared variables and ZN
of the partition function which requires that the nonzero eigenvalues occur in pairs of op Nf
N ,n=1
posite sign. Inserting the result (47) for ZN f
we see that the prefactor f =1
(m2f z2 )
is canceled from the Vandermonde (for an explicit expression in terms of determinants
N ,n=1
see (57) below). Replacing the polynomials and kernels in ZN f
by their expressions
in terms of partition functions we notice one further symmetry (for real quark masses),

N
N
f z, z , {mf }; = f z , z, {mf }; ,
(56)
that is taking the complex conjugate and at the same time exchanging z and z . These
symmetries follow directly if we write the spectral density as an integral over the joint
eigenvalue distribution. Thus the real part is symmetric about the real axis in the complex
eigenvalue plane while the imaginary part is anti-symmetric. Therefore, the eigenvalue
density on the real axis is real. However, away from the real axis it is, in general, neither
real nor positive and therefore does not define a probability density. The lack of reality
directly relates to the asymmetric way the variables z and z enter in the polynomials and
kernel respectively. The conjugation symmetry is related to the invariance of the partition
function under changing the sign of , and is valid for QCD beyond the ergodic domain (2)
as well.
5.2. Equivalence of the Toda lattice and orthogonal polynomials approach
In order to compare the result (55) from the replica limit of the Toda lattice equation to
the result from orthogonal polynomials [34] it is instructive to rewrite (55) in terms of the
weight function, polynomials and kernels. Reinserting (43), (47) and (46) we obtain
307

N
f z, z ; {mf },

...
pN (mNf )
pN (z)
pN (m1 )

..
..
..

..

.
.
.
.

pN +Nf 1 (m1 ) . . . pN +Nf 1 (mNf ) pN +Nf 1 (z)

KN (m1 , z )
...
KN (mNf , z )
KN (z, z )
,
= w(z, z ; )
det1k,lNf [pN +k1 (ml )]
(57)
again after normalizing appropriately. As we will explain next, this is precisely the result
that was obtained in [34].
It is well known from random matrix theory that the spectral density can be written
in terms of a kernel of orthogonal polynomials [62]. This also holds for complex weight
functions. However, if the weight function is not a symmetric function of z and z , we have
to use bi-orthogonal polynomials in the complex plane defined by [39,63],
(N )
N ,0
(N )
(N )
(58)
d 2 z w f z, z ; {mf }, pk f (z)ql f (z ) = rk f kl ,
C
(N )
assuming the pseudo norms rk f to be nonvanishing. (Bi-orthogonal polynomials in general no longer form a scalar product with positive definite norms.) In contrast to orthogonal
polynomials on the real line, the pk (z) and qk (z) are in general different polynomials.
Defining their kernel by
N ,0
N ,0
12
(N )
KN f (z, u ) w f z, z ; {mf }, w f u, u ; {mf },
N
1
(Nf )
j =0 rj
(Nf )
pj
(Nf )
(z)qj
(u ),
(59)
the following expression for the correlator of k complex eigenvalues holds [62]
(N )

N
f zl , zl l=1,...,k ; {mf }; = det KN f zj , zl .
1j,lk
(60)
In particular, for the spectral density we obtain

1

N
N
N ,0
f z, z ; {mf }; = w f z, z ; {mf };
(Nf )
j =0 rj
(Nf )
pj
(Nf )
(z)qj
(z ).
(61)
In [34] it was shown that the kernel for Nf flavors can be expressed as a determinant of
kernels for zero flavors which immediately leads to the result (57).
This establishes the equivalence between the Toda lattice approach and the biorthogonal polynomial approach at the level of the spectral density. Starting from the
observation that also in the case of generating functions for multi-point correlation functions the partition function ZNn=1 (zl , zl ; ) factors out, we are confident that the equiv-
308
alence between the Toda lattice approach and the bi-orthogonal polynomial approach can
be established in that case as well.
5.3. Universality
After having computed the spectral density from the replica limit of the Toda lattice
equation and compared to that from the orthogonal polynomial approach, we would like to
use the insight gained to discuss the issue of universality. First of all, we can distinguish
between two different ways to approach universality. The first way is based on the observation that different theories with the same pattern of spontaneous symmetry breaking and
a mass gap have the same low-energy limit. For example, gauge field theories and random
matrix theories with the same global symmetries are described by the same effective partition function in the microscopic scaling limit. For QCD with three colors and fundamental
quarks at nonzero chemical potential this universal partition function is given by the group
integral (14). In QCD, a mass gap exists because of confinement, and in random matrix
theory, a mass gap appears in the large N limit. Therefore, if we change the action without
affecting the global symmetries and the existence of a mass gap, the microscopic scaling
limit will remain the same. A second way to understand universality is based on a direct
calculation of correlation functions for a deformed probability distribution which does not
necessarily have the unitary invariance of the Gaussian model. In the case of deformations
that respect the unitary invariance, universality then follows from the asymptotic behavior
of the modified orthogonal polynomials. In particular, for chiral random matrix theory at
= 0 this approach has been very successful [51,55].
Let us also comment on the role of the weight function in non-Hermitian random matrix
theories. From the replica limit of the Toda lattice equation it is clear that the eigenvalue
density will be proportional to the partition function with one pair of conjugate bosonic
flavors which, according to this work, is proportional to the weight function. In general,
this weight function contains both universal and nonuniversal parts, and in the microscopic
scaling limit only the universal parts survive. If we allow for deformations of the Gaussian
weight by higher order powers the resulting weight function, and therefore the spectral density, will be modified. It is only in the microscopic scaling limit that the nonuniversal parts
of the weight function disappear and a universal spectral density will be recovered. In Hermitian matrix models the weight function at finite N is also nonuniversal but reduces to a
universal result in the microscopic scaling limit. Non-Hermitian random matrix models for
QCD at
= 0 represent a continuous one parameter deformation of the Hermitian model
at = 0, governed by 2 N . In this case, the weight function contains an additional universal piece in the microscopic scaling limit that develops into a delta function (Im(z))
for 2 N 0. The additional universal piece occurs because an eigenvalue representation can only be obtained after a nontrivial integration over the similarity transformations
that diagonalize the non-Hermitian matrices. In the Hermitian case it is trivial to obtain
the eigenvalue representation, and no additional universal piece is generated. For technical
reasons we have not been able to integrate out the similarity transformations when higher
powers are added to the Gaussian probability distribution of the model (20). We mention
in passing that for nonchiral non-Hermitian random matrix models a universality proof by
deforming the random matrix potential could be given [64].
309
6. Results and explicit examples

6.1. The partition functions in the microscopic limit
Having obtained analytical expressions for the partition functions of the random matrix
model (20) at finite N we now consider the large N limit. This limit will be taken according
to the counting scheme where mf N and 2 N are kept fixed as N and is referred to as
the microscopic limit. The rescaling of the non-Hermiticity parameter 2 with N was first
introduced in [65] as the concept of weak non-Hermiticity. As discussed in the previous
section, in the microscopic limit the random matrix partition function is given by an integral
over the Goldstone manifold, i.e., it is uniquely determined by the spontaneous breaking
of the flavor symmetries. The random matrix model can be viewed as an alternative way to
calculate the integral over the Goldstone manifold, which in the case of nonzero chemical
potential and both bosonic and fermionic quarks has not been accomplished in any other
way.
The dependence of the quark mass and the chemical potential in the integral over the
Goldstone manifold is only through the combinations mV and 2 F2 V , cf. (14) and (17)
[14]. The dimensionfull scales can be recovered from the random matrix model by making
the replacements
2mN mV ,
22 N 2 F2 V .
(62)
To make contact with the physical content of the theory we will express our results in terms
of the dimensionfull scales in the remainder of this section.
In the case we have only either fermionic or bosonic quarks and no conjugate quarks,
the Goldstone bosons are not charged with respect to the chemical potential, and we should
find the microscopic partition functions at zero chemical potential. One easily observes that
the effective partition function (14) does not depend on the chemical potential in this case.
In the microscopic limit of (33) we thus find the partition function (14) with one fermionic
flavor [10,6668]
N =1
Z f
(m; ) = I (mV ).
(63)
2 2
e F V
has been removed. This unphysical factor can be reAn overall scaling factor
moved in the random matrix model by scaling the Gaussian potential in the weight (22) by
a factor of 1 2 [34]. We did not include an explicit scale factor in the weight function,
but will assume this procedure has been done in order to arrive at the correct expressions
in the microscopic limit.
When there is one pair of conjugate fermionic quarks, the microscopic partition function is given by the result obtained by performing the integral (14) over the Goldstone
manifold [26]
2 2
Zn=1 (y, z ; ) = e2 F V
dt te2
2F 2V t 2
I (yV t)I (z V t).
(64)
For the universal partition function with a pair of conjugate bosonic quarks we find using (43)
310
Zn=1 (z, z ; ) =
2 2

|z| V
(z2 + z2 ) 2 V
1
K
,
exp
2 F2 V
82 F2
42 F2
(65)
where we have divided out the divergent factor log()/2. This is again in agreement
with the result [26] obtained by an explicit integration over the Goldstone manifold (17).
All other microscopic partition functions required to calculate the spectral density can
be obtained from the relation (50)

N ,n=1
Z f
{mf }, z, z ; =
Nf

2
N

mf z2 Z f {mf }; Zn=1 (z, z ; ).
(66)
f =1
6.2. The microscopic spectral density

In this section we write out several different examples for the eigenvalue density of
the QCD Dirac operator in the microscopic limit. We emphasize that these results are
nonperturbative analytical predictions for the QCD Dirac spectrum at nonzero chemical
potential which should be reproduced by lattice QCD simulations. In order to get more
compact expressions we will express our results in terms of the partition function (64). The
normalization of the quenched density is chosen such that there are V / eigenvalues per
unit length along the imaginary axis. The density of the projection of the eigenvalues on the
imaginary axis is therefore equal to the eigenvalue density at = 0. The other densities
reduce to the quenched density when the quark masses are taken to infinity (mf V
2 F2 V ) and we use this limit to fix the normalization.
The general expression for the spectral density is given by

N
f z, z , {mf }; =
Nf
N ,n=1
n=1
Z f
({mf }, z, z ; )
zz 2
2
,
mf z Z
(z, z ; )
N
2
Z f ({mf }; )
f =1
(67)
where the microscopic limit of
microscopic limit of
ZNn=1 (z, z ; )
N
ZN f ({mf }; )
is given in the previous subsection. The
is given by
det[(xk xk )l I (xk )]k=1,...,Nf ,l=0,...,Nf 1

N
Z f {mf }; =
,
(xk2 )
(68)
where xk = mk V . Note that with only regular fermionic quarks, has no effect on the
N ,n=1
microscopic partition function. The partition function ZN f
({mf }, z, z ; ) is given in
terms of the polynomials pN (m) and the bare kernel KN +1 (x, y) in (47). The microscopic
limit of this partition function is obtained by expressing the polynomials and the kernel
in terms of partition functions that are known in the microscopic limit. The polynomials
pN (m) can be expressed as the partition functions with one fermionic flavor (see (33)),
which in the microscopic limit does not depend on and is given by I (mV ) in (63).
In (37) the kernel KN +1 (x, y) was shown to be equal to Zn=1 which is given in (64). In
the remaining sections we will set = 0 for simplicity.
311
6.2.1. The quenched spectral density

The quenched microscopic eigenvalue density of the Dirac operator at nonzero baryon
chemical potential is given by [26]
N =0
f
=0
(z, z ; ) =
|z|2 4 V 3 22 F2 V (z
e
e
22 F2
2 +z2 ) 2 V
82 F2

K0

|z|2 2 V
n=1
Z=0
(z, z ; ).
42 F2
(69)
This result has been checked by direct numerical simulations [26] within a random matrix
model, and successfully compared to results from lattice QCD [43]. Plots of the quenched
spectral density are given in Fig. 1.
Fig. 1. The quenched eigenvalue density with a chemical potential F V = 0.1 (upper) and F V = 2.5
(lower). The density
is real and positive and follows in the other quadrants by reflection on the axes. In the
upper figure F V 1, and one can still see the individual eigenvalue distributions. In the lower figure the
repulsion of eigenvalues from the origin is apparent. The width of the strip is on the order Re[z] 22 F2 . The
normalization is such that the width times the hight is independent of .
312
By using the leading order asymptotic forms of the Bessel functions, one easily shows
that in the limit 2 F2 V 1 the support of the quenched spectrum is inside the strip
[26,49]
| Re(z)|
<1
22 F2
(70)
with a density that is equal to

N =0
f
(z, z ; ) =
=0
V 2
.
42 F2
(71)
In Fig. 1 we observe that this plateau is already present for F V = 2.5. This step in the
spectral density is associated with a phase transition of the generating function. The theory
with a pair of conjugate fermionic flavors of mass x, described by Zn=1 (x, x; ), is in the
normal phase for x/F2 = m2 /2 > 22 and in a Bose condensed phase for x/F2 =
m2 /2 < 22 [14]. For the x-integrated quenched spectral density we find (z = x + iy)
N =0
f
dx =0
(z, z ; ) =
V
,
(72)
consistent with our choice of normalization.

6.2.2. The spectral density for one flavor
The microscopic eigenvalue density with one dynamical quark of mass m and with
baryon chemical potential is given by [34]
2 2 (z2 +z2 ) 2 V
|z| V 82 F 2
|z|2 4 V 3 22 F2 V
e
e
K0
2
2
2 F
42 F2

I0 (mV )
I0 (zV )
n=1
Z (m, z ; ) Z n=1 (z, z ; )
=0
=0
I0 (mV )

n=1 (m, z ; )
I0 (zV )Z=0
Nf =0
= =0 (z, z ; ) 1
(73)
.
n=1 (z, z ; )
I0 (mV )Z=0
This
result is shown in Fig. 2 for F V = 0.1 and mV = 10, and in Fig. 3 for
F V = 2.5 and mV = 5. It is our first example of a spectral density that is complex due to the sign problem.
Since the unquenched spectral density is proportional to the quenched spectral density
its support cannot go beyond the support of the quenched spectral density. In particular
for 2 F2 V 1 the unquenched spectral density is also confined inside the strip given by
(70). As we can observe from Fig. 4 the asymptotic behavior is much more complicated
than in the quenched case. An asymptotic form of the integral (64) was derived in [37]. For
2 F2 V 1 and (x + m)/(42 F2 ) < 1, it is well approximated by
Nf =1
=0
(z, z , m; ) =
313
Fig. 2. Real and

imaginary parts of the eigenvalue density for one flavor of mass mV = 10 and chemical
potential F V = 0.1. With these values the real part of the density is quite similar to the quenched density.
Note that the scale of the imaginary part is considerably smaller than that for the real part. The imaginary part is
odd in Re[z] as well as in Im[z].
dt te2
2F 2V t 2

I0 (x iy)V t I0 (mV t)
dt te2
2F 2V t 2

I0 (x iy)V t I0 (mV t)
0
2
((xiy) +m )
1
82 F2
e
=
42 F2 V
2V
I0

m(x iy) 2 V
.
42 F2
(74)
This approximation breaks down if (x + m)/(42 F2 ) > 1 because then the saddle point
of the t-integration will be outside the interval [0, 1]. The exponential functions in the
314
Fig. 3.
Real and imaginary parts of the eigenvalue density for one flavor of mass mV = 5 and chemical potential
F V = 2.5. With these values the real part of the density deviates substantially from the quenched density.
Note that the real part of the density changes sign at z = m. In this case the scale of the imaginary part is
comparable to that of the real part.
asymptotic behavior of the Bessel functions no longer compensate each other in the additional contribution to the unquenched spectra density. Instead of a plateau for Nf = 0
we find an oscillatory contribution with an amplitude that increases exponentially with
the volume. For the total spectral density a plateau is still visible in the region where the
quenched contribution dominates. As an illustration we show in Fig. 4 the real part of the
eigenvalue density for 2 F2 V = 100 and mV = 100. The period of the oscillations is of
the order of the level spacing at = 0 whereas the amplitude is of the order exp(10) for
our choice of parameters.
6.2.3. The spectral density for two flavors
For two flavors of mass m1 and m2 at nonzero baryon chemical potential the density of
eigenvalues is given by [34]
315
Fig.
4. The real part of the eigenvalue density for one flavor of mass mV = 100 and chemical potential
F V = 10. The severe sign problem manifest itself in the strongly oscillating region of the eigenvalue density
which has amplitudes of the order exp(10). The full size of the peaks has been clipped for better illustration.
N =2
f
=0
(z, z , m1 , m2 ; )
2 2 (z2 +z2 ) 2 V
|z| V 82 F 2
|z|2 4 V 3 22 F2 V
e
e
K0
=
2
2
2 F
42 F2

I0 (m1 V )
I0 (m2 V )
I0 (zV )

m1 V I1 (m1 V ) m2 V I1 (m2 V ) zV I1 (zV )
n=1
n=1 (m , z ; )
n=1 (z, z ; )
Z=0 (m1 , z ; )
Z=0
Z=0
2

.
I0 (m1 V ) m1 V I1 (m1 V )

I0 (m2 V ) m2 V I1 (m2 V )
(75)
In Fig. 5 we show the real and imaginary parts of the spectral density for F V = 2.5
and m1 V = m2 V = 5. The spectral density is neither real nor positive.
6.2.4. The spectral density with one pair of conjugate flavors
The microscopic spectral density of the Dirac operator in QCD with one pair of conjugate quarks (with mass m and m ) is found to be (see (B.4) and [38])
2 2 (z2 +z2 ) 2 V
|z| V 82 F 2
|z|2 4 V 3 22 F2 V
e
e
K
0
22 F2
42 F2
n=1

Z (m, m ; ) Z n=1 (m, z ; )
=0
=0

Z n=1 (z, m ; ) Z n=1 (z, z ; )
n=1
(z, z , m, m ; ) =
=0
=0
=0
n=1 (m, m ; )
Z=0

n=1 (z, m ; )Z n=1 (m, z ; )
Z=0
Nf =0
=0
= =0 (z; ) 1 n=1
.
n=1 (z, z ; )
Z=0 (m, m ; )Z=0
(76)
316
Fig. 5. Real and imaginary

parts of the spectral density with two flavors of mass m1 V = m2 V = 5 and
chemical potential F V = 2.5. The density is neither real nor positive and the fluctuations are an order of
magnitude larger than the quenched density at the same value of . The spectral density has a zero at z = m1 .
The imaginary part of the spectral density is anti-symmetric about the real and imaginary axis.
From this expression it follows that the density is real, positive, and has a zero at z = m.
This is illustrated by Fig. 6, where we plot the spectral density for mV = 5 and
F V = 2.5.
Since the quark mass, m, is real, the determinant of the conjugate quark is identical to
the determinant of a quark with a chemical potential of the same magnitude but with opposite sign, cf. (13). That is, the partition function in this case is equivalent to a theory with
two degenerate flavors and nonzero isospin chemical potential, I . This version of QCD
does not have a sign problem [69] and has been explored by lattice QCD simulations [46].
6.2.5. The density with one ordinary flavor and one pair of conjugate flavors
As a final example we write out the microscopic eigenvalue density in QCD with 3 light
flavors with real masses mu = md = m and ms and chemical potentials I = s = . This
317
Fig. 6. The spectral

density with one pair of conjugate flavors of mass m1 V = m2 V = 5 and chemical
potential F V = 2.5. It is real and positive with a zero at z = m1 . Since the masses are real and equal this
corresponds to a nonzero isospin chemical potential.
theory has a sign problem due to the extra quark that isnt matched by a conjugate quark.
The phase diagram as a function of I and s was determined in [70]. The eigenvalue
density is (B.5)
N =1,n=1
(z, z , ms , m, m; )
2 2 (z2 +z2 ) 2 V
|z| V
|z|2 4 V 3 22 F2 V
82 F2
e
K
=
e
0
2
2
2
2
2 F
4 F

I0 (mV )
I0 (ms V )
I0 (zV )
n=1
Z (m, m; ) Z n=1 (ms , m; ) Z n=1 (z, m; )
=0
=0
=0

Z n=1 (m, z ; ) Z n=1 (ms , z ; ) Z n=1 (z, z ; )
=0
=0
.
=0
I0 (mV )
I0 (ms V )
n=1
Z (m, m; ) Z n=1 (ms , m; )
f
=0
=0
(77)
=0
For z = m the two top rows are identical and the first and third column are identical in the
3 3 determinant. Hence the density is zero at z = m but does not change sign. While for
z = ms the first and second column in the 3 3 determinant are identical, so the density
changes sign at z = ms . A plot of this density is shown in Fig. 7.
7. Conclusions
We have analyzed the spectrum of the QCD Dirac operator at nonzero baryon chemical potential. In the microscopic limit this spectrum is uniquely determined by the global
symmetries of the QCD partition function. This is true both for the quenched and the
unquenched theory. In both cases the spectral density of the Dirac operator can only be
obtained after the introduction of a complex conjugate pair of valence quarks resulting in
a nontrivial baryon charge matrix and Goldstone modes with nonzero baryon number. The
318
Fig. 7. The real part of the density in a theory with 3 quarks. Themasses are mu V = md V = 5 and
ms V = 10. The chemical potentials are s = I = = 2.5/(F V ). Note that the density bounces off
the real axis at z = mu and changes sign at z = ms as expected.
microscopic limit of the generating function for the QCD Dirac spectrum can therefore be
represented in two different ways. First, as an integral over the Goldstone manifold which
is determined by the pattern of symmetry breaking. Second, as the large-N limit of a partition function of an ensemble of non-Hermitian matrices with the symmetries of the QCD
partition function. In recent work, the Dirac spectrum was derived directly from the random matrix model by means of complex orthogonal polynomials. In this paper we have
obtained the Dirac spectrum from the replica limit of the Toda lattice equation. This explains that the spectral density factorizes into the product of a fermionic partition function
and the partition function for one conjugate pair of bosonic valence quarks. The bosonic
factor does not depend on the dynamical quarks and therefore goes beyond what is implied
by the Toda lattice structure. This arises as a consequence of a singularity in the partition
functions that contain a pair of conjugate bosonic quarks with the same mass. This singularity persists even for finite size matrices. In fact, it is present for an ensemble of 2 2
random matrices. Also the Toda lattice structure is valid for finite size matrices, and one can
easily verify that the correct spectral density is obtained for an ensemble of 2 2 matrices.
As a new result, we have also obtained explicit expressions for the spectral density when
the absolute value of the fermion determinant is present in the QCD partition function. In
this case a direct check from lattice simulations is possible for all values of the scaled
chemical potential, as has already been done in the fully quenched case.
The calculation of the unquenched microscopic spectral density has been a highly nontrivial test of the Toda lattice approach. We expect that it will be equally successful for
all remaining = 2 cases. For example, the method could be applied to the calculation
of GUE S-matrix fluctuations and parametric correlations [71] as well as to other nonHermitian random matrix ensembles [72]. The extension of this approach to Dyson index
= 1 and = 4 is still an open problem.
319
The spectral density can also be obtained from the replica limit of a Painlev equation that has been derived from fermionic partition functions only. On the other hand, if a
family of partition functions satisfies a Painlev equation, a Toda lattice equation can be
derived from the Bcklund transformations of the Painlev system. Therefore, all systems
in the same universality class have the same microscopic limit for the fermionic partition
functions, the bosonic partition functions and the microscopic spectral density.
Another important difference between Hermitian and non-Hermitian Gaussian random
matrix theories is that the probability density of non-Hermitian theories do not factorize into the joint eigenvalue density and the distribution of the eigenvectors. The joint
eigenvalue density is only obtained after a nontrivial integration over the similarity transformations that diagonalize the non-Hermitian matrices. This results in the appearance of
a nontrivial universal factor in the joint eigenvalue distribution, in our case a modified
K-Bessel function. The appearance of this function is a direct consequence of the chiral
symmetry of the problem.
The microscopic spectral density derived in this paper is valid for small temperatures
and all values of the dimensionless parameters 2 F2 V and mV . Within this range of parameters, where the sign problem develops and becomes severe, the effect on the spectral
density can be followed analytically. A first manifestation of the sign problem is that the
spectral density in the complex eigenvalue plane is no longer real. This occurs already in a
parameter range where lattice simulations might be feasible. For larger values of 2 F2 V
the analytical expressions for the eigenvalue density can serve as tests for numerical methods that apply to the nonperturbative regime of QCD at nonzero baryon chemical potential.
In the limit where 2 F2 V 1 the spectral density shows oscillations with a period of
the order of the level spacing and an amplitude that diverges exponentially with 2 F2 V .
This behavior, which is not present in the quenched case, should be responsible for the
breaking of chiral symmetry. This issue will be addressed in a future publication.
Acknowledgements
We wish to thank the INT in Seattle where this work was conceived. Dominique Toublan
is thanked for collaboration on the graded integrals and Poul Henrik Damgaard for discussions on orthogonal polynomials. G.A. was supported in part by a Heisenberg Fellowship,
J.C.O. was supported by US NSF grant PHY 01-39929 and US DOE grant DE-FC0201ER41180, and J.J.M.V. was supported in part by US DOE grant DE-FG-88ER40388.
Appendix A. Some details on two flavor and general partition functions

For completeness we collect here the formulas for the remaining two partition functions
with two flavors. As was observed in [38,40], partition functions that do not mix quarks or
conjugate quarks in either the numerator or denominator can be written solely as determi-
320
nants of polynomials and Cauchy transforms. We find

N

N =1, Nb =1
2
2
(x|y ; )
ZN f
x (zj x )
N =0
y
(z2 y 2 ) N =0
ZN f
j =1 j
f

hN 1 (y ) hN (y )
x
1
,

=
rN 1 y pN 1 (x) pN (x)
and
N =2
ZN b
(y1 , y2 ; )
N =0
ZN f
(A.1)

N

1
1
(y1 y2 )
(z2 y12 )(zj2 y22 ) N =0
j =1 j
f

hN 2 (y1 ) hN 1 (y1 )
1
.

=
rN 1 rN 2 (y12 y22 )(y1 y2 ) hN 2 (y2 ) hN 1 (y2 )
(A.2)
All remaining combinations of bosons and fermions and their conjugates can be obtained
from the given ones by complex conjugation.
Next we wish to comment on the relation between these formulas and in particular on
the additional pole terms which occur in the two kernels containing Cauchy transforms,
Eqs. (40) and (45). For random matrix models with real eigenvalues the kernel of the
polynomials (36) satisfies the ChristoffelDarboux identity,
KN (s, t) =
pN (s)pN 1 (t) pN (t)pN 1 (s)

,
rN 1
s2 t 2
1
s, t R.
(A.3)
For orthogonal polynomials in the complex plane (28) this relation is generally not satisfied
so that (35) and (38) are different. In the Hermitian limit 0 they are equal, due to the
relation (A.3).
The Eqs. (39) and (A.2) or (44) and (A.1) can be related if the following Christoffel
Darboux identities are valid:
1 hN 1 (s)hN 2 (t) hN 1 (t)hN 2 (s)
, s, t R,
AN 2 (s, t) =
rN 2
s2 t 2
1 hN (s)pN 1 (t) pN (t)hN 1 (s)
NN 1 (s, t) =
(A.4)
, s, t R.
rN 1
s2 t 2
This is the case for Hermitian random matrix models, but in general there are no such
ChristoffelDarboux identities for non-Hermitian random matrix theories. This identity
follows from the observation that in general the orthogonal polynomials and their Cauchy
transforms obey the same three-step recursion relation. Only the recursion involving the
lowest Cauchy transform hk=0 (y) is different from that involving the polynomial pk=0 (y),
which results in the extra integral Q(s, t) over poles in (40), or the single pole 1/(s 2 t 2 )
in (45), respectively. Thus in the Hermitian limit (39) and (A.2), and (44) and (A.1) become
equal due to their ChristoffelDarboux identities (A.4). This explains the presence of pole
terms in (39) and (44). In (45) this term can also be understood differently, as it ensures the
correct normalization of the partition functions (44) to unity at equal arguments.
Let us now turn to more general partition functions. We will briefly explain here how
our generating function (48) follows from the results of [39] (generalized to the chiral
321
ensemble). There the following expectation value of complex eigenvalues is computed as

N

Nf
2
2

f =1 (mf zj )
j =1
(zj2 y 2 )(zj2 x 2 )
(1)Nf
rN 1
Nf =0

pN 1 (m1 )

..

.

pN +N 2 (m1 )
f

NN +Nf 2 (m1 , y)

p0 (m1 )

..

.

p

Nf 2 (m1 )

0

NNf 2 (m1 , y)
...
..
.
...
...
...
..
.
...
...
...

hN 1 (x)

..

.

hN +Nf 2 (x)
AN +Nf 2 (x , y)
pN 1 (mNf )
..
.
pN +Nf 2 (mNf )
NN +Nf 2 (mNf , y)
p0 (mNf )
..
.
pNf 2 (mNf )
0
NNf 2 (mNf , y)

h0 (x)

..

.

hNf 2 (x)

p0 (x)

ANf 2 (x , y)
(A.5)
In order to simplify the denominator the determinant can be expanded with respect to the
next to last row. The remaining determinant only contains polynomials pk (mi ) and the
kernel NNf 2 (mi , y). The invariance of determinants under adding and subtracting rows
and columns can be used to first eliminate all sums in the kernels NNf 2 (mi , y). Then
all polynomials can be reduced to monomials, leading to the following Vandermonde like
determinant

m2N
m21

f
2 2

.
.
.
2 y 2

m
y
m

1
Nf

1
...
1

m4N
m2
m41
2

f
.
.
.
m

Nf
1
.
.
.

2
2
2
2

m1 y
mN y
f

..
.
.

.
.

.
.
.
..
..

..

=

.
.
.
2(Nf 2)
2(Nf 2)

m

2(N
1)
.
.
.
m
f

2(N
1)
Nf
1
m f
mN

f

1
1
1

.
.
.
2 y 2
2 2
m2 y 2
...
2
m
2

m1 y
mN y
Nf
1

f
1
1

...
2
m2 y 2
2
m y
1
Nf
Nf ({m2f })
= (1)Nf 1 N
.
f
2
2
f =1 (mf y )
(A.6)
Here we have again made use of invariance properties of the determinant before arriving
at the desired result. Substituting this in (A.5) we obtain (48).
Appendix B. The microscopic eigenvalue density with conjugate quarks
In this appendix we derive the microscopic eigenvalue density of the Dirac operator
in a theory with conjugate fermionic quarks. Even if some or all of the fermionic flavors
are conjugate quarks the Toda lattice equation holds. In the microscopic limit the partition
322
functions
N ,l,n

{mf }, {mc }, mc , z, z ;
Z f
[dA]
Nf

l

det D() + mc 2 det D() + z 2n
det D() + mf
f =1
c=1
(B.1)
eSYM (A) ,
satisfy the Toda lattice equation
N ,l,n
z z log Z f

{mf }, {mc }, mc , z, z ;
N ,l,n+1
n 2 Z f
(zz )
2
Nf ,l,n1
({mf },{mc },{mc },z,z ;)Z
({mf },{mc },{mc },z,z ;)

.
Nf ,l,n
[Z
({mf },{mc },{mc },z,z ;)]2
(B.2)
This follows by a direct extension of the argument given in Section 2.3.2 of [26]. In order
to obtain the density with l pairs of conjugate quarks we consider the replica limit of the
Toda lattice equation (B.2) with Nf = 0

l z, z , {mc }, mc ;
=
zz Zl,n=1 ({mc }, {mc }, z, z ; )Zl,n=1 ({mc }, {mc }|z, z ; )

.
2
[Zl ({mc }, {mc }; )]2
(B.3)
Using (66) we find the spectral density for a theory with l pairs of conjugate quarks

l z, z , {mc }, mc ;
=
l
2
zz 2
Z l,n=1 ({mc }, {mc }, z, z ; )
z m2c Zn=1 (z, z ; )
.
2
Zl ({mc }, {mc }; )
(B.4)
c=1
This density is positive and real for all masses and chemical potentials. This is fully consistent with having a real and positive measure in (4). We have explicitly checked that
this eigenvalue density is consistent with what we obtain using the orthogonal polynomial
method as described in [38].
Finally, we give the most general result we have obtained using this method, namely the
spectral density for a theory with Nf + l quarks and l conjugate quarks,
N ,l

z, z , {mf }, {mc }, mc ;
Nf
l
2

zz 2
z m2 2 Z n=1 (z, z ; )
z m2f
=
c
2
f =1
c=1
N ,l,n=1
Z f
({mf }, {mc }, {mc }, z, z ; )
.
N ,l
Z f ({mf }, {mc }, {mc }; )
(B.5)
323
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
I. Barbour, et al., Nucl. Phys. B 275 (1986) 296.

P. de Forcrand, O. Philipsen, Nucl. Phys. B 642 (2002) 290, hep-lat/0205016.
M. DElia, M.P. Lombardo, Phys. Rev. D 67 (2003) 014505, hep-lat/0209146.
C.R. Allton, et al., Phys. Rev. D 66 (2002) 074507, hep-lat/0204010.
F. Karsch, Prog. Theor. Phys. Suppl. 153 (2004) 106, hep-lat/0401031.
A. Gocksch, Phys. Rev. D 37 (1988) 1014.
M.A. Stephanov, Phys. Rev. Lett. 76 (1996) 4472, hep-lat/9604003.
E.V. Shuryak, J.J.M. Verbaarschot, Nucl. Phys. A 560 (1993) 306, hep-th/9212088;
J.J.M. Verbaarschot, Phys. Rev. Lett. 72 (1994) 2531, hep-th/9401059.
J.J.M. Verbaarschot, Phys. Lett. B 368 (1996) 137, hep-ph/9509369.
J. Gasser, H. Leutwyler, Ann. Phys. 158 (1984) 142;
J. Gasser, H. Leutwyler, Nucl. Phys. B 250 (1985) 465;
H. Leutwyler, Ann. Phys. 235 (1994) 165.
S. Weinberg, Phys. Rev. 166 (1968) 1568.
J.B. Kogut, M.A. Stephanov, D. Toublan, Phys. Lett. B 464 (1999) 183, hep-ph/9906346.
J.B. Kogut, M.A. Stephanov, D. Toublan, J.J.M. Verbaarschot, A. Zhitnitsky, Nucl. Phys. B 582 (2000) 477,
hep-ph/0001171.
D. Toublan, J.J.M. Verbaarschot, Int. J. Mod. Phys. B 15 (2001) 1404, hep-th/0001110.
D.T. Son, M.A. Stephanov, Phys. Rev. Lett. 86 (2001) 592, hep-ph/0005225;
K. Splittorff, D.T. Son, M.A. Stephanov, Phys. Rev. D 64 (2001) 016003, hep-ph/0012274;
J.B. Kogut, D. Toublan, Phys. Rev. D 64 (2001) 034007, hep-ph/0103271;
K. Splittorff, D. Toublan, J.J.M. Verbaarschot, Nucl. Phys. B 620 (2002) 290, hep-ph/0108040;
K. Splittorff, D. Toublan, J.J.M. Verbaarschot, Nucl. Phys. B 639 (2002) 524, hep-ph/0204076;
J. Wirstam, J.T. Lenaghan, K. Splittorff, Phys. Rev. D 67 (2003) 034021, hep-ph/0210447;
J.T. Lenaghan, F. Sannino, K. Splittorff, Phys. Rev. D 65 (2002) 054002, hep-ph/0107099.
H. Leutwyler, A. Smilga, Phys. Rev. D 46 (1992) 5607.
J. Gasser, H. Leutwyler, Phys. Lett. B 188 (1987) 477.
J.C. Osborn, D. Toublan, J.J.M. Verbaarschot, Nucl. Phys. B 540 (1999) 317, hep-th/9806110.
P.H. Damgaard, J.C. Osborn, D. Toublan, J.J.M. Verbaarschot, Nucl. Phys. B 547 (1999) 305, hepth/9811212.
G. Akemann, P.H. Damgaard, Phys. Lett. B 583 (2004) 199, hep-th/0311171.
S.F. Edwards, P.W. Anderson, J. Phys. F 5 (1975) 965.
S.F. Edwards, R.C. Jones, J. Phys. A 9 (1976) 1595;
J.J.M. Verbaarschot, M.R. Zirnbauer, Ann. Phys. 158 (1984) 78;
J.J.M. Verbaarschot, M.R. Zirnbauer, J. Phys. A 18 (1985) 1093;
A. Kamenev, M. Mezard, J. Phys. A 32 (1999) 4373, cond-mat/9901110;
A. Kamenev, M. Mezard, Phys. Rev. B 60 (1999) 3944, cond-mat/9903001;
I.V. Yurkevich, I.V. Lerner, Phys. Rev. B 60 (1999) 3955, cond-mat/9903025;
M.R. Zirnbauer, cond-mat/9903338.
E. Kanzieper, Phys. Rev. Lett. 89 (2002) 250201, cond-mat/0207745.
E. Kanzieper, Progress in Field Theory Research, Nova Science, New York, 2004, in press, condmat/0312006.
K. Splittorff, J.J.M. Verbaarschot, Phys. Rev. Lett. 90 (2003) 041601, cond-mat/0209594.
K. Splittorff, J.J.M. Verbaarschot, Nucl. Phys. B 683 (2004) 467, hep-th/0310271.
V.E. Korepin, N.M. Bogoliubov, A.G. Izergin, Quantum Inverse Scattering Method and Correlation Functions, Cambridge Univ. Press, Cambridge, 1993.
P. Forrester, Log-gases and Random Matrices, Web Book, available at http://www.ms.unimelb.edu.au/
~matpjf/matpjf.html.
K. Splittorff, J.J.M. Verbaarschot, Nucl. Phys. B 695 (2004) 84, hep-th/0402177.
K.B. Efetov, Phys. Rev. Lett. 79 (1997) 491;
K.B. Efetov, Adv. Phys. 32 (1983) 53;
K.B. Efetov, Supersymmetry in Disorder and Chaos, Cambridge Univ. Press, Cambridge, 1997.
324
[31] S. Kharchev, A. Marshakov, A. Mironov, A. Morozov, A. Zabrodin, Nucl. Phys. B 380 (1992) 181, hepth/9201013.
[32] G. Akemann, P.H. Damgaard, Nucl. Phys. B 576 (2000) 597, hep-th/9910190;
H.W. Braden, A. Mironov, A. Morozov, Phys. Lett. B 514 (2001) 293, hep-th/0105169.
[33] G. Akemann, Y.V. Fyodorov, G. Vernizzi, Nucl. Phys. B 694 (2004) 59, hep-th/0404063.
[34] J.C. Osborn, Phys. Rev. Lett. 93 (2004) 222001, hep-th/0403131.
[35] P. Di Francesco, M. Gaudin, C. Itzykson, F. Lesage, Int. J. Mod. Phys. A 9 (1994) 4257, hep-th/9401163.
[36] Y.V. Fyodorov, B.A. Khoruzhenko, H.-J. Sommers, Phys. Rev. Lett. 79 (1997) 557, cond-mat/9703152;
Y.V. Fyodorov, B.A. Khoruzhenko, H.-J. Sommers, Ann. Inst. H. Poincar 68 (1998) 449, chaodyn/9802025.
[37] G. Akemann, Phys. Rev. Lett. 89 (2002) 072002, hep-th/0204068;
G. Akemann, J. Phys. A 36 (2003) 3363, hep-th/0204246.
[38] G. Akemann, G. Vernizzi, Nucl. Phys. B 660 (2003) 532, hep-th/0212051.
[39] M.C. Bergere, hep-th/0404126.
[40] G. Akemann, A. Pottier, J. Phys. A 37 (2004) L453, math-ph/0404068.
[41] H. Markum, R. Pullirsch, T. Wettig, Phys. Rev. Lett. 83 (1999) 484, hep-lat/9906020.
[42] E. Bittner, S. Hands, H. Markum, R. Pullirsch, Prog. Theor. Phys. Suppl. 153 (2004) 295, hep-lat/0402015.
[43] G. Akemann, T. Wettig, Phys. Rev. Lett. 92 (2004) 102002, hep-lat/0308003.
[44] E. Bittner, M.P. Lombardo, H. Markum, R. Pullirsch, Nucl. Phys. B (Proc. Suppl.) 94 (2001) 445, heplat/0010018;
E. Bittner, M.P. Lombardo, H. Markum, R. Pullirsch, Nucl. Phys. B (Proc. Suppl.) 106 (2002) 468, heplat/0110048.
[45] G. Akemann, E. Bittner, M.P. Lombardo, H. Markum, R. Pullirsch, hep-lat/0409045.
[46] J.B. Kogut, D.K. Sinclair, Phys. Rev. D 66 (2002) 034505, hep-lat/0202028.
[47] A. Nakamura, T. Takaishi, Nucl. Phys. B (Proc. Suppl.) 129130 (2004) 629, hep-lat/0310052.
[48] V.L. Girko, Theory of Random Determinants, Kluwer Academic, Dordrecht, 1990.
[49] K. Splittorff, J.J.M. Verbaarschot, hep-th/0408107.
[50] A.D. Jackson, M.K. Sener, J.J.M. Verbaarschot, Nucl. Phys. B 479 (1996) 707, hep-ph/9602225.
[51] G. Akemann, P. Damgaard, U. Magnea, S. Nishigaki, Nucl. Phys. B 487 (1997) 721, hep-th/9609174.
[52] T. Guhr, T. Wettig, Nucl. Phys. B 506 (1997) 589, hep-th/9704055.
[53] A.D. Jackson, M.K. Sener, J.J.M. Verbaarschot, Nucl. Phys. B 506 (1997) 612, hep-th/9704056.
[54] P.H. Damgaard, S.M. Nishigaki, Nucl. Phys. B 518 (1998) 495, hep-th/9711023.
[55] E. Kanzieper, V. Freilikher, in: NATO Adv. Sci. Inst. Ser. C Math. Phys. Sci., vol. 531, Kluwer, Dordrecht,
1999, pp. 165211, cond-mat/9809365.
[56] D. Dalmazi, J.J.M. Verbaarschot, Nucl. Phys. B 592 (2001) 419, hep-th/0005229.
[57] G. Akemann, Y.V. Fyodorov, Nucl. Phys. B 664 (2003) 457, hep-th/0304095.
[58] M.A. Halasz, A.D. Jackson, J.J.M. Verbaarschot, Phys. Lett. B 395 (1997) 293, hep-lat/9611008.
[59] G. Akemann, P.H. Damgaard, Phys. Lett. B 432 (1998) 390, hep-th/9802174.
[60] Y.V. Fyodorov, E. Strahov, Nucl. Phys. B 647 (2002) 581, hep-th/0205215;
Y.V. Fyodorov, G. Akemann, JETP Lett. 77 (2003) 438, cond-mat/0210647.
[61] P. Zinn-Justin, Commun. Math. Phys. 194 (1998) 631, cond-mat/9705044.
[62] M.L. Mehta, Random Matrices, second ed., Academic Press, London, 1991.
[63] M.C. Bergere, hep-th/0311227.
[64] G. Akemann, Phys. Lett. B 547 (2002) 100, hep-th/0206086.
[65] Y.V. Fyodorov, B.A. Khoruzhenko, H.-J. Sommers, Phys. Lett. A 226 (1997) 46, cond-mat/9606173;
Y.V. Fyodorov, H.-J. Sommers, J. Math. Phys. 38 (1997) 1918, cond-mat/9701037.
[66] J.B. Kogut, M. Snow, M. Stone, Nucl. Phys. B 200 (1982) 211.
[67] R.C. Brower, P. Rossi, C.-I. Tan, Nucl. Phys. B 190 (1981) 699;
R.C. Brower, M. Nauenberg, Nucl. Phys. B 180 (1981) 221.
[68] A.D. Jackson, M.K. Sener, J.J.M. Verbaarschot, Phys. Lett. B 387 (1996) 355, hep-th/9605183.
[69] M. Alford, A. Kapustin, F. Wilczek, Phys. Rev. D 59 (1999) 054502, hep-lat/9807039.
[70] J.B. Kogut, D. Toublan, Phys. Rev. D 64 (2001) 034007, hep-ph/0103271.
[71] J.J.M. Verbaarschot, unpublished.
[72] D. Bernard, A. LeClair, cond-mat/0110649.
Complete CKM quark mixing via dimensional

deconstruction
P.Q. Hung a , A. Soddu b , Ngoc-Khanh Tran c
a Institute for Nuclear and Particle Physics, and Department of Physics, University of Virginia,
Charlottesville, VA 22904-4714, USA

b Department of Physics, National Taiwan University, Taipei 106, Taiwan
c Department of Physics, University of Virginia, Charlottesville, VA 22904-4714, USA
Received 18 October 2004; accepted 19 January 2005
Abstract
It is shown that the deconstruction of [SU(2)U (1)]N into [SU(2)U (1)] is capable of providing
all necessary ingredients to completely implement the complex CKM mixing of quark flavors. The
hierarchical structure of quark masses originates from the difference in the deconstructed chiral zeromode distributions in theory space, while the CP-violating phase comes from the genuinely complex
vacuum expectation value of link fields. The mixing is constructed in a specific model to satisfy
experimental bounds on quarks masses and CP violation.
PACS: 11.25.Mj; 12.15.Ff
1. Introduction
Dimensional deconstruction [1,2] is a very interesting approach to dynamically generate
the effects of extra dimensions departing from the four-dimensional (4D) renormalizable
physics at ultraviolet scale. That is, apart from having the viability in the sense of renormalizability, whatever amusing mechanisms being dynamically raised by the virtue of extra
E-mail addresses: pqh@virginia.edu (P.Q. Hung), asoddu@hep1.phys.ntu.edu.tw (A. Soddu),
nt6b@virginia.edu (N.-K. Tran).
doi:10.1016/j.nuclphysb.2005.01.027
326
P.Q. Hung et al. / Nuclear Physics B 712 (2005) 325346
dimensions (ED) now can also be easily arranged to rise dynamically in a pure 4D framework. In this paper we look specifically into two such important mechanisms of extra
dimension theories, namely the localization of matter fields in the bulk [35] and the dynamical breaking of CP symmetry by ED Wilson line [69]. Ultimately, the hybrid of
these two mechanisms is just the well-known complex mixing of fermion flavors. And it is
conceptually interesting to note that dimensional deconstruction (DD) nicely encompasses
both of these issues. In other words, complete CabibboKobayashiMaskawa (CKM) mixing can be generated naturally via dimensional deconstruction.
With the presence of extra dimensions, one has a new room to localize the matter fields
differently along the transverse directions as in the so-called split fermion scenario. Various overlaps of fermions of different flavors then induce various fermion masses observed
in nature (see, e.g., [1012]). Amazingly, the deconstruction interaction is also able to produce similar localization effects [13]. Indeed, after the spontaneous breaking of link fields,
fermions get an extra contribution to their masses via the Higgs mechanism. Fermions then
reorganize themselves into mass sequences and the lightest mass eigenstate of these towers exposes some interesting localization pattern in the theory space (also referred to as
deconstruction group index space). We will first work out the analytical expressions and
confirm the localization of these zero modes in a rather generic deconstruction set-up. The
next question to raise is how to make these light modes chiral. Imposing some kind of
chiral boundary conditions [2] is the answer again coming from the ED lessons. There is
however one more subtle point to be mentioned here. If one truly wishes to relate the ED
scenario to the dimensional deconstruction, one needs to latticize the extra dimensions to
host the deconstruction group. There comes the lattice theorys issue of fermion doubling,
and its standard remedy, such as adding to the Lagrangian a Wilson term [14] would remove half of original chiral degrees of freedom. This is the reason why most of previous
works addressing the fermionic mixing in deconstructed picture (e.g., [2,13,15]) usually
start out with only Weyl spinors. In the current work, we adopt a different and somewhat
more general 4D deconstruction approach [16] where no extra dimension is actually invoked. As a result the fermions to begin with keep a standard 4-component Dirac spinor
representation.
In any deconstruction set-up, the link fields transform non-trivially under at least two
different gauge groups. This implies a complex vacuum expectation value (VEV) for these
fields, whose phase would not be rotated away in general. After the deconstruction process,
this phase is carried over into the complex value of wave functions and wave function
overlaps of fermions. In turn, the induced complex-valued mass matrices can render a
required CP-violating phase in the well-known KM mechanism. In contrast, we note that
the generation of complex mass matrices within the split fermion scenario is a non-trivial
problem and requires rather sophisticated techniques to solve [17,18]. Interestingly, the
above CP violation induction via deconstruction can also be visualized in extra dimensional
view point. Indeed, because of having the same symmetry transformation property, DD link
field can be identified with the Wilson line pointing along a latticized transverse direction
(Appendix B), and the latter then can naturally acquire a complex VEV in the generalized
Hosotanis mechanism [69] of dynamical symmetry breaking. Apparently, the source of
CP violation in this approach comes from the complex effective Yukawa couplings so it can
be classified as hard CP violation. Nevertheless, those couplings acquire complex values
327
after the spontaneous breaking of the DD link fields. In that sense this CP violation pattern
could also be considered soft and dynamical.
This paper is presented in the following order. In Section 2.1 we give the zero mass
eigenstate of fermions obtained in the deconstructed picture, in Section 2.2 the resulting expression of mass matrix elements, and in Section 2.3 the symmetry breaking of
[SU(2) U (1)]N into [SU(2) U (1)]. In Section 3 we present the numerical fit for quark
mass spectrum and CKM matrix in a model where each standard model Higgs field is
chosen to transform under only a single deconstruction subgroup. The conclusion and comments on numerical results is given in Section 4. Appendix A provides a detailed derivation
of zero mode wave functions in 4D deconstruction using combinatoric techniques. Appendix B outlines intuitive arguments on the complexity of link field inspired by lattice models.
Appendix C presents analytical expressions for wave function overlaps used in the determination of mass matrix elements. Finally, Appendix D gives referencing values of key
physical quantities that have been used in the search algorithm (Table 1), and numerical
solution of our models parameters (Table 2).
2. Deconstruction and quark mass matrix

In this section we describe how the mixing of quark flavors arises in the DD picture.
But we first briefly recall the basic idea of the dimensional deconstruction applied to just a
single quark generation. The family replication will be restored in the later sections.
2.1. Zero-mode fermion
We begin with N copies of gauge group [SU(2) U (1)]n where n = 1, . . . , N .
To each group [SU(2) U (1)]n we associate a SU(2)n -doublet Qn , and two SU(2)n singlets Un , Dn . These fields transform non-trivially only under their corresponding group
[SU(2) U (1)]n as (2, qQ ), (1, qU ), (1, qD ) respectively, with qs denoting U (1)-charges.
Q
U
D
Finally, we use 3(N 1) scalars n1,n , n1,n
, n1,n
transforming respectively as
(2, qQ |2, qQ ), (1, qU |1, qU ), (1, qD |1, qD ) under [SU(2) U (1)]n1 [SU(2)
U (1)]n to link fermions of the same type. Because of this, scalars s are also referred
to as link fields hereafter. For the simplicity of the model, we assume a symmetry for the
Lagrangian
under the permutation of group index n.
The N
n=1 [SU(2) U (1)]n gauge-invariant Lagrangian of the fermionic sector is

N
N
N
1

Q
n i/
n Qn
Dn Qn +
Qn+1 MQ
Q
Q n
Q
L=
n,n+1
n=1
n=1
n=1
+ (Q U ) + (Q D),
(1)
where D
/ n denotes the covariant derivative associated with gauge group [SU(2) U (1)]n ,
and MQ , MU , MD are the bare masses of fermions. Ultimately, we are interested in achieving chiral fermions of standard model (SM) at low energy scale. To this aim we impose the
following chiral boundary conditions (CBC) on fermion fields [2]
Q1R = QN R = 0,
N 1,N QN,L = VQ QN 1,L ,
328
U1L = UN L = 0,
U
N
1,N UN,R = VU UN 1,R ,
D1L = DN L = 0,
D
N
1,N DN,R = VD DN 1,R .
(2)
We note that those conditions are in agreement with the gauge transformation property of
Q
fields, e.g., N 1,N QN,L and QN 1,L transform identically under the underlying gauge
groups. Essentially, these boundary conditions render one more left-handed degree of freedom over the right-handed for Q field, and the contrary holds for U and D fields. The
actual calculation will show that the zero-mode of Q field indeed is left-handed while for
U, D it is right-handed. When the link fields Q,U,D assume VEV proportional to VQ,U,D ,
above CBC become the very reminiscence of Neumann and Dirichlet boundary conditions.
In the deconstruction scenario, after the spontaneous symmetry breaking (SSB) the link
fields acquire an uniform VEV VQ,U,D respectively, independent of site index n (in accordance with the assumed permutation symmetry), and the fermions obtain new mass
structure. Using the CBC (2), the fermion mass term can be written in the chiral basis as
Q1L
Q2R
..
..
N 1,R )[MQ ]
N 1,L )[MQ ]
1L , . . . , Q
+ (Q
(Q 2R , . . . , Q
.
.
QN 1,L
QN 1,R
+ (QR,L UL,R ) + (QR,L DL,R ),
(3)
where the matrix [MQ ] of dimension (N 2) (N 1) is
VQ
0
0
[MQ ](N 2)(N 1) =
MQ
VQ
0
VQ
MQ
VQ
0
VQ
MQ
..
.
MQ
VQ
VQ
MQ VQ
(4)
By interchanging QR,L UL,R , QR,L DL,R , the matrices [MU ], [MD ] of dimension
(N 1) (N 2) can be analogously found.
By coupling the following Dirac equations for chiral fermion sets {QR } (Q2R , . . . ,
QN 1,R )T and {QL } (Q1L , . . . , QN 1,L )T
i/
{QR } [MQ ]{QL } = 0,
i/
{QL } [MQ ] {QR } = 0,
(5)
we see that [MQ

MQ ] is the squared-mass matrix for the left-handed components QL and
[MQ MQ
] for the right-handed QR . Since at low energy, we are interested only in the chiral
MQ ], [MU MU ], [MD MD
] in what
zero modes of fermions, we will work only with [MQ
follows
|VQ |2
MQ VQ
2
(VQ )
0

MQ MQ =
MQ VQ
2+
MQ
|VQ |2
2MQ VQ
)2
(VQ
2
VQ
2MQ VQ
2
VQ
2+
MQ
2|VQ |2
2MQ VQ
329
2MQ VQ
2+
MQ
2|VQ |2
..
2MQ VQ
2+
MQ
2|VQ |2
)2
(VQ
2MQ VQ
)2
(VQ
2MQ VQ
2+
MQ
2|VQ |2
+
2MQ VQ
2
(VQ )
2
VQ
2MQ VQ +
|VQ |2
2
2
M +2|V |
Q
Q
)M
(VQ +VQ
Q
(6)
[MU MU ],
[MD MD
].
and similar expressions hold for

The quantitative derivation of the
zero-eigenstates, which are identified with the SM chiral fermion, is presented in Appendix A. In this section we just concentrate on some qualitative discussion. In general the
diagonalization of matrices (6) leads to the transformation between gauge eigenstates QnL
and mass eigenstates Q mL
mL ,
QnL = [UQ ]nm Q
nL = [UQ ]mn QmL ,

Q
MQ ]
where the matrix [UQ ] diagonalizes [MQ

MQ [UQ ].
MQ MQ diag = [UQ ] MQ
(7)
(8)
The key observation, which will be analyzed in more details in Appendix B, is that VEV
VQ,U,D are generically complex and [UQ,U,D ] are truly unitary (i.e., not just orthogonal).
0L , U 0L , D 0L in Eq. (7) and
This in turn gives non-trivial phases to zero-mode fermions Q
after the SM spontaneous symmetry breaking the obtained mass matrices are complex.
0L (and U 0R , D 0R ) in the mass eigenbasis
Further, the explicit solution of zero mode Q
exhibits a very interesting localization pattern in the group index space n (see Appendix A). This in turn can serve to generate the mass hierarchy among fermion families in
a manner similar to that of ED split fermion scenario (see, e.g., [12,18]). Thus we see
that dimensional deconstruction indeed provides all necessary ingredients to construct a
complete (complex) CKM structure of fermion family mixing.
2.2. Complex mass matrix
In order to give mass to the above chiral zero-mode of fermions, we introduce Higgs
doublet fields just as in the SM. In the simplest and most evident scenario (see [15]),
there is one doublet Higgs Hn transforming as (2, qQ qD qU qQ ) under each
[SU(2) U (1)]n group. We also implement the replication of families by incorporating
330
family indices i, j = 1, . . . , 3. Another scenario to generate the (vector-like) fermion mass

hierarchy by assuming various link fields to connect arbitrary sites of the latticized fifth
dimension has been proposed in [19].
The gauge-invariant Yukawa terms read
ijU
N
(j )
D
Q (i)
n i2 Hn Un + ij
n=1
N
(j )
(i)
Q
n Hn Dn + H.c.
(9)
n=1
In order to extract the terms involving zero modes, which are the only terms relevant at low
energy limit, we rewrite (9) in the mass eigenbasis. However, this procedure depends explicitly on the specific CBCs being imposed on each of the fields Q, U , D. To be generic,
let us consider the following configuration. We assume the localization of zero modes
0L , U 0R and D 0R to be at n = 1, n = 1 and n = N , respectively. To achieve this localizaQ
tion pattern, we impose the following CBCs on these fields (see Eq. (2) and Appendix A,
Eq. (A.22))
(i)
(i)
N 1,N QN L = VQ QN 1L ,
(j )
(j )
N 1,N UN R = VU UN 1R ,
(k)
(k)
1,2
Q1R = QN R = 0,
U1L = UN L = 0,
D1L = DN L = 0,
(i)Q
(i)
(i)
(i)
(j )U
(j )
(j )
(j )
(k)
(k)
(k)D
(k)
D1R = VD D2R .
(10)
0L , U 0R and D 0R would be localized

Because of these boundary conditions, zero modes Q
at n = 1, n = 1 and n = N respectively, this also means that the first term of Eq. (9) would
represent the overlap between 2 wave function localized at the same site n = 1, while the
second term represents the overlap between wave functions localized at n = 1 and n = N .
Using (10) to eliminate
the dependent components and after the SM spontaneous symmetry
breaking Hn = (0, v/ 2 )T uniformly for all ns, we can rewrite the Yukawa term (9) as

(i)
(j )
3

VQ
VU
(i) (j )
U v
(i) U (j )
ij
Q
Q 1L U1R + (i)Q
N 1L N 1R
(j )U
2 i,j =1
N 1,N N 1,N

N
1

(i) (j )

(j
)
(i)
U +Q
U
Q
+
nL
nR
nR
nL
n=2
(k)
(i)
3

VQ
VD
v
(i) (k)
D
(i) D (k)
+ ij
Q D + (i)Q Q
N 1L N R
1L 2R
(k)D
2 i,k=1
N 1,N
1,2

N
1

(i) (k)

(i)
(k)
D +Q
D
Q
.
+
nL
nR
nR
nL
(11)
n=2
After going to the mass eigenbasis by the virtue of transformation of the type (7), keeping
only zero-mode terms and together with the assumption of universality for the Yukawa
couplings in the up and down sectors, we obtain the following effective mass terms
3
3

(i) M u U (j ) + Q
(i) M d D (k) ,
Q
ij
0R
0L
0L ik 0R
i,j =1
i,k=1
(12)
with
v
Miju = U
2

N 2
(i) (j )
UQ n,0 UU n,0
+ 1+
d
Mik
v
= D
2
331
n=1
(i)
VQ
(i)Q
N 1,N
(j )
VU
(j )U
N 1,N

(j )
(i)
UQ N 1,0 UU N 1,0 ,
(13)
N 2

(i) (k)
(i)
(i) (k)
UQ 1,0 + UQ 2,0 UD 2,0 +
UQ n,0 UD n,0
(k)
VD
(k)D
1,2
(k)
(i)
+ UQ N 1,0 UD N 1,0 +
(i)
VQ
(i)Q
N 1,N
(k)
UD N,0
n=3

.
(14)
Because all [UQ ], [UU ], [UD ] are unitary, the mass matrices M u , M d are generally complex. Thus in this simplest deconstruction approach, we might better understand the dynamical origin of CP-violation phase in the SM mass matrices. We also note that (13), (14)
0L , U 0R and D 0R are localized at n = 1, n = 1 and
represent the specific case where Q
n = N , respectively. All other localization configurations can be similarly found. Further,
when we replace link fields s in (13), (14) by their VEVs following the deconstruction,
these mass matrix elements will look much simpler (see (27), (28)).
Before moving on to give explicit expressions of these complex-valued mass matrices
in term of zero mode wave functions (Appendix A)
and perform the numerical fit, let us
briefly turn to the breaking pattern of product group N
n=1 [SU(2) U (1)]n .
2.3. Deconstructing [SU(2) U (1)]N
For the sake of completeness, in this section we will describe the breaking of [SU(2)
U (1)]N into the SM [SU(2) U (1)] gauge group by giving uniform VEVs to link fields.
The transformation and charge structure of fermions and scalar link fields have been defined in the beginning of previous section. To identify the unbroken symmetries following
the deconstruction, we look at the covariant derivative and kinetic terms of scalars
g0
g
U
U
+ iqU 0 Bn+1 n,n+1
,
Bn n,n+1
2
2
g
g
D
D
D
D
= n,n+1
iqD 0 Bn n,n+1
+ iqD 0 Bn+1 n,n+1
,
D n,n+1
2
2
U
U
= n,n+1
iqU
D n,n+1
(15)
(16)
where Bn is the gauge boson associated with U (1)n , while g0 is the common gauge coupling for all U (1)s. For Abelian groups, the opposite signs of the last two terms in (15)
U
D
(and n,n+1
) under U (1)n
(and also in (16)) originate from the opposite charges of n,n+1
U
and U (1)n+1 (so that terms like U n n,n+1 Un are gauge-invariant).
For non-Abelian groups, the similar sign reversing will hold for terms in the expression of covariant derivatives (see Eq. (21)), the nature of which also has its root in the
gauge invariance of the theory. Indeed, under the YangMills SU(2)n SU(2)n+1 gauge
332

Q
transformation (note that n,n+1 is a 2 2-matrix)

Q
n,n+1 Tn n,n+1 Tn+1

,
(17)
Qn Tn Qn ,
Qn+1 Tn+1 Qn+1 ,

i

Tn An Tn ( Tn )Tn ,
An
2
2
g0

i

Tn+1 An+1 Tn+1 ( Tn+1 )Tn+1
.
An+1
2
2
g0
(18)
(19)
(20)
The covariant derivative of n,n+1 must be formulated as follows (so that it transforms
exactly like
Q
n,n+1
in (17))

g
Q
Q
Q
Q
D n,n+1 = n,n+1 iqQ 0 Bn n,n+1 + ig0 A n n,n+1
2
2

g0

Q
Q

+ iqQ Bn+1 n,n+1 + ig0 n,n+1 An
,
2
2
(21)
where A n and Tn are respectively the gauge bosons and some 2 2-special unitary matrix
characterizing the SU(2)n transformation, while g0 is the common gauge coupling for all
SU(2)s.
Q
U,D
After the deconstruction n,n+1
VU,D , n,n+1 VQ 122 , the mass terms
for gauge bosons are generated. Specifically, we obtain as parts of kinetic terms
Q
Q
U
U
D
D
) (D n,n+1
), (D n,n+1
) (D n,n+1
), Tr[(D n,n+1 ) (D n,n+1 )] the fol(D n,n+1
lowing gauge bosons squared mass matrices
1 1
1 2
2
..
,
MB = B
.
2 1
1 1
1 1
1 2
2
.
,
..
MA = A
(22)
2 1
1 1
where, after restoring the family replication index (i = 1, 2, 3),
B =
3

1
2

2 (i) 2
2 (i) 2
VD + qQ
VQ
,
g0 2 qU2 VU(i) + qD
A =
3

2
g02 VQ(i) . (23)
Both matrices in (22) have a flat zero eigenstate. This indeed indicates the uniform breaking of [SU(2)U (1)]N into the diagonal (SM) group [SU(2)U (1)], whose gauge bosons
333
are massless and given by

N
1
B =
Bn ,
N n=1
N
1
A =
An .
N n=1
(24)
In Eq. (23) it is also shown that the pattern of symmetry breaking is not spoiled by family
replication as long as charges qU (and qD , qQ ) are independent of the site index n under a
(i)
presumed permutation symmetry (just like VU,D,Q ). Finally, by extracting the interaction
between fermions and massless gauge bosons from fermion kinetic
terms in (1)
one can
see that the couplings of the unbroken group scale as g = g0 / N and g = g0 / N , while
the charge structure (of fermions in mass eigenbasis) under this diagonal group remains
intact.
3. Fitting the models parameters

3.1. Model, parameters and numerical method
In the previous section we have outlined the process diagonalizing the squared-mass
matrix (6). The complete diagonalization process is complicated, but as we are concerned
only with the zero eigenvalue problem, the computation can be done analytically in the
MQ ] (8), the zero eigenstate

general term (see Appendix A). Since [UQ ] diagonalizes [MQ
MQ ] is just the first column of [UQ ], i.e., in the notation of Appendix A

of [MQ
(i)
(i)
UQ n,0 = xQn ,
and similarly
(j )
(j )
UU n,0 = xU n ,
(25)
(k)
(k)
UD n,0 = yDn
,
(26)
where xn s are given in (A.20) (corresponding to a zero mode localized at the end point
n = 1) and yn s in (A.24) (corresponding to a zero mode localized at the end point n = N ).
After the spontaneous symmetry breaking, the link fields acquire an uniform VEV
VQ,U,D respectively (independent of site index n). In term of xn s and yn s, the SM mass
matrices (13), (14) for up and down quark sectors become

N 1
(i) (j )
v
(j )
(i)
u
Mij = U
(27)
xQn xU n + xQN1 xU N 1 ,
2
n=1
N 1

(i) (k)
v
(i) (k)
(i)
(k)
d
Mik = D xQ1 yD2 +
(28)
xQn yDn + xQN1 yDN ,
2
n=2
where xn s, yn s are given in (A.20), (A.24), respectively. The analytical forms of (27),
(28) in term of models parameters are worked out in Appendix C, Eqs. (C.1), (C.2).
Again, let us remind ourselves that (27) represents the overlap between two wave functions localized at the same site n = 1 while (28) represents the overlap between one wave
334
function localized at n = 1 and the other at n = N . The model under consideration consists of 20 real parameters (see Table 2 and Appendix B): 3 complex VEV V s for each
complete quark
generation(Q, U, D)i (i = 1, 2 or 3), and 2 real dimensionful Yukawa
couplings U v/ 2, D v/ 2. We choose to fix N = 10 throughout.
The numerical approach to fit the parameters consists in minimizing a positive function which gets a zero value when all the predicted quantities are in the corresponding
experimental ranges [18]. The minimization procedure is based on the simulated annealing
method, which seems working better than other minimization approaches when the parameter space becomes larger [20,21]. The input referencing physical quantities are given in
Table 1 of Appendix D.
We consider eight different cases, which correspond to all the eight possible ways of
localizing the left and right components. The eight different cases are the following:
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Q, U and D localized in n = 1 denoted as (QU D1);

Q and U localized in n = 1, D localized in n = N denoted as (QU 1DN );
Q and D localized in n = 1, U localized in n = N denoted as (QD1U N );
Q localized in n = 1, U and D localized in n = N denoted as (Q1U DN );
Q, U and D localized in n = N denoted as (QU DN );
D localized in n = 1, Q and U localized in n = N denoted as (D1QU N );
U localized in n = 1, Q and D localized in n = N denoted as (U 1QDN );
U and D localized in n = 1, Q localized in n = N denoted as (U D1QN ).
We specially note that, due to the mirror complexity between CBCs (2) and (A.22), the
mass matrices obtained in the cases (1) and (5), cases (2) and (6), cases (3) and (7), cases
(4) and (8), are complex conjugate pairwise. In the result, all eight cases are inequivalent.
3.2. Numerical results
In the following we present the characteristically important numerical results for the
four cases out the eight mentioned above, for which we were able to find solutions. The
cases are referred to in the above order. For each case we give one particular, but typical,
numerical complete set of the 20 defining parameters (Table 2), the quark mass matrices
and quark mass spectra, the CKM matrix and the CP parameters. Complex phases are measured in radiant, and N = 10 for all cases. The masses are given in GeV and are evaluated
at the MZ scale. For the sake of visualization, we also present graphically the comprehensive solutions of the quark wave function profiles in the theory space (Fig. 1), the mass
spectrum (Fig. 2), the CKM matrix (Fig. 3) and the
CP parameters (Fig. 4) for the
case of all fields Q, U and D localized at the same site n = 1.
Case (1): Q, U and D localized in n = 1.
0.925e0.558i
(QU D1)
Mu
= 78.4 GeV 0.027e2.009i
0.948e0.306i
0.923e0.501i
0.027e2.046i
0.942e0.367i
0.951e0.570i
0.029e2.006i
0.973e0.280i

,
(29)
(i)
(i)
335
(i)
Fig. 1. Profiles of the absolute value of normalized wave functions |xQn |, |xU n | and |xDn | in the theory space
(2)
(N = 10) for the case with Q, U and D localized at n = 1. |xQn | with a value of
1 is less localized.
D1)
m(QU
= 0.0021 GeV,
u
(QU D1)
mt
D1)
m(QU
= 0.702 GeV,
c
= 181.1 GeV,
(30)
Md(QU D1)
0.909e1.490i
= 1.35 GeV 0.030e2.649i
0.848e2.339i
(QU D1)
= 0.0045 GeV,
(QU D1)
= 2.89 GeV.
md
mb
0.782e2.072i
0.048e0.930i
0.799e1.353i
0.960e0.992i
0.032e3.025i
0.918e1.838i

,
(31)
D1)
m(QU
= 0.106 GeV,
s
(32)
In Eqs. (29), (31) the mass matrices are written in a form that better shows deviations
from the democratic structure. In Eq. (33) we give the expression for the CKM matrix, in
Eq. (34) the values for the CP parameters and .

0.975 0.009i 0.151 0160i 0.001 0.003i
(QU D1)
= 0.015 + 0.219i 0.669 + 0.709i 0.029 + 0.024i ,
VCKM
(33)
0.003 0.009i 0.029 0.023i
0.670 + 0.742i
(QU D1) = 0.12,
(QU D1) = 0.30,
(34)
with and defined as

2
Vcd Vcb /Vcd Vcb
,
= Re Vud Vub
(35)

2
Vcd Vcb /Vcd Vcb
.
= Im Vud Vub
(36)
Case (2): Q and U localized in n = 1, D localized in n = N .

Mu(QU 1DN )
0.918e0.039i
= 66.6 GeV 0.941e0.038i
0.930e0.058i
0.609e0.590i
0.637e0.601i
0.622e0.585i
0.924e0.135i
0.946e0.132i
0.935e0.154i

,
(37)
336
Fig. 2. Solutions for the 6 quark masses in the case with Q, U and D localized at n = 1. The masses in GeV are
evaluated at the MZ scale. The range for each mass is given by the edges of the corresponding window.
1DN )
m(QU
= 0.0020 GeV,
u
(QU 1DN )
mt
= 168.3 GeV,
(QU 1DN )
Md
(QU 1DN )
md
1DN )
m(QU
= 0.687 GeV,
c
(38)
0.041e1.959i
= 23.2 GeV 0.037e1.854i
0.038e1.982i
= 0.0045 GeV,
1DN )
= 2.90 GeV,
m(QU
b
0.045e0.025i
0.042e0.003i
0.043e0.047i
0.043e2.526i
0.042e2.570i
0.043e2.605i

,
(39)
1DN )
m(QU
= 0.084 GeV,
s
(40)
337
Fig. 3. Solutions for the absolute values of the CKM matrix elements in the case with Q, U and D localized at
n = 1. The range for each element is given by the edges of the corresponding window.

(QU 1DN )
VCKM
0.975 + 0.029i
0.168 0.141i
0.003 + 0.010i
(QU 1DN ) = 0.19,
0.097 + 0.197i
0.880 + 0.420i
0.039 0.007i
0.001 + 0.003i
0.039 0.011i
0.999 + 0.011i

,
(QU 1DN ) = 0.33.
(41)
(42)
Case (5): Q, U and D localized in n = N .

Mu(QU DN )
0.887e0.494i
= 78.4 GeV 0.038e2.066i
0.895e0.410i
DN )
m(QU
= 0.0022 GeV,
u
0.881e0.478i
0.038e2.070i
0.877e0.429i
DN )
m(QU
= 0.674 GeV,
c
0.913e0.577i
0.041e2.039i
0.929e0.316i

,
(43)
338
Fig. 4. Solutions for and in the case with Q, U and D localized at n = 1.

(QU DN )
mt
= 172.6 GeV,
(44)
(QU DN )
Md
0.895e1.619i
= 1.37 GeV 0.058e2.456i
0.835e2.473i
(QU DN )
= 0.0049 GeV,
(QU DN )
= 2.90 GeV,
md
mb

(QU DN )
VCKM
0.943e
0.063e2.339i
0.893e2.468i
(45)
DN )
m(QU
= 0.106 GeV,
s
(46)
0.974 + 0.042i
0.134 0.180i
0.010 + 0.006i
(QU DN ) = 0.31,
0.776e1.668i
0.040e1.459i
0.832e0.871i
1.622i
0.046 + 0.220i
0.676 0.701i
0.022 + 0.030i
0.003 0.003i
0.020 0.033i
0.646 0.762i

,
(47)
(QU DN ) = 0.30.
(48)
Case (6): D localized in n = 1, Q and U localized in n = N .

Mu(D1QU N )
0.675e1.829i
= 71.2 GeV 0.706e1.783i
0.671e1.837i
0.824e0.198i
0.856e0.173i
0.822e0.207i
0.837e0.732i
0.868e0.698i
0.834e0.741i

,
(49)

N)
m(D1QU
= 0.0026 GeV,
u
339
N)
m(D1QU
= 0.725 GeV,
c
N)
= 169.2 GeV,
m(D1QU
t
(50)
(D1QU N )
Md
0.026e0.996i
= 26.5 GeV 0.027e0.868i
0.025e0.957i
(D1QU N )
= 0.0044 GeV,
(D1QU N )
= 2.91 GeV,
md
mb

(D1QU N )
VCKM
0.044e1.678i
0.042e1.655i
0.044e1.710i

,
(51)
N)
m(D1QU
= 0.088 GeV,
s
0.972 0.075i
0.050 0.218i
0.004 0.012i
(D1QU N ) = 0.26,
0.039e3.019i
0.037e3.017i
0.039e3.050i
(52)
0.069 + 0.213i
0.974 + 0.013i
0.038 0.014i
0.001 + 0.004i
0.039 0.016i
0.998 + 0.044i
(D1QU N ) = 0.38.

,
(53)
(54)
We are now ready for comments on the presented numerical solutions.
4. Concluding comments
In this paper we have reconstructed
the observed complex mixing of quark flavors,

starting with the product group N
[SU(2)
U (1)]n at a higher energy scale. The den=1
construction of this product group into the electroweak gauge group can indeed provide all
necessary components to generate such mixing.
We have built a specific models with 20 parameters to fit the quark mass spectrum and
the CP phase. However, the numerical fit is found only for the preferred configurations
where fermion fields Q and U are localized at the same position in the theory space. Arguably, this is because the ratio U /D of Yukawa couplings can be responsible only for
the difference in the overall scale of up and down-quark masses, while the more hierarchical internal mass spectrum of the up-quark sector (compared to that of the down-quark
sector) would still require a higher degree of overlapping.
As far as the structure of mass matrices is concerned, the deviation from democracy
is moderate. In all the cases, the mass matrices assume a hierarchy with two rows (or
two columns) having similar absolute value matrix elements, with the third row (or third
column) having different values, but still similar along that row (or that column). A quite
close mass matrix structure was found in [18], but in a different approach.
We did not perform a study of the dependence on the number of deconstruction subgroups N . We expect anyway that the fitting would be more feasible for larger N as the
wave functions and their overlaps then can be tuned more smoothly. In the other direction,
the constraint from flavor changing neutral current that sets an upper limit on the length of
extra dimension in the split fermion scenario (see, e.g., [22]) is also expected to set an upper
limit on the ratio N/V (between N and the VEV of link field) in the deconstruction theory.
We however leave a more careful analysis of these and other relevant phenomenological
issues for future publications.
340
Acknowledgements
P.Q.H. and N.-K.T. are supported in part by the US Department of Energy under Grant
No. DE-A505-89ER40518. A.S. is supported by grant NSC 93-2811-M-002-047. N.-K.T.
also acknowledges the Dissertation Year Fellowship from UVA Graduate School of Arts
and Sciences.
Appendix A. Fermion zero mode in dimensional deconstruction

In this appendix we will work out the general expression of zero eigenstate of the matrix
of the type (6). This mode plays a special role because it will be identified with the SM
chiral fermions. To simplify the writing, here we denote this zero eigenstate generally as
{x1 , x2 , . . . , xN 1 } while in Section 3 we will restore all omitted scripts Q, U, D, i, j .
A.1. Zero-mode localization at the end-point n = 1
The equation set determining the zero eigenstate (6) is
|V |2 x1 MV x2 + V 2 x3 = 0 V x1 Mx2 + V x3 = 0,

MV x1 + M 2 + |V |2 x2 2MV x3 + V 2 x4 = 0,

V 2 x1 2MV x2 + M 2 + 2|V |2 x3 2MV x4 + V 2 x5 = 0,
..
.

V 2 xN 5 2MV xN 4 + M 2 + 2|V |2 xN 3 2MV xN 2 + V 2 xN 1 = 0,

V 2 xN 4 2MV xN 3 + M 2 + 2|V |2 xN 2 + (V 2 2MV )xN 1 = 0,

V 2 xN 3 V 2 2MV xN 2 + M 2 M V + V + 2|V |2 xN 1 = 0.
(A.1)
(A.2)
(A.3)
(A.4)
(A.5)
(A.6)
After a bit of algebra, we can equivalently transform this equation set into
X1 = X2 ||2 X3 ,
(A.7)
X2 = X3 || X4 ,
..
.
(A.8)
XN 3 = XN 2 ||2 XN 1 ,
(A.9)
XN 2 = XN 1 XN 1 ,
(A.10)
where we have introduced new parameter and variables

||ei
V
|V |ei
=
,
M
M
N n1
Xn
xn
(n = 1, . . . , N 1).
(A.11)
(A.12)
341
We note that V (and ) is a complex parameter in general (see Appendix B). The new
simple recursion relation allows us to analytically determine the set {X1 , . . . , XN 1 } (and
then the zero eigenstate {x1 , . . . , xN 1 }) for any (i.e., for any real M and complex V ).
After some combinatorics1 we obtain for 1 n N 3
N 3 n k
k+1
||2
XN 1
Xn =
k
k=0
N 2 n k
k
||2 XN 2
+
k
k=0
(N 3 n k)!
k+1
||2
=
XN 1
k!(N 3 n 2k)!
k=0
(N 2 n k)!
k
||2 XN 2 ,
k!(N 2 n 2k)!
(A.13)
k=0
and for n = N 2 (see (A.10))

XN 2 = (1 )XN 1 .
(A.14)
Using the equality

m
m
m+1
+
=
,
p
p+1
p+1
(A.15)
we can rewrite (A.13) as (with 1 n N 3)

N 1 n k
k
||2
Xn =
k
k=0

N 2 n k

2 k
||
XN 1 .
k
(A.16)
k=0
Again, using another equality [24]

k p1k
(1)
sinh px = sinh x
(2 cosh x)p12k ,
k
(A.17)
k=0
we obtain for 1 n N 3 and for || <

Xn =
2XN 1
1 4||2
1
2
N n

||
sinh (N n) ||N 1n sinh (N 1 n) , (A.18)
with
cosh
1
1
cosh1
2||
2||
( > 0).
1 Another somewhat simpler solution of the above equation set is presented in [16].
(A.19)
342
For || > 12 , the expression of XN is similar to (A.18) but with hyperbolic functions (sinh
and cosh) replaced respectively by trigonometric ones (sin and cos).
Finally, from (A.12) we have altogether

xn = Cein sinh (N n) ei sinh (N 1 n) (1 n N 1), (A.20)
where C is the normalization constant determined by the normalization equation
N
1
|xn |2 = 1.
(A.21)
n=1
We note that this normalization is nothing other than the unitarity condition of the rotation
matrix U (see (8) and (25), (26)).
A.2. Zero-mode localization at the end-point n = N
The chiral boundary conditions (CBC) and the value of parameter || |V |/|M| are
two crucial factors that determine the localization pattern of the chiral zero-mode of
fermion. For, e.g., in the previous subsection we have seen that, when || < 1/2, along
Q
with CBCs Q1R = QN R = 0, N 1,N QN,L = VQ QN 1,L (2) we can localize the lefthanded zero mode of Q field around site n = 1 (A.20).
On the intuitive ground, we expect that the mirror image of (2) (apart from the requirement || < 1/2)
Q1,R = QN,R = 0,
Q 1,2 Q1,L = VQ Q2,L ,
(A.22)
would produce a left-handed zero mode of Q field localized at n = N . A similar calculation

indeed confirms this localization pattern. Specifically, if we denote yi (i = 2, . . . , N ) the
zero-mode subject to CBCs (A.22), and xj (j = 1, . . . , N 1) subject to CBCs (2) as
before, we find
yi = xN
+1i
(i = 2, . . . , N ),
or even more explicitly (see (A.20))

yn = Cei(N +1n) sinh (n 1) ei sinh (n 2)
(A.23)
(2 n N ).
(A.24)
Appendix B. The complex-valued link field VEV from broken Wilson line
Since the link fields n,n+1 transform non-trivially under two different groups, we may
expect its VEV to be complex in general. It is because in this case the VEVs phase could
not be rotated away in general. The standard and rigorous method to determine the VEV
is to write down and then minimize the corresponding potential. It turns out [23] that there
always exist ranges of potential parameters which generate complex VEV. In this appendix,
however, we just recapitulate the complexity nature of link field VEV from the latticized
extra dimension perception which is derived in [14] in details. Though the approach taken
343
in this work does not strictly stem from latticizing the fifth dimension, this perception could
still serve as the principle illustration.
To make the connection between DD theory and its latticized ED counterpart, we interpret the link field as a Wilson line connecting two neighboring branes
(n+1)a

n,n+1 exp
igy dy exp(igan ),
(B.1)
na
where n essentially is the ED component of gauge field, g and a are gauge coupling and
lattice spacing, respectively.
Following the DD symmetry breaking [SU(2) U (1)]N [SU(2) U (1)], only one
linear combination 0 of link fields remains massless at the classical level
N
1
0 =
n .
N n=1
(B.2)
In the leading order with radiative correction, by minimizing the 1-loop effective potential
of 0 , one obtains2
0 =
2k
ga N
(k N ).
(B.3)
From (B.1), (B.2), one see that in the leading order the link fields assume a uniform complex VEV

i2k
n,n+1 exp
(B.4)
.
N
Actually, this phase can be considered arbitrary.
Appendix C. Wave function overlap in theory space

In this appendix we present the analytical expressions of zero-mode wave function overlaps in the theory space, from which follow the mass matrix elements Miju,d (27), (28).
These expressions in turn were compiled using the exact solutions (A.20), (A.24) for the
wave functions. In what follows we use XX to denote the overlap of two wave functions
localized at the same site n = 1, and XY the overlap of the first wave function localized
at n = 1 and the second at n = N . All other overlap configurations can be easily found by
virtue of relation (A.23).
2 The finiteness of 1-loop effective potential requires the mass of fermionic tower be trigonometric function
of the mode number [14]. In the continuum limit such as in a S 1 /Z2 compactification, the orbifold boundary
conditions (2) can fulfill this requirement.
344
It follows from Eqs. (27), (28) that

XX =
N 1

xn(1) xn(2)
(1) (2)
+ xN
1 xN 1
n=1

C1 C2 e(N 1)(i1 i2 1 2 ) 1 N (1 +2 )
1 ei1 1 1 ei2 2
=
e
4
1 e(i1 i2 1 2 )

e(N 1)(i1 i2 1 +2 ) 1 N (1 2 )
1 ei1 1 1 ei2 +2
e
(i
i
+
)
1
2
1
2
1e
(N
1)(i
i
+
)
1
2
1
2

e
1 N (1 +2 )
1 ei1 +1 1 ei2 2
e
(i
i
+
)
1
2
1
2
1e

e(N 1)(i1 i2 +1 +2 ) 1 N (1 +2 )
i1 +1
i2 +2
+
1
e
1
e
e
1 e(i1 i2 +1 +2 )
+ C1 C2 ei(N 1)(1 2 ) sinh 1 sinh 2 ,
XY
= x1(1) y2(2)
N 1
(C.1)

(1) (2)
xn(1) yn(2) + xN
1 yN
n=2
= C1 C2 e
i1 22

sinh (N 1)1 ei1 sinh (N 2)1 sinh 2
(N 2)(i1 i2 1 2 )
e
1 i1 i2 1 2 N 1 +2
C1 C2
e
e
(i
i
)
1
2
1
2
4
1e

1 ei1 1 1 ei2 +2
+
e(N 2)(i1 i2 1 +2 ) 1 i1 i2 1 +2 N 1 2
e
e
1 e(i1 i2 1 +2 )

1 ei1 1 1 ei2 2
+
e(N 2)(i1 i2 +1 2 ) 1 i1 i2 +1 2 N 1 +2
e
e
1 e(i1 i2 +1 2 )

1 ei1 +1 1 ei2 +2
+
e(N 2)(i1 i2 +1 +2 ) 1 i1 i2 +1 +2 N 1 2
e
e
1 e(i1 i2 +1 +2 )

i1 +1
i2 2
1e
1e
+ C1 C2 ei(N 1)1 N 2 sinh 1

sinh (N 1)2 ei2 sinh (N 2)2 ,
(C.2)
where C1 , C2 are the normalization factors, which are determined also from the overlap of
the respective wave function with itself.
345
Appendix D. Numerical tables

Table 1
Central values and uncertainties for the masses of the 6 quarks evaluated
at MZ , for the two ratios mu /md and ms /md , for the absolute values
of the CKM matrix elements and the CP parameters ,

xi
xi
|ximax ximin |/2
mu
mc
mt
md
ms
mb
mu /md
ms /md
|Vud |
|Vus |
|Vub |
|Vcd |
|Vcs |
|Vcb |
|Vtd |
|Vts |
|Vtb |
2.33 103
0.45 103
0.061
13
0.66 103
0.0130
0.11
0.119
3.9
0.00075
0.0035
0.0115
0.0035
0.0008
0.003
0.005
0.0035
0.00015
0.10
0.05
0.685
181
4.69 103
0.0934
3.00
0.497
19.9
0.97485
0.2225
0.00365
0.2225
0.9740
0.041
0.009
0.0405
0.99915
0.22
0.35
Table 2
20-parameter space solutions found in 4 different cases of the model presented in Section 3.1 (N = 10 for all
cases)
Q1
Q2
Q3
U 1
U 2
U 3
D1
D2
D3
Q1
Q2
Q3
U 1
U 2
U 3
D1
D2
D3
U v/2
D v/ 2
(QUD1)
(QU1DN)
(QUDN)
(D1QUN)
2.290
0.007
1.497
0.771
0.759
0.927
0.829
0.535
1.123
1.002
3.862
1.678
1.163
1.100
1.247
0.204
3.098
0.086
78.37
1.35
0.208
0.236
0.220
0.439
1.900
0.433
0.027
0.022
0.029
0.974
0.983
0.989
1.407
14.78
5.290
0.535
4.715
9.645
66.63
23.24
2.311
0.011
1.656
0.623
0.603
0.722
0.794
0.471
1.105
1.234
4.702
1.999
1.276
1.246
1.420
0.135
2.809
0.174
78.36
1.37
0.215
0.251
0.213
0.572
0.823
0.745
0.541
0.057
0.022
0.959
0.958
0.965
0.197
13.67
5.52
0.182
3.471
9.499
71.21
26.48
346
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
N. Arkani-Hamed, A.G. Cohen, H. Georgi, Phys. Rev. Lett. 86 (2001) 4757.

C.T. Hill, S. Pokorski, J. Wang, Phys. Rev. D 64 (2001) 105005.
N. Jackiw, S. Rebbi, Phys. Rev. D 13 (1976) 3398.
N. Arkani-Hamed, M. Schmaltz, Phys. Rev. D 61 (2000) 033005.
P.Q. Hung, N-K. Tran, Phys. Rev. D 69 (2004) 064003.
N. Cosme, J.M. Frere, L.L. Honorez, Phys. Rev. D 68 (2003) 096001.
N. Cosme, J.M. Frere, Phys. Rev. D 69 (2004) 036003.
B. Gdzadkowski, J. Wudka, hep-ph/0401232.
Y. Hosotani, Phys. Lett. B 129 (1983) 193.
E.A. Mirabelli, M. Schmaltz, Phys. Rev. D 61 (2000) 113011.
G.C. Branco, A. De Gouvea, M.N. Rebelo, Phys. Lett. B 506 (2001) 115.
A. Soddu, N.-K. Tran, Phys. Rev. D 69 (2004) 015010.
W. Skiba, D. Smith, Phys. Rev. D 65 (2002) 095002.
C.T. Hill, A. Leibovich, Phys. Rev. D 66 (2002) 016006.
H. Abe, T. Kobayashi, N. Maru, K. Yoshioka, Phys. Rev. D 67 (2003) 045019.
P.Q. Hung, N.-K. Tran, Phys. Rev. D 70 (2004) 126014.
P.Q. Hung, M. Seco, Nucl. Phys. B 653 (2003) 123.
P.Q. Hung, M. Seco, A. Soddu, Nucl. Phys. B 692 (2004) 83.
S. Nojiri, S.D. Odintsov, A. Sugamoto, Phys. Lett. B 590 (2004) 239.
S. Kirkpatrick, C.D. Gelatt, M.P. Vecchi, Science 220 (1983) 671;
S. Kirkpatrick, J. Stat. Phys. 34 (1984) 975;
N. Metropolis, A. Rosenbluth, M. Rosenbluth, A. Teller, E. Teller, J. Chem. Phys. 21 (1953) 1087.
W.H. Press, S.A. Teukolosky, W.T. Vetterling, B.P. Flannery, Numerical Recipes in C, second ed., Cambridge Univ. Press, New York, 1992.
A. Delgado, A. Pomarol, M. Quiros, J. High Energy Phys. 0001 (2000) 030.
F. Bauer, M. Lindner, G. Seidl, J. High Energy Phys. 0405 (2004) 026.
I.S. Gradshteyn, I.M. Ryzhik, Tables of Integrals, Series and Products, fifth ed., Academic Press, Boston,
1994.
Neutrino quasinormal modes of a KerrNewman

de Sitter black hole
Jia-Feng Chang a,b,c , You-Gen Shen a,c,d
a Shanghai Astronomical Observatory, Chinese Academy of Sciences, Shanghai 200030, China
b Graduate School of Chinese Academy of Sciences, Beijing 100039, China
c National Astronomical Observatories, Chinese Academy of Sciences, Beijing 100012, China
d Institute of Theoretical Physics, Chinese Academy of Sciences, Beijing 100080, China
Received 25 October 2004; received in revised form 16 December 2004; accepted 22 December 2004
Abstract
Using the PshlTeller approximation, we evaluate the neutrino quasinormal modes (QNMs) of a
KerrNewmande Sitter black hole. The result shows that for a KerrNewmande Sitter black hole,
massless neutrino perturbation of large , positive m and small value of n will decay slowly.
PACS: 04.70.-s; 04.50.+h; 11.15.-q; 11.25.Hf
1. Introduction
It is well known that there are three stages during the evolution of the field perturbation in the black hole background: the initial outburst from the source of perturbation, the
quasinormal oscillations and the asymptotic tails. The frequencies and damping time of
the quasinormal oscillations called quasinormal modes (QNMs) are determined only by
the black holes parameters and independent of the initial perturbations. A great deal of
efforts have been devoted to the black holes QNMs for the possibility of direct identification of black hole existence through gravitational wave detectors in the near future [1,2].
E-mail address: ygshen@center.shao.ac.cn (Y.-G. Shen).
doi:10.1016/j.nuclphysb.2005.01.043
348
J.-F. Chang, Y.-G. Shen / Nuclear Physics B 712 (2005) 347370
The study of black holes QNMs has a long history. Most of the studies immersed in an
asymptotically flat spacetime. The discovery of the AdS/CFT [3,4] correspondence and
the expanding universe motivated the investigation of QNMs in de Sitter [5,6] and anti-de
Sitter [710] spacetime in the past several years.
Most methods in evaluating the QNMs are numerical in nature. Recently, using the
third-order WKB approximation, Cho evaluated the Dirac field QNMs of a Schwarzschild
black hole [26]. A powerful WKB scheme was devised by Schutz and Will [15], and was
extended to higher orders in [16]. Konoplya [17] extended the WKB approximation to
sixth-order and calculated the QNMs of a D-dimensional Schwarzschild black hole. Zhidenko [18] calculated low-laying QNMs of a Schwarzschildde Sitter black hole by using
sixth-order WKB approximation and the approximation by PshlTeller potential. Cardoso
[19] calculated QNMs of the near extremal Schwarzschildde Sitter black hole by using
PshlTeller approximation, which was proved to be exactly in the near extreme regime
[20]. Yoshida [21] numerically analyzed QNMs in nearly extremal Schwarzschildde Sitter spacetimes. The Kerr black hole is a more general case. It is also important to note that
the most important QNMs are the lowest ones which have smaller imaginary on the astrophysical aspect and the most important spacetimes are the asymptotically flat and now
perhaps the asymptotically de Sitter which supported by the recent observation data. So
we discuss the QNMs of a KerrNewmande Sitter black hole in this paper. Leaver [22]
developed a hybrid analytic-numerical method to calculate the QNMs of black holes and
applied to the Kerr black hole. Seidel and Iyer [23] computed the low-laying QNMs of
Kerr black holes for both scalar and gravitational perturbations by using third-order WKB
approximation. Berti et al. dealt with highly damped QNMs of Kerr black holes in [24,25].
In this paper, we evaluate the QNMs of KerrNewmande Sitter black hole for neutrino
perturbation. In Section 2 we consider the massless Dirac equations for massless neutrino
in the KerrNewmande Sitter black hole and reduced it into a set of Schrdinger-like
equations with a particular effective potential. We analyse the properties of the particular
potential in Section 3 and use the PshlTeller potential approximation to evaluate the
QNMs of massless neutrino in Section 4. Conclusions and discussions are presented in
Section 5. Throughout this paper we use units in which G = c = M = 1.
2. Massless Dirac field equation in the KerrNewmande Sitter black hole

Generally speaking, neutrino is a kind of uncharged Dirac particles without rest mass or
with tiny mass. In curved spacetime, the spinor representations of massless Dirac equations
are [11]
AB P A = 0,
(2.1)
AB Q = 0,
(2.2)
and
are two two-component spinors, the operator AB denotes the spinor
where
covariant differentiation. AB = AB , and AuB are 2 2 Hermitian matrices which sat

isfy g AB C D = AC B D , where AC and B D are antisymmetric Levi-Civita symbols,
the operator is covariant differentiation.
PA
QA
349
The metric of the KerrNewmande Sitter black hole in the BoyerLindqust coordinate
system is
ds 2 =

2 2 2
r a 2 sin2 dt 2
dr
d 2
r
2

1
2 2 r 2 + a 2 r a 2 sin2 sin2 d 2

2a
+ 2 2 r 2 + a 2 r sin2 dt d,

1
2 2
(2.3)
where
= r + ia cos ,
2 = ,

r = r 2 + a 2 1 + r 2 2Mr + Q2 ,
3
1
1
= 1 + a 2 .
= 1 + a 2 cos2 ,
3
3
Here, a and Q are the angular momentum per unit mass and electric charge of the
hole, M is the black hole mass and is the positive cosmological constant.
The contravariant component of metric tensor is

2 (r 2 +a 2 )2 a 2 sin2
a 2 2 r 2 +a 2
1

0
0
r
r
2
2
0
0
0
2
g =
.
0
0
0
2

2
2
2
2
2
2
2
a r +a
1
1
a sin
0
0
2 2 r
r
2
(2.4)
(2.5)
black
(2.6)
sin
Choose the null tetrad as follows:

2
(r + a 2 )
a
l =
,
, 1, 0,
r
r

1
n = 2 r 2 + a 2 , r , 0, a ,
2

i
1
,
ia sin , 0,
m =
sin
2

1
i
ia
sin
,
0,
.
m
=
,
sin
2
(2.7)
The above null tetrad consists of null vector, i.e.,

l l = n n = m m = 0.
(2.8)
The null vector satisfies the following pseudo-orthogonality relations

= 1,
l n = m m
l m = l m
= n m = n m
= 0.
(2.9)
350
They also satisfy metric conditions

g = l n + n l m m
m
m .
(2.10)
Set spinor basis aA = aA , in which A is the spinor component index, a is the spinor
basis index, both indices are 0 or 1.
The covariant differentiation AB A for an arbitrary spinor A can be represented as
the component along the spinor basis aA , i.e.,
c
d,
aA bB Cc AB C = a b c = a b c + da
b
where a b are ordinary spinor derivatives,

Now let
00 = l D,
01 = m ,
c
da
b
(2.11)
are spin coefficients.
11 = n ,
=m
.
10
(2.12)
Then the Dirac equations (2.1) and (2.2) can be rewritten as four coupled equations
(D + )F1 + ( + )F2 = 0,
( + )F2 + ( + )F1 = 0,

D + G2 + G1 = 0,

+ G2 + + G1 = 0,
(2.13)
where F1 , F2 , G1 , G2 are four-component spinors with F1 = P 0 , F2 = P 1 , G1 = Q1 ,
G2 = Q0 . , , , , , , , , etc., are NewmanPenrose symbols, while , , etc.,

are, respectively, the complex conjugates of , , etc. The NewmanPenrose symbols are
1
,
k = l; m l = 0,
= n; m
m
= 0,

ia sin
= l; m n =
,
= l; m m = 0,
22
r
= n; m
n = 0,
= n; m
m = 2 ,
2

ia sin
1

= n; m
l =
,
= l; n l m; m
l = 0,
2
2
2( )

1
1 dr
+ ,
= l; n n m; m
n = 2
2
4 dr

d( sin )
1
1

,
= l; n m m; m
m =
2
d
2 2 sin

1
= l; n m
m; m
m
= .
2
Take three transformations as follows:
= l; m m
=
F1 = eit eim f1 (r, ),
F2 = eit eim f2 (r, ),
(2.14)
G1 = eit eim g1 (r, ),
U1 (r, ) = f1 (r, ),
V1 (r, ) = g1 (r, ),
G2 = eit eim g2 (r, ),

V2 (r, ) = g
2 (r, ),
U2 = R+ 1 (r)S+ 1 ( ),
V1 = R+ 1 (r)S 1 ( ),
V2 = R 1 (r)S+ 1 ( ).
Eqs. (2.13) turn into
(2.15)
U2 (r, ) = f2 (r, ),
U1 = R 1 (r)S 1 ( ),
2
351
(2.16)
(2.17)
L 1 R 1 S 1 = 0,
2
2
2 2 +2 +2

2 L+1 R 1 S 1 = 0,
r D+
1 R+ 1 S + 1
2
2
2
2
2
2

+
D 0 R 1 S + 1
L R 1 S 1 = 0,
2
2
2 12 + 2 2

r D+
2 L+1 R 1 S+ 1 = 0,
1 R+ 1 S 1 +
D 0 R 1 S 1 +
(2.18)
where
iK
s dr
,
+
r
r dr
iK
s dr
Ds+ = r +
,
+
r
r dr
d( sin )
H
s
Ls =
,
+
d
sin
d( sin )
H
s
+
,
Ls = +
+
d
sin
2

K = r + a 2 am,
m
.
H = a sin
sin
By using separation of variables, Eqs. (2.18) become

D 0 R 1 = 1 R+ 1 ,
L 1 S 1 = 1 S 1 ,
2
2
2
2 2 +2

+
+
r D 1 R+ 1 = 2 R 1 ,
2 L 1 S 1 = 2 S+ 1 ,
2
2
2
2
2
2

+
D 0 R 1 = 3 R+ 1 ,
L S 1 = 3 S + 1 ,
2
2
2
2 12 2

+
r D 1 R+ 1 = 4 R 1 ,
2 L 1 S+ 1 = 4 S 1 ,
Ds = r
(2.19)
(2.20)
where 1 , 2 , 3 , 4 are the separation constants, and

1
1
1 = 3 = 2 = 4 .
2
2
(2.21)
352
Then Eqs. (2.20) can be written as

r D+
1 R+ 1 = 2R 1 ,
D0 R 1 = R+ 1 ,
2
1
2
L 1 S+ 1
2
Substituted
= 2S 1 ,
1
2
L+1 S 1 =
2S+ 1 .
2
(2.22)
2 with , 2R 1 with R 1 , Eqs. (2.22) can be written as

2
1
2
1
2
1
2
1
2
L 1 S+ 1 = S 1 ,
2
1
2
r D0+ r R+ 1 = R 1 ,
r D0 R 1 = r R+ 1 ,
2
1
2
L 1 S 1 = S+ 1 .
(2.23)
Set
1
r2 R+ 1 = P+ 1 ,
2
R 1 = P 1 ,
(2.24)
Eqs. (2.23) reduce to

1
r2 D0+ P+ 1 = P 1 ,
2
2 L 1 S+ 1 = S 1 ,
2
r2 D0 P 1 = P+ 1 ,
2
(2.25)
2 L+1 S 1 = S+ 1 .
2
(2.26)
Introducing the tortoise coordinate transformation from the radial variable r to the tortoise coordinate r which is given by
d
r d
= 2 ,
dr dr
(2.27)
where
2 = r 2 + a 2
We set
2
D0 =
r
am
.

d
+ i ,
dr
(2.28)
(2.29)
where
= ,
and
D0+
2
=
r
(2.30)

d
i .
dr
(2.31)
Eqs. (2.25) is reduced to

1

r2
d
i P+ 1 = 2 P 1 ,
2
2
dr
(2.32)
1

d
r2
+ i P 1 = 2 P+ 1 .
2
2
dr
353
(2.33)
By setting
Z = P+ 1 P 1 ,
2
(2.34)
Eqs. (2.32) and (2.33) changed to be

1
r2
d
2 Z+ = i Z ,
dr
1

r2
d
+ 2 Z = i Z+ .
dr
From Eqs. (2.35) and (2.36), we obtain the radial wave equation

2
d
2
Z = V Z ,
+
dr2
(2.35)
(2.36)
(2.37)
where
V = 2 W 2
dW
,
dr
r2
W= 2.
(2.38)
in Eqs. (2.23)(2.38) is a separation constant and is the QNMs of the black hole. We
derived the expression of at the limit of small cosmological constant and slow rotating
black hole in Appendix A. It can be written as a function of l (or j ), m, a and .
3. Properties of massless Dirac field effective potential

From the Schrdinger-like equations in Eq. (2.37), we can evaluate the QNMs. The
form of V+ and V shown in Eq. (2.38) are super-symmetric partners derived from the
same super-potential W [12]. Ref. [13] has proved that potentials related in this ways have
the same spectral of QNMs. Thus we deal with Eq. (2.37) with potential V+ in evaluating
the QNMs. The effective potential also depends on .
The QNMs are decided by the effective potential. Here we analyze the dependence
of the effective potential on parameters m, l, Q, , a and . The effective potential as
a function of r is plotted for some configurations of m, l, Q, , a and in Figs. 16.
From these figures, we can see that the dependence of V on l, , Q is stronger than on
m, a and . Because there exists cosmological a constant , the spacetime possesses two
horizons: the black hole horizon r = re and the cosmological horizon r = rc . While r varies
from re to rc , the effective potential V reduces to zero.
Because of the rotation of the Kerr black hole, the azimuthal degeneracy of magnetic
quantum number m is destroyed. On Fig. 1, we show the dependence of effective potential
354
Fig. 1. Variation of the effective potential for massless neutrino with = 0.01, = 1, a = 0.1, Q = 0.1, l = 5,
j = 92 , m = 92 , 32 , 32 , 92 .
Fig. 2. Variation of the effective potential for massless neutrino with = 0.01, = 1, a = 0.1, Q = 0.1, m = 12 ,
l = 1, 2, 3, 4, 5, j = l 1/2.
on m, and find that negative m will increase the maximum values of the effective potential
and decrease the position of the peak.
In Fig. 2, we show the dependence of the effective potential on angular momentum
number l. It is clear the peak value and position of the potential increase with l.
355
Fig. 3. Variation of the effective potential for massless neutrino with = 0.01, = 1, a = 0.1, m = 92 , l = 5,
j = 92 , Q = 0, 0.1, 0.3, 0.6, 0.9.
Fig. 4. Variation of the effective potential for massless neutrino with = 1, a = 0.1, m = 92 , l = 5, j = 92 ,
Q = 0.1, = 0, 0.01, 0.02, 0.05, 0.1.
Fig. 3 shows the dependence of the effective potential on the electric charge Q of black
hole. The electric charge of black hole will increase the peak value, decrease the position
of the peak and change the position of black hole horizon.
356
Fig. 5. Variation of the effective potential for massless neutrino with = 0.01, = 1, Q = 0.1, m = 92 , l = 5,
j = 92 , a = 0, 0.05, 0.1, 0.15, 0.2.
Fig. 6. Variation of the effective potential for massless neutrino with = 0.01, a = 0.1, Q = 0.1, m = 92 , l = 5,
j = 92 , = 1, 10, 20, 100, 200.
Fig. 4 shows the dependence of effective potential on the cosmological constant . Increasing of reduces the peak value of the effective potential, decreases the cosmological
horizon radius rc and increases the black hole horizon radius re .
357
In Fig. 5, we show the dependence of the effective potential on a. For the positive value
of m, rotation of the black hole reduce the peak value and increase the position of the
peak.
For the limit of small a and the perturbation method used in Appendix A, the effective
of rotation to the separation constant is small. When we consider the dependence of the
effective potential on , we neglect the change of . For = 5, the Fig. 6 suggests the
potential varies slowly as increases and approaches a limiting position.
4. Massless neutrino QNMS of a KerrNewmande Sitter black hole

In this section, we evaluate the QNMs by using PshlTeller potential approximation
instead of more popular WKB approximation. Konoplya [17] used sixth-order WKB approximation to calculated the QNMs of a D-dimensional Schwarzschild black hole and
compared with the result of third-order WKB approximation. Zhidenko [18] calculated
QNMs of a Schwarzschildde Sitter black hole by using sixth-order WKB approximation
and the approximation by PshlTeller potential. The results of Refs. [17,18] show that
for large l and PshlTeller potential approximation can give results agree well with the
sixth-order WKB approximation.
Proposed by Ferrari and Mashhoon [14], the PshlTeller potential approximation
method use PshlTeller approximate potential
VPT =
V0
cosh (r /b)
2
(4.1)
It contains two free parameters (V0 and b) which are used to fit the height and the second
derivative of the potential V (r ) at the maximum

1
1 d 2V
(4.2)
=
.
2V0 dr 2 rr0
b2
The QNMs of the PshlTeller potential can be evaluated analytically:

1
1
1
i ,
=
V0 b2 n +
b
4
4
(4.3)
where n is the mode number and n < l for low-laying modes. From Eq. (4.3), the real
parts of the QNMs are independent of n. This relates to how to approximates the effective
potential.
The effective potential also depends on . This will complicate matters in Eq. (4.3)
because there are dependence on both sides of the equation. Thus we cannot obtain
directly. For slowly rotating black hole with little cosmological constant , we can expand
separation constant as series of a 1, and expand the effective potential in power series
of a.
Firstly we express the position of the peak of the effective potential as series up to order
a5,
rmax = r0 + r1 a + r2 a 2 + r3 a 3 + r4 a 4 + r5 a 5 = r0 + 0 ,
(4.4)
358
and
1
1
0 = V (rmax ) = V (r0 ) + 0 V (r0 ) + 02 V (r0 ) + 03 V (4) (r0 )
2
6
1
1
5 V (6) (r0 ),
+ 04 V (5) (r0 ) +
24
120 0
(4.5)
where r0 is the position of the peak of effective potential for the nonrotating black hole.
Eq. (2.38) for the nonrotating black hole case, show that the expression of effective potential V does not depend on and can be solved independently. We evaluate the coefficients
ri s order by solving this equation. The expression of rmax contains a and unknown . We
also expand as = 0 + 1 a + 2 a 2 + 3 a 3 + 4 a 4 + 5 a 5 and plug in the expansion for
(n)
rmax , and then expand the derivation of the potential V0 performed with respect to r . We
plug all these expansions back to Eq. (4.3) and solve the coefficients i s self-consistently
order by order in a.
Now we evaluate massless neutrino QNMs of a KerrNewmande Sitter black hole by
using PshlTeller potential approximation for = 0.01, Q = 0.1, and a = 0.1, plot the
results in Fig. 7 and list them in Table 1. The results show that the real parts of the QNMs
increase with l and the magnitude of the imaginary parts increase with n.
Figures of QNMs varying with a is plotted in Fig. 8 and plot the real and imaginary
parts of QNMs as a function of a on Fig. 9. We vary a from 0 to 0.2 to satisfy the condition
a 1. Because of the spherical symmetry of a nonrotating black hole, the QNMs is the
azimuthal degenerate which can be verified in Figs. 8 and 9. a = 0 means nonrotating case,
which is ReissnerNordstrmde Sitter black hole here. Different values of m have the
Fig. 7. QNMs of massless neutrino for = 0.01, Q = 0.1, a = 0.1.
359
Table 1
QNMs of massless neutrino for = 0.01, Q = 0.1, a = 0.1
l
n m = 12
m = 12
m = 32
m = 32
1 0 0.1851 0.1002i
0.1748 0.09909i
2 0 0.3723 0.09453i
0.3639 0.09442i
0.3792 0.09474i
0.3535 0.09454i
0.3634 0.2837i
0.3808 0.2832i
0.3521 0.2853i
1 0.3729 0.2832i
3 0 0.5561 0.09302i
1 0.5564 0.2789i
2 0.5568 0.4646i
4 0 0.7399 0.09251i
0.5481 0.09301i
0.5632 0.09310i
0.5388 0.09310i
0.5694 0.09320i
0.5478 0.2792i
0.56403 0.2789i
0.5379 0.2798i
0.5707 0.2790i
0.5475 0.4656i
0.5651 0.4640i
0.5371 0.4673i
0.5726 0.4638i
0.7320 0.09251i
0.7471 0.09254i
0.7232 0.09256i
0.7536 0.09260i
1 0.7401 0.2775i
0.7318 0.2776i
0.7476 0.2774i
0.7226 0.2779i
0.7544 0.2775i
2 0.7404 0.4623i
0.7316 0.4628i
0.7484 0.4620i
0.7219 0.4637i
0.7557 0.4618i
3 0.7407 0.6469i
5 0 0.9238 0.09228i
m = 52
0.7314 0.6482i
0.7492 0.6461i
0.7213 0.6499i
0.7572 0.6455i
0.9159 0.09228i
0.9310 0.09230i
0.9073 0.09232i
0.9377 0.09233i
1 0.9239 0.2768i
0.9158 0.2769i
0.9314 0.2768i
0.9069 0.2771i
0.9383 0.2769i
2 0.9241 0.4612i
0.9156 0.4616i
0.9320 0.4611i
0.9064 0.4621i
0.9393 0.4610i
3 0.9243 0.6456i
0.9154 0.6463i
0.9326 0.6451i
0.9058 0.6474i
0.9404 0.6447i
4 0.9246 0.8298i
0.9152 0.8312i
0.9333 0.8288i
0.9053 0.8329i
0.9415 0.8281i
m = 72
m = 72
m = 92
m = 92
0.7595 0.09267i
0.7020 0.09295i
m = 52
3 0 0.5279 0.09336i
1 0.5266 0.2811i
2 0.5256 0.4699i
4 0 0.7132 0.09269i
1 0.7123 0.2785i
0.7606 0.2777i
0.7008 0.2795i
2 0.7113 0.4651i
0.7625 0.4619i
0.6995 0.4672i
3 0.7104 0.6523i
5 0 0.8979 0.09240i
1 0.8973 0.2774i
0.7645 0.6452i
0.6987 0.6554i
0.9439 0.09238i
0.8875 0.09254i
0.9497 0.09244i
0.8761 0.09277i
0.9447 0.2770i
0.8867 0.2780i
0.9507 0.2771i
0.8750 0.2788i
2 0.8964 0.4629i
0.9460 0.4610i
0.8855 0.4641i
0.9523 0.4611i
0.8736 0.4658i
3 0.8955 0.6489i
0.9476 0.6445i
0.8844 0.6509i
0.9543 0.6445i
0.8725 0.6535i
4 0.8948 0.8351i
0.9492 0.8275i
0.8836 0.8379i
0.9564 0.8272i
0.8717 0.8414i
same QNMs in Figs. 8 and 9. They also clearly display the split of azimuthal degeneracy
as a increasing from 0 to 0.2.
From Fig. 9, we see that for a rotating black hole, the real parts of QNMs is also related
to the n though we use the PshlTeller potential approximation. For a slowly rotating
KerrNewmande Sitter black hole with little value of cosmological constant , the separation constant can be written as Eq. (A.17) and the function of which relates to n.
As to the increasing of a, the real parts of QNMs increase for positive m and decrease for
negative m, while the split of real parts for negative m is bigger than the positive case with
the same magnitude of m. The split of real parts for n increase with a and the magnitude of
360
Fig. 8. QNMs of massless neutrino vary with a for = 0.01, Q = 0.1, l = 3, j = 52 , m = 52 , 32 , 12 .
m. n increases the real parts for the positive m and decreases the imaginary parts. The real
parts split of n for positive m is bigger than imaginary case. For the negative m, the larger
magnitude of m change the imaginary parts more. The rotation increases the magnitude of
imaginary parts and the split of different values of m. For the positive m case, it is more
complex but the Fig. 9 shows that the imaginary parts of different positive m for different
n trend to the same values and the tendency is more clearly for large n.
We plot the image of QNMs varying with Q in Fig. 10 and plot the real and imaginary
parts of QNMs as a function of Q on Fig. 11. The real parts of QNMs increase with Q, the
split of different values of n decrease first and increase later. For example, while m = 52
the real parts of n = 0 is larger than n = 2 for Q = 0 and the other way round for Q = 0.9.
The split of imaginary parts of different m for the same n increase with Q.
In Figs. 12 and 13 we plot image of QNMs varying with cosmological constant and
the real and imaginary parts of QNMs as a function of . Fig. 4 shows that influence the
effective potential more than other parameters. The real parts of QNMs and the magnitude
of imaginary parts decrease with . Just like varying with Q the split of different values of
n decrease first and increase later. The split of imaginary parts of different m for the same
n increase with Q. For a sufficient large , the imaginary parts split of different m will be
larger than that of different n, which means that in the limit of the near extreme term for
a slowly rotating black hole, the imaginary parts of the QNMs are mostly determined by
m rather than n. This is why in Fig. 12 the lines of m = 52 and m = 52 seem to approach
two dots for different m.
361
Fig. 9. Real and imaginary parts of the massless neutrino QNMs as a function of a for = 0.01, Q = 0.1, l = 3,
j = 52 , m = 52 , 32 , 12 .
362
Fig. 9. (continued)
363
Fig. 10. QNMs of massless neutrino vary with Q for = 0.01, a = 0.1, l = 3, j = 52 , m = 52 .
5. Conclusions and discussions

We have evaluated low-laying massless neutrino QNMs of a KerrNewmande Sitter
black hole by using PshlTeller potential approximation. We adopt a further approximation by making perturbative expansions for all the quantities in powers of parameter a.
In general, the real parts of QNMs increase with l. The magnitude of the imaginary parts
increase with n. A character feature of the QNMs of rotating black holes is the split of the
azimuthal degeneracy for different values m. This is clearly displayed in Figs. 79. For a
KerrNewmande Sitter black hole, massless neutrino perturbation of large , positive m
and small value of n will decay slowly.
We can expand this methods to others black holes. All these works will enrich our
knowledge about QNMs of different kind of black holes and give direct identification to
distinguish the kind of black holes, through gravitational wave detectors in future.
Acknowledgements
The work has been supported by the National Natural Science Foundation of China
(Grant No. 10273017). J.F. Chang thanks Dr. Xian-Hui Ge for his zealous help during the
work.
364
Fig. 11. Real and imaginary parts of the massless neutrino QNMs as a function of Q for = 0.01, a = 0.1, l = 3,
j = 52 , m = 52 .
365
Fig. 12. QNMs of massless neutrino vary with for Q = 0.1, a = 0.1, l = 3, j = 52 , m = 52 .
Appendix A. Angular eigenfunctions and eigenvalues

The angular equations (2.26) can be combined into
1
1

2 L 1 2 L+1 S 1 + 2 S 1 = 0,
2
(A.1)
and a similar equation for S+ 1 obtained by replacing by . For small value of

2
cosmological constant , we expand Eq. (A.1) as series of , say by first-order expansion
and

1 2
a cos2 D0 + D2 + 2 S 1 = 0,
D 0 + D1 +
(A.2)
2
3
where
(m 12 cos2 )2 1
d2
d
+
cot
,
d
2
d 2
sin2
2 2
2
D1 2am a cos a sin ,
2
2
d
1
2
a 2 m2 a 3 cos a 4 2 sin2
D2 a 2 cos sin
3
d
3
3
3
1 2 2
2 3
1 4 2 2
4 3
2
+ a m + a sin a m cos + a sin cos2
3
6
3
3
1 3
1 2
2
2
a cos sin a cos .
3
2
D0
(A.3)
366
Fig. 13. Real and imaginary parts of the massless neutrino QNMs as a function of for Q = 0.1, a = 0.1, l = 3,
j = 52 , m = 52 .
The operator D0 has no dependence on a and the solution for S 1 of equation

2
D0 S 1 = E S 1
2
(A.4)
367
can be written in terms of the standard spin-weighted spherical harmonics (Newman and
Penrose, 1966 [27]; Goldberg et al., 1967 [28])

1 2
im
2
= 1 Yj m (, ),
E = j+
,
S 1 (cos )e
(A.5)
2
2
2
with positive integer l. In general, the function s Yj m (, ) are defined in
where j = (2l1)
2
the following ways:
1

2j + 1 2 j
Dsm (, , 0)
s Yj m (, ) =
4
1

2j im
(2j + 1) (j + m)! (j m)! 2
sin
=
e
4
(j + s)! (j s)!
2

j s j + s
2n+sm
(1)j sn cot
,
(A.6)
n
n+s m
2
n
j
where Dsm (, , 0) are the elements of the matrix representations of the rotation group
which satisfy
j
D11 m1 (, , )D22 m2 (, , )

j1 j2 1 2 |j 1 + 2 j1 j2 m1 m2 |j m1 + m2
=
j
j
D1 +2 ,m1 +m2 (, , ),
j
d D11 m1 (, , )D22 m2 (, , ) =
j
(A.7)
8 2
j j m m ,
2j1 + 1 1 2 1 2 1 2
where j1 j2 m1 m2 |j m1 + m2 are the ClebschGordon coefficients and
j = j1 + j2 , j1 + j2 1, . . . , |j1 j2 |,
2 2

d d d sin d.
0
(A.8)
(A.9)
(A.10)
The function s Yj m (, ) satisfies the parity and the orthogonality relations
s+m
s Yj m (, ),
s Yj m (, ) = (1)
s Yj m (, )s Yj m (, ) d
= j m j m ,
(A.11)
(A.12)
2
d =
sin d d.
(A.13)
=0 =0
With the customary definition,

1

(2j + 1) (j m)! 2
im
Y
(,
)
=
,
s jm
s Pj m ( )e
4 (j + m)!
(A.14)
368
the operators L+
s and L+s , for a = 0, are raising and lowering operators:

1
( m csc s cot )s Pj m = (j s)(j + s + 1) 2 s+1 Pj m ,
(A.15)

1
( + m csc + s cot )s Pj m = + (j + s)(j s + 1) 2 s1 Pj m .
(A.16)
For small values of a and , we can view H = D1 + ( 13 a 2 cos2 D0 + D2 ) as a

perturbation operator and obtain the results by using ordinary perturbation theory, say by
the first-order expansion

1
1 2
1
1 2
2
2
a cos D0 + D2 j m + , (A.17)
= j+
j mD1 +
2
2
3
2
where

1
1
1 2
2
j mD1 +
a cos D0 + D2 j m
2
3
2

1 2
2
a cos D0 + D2 1 Yj m d
1 Yj m D1 +
2
2
3

4 3
2 2 2
1 1
= 2am +
a m a m j m j m
3
3
2
2

1

2 4 2 1 2
1
a 2 2 +
a a j m sin2 j m
3
6
2
2

2
1 2
2 3
1 2
1
1 2 1
+ a j+
a m a j m cos j m
3
2
3
2
2
2

1
1
1
a + a 3 j m cos j m
3
2
2

1
2 2
1

a j m cos sin j m
3
2
2

1 4 2
1 2
2 1
+ a j m sin cos j m
3
2
2

1
1
1
a 3 j m cos sin2 j m .
3
2
2

1
1
1 1
m
1
j m cos j m = j 1m0|j m j 1 0j =
,
2
2
2
2
2 j (j + 1)

1
1
1 2
1 1
j m cos2 j m = + j 2m0|j m j 2 0j
2
2
3 3
2
2
(A.18)
(A.19)
1 2 [3m2 j (j + 1)][ 34 j (j + 1)]

+
,
3 3
(2j 1)j (j + 1)(2j + 3)
(A.20)
(A.21)
369

1
1
2 2
1 1
j m sin2 j m = j 2m0|j m j 2 0j
2
2
3 3
2
2
2 2 [3m2 j (j + 1)][ 34 j (j + 1)]
,
3 3
(2j 1)j (j + 1)(2j + 3)

1 2
2 1
j m cos sin j m
2
2

2
1 1
8
1 1
2
+ j 2m0|j m j 2 0j j 4m0|j m j 4 0j
=
15 21
2
2
35
2
2
=
(A.22)
2 [3m2 j (j + 1)][ 34 j (j + 1)]

8 (2j + 1)(2j 4)!
2
+
15 21
(2j 1)j (j + 1)(2j + 3)
35
(2j + 5)!

(j + m)(j + m 1) (j + m 2)(6j 8m + 3) 9(j m)(j 3m)

(j m)(j m 1) 9(j + m)(j + 3m) (j m 2)(6j + 8m + 3)

1
3
1
3
1
j
j
(6j 1) 9 j
j
j+
2
2
2
2
2

1
3
1
3
5
j
(A.23)
j
9 j+
j+
j
(6j + 7) ,
2
2
2
2
2

1
2 1
j m cos sin j m
2
2

2
1 1
2
1 1
= j 1m0|j m j 1 0j j 3m0|j m j 3 0j
5
2
2
5
2
2
=
1
8 (2j + 1)(2j 3)!
m
5 j (j + 1) 5
(2j + 4)!

(j + m)(j + m 1)(4j 5m + 1) (j m)(j m 1)(4j + 5m + 1)

1
3
1
3
7
1
j
4j
j
j
4j +
, (A.24)
j+
2
2
2
2
2
2

1
1
j m sin cos j m
2
2

1
2
1
1 1
j+
=
j 2m0|j m j 2 1j
2
2
3
2
2

1
1
3
3
2
j
j+
j 2m0|j m j 2 1j
+
3
2
2
2
2
=
3m2 j (j + 1)
.
4j (j + 1)
(A.25)
370
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
K. Kokkotas, B. Schmidt, Living Rev. Relativ. 2 (1999) 2.

H.P. Nollert, Class. Quantum Grav. 16 (1999) R159.
D. Birmingham, I. Sachs, S.N. Solodukhin, Phys. Rev. Lett. 88 (2002) 151301.
V. Cardoso, R. Konoplya, J.P.S. Lemos, Phys. Rev. D 68 (2003) 044024.
P. Brady, C. Chambers, W. Krivan, P. Laguna, Phys. Rev. D 55 (1997) 7538.
P. Brady, C. Chambers, W. Laarakkers, E. Poison, Phys. Rev. D 60 (1999) 064003.
D.T. Horowitz, V.E. Hubeny, Phys. Rev. D 62 (2000) 024027.
V. Cardoso, J.P. Lemos, Phys. Rev. D 63 (2001) 124015.
I.G. Moss, J.P. Norman, Class. Quantum Grav. 19 (2002) 2323.
S. Chandrasekhar, The Relativistic Quantum Mechanics, Clarendon, Oxford, 1983.
F. Cooper, A. Khare, U. Sukhatme, Phys. Rep. 251 (1995) 267.
A. Anderson, R.H. Price, Phys. Rev. D 43 (1991) 3147.
V. Ferrari, B. Mashhoon, Phys. Rev. D 30 (1984) 295.
B.F. Schutz, C.M. Will, Astrophys. J. 291 (1985) L33.
S. Iyer, C.M. Will, Phys. Rev. D 35 (1987) 3621.
R.A. Konoplya, Phys. Rev. D 68 (2003) 024018.
A. Zhidenko, Class. Quantum Grav. 21 (2004) 273.
V. Cardoso, J. Natario, R. Schiappa, J. Math. Phys. 45 (2004) 4698.
S. Yoshida, T. Futamase, Phys. Rev. D 69 (2004) 064025.
E. Leaver, Proc. R. Soc. London A 402 (1985) 292.
E. Seidel, S. Iyer, Phys. Rev. D 41 (1990) 374.
E. Berti, V. Cardoso, K. Kokkotas, H. Onozawa, Phys. Rev. D 68 (2003) 124018.
E. Berti, V. Cardoso, S. Yoshida, Phys. Rev. D 69 (2004) 124018.
H.T. Cho, Phys. Rev. D 68 (2003) 024003.
E. Newman, R. Penrose, J. Math. Phys. 7 (1966) 863.
J.N. Goldberg, A.J. Macfarlane, E.T. Newman, F. Rohrlich, E.C.G. Sudarsan, J. Math. Phys. 8 (1966) 2155.
Giant gravitons and fuzzy CP2

Bert Janssen a , Yolanda Lozano b , Diego Rodrguez-Gmez b,c
a Departamento de Fsica Terica y del Cosmos and Centro Andaluz de Fsica de Partculas Elementales,
Universidad de Granada, 18071 Granada, Spain

b Departamento de Fsica, Universidad de Oviedo, Avda. Calvo Sotelo 18, 33007 Oviedo, Spain
c Queen Mary, University of London, Mile end Road, London E1 4NS, UK
Abstract
In this article we describe the giant graviton configurations in AdSm S n backgrounds that involve
5-spheres, namely, the giant graviton in AdS4 S 7 and the dual giant graviton in AdS7 S 4 , in terms
of dielectric gravitational waves. Thus, we conclude the programme initiated in [hep-th/02071990]
and pursued in [hep-th/0303183] and [hep-th/0406148] towards the microscopical description of
giant gravitons in AdSm S n spacetimes. In our construction the gravitational waves expand due to
Myers dielectric effect onto fuzzy 5-spheres which are described as S 1 bundles over fuzzy CP2 .
These fuzzy manifolds appear as solutions of the matrix model that comes up as the action for Mtheory gravitational waves. The validity of our description is checked by confirming the agreement
with the Abelian description in terms of a spherical M5-brane when the number of waves goes to
infinity.
PACS: 11.25.-w
1. Introduction
It is by now well known that giant gravitons can be described microscopically in terms
of dielectric [1] gravitational waves [26]. In AdSm S n spacetimes the gravitational waves
E-mail addresses: bjanssen@ugr.es (B. Janssen), yolanda@string1.ciencias.uniovi.es (Y. Lozano),
diego@fisi35.ciencias.uniovi.es (D. Rodrguez-Gmez).
doi:10.1016/j.nuclphysb.2005.01.039
372
B. Janssen et al. / Nuclear Physics B 712 (2005) 371391
expand into a fuzzy S (n2) brane included in S n , for the genuine giant graviton, or into a
fuzzy S (m2) brane included in AdSm , for the dual giant graviton, both carrying angular
momentum on the spherical part of the geometry. For the giant graviton in AdS7 S 4 and
the dual giant graviton in AdS4 S 7 these fuzzy spheres are ordinary non-commutative S 2
[4], whereas for both the giant and dual giant gravitons in AdS5 S 5 the corresponding
fuzzy 3-spheres are defined as S 1 bundles over non-commutative S 2 base manifolds [5].
In all cases perfect agreement is found, when the number of waves goes to infinity, with the
(Abelian) description in [79] in terms of spherical test branes. This agreement provides
the strongest support to the (non-Abelian) dielectric constructions of [4,5].
The key point in the non-Abelian description of giant gravitons is the construction of
the action for the system of coincident waves, and the identification of the dielectric and
magnetic moment couplings responsible of their expansion. Given that AdSm S n is not
weakly curved one cannot use linearised matrix theory results as those in [10,11].
The action appropriate to describe coincident gravitational waves in non-weakly curved
M-theory backgrounds was constructed in [4]. This action contains dielectric and magnetic
moment couplings to the 3-form potential of eleven-dimensional supergravity, which are
responsible for the expansion of the waves into a dielectric or magnetic moment M2-brane
with the topology of a fuzzy S 2 , which constitute, respectively, the dual giant and giant
graviton configurations in the AdS4 S 7 and AdS7 S 4 backgrounds. Using this action
one can also describe coincident type IIB gravitational waves in non-weakly curved AdS
type backgrounds, after reduction and T-duality [5]. In the type IIB action the T-duality
direction occurs as an special isometric direction, and this turns out to be essential in the
construction of the fuzzy manifold associated to both giant and dual giant graviton configurations in AdS5 S 5 . Describing the fuzzy 3-sphere as an S 1 bundle over a fuzzy
2-sphere, the isometric direction is precisely the coordinate along the fibre. This same action was later used in [6] to describe the giant graviton solutions in the AdS3 S 3 T 4
type IIB background in terms of expanding waves. In this case the waves expand into a
fuzzy cylinder whose basis is contained either in S 3 or in AdS3 . In all these cases the
agreement between these descriptions and the Abelian descriptions of [79] for large number of waves provides the strongest support for the validity of the non-Abelian actions for
coincident waves.1
In this paper we would like to complete the programme initiated in [4], and pursued in
[5] and [6], with the microscopical description of the giant graviton in AdS4 S 7 and the
dual giant graviton in AdS7 S 4 . One expects that microscopically these gravitons will
be described in terms of a magnetic moment or dielectric M5-brane with the topology of
a fuzzy 5-sphere. Fuzzy S n with n > 2 are however quite complicated technically. The
general strategy is to identify them as subspaces of suitable spaces (spaces which admit
a symplectic structure) and to introduce conditions to restrict the functions to be on the
sphere [1214]. We will see that in our construction the fuzzy S 5 is simply defined as an
S 1 bundle over a fuzzy CP2 , in very much the same way the fuzzy S 3 in [5] was defined
as an S 1 bundle over a fuzzy S 2 . Again the coordinate along the S 1 fibre is, as in the fuzzy
1 Of course, together with the fact that the action for M-theory waves reduces to Myers action for D0-branes
when they propagate along the eleventh direction. This was in fact the key ingredient in [4] in order to extend the
linearised matrix theory result to more general backgrounds. See this reference for more details.
373
S 3 of [5], an isometric direction present in the action describing the waves. The existence
of this direction is, moreover, crucial in order to construct the right dielectric couplings
that will cause the expansion of the waves. As we will discuss the expanded 5-brane will
be a longitudinal brane wrapped on this direction.
The fuzzy CP2 has been extensively studied in the literature (see, for instance, [15
22]). In the context of Myers dielectric effect it was studied in [23,24]. CP2 is the coset
manifold SU(3)/U (2). G/H coset manifolds can be described as fuzzy surfaces if H
is the isotropy group of the lowest weight state of a given irreducible representation of
G [23,25]. When G = SU(2) all lowest weight vectors have isotropy group U (1), and
therefore one describes a fuzzy SU(2)/U (1), i.e., a fuzzy S 2 for any choice of irreducible
representation. In the case of SU(3) irreducible representations can be parametrised by
two integers (n, m), corresponding to the number of fundamental and anti-fundamental
indices. The lowest weight vector in the representations (n, 0) or (0, n) has isotropy group
U (2), whereas for any other irreducible representation it has isotropy group U (1) U (1).
Therefore, choosing a (n, 0) or a (0, n) irreducible representation one describes a fuzzy
CP2 .
We will use this result to describe the fuzzy S 5 , associated to the giant graviton in
AdS4 S 7 and the dual giant graviton in AdS7 S 4 , as an S 1 bundle over a fuzzy CP2
base manifold. We will then use the action for coincident M-theory gravitational waves to
find the corresponding ground state configuration. A key ingredient in this construction is
the identification in the action of the direction along the S 1 bundle. For this purpose we
will start in Section 2 by recalling some properties of the action for coincident M-theory
gravitational waves constructed in [4]. Then in Section 3 we will use this action to describe
microscopically the giant graviton in AdS4 S 7 . We will see that the corresponding macroscopical description is in terms of a longitudinal M5-brane with S 5 topology. In Section 4
we present the analogous description for the dual giant graviton in AdS7 S 4 . In both cases
we show the explicit agreement with the macroscopical description in [7,8] for large number of gravitons. In Section 5 we present our conclusions, where we discuss the connection
between our solution and other 5-brane solutions to matrix theory actions previously found
in the literature, as well as the supersymmetry properties of our configurations.
2. The action for M-theory gravitational waves

The action for coincident M-theory gravitational waves constructed in [4] is given by:
S = S BI + S CS
(2.1)
with BI action given by

i
BI
S = T0 d STr k 1 P E00 + E0i Q1 k E kj Ej 0 det Q ,
(2.2)
where

E = G + k 1 ik C (3) ,
G = g
k k
,
k2
(2.3)
374
and

Qij = ji + ik X i , X k Ekj ,
and CS action given by

1

S CS = T0 d STr P k 2 k (1) + iP (iX iX )C (3) + P (iX iX )2 ik C (6)
2

i
P (iX iX )3 ik N (8) ,
6
(2.4)
(2.5)
where ik N (8) denotes the KaluzaKlein monopole potential [26]. In this action k is an
Abelian Killing vector that points on the direction of propagation of the waves. This direction is isometric, because the background fields are either contracted with the Killing
vector, so that any component along the isometric direction of the contracted field vanishes, or pulled back in the worldvolume with covariant derivatives relative to the isometry
(see [4] for their explicit definition).2 To understand why this is so we need to recall the
construction of this action.
Expression (2.1) was obtained by uplifting to eleven dimensions the action for type IIA
gravitational waves derived in [27] using matrix string theory in a weakly curved background, and then going beyond the weakly curved background approximation by demanding agreement with Myers action for D0-branes when the waves propagate along
the eleventh direction.
In the action for type IIA waves the circle in which one matrix theory is compactified
in order to construct matrix string theory cannot be decompactified in the non-Abelian
case [27]. In fact, the action exhibits a U (1) isometry associated to translations along this
direction, which by construction is also the direction on which the waves propagate. A
simple way to see this is to recall that the last operation in the 911 flip involved in the
construction of matrix string theory is a T-duality from fundamental strings wound around
the 9th direction. Accordingly, in the action we find a minimal coupling to g9 /g99 , which
is the momentum operator k 2 k in adapted coordinates. Therefore, by construction, the
action (2.1) is designed to describe BPS waves with momentum charge along the compact
isometric direction. It is important to mention that in the Abelian limit, when all dielectric
couplings and U (N ) covariant derivatives3 disappear, (2.1) can be Legendre transformed
into an action in which the dependence on the isometric direction has been restored. This
action is precisely the usual action for a massless particle written in terms of an auxiliary metric (see [4] and [27] for the details), where no information remains about the
momentum charge carried by the particle.
Let us now look at the couplings to the 3-form potential of eleven-dimensional supergravity. We clearly find a dipole coupling in the CS part of the action and a magnetic
moment coupling
i k (3)
X , X ik C kj ,
(2.6)
2 The reduced metric G
appearing in (2.3) is in fact defined such that its pull-back with ordinary derivatives
equals the pull-back of g with these covariant derivatives.
3 Which are of course implicit in the pull-backs of the non-Abelian action (2.1).
375
in the BI part.4 These couplings play a crucial role in the microscopical description of the
AdS4 S 7 dual giant graviton and the AdS7 S 4 giant graviton, respectively.
Let us parametrise the AdSm S n background as

r2
dr 2
2
2
dt 2 +
ds = 1 +
+ r 2 dm2
1 + r 2 /L 2
L 2

2
,
+ L2 d 2 + cos2 d 2 + sin2 dn2
(2.7)
(m1)
Ct1 ...m2 =
r m1
g ,
L
(n1)
C1 ...n2 = an Ln1 sinn1 g ,
(2.8)
where a4 = a5 = 1, a7 = 1, i (i ) parametrise the S m2 (S n2 ) contained in AdSm (S n )

as

2
2
= d12 + sin2 1 d22 + sin2 2 + sin2 m3 dm2
dm2
(2.9)
m2
n2
(similarly for i ) and g ( g ) denotes the volume element on the unit S
(S
).
4
Consider the AdS7 S background and take r = 0, = const and time-dependent,
i.e., the ansatz for the giant graviton configuration. In this case there is a non-vanishing
3-form potential
(3)
C1 2 = L3 sin3 g
(2.10)
which (when rewritten in terms of Cartesian coordinates) clearly couples as in (2.6), given
that the gravitons carry P angular momentum, which identifies with the isometric
direction in the action (so that k = ). On the other hand, taking the AdS4 S 7 background and the dual giant graviton ansatz, = 0, r = const and time-dependent, the
non-vanishing 3-form potential is
(3)
Ct1 2 =
r3
g
L
(2.11)
which couples through (2.5). The detailed computations of the potentials associated to
these configurations were performed in [4], and perfect agreement was found for large
number of gravitons with the macroscopical calculations in [7,8].
Consider now the backgrounds in which the gravitons expand into M5-branes. Taking
the giant graviton ansatz in the AdS4 S 7 background one finds a non-vanishing 6-form
potential
(6)
C1 ...5 = L6 sin6 g .
(2.12)
Clearly this potential does not couple in the action for the system of waves if we identify
with the isometric direction, given that in the pull-back involved in

d P (iX iX )2 ik C (6) ,
(2.13)
4 Let us stress that through its contraction with the Killing vector the 3-form potential acquires the necessary
rank to couple in the BI action.
376
only is time-dependent, and this component is already taken through the interior product
with k . Similarly, the dual giant graviton ansatz in AdS7 S 4 yields
(6)
Ct1 ...5 =
r6
g ,
L
(2.14)
which again does not couple in the action, this time because the quadrupolar coupling
(2.13) has component and therefore is not of the form (2.14).
A puzzle then arises regarding the microscopical description of these giant graviton configurations in terms of dielectric gravitational waves. By analogy with other expanded configurations one would expect the gravitons to expand into fuzzy 5-spheres due
to quadrupolar electric or magnetic moment couplings to the 6-form potential. Since a
5-sphere has 5 relative dimensions with respect to a point-like object, the 6-form potential
has to be contracted as well with the Killing direction in order to be able to couple to a
one-dimensional worldvolume. However, we have seen that if this Killing direction is the
direction of propagation the only coupling of this form present in the action for M-theory
waves (2.1) vanishes for the AdS backgrounds that we want to study.
In order to find a possible solution to this puzzle let us recall the microscopical description of the giant graviton configurations in the AdS5 S 5 background [5]. These
configurations correspond microscopically to type IIB gravitational waves expanding into
S 3 D3-branes due to their dielectric or magnetic moment interaction with the 4-form RR
potential of the background. This potential is
(4)
C1 2 3 = L4 sin4 g ,
(2.15)
for the giant graviton, and
r4
(2.16)
g
L
for the dual giant graviton. Consider now that the action for type IIB waves contained the
coupling that one would naturally expect to involve the 4-form potential:

d P (iX iX )ik C (4) .
(2.17)
(4)
Ct1 2 3 =
By the same arguments above one easily checks that this coupling vanishes both for the
giant and dual giant graviton potentials.
The action describing type IIB waves constructed in [5] contains however a second
isometric direction, with Killing vector l . This direction is the direction in which one
performs the T-duality transformation that has to be made in order to obtain the action
for type IIB waves from the (reduction of the) action for M-theory waves. In the Abelian
limit one can interpret the resulting action as a dimensional reduction over the T-duality
direction, as one usually does, but this cannot be done in the non-Abelian case, in part due
to the presence of non-trivial dielectric couplings. One finds, in particular, the following
coupling in the CS part of the action

d P (iX iX )il C (4) ,
(2.18)
377
together with a magnetic moment coupling to the 4-form potential in the BI part of the
action,

i k
X , X ik il C (4) kj .
(2.19)
This isometric action for type IIB waves is therefore valid to describe waves propagating
in backgrounds which contain a U (1) isometric direction. This is the case for the AdS5 S 5
background, where the U (1) isometry is that corresponding to the translations along the
S 1 -fibre in the description of the 3-sphere (contained in S 5 (for the giant graviton) or AdS5
(for the dual giant graviton)) as an S 1 -fibre over an S 2 base manifold. In fact rewriting the
potentials (2.15) and (2.16) in adapted coordinates to this isometry it is easy to see that
the coupling (2.18) is non-vanishing for the 4-form potential associated to the dual giant
graviton and the coupling (2.19) for the one associated to the giant graviton. Indeed the
detailed computation of the corresponding non-Abelian potentials shows perfect agreement
with the macroscopical calculations in [7,9] for large number of gravitons. Let us stress
however that the right dielectric couplings that cause the expansion of the gravitons can
only be constructed in spacetimes with a U (1) isometry.
The discussion above suggests that something similar can be happening for the gravitons expanding into 5-spheres that we are considering in this article. The S 5 can similarly
be described as a U (1) bundle, in this case over the two-dimensional complex projective
space, CP2 . Therefore, there is a U (1) isometry in the background that could allow the
construction of further dielectric couplings.
However, the action that we know for M-theory waves contains only one isometric direction, that we naturally identified with the direction of propagation of the waves, since, as
we discussed, they are minimally coupled to the momentum operator in this direction. Consider instead that we identified this direction with the U (1) fibre of the U (1)-decomposition
of the 5-sphere. In this case one can see that the coupling (2.13) is non-vanishing both for
the giant and dual giant graviton potentials. We will see this in detail in the next sections,
when we write S 5 as an S 1 fibre over CP2 in adapted coordinates. Denoting the coordinate adapted to the isometry, (2.13) becomes

(6)
(6)
d X l X k X j X i Cij kl0 + Cij kl ,
(2.20)
and one easily finds that the first term is non-vanishing for the dual giant graviton whereas
the second one is non-vanishing for the giant graviton.
We then propose to use the action (2.1) to describe gravitons propagating in a spacetime
with a compact isometric direction, , with, by construction, non-vanishing momentumcharge along that direction, P , and with a non-zero velocity along a different transverse
as implied by the giant and dual giant gravitons anstze. Clearly, in order
direction, ,
to describe giant graviton configurations, which only carry momentum P , we will have
to set the momentum charge P to zero at the end of the calculation. We will see that
this calculation matches exactly the macroscopical calculation in [7,8] for large number of
gravitons, which will provide the strongest check for the validity of our proposal.
Of course, a more direct microscopical description of giant gravitons with momentum
P living in a spacetime with an isometric direction would be in terms of an effective
action containing both and as isometric directions, and with momentum charge only
378
with respect to the second one. As we have seen this type of action exists in the type IIB
theory. However, the construction of a similar type of action for M-theory waves cannot
be made based on duality arguments. One possibility would be to start from the two isometric action for type IIB waves, T-dualize and uplift to eleven dimensions. The detailed
calculation shows that in the non-Abelian case both the T-duality direction and the eleventh
dimension become isometric directions in the M-theory action. Therefore, the resulting action is adequate for the study of M-theory gravitational waves propagating in a spacetime
with three U (1) directions (the other isometric direction is the direction of propagation of
the waves). These isometries are however not present in the M-theory backgrounds that we
want to study.
Let us finally remark that an action for M-theory gravitational waves with two isometric
directions would have to be highly non-perturbative. This would be in the same spirit of
[3], where it is argued that the 5-brane cannot appear as a classical solution to the pp-wave
matrix model because the scaling of its radius with the coupling constant is more nonperturbative than the one corresponding to a classical solution. Coming back to our action,
if the direction of propagation is also isometric, we cannot have a magnetic coupling to
the 6-form potential like the second one in (2.20). Therefore the only way to find such a
magnetic coupling is in the BI action, through something like

Qij = ji + + X i , X k X l , X m ik il C (6) klmj .
(2.21)
However, quadratic couplings of this sort are by no means predicted by T-duality (plus the
uplift to M-theory).5 Clearly this is due to the fact that an action containing this kind of
couplings must be highly non-perturbative. Therefore it could only be derived la Myers
from a non-perturbative action for coincident D-branes. Although such a non-perturbative
action is known for a single brane (it is well known that worldvolume duality of the BI
vector yields the action which is valid in the strong coupling regime) it is not known for
coincident branes. Therefore, we seem to be stuck with some fundamental problem in
D-brane actions.
In the next two sections we show how the action (2.1) can be used to correctly describe
microscopically the giant graviton in AdS4 S 7 and the dual giant graviton in AdS7 S 4 .
We will compare in both cases to the corresponding macroscopical descriptions and see
that there is perfect agreement for large number of gravitons.
3. The giant graviton in AdS4 S 7

The giant graviton solution of [7] in this spacetime is in terms of an M5-brane with the
topology of a 5-sphere contained in S 7 , carrying angular momentum along the -direction
5 This is not the case for the type IIB waves, where the corresponding coupling is a dipole coupling

Qij = ji + i X i , X k ik il C (4) kj ,
neatly predicted by T-duality from the action for type IIA waves (see [5] for the details).
379
and magnetic moment with respect to the 6-form potential of the background:
(6)
C1 ...5 = L6 sin6 g ,
(3.1)
where i are the angles parametrising the unit 5-sphere in S 7 :

d52 = d12 + sin2 1 d22 + sin2 2 d32 + sin2 3 d42 + sin2 4 d52 .
(3.2)
Taking the ansatz r = 0, = constant, = ( ) in (2.7) we expect to find the gravitons

expanding into a non-commutative S 5 with radius L sin . As we have mentioned in the
previous section, describing the S 5 as a U (1) bundle over the two-dimensional complex
projective space, CP2 , it is natural to identify the isometry of the action for the coincident
gravitons with the U (1) coordinate.
3.1. S 5 as a U (1)-bundle over CP2
It is well known that the 5-sphere can be described as a U (1) bundle over the twodimensional complex projective space. Here we will review some details of this construction and introduce a convenient set of adapted coordinates to the U (1) isometry. We will
mainly follow references [28,29]. The reader is referred to those references for more details.
The unit S 5 can be represented as a submanifold of C3 with coordinates (z0 , z1 , z2 )
satisfying z 0 z0 + z 1 z1 + z 2 z2 = 1. This is invariant under zi zi ei , and CP2 is the space
of orbits under the action of this circle group. The projection of points in S 5 onto these
orbits is the U (1)-fibration of S 5 . Setting:
1 =
z1
,
z0
2 =
z2
,
z0
z0 = |z0 |ei ,
(3.3)
and defining:
A=
1

i
1 + |1 |2 + |2 |2
1 d1 + 2 d2 c.c. ,
2
(3.4)
the metric on the S 5 may be written as

1

d52 = (d A)2 + 1 +
|k |2
|di |2
k
2

1+
|k |2
i j d i dj
k
= (d A)
i,j
2
+ dsCP
2,
(3.5)
since CP2 is the projection orthogonal to the vector / . ds 2

2
for CP :
CP2
is the FubiniStudy metric
380

2
dsCP
2
= 1+
|k |
2K
i j

|di | 1 +
2
|k |
i j d i dj
i,j
d i d j ,
(3.6)
with K = log(1 + | 1 |2 + | 2 |2 ). Therefore CP2 has a Khler structure with Khler form
J = i K.
The field strength F = dA = 2J is a solution of Maxwells equations, the
so-called
electromagnetic
instanton of [30]. It is self-dual and satisfies that its integral

F F associated with the second Chern class is equal to 4 2 . This solution to Maxwells
equations will in fact play a role in the macroscopical description of giant gravitons of
Sections 3.4 and 4.1.
One can obtain a real four dimensional metric on CP2 by defining coordinates
(1 , 2 , , 3 ), 0 1 /2, 0 2 , 0 4 , 0 3 2 , as [29]:
2 i(+3 )/2
e
,
2
2
2 = tan 1 sin ei(3 )/2 ,
2
1 = tan 1 cos
(3.7)
to give
2
2
dsCP
2 = d1 +

1 2 2
sin 1 cos 1 (d + cos 2 d3 )2 + d22 + sin2 2 d32 .
4
(3.8)
In these coordinates the connection A defined in (3.4) is given by

1
A = sin2 1 (d + cos 2 d3 )
2
(3.9)
and the 6-form potential of the background reads:

1
(6)
C
= L6 sin6 sin3 1 sin 2 cos 1 .
1 2 3
8
(3.10)
Clearly, the background is isometric in the -direction, and this is the direction that we
are going to identify with the isometric direction in the action (2.1), i.e., k = .
Let us now make the non-commutative ansatz for the 5-sphere. Inspired by the results
of [5] for AdS5 S 5 , where it was found that the non-commutative manifold on which the
giant (and dual giant) graviton expands is defined as an S 1 bundle over a non-commutative
S 2 , we make the ansatz that the non-commutative manifold onto which the giant graviton
in AdS4 S 7 expands is defined as an S 1 bundle over a non-commutative CP2 .
3.2. The fuzzy CP2
In this subsection we review some basic properties about the fuzzy CP2 . The fuzzy CP2
has been extensively studied in the literature (see, for instance, [1523]). In this section we
will mainly follow the notation in [17].
381
CP2 is the coset manifold SU(3)/U (2), and can be defined as the submanifold of R8
determined by the constraints:
8

x i x i = 1,
i=1
1
d ij k x j x k = x i ,
3
(3.11)
where d ij k are the components of the totally symmetric SU(3)-invariant tensor defined by

2
i j = ij + d ij k + if ij k k ,
3
(3.12)
where i , i = 1, . . . , 8, are the Gell-Mann matrices. In this set of constraints only four are
independent (the first one, for instance, is a consequence of the rest), therefore they define
a four-dimensional manifold.
A matrix level definition of the fuzzy CP2 can be obtained by imposing the conditions
(3.11) at the level of matrices. Defining a set of coordinates X i , i = 1, . . . , 8, as
1
Ti
Xi =
CN
(3.13)
with T i the generators of SU(3) in an N -dimensional irreducible representation and CN the

quadratic Casimir of SU(3) in this representation, the first constraint in (3.11) is trivially
satisfied through the quadratic Casimir of the group
8

i=1
Xi Xi =
8
1 i i
1
T T =
CN = 1,
CN
CN
(3.14)
i=1
whereas the rest of the constraints are satisfied for any n (see Appendix A in [17]) if the
X i s are taken in the (n, 0) or (0, n) representations of SU(3), parametrising the irreducible
representations of SU(3) by two integers (n, m) corresponding to the number of fundamental and anti-fundamental indices. They become at the level of matrices
n/3 + 1/2 i
X.
d ij k X j X k =
1 2
n
+
n
3
(3.15)
Therefore, to describe the fuzzy CP2 the non-commuting coordinates X i have to be

taken in the (n, 0) or (0, n) irreducible representations of SU(3). This is in agreement
with the fact that G/H cosets can be made fuzzy if H is the isotropy group of the lowest weight state of a given irreducible representation of G [23,25]. Therefore, different
irreducible representations, having associated different isotropy subgroups, can give rise
to different cosets G/H . CP2 has G = SU(3) and H = U (2), and this is precisely the
isotropy subgroup of the SU(3) irreducible representations (n, 0) and (0, n). Any other
choice of (n, m) has isotropy group U (1) U (1), and therefore yields to a different coset,
SU(3)/(U (1) U (1)).
It will be useful later to know that the irreducible representations (n, 0), (0, n) have
dimension N given by
1
N = (n + 1)(n + 2)
2
(3.16)
382
and quadratic Casimir

1
CN = n2 + n.
(3.17)
3
CP2 can be embedded in R8 using coherent state techniques [31] (see also Appendix B
in [23]). In our coordinates (3.8) we have:
1
2
+ 3
sin 21 cos
cos
,
2
2
2
1
2
+ 3
X 2 = sin 21 cos
sin
,
2
2

2
2
1
1 ,
X 3 = sin2 1 1 + cos2
2
2
3
1
2
X 4 = sin 21 sin
cos
,
2
2
2
3
1
2
X 5 = sin 21 sin
sin
,
2
2
2
1
X 6 = sin2 1 sin 2 cos 3 ,
2
1
7
X = sin2 1 sin 2 sin 3 ,
2

1
8
2
2 2
1 ,
X = 3 sin 1 sin
2
2 3
X1 =
(3.18)
for which
8

i 2

1
dX = d12 + sin2 1 cos2 1 (d + cos 2 d3 )2 + d22 + sin2 2 d32
4
i=1
2
= dsCP
2.
(3.19)
We also have that

8

i 2 1
X = ,
3
(3.20)
i=1
so that in order to fulfill this constraint we will have to slightly modify our definition (3.13).
We then get, in the (n, 0) or (0, n) representations
1
1
Xi =
T i,
Ti =
3 CN
n2 + 3n
i = 1, . . . , 8.
With this normalisation the commutation relations of the X i become

i j
i
f ij k X k ,
X ,X =
2
n + 3n
with f ij k the structure constants of SU(3) in the notation
i j
, = 2if ij k k .
(3.21)
(3.22)
(3.23)
383
3.3. The microscopical description

Let us now take the giant graviton ansatz, r = 0, = const, = ( ) in the AdS4 S 7
background. We find, in Cartesian coordinates
ds 2 = dt 2 + L2 cos2 d 2
2
2

2

+ L2 sin2 d A + dX 1 + + dX 8
(3.24)
Cij kl = 2L6 sin6 f [ij m f kl]n X m X n .
(3.25)
and
(6)
Taking now
k =
in the action (2.1) we have that
E00 = 1 + L2 cos2 2 ,
L3 sin3 ij k k
f X , i, j = 1, . . . , 8,
Qij = ji
n2 + 3n
and substituting in the action we find

1
BI
S = T0 d STr
1 L2 cos2 2
L sin

3 L6 sin6 2
9 L12 sin12 2 2
X +
1+
X X + .
2 n2 + 3n
16 (n2 + 3n)2
k = L sin ,
(3.26)
(3.27)
Here we have dropped those contributions to det Q that will vanish when taking the symmetrised trace, and ignored higher powers of n2 + 3n which will vanish in the large N
( n ) limit. These terms on the other hand cannot be nicely arranged into higher
powers of the quadratic Casimir without explicit use of the constraints (3.15). Up to order
n4 we have that

3 L6 sin6 2
9 L12 sin12 2 2
3 L6 sin6 2
(3.28)
X
X .
1+
+
X
X
+
=
1
+
2 n2 + 3n
16 (n2 + 3n)2
4 n2 + 3n
We also have for the CS part of the action:

N T0 L6 sin6
T0
CS
2
(6)
= d
S =
(3.29)
d STr P (iX iX ) ik C
.
2
4 n2 + 3n
It is then easy to compute the symmetrised trace to finally arrive, in Hamiltonian formalism, to

2
N 2 T02
P
1 L6 sin6
N T0
4

2
6
1+
1 + tan 1
L sin + 2 2
H=
L
2 n2 + 3n
4(n2 + 3n)P
P sin
(3.30)
where P is a conserved quantity given that is cyclic in the Lagrangian. We should stress
that this expression is an approximated expression in which higher powers of n2 + 3n
vanishing in the large n limit have been omitted.
384
In order to describe giant graviton configurations we must set to zero the momentum
along the -direction, P . Recall however that, by construction, the N gravitons carry momentum P = N T0 . The difference between P being zero or not is merely a coordinate
transformation, a boost in . However, how to perform coordinate transformations in nonAbelian actions is an open problem [3235]. In order to clarify how the limit P 0 must
be taken we study in the next subsection the macroscopical description of this microscopical configuration, in terms of a spherical M5-brane carrying both P and P charges. The
agreement between the two descriptions in the large N limit will make clear that in the
limit P 0 the last term in the Hamiltonian should vanish. On the other hand, the term
N T0
,
4(n2 + 3n)P
(3.31)
whose presence is in fact crucial in order to find the giant graviton configuration, remains
finite, since N = (n + 1)(n + 2)/2 for (n, 0) and (0, n) irrespectively, so that numerator
and denominator in (3.31) scale with the same power of n. In fact, in the N ( n
) limit both terms are compensated, and the finite result T0 /8P is reached, in perfect
agreement with the macroscopical calculation. This compensation does not however take
place in the last term in the Hamiltonian.
Therefore, in order to describe giant graviton configurations we minimize

2
P
N T0
4
2
6
Hmic =
(3.32)
1 + tan 1
L sin
L
4(n2 + 3n)P
with respect to . We then find two solutions with energy P /L: sin = 0, which corresponds to the point-like graviton, and

sin =
4(n2 + 3n)P
N T0 L6
1/4
(3.33)
which corresponds to the giant graviton, with radius

R=
4(n2 + 3n)P
N T0 L2
1/4
.
(3.34)
Clearly, for the giant graviton

P
N T 0 L6
4(n2 + 3n)
(3.35)
which provides the microscopical bound to the angular momentum predicted by the stringy
exclusion principle. When the number of gravitons is large we find that
P
T 0 L6
= N
8
(3.36)
with N the units of 6-form flux on the 5-sphere in the macroscopical description of [7,8],
given by N = A5 T5 L6 with A5 the area of the unit 5-sphere. We therefore find perfect
385
agreement with the bound found in [7]. The same holds true for the radius of the configuration, which for large number of gravitons is given by

R=
8P
T0 L2
1/4
=L
P
N
1/4
(3.37)
exactly as in [7,8].
In this section we have achieved the right microscopical description of the giant graviton in AdS4 S 7 in terms of gravitons expanding into a fuzzy 5-sphere, defined as a
U (1) bundle over the fuzzy CP2 . We have seen that the coordinate along the fibre must
be isometric in the action, and this has forced our choice of k pointing on that direction, which has in turn introduced, by construction, a non-vanishing momentum P on
the configuration. The macroscopical description of such a configuration should then be
in terms of a spherical M5-brane carrying both P and P charges. We will perform the
detailed macroscopical description in these terms in the next subsection. The agreement
between the two descriptions will provide further justification to the limit taken to arrive at
expression (3.32) as the right microscopical Hamiltonian describing gravitons with angular
momentum P .
3.4. The macroscopical calculation
The simplest way to describe an M5-brane living in a spacetime with a U (1) isometry
and carrying momentum along that direction is by uplifting a D4-brane to M-theory keeping the eleventh direction compact. The resulting brane is therefore a longitudinal 5-brane.
Doing this we obtain the action adequate to describe an M5-brane whose worldvolume
contains an isometric direction. Clearly this applies to the M5-brane with the topology of
an S 5 . In this case the isometric direction is the coordinate along the fibre in the decomposition of the S 5 as a U (1) fibre over CP2 . The worldvolume of such a brane is therefore
effectively four-dimensional, and locally that of a CP2 .
For the non-vanishing eleven-dimensional fields involved in our AdS backgrounds we
find the following action for a wrapped M5-brane:

1

S = T4 d 5 k det G + k 1 F P ik C (6) P k 2 k (1) F F .
2
(3.38)
Here F is the field strength associated to M2-branes wrapped on the isometric direction
ending on the M5-brane, since the uplifting of the BI field strength of the D4-brane, F +
B (2) , to M-theory gives F + ik C (3) .6 G is the reduced metric defined in (2.3) and we
have denoted T4 the tension of the brane to explicitly take into account that its spatial
worldvolume is 4-dimensional.
As we discussed in Section 2 k 2 k (1) is identified with the momentum operator along
the isometric direction. Therefore we can switch on momentum charge on the M5-brane
6 In the action (3.38) we have set C (3) to zero, which is valid for our particular backgrounds.
386
by choosing a non-vanishing field strength such that

F F = 8 2 N,
(3.39)
CP2
since then

2 (1)

T4
4
(3.40)
F F = N T0 d P k 2 k (1) .
d d P k k
2
With F satisfying (3.39) we are therefore dissolving N gravitons, propagating along the
isometric direction, in the worldvolume of the 5-brane. The field satisfying this condition
is in fact the electromagnetic instanton that we discussed in Section 3.1, which now must
have instanton number equal to twice the number of gravitons. This gives

N
sin 21 d1 d sin 21 cos 2 d1 d3
F=
2

+ sin2 1 sin 2 d2 d3 .
(3.41)
Identifying the isometric direction in the action (3.38) with in (3.5) and integrating
over the spatial worldvolume of the M5-brane we arrive at

4 4

L sin
N
+
S = 4 2 T4 d L sin 1 L2 cos2 2
8
L2 sin2
1
L6 sin6
(3.42)
8
and, in Hamiltonian formalism, to

4 2
2
6
16 4 T42 N 2
P
L6 sin6
1 + tan2 1 T4 L sin
1
+
.
H=
+
L
2P
4N
P2 sin2
(3.43)
Comparing with the Hamiltonian constructed in [7,8], which describes a spherical 5-brane
with momentum P , we see that a non-vanishing momentum along the -direction translates into an additional piece depending on N inside the squared root. Comparing this
expression with the microscopical Hamiltonian (3.30) we find that they exactly agree in
the large N limit, when N n2 /2. In the macroscopical Hamiltonian (3.43) it is clearer
however that the limit of zero momentum along the fibre direction is reached when there are
zero gravitons propagating in the direction dissolved in the worldvolume of the 5-brane,
which neatly sets to zero the last term in the Hamiltonian. This further justifies the elimination of the corresponding term in the microscopical Hamiltonian (3.30) in Section 3.3.
4. The dual giant graviton in AdS7 S 4
4.1. The microscopical description
In this section we briefly describe the microscopical description of the dual giant graviton in AdS7 S 4 . One expects that microscopically the gravitons expand into a 5-sphere
387
(6)
with radius r due to the coupling to the 6-form potential Ct1 ...5 . As before, we describe
the fuzzy 5-sphere as a U (1) bundle over the fuzzy CP2 , and we embed the CP2 in R8 .
We then have for the dual giant graviton ansatz:

2
2

r2
2
ds = 1 +
dt 2 + L2 d 2 + r 2 (d A)2 + dX 1 + + dX 8 ,
2
L
6
r
(6)
C0ij kl = 2 f [ij m f kl]n X m X n .
(4.44)
L
Substituting in the action (2.1) we have that

r2
+ L2 2 ,
k = r,
E00 = 1 +
L 2
r3
Qij = ji
f ij k X k ,
n2 + 3n
and

S = T0

i, j = 1, . . . , 8
(4.45)

1
r2
1+
L2 2
d STr
r
L 2

r 12
N
3 r6
9
r6
2
2
2
X +
1+
X X +
.
2 n2 + 3n
16 (n2 + 3n)2
L 4(n2 + 3n)
(4.46)
n4
we arrive at the Hamiltonian

Up to order

2

N 2 T02 r 10
N 2 T02
P
r2
1 r6
1+
1
+
H=
+
+
2 n2 + 3n
L2
16(n2 + 3n)2
r2
L 2
r6
N T0
.
4(n2 + 3n) L
(4.47)
Now we have to take the limit P = 0 which amounts to setting to zero the last term in the
squared-root, as we discussed in detail when we studied the giant graviton in AdS4 S 7 .
We are then left with

2

N 2 T02 r 10
P
r6
r2
N T0
,
+
Hmic =
(4.48)
1+
2
2
2
2
L
16(n + 3n)
4(n + 3n) L
L 2
which again is an approximated expression in which higher order powers of n2 + 3n vanishing in the large n limit have been omitted. Minimizing with respect to r we find two
solutions with energy P /L: r = 0, which corresponds to the point-like graviton, and

r=
4(n2 + 3n)P
N T0 LL
1/4
(4.49)
388
which corresponds to the dual giant graviton. When N

8P 1/4 P 1/4
=L
r
N
T0 LL
(4.50)
in agreement with the result in [8].

4.2. The macroscopical calculation
Using the action (3.38) we can easily describe the previous configurations in terms of
an M5-brane wrapped on the -direction and carrying P and P momentum charges.
P is switched on by dissolving N gravitons with momentum on this direction in the
worldvolume of the M5-brane. We find

4

N
1 r6
r2
r
2
2
2
+ 2
,
L
1+
S = 4 T4 d r
(4.51)
8
8 L
r
L 2
and, in Hamiltonian formalism

4

2
P2
r
N 2 2 T4 r 6
r
4T 2r 2
+
.
+
16
H = 1+
4
8
2 L
r2
L 2 L2
(4.52)
Clearly the condition P = 0 is met for N = 0 gravitons dissolved in the worldvolume,

which amounts to setting to zero the N -term in the squared-root. Comparing to (4.47) this
further justifies the limit taken in that expression, yielding to the microscopical potential
(4.48). In fact, when N ( n ) we find perfect agreement between (4.52) and
the microscopical potential (4.48).
5. Conclusions
We have shown that the giant graviton configurations in AdSm S n backgrounds that
involve 5-spheres are described microscopically in terms of gravitational waves expanding into fuzzy 5-spheres which are defined as S 1 bundles over fuzzy CP2 . The explicit
construction can be done due to the fact that the action used to describe the system of coincident waves contains a special U (1) isometric direction that can be identified with the
U (1) fibre.
In this description the gravitons expand into a longitudinal M5-brane which has four
manifest dimensions and one wrapped on the U (1) direction. This brane carries quadrupolar magnetic moment with respect to the 6-form potential of the background in the
AdS4 S 7 case, or quadrupolar electric moment in the AdS7 S 4 case. In both cases
it has a non-vanishing angular momentum along the spherical part of the geometry, P .
The details of the construction show that there is as well a non-vanishing momentum along
the U (1) direction, which has to be set to zero to find the right point-like graviton and
giant graviton configurations in the background. We have seen that in that case not only
the radii of the giant gravitons but also the Hamiltonian that they minimize agree with the
macroscopical results in [7,8] for large number of gravitons.
389
Gravitational waves propagating both along the spherical part of the geometry (the
direction) and the compact U (1) direction can be described macroscopically in terms of
a longitudinal M5-brane with velocity wrapped on the isometric direction. The action
associated to this brane can be easily constructed by just uplifting the action
of the D4brane to M-theory, while maintaining the eleventh direction compact. The C (1) F F
term in the CS part of the D4-brane action is then uplifted to k 2 k (1) F F , with F now
associated to wrapped M2-branes ending on the M5-brane. Therefore, a momentum charge
along the compact direction is simply switched on by taking F with non-vanishing instanton number. The comparison between this description and our microscopical description
shows exact agreement for large number of gravitons. Moreover, this comparison can be
used to clarify the right way to set to zero the momentum along the compact direction to
finally obtain the correct giant graviton configurations.
Our action for M-theory waves, therefore, provides an explicit matrix action which
is solved by some sort of non-commutative 5-sphere. Moreover, although we have not
checked the supersymmetry properties of our configurations, the agreement with the
macroscopical description of [7,8] suggests that they should occur as BPS solutions preserving the same half of the supersymmetries as the point-like graviton [8]. To our knowledge this would be the first example of a physical matrix model, coming up as the action
for coincident M-theory gravitational waves, admitting some fuzzy 5-sphere as a supersymmetry preserving solution.
Let us stress that the fuzzy 5-sphere that we have constructed is defined as an S 1 bundle over the fuzzy CP2 , and is therefore different from previous fuzzy 5-spheres discussed
in the literature [1214]. In particular, our solution does not show SO(6) covariance, this
invariance being broken down to SU(3) U (1), whereas this is the case for the fuzzy
5-sphere in [1214]. The SO(6) invariance might still be present in a non-manifest way, after all the SO(6) covariance of the classical 5-sphere is also not explicit when it is described
as an S 1 -bundle over CP2 . Another difference is that our solution approaches neatly the
classical S 5 in the large N limit, where all the non-commutativity disappears. This is not
the case for the fuzzy 5-sphere in [1214]. Indeed, the right dependence of the radius of the
1/4
5-sphere giant graviton with P , R P , is only achieved within the S 1 bundle over CP2
description, the corresponding dependence of the fuzzy 5-sphere of [1214] being given
1/5
by R P . Another difference is that our fuzzy S 5 inherits its symplectic structure
from the Khler form of the fuzzy CP2 , whereas in the construction in [12,13] the bundle
structure corresponds to a CP3 base and a CP2 fibre. Therefore, there are clear differences
between the two constructions.
Non-supersymmetric longitudinal M5-branes with CP2 S 1 topology have been obtained as explicit solutions of matrix theory in [15]. Our longitudinal M5-branes, although
similar in the explicit construction, have S 5 topology, once the necessary twist in the fibre
is taken into account. This twist should provide the global extension of the local residual
supersymmetry found in [15], in terms of spinors charged under the gauge potential whose
field strength is the Khler form (see [29,36]).
Longitudinal 5-branes with other topologies have also been shown to arise as solutions
to matrix theory in [3740] (see also the fuzzy funnel solution in [41]). In general, to find
these solutions it is necessary to include additional ChernSimons terms or mass terms.
390
Very recently [42] there has been some speculation on how the fuzzy 5-sphere of [12,13]
might appear as a solution to the pp-wave matrix model of [3]. It would be interesting to
elucidate the relation between the new ChernSimons coupling conjectured in this reference and the dielectric couplings constructed in this paper once the Penrose limit is taken.
This would help clarifying if indeed the resulting matrix action would allow for transverse
5-brane solutions [43].
Acknowledgements
We would like to thank J.M. Figueroa-OFarrill and S. Ramgoolam for useful discussions. The work of B.J. is done as part of the program Ramon y Cajal of the MEC
(Spain). He was also partially supported by the MEC under contract FIS 2004-06823 and
by the Junta de Andaluca group FQM 101. The work of Y.L. and D.R.-G. has been partially supported by CICYT grant BFM2003-00313 (Spain). D.R.-G. was supported in part
by a FPU. Fellowship from MEC. He would like to thank the String Theory group at Queen
Mary College (London) for its hospitality while part of this work was done.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
R.C. Myers, JHEP 9912 (1999) 022, hep-th/9910053.

S.R. Das, S.P. Trivedi, S. Vaidya, JHEP 0010 (2000) 037, hep-th/0008203.
D. Berenstein, J. Maldacena, H. Nastase, JHEP 0204 (2002) 013, hep-th/0202021.
B. Janssen, Y. Lozano, Nucl. Phys. B 658 (2003) 281, hep-th/0207199.
B. Janssen, Y. Lozano, D. Rodrguez-Gmez, Nucl. Phys. B 669 (2003) 363, hep-th/0303183.
B. Janssen, Y. Lozano, D. Rodrguez-Gmez, Giant gravitons in AdS3 S 3 T 4 as fuzzy cylinders, hepth/0406148.
J. McGreevy, L. Susskind, N. Toumbas, JHEP 0006 (2000) 008, hep-th/0003075.
M.T. Grisaru, R.C. Myers, O. Tafjord, JHEP 0008 (2000) 040, hep-th/0008015.
A. Hashimoto, S. Hirano, N. Itzhaki, JHEP 0008 (2000) 051, hep-th/0008016.
W. Taylor, M. Van Raamsdonk, Nucl. Phys. B 558 (1999) 63, hep-th/9904095.
W. Taylor, M. Van Raamsdonk, Nucl. Phys. B 573 (2000) 703, hep-th/9910052.
S. Ramgoolam, Nucl. Phys. B 610 (2001) 461, hep-th/0105006.
S. Ramgoolam, JHEP 0210 (2002) 064, hep-th/0207111.
M.M. Sheikh-Jabbari, JHEP 0409 (2004) 017, hep-th/0406214.
V.P. Nair, S. Randjbar-Daemi, Nucl. Phys. B 533 (1998) 333, hep-th/9802187.
H. Grosse, A. Strohmaier, Lett. Math. Phys. 48 (1999) 163, hep-th/9902138.
G. Alexanian, A.P. Balachandran, G. Immirzi, B. Ydri, J. Geom. Phys. 42 (2002) 28, hep-th/0103023.
Y. Kitazawa, Nucl. Phys. B 642 (2002) 210, hep-th/0207115.
U. Carow-Watamura, H. Steinacker, S. Watamura, Monopole bundles over fuzzy complex projective spaces,
hep-th/0404130.
T. Azuma, S. Bal, K. Nagao, J. Nishimura, Dynamical aspects of the fuzzy CP2 in the large N reduced
model with a cubic term, hep-th/0405277.
H. Grosse, H. Steinacker, Finite gauge theory on fuzzy CP2 , hep-th/0407089.
D. Karabali, V.P. Nair, S. Randjbar-Daemi, Fuzzy spaces, the (M)atrix model and the quantum hall effect,
hep-th/0407007.
S.P. Trivedi, S. Vaidya, JHEP 0009 (2000) 041, hep-th/0007011.
G.G. Alexanian, A.P. Balachandran, P.J. Silva, Obstruction to D-brane topology change, hep-th/0207052.
391
[25] J. Madore, An Introduction to Noncommutative Differential Geometry and its Applications, Cambridge
Univ. Press, Cambridge, 1995.
[26] E. Bergshoeff, E. Eyras, Y. Lozano, Phys. Lett. B 430 (1998) 77, hep-th/9802199.
[27] B. Janssen, Y. Lozano, Nucl. Phys. B 643 (2002) 399, hep-th/0205254.
[28] G.W. Gibbons, C.N. Pope, Commun. Math. Phys. 61 (1978) 239.
[29] C.N. Pope, Phys. Lett. B 97 (1980) 417.
[30] A. Trautman, Int. J. Theor. Phys. 16 (1977) 561.
[31] A. Perelomov, Generalized Coherent States and Their Applications, Springer-Verlag, Berlin, 1986.
[32] J. de Boer, K. Schalm, JHEP 0302 (2003) 041, hep-th/0108161.
[33] S.F. Hassan, N = 1 worldsheet boundary couplings and covariance of non-Abelian worldvolume theory,
hep-th/0308201.
[34] J. de Boer, K. Schalm, J. Wijnhout, Ann. Phys. 313 (2004) 425, hep-th/0310150.
[35] D. Brecher, K. Furuuchi, H. Ling, M. Van Raamsdonk, JHEP 0406 (2004) 020, hep-th/0403289.
[36] M.J. Duff, H. L, C.N. Pope, Nucl. Phys. B 532 (1998) 181, hep-th/9803061.
[37] J. Castelino, S. Lee, W. Taylor, Nucl. Phys. B 526 (1998) 334, hep-th/9712105.
[38] P.M. Ho, JHEP 0012 (2000) 015, hep-th/0010165.
[39] Y. Kimura, Nucl. Phys. B 664 (2003) 512, hep-th/0301055.
[40] Y. Kimura, Nucl. Phys. B 637 (2002) 177, hep-th/0204256.
[41] N.R. Constable, R.C. Myers, O. Tafjord, JHEP 0106 (2001) 023, hep-th/0102080.
[42] H. Nastase, On fuzzy spheres and (M)atrix actions, hep-th/0410137.
[43] J. Maldacena, M.M. Sheikh-Jabbari, M. Van Raamsdonk, JHEP 0301 (2003) 038, hep-th/0211139.
Three-neutrino oscillations of atmospheric

neutrinos, 13, neutrino mass hierarchy and iron
magnetized detectors
Sergio Palomares-Ruiz a,b , S.T. Petcov c,1
a Department of Physics and Astronomy, UCLA, Los Angeles, CA 90095, USA
b Department of Physics and Astronomy, Vanderbilt University, Nashville, TN 37235, USA
c Scuola Internazionale Superiore di Studi Avanzati and Istituto Nazionale di Fisica Nucleare,
I-34014 Trieste, Italy

Received 12 July 2004; received in revised form 9 December 2004; accepted 27 January 2005
Abstract
We derive predictions for the Nadir angle (n ) dependence of the ratio N /N+ of the rates
of the and + multi-GeV events, and for the + event rate asymmetry, A + =
[N ( ) N (+ )]/[N ( ) + N (+ )], in iron-magnetized calorimeter detectors (MINOS, INO,
etc.) in the case of 3-neutrino oscillations of the atmospheric and , driven by one neutrino
mass squared difference, |m231 | (2.03.0) 103 eV2 m221 . The asymmetry A + (the
ratio N /N+ ) is shown to be particularly sensitive to the Earth matter effects in the atmospheric
neutrino oscillations, and thus to the values of sin2 13 and sin2 23 , 13 and 23 being the neutrino
mixing angle limited by the CHOOZ and Palo Verde experiments and that responsible for the dominant atmospheric ( ) oscillations. It is also very sensitive to the type of neutrino
mass spectrum which can be with normal (m231 > 0) or with inverted (m231 < 0) hierarchy. We
find that for sin2 23 0.50, sin2 213 0.06 and |m231 | = (23) 103 eV2 , the Earth matter
in the
effects produce a relative difference between the integrated asymmetries A + and A 2
+
mantle (cos n = 0.300.84) and core (cos n = 0.841.0) bins, which is bigger in absolute value
than approximately 15%, can reach the values of (3035)%, and thus can be sufficiently large
E-mail address: sergiopr@physics.ucla.edu (S. Palomares-Ruiz).
1 Also at: Institute of Nuclear Research and Nuclear Energy, Bulgarian Academy of Sciences, 1784 Sofia,
Bulgaria.
doi:10.1016/j.nuclphysb.2005.01.045
S. Palomares-Ruiz, S.T. Petcov / Nuclear Physics B 712 (2005) 392410
393
to be observable. The sign of the indicated asymmetry difference is anticorrelated with the sign of
m231 . An observation of the Earth matter effects in the Nadir angle distribution of the asymmetry A + (ratio N /N+ ) would clearly indicate that sin2 213 0.06 and sin2 23 0.50, and
would lead to the determination of the sign of m231 .
PACS: 14.60.Pq; 29.40.Vj
1. Introduction
There has been a remarkable progress in the studies of neutrino oscillations in the last
several years. The experiments with solar, atmospheric and reactor neutrinos [17] have
provided compelling evidences for the existence of neutrino oscillations driven by nonzero
neutrino masses and neutrino mixing. Evidences for oscillations of neutrinos were obtained
also in the first long baseline accelerator neutrino experiment K2K [8].
It was predicted already in 1967 [9] that the existence of solar e oscillations would
cause a deficit of solar e detected on Earth. The hypothesis of solar neutrino oscillations,
which in one variety or another were considered starting from the late 60s as the most
natural explanation of the observed [1,2] solar e deficit (see, e.g., Refs. [912]), has received a convincing confirmation from the measurement of the solar neutrino flux through
the neutral current reaction on deuterium by the SNO experiment [4,5], and by the first results of the KamLAND (KL) experiment [7] (see also, e.g., Ref. [13]). The analysis of the
solar neutrino data obtained by Homestake, SAGE, GALLEX/GNO, Super-Kamiokande
(SK) and SNO experiments showed (see, e.g., Ref. [4]) that the data favor the large mixing angle (LMA) MSW solution [14,15] of the solar neutrino problem. The first results
of the KL reactor experiment [7] have confirmed (under the plausible assumption of CPTinvariance) the LMA MSW solution, establishing it essentially as a unique solution of the
solar neutrino problem.
The latest contributions to these magnificent progress are the new SK data on the L/Edependence of multi-GeV -like atmospheric neutrino events [16], L and E being the
distance traveled by neutrinos and the energy, and the new more precise spectrum data
of the KL and K2K experiments [17,18]. For the first time these data exhibit directly the
effects of the oscillatory dependence on L/E and E of the probabilities of -oscillations
in vacuum [19]. We begin to see the oscillations of neutrinos.
The interpretation of the solar and atmospheric neutrino, and of KL data in terms of oscillations requires the existence of 3-neutrino mixing in the weak charged lepton current:
lL =
3
Ulj j L .
(1)
j =1
Here lL , l = e, , , are the three left-handed flavor neutrino fields, j L is the left-handed
field of the neutrino j having a mass mj and U is the PontecorvoMakiNakagawa
394
Sakata (PMNS) neutrino mixing matrix [20],
Ue1 Ue2 Ue3

U = U1 U2 U3
U 2 U 3
c12 c13
= s12 c23 c12 s23 s13 ei
s12 s23 c12 c23 s13 ei
U 1
s12 c13
c12 c23 s12 s23 s13 ei
c12 s23 s12 c23 s13 ei
s13 ei
s23 c13 ,
c23 c13
(2)
where we have used a standard parametrization of U with the usual notations, sij sin ij ,
cij cos ij , and is the Dirac CP-violation phase.2 If one identifies m221 > 0 and
m231 (or m232 ) with the neutrino mass squared differences which drive the solar and
atmospheric neutrino oscillations, the data suggest that |m231(2) | m221 . In this case
m231
= m232 , 12 and 23 , represent the neutrino mixing angles responsible for the solar
and the dominant atmospheric neutrino oscillations, 12 , 23 , while 13 is the angle limited
by the data from the CHOOZ and Palo Verde experiments [24,25].
The existing atmospheric neutrino data is essentially insensitive to 13 obeying the
CHOOZ limit [26]. The Super-Kamiokande and K2K data are best described in terms
of dominant ( ) vacuum oscillations. The best fit values and the 99.73%
C.L. allowed ranges of the atmospheric neutrino oscillation parameters read [26]:

m2 = 2.1 103 eV2 ,
sin2 223 = 1.0,
31

m2 = (1.53.4) 103 eV2 ,
sin2 223 0.85.
(3)
31
The values of |m231 | and sin2 223 , deduced from the SK analysis of the L/E dependence
of the observed -like atmospheric neutrino events [16], are comfortably compatible with
the values obtained in the analysis of the global SK atmospheric neutrino data quoted
above.
The sign of m231 and of cos 223 when sin2 223 = 1.0 cannot be determined using the
existing data. The two possibilities, m231 > 0 or m231 < 0, correspond to two different
types of neutrino mass spectrum: with normal hierarchy (NH), m1 < m2 < m3 , and with
inverted hierarchy (IH), m3 < m1 < m2 . The fact that the sign of cos 223 is not determined
when sin2 223 = 1.0 implies that when, e.g., sin2 223 = 0.92, two values of sin2 23 are
possible, sin2 23
= 0.64 or 0.36. We will use in our further discussion as illustrative the
values |m231 | = (2.0; 3.0) 103 eV2 and sin2 223 = 0.92; 1.0.
The combined 2-neutrino oscillation analysis of the solar neutrino and the new KL
766.3 Ty spectrum data shows [17,27] that the -oscillation parameters lie in the lowLMA region:

5
eV2 ,
tan2 21 = 0.40+0.09
m221 = 7.9+0.6
(4)
0.5 10
0.07 .
The value of m221 is determined with a remarkably high precision. The high-LMA solution (see, e.g., Ref. [13]) is excluded at more than 3 . Maximal -mixing is ruled out at
2 We have not written explicitly the two possible Majorana CP-violation phases [21,22] which do not enter
into the expressions for the oscillation probabilities of interest [21,23].
395
6 . The 3- oscillations of the solar e depend in the case of interest, |m231 | m221 ,
not only on m221 and 12 , but also on 13 . A combined 3- oscillation analysis of the solar neutrino, CHOOZ and KamLAND data shows [27] that for sin2 13 0.02 the allowed
ranges of the solar neutrino oscillation parameters do not differ substantially from those derived in the 2-neutrino oscillation analyzes (see, e.g., Refs. [17,28]). The best fit values, for
instance, read [27]: m221 = 8.0 105 eV2 , sin2 21 = 0.28. The best fit value of sin2 13
was found to be nonzero, sin2 13 = 0.004, but not at statistically significant level [27]. Using the 3 allowed range of |m231 | [26] and performing a combined analysis of the solar
neutrino, CHOOZ and KamLAND data, one finds [27]:
sin2 13 < 0.05,
99.73% C.L.
(5)
Similar constraint is obtained from a global 3- oscillation analysis of the data [28].
It is difficult to overestimate the importance of getting more precise information about
the value of the mixing angle 13 , of determining the sign of m231 , or the type of the
neutrino mass spectrum (with normal or inverted hierarchy), and of measuring the value of
sin2 23 with a higher precision, for the future progress in the studies of neutrino mixing3
(see, e.g., Ref. [29] and the references quoted therein).
In the present article we study possibilities to obtain information on the value of sin2 13
and on the sign of m231 using the data on atmospheric neutrinos, which can be obtained
in experiments with detectors able to measure the charge of the muon produced in the
charged current (CC) reaction by atmospheric or . It is a natural continuation of
our similar study for water-Cerenkov

detectors [34]. In the experiments with muon charge
identification it will be possible to distinguish between the and induced events.
As is well known, the water-Cerenkov

detectors do not have such a capability. Among
the operating detectors, MINOS has muon charge identification capabilities for multi-GeV
muons [35]. The MINOS experiment is currently collecting atmospheric neutrino data. The
detector has relatively small mass, but after 5 years of data-taking it is expected to collect
about 440 atmospheric and about 260 atmospheric multi-GeV events. There are also
plans to build a 3050 kton magnetized tracking iron calorimeter detector in India within
the India-based Neutrino Observatory (INO) project [36]. The INO detector will be based
on MONOLITH design [37]. The primary goal is to study the oscillations of atmospheric
and . This detector is planned to have efficient muon charge identification, high muon
energy resolution ( 5%) and muon energy threshold of about 2 GeV. It will accumulate
sufficiently high statistics of atmospheric and induced events in several years, which
would permit to search for effects of the subdominant e (e ) and e
( e ) transitions.
The subdominant e ( e ) and e ( ) ( e ( ) ) oscillations of the
multi-GeV atmospheric neutrinos of interest should exist and their effects could be observable if 3-flavor-neutrino mixing takes place in vacuum, i.e., if sin2 213 = 0, and if sin2 213
3 For a detailed discussion of the prospects to improve the sensitivity to the value of sin2 at least by a
13
factor (510), i.e., to values of 0.005 or smaller, in future experiments with reactor and accelerator neutrinos,
and to determine the sign of m231 in future experiments with accelerator neutrinos, see, e.g., Refs. [30,31]. If
neutrinos with definite mass are Majorana particles, information about the sign(m231 ) could be obtained also by
measuring the effective neutrino Majorana mass in neutrinoless double -decay experiments [32,33].
396
is sufficiently large [3840]. These transitions are driven by m231 . The probabilities of the
transitions contain sin2 23 as factor which determines their maximal value (see further).
For a given sign of m231 , the Earth matter affects differently the transitions of neutrinos
and antineutrinos: for m231 > 0, the e ( e ) and e ( ) ( e ( ) )
transitions of the multi-GeV atmospheric neutrinos, are amplified (suppressed), while if
m231 < 0, the transitions of neutrinos are suppressed and those of antineutrinos are enhanced.4 The effects of the enhancement can be substantial for sin2 13 0.01. Since the
fluxes of multi-GeV and e ( and e ) differ considerably (by a factor (2.64.5) for
E, (210) GeV), the Earth matter effects can therefore create a substantial difference
between the and + rates of events, produced by the atmospheric and in MINOS, INO, or any other detector of the same type. This difference can be relatively large
and observable in the samples of events (E (210) GeV), in which the are produced by atmospheric s with relatively large path length in the Earth, crossing deeply the
Earth mantle [41,42] or the mantle and the core [3840,4345]. These are the neutrinos for
which [34] cos n (0.30.4), n being the Nadir angle characterizing the trajectory in
the Earth.
An observable which is sensitive to the Earth matter effects, and thus to the value of
sin2 13 and sign(m231 ), as well as to sin2 23 , is the Nadir angle (n ) distribution of the
ratio N ( )/N(+ ) of the multi-GeV and + event rates, or, equivalently, of the
+ event rate asymmetry A + = [N ( ) N (+ )]/[N ( ) + N (+ )]. We have
obtained predictions for the n -distribution of A + in the case of 3- oscillations of the
atmospheric , , e and e , both for j mass spectra with normal (m231 > 0) and inverted (m231 < 0) hierarchy, (A3
) , (A3
) , and for sin2 23 = 0.64; 0.50; 0.36.
+ NH
+ IH
These are compared with the predicted n -distributions of the same asymmetry in the case
of 2-neutrino and oscillations of the atmospheric and (i.e., for
. Predictions for the three types of asymmetries of the suitably insin2 13 = 0), A2
+
tegrated n -distributions of the and + multi-GeV event rates are also given. Our
results show, in particular, that if the effects of the Earth matter enhanced transitions of
atmospheric neutrinos, e and e , or e and e , are observed in
MINOS and/or INO experiments, that would imply that sin2 213 0.05, sin2 23 0.50
and would permit to determine the sign(m231 ) and thus to answer the fundamental question about the type of hierarchynormal or inverted, the neutrino mass spectrum has.
Let us note that the Earth matter effects in atmospheric neutrino oscillations have been
widely studied (for recent detailed analyzes see, e.g., Refs. [34,40,46] which contain also a
rather complete list of references to earlier work on the subject). A rather detailed analysis
for the MONOLITH detector has been performed in Ref. [47]. A large number of stud
ies have been done for water-Cerenkov
detectors. In Ref. [34] the magnitude of the Earth
matter effects in the Nadir angle distribution of the ratio of the multi-GeV -like and
e-like events, measured in water-Cerenkov

detectors, N /Ne , has been investigated. This
n -distribution is the observable most sensitive to the matter effects of interest. In general,
the matter effects in the Nadir angle distribution of the ratio of the multi-GeV -like and
4 If and with energies E
, 2 GeV take part in 2-neutrino and oscillations, one

would have P ( ) = P ( ).
397
e-like events, N /Ne , which can be measured in the Super-Kamiokande or other water
Cerenkov
detectors, are smaller than the matter effects in the Nadir angle distribution of
the ratio of the multi-GeV and + events, N ( )/N (+ ), which can be measured
in atmospheric neutrino experiments with sufficiently good muon charge identification
(MINOS, INO, etc.). In the case of water-Cerenkov

detectors, approximately 2/3 of the
rate of the multi-GeV -like events is due to , and 1/3 is due to ; similar partition
is valid for the multi-GeV e-like events. Depending on the sign of m231 , the matter effects
enhance either the neutrino transitions, e and e , or the antineutrino transitions, e and e , but not both types of transitions. Correspondingly, because
detectors,5 only
the ,e - and ,e -induced events are indistinguishable in water-Cerenkov
2/3 or 1/3 of the events in the multi-GeV -like and e-like samples collected in these
detectors are due to neutrinos whose transitions can be enhanced by matter effects. This
effectively reduces the magnitude of the matter effects in the samples of multi-GeV -like
and e-like events. Obviously, such a dilution of the magnitude of the matter effects does
not take place in the samples of the multi-GeV and + events, which can be collected
in the experiments with muon charge identification.
2. Subdominant 3- oscillations of multi-GeV atmospheric neutrinos in the Earth

In the present section we briefly review the physics of the subdominant 3-neutrino oscillations of the multi-GeV atmospheric neutrinos in the Earth (see, e.g., Ref. [34] for further
details). Under the condition |m231 | m221 , which the neutrino mass squared differences determined from the existing atmospheric and solar neutrino and KamLAND data
satisfy, the relevant 3-neutrino e ( e ) and e ( ) ( e ( ) ) transition
probabilities reduce effectively to a 2-neutrino transition probability [49] with m231 and
13 playing the role of the relevant 2-neutrino oscillation parameters. The 3-neutrino oscillation probabilities of interest for atmospheric e, having energy E and crossing the Earth
along a trajectory characterized by a Nadir angle n , have the following form [49]:
P3 (e e )
= 1 P2 ,
2
2
P3 (e )
P2 ,
P3 (e )
P2 ,
= P3 ( e )
= s23
= c23

4
2 2
i
P3 ( ) = 1 s P2 2c s 1 Re e A2 ( ) ,
(7)
P3 ( ) = 1 P3 ( ) P3 ( e ).
(9)
23
23 23
(6)
(8)
Here P2 P2 (m231 , 13 ; E, n ) is the probability of 2-neutrino e oscillations in

the Earth, where = s23 + c23 [49], and and A2 ( ) A2 are known phase
and 2-neutrino transition probability amplitude [38,43].
The fluxes of atmospheric e, of energy E, which reach the detector after crossing the
Earth along a given trajectory specified by the value of n , e, (E, n ), are given by the
5 As was pointed out in Ref. [48], an event-by-event distinction of s could be carried out by observing
the proton above Cerenkov

threshold in quasi-elastic events. These proton events are rare, though very distinctive
when they do occur. This could also be useful for evaluating charge-asymmetries.
398
following expressions in the case of the 3-neutrino oscillations under discussion [43]:

2

e (E, n )
(10)
r 1 P2 ,
= 0e 1 + s23

1
4
2
(E, n )
r
1 P2
s23
= 0 1 + s23

2 2
2c23
s23 1 Re ei A2 ( ) ,
(11)
where 0e() = 0e() (E, n ) is the e() flux in the absence of neutrino oscillations and
r r(E, n )
0 (E, n )
0e (E, n )
(12)
The interpretation of the SK atmospheric neutrino data in terms of oscillations

2 to lie approximately in the interval (0.300.70), with 0.5 being
requires the parameter s23
the statistically preferred value. For the predicted ratio r(E, n ) of the atmospheric and
e fluxes for (i) the Earth core crossing and (ii) only mantle crossing neutrinos, having trajectories for which 0.3 cos n 1.0, one has [5052]: r(E, n )
= (2.64.5) for neutrinos
giving the main contribution to the multi-GeV samples. Thus, for the multi-GeV neutri4 [1 (s 2 r(E, ))1 ] 0.060.14 (0.160.27) and (s 2 r(E, ) 1)
nos one finds s23
=
=
n
n
23
23
2
0.31.3 (0.661.9) for s23 = 0.5 (0.64). The effects of interest are much larger for the
multi-GeV neutrinos than for the sub-GeV neutrinos6 [43].
The same conclusions are valid for the effects of oscillations on the fluxes of, and event
rates due to, atmospheric antineutrinos e and . The formulae for antineutrino fluxes and
oscillation probabilities are analogous to those for neutrinos (see Refs. [34,43]).
Eqs. (6)(8), (10)(11) and the similar equations for antineutrinos imply that in the case
under study the effects of the e , e , and e ( ) , e ( ) , oscillations
2 and are maximal for the largest allowed value of s 2 ,
(i) increase with the increase of s23
23
and (ii) in the case of the multi-GeV samples, for m231 > 0 they lead to a decrease of the
event rate, while if m231 < 0, the + event rate will decrease.
2.1. Enhancing mechanisms
As is well known, the Earth density distribution in the existing Earth models is assumed to be spherically symmetric and there are two major density structuresthe core
and the mantle, and a certain number of substructures (shells or layers). The core radius
and the depth of the mantle are known with a rather good precision and these data are
incorporated in the Earth modelsthe Stacey 1977 and the more recent PREM models
[53,54], which are widely used in the calculations of the probabilities of -oscillations in
6 Indeed, for the neutrinos giving contribution to the sub-GeV samples of Super-Kamiokande events
2 r(E, ) 1) 0,
(2.02.5). If s 2 = 0.5 and r(E, n )
one has [5052] r(E, n ) =
= 2.0, we have (s23
=
n
23
2
1
((s23 r(E, n )) 1)
= 0, and the possible effects of the e and e ( ) transitions on the e and
fluxes, and correspondingly in the sub-GeV e-like and -like samples of events, would be strongly suppressed
independently of the values of the corresponding transition probabilities.
399
the Earth.7 The mean electron number densities in the mantle and in the core read [54]:
N eman
= 2.2NA cm3 , N ec
= 5.4NA cm3 , mN and NA being the nucleon mass and Avogadro number. Numerical calculations have shown [38,56] (see also Ref. [57]) that, e.g.,
the e oscillation probability of interest, calculated within the two-layer model of the
Earth8 with N eman and N ec for a given neutrino trajectory determined using the PREM (or
the Stacey) model, reproduces with a remarkably high precision the corresponding probability, calculated by solving numerically the relevant system of evolution equations with
the much more sophisticated Earth density profile of the PREM (or Stacey) model.
In the two-layer model, the oscillations of atmospheric neutrinos crossing only the Earth
mantle (but not the Earth core), correspond to oscillations in matter with constant density.
The relevant expressions for P2 , and A2 ( ) are given by (see, e.g., Ref. [34])

2

2
2 M L
P2 m31 , 13 ; E, n = sin
(13)
sin2 2m ,
4E

M 2 L
1 m231
L + 2GF N eman L
,
=
2 2E
2E
M 2 L

A2 ( ) = 1 + ei 2E 1 cos2 m .
(14)
Here M 2 is the mass difference between the two mass-eigenstate neutrinos in the mantle,
m is the mixing angle in the mantle which coincides with 13 in vacuum, and L is the distance the neutrino travels in the mantle. For a -trajectory in the mantle, which is specified
by a given n , we have L = 2R cos n , where R = 6371 km is the Earth radius.9
It follows from Eqs. (10) and (11) that, e.g., for m231 > 0 the oscillation effects of
interest will be maximal if P2
= 1. The latter is possible provided (i) the well-known
2L
resonance condition [15,58], leading to sin2 2m

= 1, is fulfilled, and (ii) cos( M
2E ) = 1.
man
Given the value of Ne , the first condition determines the -energy at which P2 can be
enhanced: Eres
= 6.6m231 [103 eV2 ]N eman [NA cm3 ] cos 213 GeV. If the first condition
is satisfied, the second determines the length of the path L for which one can have P2
= 1.
Taking m231
= (2.03.0) 103 eV2 , N eman
= 2NA cm3 and cos 213
= 1 one finds that
Eres
= (6.610.0) GeV and that for sin2 13 = 0.05 (0.025), one can have P2
= 1 only if
L
= 8000 (10 000) km. Thus, the Earth matter effects can amplify P2 significantly when
the neutrinos cross only the mantle [42] (i) for E (611) GeV, and (ii) for cos n 0.3.
In the case of atmospheric neutrinos crossing the Earth core, new resonant effects become apparent. For sin2 13 < 0.05 and m231 > 0, we can have P2
= 1 only due to the
effect of maximal constructive interference between the amplitudes of the e transitions in the Earth mantle and in the Earth core [38,4345]. The effect differs from the
7 According to the Earth models [53,54], the core has a radius R = 3485.7 km, the Earth mantle depth
c
is approximately Rman = 2885.3 km, and the Earth radius is R = 6371 km. The mean values of the matter
densities and the electron fraction numbers in the mantle and in the core read, respectively: man
= 4.5 g/cm3 ,
c
= 11.5 g/cm3 , and [55] Yeman = 0.49, Yec = 0.467.
8 In the two-layer model of the Earth the electron number densities in the mantle and in the core are assumed
to be constant.
9 Neutrinos cross only the Earth mantle on the way to the detector if 33.17 .
n
400
MSW one [38] and the enhancement happens in the case of interest at a value of the energy between the resonance energies corresponding to the density in the mantle and that
of the core. The mantlecore enhancement effect is caused by the existence (for a given
-trajectory through the Earth core) of points of resonance-like total neutrino conversion,
P2 = 1, in the corresponding space of -oscillation parameters [44,45]. The points where
P2 = 1 are determined by the conditions [44,45]:

cos 2m
cos 2m

,
tan =
,
tan
(15)

cos(2m 4m )
cos 2m cos(2m 4m )
where the signs are correlated and cos 2m cos(2m 4m ) 0. In Eq. (15) 2 and 2
are the oscillation phases (phase differences) accumulated by the (two) neutrino states
after crossing respectively the first mantle layer and the core, and m is the -mixing
angle in the core. A rather complete set of values of m231 /E and sin2 213 for which
both conditions in Eq. (15) hold and P2 = 1 was found in Ref. [45]. The location of
these points determines the regions where P2 is large, P2 0.5. For sin2 13 < 0.05,
there are two sets of values of m231 and sin2 13 for which Eq. (15) is fulfilled and
P2 = 1. These two solutions of Eq. (15) occur for, e.g., n = 0; 13 ; 23 , at (1) sin2 213 =
0.034; 0.039; 0.051, and at (2) sin2 213 = 0.15; 0.17; 0.22 (see Table 2 in Ref. [45]).
For m231 = 2.0(3.0) 103 eV2 , P2 = 1 occurs in the case of the first solution10 at
E
= (2.83.1) GeV (E
= (4.24.7) GeV).
The effects of the mantle-core enhancement of P2 (or P2 ) increase rapidly with
sin2 213 as long as sin2 213 0.06, and should exhibit a rather weak dependence on
sin2 213 for 0.06 sin2 213 < 0.19. If 3- oscillations of atmospheric neutrinos take
place, the magnitude of the matter effects in the multi-GeV -like and e-like event samples,
produced by neutrinos crossing the Earth core, should be larger than in the event samples
due to neutrinos crossing only the Earth mantle (but not the core). This is a consequence
of the fact that in the energy range of interest the atmospheric neutrino fluxes decrease
rather rapidly with energyapproximately as E 2.7 , while the -interaction cross section
rises only linearly with E, and that the maximum of P2 (or P2 ) due to the mantle-core
interference effect takes place at approximately two times smaller energies than that due to
the MSW effect for neutrinos crossing only the Earth mantle (e.g., at E
= (3.53.9) GeV
and E
= 8.3 GeV, respectively, for m231 = 2.5 103 eV2 ).
3. Results
It follows from the preceding analysis that in the case of detectors with muon charge
identification, as observable which is most sensitive to the Earth matter effects, and thus to
the value of sin2 13 and the sign of m231 , as well as to sin2 23 , we can consider the Nadir
angle (n ) distribution of the ratio N ( )/N (+ ) of the multi-GeV and + event
10 The first solution corresponds to [38] cos 2
4 ) = 1. The
= 1, cos 2
= 1 and [45] sin2 (2m
m
enhancement effect in this case was called neutrino oscillation length resonance (NOLR) in [38].
401
rates, or equivalently the Nadir angle distribution of the + event rate asymmetry
A + =
N ( ) N (+ )
.
N ( ) + N (+ )
(16)
We have obtained predictions for the cos n distribution of the ratio N ( )/N (+ ) and the
asymmetry A + in the case of 3-neutrino oscillations of the atmospheric , , e and
e , both for neutrino mass spectra with normal (m231 > 0) and inverted (m231 < 0) hierarchy, and for sin2 23 = 0.64; 0.50; 0.36, and sin2 213 = 0.05; 0.10. These are compared
with the predicted Nadir angle distributions of the same ratio and asymmetry in the case
of 2-neutrino (sin2 13 = 0) vacuum and oscillations of the atmospheric
.
and , A2
+
In the calculations we have used the predictions for the Nadir angle and energy distributions of the atmospheric neutrino fluxes given in Ref. [52]. The interactions of the atmospheric neutrinos are described by taking into account only the and deep inelastic
scattering (DIS) cross sections. The latter are calculated using the GRV94 parton distributions given in Ref. [59]. We present here results for the asymmetry A + .11 They are
shown graphically in Figs. 17. The figures correspond to three different intervals of integration over the energies of the atmospheric and , and of the and + they produce
in the detector, E = [2, 10], [2, 20], [5, 20] GeV, and thus to three different possible event
as functions of cos n for
samples. Figs. 1, 2 show the asymmetries A + and A2
+
two reference values of |m231 |, m231 = 2 103 eV2 and m231 = 3 103 eV2 ,
while in Figs. 37 we present results for the asymmetries in the rates of the multi-GeV
and + events, integrated over cos n in the intervals [0.30, 0.84] (mantle bin) and
. The dependence of the latter on sin2 213 for
[0.84, 10] (core bin), A + and A 2
+
m231 = 2 103 eV2 , and on m231 for sin2 213 = 0.10, is shown for three values of
sin2 23 = 0.36; 0.50; 0.64. The numerical calculations are performed assuming perfect
reconstruction of the direction and the energy of the atmospheric neutrinos. In realistic detectors the spread in n and E is not negligible [47]. This is mainly due to the scattering
process, and thus would not improve much with future detectors.12 This implies that the
fine structure details in Figs. 1 and 2 will be smeared out in an actual measurement. Nevertheless, Figs. 1 and 2 show in a clear way the resonance effects explained in the previous
section and indicate the ranges in cos n where these effects can be large. The indicated
approximations have a much smaller impact on the asymmetries of the integrated rates,
shown in Figs. 37.
As Figs. 17 indicate, the Earth matter effects can produce a noticeable deviations of
at cos n 0.3. As a
A + from the 2-neutrino vacuum oscillation asymmetry A2
+
11 It is interesting to note that the ratio N ( )/N (+ ) exhibits essentially the same dependence on cos
n
as the asymmetry A + . This is a consequence of the special form of the dependence of A + on
N ( )/N (+ ) and of the fact that typically one finds N ( )/N (+ ) (1.52.4) for the ranges of the values of
the parameters of interest. Correspondingly, the following approximate relation holds (within 20% and typically
with much higher precision) for the range of values of the parameters of interest: N ( )/N (+ )
= 6A + .
12 Future iron-magnetized calorimeter detectors would allow to reconstruct the initial neutrino energy and
direction with a resolution of 20% and 5 10 , respectively [3537].
402
Fig. 1. The Nadir angle distribution of the charge asymmetry, A + , Eq. (16), of the multi-GeV and +
event rates, integrated over the neutrino (and muon) energy in the interval E = (2.010.0) GeV, in the cases
(i) of 2-neutrino and oscillations in vacuum of the atmospheric and and no e and e
oscillations, A2
(solid lines), (ii) 3-neutrino oscillations of , e and e in the Earth and neutrino mass
+
)
(dashed lines), or with inverted hierarchy, (A3
) (dotted
spectrum with normal hierarchy (A3
+ NH
+ IH
lines). The results shown are for |m231 | = 2 103 eV2 , sin2 23 = 0.36 (upper panels); 0.50 (middle panels);
0.64 (lower panels), and sin2 213 = 0.05 (left panels); 0.10 (right panels).
quantitative measure of the magnitude of the matter effects one can use the deviation of
the asymmetry A + in the case of 3-neutrino oscillations, sin2 213 = 0, sin2 213
, predicted in the case of 2-neutrino oscillations, i.e., for
0.04, from the asymmetry, A2
+
sin2 213 = 0, or the relative difference between the two asymmetries,
=
A + A2
+
A2
+
(17)
The magnitude of the matter effects, or the relative difference , depends critically on
the value of sin2 23 : || increases rapidly with the increasing of sin2 23 . This is clearly
seen in Figs. 17. The matter effects in A + are hardly observable for sin2 23 0.30,
though.
It follows from Figs. 1, 2 that the deviations of the asymmetry A + from the 2, are maximal typically (i) in the core bin, cos n =
neutrino oscillation one, A2
+
403
Fig. 2. The same as in Fig. 1, but for and + event rates integrated over the neutrino (and muon) energy in
the interval E = (5.020.0) GeV and for |m231 | = 3 103 eV2 .
[0.841.0], and (ii) at cos n 0.50 in the mantle bin, cos n 0.84. For m231 = 2
103 eV2 , sin2 23 = 0.50 (0.64) and E = [2, 10] GeV (Fig. 1), we have at cos n 0.50
0.29, while the matter effects lead to A +
and sin2 213 = 0.05: A2
= 0.25 (0.24).
+ =
This corresponds to a negative relative difference between A( + ) and A2 ( + ),

< 0, and || 14% (17%). For sin2 213 = 0.10, one finds 28% (34%). The
relative difference in the core bin is also negative and || has similar or larger values. For E in the interval E = [5, 20] GeV (Fig. 2), we get at cos n 0.50 for m231 =
0.25, A +
3 103 eV2 , sin2 213 = 0.10 and sin2 23 = 0.64: A2
= 0.075, and
+ =
a relative difference between the two asymmetries = 70%.

Consider next the asymmetries in the rates the multi-GeV and + events, integrated
over cos n in the intervals [0.30, 0.84] (mantle bin) and [0.84, 10] (core bin), A + and
A 2
. As can be seen from Figs. 3, 5 and 7, for cos n 0.84, i.e., in the mantle bin, the
+
asymmetry difference || increases practically linearly with sin2 213 . On the other hand,
from Figs. 4 and 6, we see that in the case of 0.84 cos n 1.0, i.e., in the core bin, ||
increases rapidly with sin2 213 until the latter reaches the value of sin2 213
= 0.06. For
values of sin2 213
= (0.060.15), is essentially independent of sin2 213 and is given
by its value at sin2 213
= 0.06. The magnitude of the asymmetry difference depends
weakly on m231 taking values in the interval (23) 103 eV2 , as long as the energy
integration interval is sufficiently wide to include the energy regions where the Earth mat-
404
Fig. 3. The charge asymmetry A + of the multi-GeV and + event rates, integrated over the neutrino
(and muon) energy in the interval E = (2.010.0) GeV and over the Nadir angle in the interval corresponding to
0.30 cos n 0.84 (mantle bin), as function (i) of sin2 213 for |m231 | = 2103 eV2 (left panels), and (ii) of
(solid lines), (A3
)
(dashed lines) and (A3
)
|m231 | for sin2 213 = 0.10 (right panels): A2
+ NH
+ IH
+
(dotted lines). The results shown are obtained for sin2 23 = 0.36 (upper panels); 0.50 (middle panels); 0.64
(lower panels).
ter effects enhance strongly the subdominant transition probabilities. If this is not the case,
a noticeable dependence on m231 can be present. This is illustrated, e.g., in Fig. 3, which
corresponds to E = [2, 10] GeV. The asymmetry difference in the mantle bin diminishes
monotonically as |m231 | increases starting from the value of 1.3 103 eV2 and becomes rather small at |m231 | 3 103 eV2 . This behavior can be easily understood:
for |m231 |
= 2 103 eV2 , the region of enhancement of the subdominant neutrino oscillations lies in the region of energy integration, while for |m231 | > 3 103 eV2 the
enhancement region is practically outside the region of integration over the neutrino energy.
For the ranges considered of the three oscillation parameters, sin2 23 , sin2 213 and
|m231 |, the magnitude of the asymmetry difference || depends weakly on the maximal
neutrino (and muon) energy, Emax , for the chosen event sample as long as Emax 10 GeV.
By increasing the minimal energy of the neutrinos contributing to the event sample, Emin ,
from 2 GeV to, e.g., 5 GeV, one could diminish the asymmetry in the core bin substantially
(Fig. 2). As outlined in the previous section, the resonances for neutrinos crossing the core
happen at lower energies than for those traveling through the mantle, and therefore, in that
405
Fig. 4. The same as in Fig. 3, but for and + event rates integrated over the Nadir angle in the interval
corresponding to 0.84 cos n 1.00 (core bin).
case, a large fraction of the region of enhancement is not included within the interval of
integration.
For sin2 23 0.50, sin2 213 0.06 and |m231 | = (23) 103 eV2 , the Earth matter
effects produce an integrated asymmetry difference || which is bigger than approximately
15%, can reach the values of (3035)% (Figs. 37), and thus can be sufficiently large to
be observable in iron-magnetized detectors. As Figs. 17 clearly show, the sign of the relative difference of the integrated asymmetries is anticorrelated with the sign of m231 : for
, while for m231 < 0, the inequality A + > A2
m231 > 0 we have A + < A2
+
+
holds.13 Therefore the measurement of A + in this type of detectors could provide a direct information on the sign of m231 .
Our results show that the Earth matter effects in the Nadir angle distribution of the ratio N ( )/N(+ ) of the rates of multi-GeV and + events, or equivalently in the
Nadir angle distribution of the + event rate asymmetry A + , Eq. (16), can be
sufficiently large to be observable in the current and planned experiments with iron mag13 Note that A2
> 0. This is a consequence of the fact that the DIS cross section is approximately by
+
a factor 2 bigger than the DIS cross section and that the fluxes of atmospheric and , (e and e ) do not
differ considerably.
406
Fig. 5. The same as in Fig. 3, but for |m231 | = 3 103 eV2 and and + event rates, integrated over the
neutrino (and muon) energy in the interval E = (2.020.0) GeV.
netized calorimeter detectors which have muon charge identification capabilities (MINOS,
INO, etc.).
4. Conclusions
We have studied the possibilities to obtain information on the values of sin2 13 and
sin 23 , and on the sign of m231 using the data on atmospheric neutrinos, which can be
obtained in experiments with detectors able to measure the charge of the muon produced
in the charged current (CC) reaction by atmospheric or (MINOS, INO, etc.). The
indicated oscillation parameters control the magnitude of the Earth matter effects in the
subdominant oscillations, e (e ) and e ( e ), of the multi-GeV
(E (210) GeV) atmospheric neutrinos. As observable which is most sensitive to the
Earth matter effects, and thus to the value of sin2 13 and the sign of m231 , as well as to
sin2 23 , we have considered the Nadir angle (n ) distribution of the ratio N ( )/N (+ )
of the multi-GeV and + event rates, and the corresponding + event rate asymmetry A + , Eq. (16). The systematic uncertainty, in particular, in the Nadir angle dependence of N ( )/N(+ ) and of the asymmetry A + , can be smaller than those on the
measured Nadir angle distributions of the rates of and + events, N ( ) and N (+ ).
We have obtained predictions for the cos n distribution of the asymmetry A + (and of
2
407
Fig. 6. The same as in Fig. 3, but for |m231 | = 3 103 eV2 and and + event rates, integrated over
the neutrino (and muon) energy in the interval E = (2.020.0) GeV and over the Nadir angle in the interval
corresponding to 0.84 cos n 1.00 (core bin).
the ratio N ( )/N(+ )) in the case of 3-neutrino oscillations of the atmospheric , ,

e and e , both for neutrino mass spectra with normal (m231 > 0) and inverted (m231 < 0)
hierarchy, and for sin2 23 = 0.64; 0.50; 0.36, and sin2 213 = 0.05; 0.10. These are compared with the predicted Nadir angle distribution of the same ratio and asymmetry in the
case of 2-neutrino (sin2 13 = 0) vacuum and oscillations of the atmospheric and , A2
.
+
Our results are summarized in Figs. 17. Figs. 1, 2 show the dependence of the asymmetries A + and A2
on cos n for two reference values of |m231 |, m231 =
+
2 103 eV2 and m231 = 3 103 eV2 , and for two possible ranges of energies of the
atmospheric neutrinos, contributing to the event rates of interest, E = [2, 10], [5, 20] GeV.
In Figs. 37 we present results for the asymmetries in the rates of the multi-GeV
and + events, integrated over cos n in the intervals [0.30, 0.84] (mantle bin) and
[0.84, 10] (core bin) for E = [2, 10], [2, 20], [5, 20] GeV. We find that for sin2 23 0.50,
sin2 213 0.06 and |m231 | = (23) 103 eV2 , the Earth matter effects produce a relawhich is bigger in
tive difference between the integrated asymmetries A + and A 2
+
absolute value than approximately 15%, can reach the values of (3035)% (Figs. 37),
and thus can be sufficiently large to be observable. As our results show (Figs. 37), the
sign of the indicated asymmetry difference, (A + A 2
), is directly related to the
+
408
Fig. 7. The same as in Fig. 3, but for and + event rates integrated over the neutrino (and muon) energy in
the interval E = (5.020.0) GeV, and for |m231 | = 3 103 eV2 .
) < 0, while if m231 < 0 then

sign of m231 : for m231 > 0 we have (A + A 2
+
(A + A 2 + ) > 0. Therefore the measurement of the Nadir angle dependence of

A + , or of the value of A + in the mantle and/or in the core bins, can provide a direct
information on the sign of m231 , i.e., on the neutrino mass hierarchy.
To summarize, the studies of the oscillations of the multi-GeV atmospheric and in
experiments with detectors having good muon charge identification capabilities (MINOS,
INO, etc.), can provide fundamental information on the values of sin2 13 and sin2 23 , and
on the sign of m231 , i.e., on the neutrino mass hierarchy.
Acknowledgements
We are indebted to J. Bernabu, T. Kajita, A. Mann and S. Wojcicki for useful discussions. S.T.P. would like to thank Prof. T. Kugo, Prof. M. Nojiri and the other members
of the Yukawa Institute for Theoretical Physics (YITP), Kyoto, Japan, where part of the
work on this article was done, for the kind hospitality extended to him. S.P.-R. would like
to thank the Theory Division at CERN for hospitality during the final completion of this
work. This work is supported in part by the Italian INFN under the programs Fisica Astroparticellare (S.T.P.) and by NASA grant NAG5-13399 (S.P.-R.).
409
References
[1] B.T. Cleveland, et al., Astrophys. J. 496 (1998) 505;
Kamiokande Collaboration, Y. Fukuda, et al., Phys. Rev. Lett. 77 (1996) 1683;
SAGE Collaboration, J.N. Abdurashitov, et al., J. Exp. Theor. Phys. 95 (2002) 181;
GALLEX Collaboration, W. Hampel, et al., Phys. Lett. B 447 (1999) 127;
T. Kirsten, Nucl. Phys. B (Proc. Suppl.) 118 (2003) 33.
[2] Super-Kamiokande Collaboration, S. Fukuda, et al., Phys. Rev. Lett. 86 (2001) 5656.
[3] SNO Collaboration, Q.R. Ahmad, et al., Phys. Rev. Lett. 87 (2001) 071301.
[4] Q.R. Ahmad, et al., Phys. Rev. Lett. 89 (2002) 011301, 011302.
[5] SNO Collaboration, S.N. Ahmed, et al., Phys. Rev. Lett. 92 (2004) 181301.
[6] Super-Kamiokande Collaboration, Y. Fukuda, et al., Phys. Rev. Lett. 81 (1998) 1562.
[7] KamLAND Collaboration, K. Eguchi, et al., Phys. Rev. Lett. 90 (2003) 021802.
[8] K2K Collaboration, M.H. Ahn, et al., Phys. Rev. Lett. 90 (2003) 041801.
[9] B. Pontecorvo, Zh. Eksp. Teor. Fiz. 53 (1967) 1717, Sov. Phys. JETP 26 (1968) 984.
[10] S.M. Bilenky, B. Pontecorvo, Phys. Rep. 41 (1978) 225.
[11] S.M. Bilenky, S.T. Petcov, Rev. Mod. Phys. 59 (1987) 671.
[12] S.T. Petcov, in: H. Gausterer, C.B. Lang (Eds.), Lecture Notes in Physics, vol. 512, Springer, Berlin, 1998,
p. 281, hep-ph/9806466.
[13] A. Bandyopadhyay, et al., Phys. Lett. B 581 (2004) 62.
[14] L. Wolfenstein, Phys. Rev. D 17 (1978) 2369;
L. Wolfenstein, Phys. Rev. D 20 (1979) 2634.
[15] S.P. Mikheyev, A.Yu. Smirnov, Yad. Fiz. 42 (1985) 1441.
[16] Super-Kamiokande Collaboration, Y. Ashie, et al., Phys. Rev. Lett. 93 (2004) 101801.
[17] KamLAND Collaboration, T. Araki, et al., hep-ex/0406035.
[18] K2K Collaboration, E. Aliu, et al., hep-ex/0411038.
[19] V. Gribov, B. Pontecorvo, Phys. Lett. B 28 (1969) 493.
[20] B. Pontecorvo, Zh. Eksp. Teor. Fiz. 33 (1957) 549;
B. Pontecorvo, Zh. Eksp. Teor. Fiz. 34 (1958) 247;
Z. Maki, M. Nakagawa, S. Sakata, Prog. Theor. Phys. 28 (1962) 870.
[21] S.M. Bilenky, et al., Phys. Lett. B 94 (1980) 495.
[22] M. Doi, et al., Phys. Lett. B 102 (1981) 323;
J. Bernabu, P. Pascual, Nucl. Phys. B 228 (1983) 21.
[23] P. Langacker, et al., Nucl. Phys. B 282 (1987) 589.
[24] M. Apollonio, et al., Phys. Lett. B 466 (1999) 415.
[25] F. Boehm, et al., Phys. Rev. Lett. 84 (2000) 3764;
F. Boehm, et al., Phys. Rev. D 62 (2000) 072002.
[26] E. Kearns, Talk given at 04 International Conference, Paris, France, 1419 June 2004.
[27] A. Bandyopadhyay, et al., hep-ph/0406328.
[28] J.N. Bahcall, M.C. Gonzalez-Garcia, C. Pena-Garay, JHEP 0408 (2004) 016;
M. Maltoni, et al., hep-ph/0405172.
[29] S.T. Petcov, Talk given at 04 International Conference, Paris, France, 1419 June 2004, http://
neutrino2004.in2p3.fr.
[30] K. Anderson, et al., hep-ex/0402041.
[31] P. Huber, et al., hep-ph/0403068;
K. McConnel, M.H. Shaevitz, hep-ex/0409028.
[32] S.M. Bilenky, S. Pascoli, S.T. Petcov, Phys. Rev. D 64 (2001) 053010.
[33] S. Pascoli, S.T. Petcov, Phys. Lett. B 544 (2002) 239;
S. Pascoli, S.T. Petcov, Phys. Lett. B 580 (2004) 280;
S. Pascoli, S.T. Petcov, W. Rodejohann, Phys. Lett. B 558 (2003) 141.
[34] J. Bernabu, S. Palomares-Ruiz, S.T. Petcov, Nucl. Phys. B 669 (2003) 255.
[35] D. Michael, et al., Nucl. Phys. B (Proc. Suppl.) 118 (2003) 189.
[36] G. Rajasekaran, hep-ph/0402246.
410
[37] MONOLITH Collaboration, N.Y. Agafonova, et al., Proposal LNGS-P26-2000, http://castore.mi.infn.

it/~monolith/.
[38] S.T. Petcov, Phys. Lett. B 434 (1998) 321;
S.T. Petcov, Phys. Lett. B 444 (1998) 584, Erratum.
[39] J. Bernabu, et al., Phys. Lett. B 531 (2002) 90;
S. Palomares-Ruiz, J. Bernabu, hep-ph/0312038.
[40] M.C. Gonzlez-Garca, M. Maltoni, Eur. Phys. J. C 26 (2003) 417.
[41] M. Freund, et al., Nucl. Phys. B 578 (2000) 27.
[42] M.C. Bauls, G. Barenboim, J. Bernabu, Phys. Lett. B 513 (2001) 391;
J. Bernabu, S. Palomares-Ruiz, hep-ph/0112002;
J. Bernabu, S. Palomares-Ruiz, hep-ph/0201090.
[43] S.T. Petcov, hep-ph/9809587;
S.T. Petcov, hep-ph/9811205;
S.T. Petcov, hep-ph/9907216;
M.V. Chizhov, M. Maris, S.T. Petcov, hep-ph/9810501.
[44] M.V. Chizhov, S.T. Petcov, Phys. Rev. Lett. 83 (1999) 1096;
M.V. Chizhov, S.T. Petcov, Phys. Rev. Lett. 85 (2000) 3979.
[45] M.V. Chizhov, S.T. Petcov, Phys. Rev. D 63 (2001) 073003.
[46] Super-Kamiokande Collaboration, T. Kajita, et al., Talk given at the International Workshop NOON2004,
Tokyo, Japan, 1115 February 2004.
[47] T. Tabarelli de Fatis, Eur. Phys. J. C 24 (2002) 43.
[48] J.F. Beacom, S. Palomares-Ruiz, Phys. Rev. D 67 (2003) 093001.
[49] S.T. Petcov, Phys. Lett. B 214 (1988) 259.
[50] M. Honda, et al., Phys. Rev. D 52 (1995) 4985.
[51] V. Agraval, et al., Phys. Rev. D 53 (1996) 1314.
[52] G. Fiorentini, V.A. Naumov, F.L. Villante, Phys. Lett. B 510 (2001) 173.
[53] F.D. Stacey, Physics of the Earth, second ed., Wiley, London, 1977.
[54] A.D. Dziewonski, D.L. Anderson, Phys. Earth Planet. Inter. 25 (1981) 297.
[55] M. Maris, S.T. Petcov, Phys. Rev. D 56 (1997) 7444.
[56] M. Maris, S.T. Petcov, Study performed in NovemberDecember 1997, unpublished.
[57] P.I. Krastev, S.T. Petcov, Phys. Lett. B 205 (1988) 84.
[58] V. Barger, et al., Phys. Rev. D 22 (1980) 2718.
[59] M. Gluck, E. Reya, A. Vogt, Z. Phys. C 67 (1995) 433.
The polarised photon g1 sum rule at the linear

collider and high luminosity B factories
G.M. Shore
Department of Physics, University of Wales, Swansea, Swansea SA2 8PP, UK
Abstract
The sum rule for the first moment of the polarised (virtual) photon structure function g1 (x, Q2 ;
2
K ) is revisited in the light of proposals for future e+ e colliders. The sum rule exhibits an array
of phenomena characteristic of QCD: for real photons (K 2 = 0) electromagnetic gauge invariance
constrains the first moment to vanish; the limit for asymptotic photon virtuality (m2 K 2 Q2 )
is governed by the electromagnetic UA (1) axial anomaly and the approach to asymptopia by the
gluonic anomaly; for intermediate values of K 2 , it reflects the realisation of chiral symmetry and is
determined by the off-shell radiative couplings of the pseudoscalar mesons; finally, like many polar
isation phenomena in QCD, the first moment of g1 involves the gluon topological susceptibility. In
this paper, we review the original sum rule proposed by Narison, Shore and Veneziano and extend
the relation with pseudoscalar mesons. The possibility of measuring the sum rule in future polarised
e+ e colliders is then considered in detail, focusing on the International Linear Collider (ILC) and
high luminosity B factories. We conclude that all the above features of the sum rule should be accessible at a polarised collider with the characteristics of SuperKEKB.
PACS: 12.38.-5; 13.60.-r; 13.88.+e
This research is supported in part by PPARC grant PP/G/O/2002/00470.

E-mail address: g.m.shore@swansea.ac.uk (G.M. Shore).
doi:10.1016/j.nuclphysb.2005.01.026
412
G.M. Shore / Nuclear Physics B 712 (2005) 411429
1. Introduction
The sum rule for the first moment of the polarised photon structure function g1 (x, Q2 ;
K 2 ) provides a window into many features of QCD dynamics, including the gluonic axial
anomaly and the realisation of chiral symmetry. This sum rule was first proposed by Narison, Shore and Veneziano in 1992 [1] as part of a series of investigations into gluonic and
anomaly-dependent phenomena in QCD, notably the origin of the spin of the proton supp
pression observed in the first moment of the polarised proton structure function g1 [27].
At that time, however, the details of the sum rule were out of reach of contemporary col
liders since the spin asymmetries which need to be measured to determine g1 (x, Q2 ; K 2 )
require exceedingly large luminosities.
Since that time, collider technology has moved on and plans are now well advanced for
machines capable of integrated annual luminosities in the regime of inverse attobarns. It is
therefore appropriate to revisit the g1 sum rule and investigate whether this new generation
of colliders will be able to measure the full array of QCD phenomena encoded in it.
We focus on two future machines. First, the International Linear Collider (ILC) has
recently passed the technology choice phase and agreement has been found to proceed with
the cold, i.e., superconducting magnet, design. If international agreement is forthcoming,
it is hoped that a linear e+ e collider with a CM energy of at least 500 GeV will be
commissioned around 2015. The projected luminosity for the ILC is around 1034 cm2 s1 ,
corresponding to an annual integrated luminosity of order 0.1 ab1 [8,9]. Of course, for
our purposes, the collider would have to be run in polarised mode.
The second machine we consider in detail is SuperKEKB. It was already noted in
Ref. [1] that high-luminosity B factories were the colliders of choice for measuring the
g1 sum rule, since high energy is not in itself an advantage but ultra-high luminosity is
essential. The proposed upgrade of KEKB to SuperKEKB [10] envisages an 8 GeV (e )
on 3.5 GeV (e+ ) collider with a target luminosity 5 1035 cm2 s1 , corresponding to
5 ab1 annual integrated luminosity. As we shall show, if this machine were run with po
larised beams, this luminosity would allow the full details of the off-shell first moment g1
sum rule to be measured.
We come to these experimental considerations in Section 5. We begin though with a
brief review of the derivation of the first moment sum rule itself, emphasising that, as appropriate for a measurement in e+ e collisions (as opposed to doing two-photon physics
using real back-scattered laser photons as the target), we are determining the off-shell struc
ture function g1 (x, Q2 ; K 2 ), where K 2 is the virtuality of the off-shell target photon in
DIS. Almost all the interesting QCD physics resides in the K 2 -dependence of the sum rule.
An important point is that we should therefore avoid an unnecessary use of the equivalent photon formalism [11,12] in establishing the sum rule. Moreover, all our results will
be formulated in QCD field-theoretic terms, focusing on the OPE and current correlation
functions, rather than using parton language. This makes the important non-perturbative
results far more transparent. (For a selection of reviews and recent papers on g1 from a
parton perspective, see, e.g., [1220].)
Having reviewed the basic sum rule, in Section 3 we study its asymptotic properties
for momenta K 2 0 and K 2 m2 . The first moment is known [1,21,22] to vanish for real
photons as a consequence of electromagnetic gauge invariance (current conservation). For
413
photon virtualities well above the relevant hadronic scale of m2 (but still of course, in the
DIS regime K 2 Q2 ), dx g1 (x, Q2 ; K 2 ) tends to a value fixed by the electromagnetic

axial UA (1) anomaly. Moreover, the approach to this asymptotic region is governed by the
gluonic contribution to the anomaly, so there is much of theoretical interest even in this
essentially perturbative regime.
For intermediate virtualities, K 2 O(m2 ), the first moment depends on the explicit
momentum-dependence of the form factors specifying the three-current AVV correlation
function involving the hadronic axial UA (1) current and two electromagnetic currents. This
is an important non-perturbative object in QCD and the ability to measure it explicitly for
a range of photon momenta would provide an interesting window into chiral symmetry
breaking and associated QCD phenomena. The reasons for this are explored in considerable detail in Ref. [23], a companion paper to Ref. [1]. In particular, we can show that these
form factors are essentially the off-shell couplings of the pseudoscalar mesons , , to
photons, whose on-shell limits are determined by the radiative decays , , . The
radiative pion decay has played a distinguished role in establishing the reality of anomalies and the nature of QCD (in particular, by providing a direct measure of the number
of colours). In the flavour singlet sector, the theory is even more interesting as it involves
in an essential way the gluonic axial UA (1) anomaly and the gluon content of the me
son [24]. In Section 4, therefore, we explore the connection between the g1 sum rule and
radiative pseudoscalar decays, extending the results of Ref. [1] to incorporate the analysis
developed in our papers [2527]. One theoretically interesting feature is the link with the
gluon topological susceptibility, which plays a key role in many polarised QCD phenomena, notably the spin of the proton. This section may be read in conjunction with another
paper, Ref. [28], in which we revisit our results [2527] for radiative pseudoscalar decays
and their relation with the topological susceptibility and WittenVeneziano formula [29,
30], and derive explicit experimental values for the pseudoscalar decay constants which
may be (carefully) compared with large Nc chiral Lagrangians [31,32] (see also Ref. [33]).
Having reviewed and developed the theory of the g1 first moment sum rule, we then
return in Section 5 to the experimental question of whether it can be measured, including
the full range of K 2 -dependence, in the forthcoming generation of high-luminosity colliders. Our conclusion is that the ILC is marginal for this purpose, but that a polarised collider
with the energy and luminosity of SuperKEKB would be able to uncover the full dynamical
richness of the sum rule.
2. The sum rule for
dx g1 (x, Q2 ; K 2 )
We are concerned with the process e+ e e+ e X, which at sufficiently high energy

is dominated by the two-photon interaction shown in Fig. 1. The deep-inelastic limit is
characterised by Q2 , e , with1 xe = Q2 /2e and x = Q2 /2 fixed, where (see
1 We have made a number of changes of notation compared to Ref. [1]. The dictionary is K 2 2 , ,
e
, xe x, x y. The standard DIS notation (, x) therefore refers here to the target photon, rather than the
target electron as in Ref. [1].
414
Fig. 1. Kinematics for the two-photon DIS process e+ e e+ e X.
Fig. 1 for definitions of the momenta) Q2 = q 2 , K 2 = k 2 , e = p2 .q, = k.q and s =

(p1 + p2 )2 . We also consider the target photon to be relatively soft, K 2 Q2 .
Verifying the first moment sum rule for the polarised structure function g1 (x, Q2 ; K 2 )
requires studying the spin asymmetries in cross-sections which are differential with respect
to Q2 , K 2 and xe . Experimentally, these are determined from
xe =
E1 sin2
1
2
E1 cos2 21
Q2 = 4EE1 sin2
x=
1
,
2
Q2
,
Q2 + W 2
K 2 EE2 22 .
(2.1)
Here, E1 (E2 ) and 1 (2 ) are the energy and scattering angle of the hard-scattered (target)
electron and W is the invariant hadronic mass. For the values K 2 m2 of interest in the
sum rule, the target electron is nearly-forward and 2 is very small. If it can be tagged,
then the virtuality K 2 is simply determined from Eq. (2.1); otherwise K 2 can be inferred
indirectly from a measurement of the total hadronic energy.
A systematic presentation of the relations between cross-section moments and structure
functions from first principles may be found in Ref. [1], so here we shall only display
some key formulae. Electron structure functions F2e (xe , Q2 ), FLe (xe , Q2 ) and g1e (xe , Q2 )
are introduced in the analogous way to ordinary nucleon structure functions and are related
to the spin-dependent cross-sections as follows:
= 2
21
dQ2
Q2
= 2
21
1
0
s
0
dQ2
Q2

2
dxe
1 Q2
e xe s
e1Q
F2
FL
,
1+
2 xe s
2 xe s
xe2
Q2
1
0

1 Q2
dxe e
,
g1 1
xe
2 xe s
(2.2)
(2.3)
415
where = 12 (++ + + ) and

= 12 (++ + ) with +, referring to the electron
helicities. The parameter Q2 /xe s 1 and only leading order terms are retained below. is
the fine structure constant.
The photon structure functions themselves may be defined in the standard way in terms
of the matrix elements of the off-shell matrix elements

(k, )Jem (q)Jem (q) (k, )
of electromagnetic currents in the DIS limit. We can readily show that the electron structure functions introduced above can be expressed as convolutions of the photon structure
functions with appropriate AltarelliParisi splitting functions. In particular, we have
F2e
xe , Q2 =
2
dK 2
K2
xe , Q2 =
2

0

xe
dx xe
P e
F2 x, Q2 ; K 2 ,
x x
x
(2.4)

xe
dx
g1 x, Q2 ; K 2 ,
P e
x
x
(2.5)
xe
g1e
1
dK 2
K2
1
xe
where
P e (z) =
1 + (1 z)2 ,
z
P e (z) = (2 z).
(2.6)
These results allow us to write relations that link the x-moments of the photon structure
functions themselves to the xe -moments of the cross-section. These key expressions, which
we return to in Section 5, are:
1
dxe xen
0
1
1
n
dz z P e (z)
0
1
dxe xen
0
d 3
1
= 3 4 2
dQ2 dxe dK 2
Q K
d 3
1
= 3 2 2
2
2
dQ dxe dK
sQ K
dx x n1 F2 x, Q2 ; K 2 ,
1
(2.7)
1
dz zn1
P e (z)
0
dx x n1 g1 x, Q2 ; K 2 .
(2.8)
The integrals factorise, so we see that the nth x-moments of the photon structure functions
are given by the (n + 1)st xe -moments of the fully differential cross-sections.
The expressions (2.4), (2.5) are derived from first principles using only Feynman diagram rules and the operator product expansion. Despite the appearance of the AP splitting
functions, no parton technology is used. Most importantly, it is not necessary to resort to
the equivalent photon approximation [11,12] (see Ref. [1] for a careful discussion of this
point), so we can be certain that our results are accurate throughout the required range
of target photon momenta K 2 . The relevant OPE is for the product of electromagnetic
416
currents:
iJem (q)Jem (q)
Q2
2n
1 ...n
q qn i1 q E1a,n (Q2 )Ra,n
(0)
2n 2
Q
a
n=1,odd
+ even-parity operators,
(2.9)
1 ...n
Ra,n
(0)
E1a,n (Q2 )
are Wilson coefficients and

are the complete set of oddwhere the
parity, twist-2 operators in QCD. (We only indicate here the odd-parity operators, which
contribute to g1 , to establish notation; the even-parity operators contribute to F2 and FL .

See Ref. [1] for full details and the explicit forms of the relevant operators.) Their form
factors in the 3-point correlation functions with the photon fields A (k) are:
1 ...n
(0)A (k)A (k)|0 =
0|Ra,n
1 2

k k n i1 k R a,n (K 2 )
K4
(n 1, odd).
(2.10)
The key result now is the relation between the moments of the structure functions and
the Wilson coefficients and form factors from the OPE. We can show [1]
1
a,n
2
dx x n1 g1 x, Q2 ; K 2 =
E1 Q R a,n K 2 ,
(2.11)
with similar expressions for F2 and FL . We are of course primarily concerned with the n =
1 moment of g1 (x, Q2 ; K 2 ). In this case, the only relevant Ra,1 operator is the hadronic
r ,
UA (1) axial current J5
r
r 5 ,
= T
Ra,1 J5
(2.12)
where T r are the SU(3) generators (including the flavour singlet T 0 = 1), and are quark
fields. Clearly, only the diagonal generators a = 3, 8, 0 contribute. The form factor is therefore defined as

1
r
(0)A (k)A (k)|0 = 4 i k R r,1 K 2
0|J5
(2.13)
K
and the essential first moment sum rule is
1
0
r,1

dx g1 x, Q2 ; K 2 =
E1 Q2 R r,1 K 2 .
(2.14)
r=3,8,0
To sum up this section, we have established in Eq. (2.8) how to measure the first moment
of the polarised photon structure function g1 (x, Q2 ; K 2 ) in terms of the spin asymmetry of
the differential cross-section, d 3
/dQ2 dxe dK 2 . The kinematic variables Q2 and xe are
found from the energy and scattering angle of the hard-scattered electron, while the target
photon virtuality K 2 is most easily found by tagging the nearly-forward electron. On the
theory side, Eq. (2.14) determines the first moment of g1 (x, Q2 ; K 2 ) in terms of a perturbatively known Wilson coefficient E1r,1 (Q2 ) and a non-perturbative form factor R r,1 (K 2 )
characterising the AVV 3-current correlation function involving the hadronic UA (1) axial
current and two electromagnetic currents.
417
3. The AVV correlation function, anomalies and momentum-dependence

The first element of the sum rule is the Wilson coefficient E1r,1 (Q2 ). The Q2 dependence is governed by the RGE and the solution is well known. For QCD with Nc = 3
and Nf = 3 active flavours (results for the arbitrary nth moments and general Nc , Nf are
quoted in Ref. [1]), we have

s (Q2 )
(r = 3, 8),
E1r,1 = c(r) 1
(3.1)
t

s (Q2 )
0,1
(0)

1
E1 = c exp
(3.2)
,
dt s (t )
t = 12 ln Q2 /2 . The coefficients are fixed by the quark charges
1/(3 3 ) and c(0) = 2/9. Since the singlet axial current is not
where
and are c(3) = 1/3,
conserved because
c(8) =
of the gluonic contribution to the UA (1) anomaly, it is associated with an anomalous dimension , which enters into the momentum-dependence of the singlet Wilson coefficient.
Explicitly,
= 0
s
2
1 s 2 + ,
4
(4)
where 0 = 0 and 1 = 3/4. Notice the important result that the expansion begins only at
O(s2 ). The beta function, which determines the running of the QCD coupling s (Q2 ), is
similarly given by
= 0
s2
3
1 s 2 +
4
(4)
with 0 = 18.
Now consider the amputated AVV correlation function in Eq. (2.13), defined with the
(em)
electromagnetic current J . We first allow the axial current momentum p to be non-zero
and subsequently take the required limit p 0. There are two important Ward identities.
Electromagnetic current conservation implies

(em)
r
ik1
0|J5
(3.3)
(p)J (k1 )J(em) (k2 )|0 = 0 sim for k2 .
The (anomalous) chiral Ward identity, which follows from the usual anomalous conservation law (for Nf = 3)

(3.4)
F F ,
8
G , with G the gluon field strength and F
r 5 and Q = s tr G
where 5r = T
8
the electromagnetic field strength, is
r
J5
= Mrs 5s + 6Qr0 + a (r)
r
ip
0|J5
(p)J(em) (k1 )J(em) (k2 )|0 Mrt
0|5t (p)J(em) (k1 )J(em) (k2 )|0
(em)
r0 6
0|Q(p)J
(k1 )J(em) (k2 )|0 +
1 (r)
a k1 k2 = 0.
8 2
(3.5)
418
The notation used for the quark masses follows Ref. [25]:

diag(mu , md , ms ) =
mr T r ,
r=3,8,0
then Mrt = drst ms , where drst are the usual SU(3) d-symbols. The term involving the
gluon topological density Q is the gluonic UA (1) anomaly and only appears in the flavour
singlet case, while the final term arises because of the electromagnetic UA (1) anomaly
whose
depends on the quark charges. With our normalisations, a (3) = 1, a (8) =
strength
(0)
1/ 3 and a = 4. (Notice that this notation differs from Ref. [1]. a (r) here is 2Nc times
r of Refs. [2527].)
the a (r) of Ref. [1] but coincides with the aem
Now define form factors for the correlation functions appearing above:
(em)
r
i
0|J5
(p)J
(k1 )J(em) (k2 )|0
= Ar1 k1 + Ar2 k2
+ Ar3 k1 k2 k2 + Ar4 k1 k2 k1
+ Ar5 k1 k2 k1 + Ar6 k1 k2 k2 ,
(3.6)
where the six form factors are functions of the invariant momenta, i.e., Ari = Ari (p 2 , k12 , k22 ),
(em)
Mrt
0|5t (p)J (k1 )J(em) (k2 )|0 = D r p 2 , k12 , k22 k1 k2 ,
(3.7)

(em)
6
0|Q(p)J (k1 )J(em) (k2 )|0 = B p 2 , k12 , k22 k1 k2 .
(3.8)
With these definitions, it follows immediately from Eq. (2.13) that the second element of
the sum rule, i.e., the form factor R r,1 is just

R r,1 K 2 = 4 Ar1 Ar2 K 2

1 (r)
,
= 4 D r K 2 + r0 B K 2
(3.9)
a
8 2
where we have used the notation D r (K 2 ) = D r (0, k 2 , k 2 ), etc. The non-perturbative QCD
dynamics governing the first moment sum rule is therefore encoded in these 3-current form
factors.
In the next section, we discuss the K 2 -dependence of the sum rule in terms of these
form factors. However, without any detailed knowledge of their non-perturbative features,
we can already determine the first moment of g1 in the limit K 2 = 0, corresponding to real
photons, and in the asymptotic region K 2 m2 .
The first observation is that electromagnetic current conservation requires R r,1 (K 2 ) to
vanish at K 2 = 0. Substituting the form factor expansion (3.6) into the Ward identity (3.3),
and taking p = 0, we find

1
2
p k12 k22 ,
2

r
r 2
r1
A2 = A4 k1 + A6 p 2 k12 k22 .
2
Ar1 = Ar3 k22 + Ar5
(3.10)
419
Provided none of the form factors have singularities at p 2 = 0 (or K 2 = 0), as is the case
away from the chiral limit, then it follows immediately that both Ar1 (K 2 ) and Ar2 (K 2 ) are
of O(K 2 ) for small photon virtuality. (The chiral limit is subtle and is discussed in detail
in Ref. [23].) We therefore establish R r,1 (0) = 0 and therefore [1,21,22]
1
dx g1 x, Q2 ; K 2 = 0 = 0.
(3.11)
Next, we consider the asymptotic limit of large K 2 , while still keeping in the DIS regime
of K 2 Q2 . For this, we need the large K 2 limit of the form factors Ari (K 2 ), D r (K 2 ) and
B(K 2 ), which can be obtained using the renormalisation group. The flavour non-singlet
(r = 3, 8) and singlet (r = 0) cases are different. In the non-singlet case, since the axial
current is conserved, the form factors Ari (K 2 ) satisfy a homogeneous RGE (see Ref. [1]
for explicit details), with the standard solution

Ari K 2 ; s (); m = Ari 2 ; s (t); et m(t) ,
(3.12)
2
(contrast with the t in the Wilson coeffiwhere is an RG reference scale, t = 12 ln K

2
cient expressions, which refers to the scale Q2 ), s (t) and m(t) are running couplings and
m generically denotes the individual quark masses mr . The large K 2 limit of Ar1 Ar2 is
therefore obtained from the correlation function evaluated at weak coupling in the chiral
limit. In this limit, D r is clearly zero, so recalling Eq. (3.9), we conclude that in the flavour
non-singlet sector,
R r,1 K 2 = a (r)
(3.13)
(r = 3, 8).
2
The asymptotic value of the form factor is therefore determined by the electromagnetic
UA (1) anomaly coefficient.
In the flavour singlet case, however, the 3-current correlation function satisfies an inhomogeneous RGE with anomalous dimension because of the anomalous non-conservation
of the singlet axial current. In this case, therefore,
A0i
t

0
2

K ; s (); m = exp dt s (t ) Ai ; s (t); et m(t) .
(3.14)
Here, we need both the form factors D 0 and F 0 to evaluate the r.h.s. At weak coupling,
the correlation function involving the topological charge Q is of O(s2 ), and so contributes
only at the same order as other neglected terms. So once again, the asymptotic limit is controlled simply by the anomaly coefficient. However, this time we also need the anomalous
dimension term, and the final result is
t

2
1 (0)

exp dt s (t ) .
R0,1 K = a
(3.15)
2
420
The asymptotic form for the g1 sum rule is finally obtained by putting together
Eqs. (3.13), (3.15) for the form factors with Eqs. (3.1), (3.2) for the Wilson coefficients.
This gives:
1
dx g1 x, Q2 ; K 2

E1r,1 Q2 R r,1 K 2
(3.16)
r=3,8,0
t(Q)

1
s (Q2 )
(3) (3)
(8) (8)
(0) (0)

=
1
dt s (t )
,
c a + c a + c a exp
2
t (K)
Q2
,
2
K2
(3.17)
s (t)
4
t (K) = 12 ln 2 . Substituting
10 t for the
with the obvious notation t (Q) = 12 ln
running couplings and reorganising terms, we obtain the final form of the sum rule:
1
dx g1 x, Q2 ; K 2

4
1
1
16
1
2
.
1
+
(3.18)
3
9 ln Q2 /2 81 ln Q2 /2 ln K 2 /2

Notice that the overall normalisation factor is Nc f ef4 , proportional to the sum of the
fourth power of the quark charges ef , corresponding to the lowest order box diagram con
tributing to g1 .
The key physics in Eqs. (3.17), (3.18) is that the first moment of g1 (x, Q2 ; K 2 ) in
2
2
2
the asymptotic limit m K Q for the target photon virtuality is governed by the
quark charges, with a flavour-dependence reflecting the electromagnetic UA (1) anomaly
coefficients. The approach to this asymptotic value depends on logarithmic corrections
given by the anomalous dimension arising from the gluonic UA (1) anomaly in the flavour
singlet current.
In between the limits K 2 = 0 and K 2 m2 , the sum rule depends on form factors
F r (K 2 ) as follows (substituting for c(r) and a (r) compared to Eq. (3.17)):
=
1
0
dx g1 x, Q2 ; K 2

s (Q2 )
1
1
=
18

t(Q)

3F 3 K 2 + F 8 K 2 + 8F 0 K 2 ; 2 = K 2 exp
dt s (t )
.
t (K)
(3.19)
421
The form factors F r (K 2 ) interpolate between 0 for K 2 = 0 and 1 for asymptotically large
K 2 . Notice that in the anomalous singlet sector, we have to specify the renormalisation
scalewith the anomalous dimension factor as shown, the form factor F 0 (K 2 ) must be
evaluated at 2 = K 2 .
These form factors are simply written in terms of those defined in Eqs. (3.6)(3.8).
Explicitly,

2

1 (r) 1
r
r
A1 Ar2 K 2
a
F K =
8 2

1 (r) 1 r
2
a
D K + r0 B K 2 .
=1
(3.20)
2
8
The full K 2 -dependence of the sum rule is therefore governed by the AVV correlation
function, or alternatively, the correlation functions in the corresponding Ward identity. The
sum rule therefore gives an experimental measure of these correlators, which are sensitive
to the realisation of chiral symmetry in QCD [23] and the gluonic UA (1) anomaly. A first
principles calculation of these correlation functions in QCD, if it were possible, would
therefore give a complete prediction for the first moment sum rule.
4. UA (1) PCAC, , , , and the gluon topological susceptibility
In order to gain some more insight into the non-perturbative behaviour of the first mo
ment of g1 (x, Q2 ; K 2 ), we can use the ideas of PCAC and spontaneously broken chiral
symmetry to rewrite the sum rule in terms of the off-shell couplings for radiative decays of
the pseudo-Goldstone bosons , and , since these are also controlled by the AVV correlation function in QCD. Of course, the gluonic anomaly makes the application of PCAC
to the UA (1) sector both interesting and subtle. In particular, the sum rule is sensitive to
the gluon topological susceptibility, which plays a key role in many polarisation-dependent
phenomena in QCD.
The link between the AVV correlation function and radiative pseudoscalar decays arises
by writing the close analogue of the axial Ward identity (3.5) involving photon states:
r
| = Mrt
0|5t | + 6r0
0|Q| + a (r)
0|F F | , (4.1)
ip
0|J5
8
and assuming pseudoscalar dominance of the matrix elements to rewrite them in terms
of the radiative couplings g , g and g . However, because of the anomaly, the
relation of the operators 5r and Q to the physical pseudoscalars , , is not entirely
straightforward and it is best to make a change of variables to operators which are more
appropriate as interpolating fields for the pseudoscalar particles. This approach to UA (1)
PCAC and radiative pseudoscalar decays is described in detail in Refs. [2527]. See also
Ref. [28] for an analysis of experimental values for the various decay constants which arise.
The result is the following set of expressions for the form factors F r (K 2 ) in terms of
the off-shell radiative pseudoscalar couplings for photon virtuality K 2 :
1
f g K 2 ,
F 3 K2 = 1
422

1 1 8

F K =1
f g K 2 + f 8 g K 2 ,
3
1

0

F 0 K 2 ; 2 = 1 4
f g K 2 + f 0 g K 2 + 6AgG K 2 ; 2 .
(4.2)
We now discuss what insight this new representation gives into the momentumdependence of the form factors F r (K 2 ). The first striking observation is that the first
moment of g1 (x, Q2 ; K 2 ) for an off-shell photon target involves the gluon topological
susceptibility, as is characteristic of many polarisation phenomena in QCD. This arises
through the dependence of the singlet form factor F 0 (K 2 ) on the non-perturbative constant
A which controls the topological susceptibility. For non-vanishing quark masses [34]:

1

1
(0)
QQ = A 1 A
(4.3)
.
mq
qq
q
8
The corresponding radiative coupling gG has a clear theoretical interpretation as the

coupling of the two-photon state to a glueball-like operator G orthogonal to the physical
. It does not, however, necessarily refer to a physical particle state (see, e.g., Refs. [25
28] for a further discussion), so we do not have a clear intuition about its momentumdependence.
Although these anomalous contributions are interesting from a theoretical perspective,
in practice they may not be so significant for the sum rule. Arguments based on the 1/Nc
expansion or OZI rule, carefully applied to the flavour singlet channel, suggest that the
contribution of the 6AgG term on the l.h.s. of Eq. (4.2) is subdominant. An explicit fit
[28] of the decay constants and couplings in Eqs. (4.2) indeed confirms that, for on-shell
photons, the relative contribution of this term is around 20%.
The main result implied by Eqs. (4.2) is that the momentum-dependence of the form factors in the sum rule is determined by the non-perturbative couplings g (K 2 ), g (K 2 )
and g (K 2 ). The relevant mass scale determining the crossover from F r (0) = 0 to
F r () = 1 is therefore given by the non-perturbative scale in the photon channel of the
pseudoscalar radiative coupling. Well-established ideas invoking vector meson dominance
(VMD) imply that for F 3 (K 2 ) this scale is m2 (equivalently m2 , m2 for the other flavours).
Given that QCD spontaneously breaks chiral symmetry, we therefore expect that the form
factors interpolate smoothly between 0 and 1 with a crossover scale characterised by the
vector meson masses. This is in sharp contrast to a perturbative QCD picture, in which
this scale would correspond instead to the light quark masses [35]. (See Ref. [23] for an
extensive discussion of this point.2 ) The K 2 -dependence of the first moment sum rule for
g1 (x, Q2 ; K 2 ) is therefore a clear signal of chiral symmetry breaking.

We can try to justify this VMD prediction directly from QCD field theory as follows.
(See also the closely related analyses of the AVV correlation functions and radiative and
2 Of course, for final states characteristic of heavy quarks (c, b) with mass >
QCD , the crossover scale would
simply be the quark mass itself, m2c,b . Similarly, for leptonic final states which probe the QED structure of the
photon.
423
leptonic pseudoscalar decays in Refs. [3638] (and references therein). Comprehensive

reviews of relevant QCD sum rule results may be found in Refs. [39,40].) Once we have
related the form factors to the pseudoscalar radiative couplings, we can use the OPE for the
two electromagnetic currents in, for example, the matrix element
|J(em) (k)J(em) (k)|0
for large K 2 (compare Eq. (2.9)) to write
1 3,1 2
3
(4.4)
E (K )
|J5
(0)|0 + .
K2 1
Comparing with the definitions of the form factors and couplings, we find to leading order,
(em)
|J
(k)J(em) (k)|0 = 2 k

(4)2 2 1
F 3 K2 = 1
(4.5)
f 2 + .
3
K
Comparing this large K 2 behaviour with a simple interpolation formula such as F 3 (K 2 )
K 2 /(K 2 + M 2 ), we would identify the characteristic crossover mass scale as M 2
(4)2 2
2
3 f , which is numerically m . This estimate is therefore consistent with the VMD
picture.
This completes our discussion of the non-perturbative QCD dynamics behind the
momentum-dependence of the g1 (x, Q2 ; K 2 ) sum rule. In the next section, we move on

to discuss the experimental question of whether all this can actually be directly measured
in DIS experiments at e+ e colliders.
5. Cross-sections and spin asymmetries at the ILC and SuperKEKB

The spin-dependent cross-sections for the two-photon DIS process e+ e e+ e X
defined in Section 2 are given in Ref. [1] as
Q2
1
4
K2
x max
x max
a 2 log min
log max
log emin log
2
2
2 Qmin
xe
xe
Kmin
(5.1)
and
2
Q2
Q2
Kmax
xemax
x max
4 1
.
log
log
log
log
b log max
2
2 s
xe
2
xemin
Q2min
Kmin
(5.2)
In these expressions, we have included experimental cuts on the maximum and minimum
values of the kinematical variables Q2 , K 2 , xe and x.
Q2 is the geometric mean of Q2max
and Q2min (similarly for
xe ). The constants a and b are approximations to the functions
a(x) and b(x) given by inverse Mellin transforms of the moments an and bn corresponding
to the higher spin operators in the OPEs for F2 and g1 . Numerically, a b 1.5.
The spin asymmetry is therefore

Q2min 1
1 Q2min
Q2max
Q2max
log 2

log 2
(5.3)
1 + log
.
2 s
2
Qmin
In order to extract information on the g1 structure function from

, we need this spin
asymmetry
to be large. More precisely, for a statistically significant result, we require
/ 1/ L , where L is the integrated luminosity [1].
424
Experimentally, the accelerator design specifies the CM energy s and luminosity L, but
we can then choose the cuts on the kinematic variables, subject of course to detector constraints, in order to maximise the measured cross sections and spin asymmetries necessary
to determine g1 . The relevant cuts are on Q2 , K 2 , e and . The upper cut on Q2 is limited
by the detector acceptance and we take Q2max s/4. For the lower cut, Q2min , we have to
be within the DIS region but otherwise will keep this as a free parameter to be varied to
try and obtain the most statistically significant measurement. For K 2 , we set a lower cut
2 = 1 GeV2 in the total
at K 2 m2e and vary to an upper limit well above m2 , taking Kmax
1
2
2
2
cross-section estimates. Since = 2 (Q + W ), and Wmin is small, we choose the following cuts on the Bjorken variables: emin = min = 12 Q2min and emax = max = 12 s. Inserting
these cuts into the formulae for the total cross-section and spin asymmetry, we have

2
Q2min
s
8 1
0.5 10
(5.4)
log 2 log 2
Q2min
Qmin
and

Q2min 1
1 Q2min
s
s
log 2
=
log
1 + log
.
2 s
42
4Q2min
(5.5)
As noted in Section 2, the nth x-moment of g1 is determined by the (n+1)st xe -moment

of
. In particular, from Eq. (2.8) we have
1
0
d 3
1
3
dxe xe
= 3 2 2
2
2
2 sQ K
dQ dxe dK
1
dx g1 x, Q2 ; K 2 .
(5.6)
Comparing with the first moment sum rule (3.19), we can therefore determine the form
factors F r (K 2 ) if we can measure the K 2 -dependence of the fully differential cross-section
d 3
/dQ2 dxe dK 2 .
We now discuss whether such measurements are feasible at present and future e+ e colliders. For this purpose, we consider two accelerators in detailthe International Linear
Collider (ILC) and the proposed high-luminosity B factory SuperKEKB. The contrasting
machine parameters illustrate clearly the main issues involved in measuring the first moment sum rule.
At the time the sum rule was proposed in Ref. [1], the luminosity available from the
then current accelerators was inadequate for measuring the sum rule. For example, for a
polarised version of LEP operating at s = 104 GeV2 with an annual integrated luminosity of L = 100 pb1 , and choosing the cut at Q2min = 10 GeV2 , we find 35 pb and
/ 0.01. However, thecorresponding annual event rate would be 3.5 103 and the
statistical significance only L
/ 0.5, so even a reliable measurement of the spin
asymmetry could not be made.
Clearly, a hugely increased luminosity is required and this has now become available
with proposals for machines with projected annual integrated luminosities measured in
inverse attobarns. However, as noted in Ref. [1], if this increased luminosity is associated
with increased CM energy, then the 1/s factor in the spin asymmetry (5.2) sharply reduces
425
(a)
(b)
Fig. 2. The first graph (a) shows the fall in total cross-section (in pb) at the ILC as the experimental cut Q2min
is varied from 1 to 100 GeV2 . The second graph (b) shows the spin asymmetry
/ rising over the same range
of Q2min .
the possibility of extracting a measurement of g1 . For this reason, it was already recognised
in Ref. [1] that the best future colliders for studying the sum rule would be high-luminosity
B factories.
Now consider the accelerator parameters of the ILC. This will operate initially at a CM
energy of 500 GeV (s = 2.5 105 GeV2 ) with a projected luminosity of 1034 cm2 s1 ,
corresponding to an annual integrated luminosity of 0.1 ab1 [8,9]. For simplicity, we
assume here that this luminosity could be achieved with the ILC running with polarised
beams. We now analyse how to optimise a measurement of g1 by varying the experimental

cuts, in particular Q2min . In Fig. 2, we have plotted the cross-section (in pb) and spin
asymmetry
/ for values of Q2min varying from 1 to 100 GeV2 . The sharp fall-off in
the cross-section is determined by the 1/Q2min factor in Eq. (5.4). On the other hand, in
this range, the spin asymmetry rises with increasing Q2min because of the corresponding
factor in Eq. (5.5). The absolute value of
/ is kept relatively small because of the
(1/s)-dependence of the spin asymmetry on the CM energy. Optimising the cut on Q2min
is therefore a balance between
keeping the total event rate high and maximising the spin
asymmetry. If we plot L
/ for this same range of Q2min (Fig. 3), we see that it
rises monotonicallyin fact it reaches a maximum only at Q2min 103 GeV2 where the
cross-section has fallen to a mere 0.5 pb.
426
Fig. 3. Plot of the statistical measure
L
/ for Q2min between 1 and 100 GeV2 at the ILC.
As a reasonable compromise between event rate and spin asymmetry, we could choose
to take the cut at Q2min 50 GeV2 . This corresponds to 15 pb and
/ 0.002. The
annual event rate is 1.5 106 with L

/ 3, which would allow a measurement of
the spin asymmetry itself. However, as we see from Eq. (2.8), this determines only the
n = 0 moment of g1 , integrated over K 2 . A detailed study of the first moment sum rule
itself would require a much greater
/ .
This leads us to consider instead the new generation of ultra-high luminosity e+ e colliders. Although these are envisaged as B factories, these colliders operating with polarised
beams would, as we now show, be extremely valuable for studying polarisation phenomena in QCD. As an example of this class, we take the proposed SuperKEKB collider. (The
analysis for PEPII is very similar, the main difference being the additional ten-fold increase
in luminosity in the current SuperKEKB proposals.)
SuperKEKB is an asymmetric e+ e collider with s = 132 GeV2 , corresponding to
electron and positron beams of 8 and 3.5 GeV, respectively. The design luminosity is 5
1035 cm2 s1 , which gives an annual integrated luminosity of 5 ab1 [10]. To see the
effects of the experimental cut on Q2min in this case, we again plot the total cross-section
and the spin asymmetry in Fig. 4, this time for the range of Q2min from 1 to 10 GeV2 .
As before, in this range is falling like 1/Q2min while
/ rises to what is actually a
maximum at 1/Q2min = 10 GeV2 . On the other hand, the statistical significance L

/
falls monotonically, though is orders of magnitude improved on the corresponding plot for
even the ILC (Fig. 5).
Taking Q2min = 5 GeV2 , we find 12.5 pb with spin asymmetry
/ 0.1. The
annual event rate is therefore 6.25 107 , with L

/ 750. This combination of a
very high event rate and the large 10% spin asymmetry means that SuperKEKB has the
potential not only to measure

but to access the full first moment sum rule for g1 itself.
1
Recall from Eq. (5.6) that to measure 0 dx g1 (x, Q2 ; K 2 ) we need not just
but the
fully differential cross-section w.r.t. not only xe and Q2 , but also K 2 if the interesting
non-perturbative QCD physics is to be accessed. To measure this, we need to divide the
data into sufficiently fine K 2 bins in order to plot the explicit K 2 -dependence of g1 , while
still maintaining the statistical significance of the asymmetry. The ultra-high luminosity of
SuperKEKB ensures that the event rate is sufficient, while its moderate CM energy means
that the crucial spin asymmetry is not overly suppressed by its (1/s)-dependence.
427
(a)
(b)
Fig. 4. The first graph (a) shows the total cross-section (in pb) at SuperKEKB as the experimental cut Q2min is
varied from 1 to 10 GeV2 . The second graph (b) shows the spin asymmetry
/ over the same range of Q2min .
Fig. 5. Plot of the statistical measure
L
/ for Q2min between 1 and 10 GeV2 at SuperKEKB.
Our conclusion is that the new generation of ultra-high luminosity, moderate energy
e+ e colliders, currently conceived as B factories, could also be uniquely sensitive to
important QCD physics if run with polarised beams. In particular, they appear to be the
only accelerators capable of accessing the full physics content of the sum rule for the first
moment of the polarised structure function g1 (x, Q2 ; K 2 ). The richness of this physics, in
particular the realisation of chiral symmetry breaking, the manifestations of the axial UA (1)
anomaly and the role of non-perturbative gluon dynamics, provides a strong motivation
428
for giving serious consideration to an attempt to measure the g1 sum rule at these new
colliders.
Acknowledgements
I would like to thank S. Narison and G. Veneziano for their collaboration on Ref. [1]
and their helpful comments on this paper.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
S. Narison, G.M. Shore, G. Veneziano, Nucl. Phys. B 391 (1993) 69.

G.M. Shore, G. Veneziano, Phys. Lett. B 244 (1990) 75.
G.M. Shore, G. Veneziano, Nucl. Phys. B 381 (1992) 23.
D. de Florian, G.M. Shore, G. Veneziano, Target fragmentation at polarized HERA: a test of universal
topological charge screening in QCD, in: Proceedings of Conference Physics With Polarized Protons at
HERA, DESY, Hamburg, 1997, pp. 194201, hep-ph/9711353.
F. Richard, et al., TESLA: The superconducting electronpositron linear collider with an integrated X-ray
laser laboratory, Technical Design Report, Part I, hep-ph/0106314.
M. Woods, et al., Luminosity, energy and polarization studies for the linear collider, in: Proceedings of
5th International Workshop on ElectronElectron Interactions at TeV Energies, Santa Cruz, CA, 2003,
physics/0403037.
A.G. Akeroyd, et al., SuperKEKB Physics Working Group, Physics at super B factory, hep-ex/0406071.
Ch. Berger, W. Wagner, Phys. Rep. 146 (1987) 1.
R. Nisius, Phys. Rep. 332 (2000) 165.
M. Stratmann, Nucl. Phys. B (Proc. Suppl.) 82 (2000) 400.
M. Stratmann, W. Vogelsang, Phys. Lett. B 386 (1996) 370.
J. Kwiecinski, B. Ziaja, Nucl. Instrum. Methods A 472 (2001) 229.
M. Glck, E. Reya, C. Sieg, Phys. Lett. B 503 (2001) 285;
M. Glck, E. Reya, C. Sieg, Eur. Phys. J. C 20 (2001) 271.
H. Baba, K. Sasaki, T. Uematsu, Phys. Rev. D 68 (2003) 054025.
T. Uematsu, Nucl. Phys. B (Proc. Suppl.) 116 (2003) 136.
K. Sasaki, T. Uematsu, Nucl. Phys. B (Proc. Suppl.) 135 (2004) 178.
S.D. Bass, The spin structure of the proton, hep-ph/0411005.
S.D. Bass, Int. J. Mod. Phys. A 7 (1992) 6039.
S.D. Bass, S.J. Brodsky, I. Schmidt, Phys. Lett. B 437 (1998) 417.
G.M. Shore, G. Veneziano, Mod. Phys. Lett. A 8 (1993) 373.
G.M. Shore, Nucl. Phys. B 569 (2000) 107.
G.M. Shore, Nucl. Phys. B (Proc. Suppl.) 86 (2000) 368.
G.M. Shore, Phys. Scr., T 99 (2002) 84.
G.M. Shore, Pseudoscalar meson decay constants and the topological susceptibility, SWAT 04-418, in preparation.
E. Witten, Nucl. Phys. B 156 (1979) 269.
G. Veneziano, Nucl. Phys. B 159 (1979) 213.
P. Herrera-Siklody, J.I. Latorre, P. Pascual, J. Taron, Nucl. Phys. B 497 (1997) 345;
P. Herrera-Siklody, J.I. Latorre, P. Pascual, J. Taron, Phys. Lett. B 419 (1998) 326.
R. Kaiser, H. Leutwyler, Eur. Phys. J. C 17 (2000) 623.
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
T. Feldmann, Int. J. Mod. Phys. A 15 (2000) 159.

P. Di Vecchia, G. Veneziano, Nucl. Phys. B 171 (1980) 253.
R.D. Carlitz, J.C. Collins, A.H. Mueller, Phys. Lett. B 214 (1988) 229.
S. Narison, N. Paver, Z. Phys. C 22 (1984) 69.
M. Knecht, S. Peris, M. Perrottet, E. de Rafael, Phys. Rev. Lett. 83 (1999) 5230.
P.D. Ruiz-Femenia, A. Pich, J. Portols, Nucl. Phys. B (Proc. Suppl.) 133 (2004) 215.
S. Narison, QCD Spectral Sum Rules, World Scientific, Singapore, 1984.
S. Narison, QCD as a Theory of Hadrons, Cambridge Univ. Press, Cambridge, 2004.
429
Nuclear Physics B 712 [FS] (2005) 433512
D-branes and BCFT in Hpp-wave backgrounds

Giuseppe DAppollonio a,b , Elias Kiritsis c,d
a Department of Mathematics, Kings College, The Strand, London WC2R 2LS, UK
b LPTHE, Universit Paris VI, 4 pl Jussieu, 75252 Paris cedex 05, France
c CPHT, Ecole Polytechnique, UMR du CNRS 7644, 91128 Palaiseau, France
d Department of Physics, University of Crete, 71003 Heraklion, Greece
Abstract
In this paper we study two classes of symmetric D-branes in the NappiWitten gravitational wave,
namely D2- and S1-branes. We solve the sewing constraints and determine the bulk-boundary couplings and the boundary three-point couplings. For the D2-brane our solution gives the first explicit
results for the structure constants of the twisted symmetric branes in a WZW model. We also compute the boundary four-point functions, providing examples of open string four-point amplitudes in
a curved background. We finally discuss the annulus amplitudes, the relation with branes in AdS3
and in S 3 and the analogy between the open string couplings in the H4 model and the couplings for
magnetized and intersecting branes.
PACS: 11.23.-w; 11.25.Hf
1. Introduction
The study of gravitational waves as string theory backgrounds began more than fifteen
years ago [17]. They were proposed as the most convenient starting point for extending
the analysis of the properties of string theory from the familiar vacua given by the product
of flat space and a compact manifold to the less explored curved, non-compact space
E-mail addresses: giuseppe@mth.kcl.ac.uk (G. DAppollonio), kiritsis@cpht.polytechnique.fr (E. Kiritsis).
doi:10.1016/j.nuclphysb.2005.01.020
434
G. DAppollonio, E. Kiritsis / Nuclear Physics B 712 [FS] (2005) 433512
times. The main reason was that already from the point of view of general relativity the
gravitational waves are some of the simplest time-dependent backgrounds. They admit
a covariantly constant null Killing vector, most of their curvature invariants vanish and
there is no particle creation. Another distinctive feature, which is particularly relevant for
string theory, is that it is always possible to fix the light-cone gauge for the quantization of
the world-sheet action. Recently it was also realized that the gravitational waves play an
important role in the study of gauge/string duality.
Motivated by the notion of Penrose limits [810], it was argued [11] that such backgrounds are dual to modified large-N limits of gauge theories. This observation has opened
new avenues in understanding stringy aspects of the gauge theory/string theory duality. Indeed, the GreenSchwarz action for the pp-wave that is obtained by the Penrose limit of
AdS5 S 5 , can be quantized in the light-cone gauge, even though there is a non-trivial
RR flux [12,13]. A lot of progress has been made since, and this is reviewed in [1417]. It
should be noted however that at the moment there is no generally accepted theory of the full
holographic correspondence, although several proposals have been put forward [1822].
Given the relevance for both the study of string theory in non-compact curved space
times and the gauge theory/string theory duality, we believe that it is important to obtain a
clear and detailed understanding of the string dynamics at least in some particular gravitational wave backgrounds. With this aim in mind, it is natural that our choice falls upon a
class of gravitational waves that are supported by a NSNS flux and have an exact CFT description as WZW models. This is the class of the WZW models based on the Heisenberg
groups H2n+2 , n 1. The first example in this family was discovered by Nappi and Witten
[7] and the others were introduced in [23,24]. These models, unlike those with the same
metric but supported by a RR flux, can be quantized in a covariant way using standard
perturbative string theory techniques. The presence of the affine symmetry algebra then
imposes additional constraints that can lead to the complete solution of the model.1 Since
the study of string and brane dynamics in non-compact curved backgrounds is a difficult
arena, all models that can be solved exactly are a source of useful information. Unfortunately, only very few examples are available and they essentially amount to the Liouville
model [2729], to AdS3 [30,31] and its cosets [3234].
In [35] we added another entry to this list, solving the H4 model, which describes the
propagation of a string in a four-dimensional gravitational wave. The structure of the closed
string spectrum turned out to be very similar to the one established for AdS3 [31]. It can be
organized in highest-weight and spectral-flowed representations of the affine H 4 algebra
and there are two distinct classes of states. For generic values of the light-cone momentum
p, the states belong to the discrete representations of the H 4 algebra. They correspond
to short strings that are confined by the background fields in closed orbits in the plane
transverse to the two light-cone coordinates. Whenever p Z, with a parameter of
the pp-wave metric, the states belong to the continuous representations of the H 4 algebra
and correspond to long strings that move freely in the transverse plane.
In [35], we computed all the three- and four-point correlation functions of primary vertex operators, thus providing all the structure constants for this non-compact WZW model.
1 See [25,26] for a study of other interesting pp-waves without affine symmetry algebras.
435
We also showed explicitly that the spectral-flowed representations are necessary for the
consistency of the model since they appear in the intermediate channel of four-point amplitudes with external highest-weight states. In [36] we performed a similar analysis for
the H6 model. This model already displays all the new features of the higher-dimensional
cases, namely the existence of enhanced symmetry points and the necessity of introducing
representations that satisfy a modified highest-weight condition, which generalize the concept of spectral-flowed representations. The H6 model is also relevant for the AdS3 /CFT 2
correspondence, being the Penrose limit of AdS3 S 3 .
As conformal field theories, the H4 model and its higher-dimensional versions deserve
attention not only because of the rich and interesting structure we have just outlined but
also because there are several relations between them and other important models. Most of
these relations follow directly from the original idea of Penrose of considering the gravitational waves as limits of other spacetimes. From the point of view of the world-sheet
-model, the Penrose limit that connects two backgrounds both having an exact description as WZW models can be interpreted as a contraction of the underlying current algebra
[37]. As such, the H4 model captures the limiting behavior of backgrounds of the form
R S 3 and AdS3 S 1 . In [35] we analyzed the contraction of R SU(2)k to H4 and in
[36] the contraction of SL(2, R)k SU(2)k to H6 .
The structure of the algebra changes drastically in the contraction process. The structure of the spacetime is also drastically changing during the Penrose limit. Despite this,
we have shown that it is possible to take the limit of the CFT operators and of the dynamical
quantities such as the correlation functions in a controlled way. Another interesting relation stems from the free-field realization of the H4 model introduced in [38,39]. The H4
primary vertex operators can be represented using the twist fields of the orbifold obtained
as the quotient of the plane by a rotation and a dictionary can be established connecting
the amplitudes computed in the H4 model and the amplitudes computed in the orbifold
CFT [35]. Very similar correlators arise in the study of closed strings in Misner space, the
quotient of a two-dimensional spacetime by a boost and were discussed in [40,41]. In
fact, using the formal analogy with open strings in a constant electric field, it was argued
in [40,41] that the twisted sectors of this Lorentzian orbifold contain physical scattering
states whose condensation could be relevant for the resolution of the space-like singularity.
In this paper we complete our analysis of the H4 model by studying D-branes in the
NappiWitten gravitational wave and the dynamics of their open string excitations. The
dynamics of open strings in curved spacetime is even less understood than its closed
counterpart and again we have at our disposal a very limited number of exactly solved
examples. What is typically accessible, are the boundary states that have been studied for
the Liouville branes [42,43], for branes in AdS3 [4446] and for the 2d black hole [47,
48]. For all the other quantities such as the bulk-boundary and the three-point boundary
couplings only partial results exist [46] and their computation proved to be an extremely
difficult task. The only exception is the Liouville model for which the complete solution is
available [42,43,4951]. In this paper we will provide the complete solution for the BCFT
pertaining to the two classes of symmetric branes of the H4 model.
D-branes in pp-wave backgrounds have already been the object of several studies and
we summarize here only the main results. D-branes in RR supported pp-waves have been
discussed in the light-cone gauge and various aspects of their physics have been analyzed
436
[5258]. Interesting world-volume theories have been argued to exist on such branes [59].
D-branes in NSNS supported pp-wave have also been studied, since they are relevant for
the Penrose limits of little string theory and, unlike the RR supported pp-waves, they are
amenable to study using boundary CFT methods [6066]. Our aim in this paper is not to
describe the most general brane configuration in the NappiWitten gravitational wave. We
focus instead on two particular classes of branes which preserve half of the background
isometries and we clarify the closed and open string dynamics in full detail.
The H4 model has two families of symmetric D-branes [60,61], namely D2- and S1branes. We solve in both cases the consistency BCFT conditions [78,79] and obtain the
BCFT data, that is, the bulk-boundary and the three-point boundary couplings. The bulkboundary couplings for the D2-branes can be found in Eqs. (5.1.6) and (5.1.9) while the
three-point boundary couplings are in Eqs. (5.1.12), (5.1.16), (5.1.22) and (5.1.24). The
bulk-boundary couplings for the S1-branes are in Eq. (5.2.7) while the boundary threepoint couplings can be found in Eqs. (5.2.22), (5.2.25) and (5.2.27). To our knowledge,
with the notable exception of the Liouville model [42,43,4951], this is the first complete
tree-level solution of D-brane dynamics in a curved non-compact background.
The D2-branes are the twisted symmetric branes of the H4 model. Their world-volume
covers the two light-cone directions and one direction in the transverse plane. The induced
metric is that of a pp-wave in one dimension less and they also carry a null electromagnetic
flux. As such, they provide an interesting example of curved branes in a curved spacetime.
The spectrum of open strings starting and ending on the same brane or stretched between
different branes contains all the representations of the H 4 algebra.
There are many similarities between the D2-branes in H4 and the AdS2 -branes in AdS3 .
This is not surprising since they can be considered as Penrose limits of specific branes in
AdS3 S 1 or in R S 3 . More precisely, if we start from R S 3 they arise from S 2 -branes
with Neumann boundary condition in time while if we start from AdS3 S 1 they arise
from the AdS2 -branes. The relation between the H4 vertex operators and the orbifold twist
fields leads in this case to an analogy between the D2-branes in the NappiWitten wave
and configurations of intersecting branes in flat space [67].
The S1-branes are the untwisted symmetric branes of the H4 model. They have Dirichlet
boundary conditions on the two light-cone coordinates and their world-volume covers the
transverse plane with an induced flat Euclidean metric. They are also supported by a worldvolume electric field whose magnitude determines their localization in one of the light-cone
coordinates. Having a non-trivial boundary condition along the real time direction they
are examples of S-branes [68]. In fact, the Penrose limit relates the S1-branes to either
S 2 -branes with a Dirichlet boundary condition in time in R S 3 or to the H2 -branes in
AdS3 S 1 .
The world-volume of the S1-branes shrinks to a point whenever their light-cone position is given by u = 2n. For these values of the coordinate u, there is another class of
symmetric branes with a cylindrical world-volume. These branes extend along the lightcone direction v and have a fixed radius in the transverse plane. We will not discuss them
in detail in this paper. We also mention that the S1-branes are a special case of a more
general class of non-symmetric branes that we discuss from the DBI point of view. Finally,
the behavior of the open strings attached to the S1-branes is very similar to the behavior of
open strings ending on magnetized branes in flat space [69].
437
Our solution of the H4 model with boundary should be useful not only to improve our
understanding of the closed and open string dynamics in curved spacetimes but also to
clarify some properties of both compact and non-compact WZW models. Indeed, as it is
widely appreciated by now, only when studied in the presence of a boundary a conformal
field theory reveals its full richness. Among other results we provide the first example of
structure constants for twisted symmetric branes in a WZW model (the D2-branes) and of
open four-point functions in a curved background.
While the physical interpretation of the amplitudes in the presence of the D2-branes is
straightforward, the interpretation of the amplitudes for the S1-branes is less evident, in
particular when they involve open string states stretched between different S-branes which
can be put on shell. They might play the role of boundary conditions at spatial infinity
but at fixed time (specified by the S-brane in question). In fact thinking of an S-brane as
a standard soliton supported by a scalar [68], we can imagine that they appear because
of the special initial state of the scalars. In this context, the open string insertions can be
interpreted as small variations on the initial data that creates the branes. It still remains
to be seen whether such a setup may be realized in a problem with physical interest and
whether the S-branes can be relevant for cosmology.
In this paper, we also discuss the annulus amplitudes. Our results, even though suggestive, are not conclusive and the relation between the open and the closed string channel of
the annulus certainly deserves further study.
The structure of this paper is the following: In Section 2 we review the geometry of the
symmetric branes of the H4 WZW model, first considered in [60,61]. We also discuss the
relationship between branes in the NappiWitten gravitational wave and branes in AdS3
S 1 and R S 3 . In Section 3 we evaluate the bulk-boundary couplings using the semiclassical wave functions. In Section 4 we discuss the spectrum of the boundary operators.
In Section 5 we solve the CardyLewellen constraints [78,79] and display the structure
constants of the boundary theory. In Section 6 we discuss the four-point amplitudes. In
Section 7 we analyze the annulus amplitudes and discuss the contraction of the R SU(2)k
WZW model with boundary. In Section 8 we analyze the physics of the branes in the
NappiWitten gravitational wave using the DiracBornInfeld action. Finally in Section 9
we suggest some interesting lines of further research. Several technical details are collected
in the appendices.
2. Branes in H4
The NappiWitten gravitational wave [7] is a curved homogeneous Lorentzian space.
The metric
2 r 2 2
du + dr 2 + r 2 d 2 ,
(2.1)
4
solves the Einstein equations with a constant null stress-energy tensor, provided in our case
by the 2-form field-strength
ds 2 = 2 du dv
H = r dr d du.
(2.2)
438
The light-cone and the radial coordinates are related to the Cartesian ones by u =
t+x
,
2
v=
and rei = + i . As given before, the metric is in the so-called Brinkman form.
The change of coordinates

2
u = x+,
y1 + y22 sin x + ,
v = x +
8
x +
x +
,
= y2 sin
,
= y1 sin
(2.3)
2
2
gives the metric in Rosen form
tx
ds 2 = 2 dx + dx + sin2

x + 2
dy1 + dy22 ,
2
x +
dy1 dy2 dx + .
(2.4)
2
In the following, both Brinkman and Rosen coordinates will be useful. The twodimensional -model that describes the propagation of a string in this background is a
WZW model based on the Heisenberg group H4 [7]. The commutation relations are
+

P , P = 2iK,
(2.5)
J, P = iP ,
H = sin2
where the generators J and K are anti-hermitian and (P + ) = P . Even though the group
is not semi-simple, there is a non-degenerate invariant symmetric form given by

2J, K = P + , P ,
(2.6)
which can be used to express the stress-energy tensor as a bilinear in the currents. For a
detailed discussion of this model we refer the reader to [35].
Since this gravitational wave is a WZW model, we can study in considerable detail the
symmetric branes, that is the branes that preserve a linear combination of the left and right
affine algebras. The symmetric branes fall in families which are in one-to-one correspondence with the automorphisms of the current algebra. Given such an automorphism ,
the relevant boundary CFT is defined by the following boundary conditions on the affine
currents

a
J (z) Ja (z) z=z = 0.
(2.7)
Equivalently we can introduce for each symmetric brane a boundary state |B that satisfies
a
a
Jm + Jm
(2.8)
|B = 0.
The geometry of the symmetric branes in a group manifold has a simple and elegant description. Their world-volume coincides with the (twisted) conjugacy classes [70,71]

Cg = G h1 gh, h G ,
(2.9)
where G is the group automorphism induced by . A generic automorphism can be
written as the composition of the adjoint action of a group element g0 and of an outer
automorphism
= Adg0 .
(2.10)
439
Since two families of branes that differ only in the choice of the inner automorphism Adg0
are related by the left action of the group, we can set without loss of generality g0 = 1 and
restrict our attention to families of branes associated with distinct outer automorphisms .
The H4 algebra admits a non-trivial outer automorphism which acts on the currents as
charge conjugation

P = P ,
(2.11)
(J ) = J,
(K) = K.
As such, we have two families of symmetric branes, wrapped on the conjugacy classes Cg
and on the twisted conjugacy classes Cg , respectively. In the following, we will briefly
review their geometry which was first discussed in [60,61].
The H4 conjugacy classes Cg (u, ) are characterized by two parameters, the constant
value of the coordinate u and the constant value of the invariant
u
r 2
cot
.
(2.12)
4
2
Their geometric description is particularly simple in Rosen coordinates where the brane
world-volume coincides with the wave-fronts since x = . These branes are thus Euclidean two-planes with an x + -dependent scale factor and a two-form field-strength
=v
sin x +
(2.13)
.
2
As usual, F is the gauge invariant combination that appears in the DiracBornInfeld
action. These branes have a non-trivial boundary condition on the real time coordinate and
can be called, following the modern terminology, S1-branes [68]. As we will show at the
end of this section, they are related by the Penrose limit to the H2 -branes in AdS3 S 1 or
to the S 2 -branes in R S 3 with a Dirichlet boundary condition along the time direction.
The brane world-volume degenerates to a point whenever x + = 2n, n Z. Indeed if
we start from x + = 2n and change the value of x + until we reach x + = + 2n, we
interpolate between a point-like world-volume and a flat, infinite two-dimensional worldvolume. This is very similar to what happens in flat space when we turn on a magnetic field
on a brane and send its field-strength to infinity. As we will show in more detail later, there
are several analogies between the untwisted symmetric branes of the H4 model and branes
in flat space with a magnetic field on the world-volume. In particular, the open strings
stretched between two S1-branes behave very similarly to the open strings in a magnetic
field [69].
In Brinkman coordinates, the metric induced on the two-dimensional world-volume is
trivial and the flux is
u
.
Fr = Br + 2 Fr = r cot
(2.14)
2
When 2n < u < + 2n we can parameterize the world-volume with (r, ) and consider the brane as a point-like object which appears at the point (u, v = , r = 0) and then
moves away along the x axis at the velocity of the light while simultaneously expanding
in a circle in the transverse plane according to Eq. (2.12). When + 2n < u < 2n we
have the time-reversed process, where an infinite circle coming from spatial infinity in the
F12 B12 + 2 F12 =
440
Table 1
u = 2 n
u = 2 n
u = 2 n
v, r = 0
r = 0
S1-branes
S(1)-branes
Null branes
transverse plane shrinks to a point. Finally when u = + 2n the brane is a two-plane

with a fixed light-cone position also in Brinkman coordinates.
When u = 2n, the geometry of the conjugacy classes changes. In Rosen coordinates
we notice that the two-dimensional planes degenerate to points. However, the transformation (2.3) is singular when u = 2n. Indeed, the analysis of the conjugacy classes shows
that there are other possibilities for the symmetric branes at u = 2n: when r = 0 we
have points with a fixed value of v and when r = 0 we have a cylinder of radius r extended
along the null direction v. The geometry of the conjugacy classes is summarized in Table 1.
In this paper we will often denote the two parameters that identify an S1-branes with
a single letter a (ua , a ). It would be interesting to identify all these branes as bound
states of some set of elementary branes. For instance, the D2-branes in S 3 were shown to
arise as bound states of D0-branes [72]. In a similar spirit and with similar techniques we
could choose as fundamental branes the point-like branes, that is the degenerate conjugacy
classes for u = 2n and try to identify the S1-branes and the cylindrical branes as bound
states of them.
The second class of symmetric branes we are interested in, wrap the twisted conjugacy
classes Cg () which are parameterized by a single invariant
= r cos .
(2.15)
In Brinkman coordinates, they have a simple description as D2-branes whose worldvolume extends along the (u, v, ) directions and is localized in the direction. They have
a non-trivial induced metric, which describes a gravitational wave, and a null world-volume
flux Fu =
2 . The D2-branes of the H4 model thus provide an interesting example of
curved branes in a curved spacetime.
The NappiWitten gravitational wave is the Penrose limit of two simple and interesting
spacetimes, RS 3 and AdS3 S 1 . For both spacetimes there is an exact CFT description
in terms of WZW models based respectively on R SU(2)k and SL(2, R)k U (1), where
the level k is related to the radius of curvature. From the CFT point of view, the existence
of the Penrose limit relating the NappiWitten wave and R S 3 or AdS3 S 1 corresponds
to the fact that the H4 current algebra is a contraction of the current algebras underlying
the two original spacetimes [37].
We close this section by discussing the relation implied by the Penrose limit between
the symmetric branes of the H4 WZW model and the symmetric branes in R SU(2)k and
in SL(2, R)k U (1). The latter branes have been studied in [70,71,73,74]. In the first case,
we will see that the S1-branes arise from the S 2 -branes in S 3 with a Dirichlet boundary
condition in the time direction and that the D2-branes arise from rotated S 2 -branes in S 3
with a Neumann boundary condition in the time direction. In the second case, we will
identify the D2-branes as the limit of the AdS2 -branes in AdS3 with a Neumann boundary
condition in S 1 - and the S1-branes as the limit of the H2 -branes in AdS3 with a Dirichlet
441
boundary condition in S 1 . A similar discussion of the Penrose limit applied simultaneously

to the spacetime and to a brane contained in it can be found in [75].
We start with R S 3 . Using the following standard parametrization for the SU(2) group
manifold

cos ei i sin ei
g(, , ) =
(2.16)
,
i sin ei cos ei
the metric and the two-form field-strength read

ds 2 = k dt 2 + cos2 d 2 + d 2 + sin2 d 2 ,
H = k sin 2 .
(2.17)
Along the time direction we can impose either Neumann or Dirichlet boundary conditions. As for the symmetric branes in S 3 , they wrap the SU(2) conjugacy classes which are
two-spheres characterized by a constant value of tr(g) = 2 cos cos . Since SU(2) does
not have any external automorphism, the other possible symmetric branes are two-spheres
) SU(2) and characterized by a conshifted by the action of a group element R(,
,
stant value of tr(Rg). In order to take the Penrose limit we first perform the change of
variables
=
u 2v
,
2
k
t=
u
,
2
r
= ,
k
= ,
(2.18)
and then send k . In the limit

cos cos cos
u
u 2
+
sin
.
2
k
2
(2.19)
We observe that an S 2 -brane with a Dirichlet boundary condition along the time direction
becomes an S1-brane labeled by the parameters u and . Note that we have obtained the
untwisted H4 -branes starting from branes whose world-volume does not contain the null
geodesic used to define the Penrose limit. It is then natural to consider the limit of branes
whose world-volume contains the null geodesic. For this purpose, we consider a family of
S 2 -branes rotated by R(0, 0, /2) and with a Neumann boundary condition along the time
direction. In the limit, the parameter that characterizes the new family of branes behaves
as follows
r
sin cos cos ,
(2.20)
k
and they become the D2-branes of the H4 model. We can proceed in a similar way for the
Penrose limit of AdS3 S 1 and of its symmetric branes. We write the background in global
coordinates

H = k sinh 2,
ds 2 = k cosh2 d 2 + d 2 + sinh2 d 2 + k dx 2 ,
(2.21)
and define the Penrose limit introducing
=
u 2v
+
,
2
k
x=
u
,
2
r
= ,
k
(2.22)
442
and then sending k to infinity. The conjugacy classes are characterized by a constant value
of cosh cos . In the limit
u
u 2
sin
.
(2.23)
2
k
2
We observe that an untwisted symmetric brane of AdS3 with a Dirichlet boundary condition
in the S 1 factor gives rise to an S1-brane in H4 . Note that for large k, the classes we are
considering are precisely those with | cosh cos | 1. This is precisely the H2 -branes and
the degenerate branes in AdS3 [74]. The null geodesic used in the limit is not contained in
the brane world-volume. On the other hand, the twisted conjugacy class with a Neumann
boundary condition in the S 1 factor, are characterized by a constant value of sinh cos
and contains the null geodesic. Since in the limit
cosh cos cos
r
sinh cos cos ,
k
(2.24)
we observe that the D2-branes of the H4 model are the Penrose limit of the AdS2 -branes.
A more detailed description of the Penrose limit for the different types of branes, using coordinate systems adapted to their world-volumes, can be found in Appendix E. In
Section 7 we will extend the analysis performed in [35] and discuss the contraction of the
boundary R SU(2)k WZW model which is the world-sheet analogue of the Penrose limit
applied simultaneously to the spacetime and to the brane contained in it.
3. Review of the bulk spectrum and semi-classical analysis

The structure of the Hilbert space of the H4 WZW model [35] is very similar to the
structure of the Hilbert space of the SL(2, R)k WZW model, clarified by Maldacena and
Ooguri [31]. Together with the standard highest-weight representations of the affine algebra, restricted by a unitarity constraint, there are other representations that satisfy a
modified highest-weight condition. These new representations are related to the standard
ones by the operation of spectral flow, which is an automorphism of the current algebra.
In our case there are three classes of highest-weight affine representations, reviewed in
Appendix A. To each affine representations we associate a primary chiral vertex operator
(z, x),
p,
0
s,
(z, x),
0 < p < 1, R,
s > 0, [/2, /2).
(3.1)
vertex operators, p is the eigenvalue of K and the highest (lowest) eigenFor the p,
0 vertex operators, s is related to the Casimir of the representation

value of J . For the s,
and is the fractional part of the eigenvalues of J . Here z is a coordinate on the world-sheet
and x a charge variable we introduced to keep track of the infinite number of components
of the H4 representations. On the charge variables the H4 algebra is realized in terms of
the operators
P = 2x ,
J = i( xx ),
K = ip,
P = 2px,
(3.2)
443
when considering its action on p,

and by the operators
P + = sx,
P =
s
,
x
J = i( + xx ),
K = 0,
(3.3)
0 . States with p = p + w with 0 < p < 1 and w N

when considering its action on s,
) while states with p = w

fall into spectral-flowed discrete representations w (
p,

0 ). We also recall
with w Z fall into spectral-flowed continuous representations w (s,
the relation
+
= 1p,
.
1 p,
(3.4)
We will denote the image under spectral flow of a representation and the corresponding
vertex operators either by w [ ], as we did before, or by the introduction of a further index ;w . Finally the operator content of the bulk theory is given by the charge conjugation
modular invariant. We then have the operators
(z, z |x, x)
= p,
(z, x)p,
(z, x),
p,

0
0
0
1
,
s,
= s,
(z, z |x, x)
(z, x)s, z , x
(3.5)
as well as their images under an equal amount of spectral flow in the left and right sectors.
The currents that generate the affine algebra preserved by the boundary conditions are
given by the combination J + (J). A more detailed description of the representation
theory of the affine H 4 algebra can be found, for instance, in [35].
Before performing the exact CFT analysis we will try to clarify the physics of the model
in a semi-classical approximation. In doing so we will gain some intuition about the spectrum of the boundary operators and on the form of the bulk-boundary and the three-point
boundary couplings. We use the following parametrization for the group elements
u
g = e2Je
i w
P + iw P +
2
2
e 2 J +2vK ,
(3.6)
where w = rei = + i . The isometry generators are K = K = 12 v and

i
J = u (w w ),
2
u
u
ei 2
ei 2
+
+
P = [4i + wv ],
P = [4i wv ],
2 2
2 2
J = u
(w w ),
2
ei 2
P = [4i w
v ],
2 2
ei 2
v ].
P = [4i + w
2 2
To each vertex operator we can associate a semi-classical wave function. For the discrete
representations they are
+
p,
= e2ipv+i u
=e
p,
iu
iu
p
2 px xe
xe
2 +ip wxe
iu
2 w wipw
iu
iu
2
2
2ipv+i u p
ip w xe
px xe
iu
2 w w+ipwxe
(3.7)
444
where p > 0, R and x, x are two independent charge variables. The states that belong
to these representations are confined in periodic orbits in the transverse plane, familiar
from the quantum mechanical problem of a charged particle in a magnetic field. For the
continuous representations we have
0
s,
= e
ij u+ is [ wx e
2
u
u
i 2
i
+w xe
2 ]

(x x)
n einu ,
(3.8)
nZ
where s 0, [/2, /2) and x = ei , x = ei are two independent phases. The states
that belong to the continuous representations can move freely in the transverse plane: the
parameter s is the modulus of the momentum and we can identify = 2 with its phase.
These wave functions can be expanded in modes which represent the different components
of the H4 H4 representations. It is also easy to compute semi-classical expressions for
the bulk two- and three-point functions.
We can proceed in a similar way for the states confined on the brane world-volume. In
the case of the D2-branes, according to the boundary conditions (2.7), the generators of the
unbroken background isometries are
K K = v ,
J J = 2u ,
i u2
e
P + P = [2 i v ].
2
(3.9)
They satisfy the commutation relation of the Heisenberg algebra and the brane spectrum,
exactly as the bulk spectrum, can be organized in terms of H4 representations. We then introduce three types of vertex operators for the open strings that live on a D2-brane localized
in the direction
p, = eipv+i 2 u
i 2 u+ i s
iu
p 2
2 p x 2 eiu
4 +p xe
2
p, = eipv+i 2 u
s, = e
p > 0, R,
iu
2
p 2
2 iu
p
4 p xe
2 x e
nZ
x n ein
u
2
p > 0, R,

.
s R, ,
2 2
,
(3.10)
The operators that act on the charge variable x are given by (3.2) and (3.3). The wave
functions for the discrete representations are easily recognized as the generating functions
for the Hermite polynomials. The open string states, as it was already the case for the closed
string states, are trapped in periodic orbits in the direction of the brane world-volume
unless their light-cone momentum is an integer. Note that for the boundary operators in the
continuous representations s can be an arbitrary real number.
Semi-classical expressions for the couplings between the bulk and the boundary operators can now be computed as overlap integrals of the corresponding wave functions
on the brane world-volume. As a first example consider the bulk-boundary coupling
p,
. It is easily evaluated as
q,
1
p,
1 q,2
=e
px2 (x1 +x1 )
||
(x1 x1 )
e
where

p 2
2
445
||
2 i ||
[p ] 2
||
p 2 2 ||!

H|| p ,
du dv d d ( ),
(3.11)
(3.12)
denotes a spacetime integral restricted to the brane world-volume. The integral is non-zero
only when q = 2p and || N where = 21 2 . The other possible bulk-boundary
coupling is given by

0
s,
1 t,2
x 2 m
= 8(ix2 )
t 2s cos(2 1 )
2
x1 x1
mZ
2is sin(2 1 )
(3.13)
where xi = eii . Since the D2-branes are invariant under translations along the two lightcone directions, only operators with p = 0 and = 0, 1/2 couple to their world-volume.
Their one-point functions can be derived from the previous bulk-boundary coupling upon
setting t = 0 and integrating over 2
cos s

0
s,0
= 8
( ) + ( ) ,
s

0
sin s
( ) + ( ) .
s,1/2 = 8i
sx

(3.14)
The discussion for the S1-branes is very similar. The relevant isometry generators are
K + K = 0 and
J + J = i(w w ),
u
.
P + P = 2 2 sin
2
u
P + + P + = 2 2 sin
,
2
(3.15)
They realize the algebra of the rigid motions of the plane and as a consequence the open
string states that live on an S1-brane labeled by a (ua , a ) can only belong to the continuous representations of the Heisenberg algebra. The semi-classical wave functions are
saa = e
s u
2 2 sin 2 a
w
(wx
x)
(3.16)
The action of the zero-modes of the currents on the charge variable x is always given by
(3.3). As we did for the D2-branes we can now extract the bulk-boundary couplings from
the overlap of the wave functions
446
p,
(x ) aa (x2 )
1 s
ua ,a
1 +x x
1 2
8i
ua 2ipa +i( 2 )ua px1 x1 2 ( x 1
2
=
sin
e
p
2
s
) 4p
(1i cot u2 a )
.
(3.17)
We see that all the discrete representations as well as the identity field have a non-vanishing
one-point function, as expected since the S1-branes brake the translational invariance in the
light-cone directions. When u = 2n the geometry of the conjugacy classes changes. The
vertex operators have a non-vanishing one-point function only in the presence of the
S(1)-branes, localized at the origin of the transverse plane and at arbitrary positions in
the light-cone directions

8 2ipv+i upx x
p, u,v =
e
.
p
(3.18)
Translational invariance in the transverse plane being now broken, also all the continuous
representations have a non-vanishing coupling
0
s, = 2ei u ( + ).
(3.19)
If we consider instead the cylindrical branes, extended along the v direction and with a
0 vertex operators have a non-vanishing
fixed radius r in the transverse plane, only the s,
coupling
0

s, u,r = ei u J0 2sr ( + ).
(3.20)
In the following we will not discuss in detail the cylindrical branes even though it would
be interesting to extend our exact CFT analysis to them as well.
4. Spectrum of the boundary operators

In the previous section we discussed the semi-classical open string spectrum for the two
families of symmetric branes we are studying in this paper. According to the semi-classical
analysis, the states of open strings that live on a D2-brane, form all possible representations
of the H 4 algebra. The states of open strings that live on an S1-brane, can only belong to
the continuous representations. In this section, we provide a detailed description of the
spectrum of the open strings and extend the analysis also to the open strings that end on
different branes. In close analogy with the bulk primary vertex operators, we introduce for
each representation of the affine algebra, a boundary primary vertex operator ab (t, x),
which depends on the insertion point on the real axis t and on a charge variable x. The two
upper indices label the two branes on which the open string ends. This is the same as the
two boundary conditions the vertex operator interpolates between.
Let us start with a D2-brane localized at = 0. The space of states H00 contains in this
case all possible representations of the H 4 algebra

w p, , 0 < p < 1, R, w N,

w s, , s R, [/2, /2), w Z.
447
(4.1)
This is very similar to what happens for the AdS2 -brane localized at = 0 in AdS3 , where
the spectrum is also given by the holomorphic square root of the bulk spectrum [76]. Similarly to that case, an observation about the spectral flow is in order [76]. Given a solution
g(u0 , v0 , r0 ei0 ) H4 of the classical equation of motion, a new solution can be generated
by the action of the spectral flow
w +

w

w [g] u, v, rei e J g u0 , v0 , r0 ei0 e J ,
where
(4.2)
= . Therefore
u = u0 +
2w
,
v = v0 ,
rei = r0 ei(0 +w ) .
(4.3)
As we can see from (4.3), when = 0 only the spectral flow by an even integer is a
symmetry of the D2-brane spectrum. Spectral flow by an odd integer maps a string living
on a brane sitting at to a string stretched between a brane at and a brane at . As
a consequence, we cannot use in the general case the spectral flow by an odd integer, to
generate the complete brane spectrum, as we did in (4.1) for a brane sitting at the origin.
However, using the relation (3.4), it is easy to verify that for the discrete representations,
it is enough to consider the spectral flow by an even integer in order to obtain all possible
values of the light-cone momentum. For instance, a state carrying a light-cone momentum
p + 2w 1, belongs to the representation 2w [(1p), ].

For the continuous representations, the spectral flow by an even integer is not enough.
We have to proceed in a different way [76]. We start with the vertex operator s, and
take its image under the spectral flow by an odd integer 2w + 1. In this way, as explained
before, we obtain the vertex operator pertaining to a string ending on the brane at and
carrying an odd light-cone momentum. This asymmetry between the even and the odd
spectral-flowed continuous representations will be also manifest in the annulus amplitude,
as we will see in Section 7. Summarizing, the spectrum of a brane localized at = 0, is
given by

2w p, , 0 < p < 1, R, w Z,

2w s, , s R, [/2, /2), w Z,

2w+1 s, , s R, [/2, /2), w Z.
(4.4)
The same structure of the spectrum holds for the strings ending on two different branes
localized at 1 and 2 , respectively. The only difference is that the possible values of the
parameter s that label the continuous representations are now constrained by 2 2 s 2
[1 eiw 2 ]2 . The minimal value of s simply reflects the tension of the string stretched
between the two branes. The fact that the bound depends on whether the amount of spectral
flow is even or odd nicely reflects the different behavior of the continuous representations
under the action of the spectral flow.
448
We now turn to the S1-branes. As in the previous section, we use the short-hand notation
a (ua , a ) for the boundary labels. The S1-branes are Cardy branes. Therefore, a relation
can be established between the parameters labeling the conjugacy classes and the quantum
numbers of the H 4 representations
u = 2(p + w),
2 = (2 2p 1).
(4.5)
As usual, 0 < p < 1 and w N. We will derive this relation in Section 7, both by studying
the annulus amplitudes of the H4 model and by taking the Penrose limit of the R SU(2)k
WZW model. In close analogy with the D-branes in flat space, the quantum numbers p
and together with the spectral flow parameter w, are related to the distance along the u
and the v direction, respectively. Using the relation (4.5), we can associate to a brane with
labels (ua , a ) the parameters pa , a and wa . This is useful because as it is the case for the
Cardy branes in a RCFT, the spectrum of open strings ab stretched from the brane b to
the brane a is encoded in the fusion product wb [pb ,b ] wa [pa ,a ] of the two
corresponding chiral vertex operators.
As mentioned earlier, the open strings ending on the same brane belong to the continuous representations of the H 4 algebra. The Hilbert space decomposes as

Haa =
ds s Vs,0;0 ,
(4.6)
0
aa . The open strings that end on two differand the corresponding vertex operators are s,0;0
ent S1-branes with labels a and b can belong to any of the highest-weight representations
of the H 4 algebra as well as to their images under spectral flow. The precise representation
depends on the distance between the two branes along the u direction. We introduce the
following notation: when pb pa > 0 we set pb pa = p ab + w ab , with 0 < p ab < 1 and
w ab N. We also define ab = b a . The brane spectrum is then given by
Hab =
Vpab ,ab n;wab ,
(4.7)
n=0
and the vertex operators are pabab ,ab n;wab . Similarly when pb pa < 0 we set pb pa =
p ab + w ab with 1 < p ab < 0, wab N. The brane spectrum is then given by
Hab =
Vpab ,ab +n;wab ,
(4.8)
n=0
and the vertex operators are pabab ,ab +n;wab , n N. Finally when pb pa is an integer we
set pb pa = w ab and we have

Hab =
0
ds Vs,ab ;wab .
(4.9)
449
The vertex operators read s,abab ;wab . We assume that for our non-compact WZW model,
the space Hab decomposes as expected from the fusion rules. This assumption will be
confirmed by the complete solution of the model that is presented in the next section.
For the S1-branes, the spectral-flowed representations appear whenever the distance
along the u direction between two branes exceeds 2
. In fact, the action of the spectral
flow amounts to
w +

w

w [g] u, v, rei e J g u0 , v0 , r0 ei0 e J ,
(4.10)
that is
u = u0 +
2w
,
v = v0 ,
rei = r0 ei(0 w ) .
(4.11)
Note that in this case, the spectral flow is a symmetry for every integer w and maps a
string stretched between two branes localized at ua and ub to a string stretched from ua
to ub + 2w. In the following, we will derive most of our results assuming that all the
representations are highest-weight representations of the current algebra. It is however not
difficult to extend our results to amplitudes involving spectral-flowed states, using the freefield realization [38,39], as we did for the closed string amplitudes in [35].
5. Structure constants
A boundary conformal field theory is completely specified by three sets of structure
constants: the couplings between three bulk or three boundary fields and the couplings
between one bulk and one boundary field. These structure constants satisfy a set of factorization constraints first derived by Cardy and Lewellen [78,79] (see also [8082]). For the
sake of clarity, we will briefly review the sewing constraints for a generic CFT and in the
next section we will write them explicitly for the NappiWitten model.
For a CFT defined on the upper-half plane, there are two sets of fields. The first set
contains the bulk fields i, (z, z ), inserted in the interior of the upper-half plane and characterized by the quantum numbers (i, ). These quantum numbers specify the representations
of the left and right chiral algebras. The second set contains the boundary fields iab (t),
inserted on the boundary (here the real axis). They are characterized by two boundary conditions a and b. They are also characterized by the quantum number i, which labels the
representations of the linear combination of the left and right affine algebras left unbroken
by the boundary conditions. There are three sets of OPEs (we adopt with minor modifications the notation used in [81])

(z1 z2 )hk hi hj (z1 z 2 )hk h h C(i,),(j,) (k,k)

i, (z1 , z 1 )j, (z2 , z 2 )
k
k,k (z2 , z 2 ),
i, (t + iy)

j
(2y)hj hi h a B (i,) jaa (t),
j
(5.1)
(5.2)
450
iab (t1 )jbc (t2 )

(t1 t2 )hk hi hj Cijabc,k kac (t2 ),
(5.3)
where t1 < t2 . Here and in the following, the symbol 1 will always denote the identity field
and 1a the one-point function of the identity with boundary condition a on the real axis.
Indexes are raised and lowered using
dijab = Cijaba,1 1a = Cjbab,1
1b
i
(5.4)
aaa,1
and for consistency we require that C11
= 1. Any correlation function on a Riemann
surface with boundaries and with arbitrary insertions of bulk and boundary operators can
be constructed by sewing together the basic amplitudes that correspond to the structure
constants displayed in (5.1)(5.3). The sewing constraints follow from the requirement that
all possible ways of decomposing a given amplitude into the basic amplitudes lead to the
same answer.The resulting constraints involve, besides the structure constants,
the fusing

matrices Fpq ji kl and the modular S matrix. The fusing matrices Fpq ji kl , by definition,
implement the duality transformations of the conformal blocks pertaining to the four-point
amplitudes. Our conventions for the fusing matrices can be found in Appendix B.
Sonoda [83] has analyzed the sewing constraints for Riemann surfaces without boundaries, and proved that the CFT correlation functions are unambiguously defined provided
that the four-point functions on the sphere are crossing symmetric and that the one-point
functions on the torus are modular covariant. Cardy and Lewellen extended these results to
Riemann surfaces with boundaries [78,79]. They proved that all the amplitudes are unambiguously defined provided that the structure constants satisfy four additional constraints.
The first constraint is a quadratic constraint that follows from two different factorization
limits of the bulk two-point function i (z1 )j (z2 )a and reads

( ) i
a l a l aaa,1
.
Bi Bj C
=
Cij k a B 1k Fkl
(5.5)
ll
( ) j
k
Here l is the representation conjugate to l, and represents the action of an external automorphism on the representations of the chiral algebra. A non-trivial has to be taken
into account when considering for instance the symmetric branes of a WZW model. This
constraint is a particular case of a more general one that follows from the factorization of
a three-point function with two bulk and one boundary fields.
The second constraint follows from two distinct factorization limits (t2 t3 and t1 t2 ,
respectively) of the four-point boundary correlator

ab
i (t1 )jbc (t2 )kcd (t3 )lda (t4 ) ,
(5.6)
and it reads
abd,l ada,1
Cjbcd,n
Cin
C
k
ll

cda,r aca,1
Cijabc,r Ckl
Cr r Frn
j
i

k
.
l
j
(5.7)
The third constraint, relates the bulk-boundary couplings a B i and the boundary threepoint couplings Cijabc,k . It follows from the requirement of locality for a bulk field in a
three-point function with two boundary fields

i (z)jab (t1 )kba (t2 ) .
451
(5.8)
It can be written as follows

r aba,r aaa,1
hr +hl
a
b l abb,k aba,1
B i Cj l C
=
B i Cj k Cr r ei(2hn hj hk 2hi + 2 )
kk
r,n
Frn
k
j

i
( )
Fnl
( )
j

i
.
k
(5.9)
The fourth and last constraint involves the boundary one-point functions on the cylinder.
When the boundary field is the identity, this constraint reduces to the Cardy constraint
which relates the open and the closed string channel of the vacuum annulus amplitude. We
postpone the analysis of this constraint to Section 7.
j
In the following, we will determine the bulk-boundary couplings a B i and the boundary
three-point couplings Cijabc
k for the two classes of symmetric branes of the H4 WZW model.
We will use as an input the structure constants and the fusing matrices of the model, as
computed in [35]. Here, we only recall the two and the three-point couplings, while the
fusing matrices are collected for convenience of the reader in Appendix B. In order to
write our formulae in a simple way, we shall leave in the following the dependence on
the world-sheet variables zi and ti understood. We will write each bulk amplitude as the
which contain the dependence on
product of three terms, two kinematical parts K and K,
the charge variables of the left and right chiral algebras, and a dynamical part C. The form
of K and K for the two- and the three-point functions is completely fixed by the current
algebra Ward identities.2 Another convention we will commonly use is
=
i ,
(5.10)
i=1
for an n-point amplitude with primary vertex operators carrying the labels i . The nontrivial two-point functions are

+
p , p , = (p1 p2 )(1 + 2 )ep(x1 x2 +x1 +x2 ) ,
1
s01 ,1 s02 ,2
(s1 s2 )
(2)2 (1 2 )( 1 2 ),
s1
= 0, 1.
(5.11)
The bulk three-point couplings between three discrete representations are

1
1 (p1 + p2 ) 2 +n
,
n! (p1 ) (p2 )

1 +n
2
(p1 )
1
,
C(p1 ,1 ),(p2 ,2 ) (p1 p2 ,1 +2 n) =
n! (p2 ) (p1 p2 )
C(p1 ,1 ),(p2 ,2 ) (p1 +p2 ,1 +2 +n) =
(5.12)
p1 > p2 ,
(5.13)
2 The factor K for the three-point function, is the group-theoretic ClebschGordan coefficient of the classical
algebra H4 .
452
where (x) = (x)/ (1 x). The corresponding kinematical factors are
K(p1 ,1 ),(p2 ,2 ),(p3 ,3 ) = ex3 (p1 x1 +p2 x2 ) (x1 x2 ) ,
K(p1 ,1 ),(p2 ,2 ),(p3 ,3 ) = ex1 (p2 x2 +p3 x3 ) (x2 x3 ) ,
(5.14)
The coupling between one continuous and two discrete

with similar expressions for K.
representations is
C(p,1 ),(p,2 ) (s,{1 +2 }) = e
where (x) =
d ln (x)
dx
s2
2 (p)
(5.15)
and we defined
(p) (p) + (1 p) 2(1).
(5.16)
We also have
K(p,1 ),(p,2 ),(s,3 ) = e
px1 x2 s (x2 x3 + x1 )
2
3 x .
3
(5.17)
Finally, the coupling between three 0 vertex operators simply reflects the conservation of
the momentum in the transverse plane. Therefore it is non-zero only when
s32 = s12 + s22 2s1 s2 cos ,
s3 ei = s1 s2 ei .
(5.18)
It can be written as
(s , ),(s , ),(s , )
KKC
1 1
2 2
3 3
21 )(31 + 31 )(21 )(31 )
4
(21 +

= (2) (x1 x1 )
,
4s12 s22 (s32 s12 s22 )2
(5.19)
where ij = i j and Z.
We have now at our disposal all the information required to solve the H4 WZW model
with boundary. We will consider first the D2-branes and use the sewing constraints to fix
their structure constants. We will then study the S1-branes that, as we already explained,
are the Cardy branes of the model. In this case we compute the bulk-boundary couplings
and verify that, for our non-compact WZW model, the boundary three-point couplings are
also proportional to the fusing matrices. More details concerning the steps leading to the
solution, can be found in Appendix D.
5.1. The D2-branes
In this section we compute the structure constants for the maximally symmetric D2branes of the NappiWitten model. From the physical point of view, the couplings we
derive are important since they provide examples of open and closed string interactions
in a non-compact, curved spacetime. They are also interesting from a more formal point
of view, since they represent the first example of the structure constants for the twisted
symmetric branes of a WZW model. They could be a useful guide, leading to a general
answer, extending the one that already exists for the Cardy branes. We will first derive the
bulk-boundary couplings and then the boundary three-point couplings for strings ending
453
on the same D2-brane at = 0. Once these couplings are obtained, it is easy to generalize
the solution to strings ending on different branes at arbitrary position in the direction.
The boundary two-point functions are

1 2

1 ,1
11 epx1 x2 ,
p , (x1 )p 2,1 (x2 ) = C(p1 ,2 ),(p
1 1
2 2
1 1
2 ,2 )

1 2

1 ,1
11 2(s1 + s2 )(2 1 )x1 1 2 .
s , (x1 )s 2,1 (x2 ) = C(s1 ,2 ),(s
1 1
2 2
1 1
2 ,2 )
(5.1.1)
The second two-point function is non-zero only when 1 + 2 = 0, 1. The bulk-boundary
couplings have the following form

(2p, )
p, (x1 )2p, (x2 ) = epx2 (x1 +x1 ) (x1 x1 ) B (p,) 2 1 .

(5.1.2)
1
Here = 21 2 and the coupling is non-zero only when 2 = 21 n, n N. In

0 vertex operators, it is convenient to introduce two
order to write the couplings for the s,
new angles defined by

= + ,
= ,
0 2, .
(5.1.3)
We obtain
0

s,1 (x1 )t, (x2 )

2
x 2 m
2
= x2
x1 x1
mZ

(t, )
(t, )
(2 1 ) B (s,2 )+ + (2 1 + ) B (s,2 ) 1 ,
1
(5.1.4)
where = 21 2 Z. We set t = 2s cos with [0, ) and we introduced a further

(t, )
subscript in B (s,2 ) in order to distinguish the coefficients of the two delta-functions. The
1
coupling with the identity can be obtained by setting t = 0 and integrating over 2

0

s,1 (x) = (ix) ( ) + ( ) B 1s + + (1) B 1s 1 .
(5.1.5)
There are two non-trivial bulk two-point functions that may be used to derive the constraint
0 0 and +
for the bulk-boundary couplings, namely s,
. The first one is
1 t,2
p,1 p,2
(t, )
very similar to the corresponding amplitude in flat space. Indeed, the form of B s, 2
1
turns out to be very simple
(t,2 )
B s,
1
ei 2s sin
,
=
2s| sin |
(5.1.6)
where t = 2s cos and [0, ). The coupling with the identity can then be written as
follows
0

cos( 2s)
s,1 (x) = ( ) + ( )
, = 0,
s
0
sin( 2s)

s,1 (x) =
(5.1.7)
( ) + ( )
, = 1.
x
s
454
The second correlator leads to the following constraint
(2p,2 +n)
B (p, ) 1
1

=
(2p,2 n+)
B (p, ) 2
2
,1
C(2p,2 +n),(2p,2 n+)

1
ds sC(p,1 ),(p,2 ) (s,) B 1s F(s,),(2p,21 +n)
(p, 1 )
(p, 2 )

(p, 1 )
,
(p, 2 )
(5.1.8)
which is solved by
(2p,2 n)
B (p,) 1

n 1
1 (2p) 2 + 4 2
1 in 2
4
cot p
= n
e 2 cot p
2 (p)
2 2 2 n!

.
Hn
cot p
(5.1.9)
This exact result and the semi-classical computation (3.11) agree in the limit p 1.
Actually the two results differ only in that p has to be replaced by tan(p) in the
argument of both the exponential and the Hermite polynomials and that the overall powers
of p have to become powers of (p). Note also the similarity of this coupling with the
square-root of a bulk coupling of the form C++ (5.12), as expected on general grounds.
Finally, it is interesting to remark that the coupling vanishes when p = 1/2 and it has
to be replaced by the coupling with a spectral-flowed boundary operator. These couplings
can be computed either using the free-field realization of [35,38,39] or by studying the
factorization of a four-point amplitude with suitable external momenta.
We now proceed to the computation of the boundary three-point couplings. The D2branes of the H4 WZW model are not Cardy branes and in the absence of an one-toone correspondence between the brane parameters and the representations of the chiral
algebra, there is no natural relation between the boundary three-point couplings and the
fusing matrices. In the absence of any ansatz, we have to solve directly the constraints.
Fortunately, at least the couplings between open strings living on the brane at = 0 are
simple. They are given by
000,r
= (s + t r),
Cs,t
000,s
C(p,
=e
),(p, )
s2
4 (p)

n+1
2 4
2 1/4
(p1 )
000,(p1 p2 ,1 +2 n)
,
C(p , ),(p , )
=
n!!
(p2 ) (p1 p2 )
1 1
2 2
0,
1
n 2N,
(5.1.10)
n 2N + 1,
and we see that they are very similar to the square root of the bulk couplings. The kinematical parts can be easily obtained from (5.14) and (5.17). The fact that the last coupling
vanishes whenever n 2N + 1 is easy to understand if we perform a semi-classical computation using the wave functions (3.10).
In order to find the solution for branes sitting at arbitrary positions, we rely once more
on the fact that the boundary vertex operators in the continuous representations are very
similar to the standard tachyonic vertex operators in flat space. If we also take into account
the finite length of the open string stretched between two branes at i and j , we conclude
455
that the quantity that is conserved in the interactions of the s,i j vertex operators is not s
but

s sign(s) s 2 b2 ij2 , |s| b|ij |, ij = i j , b2 = 1/2 2 .
(5.1.11)
As a consequence, we expect that the general form of the coupling between three vertex
operators in the continuous representations is
2 3
Cstr1
= (s + t + r ).
(5.1.12)
Consider now the correlator

1 2
2 3

(2)(s,3 4) (3)(t,4 1) (4) ,
(p, ) (1)(p,
)
1
(5.1.13)
where for clarity the numbers in the round brackets stand for both the charge variables and
the worldsheet coordinates of the corresponding vertex operators. This correlator factorizes
on a single block in the s-channel and leads to the sewing constraint

4 2 3
C(p,
C 1 2 4
+n),(p, ),(s, ) (p, ),(p, + n),(t, )
2
st
1 2 3
= C(p,
e 2 ( (p) cos i cot p sin )i(n+)+i ,
1 ),(p,2 ),(r,)
(5.1.14)
where
stei = (s + ib34 )(t ib41 ),
rei = s + tei .
(5.1.15)
Note that according to (5.1.12), when raising a continuous index s we must multiply the
coupling by s/s . From the previous constraint, we may read the coupling between one
continuous and two discrete representations
1 +2 3

ibs
s2
s
s
1 2 3 ,s
sin
p 2 +i 2 cot p s b(1 +3 )+ 4 (p) ,
C(p,
=
e
1 ),(p,2 )
s s + ib13
1 +2 3

2
ibs
s
s
1 2 3 ,s
i cot p,s b(1 +3 )+ s4 (p)
C(p, ),(p, ) =
e sin p 2 2
, (5.1.16)
1
2
s s ib13
where |s| b|13 |. Actually, the phase proportional to 2 is not fixed by Eq. (5.1.14) but
by the constraint associated with the correlator
1 2

2 3

4 1
(p, ) (1)(p,
(5.1.17)
(2)(p,3 4 ) (3)(p,
(4) ,
)
)
1
which reads
,(s, )
s
2 3 4
C(p,
C 4 1 2
+ (1) (s s)
),(p, ) (p, ),(p, ),(s, )
2

=
|13 |
st
s s 2 t 2 (p)
dt
J
e 2
sin p
sin p

1 2 3 ,(t,t )

C(p,
C 3 4 1
+ (1) (t t) .
),(p, ) (p, ),(p, ),(t,t )
1
(5.1.18)
456
In order to verify that the couplings in (5.1.16) solve the previous constraint, one needs the
integral (H.14). At this point the correlator

1 2
2 3

4 1
(2)(q,3 4) (3)(q,
(4) ,
(p, ) (1)(p,
(5.1.19)
)
)
1
gives a constraint whose unknowns are the couplings between three discrete representations. The constraint is
,((pq),2 +3 +n)
2 3 4
C(p,
),(q, )
2
,(q, )
1 2 4
4
C(p,
),((pq), + +n)
1

=
ds F(s,s ),((pq),2 +3 +n)
|13 |
(p, 1 )
(q, 4 )
(p, 2 )
(q, 3 )

1 2 3 ,(s,s ) 3 4 1
C
+ (1) (s s) ,
C(p,
),(p, ) (q, ),(q, ),(s,s )
1
and it involves the following fusing matrix

(p, 1 ) (p, 2 )
F(s,s ),((pq),2 +3 +n)
(q, 4 ) (q, 3 )

+1
n
(p)
1 (p)(1 q)
=
n!
(p q)
(q) (p q)
2

2
s sin (p q)
s
s2 ((p)+(1q)2(1))
.

e
Ln
2 sin p sin q
2
(5.1.20)
(5.1.21)
The integral on the right-hand side can be evaluated using (H.13). We obtain
n+1
2 4
Q2
(p)
e 2 Hn (Q),
n
1
2
2 2 n! (q) (p q)
n+1

1
1
2 4
Q2
(p)
2 2 4 in
1 2 3 ,((pq),1 +2 n)
C(p,
(5.1.22)
=
e 2 Hn (Q),
n
1 ),(q,2 )
2 2 n! (q) (p q)
1
,((pq),1 +2 +n)
1 2 3
C(p,
),(q, )
2 2 4 in
where
Q=
(1 sin q + 2 sin (p q) 3 sin p)

.
2 sin p sin (p q) sin q
(5.1.23)
Similarly
n 1

1
1
2 2 4 i n (p + q) 2 + 4 Q2
=
e 2 Hn (Q),
n
2 2 n! (p) (q)
n 1

1
1
2 2 4 i n (p + q) 2 + 4 Q2
1 2 3 ,(pq,1 +2 n)
C(p, ),(q, )
=
e 2 Hn (Q),
n
1
2
2 2 n! (p) (q)
1 2 3 ,(p+q,1 +2 +n)
C(p,
1 ),(q,2 )
(5.1.24)
where
Q=
(1 sin q 2 sin (p + q) + 3 sin p)

.
2 sin p sin (p + q) sin q
(5.1.25)
457
Fig. 1. World-sheet instanton contribution to the couplings for intersecting brane models.
The kinematical part is similar to the one given in (5.14). We did not check the constraint
(5.9) for this family of branes.
There are many similarities between the H4 WZW model and the orbifold CFT of a
plane with points identified by a rotation, stemming from the free field realization found in
[38,39]. It is therefore worth to compare the couplings derived in this section with the couplings for intersecting branes in toroidal compactifications, discussed in [9092]. Actually,
we can start directly from branes at angles in flat space. The boundary three-point couplings contains a quantum part that can be computed using orbifold twist fields [91,92] and
that coincides with (5.1.22) and (5.1.24) with n = Q = 0. They also receive contributions
from disc world-sheet instantons that behave as
Cij k e
Aij k
2
(5.1.26)
where Aij k is the area of the triangle formed by the three intersecting branes (see Fig. 1).
Consider first three branes intersecting at the origin. In this case Aij k = 0. The couplings
in the H4 model in this case contain no exponential contribution depending on the position
of the branes. We now move each brane parallel to itself a distance di from the origin. If
we call ij the angle between the brane i and the brane j , the area of the triangle is
Aij k =
[d1 sin 23 + d2 sin 13 d3 sin 12 ]2

.
2 sin 12 sin 13 sin 23
(5.1.27)
We recognize that the instanton contribution (5.1.26) coincides with the exponential term
in (5.1.23) and (5.1.25) upon setting di = i and identifying the angles ij with the lighti j
cone momentum carried by the vertex operators p,
.
Finally there is an interesting limit to consider. The limit in the H4 model is

the analogue of the limit in which an AdS2 -brane is moved towards the boundary
of AdS3 . In this limit, as discussed in [76], one obtains a so-called NCOS theory [85,86],
which is a theory of open strings decoupled from the closed string sector. In our case however there is an important difference: after the Penrose limit, the world-volume flux is null
458
and therefore there is no notion of a critical electric field. In this respect, the world-volume
theories of the D2-branes provide examples of theories with light-like non-commutativity
[87]. We may also observe that in the limit , only the bulk continuous representations remain coupled to the brane. The couplings in (5.1.22) and (5.1.24) between open
strings in discrete representations are exponentially suppressed. Thus the strings interact in
this limit only through the exchange of states in the continuous representations.
5.2. The S1-branes
In this section we provide a detailed solution for the structure constants of the S1-branes.
We start with the boundary two-point functions for open strings ending on two different
branes a and b, sitting at different positions in the u direction. As explained in Section 4,
the vertex operators for the open strings stretched between the two branes belong to the
Vp+ab ,ab n representations, n N, when pb > pa and to the Vpab ,ab +n representations,
n N, when pb < pa . We will use the short-hand notation pabab ,ab n = pabab ,n where the
sign in ab n is fixed accordingly to the sign of p ab . The two-point functions are

ab

ab
pab ,n (t1 , x1 )pbaba ,n (t2 , x2 ) a = Cnaba,1
(5.2.1)
1a e|p |x1 x2 n1 ,n2 .
1 ;n2
1
When the two branes are at the same position in the u direction, the open strings ending on
them belong to the continuous representations. In particular, when a = b we have

(s1 s2 )
(t1 , x1 )saa
(t2 , x2 ) a = Csaaa,1
1a 2(1 2 )
.
saa
(5.2.2)
1
2
1 s2
s1
The Ward identities completely fix the dependence of the bulk-boundary couplings on
the charge variables
x

+
px x s (x x + 1 )
p, (z, x1 )saa (t, x2 ) a = e 1 1 2 1 2 x2 a B sp, 1a ,

x

px1 x1 s (x1 x2 + x1 ) a s
aa
2
2
(z,
x
)
(t,
x
)
=
e
B p, 1a ,
p,
1
2
s
a

0
aa
2
s1 , (z, x1 )s2 (t, x2 ) a = 8 (1 1 22 + )( + 1 + 1 ) a B ss2 , 1a .
(5.2.3)
The last coupling is non-zero only when 0 s2 2s1 and we set
s2 = 2s1 sin .
(5.2.4)
2
The bulk one-point functions with the identity, are particular cases of the previous expressions

p, (z, x) a = epx x a B 1p, 1a
0

s, (z, x) a = 2( + )
a B 1s, 1a .
(5.2.5)
We can fix all the bulk-boundary structure constants by studying the factorization of the
three kinds of bulk two-point functions, namely p+ q+ , p+ q (where we have to
distinguish between p > q and p = q) and p+ s0 . Using the bulk three-point couplings
459
in (5.12)(5.15) and the fusing matrices in Appendix B, the sewing constraints can be
written as follows

s2
s2 a 1
a s
a s
((p)+(q)2(1))
B p,1 B q,2 = p,q e 2
Ln p,q
B p+q,1 +2 +n ,
2
n=0

s2
s2 a 1
a s
a s
((q)+(1p)2(1))
2
B p,1 B q,2 = p,q e
Ln p,q
B pq,1 +2 n ,
2
n=0
B sp,1 a B sp,2 =
s2
e 2 ((p)+(1p)2(1))
sin p

dt tJ0
st
sin p
B 1t,1 +2 ,
0
a
B sp,2 a B ss2 , =
1
is12 sin
2 tan(p)
s12 (1cos )

e
((p)+(1p)2(1))
2
e
ein a B 1p,1 +2 +n ,
2
s1 sin
nZ
(5.2.6)
where p,q = cot p + cot q and as before s22 = 2s12 (1 cos ). The solution is

a
B sp,
B ss2 , =
is
u
e2ip+i u s4 [(p)+(1p)2(1)] 4 tan(p)
tan( 2 )
e
,
sin p 1 eiu
ei u
2(u ),
s12 sin
s22 = 2s12 (1 cos ).
(5.2.7)
Note the similarity between the first coupling and a bulk three-point coupling of the form
C+0 (5.15), as expected on general grounds. The second coupling can also be written as
a
B ss2 , = ei u
1
(s2 (u) s2 )
,
s2
s22 (u) = 4s12 sin2
u
.
2
(5.2.8)
Finally the one-point functions with the identity, relevant for the construction of the boundary states, are a particular case of the previous couplings and read

iu
e2ip+i u
i
a 0
eip+i u 2 ,
B p, =
=
u
iu
sin p 1 e
2 sin 2 sin p
a
B 0s, =
ei u
4 sin2 ( u
2 )
(s)
.
s
(5.2.9)
Whenever u = + 2n the couplings in (5.2.7) simplify, since the brane is trivially embedded in the spacetime. On the other hand, whenever u = 2n, the behavior of the
couplings in (5.2.7) reflects the change in the geometry of the branes. We have first to multiply the structure constants by the one-point function of the identity 1a = sin(ua /2).
We then observe that the first coupling in (5.2.7) becomes non-trivial only for s = 0, as
expected since only these representations live on the S(1)-brane world-volume. As for
the couplings of the continuous representations, they are now non-zero for every value of
s1 while s2 has to vanish. One can think of this process as an interpolation between Neumann and Dirichlet boundary conditions induced by the flux on the brane world-volume.
460
In the limit p 1 these couplings reproduce the semi-classical expressions derived in

Section 3.
We determine now the boundary three-point couplings. Even though the explicit form
of these couplings is slightly more involved than the couplings we computed in the previous section for the D2-branes, our task is in this case simplified since the S1-branes are
the Cardy branes of the H4 WZW model. The three-point boundary couplings for Cardy
boundary conditions in a RCFT, due to the one-to-one correspondence between the boundary labels and the representations of the chiral algebra, can be expressed using the fusing
matrices. This was first realized in the case of the Virasoro minimal models [81]. Indeed,
one can verify that setting

i j
abc,k
,
Fbk
Cij
(5.2.10)
a c
the constraint (5.7) is satisfied as a consequence of the pentagon relation between the fusing
matrices [88]. The previous relation proved very useful in order to study the effective field
theory on the brane world-volume. Indeed, the fusing matrices of a WZW model based on
the group G coincide with the Racah coefficients of the quantum group algebra Uq (G),
2i
where q = e k+g with k the level and g the dual Coxeter number. In the limit q 1, the
quantum Racah coefficients reduce to the classical ones, a fact that has been exploited
in [72,84] to study the non-commutative geometry on the brane world-volume. Similar
observations were made also for some non-compact CFTs, most notably for the Euclidean
version of AdS3 [46] and the Liouville model [51]. We now verify that the relation between
the fusing matrices and the three-point boundary couplings (5.2.10) remains valid also for
the non-compact H4 WZW model.
In the following, we will repeatedly use the relation (4.5) between the brane parameters
and the H4 quantum numbers. As it was the case for the bulk theory, we can distinguish
between various types of boundary three-point couplings. Consider three S1-branes with
labels a, b and c. When pc > pb > pa , we have a boundary coupling similar to a bulk
C++ coupling
ab

pab ,n (t1 , x1 )pbcbc ,n (t2 , x2 )pcaca ,n3 (t3 , x3 ) a
1
ab
bc
3 C aca,1 1 e x3 (|p |x1 +|p |x2 ) (x
= Cnabc,n
a
1
n3 n 3
1 n2
x2 ) ,
(5.2.11)
where = n1 + n2 n3 . When pb > pa = pc , we have a boundary coupling similar to a

bulk C+0 coupling
ab

pab ,n (t1 , x1 )pbcbc ,n (t2 , x2 )s,caca (t3 , x3 ) a
1
|p ab |x1 x2 s (x2 x3 + x1 )
aca,1
2
3 x ,
= Cnabc,s
Css
1a e
3
1 n2
(5.2.12)
where = n1 n2 . Finally, when pa = pb = pc we have a boundary coupling similar to a

bulk C000 coupling
ab

s1 (t1 , x1 )sbc
(t2 , x2 )sca
(t3 , x3 ) a
2
3
3 C aca,1 1 (2)2 ( )( + ),
= Csabc,s
a
2
3
1
2
s3 s3
1 s2
(5.2.13)
461
with
s32 = s12 + s22 2s1 s2 cos ,
ei =
s1 s2 ei
.
s3
(5.2.14)
The other configurations can be obtained by permuting the fields in the previous expressions. It is also useful to remember that for the continuous representations, raising an index
is equivalent to (x, ) (x, ) while for the discrete representation it corresponds to
(x, ) (x , ).
We adopt now the following strategy. We assume that the three-point boundary couplings are given by the fusing matrices, up to a choice of normalization for the boundary
fields. We then use the constraints pertaining to the correlators ab ba ab ba and
ab ba aa aa to fix a convenient normalization for both the discrete pabab ,n and the
continuous saa vertex operators. Finally we verify that the couplings determined in this
way solve the other constraints as well.
Let us start with the three-point coupling between two open strings stretched between
a pair of S1-branes with labels a and b and an open string that lives on the world-volume
of one of the two S1-branes. This coupling involves two discrete and one continuous representation and is related to the following fusing matrix

aba,s
C(p
ab ,n ),(p ab ,n )
1
2
= aba (n1 , n2 )F(pb ,2 a n2 ),s
(p ab , 1 )
(pa , a )

(p ab , 2 )
,
(pa , a )
(5.2.15)
with pb > pa , 1 = ab n1 and 2 = ba + n2 . Similarly,

bab,s
C(p
ab ,n1 ),(pab ,n2 )
= bab (n1 , n2 )F(pa ,2 b +n2 ),s
(pab , 1 )
(pb , b )

(pab , 2 )
,
(pb , b )
(5.2.16)
with 1 = ba + n1 and 2 = ab n2 . Here aba and bab are proportionality factors that
can depend on all the quantum numbers involved even though we only emphasized their
dependence on the labels n1 and n2 . The factorization constraint for ab ba ab ba
reads
bab,(s,{ac })
bab,(s,{ac })
C
1b
ab
ba
ab
2 ),(p, n3 ) (p, +n4 ),(p, n1 )
C(p,ba +n

=
dt te
s 2 t 2
2 (p)

st
J
sin p
sin p
0
aba,(t,{ac })
aba,(t,{ac })
ba +n ) C(p,ab n ),(p,bc +n ) 1a ,
),(p,
1
2
3
4
C(p,ab n
(5.2.17)
where p = pb pa > 0 and = n1 + n3 n2 n4 . Using the integral (H.8) one can verify
that the constraint is satisfied if the following relation holds between the proportionality
factors aba
462
1a aba (n1 , n2 )aba (n3 , n4 ) sin pb = 1b bab (n2 , n3 )bab (n4 , n1 ) sin pa .
(5.2.18)
Combining the previous equation with the continuity of the two-point function on the disc
1a aba (n, n) sin pb = 1b bab (n, n) sin pa ,
(5.2.19)
we obtain aba (n, n) = bab (n, n). A convenient choice for the normalization is

n2

(1 pa )(pb ) n1 n2 +1
(pb )
1
,
n2 ! (pa ) (p)
(p)

n1

(1 pa )(pb ) n2 n1 +1
(pb )
1
bab (n1 , n2 ) =
.
n1 ! (pa ) (p)
(p)
aba (n1 , n2 ) =
(5.2.20)
Once we make this choice, (5.2.18) implies the following relation for the one-point function of the identity
1a sin pb = 1b sin pa ,
(5.2.21)
which is satisfied if 1a sin pa . This is indeed the case as we will se in Section 7. We
may now write

n1 n2 2
s
=
e 2 [(p)(1) 2 cot pa ]
sin pa
2 sin pa
2

s
(cot p + cot pa ) ,
Lnn12 n2
2

n2 n1 2
s
bab,s
=
e 2 [(p)(1)+ 2 cot pb ]
C(p,n
1 ),(p,n2 )
sin pb
2 sin pb

2
n2 n1 s
Ln1
(cot p cot pb ) ,
2
aba,s
C(p,n
1 ),(p,n2 )
(5.2.22)
with p = pb pa > 0 and Lm

n (x) a generalized Laguerre polynomial. Note the different
behavior of the first of these couplings for pa 0 and pb 0. In the first case it is nonzero only for s = 0 and n1 = n2 while it remains essentially unchanged in the second case.
This is as expected since only the identity exists on the brane at ua = 0 while the open
string spectrum for the brane at u = 2pb contains all the representations sbb , s 0.
Consider now the coupling between three open strings that live on the same brane. This
is a coupling between three continuous representations. In terms of the fusing matrices we
have

s
t
aaa,r
Cs,t = aaa F(pa ,a ),r
(5.2.23)
.
(pa , a ) (pa , a )
aaa,1
In this case we set aaa = 1 which is equivalent to Cs,s
= 1. The constraint for
ab
ba
aa
aa
is

aba,r
C(p,
ab n
1 ),(p,
ab +n )
2
463
aaa,r
Cs,t
st
1
e 2 [ (p) cos +i sin cot p]i
st sin

baa,(p,ab +n) aba,t
ei(n1 n)() C(p,ab +n ),s C(p,

ab n
1 ),(p,
ab +n)
(5.2.24)
n=0
i
where p = pb pa > 0, = n1 n2 , r 2 = s 2 + t 2 2st cos and ei = ster . In order

to verify that this constraint is satisfied, one can use the series (H.11). The result is
aaa,r
=
Cts
ist sin
1
e 2 tan pa ,
st sin
(5.2.25)
with r 2 = s 2 + t 2 2ts cos . For r = 0, = 0 the three-point function reproduces the

two-point function in (5.2.2).
At this point, we have completely fixed the normalization of the boundary vertex operators. Thus, the factorization constraints for the other four-point amplitudes represent a
consistency check on the solution.
The constraint for ab ba ad da is slightly more complicated and reads
bad,((pq),bd +n)
abd,(q,ad n4 )
C ab
ad
bd
2 ),(q, n3 ) (p, n1 ),((pq), +n)
C(p,ba +n

n+n4 n1
(p)(1 q) +1 sin pd
(q)
=
(n n2 + n3 )!
(p q)
(p) (q p)
2 /2 2

s
s
ds s
e 2 ((p)+(1p)2(1))
2
1
Ln+n4 n1
aba,s
C(p,
ab n

s 2
(cot q cot p)
2
1 ),(p,
ab +n )
2
ada,s
C(q,
ad n
3 ),(q,
ad +n )
4
(5.2.26)
where p = pb pa > 0, q = pd pa > 0 and = n1 + n3 n2 n4 . In order to verify

that this constraint is satisfied one can use the integral (H.12). Consider now the couplings
between states in discrete representations. Using the relation (5.2.10) we can write

abc,(p ac , )
3
= F(pb ,2 c +n2 ),(pac ,1 +2 +k)
C(pab , )(pbc
, )
1
(p ab , 1 )
(pa , a )

(p bc , 2 )
,
(pc , c )
where 1 = ab n1 , 2 = bc n2 , 3 = ac n3 and k = n1 + n2 n3 0. Similar

expressions hold for the other couplings between discrete representations and can be found
in Appendix D.
For completeness we display here the explicit form of one of these couplings
464

abc,(pq, )
C(p, )(q,3 )
1

k+1
2
(p)
sin (p q)
(n2 + n3 )!
=
n3 !(n2 + n3 n1 )! sin pc
(q) (p q)
k+1

2
sin pa k
sin q
F (n2 , n1 n2 n3 , n2 n3 , ),
sin p sin (p q)
(5.2.27)
where 1 = ab n1 , 2 = bc + n2 , 3 = ac n3 and
=
sin p sin pc
.
sin q sin pa
(5.2.28)
Note that k n3 n1 + n2 0. There are other constraints we should verify. We also

checked the constraints following from amplitudes with one bulk and two boundary operators while we did not check the constraint related to the amplitude ab bc cd da .
It is interesting to compare the coupling in (5.2.25) with the coupling between open
tachyon vertex operators on a two-torus with a magnetic field B [69,89]. Consider two free
bosonic fields X1 and X2 with the boundary conditions
X1 + F X2 = 0,
X2 F X1 = 0,
(5.2.29)
where F = 2 B. In this case the momenta are measured using the open string metric
Gij =
1
ij ,
1 + F2
(5.2.30)

and the conformal dimension of a boundary tachyon vertex operator ei pX is

p2
.
1 + F2
Moreover in the OPE there is a momentum dependent phase
h=
ei pX(t1 ) ei qX(t2 ) (t1 t2 )2 G
ij p q
i j
ei
ij
2
2)
pi qj i(p+
q )X(t
(5.2.31)
(5.2.32)
where the deformation parameter is

ij =
2 F ij
.
1 + F2
(5.2.33)
Comparing the conformal dimension of saa with (5.2.31) we see that p 2 = s 2 (1 + F 2 ).

Thus comparing the phase in (5.2.32) with the one in (5.2.25) we can identify
u
F (u) = cot
(5.2.34)
,
2
which is the expected result. Note that the magnetic field vanishes for u = + 2n, which
corresponds to the flat S1-brane or equivalently to Neumann boundary conditions on X1
and X2 . Changing the value of u we get the mixed NeumannDirichlet boundary conditions in (5.2.29) until we reach u = 2n, where the field-strength diverges and therefore
the boundary conditions become pure Dirichlet. In fact, precisely for these values of the
465
coordinate u, the two-dimensional conjugacy classes degenerate to points. According to

the analogy with open strings in a magnetic field, the strings that live on the brane worldvolume belong to the continuous representations since their ends are both subject to the
same magnetic field and then they behave as free strings. On the other hand, the strings
stretched between two different branes feel generically different magnetic fields and the
corresponding vertex operators are therefore twist fields or, in H4 terminology, they belong to the discrete representations. The twist for a string stretched from brane b to brane
a is given by [69]
ub ua
1
arctan F (ub ) arctan F (ua ) =
(5.2.35)
,
2
as expected. The effective field theory on the world-volume of an S1-brane is the limit of
the non-commutative field theory on the fuzzy sphere pertaining to the S 2 -branes in S 3 :
the volume and the magnetic flux are both scaled to infinity in order to obtain the noncommutative plane in the limit.
ab =
6. Four-point amplitudes
In the previous section we derived all the structure constants for the two families of
boundary CFTs that describe the D2- and the S1-branes of the H4 WZW model. These
are the essential ingredients for the solution of the models. We may now compute arbitrary
correlation functions by sewing together the basic one, two and three-point amplitudes.
Here, we are going to discuss in this section only those disc amplitudes that can be expressed in terms of the four-point conformal blocks, namely amplitudes containing either
two bulk fields or one bulk and two boundary fields or four boundary fields
.
In the following, we will provide some examples for each type of amplitude, both for
the D2- and the S1-branes. There are of course many other amplitudes one could consider,
besides the few we are going to discuss here. In general, one can write for them a decomposition in terms of our conformal blocks and structure constants but we expect that using
series and integrals more general than the ones we use in this paper one should be able to
find a closed form for many of them. It will be interesting to analyze their properties in
detail, both from the CFT and the string theory point of view.
As in the previous section, in order to avoid writing unnecessarily large formulae, we
denote with a single number in round brackets all the variables a vertex operator depends
on. This includes its insertion point and the charge variables. The four-point conformal
blocks we will need in the following computations are displayed in Appendix C. We choose
the gauge
A(z1 , z2 , z3 , z4 ) =
4

j >i=1
zij3
hi hj
A(z),
z=
z12 z34
,
z13 z24
(6.1)

where h = 4i=1 hi , zij = zi zj and the zi can represent the insertion points of either a

bulk or a boundary vertex operator. Finally we set = 4i=1 i .
466
6.1. The D2-branes

For the D2-branes we only display a very simple amplitude

2 1

2 1
A = p,1 2 (1)p,
(2)p,1 2 (3)p,
(4) .
(6.1.1)
In this case the integral over the conformal blocks can be performed explicitly and we
obtain

1
2
2
2 sin p
12
14 p(x1 x2 +x3 x4 ) x1 x3
A(z) = z (1 z) e
x4 x2
c1 (z)c1 (1 z)

x sin p
x sin p
xpzxz(1z) ln c1 (z) 2c (z)c (1z)
1
1
,
e
I/2
(6.1.2)
2c1 (z)c1 (1 z)
where c1 (z) = F (p, 1 p, 1, z), x = (x1 x3 )(x2 x4 ) and I is a modified Bessel function. Moreover
h
12 = h1 + h2 + p 2 + (1 2 )p p,
3
h
14 = h1 + h4 + p 2 + (1 4 )p p.
(6.1.3)
3
Even though this amplitude does not depend on the brane parameters i , it is interesting
since we can expand it in powers of the charge variables and compare the correlator of
the ground states of the H4 representations with the correlator of open strings living at the
intersection of two branes at an angle [91,92], which can be described by boundary twist
fields. The two expressions indeed coincide upon identifying the light-cone momentum
p with the angle formed by the two branes, as expected given the relationship between
the primary vertex operators in the NappiWitten background and the orbifold twist fields
[35,38,39]. Our open string amplitudes in the gravitational wave can be thought of as generating functions for the correlators of arbitrarily excited boundary twist fields in orbifold
models or equivalently of open strings living at the intersection of a configuration of branes
at angles.
6.2. The S1-branes
For the S1-branes we provide several explicit examples of four-point amplitudes. As
we mentioned in the introduction, the spacetime interpretation of these correlation functions deserves further investigation, in particular those involving on-shell open string states
stretched between branes localized at different positions in the u direction.
We start from the bulk two-point functions on the disc. This gives the first correction to
the propagation of the closed strings due to the presence of the S-brane. Using the results
of Section 5 and the conformal blocks in Appendix C we obtain

p,
q,
1
12
14 ei(pq)a +i(1 +2 )ua
px1 x1 qx2 x2 z (1 z)
=
1
e
a
a
1 eiua
(p) (q)
ex(1z)qxz(1z) ln B(z)
,
B(z)
(6.2.1)
467
where (ua , a ) are the brane parameters,

B(z) =
F (q, 1 p, 1 p + q, z)
F (p, 1 q, 1 q + p, z)
eiua zpq
, (6.2.2)
(p)(1 p + q)
(q)(1 q + p)
and
z=
|z1 z2 |2
,
|z1 z 2 |2
x = (x1 x2 )(x2 x1 ).
(6.2.3)
Here
12 = h1 + h2
h
+ pq 2 p + 1 q q,
3
h
+ p 2 + 21 p p.
(6.2.4)
3
When p = q only the identity couples to the brane in the closed channel and the correlator
simplifies
14 = 2h1
p,
p,
1
= 1a epx1 x1 qx2 x2

a
ei(1 +2 )ua z12 (1 z)14

c1 (z)
4 sin2 u2a
ex(1z)pxz(1z) ln c1 (z) ,
(6.2.5)
where
c1 (z) = F (p, 1 p, 1, z).
(6.2.6)
We also display the closely related amplitude

+
+
p,
q,
1
12
14
px1 x1 qx2 x2 z (1 z)
=
1
e
a
a
(1 p) (q)
ei(p+q)+i(1 +2 )ua ex(1z) ln D(z)

,
1 eiua
D(z)
(6.2.7)
where
D(z) =
F (1 p, 1 q, 2 p q, z)
F (p, q, p + q, z)
eiua z1pq
(p + q) (1 p)
(p q) (q)
(6.2.8)
and
z=
|z1 z2 |2
,
|z1 z 2 |2
x = (x1 x2 )(x1 x2 ).
(6.2.9)
In this case
12 = h1 + h2
14 = 2h1
h
+ (1 p 1 )(p + q) (1 + 2 )p q
3
h
+ pq + (1 + 2 )p p.
3
(6.2.10)
468
We consider finally
+

0
p, s,
2 a = 1a
eipa +i(1 +2 )u px1 x1 s (x1 x2 +x1 x2 + xx1 + xx1 )

2
2
2
e
sin p
1 eiua
h
zh1 3 p2 (1 z)s

n
xzp eiua ,
2 h
3
s2
2 ((p)+(1p)2(1))
exa(z)
b(z)
x
(6.2.11)
nZ
where x = x2 x2 and
a(z) =
s2
F (p, 1, 1 + p, z),
2p
b(z) =
s2
F (1 p, 1, 2 p, z).
2(1 p)
(6.2.12)
We now turn to the four-point open string amplitudes. The cross ratio in this case is
t34
. The first amplitude we consider is
z = tt12
13 t24
ab

ba
(1)(p,
(2)saa
(3)saa
(4) ,
An1 ,n2 ,s1 ,s2 (p,
1
2
ab n )
ba +n )
1
p = pb pa > 0.
(6.2.13)
It describes the correlation between two open strings stretched between the branes a and b
and two open strings ending on the brane a. We obtain
x
px x 1 (
s1
s2
) 2 (s1 x3 +s2 x4 ) n1 n2
2
x3
(1 w)12 w 14
An1 ,n2 ,s1 ,s2 = 1a e 1 2 2 x3 x4

s2 + s1 xw p n1 n2
2
s12 +s22
4 [2(p)2(1)
cot pa ]
n1 n2 +1
s1 s2 xa(w)
(cot p+cot pa )wp
wb(w)
2 [ p + x(1p) +
x
sin pa

s1 s2 x s1 s2 up
n1 n2
2
2
(cot p + cot pa ) s1 + s2 +
,
Ln2
+
2
wp
x
(6.2.14)
where w = 1 z and
a(w) = F (p, 1, 1 + p, w),
Moreover x =
x4
x3
b(w) = F (1 p, 1, 2 p, w).
(6.2.15)
and
s12 + s22 h
h
,
14 = h1 p 4 .
(6.2.16)
2
3
3
Another interesting amplitude is
ab

ba
ab
ba
An1 ,n2 ,n3 ,n4 (p,
(1)(p,
(2)(p,
(3)(p,
(4) ,
ab n )
ba +n )
ab n )
ba +n )
12 =
(6.2.17)
469
which computes the correlation between four open strings stretched between the a- and the
b-branes. When n1 = n3 = n, n2 = n4 = m we can use the integral (H.9) and the result is
An,m,n,m = ep(x1 x2 +x3 x4 ) (x1 x3 )2(nm) z12 (1 z)14
n! 1a
m! sin pa

sin p sin pa
sin p 2(nm)+1 xpz xp2 xz(1z) log c1 (z)+ xc
1 (z)R (z)
e
R (z)

m

(m l + 1/2)(l + 1/2) R+ (z) 2l
(1)l
(l + 1 + n m)(m l)! R (z)
l=0

x sin p sin pa sin pb
2(nm)
,
L2l
R+ (z)R (z)

(6.2.18)
where
R (z) = c1 (z) sin pb c1 (1 z) sin pa ,
c1 (z) = F (p, 1 p, 1, z).
(6.2.19)
Moreover x = (x1 x3 )(x2 x4 ) and

h
+ p 2 + 2ab n m p p,
3

h
14 = h1 + h4 + p 2 + 2ab n m p p.
(6.2.20)
3
For the D2-branes we showed that the H4 boundary amplitudes can be considered as
generating functions for the open string amplitudes in models with intersecting branes. In
the case of the S1-branes one can show in exactly the same way that the boundary amplitudes are generating functions for the open string amplitudes in models with magnetized
branes. For instance, the amplitude between the ground states of the H 4 representations,
which can be easily extracted from (6.2.18), coincides when pa = n = m = 0 with the
amplitude computed in [93]. This result is not surprising since the magnetized and the
intersecting branes in toroidal compactifications are related by the operation of charge
conjugation, as it is also the case for the S1- and the D2-branes in the NappiWitten gravitational wave.
For the sake of completeness, we also present a correlator with one bulk and two boundary vertex operators. There are four types of such correlators
+ aa aa
0 ab ba
0 aa aa
+ ab ba
,
,
,
,
(6.2.21)
12 = h1 + h2
and the example we chose is

+ ab

ba
p, q,ab n q,
ba +n2
1
+1 ipa +i ua

e
= 1a
epx1 x1 qx3 x4 (x1 x3 ) z12 (1 z)14
sin p sin pa
1 eiua

pq
D(z)n2 sin pb x(1 z)pq
x[g1 (z)+ (1z)
c1 (z)B(z) ]
.
e
L
(6.2.22)
B(z)n1 +1 n2 sin q sin pa B(z)D(z)
470
z1 )
The cross-ratio in this case is z = (z(t13
t4 ) and we have also defined

(p) i
B(z) = (1 pa ) (q)
+
cot p cot pa c1 (z) c2 (z),
2
2

(p) i
+
cot p cot pa c1 (z) c2 (z),
D(z) = (pa ) (1 q)
2
2
(6.2.23)
and
h
+ p 2 + 21 p p,
3

h
14 = h1 + h4 + pq ba + n2 p + 1 q q.
3
12 = 2h1
(6.2.24)
7. Annulus amplitudes
All the amplitudes we discussed so far, were defined on the disc. The consistency conditions of a boundary CFT impose a constraint also on the one-point functions of the
boundary fields on the annulus. When the boundary field is the identity, this additional
constraint reduces to the Cardy constraint, which interchanges the two equivalent interpretations of the annulus diagram. The first is the partition function of the boundary CFT,
when time is running along the boundary. The other is as a tree-level amplitude for the
propagation of the bulk states when time is running perpendicular to the boundary. In this
second case the boundary conditions are specified by the introduction of two boundary
states. Passing from one description to the other requires an S modular transformation
t 1/t, where t is the modulus of the annulus.
The Cardy constraint has been at the origin of many important insights concerning the
operator content of a rational boundary CFT [77,80,82]. It has also been exploited for the
investigation of some non-compact models [4448]. The best way to analyze the annulus constraint is to introduce characters for the representations of the chiral algebra and
study their modular transformations. This is a non-trivial problem for conformal -models
describing non-compact curved spacetimes or time-dependent branes [47,48,9496].
The characters of a generic representation of the affine H 4 algebra are defined as
follows

c
(z, v| ) = trH q L0 24 e2(zJ0 +vK0 ) .
(7.1)
For the w [Vp,

] representations we have
h
+ w (1+w)
1
iq p,;w 2
e2iz(w 2 )iw(p+w)v .
p,;w (z, v| ) =
( )1 (z| )
+
= 1p,
, as required by (3.4). It is therefore convenient
Note that p,
;1w
;w
+
everything in terms of the p,;w characters, letting w Z. It is useful to define
r = p + w,
t =
1r
w,
2
(7.2)
to express
(7.3)
471
and rewrite the character as

+
p,
(z, v| ) =
;w
r
ieiw
e2i rt e2irv2iz(t 2 ) .
( )1 (z| )
(7.4)
We can derive the following S modular transformation

z v 1
,
p+ , ,w
1 1 1
=e
1

2i vz

dp2
w2 Z 0
d 2 S(p1 ,1 ,w1 );(p2 ,2 ,w2 ) p+ , ,w (z, v, ),

2
(7.5)
where
S(p1 ,1 ,w1 );(p2 ,2 ,w2 )
= ei(w1 w2 1) e2i(p2 +w2 )(1
1p1 w1
1p2 w2
w1 )+2i(p1 +w1 )(2
w2 )
2
2
= ei(w1 w2 1) e2ir1 t2 +2it1 r2 .
(7.6)
The characters of the w [Vs,0 ] representations are

s2
s,0 ;w (z| ) =
q 2 2ik
e
(z + w + k),
4 ( )
(7.7)
kZ
and they have the following modular transformation properties

z 1
0
s1 ,1 ;w1

=
1/2

w2 Z 1/2

d 2
0
s2 ds2 S(s
0
(z| ),
1 ,1 ;w1 );(s2 ,2 ;w2 ) s2 ,2 ;w2
(7.8)
where
0
= 2ie2i(w2 1 +w1 2 ) J0 (2s1 s2 ).
S(s
1 ,1 ;w1 );(s2 ,2 ;w2 )
(7.9)
Note that when s = 0, all the representations with R are inequivalent and their base is
one-dimensional. The corresponding characters are
0
0,
;w (z| ) =
e2i (z+w )
.
4 ( )
(7.10)
The torus vacuum amplitude of the NappiWitten gravitational wave [39,97] can be
expressed in terms of the H 4 characters. The discrete series contribution to the closed
string partition function is given by
472
Z + (, z; , z )

c
c
= Tr q L0 24 e2zJ0 q L0 24 e2 z J0 =
1
d
dp
0
1
| 1 |2
1

d
dp
e22 [(p+ 2 )

+

p,;w
2
(z, )
wZ
1 2
( 12 w)2 ]+4
Im z(w 12 )
(7.11)
wZ
Changing variable in each term of the sum l = w 12 and setting p = p + w we obtain

1
Z + =
|1 |2
d l

d p e22
[(p+
)2 2 ]+4
2 (Im z)
2
ie
Im z
=
,
22 |1 |2
(7.12)
where we performed the rotation p i p in order to evaluate the Gaussian integral. Similarly the contribution of the type-0 characters is
Z = V2
0
1/2

wZ 1/2

d

2
ds s s,0 ;w (z| ) =

V2
(z + w + k)2 ,
42 4 4
w,kZ
(7.13)
where V2 is the volume of the transverse plane. This additional volume factor is a consequence of the fact that the states that belong to the continuous representations can move
freely in the transverse plane and their wave functions are only delta function normalizable. On the other hand, the discrete states are confined around the origin of the transverse
plane and have normalizable wave functions.
We turn now to the annulus amplitudes. In the closed channel they can be constructed
using suitable boundary states. In the open channel they encode the spectrum of the open
strings ending on the two given branes. We will display the amplitudes in both channels and
investigate how they are related by the S modular transformation. In the following we will
make several assumptions and formal manipulations and the results we obtain even though
apparently consistent are not rigorous and we think that the modular properties of these
amplitudes deserve further study. We will use the short-hand notation (z| ) and
(z/ |1/ ) . Boundary states for the H4 WZW model3 can be easily constructed
using the bulk-boundary couplings derived in Section 5. The boundary state for a D2-brane
localized at is
1/4
| = 1 V2
1/4
ds s B s, |s, ; 0
=0,1/2 0

ds cos 2s |s, 0; 0 + i sin 2s |s, 1/2; 0 .
= 1 V2
3 Boundary states for the H WZW model were also considered in the recent paper [98].
4
(7.14)
473
1/4
Here we have a factor of V2 , since the boundary continuous representations correspond

to one-dimensional waves. The annulus amplitude in the closed string channel for two
branes localized at 1 and 2 then reads
1/2
A 1 2 = 11 12 V2
= 11 12 V2
+ sin
21 s sin
B s,
=0,1/2
0
1/2
ds s 2
B s, s,0
(7.15)

0
ds cos 21 s cos 22 s s,0;0

0
.
22 s s,1/2;0
(7.16)
On the other hand, since the spectrum of the BCFT contains all the H 4 representations, the
annulus amplitude in the open channel reads
A1 2 =
1

w=0 0

dp
1/2
+
d p,
;w
1/2
+ 2V2
d s (w)
wZ 0
d s,0 ;w ,
(7.17)
1/2
(w) 2
iw | in order
where s 2 (w) = s 2 + [ 2
2
2 ] . We introduced the quantity (w) = |1 e
to specify the domain of the integral in s. The contribution of the continuous representation
can also be written as
1/2
A1 2 = 11 12 V2
1/2
d
dt
wZ 2 (w)
2
1/2
t
t2 [
2 (w) 2
2 ]
t,0;w .
(7.18)
Note that the even and odd spectral-flowed continuous representations appear with different
ranges of integration in the partition function, a remnant after the Penrose limit of the
different density of the corresponding states for the AdS2 -branes in AdS3 [76].
If we compute the modular transformation of the transverse annulus using the S matrix
in (7.9), we correctly reproduce the spectrum of the continuous representations in the direct
annulus. It is less clear how the discrete contribution should appear. By comparing the
normalization of the annulus amplitude in the direct and in transverse channel we can fix
the one-point function of the identity. We have
1 = 2.
(7.19)
The discussion for the S1-branes is similar. The boundary state for an S1-brane labeled by
(u, ) is
1

dp
d u, B p,;w |p, , w
|u, = 1(u,)
wZ
1/2
+ V2
1/2
1/2

d
ds s
0
u,
B s,;w |s, , w .
(7.20)
474
The bulk-boundary couplings with the identity are

e2i(p+w)+iu(w)+iw
u,
B p,;w =
,
sin p
1 eiu
(7.21)
and we recall the relations u = 2(q + a) and 2 = (2q + 2l 1). The annulus amplitude
in the closed channel reads
A 12 = 1a 1b
1

wZ 0
+ 1a 1b V2
dp
wZ
e2i(p+w)(1 2 )+i(w)(u1 u2 ) +
p,;w
sin p
(1 eiu1 )(1 eiu2 )
ei(w)(u1 u2 )+2iw(1 2 )
16 sin2 (u1 /2) sin2 (u2 /2)
0
0,
;w ,
(7.22)
where we used the fact that R also for the continuous representations when s = 0. In
the previous expression we separated the contribution of the discrete and the continuous
representations, which are weighted by different volume factors. In fact the first term has to
be considered as a regularized term without the divergences due to the almost delocalized
states in the transverse plane with p w, w Z and the second term as the term containing all these divergences, since the continuous representations capture the behavior of
the discrete representations when p approaches an integer value. What this means is that
when manipulating the term containing the discrete representations, whenever a constraint
arises forcing to evaluate it at the boundary of the interval 0 < p < 1, the corresponding
contribution should be discarded and only the contribution coming from the continuous
representations in the second term of (7.22) retained. Equivalently, we could keep only the
first term in (7.22) and take into account the divergent behavior of the integrand when p
becomes an integer. We will show in the following that both points of view lead to the same
result.
Let us transform the amplitude (7.22) to the open string channel using the S matrix in
(7.6), (7.9) writing for instance
+
=
p,
;w
dq
+
d l S(p,;w);(q,l;a)
.
q,l;a
(7.23)
aZ 0
We perform first the integration over which gives the constraint

u2 u1
(7.24)
.
2
Let us suppose for the moment that u2 u1
/ 2Z so that only the discrete representations
contribute. We can recombine the integral over p and the sum over w in a single integral
over p = p + w
q +a=
i(u1 u2 )
1a 1b e 2
A12 =
(1 eiu1 )(1 eiu2 )
d l
d p
12 +2(l+q)) +
ei p(2
.
q,l;a
sin p
(7.25)
475
We now need a prescription to perform the integral over p.

The prescription that reproduces in the open channel the spectrum of the boundary operators is the following
p p + (1)a i.
(7.26)
Consider for instance the case a even. We can expand
e2ip(n+ 2 ) ,
= 2i
sin (p + i)
(7.27)
n=0
and then perform the integral over p that gives the constraint
l = 1 + 2 n,
n N.
(7.28)
Therefore
i(u1 u2 )

2i1a 1b e 2
+
q,
.
A12 =
1 +2 n;a
iu
iu
1
2
(1 e )(1 e
)
(7.29)
n=0
This is precisely the expected result for the annulus amplitude in the open string channel.
The reason is that when u2 u1 = 2(q + a) with 0 < q < 1 and 2 1 = (q + ), the
open string spectrum only contains discrete spectral-flowed representations
A12 =
+
q,
.
n;a
(7.30)
n=0
From the comparison of the amplitudes in the open and closed string channel, we can fix
the last structure constants required for the complete solution of the boundary H4 model,
namely the one-point functions of the identity. We obtain

u
2
sin
.
1u, =
(7.31)
2
When u2 u1 = 2k with k Z, the constraint (7.24) has two solutions, q = 0, a = k
and q = 1, a = k 1. We have two options. The first is to think as proposed before, that
the term that contains the discrete representations is a regularized term. In this case, it does
not contribute while the term containing the continuous representations after the modular
transformation gives
i(u1 u2 )
V2
2i1a 1b e 2
A12 =
iu
iu
(1 e 1 )(1 e 2 ) sin2 u2

ds ss,0 2 1 ;k .
(7.32)
This is the expected result, since whenever u2 u1 = 2k and 2 1 = the open

string states belong to the continuous spectral-flowed representations
A12 =
V2
sin2
u
2
ds ss,0 ;k .
0
(7.33)
476
The second option is to think of the continuous representations as already included

in the divergent behavior of the discrete representations for p close to an integer. It is
instructive to derive (7.32) once more, adopting this point of view, and to show explicitly
its equivalence with the previous one. In this case, we have to extract (7.32) from the
integral over the discrete representations when the integrand is evaluated at the extrema of
the interval. Performing the same steps as before and keeping both contributions q = 0,
a = k and q = 1, a = k 1, we obtain
i(u1 u2 )

2i1a 1b e 2
+
+
1,
0,
.
A12 =
n;k
+
+n;k1
iu
iu
1
2
1
2
1
2
(1 e )(1 e
)
(7.34)
n=0
Using the explicit form of the + characters and sending k k + i, the previous expression can be rewritten as follows for 0
i(u u )
1 2
1 1 2i(2 1 )m
2i1a 1b e 2
A12 =
e
(z + k + m).
(1 eiu1 )(1 eiu2 ) 4 ( )
(7.35)
mZ
We should now interpret the divergence 1/ as due to the infinite volume of the transverse
plane (consistently with the power we have of the modular parameter which is the one
commonly associated with two non-compact directions)
lim
0
V2
1
open
=
= V2 ,
sin2 u2
(7.36)
which is the volume measured using the open string metric [89]
1
1
G1
op (gcl + F) gcl (gcl F) .
(7.37)
In this way, we obtain again the amplitude displayed in (7.32).

As we mentioned earlier, our aim in this section is not to provide a rigorous discussion of the modular transformation properties of the annulus amplitudes, but rather to give
a plausible suggestion about how the standard openclosed duality should work for the
branes of the H4 model.
7.1. Contraction of the BCFT
The Penrose limit that connects the NappiWitten gravitational wave and R S 3 or
AdS3 S 1 can also be extended to the branes contained in these spacetimes, as described
in Section 2. In [35] we gave a detailed description of the world-sheet equivalent of the
Penrose limit in spacetime. This is the contraction of the R SU(2)k WZW model to the
H4 WZW model. It is interesting to perform the same contraction for the boundary CFT.
Here we shall comment on how the BCFT describing the H4 -branes arises as the limit
of the BCFT describing the S 2 -branes in S 3 . We first recall how to derive the affine H 4
characters from the contraction of the U (1) SU(2)k characters [35]. It is convenient to
write the latter as follows

+

l,n
,
+ l,n
lk =
(7.1.1)
nZ
477
where l is the spin of the representations and
(z| ) =
l,n
2l+1 2
2l+1
1 2i(k+2)[(n+ 2(k+2)
) (n+ 2(k+2)
)z]
e
.
i1
(7.1.2)
The U (1) characters are

Q (z| ) =
e2i
Q2
k +2izQ
(7.1.3)
The characters of the original CFT become the H4 characters if we send the level k to
infinity and scale simultaneously the spin l and the charge Q in a correlated way
(z| ) (z, v| ).
Q (z 2v/k| )l,n
(7.1.4)
More precisely for the discrete representations we obtain

+
p,
,
k(n+ p ) l,n
;2n
2
n 0,
+
+
k(m+ p ) l,m
1p,
,
;2m1
2
+
k(n+ p ) l,n
p,
, n 0,
;2n
2
k(m+ p ) l,m
1p,
,
;2m1
2
m 1,
m 1,
(7.1.5)
with l = k2 p . The characters for the continuous representations require a different

scaling, namely

+
k

0
s,
kn+ l,n + l,n s,;2n , n Z, l =
2

+

k
k
0
s.
k(n+1/2)+ l,n + l,n1 s,;2n1 , n Z, l =
(7.1.6)
2
2
In the following, we will only discuss the contraction involving the SU(2)k WZW model.
Similar relations however, can be written for the SL(2, R)k characters. The boundary state
for an S 2 -brane in S 3 reads
k/2

Sil
l
|i, R =
(, , )|l, m, n.
Dm,n
S
0l
l=0
(7.1.7)
Here, we are considering the general case where the gluing condition (2.7) also involves
l (, , ) are the matrix
the adjoint action of a group element R(, , ) SU(2). The Dmn
elements of R in the spin l representation and can be expressed in terms of the Jacobi
polynomials
l
l
(, , ) = i mn ei(n+m)i(nm) Pmn
(cos ).
Dmn
Finally Sil is the modular S matrix of the affine SU(2)k algebra

2
sin
(2i + 1)(2l + 1) .
Sil =
k+2
k+2
(7.1.8)
(7.1.9)
478
Table 2
+
p,
,2n
+
p,
,2m1
p,
,2n
p,
,2m1
u = 2(p + 2n)
2 = (2p + 2 1)
k(n+ p ) k p
u = 2(p + 2m 1)
2 = (2p + 2 1)
u = 2(p + 2n)
2 = (2p 2 1)
k(n+ p ) k p+
u = 2(p + 2m 1)
2 = (2p 2 1)
k(m 1p
2 )
2
k (1p)+
2
(1p)
k(m 1p
2 ) 2
Using the fact that for R = 1 the label i of the branes is related to the coordinate by
(2i + 1)
k
(7.1.10)
, i = 0, . . . , ,
k+2
2
we can derive the relation between the S1-brane parameters and the quantum numbers of
the H 4 representations that is inherited from the original relations between the S 2 -brane
parameters and the spin of SU(2). The results are summarized in Table 2. In the first column
we listed the discrete H4 representations. In the second and in the third column we show the
labels of the corresponding branes. Finally, in the fourth column, we list the U (1) SU(2)k
representations they originate from in the Penrose limit
In a similar fashion, one may show that the label of the D2-branes is related to the
label l of the original S 2 -brane by
k
l= + k .
(7.1.11)
4
2
i =
We may now write down the annulus amplitudes for the brane configurations in R S 3 ,
whose Penrose limit is one of the branes in H4 , as explained in Section 2. We will show
that in the limit, the direct and the transverse annulus amplitudes become the corresponding
amplitudes for the H4 model we discussed in the previous section. In order to do this, we
have to first understand for each amplitude how we can scale the quantum numbers of the
original representations and then restrict our attention to states that have finite charges and
conformal dimension in the limit.
For the D2-branes, we start with a brane with Neumann boundary conditions along the
time direction. The original annulus amplitude in the open string channel is

Al1 ,l2 =
dQ
min(l1 +l
2 ,kl1 l2 )
Q l .
(7.1.12)
l=|l1 l2 |
The brane labels have to be scaled in the limit as li = k4 + 2k i , i = 1, 2, and therefore the
range of the possible SU(2)k representations is
k
k
k
(7.1.13)
|1 2 | l
|1 + 2 |.
2
2 2
Since there is a lower bound in the range of l, we have to slightly change the way we scale
+
we simply set
the spin in the Penrose limit. For the representations p,
;2a

|1 2 |
|1 2 |
p
k
,
Q = k w +
+ k
,
l= p+ k
(7.1.14)
2
2
2
2
479
and similarly for all the other discrete representations. For the continuous representations,
the lower bound in l shifts the lower bound for the integral in s. It is easy to se that
|1 + 2 |
|1 2 |
, for s,0 ;2w ,
, for s,0 ;2w+1 .
s
(7.1.15)
2
2
We observe that the different behavior of the even and odd spectral-flowed continuous representations, arises also in a very transparent way from the contraction of SU(2)k . Note that

2 )2
, which
the minimal conformal dimension for the vertex operators s,2;01 is h = (12
2
can be ascribed to the tension of the string stretched between the two branes, as expected.
Since the original amplitude contains arbitrary U (1) charges, in the limit we obtain an
amplitude that contains all possible H4 representations
s
A1 2
1

dp
w=0 0
1/2
+
+ 2V2
d p,
+
;w
p,;w
wZ 0
1/2
d s
d s,0 ;w .
1/2
(7.1.16)
In the transverse channel we can reason in the same way. The original amplitude is
A l1 l2 = 0
k/2

S l1 l S l2 l
l .
S0l
(7.1.17)
l=0
Since now all the states have zero U (1) charge, according to (7.1.5)(7.1.6) we can only
obtain in the limit the highest-weight continuous representations.
The discussion for the S1-branes is similar. We label the branes with their position in
time u and with the SU(2) spin l. In the open string channel, the original amplitude is
A(u1 ,l1 )(u2 ,l2 ) = u2 u1
4
min(l1 +l
2 ,kl1 l2 )
l .
(7.1.18)
l=|l1 l2 |
We scale li = k2 pi i , i = 1, 2. As before we have to distinguish two cases. When u2 u1

is not an integer multiple of 2 , we may write

p
u2 u1
=k w+
,
(7.1.19)
4
2
with 0 < p < 1. We then have to scale l as l = k2 (p2 p1 ) and the possible values of
follow from the original range of l
2 1 ,
(7.1.20)
in integer steps and therefore = 2 1 n, n N, as expected. On the other hand, when

u2 u1 = 2(kw ), we may use the relations (7.1.6). In the limit, we obtain an annulus
amplitude that only involves the continuous representations.
In the closed string channel, the original amplitude is
A (u1 ,l1 )(u2 ,l2 ) =
dQ eiQ(u2 u1 ) Q
k/2

S l1 l S l2 l
l .
S0l
l=0
(7.1.21)
480
As in the amplitude (7.1.12), we have again arbitrary U (1) charges. We thus obtain in the
limit all possible H4 representations. Therefore, the annulus amplitudes, both in the closed
and in the open string channel, reproduce in the limit the results we expect for the D2- and
the S1-branes. It would be very interesting to pursue this line of thinking, in order to gain
a more detailed understanding of the contraction of the boundary CFT.
8. The DBI approach

We will study here the DBI approach for the branes described in this paper. This approach although in most cases approximate, has the advantage of an obvious geometric
interpretation. We will also be able to provide an independent confirmation of the spectrum of fluctuations for the H4 -branes and justify some of the assumptions made during
the solution of the BCFT. In the bosonic case the lowest state is the tachyon. In anticipation
of the supersymmetric case we will use the bosonic part of the supersymmetric DBI action
that describes the dynamics of the massless modes. To simplify the formulae, we use
here the background (2.1) and (2.2) with = 2 so that the metric and antisymmetric tensor
read
ds 2 = 2 du dv r 2 du2 + dr 2 + r 2 d 2 ,
Br = 2ur.
(8.1)
We can put back by rescaling u u/2, v 2v/.

8.1. The S1-branes and the spectrum of their fluctuations
We will find a class of solutions to the DBI equations that will contain as special cases
the S1-branes discussed in this paper. The S1 will have Dirichlet boundary conditions on
the u, v coordinates. We choose a static gauge where the brane world-volume is parameterized by r, . The induced metric and antisymmetric tensor is
g rr = 1 2u v r 2 u 2 ,
g r = u v uv
r 2 u u,
g = r 2 r 2 u 2 2u v,
B r = 2ur.
(8.1.1)
(8.1.2)
In the formulae above, a dot stands for a derivative and a prime for an r derivative.
We can directly evaluate the NambuGotoDiracBornInfeld Lagrangian as

2 + (2ur + F )2 ,
L = det(g + B + F ) = g rr g g r
(8.1.3)
r
where Fr is the world-volume gauge field strength. The equations of motion for the gauge
field can be integrated to
Er
E
2ur + Fr
1 2u v r 2 u 2 ,
= 2ur + Fr =
(8.1.4)
L
2
4 E2
where E is a constant (the electric field). The u, v equations are
(r 2 u + v)
g rr (r 2 u + v )g r
(r 2 u + v)
g r + (r 2 u + v )g
+ r
= Er,
L
L
(8.1.5)
481
u g rr u g r
u g r u g
r
= 0.
(8.1.6)
L
L
We will from now on consider a rotationally invariant ansatz. Dropping -derivatives the
equations simplify to

2

1 2
B

r u + v g = Er + A L, u g = L,
(8.1.7)
2
2
with A, B integration constants and L from (8.1.3) given by

2r
L=
1 2u v r 2 u 2 .
4 E2
(8.1.8)
Massaging (8.1.7) and (8.1.8) we obtain

B2
v
Er2 + 2A

2 2
2
1
2u
,
r
v
r
u
+
=
,
u
B
4 E2
2A (B + E)r 2
B2
u,
.
v =
u 2 =
B
(4 (B + E)2 )r 2 + 4AB
r 2 u 2 =
(8.1.9)
(8.1.10)
We will look here for solutions where u is a constant corresponding to the symmetric S1branes discussed in this paper. u = constant implies that B = 0 and

A
2
1
Er2
2A
v = Er +
v = v0
+
log r. (8.1.11)
2
2
2
r
4E
2 4E
4 E2
Since the class variable is = 2v sin u r 2 cos u, we learn that the symmetric S1-branes
have also A = 0. Comparison with (2.12) and (2.14) gives
E
cot u0 =
,
4 E2
= 2v0 sin u0 .
(8.1.12)
The fluctuations around the classical embedding are in one-to-one correspondence with
on-shell open marginal deformations. Although for a single S1-brane this is not very rich,
it is still useful to verify it explicitly. The richer case of two branes at a non-zero distance
in light-cone is much harder to analyse and we will not do it here.
In Appendix F we analyse the action for the fluctuations u, V , F around the S1 solution
that turns out to be
2 2

4 E2
8ur
+ r 2 u V + 32 r u
.
4 E2 F +
L2 =
+
8
u
V
16r
4 E2
4 E2
(8.1.13)
The equations of motion that ensue are
2u = 0
1
1
(ru ) + 2 u = 0,
r
r
1
1
4
(2u + C),
2V = (rV ) + 2 V =
r
r
4 E2
(8.1.14)
(8.1.15)
482
with the gauge field satisfying

F + 2ur
E2 2
2C
r
r u =
4 E2
(4 E 2 )
(8.1.16)
with C a constant.
The regular solution of (8.1.14) is u = u0 constant. On the other hand, the solution of
(8.1.15) is
V = V0
2u0 + C 2
r ,
4 E2
(8.1.17)
In order to be regular at r = , the electric field fluctuation and the u fluctuation must be
related by
C = 2u0 .
(8.1.18)
This is indeed implied by (8.1.12). In the non-symmetric case where (8.1.12) is no longer
valid, (8.1.18) must still be in effect for the fluctuations to be continuum normalizable and
thus physical states of the theory.
The two physical states obtained correspond to K1 |s = 0 and J1 |s = 0 in the
K |s = 0 and J
bosonic case and 1/2
1/2 |s = 0 in the supersymmetric case [99] in accordance with the BCFT discussion. Note that here, unlike the D-brane case, including
the contribution of the additional coordinates of the ten-dimensional string theory does not
change our results. The reason is that the world-sheet is Euclidean and the physical states
conditions are very restrictive, implying the vanishing of all momenta.
8.2. The D2-branes and the spectrum of their fluctuations
The Cartesian coordinates on the plane (x, y) are more convenient here. The metric and
antisymmetric tensor read

ds 2 = 2 du dv x 2 + y 2 du2 + dx 2 + dy 2 ,
(8.2.1)
Bxy = 2u.
Putting Dirichlet boundary conditions on y we obtain the following induced metric

d s 2 = x 2 y 2 + yu2 du2 + yv2 dv 2 2(1 yu yv ) du dv + 2yu yx du dx

+ 2yv yx dv dx + 1 + yx2 dx 2 ,
(8.2.2)
while
B = 2u(yu dx du + yv dx dv).
The action is
SD2 =
with
dx du dv

1 + L2 ,
(8.2.3)
(8.2.4)

L2 = 2yu yv + x 2 + y 2 yv2 + (2uyv + Fvx )2 + yx2

2 2
2
1 + yv2
2(2uyu + Fux )(2uyv + Fvx ) Fvx
yu Fuv
2 2
yv + 2Fuv Fux yv yx 2Fuv Fvx yu yx .
+ 2Fux Fvx yu yv + Fux
(8.2.5)
483
We will now search for solutions where the D2-brane is sitting at y = y0 constant that
contain the symmetric D2 solutions studied in this paper.
Setting y = y0 we obtain the following equations to be solved

2

Fux
Fuv
2 Fvx
+ x y0 + x
= 0,
u
(8.2.6)
L
L
L

2

Fux
Fvx
2 Fvx
u
v y0 + x
= 0,
L
L
L
v
(8.2.7)
Fuv
Fvx
+ x
= 0,
L
L
with
L=
(8.2.8)

2 2F F + x 2 + y 2 F 2
1 Fuv
ux vx
vx
0
(8.2.9)
while the y equation gives4

Fvx = 0.
(8.2.10)
Eqs. (8.2.6)(8.2.8) are then solved by

Fuv = fuv = constant,
Fux = y0 + fux = constant.
(8.2.11)
The symmetric solution corresponds to fux = fuv = 0. The gauge field can be dualized to
a scalar here as follows

Fuv
Fux
Fvx
= x A,
= u A y02 + x 2 v A,
= v A.
(8.2.12)
L
L
L
Solving for the gauge field strength we obtain
Fuv =
x A
,
L
Fux =
u A (x 2 + y02 )v A
,
L
Fvx =
v A
.
L
(8.2.13)
Such expressions solve Eqs. (8.2.6)(8.2.7) but now the Bianchi identity gives
x
u A + (x 2 + y02 )v A
x A
v A
u
+ v
= 0,
L
L
L
where
L =

1 2u Av A + (x A)2 + x 2 + y02 (v A)2 .
(8.2.14)
(8.2.15)
Using (8.2.10), (8.2.14) becomes

x
x A
= 0,
L
v A = 0.
(8.2.16)
4 There is another possibility here, namely F = 2/y , but this does not correspond to a symmetric solution.
vx
0
484
Thus the A corresponding to our previous solution is

A=
fuv x + (fux y0 )u

.
2
1 fuv
(8.2.17)
We will now study the spectrum of fluctuations around the simplest solution y = y0 , Fvx =
Fuv = 0, Fux = y0 . Setting y y0 + y, Fvx Fvx , Fuv Fuv , Fux y0 + Fux and
expanding the action to quadratic order we obtain

1
2
+ x 2 (Fvx 2uyv )2
S2 =
2u yv y + (x y)2 + x 2 (v y)2 Fuv
2

2(Fux 2uyu )(Fvx 2uyv ) .
(8.2.18)
The ensuing equations of motion are
2u v y x2 y x 2 v2 y 2(Fvx 2uyv ) = 0,
(8.2.19)

u Fuv + x x 2 (Fvx 2uyv ) Fux + 2uyu = 0,
(8.2.20)

u (Fvx 2uyv ) v x 2 (Fvx 2uyv ) Fux + 2uyu = 0,
(8.2.21)
v Fuv + x (Fvx 2uyv ) = 0.
(8.2.22)
Introducing a dual scalar field A by

Fuv = x A,
Fux = 2uyu + u A x 2 v A,
Fvx = v A + 2uyv ,
(8.2.23)
the equations read

2A = 2yv2 y = 2v A,
(8.2.24)
where
2 = 2u v x2 x 2 v2 .
In terms of the dual variable, the quadratic action can be written as

1
1
S2 = du dv dx AA + yy 2Av y .
2
2
Defining a new complex scalar field as = (A + iy)eiu we find

1
1
S2 = du dv dx + .
4
4
(8.2.25)
(8.2.26)
(8.2.27)
Thus, is a massless scalar. Its solutions are in one-to-one correspondence with the discrete and the continuous representations in accordance with the BCFT discussion.
Since here the world-volume has Minkowski signature it is the eigenvalues of the Laplacians that are relevant when we include 6 extra flat coordinates in order to study strung
theory in the H4 R6 background. Thus we need to solve
= E.
(8.2.28)
485
We parameterize,
= eip u+ip+ v z,
and z satisfies the harmonic oscillator equation
2

2 2
x + p+
x z = (2p+ p+ E)z,
(8.2.29)
(8.2.30)
with quantized values of p . For p+ = 0 the equation becomes

x2 z = Ez,
(8.2.31)
and the solutions for z are plane waves in one dimension.

Thus, the spectrum is in agreement with the BCFT findings in Section 4.
9. Conclusions and generalizations

In this paper we provided the complete solution for the BCFT pertaining to the two
classes of symmetric branes of the H4 model, the D2- and the S1-branes. In both cases we
solved the consistency BCFT conditions [78,79] and obtained the BCFT data, namely the
bulk-boundary and the three-point boundary couplings.
The bulk-boundary couplings for the D2-branes can be found in Eqs. (5.1.6) and
(5.1.9) while the three-point boundary couplings are in Eqs. (5.1.12), (5.1.16), (5.1.22)
and (5.1.24). The bulk-boundary couplings for the S1-branes are in Eq. (5.2.7) while the
boundary three-point couplings can be found in Eqs. (5.2.22), (5.2.25) and (5.2.27). To our
knowledge, with the notable exception of the Liouville model [42,43,4951], this is the first
complete tree-level solution of D-brane dynamics in a curved non-compact background.
Our solution of the H4 model with and without a boundary should help to clarify the
properties of the non-compact WZW models and the closed and open string dynamics
in curved spacetimes. Among other results we provided the first example of structure
constants for twisted symmetric branes in a WZW model (the D2-branes) and of open
four-point functions in a curved background.
There are two aspects of our work we think deserve further study. The first is to perform
a more detailed analysis of the four-point amplitudes and the second to clarify the relation
between the open and closed string channel of the annulus amplitudes. There are also
several other issues it would be worth pursuing and we mention here a few. One is the
study of the symmetric branes of the other WZW models based on the Heisenberg groups
H2+2n , n 2. Their generators satisfy the following commutation relations

+

Pi , Pi = 2ii K,
(9.1)
J, Pi = ii Pi ,
with i = 1, . . . , n. It will be interesting to generalize our results to the higher-dimensional
analogues of the D2- and the S1-branes as well as to extend them to encompass other
classes of symmetric branes. In fact whenever two or more of the i parameters in (9.1) coincide, the higher-dimensional Heisenberg algebras have additional outer automorphisms
which permute the corresponding pairs of Pi generators. The existence of additional outer
automorphisms parallels the enhancement of the isometry group of these pp-wave backgrounds when some of the i parameters coincide [36]. As a consequence these models
486
display a richer set of symmetric branes, some of them similar to the oblique branes discussed in [54,56,75].
It should also be possible to study less symmetric branes which can be obtained by
performing a T -duality along the Cartan torus, following [100]. Also the supersymmetric
H2n+2 WZW models should be analyzed and brane configurations preserving some or none
of the bulk supersymmetries. An interesting brane is the H4 -brane in the H6 gravitational
wave [60,61], the Penrose limit of the AdS2 S 2 -brane in AdS3 S 3 . The dynamics of
the open strings ending on this brane should be described by a direct generalization of our
results for the D2-branes.
Acknowledgements
The authors are grateful to Costas Bachas and Volker Schomerus for several discussions.
G.D. would like to thank the Erwin Schrdinger International Institute in Vienna for the
warm hospitality provided to him during the program String theory in curved backgrounds
and BCFT and the workshop String theory on non-compact and time-dependent backgrounds, 718 June 2004. He is particularly grateful to the organizers of the program for the
brilliant atmosphere they created and for giving him the opportunity to present this work.
Most of the present work was done while G.D. was supported by an European Commission
Marie Curie Individual Postdoctoral Fellowship at the LPTHE, Paris, contract HMPF CT
2002-01908. G.D. also acknowledge support by the PPARC grant PPA/G/O/2002/00475
and by the EU network HPRN-CT-2000-00122. The work of E.K. was partially supported
by INTAS grant, 03-51-6346, RTN contracts MRTN-CT-2004-005104 and MRTN-CT2004-503369 and by a European Union Excellence Grant, MEXT-CT-2003-509661.
4 representations
Appendix A. H
The Heisenberg group H4 has three types of unitary representations.
+
(1) Lowest-weight representations Vp,
, where p > 0. They are constructed start
+
ing from a state |p, which satisfies P |p, = 0, K|p, = ip|p, and J |p, =
i |p, . The spectrum of J is given by { + n}, n N and the value of the Casimir is
C = 2p + p.
(2) Highest-weight representations Vp,

, where p > 0. They are constructed starting
from a state |p, which satisfies P |p, = 0, K|p, = ip|p, and J |p, =
i |p, . The spectrum of J is given by { n}, n N, and the value of the Casimir is
+
C = 2p + p. The representation Vp,
is the representation conjugate to Vp,
.
(3) Continuous representations Vs,0 with p = 0. These representations are characterized by K|s, = 0, J |s, = i |s, and P |s, = 0. The spectrum of J is then
given by { + n}, with n Z and || 12 . The value of the Casimir is C = s 2 . The onedimensional representation can be considered as a particular continuous representation,
where the charges s and are zero.
487
For the study of the H4 WZW model, three types of highest-weight representations of
based on Vp,
representhe affine H 4 algebra will be relevant. Affine representations Vp,
tations of the horizontal algebra, with conformal dimension

p
(A.1)
(1 p),
2
and affine representations Vs0 based on Vs0 representations, with conformal dimension
h = p +
s2
(A.2)
.
2
Highest-weight representations of the current algebra lead to a string spectrum free from
negative norm states only if they satisfy the constraint
h=
0 < p < 1.
(A.3)
States with larger values of p belong to new representations resulting from spectral flow of
the original highest-weight representations [31,38]. In fact the spectral-flowed representations are highest-weight representations of an isomorphic algebra whose modes are related
to the original ones by
+
Pn+ = Pnw
,
Pn = Pn+w
,
Jn = Jn ,
L n = Ln iwJn .
K n = Kn iwn,0 ,
(A.4)
An important piece of information for understanding the structure of the three-point couplings is provided by the decomposition of the tensor products of the H4 representations
Vp+ , Vp+ , =
1
Vp+ , Vp , =
1
Vp+ +p
Vp+ +p
p1 > p2 ,
Vp +p
p1 < p2 ,
2 ,1 +2 +n
n=0
2 ,1 +2 n
n=0
Vp+ , Vp , =
1
Vp,
Vp,
=
2 ,1 +2 +n
n=0

s ds Vs,01 +2 ,
Vp+ , Vs,02 =
1
Vp+ +p
1
n=
2 ,1 +2 +n
(A.5)
The fusion rules for the primary vertex operators of the H4 model can be obtained from
the previous tensor products. When the representations involved are spectral-flowed representations, one has to use the relation [101]
w1 [1 ] w2 [2 ] = w1 +w2 [1 2 ].
(A.6)
488
Appendix B. Fusing matrices

ij kl
Consider the correlator i (z1 )j (z2 )k (z3 )l (z4 ) and let Fp (z) denote the conforlij k
mal blocks in the s-channel z1 z2 and Fq (1 z) the conformal blocks in the u-channel
z12 z34
z1 z4 , where z = z13 z24 . We use the following convention for the fusing matrices

j k
ij kl
lij k
Fq (1 z).
Fp (z) =
(B.1)
Fpq
i l
q
Fpq defines a linear transformation

j k
p
q
i
: Vjp
Fpq
Vkl Vqli Vj k ,
i l
where Vjik is the space of the three-point couplings. Moreover,

j k
l k
Fqr
= r,s .
Fpq
i l
i j
(B.2)
(B.3)
Since in our non-compact CFT the conformal blocks are labeled either by discrete or continuous indexes, in the previous expressions we will have a sum or an integral, according
to the case. The following are thefusing matrices we used in Section 5 to compute the
structure constants. We set = 4i=1 i . For correlators of the form +++ we have

(p4 , 4 ) (p1 , 1 )
F(p1 +p2 ,1 +2 +n),(p1 p4 ,1 +4 +m)
(p3 , 3 ) (p2 , 2 )

(p2 + p3 )(p1 + p2 )
( n)!
=
m!( n m + 1)
(p2 )(p4 )
n
m

(p2 )(p4 )
(p2 )(p4 )
(p1 + p2 )(1 p1 )(p3 )

(p2 + p3 )(p3 )(1 p1 )
F (n, m, n m + 1, ),
(B.4)
where
=
sin p2 sin p4
.
sin p1 sin p3
(B.5)
For correlators of the form ++ we have

(p4 , 4 ) (p1 , 1 )
F(p1 p2 ,1 +2 n),(p1 p4 ,1 +4 m)
(p3 , 3 ) (p2 , 2 )

m
(p2 )(1 p4 )
(m + n + )! (p1 p2 )(p2 )(1 p4 ) n
=
m!(m + )!
(p1 )(1 p3 )
(p2 p3 )(p3 )(1 p1 )
+1

(p2 )(1 p4 )
F (n, m, n m , ),
(B.6)
(1 p1 + p2 )(p2 p3 )
where
=
sin p2 sin p4
.
sin p1 sin p3
(B.7)
489

(q, 4 ) (p, 1 )
(q, 3 ) (p, 2 )

n 2
(p)
s 2
(p)(1 q) +1
1
=
(n + )!
(p q)
(q) (p q)
2
2

s2
s
e 2 ((p)+(1q)2(1)) Ln
(ctgq ctgp) .
2
F(s,{1 +2 }),(pq,1 +4 n)
(B.8)

(p, 2 ) (p, 1 )
F(pq,1 +4 n),(s,{3 +4 })
(q, 3 ) (q, 4 )
+1

(q) (p q) n s 2 2 s 2 ((q)+(1p)2(1))
(q)(1 p)
e2
= n!
(1 p + q)
(p)
2
2

s
(ctgq ctgp) .
Ln
(B.9)
2

(p, 1 )
(p, 2 )

t 2 s 2
st
e 2 ((p)+(1p)2(1)) J
.
=
sin p
sin p
F(s,{1 +2 }),(t,{1 +4 })
(p, 4 )
(p, 3 )
Similar expressions hold for correlators of the form ++.

(p4 , 4 ) (p1 , 1 )
F(p1 +p2 ,1 +2 +n),(p1 p4 ,1 +4 m)
(p3 , 3 ) (p2 , 2 )

n
(p1 )(p3 )
(m + n + )!
=
m!(m + )! (p1 + p2 )(1 p2 )(1 p4 )
m
+1

(p1 )(p3 )
(p1 )(p3 )
(p3 p2 )(p2 )(p4 )

(p1 + p2 )(p3 p2 )
F (n + , m, n m, ),
(B.10)
(B.11)
where
=
sin p1 sin p3
.
sin p2 sin p4
(B.12)

F(s,{1 +2 }),(p+q,1 +4 +n)
(p, 4 )
(q, 3 )
+1
(p, 1 )
(q, 2 )
n 2 2

(p)
s 2 s ((p)+(q)2(1))
(p)(q)
e2
= (n )!
(p + q)
(q) (p + q)
2
2

s
(ctgp + ctgq) .
Ln
(B.13)
2
490

(q, 2 )
(p, 1 )
(q, 3 ) (p, 4 )

1 (1 q)(1 p) +1 (p + q) n s 2 2 s 2 ((1q)+(1p)2(1))
=
e 2
n!
(1 p q)
(p) (q)
2
2

s
(ctgp + ctgq) .
Ln
(B.14)
2
F(p+q,1 +2 +n),(s,{3 +4 })
For correlators of the form +0 0 we have

s s
is s sin
(s4 , 4 ) (p, 1 )
3 4 cos (p)+ 2 3tan4p in+i
F(s,{1 +2 }),(p,1 +4 +n)
=e 2
,
(s3 , 3 ) (p, 2 )
(B.15)

F(p,1 +4 +n)(s,{1 +2 })
=
1
s3 s4 sin
s3 s4
2
(p, 2 )
(s3 , 3 )
cos (p)
(p, 1 )
(s4 , 4 )
is3 s4 sin
2 tan p +ini
(B.16)
where (p) = (p) + (1 p) 2(1), s 2 = s32 + s42 + 2s3 s4 cos and ei =
s3 +s4 ei
.
s
Appendix C. Bases of conformal blocks

In this appendix we collect various bases of conformal blocks for correlators of the
form ++, ++ and +0 0. Using the global conformal and H4 symmetries,
the four-point amplitudes can be written as follows
4

A4 (zi , xi ; z i , xi ) =
2h
xi )
|zij |2hi +2hj 3 K(xi )K(
Fn (z)Fn (z).
(C.1)
j >i=1
The kinematical functions K and K are completely fixed by the Ward identities of the left
and right H 4 algebras and we chose the standard gauge for the global conformal transz34
formations. The conformal blocks Fn (z, x) thus depend only on the cross-ratio z = zz12
4 13 z24
and a suitable combination x of the four charge variables. In the following = i=1 i .
Consider first the correlator
+

p , (z1 , z 1 , x1 , x1 )p+ , (z2 , z 2 , x2 , x2 )p+ , (z3 , z 3 , x3 , x3 )p , (z4 , z 4 , x4 , x4 ) .
1
(C.2)
The H4 Ward identities require
p1 + p 2 + p 3 = p4 ,
(C.3)
and give the function K

K(xi ) = ex4 (p1 x1 +p2 x2 +p3 x3 ) (x3 x1 ) ,
(C.4)
491
as well as the invariant combination

x=
x 2 x1
.
x3 x1
(C.5)
We set F = z12 (1 z)14 F (z, x) where

h
p1 p2 2 p1 1 p2 ,
3
h
14 = h1 + h4 + p1 p4 4 p1 + 1 p4 p1 + (p2 + p3 ).
3
We then arrive at the following form for the KZ equation
12 = h1 + h2
z F (z, x) =

1
p1 x + p2 x(1 x)x p2 x F (z, x)
z

1
(1 x)(p2 x + p3 )x p2 (1 x) F (z, x).

1z
(C.6)
(C.7)
The correlator vanishes when < 0. In the s-channel we have the p+ +p , + +n repre1
2 1
2
sentations with 0 n . The conformal blocks are

n
Fn (z, x) = f n (z, x) g(z, x)
(C.8)
,
where
p3
z1p1 p2 0 (z) xzp1 p2 1 (z),
1 p1 p 2
xp2
1 (z),
g(z, x) = 0 (z)
p1 + p 2
f (z, x) =
(C.9)
and
0 (z) = F (1 p1 , 1 + p3 , 2 p1 p2 , z),
1 (z) = F (1 p1 , p3 , 1 p1 p2 , z),
0 (z) = F (p2 , p4 , p1 + p2 , z),
1 (z) = F (1 + p2 , p4 , 1 + p1 + p2 , z).
In the u-channel we have the representations
formal blocks read

m
Fm (u, x) = fm (u, x) g(u,
x) ,
(C.10)
p p , + m ,
4
1 1
2
0 m , and their con(C.11)
where

f(u, x) = up2 p3 0 (u) x 1 (u) ,
xp2
p3
0 (u) +
1 (u),
g(z,
x) =
p 2 + p3
p 2 + p3
and
(C.12)
492
0 (u) = F (p1 , p3 , 1 p2 p3 , u),

1 (u) = F (1 p3 , p1 , 1 p2 p3 , u),
0 (u) = F (p2 , p4 , 1 + p2 + p3 , u),
1 (u) = F (1 + p2 , p4 , 1 + p2 + p3 , u).
(C.13)
Consider now the correlator

+

p , (z1 , z 1 , x1 , x1 )p , (z2 , z 2 , x2 , x2 )p+ , (z3 , z 3 , x3 , x3 )p , (z4 , z 4 , x4 , x4 ) .
1
(C.14)
The H4 Ward identities require
p 1 + p 3 = p 2 + p4 ,
(C.15)
and give the function K

K(xi ) = ep2 x1 x2 p3 x3 x4 (p1 p2 )x1 x4 (x1 x3 ) .
(C.16)
In this case the invariant combination is

x = (x1 x3 )(x2 x4 ).
We set F
= z12 (1 z)14 F (z, x)
(C.17)
where
h
+ p1 p2 2 p1 + 1 p2 p2 ,
3
h
14 = h1 + h4 + p1 p4 4 p1 + 1 p4 p4 .
3

z(1 z)z F (z, x) = xx2 + (p1 p2 )x + 1 + x F (z, x)

+ z (p1 + p3 )xx + xp2 p3 (1 + )p3 F (z, x).
12 = h1 + h2
(C.18)
(C.19)
When p1 > p2 in the s-channel we have p+ p , + n with n 0 for 0 and n =

1
2 1
2
m with m 0 for 0. The conformal blocks are
Fn (z, x) = n

exg1 (z)
L x (z) (z)n ,
(f1 (z))1+ n
n N,
(C.20)
where Ln is the nth generalized Laguerre polynomial,

f2 (z)
,
(z) = z(1 z) ln ,
f1 (z)
g1 (z) = zp3 z(1 z) ln f1 ,
(z) =
n =
n!
,
(p1 p2 )n
(C.21)
and
f1 (z) = F (p3 , 1 p1 , 1 p1 + p2 , z),
f2 (z) = zp1 p2 F (p4 , 1 p2 , 1 p2 + p1 , z).
(C.22)
493
When p1 < p2 , the intermediate states belong to the p p , + +n representation with

2
1 1
2
n = m + with m 0 for 0 and n 0 for 0. The conformal blocks are very
similar. Finally when p1 = p2 = p and p3 = p4 = q the intermediate representation is
0 and the conformal blocks read
s,
Fs (z, x) =

s2
exg1 (z)
2 J s 2x ,
2 (z) xz(1 z)
e
(c1 (z))1+
(C.23)
where
c2 (z)
,
c1 (z)
(z) =
= z(1 z)(z),
(C.24)
and
c1 (z) = F (q, 1 p, 1, z),

c2 (z) = ln z + 2(1) (q) (1 p) c1 (z)

(q)n (1 p)n
(q + n) + (1 p + n) 2(n + 1) zn ,
+
n!2
(C.25)
n=0
where
(a)n
(a + n)
.
(a)
(C.26)
Moreover
g1 (z) = qz z(1 z)z ln c1 .
(C.27)
In the u-channel when p1 > p4 we have the representations p+ p , + n with n N

1
4 1
4
for 0 and n = m , m 0 for 0. The conformal blocks are
Fn (u, x) = n

exg1 (u)
L x (u) (u)n ,
(f1 (u))1+ n
n N,
(C.28)
where
f2 (u)
,
(u) = u(1 u) ln ,
f1 (u)
g1 (u) = (1 u)p3 + u(1 u) ln f1 ,
(u) =
n =
n!
,
(p2 p3 )n
(C.29)
and
f1 (u) = F (p3 , 1 p1 , 1 p2 + p3 , u),
f2 (u) = up2 p3 F (p2 , 1 p4 , 1 + p2 p3 , u).
(C.30)
When p1 < p4 , the intermediate states belong to the p p , + +n representation with

4
1 1
4
n = m + with m 0 for 0 and n 0 for 0. The conformal blocks are very
similar. Finally, when p1 = p4 = p and p2 = p3 = q the intermediate representation is
494
0 and the conformal blocks read

s,
Fs (u, x) =

s2
exg1 (u)
e 2 (u) xu(1 u) 2 J s 2x ,
1+
(c1 (u))
(C.31)
where
(u) =
c2 (u)
,
c1 (u)
= u(1 u)(u),
(C.32)
and
c1 (u) = F (q, 1 p, 1, u),

c2 (u) = ln u + 2(1) (q) (1 p) c1 (u)
+

(q)n (1 p)n
(q + n) + (1 p + n) 2(n + 1) un .
2
n!
(C.33)
n=0
Moreover
g1 (u) = (1 q)u + u(1 u)u ln c1 .
(C.34)
We will also need correlators of the form

+

p , (z1 , z 1 , x1 , x1 )p+ , (z2 , z 2 , x2 , x2 )p , (z3 , z 3 , x3 , x3 )p , (z4 , z 4 , x4 , x4 ) .
1
(C.35)
In this case the H4 symmetry requires
p 1 + p 2 = p 3 + p4 ,
(C.36)
and gives
K(xi ) = ep3 x1 x3 p2 x2 x4 (p1 p3 )x1 x4 (x1 x2 ) ,
(C.37)
as well as x = (x1 x2 )(x3 x4 ). Proceeding as before we pass to the conformal blocks

and we set F = z12 (1 z)14 F (z, x) where
12 = 12 14 (1 + )p2
h
= h1 + h2 + (1 p1 )(p1 + p2 ) + p1 (3 + 4 ) 1 (p1 + p2 )
3
(1 + )p2 ,
h
14 = 14 = h1 + h4 + p1 p4 4 p1 + 1 p4 p4 .
3
z(1 z)z Fn (z, x)

= z xx2 + (p1 p3 )x + 1 + x (1 + )p2 Fn (z, x)

+ (p1 + p2 )xx + xp2 p3 Fn (z, x).
(C.38)
(C.39)
495
In the s-channel, the representations p+ +p , + +n with n N when 0 and n N

1
2 1
2
when 0. In the first case with m = n the conformal blocks are
Fm (z, x) = m

exg1 (z)
Lm x (z) (z)m ,
1+
(f1 (z))
m N,
(C.40)
where
f2 (z)
,
(z) = (1 z) ln ,
f1 (z)
g1 (z) = p2 (1 z) ln f1 ,
(z) =
m =
m!
,
(1 p1 p2 )m
(C.41)
and
f1 (z) = F (p2 , p4 , p1 + p2 , z),
f2 (z) = z1p1 p2 F (1 p1 , 1 p3 , 2 p1 p2 , z).
(C.42)
When 0 the conformal blocks are given by the same expression except that now n 0.
Using
Ln (x) =
n!
(x) L
n (x),
(n )!
(C.43)
and the Wronskian

W (f1 , f2 ) = (1 c)zc (1 z)cab1 ,
(C.44)
they can be rewritten as

Fn (z, x) = z(p1 +p2 ) (1 z)(p1 +p4 ) x n

exg1 (z)
L|| x (z) (z)n ,
(f1 (z))1+|| n
n N.
(C.45)
In the u-channel, when p1 > p4 , we have the representations

mal blocks are
Fn (z, x) = n

exb1 (z)
L x (z) (z)n ,
(a1 (z))1+ n
p+ p , + n .
1
4 1
4
n N,
The confor-
(C.46)
where
a2 (z)
,
(z) = u ln ,
a1 (z)
b1 (z) = p2 + u ln a1 ,
(z) =
n =
n!
,
(p3 p2 )n
(C.47)
and
a1 (z) = F (p2 , p4 , 1 + p2 p3 , u),
a2 (z) = up3 p2 F (p3 , p1 , 1 p2 + p3 , u).
(C.48)
Here n 0 when 0 and n = m with m 0 when 0. When p1 < p4 we have the

representations p p , + +n with n 0 when 0 and n = m + with m 0 when
4
496
0. The conformal blocks are similar to the ones already displayed. Finally when p1 =
p4 = p and p2 = p3 = q the intermediate states belong to the continuous representations
0
s,{
. Let us now turn to correlators of the form
+ }
1

+
0
0
p,
(z , z , x , x )p,
(z , z , x , x )s,
3 (z3 , z 3 , x3 , x 3 )t,4 (z4 , z 4 , x4 , x 4 ) .
1 1 1 1
2 2 2 2
1
(C.49)
In this case
K(xi ) = e
px1 x2 1 ( xs + xt ) 2 (sx3 +tx4 ) n1 n2

2 3
2
4
x3
,
(C.50)
+
in the
and x = xx43 . The conformal blocks corresponding to the propagation of (p,
1 +4 +n)
14
12
u-channel are F = u (1 u) F (u, x) where
12 =
s2 + t 2 h
,
2
3
14 = h1
h
p 4 .
3
(C.51)
They solve the following KZ equation

1
1
stx
1 st
Fn (u, x)
x+
Fn (u, x).
u Fn (u, x) = pxx +
u
2
1u 2
x
(C.52)
Their explicit form is

Fn (u, x) = x n unp e
ub(u)
st2 ( xa(u)
p + x(1p) )
(C.53)
where
a(u) = F (p, 1, 1 + p, u),
b(u) = F (1 p, 1, 2 p, u).
(C.54)
Similarly the blocks pertaining to p,

are given by F+m . In the s-channel the
+ m
blocks for the representation r0 with

r 2 = s 2 + t 2 + 2st cos ,
ei =
s + tei
,
r
[0, 2),
(C.55)
ein Fn (u, x).
(C.56)
are
st
Fr (z, x) = e 2 [cos (p)i sin cot p]+i

nZ
Appendix D. Sewing constraints

In this appendix we outline with an example the main steps that are necessary in order
to verify that the structure constants given in Section 5 solve the sewing constraints. We
consider the bulk-boundary couplings for the S1-branes and study the factorization of the
following bulk two-point functions: p+ q+ , p+ q and p+ s0 . The first correlator
gives
aaa,1
B sp,1 a B sq,2 Css
=
(p+q, + +n) a
1
2
C(p, );(q,
)
1
n=0
497
B 1p+q,1 +2 +n
F(p+q,1 +2 +n),s

(p, 1 )
.
(q, 2 )
(p, 1 )
(q, 2 )
(D.1)
The second correlator gives

a
aaa,1
B sp,1 a B sq,2 Css
=
(pq, + n) a
1
2
C(p, );(q,
)
1
n=0
B 1pq,1 +2 n
F(pq,1 +2 n),s

(p, 1 ) (p, 1 )
,
(q, 2 ) (q, 2 )
(D.2)
when p > q and

a
aaa,1
B sp,1 a B sp,2 Css
(t,{ + })
a 1
2
dt tC(p,1);(p,
B t,{1 +2 }
)
Ft,(s,{1 +2 })
(p, 1 )
(p, 2 )

(p, 1 )
,
(p, 2 )
(D.3)

(p, 1 )
.
(s2 , 2 )
(D.4)
when p = q. Finally the third correlator gives

a
aaa,1
B sp,1 a B ss2 ,2 Css
=
(p, + +n)
C(p,1);(s2 , ) a B 1p,1 +2 +n
1
n=0
F(p,1 +2 +n),s
(p, 1 )
(s2 , 2 )
Using the bulk three-point couplings in (5.12)(5.15) and the fusing matrices in Appendix B
the previous equations become

s2 a 1
Ln p,q
B p+q,1 +2 +n ,
2
n=0

s2
s2 a 1
= p,q e 2 ((q)+(1p)2(1))
Ln p,q
B pq,1 +2 n ,
2
B sp,1 a B sq,2 =
B sp,1 a B sq,2
s2
p,q e 2 ((p)+(q)2(1))
n=0
B sp,1 a B sp,2
s2
=
e 2 ((p)+(1p)2(1))
sin p

dt tJ0
st
sin p
B 1t,1 +2 ,
0
a
B sp,2 a B ss2 , =
1
is12 sin
2 tan(p)
s12 (1cos )

e
((p)+(1p)2(1))
2
e
ein a B 1p,1 +2 +n ,
s12 sin
nZ
where p,q = cot p + cot q and
s22
= 2s12 (1 cos ).
(D.5)
498
We make the following ansatz

s2
a s
s
e2ip+ 4 [(p)+(1p)2(1)] bp,
B p, =
(u)
sin p
(D.6)
s
s
(u) = einu bp,
(u). The constraints then simplify
with bp,
+n
s
s
bp,
1 (u)bq,2 (u) = e
i[cot(p)cot(q)]s
u

s
s
bp,
1 (u)bp,2 (u) =
dt tJ0
st
sin p
1
bpq,
(u)
+
1
4 tan( 2 )
1 eiu
B 1t,1 +2 ,
0
is12 sin
s2
(u)a B ss2 ,
bp,
e 2 tan(p) in 1
=
e bp,1 +2 +n (u),
s12 sin nZ
s
(u) = e
bp,
and are solved by

Therefore

a s
B p, =
i u
is 2
u
4 tan(p) tan( 2 )
(D.7)
is
u
e2ip+i u s4 [(p)+(1p)2(1)] 4 tan(p)
tan( 2 )
e
,
sin p 1 eiu
ei u
2(u ), t 2 = 2s 2 (1 cos ).
(D.8)
s 2 sin
Here we show some explicit examples of the relation (5.2.10) for the S1-branes of the
H4 model. We have

(p, 1 )
(q, 2 )
abc,(p+q, )
,
C(p, )(q, )3 = F(pb ,2 c +n2 ),(p+q,1 +2 +k)
1
2
(pa , a ) (pc , c )
a
B ts, =
where 1 = ab n1 , 2 = bc n2 , 3 = ac n3 and k = n1 + n2 n3 0. Similarly

(p, 1 )
(q, 2 )
abc,(pq,3 )
C(p, )(q, ) = F(pb ,2 c n2 ),(pq,1 +2 k)
, p > q,
1
2
(pa , a ) (pc , c )
where 1 = ab n1 , 2 = bc + n2 , 3 = ac n3 and k = n3 + n2 n1 0. We also have

(p, 1 )
(q, 2 )
abc,((qp),3 )
C(p, )(q, ) = F(pb ,2 c n2 ),(pq,1 +2 +k)
,
1
2
(pa , a ) (pc , c )
p < q,
where 1 = ab n1 , 2 = bc + n2 , 3 = ac + n3 and k = n3 + n1 n2 0.
Appendix E. Penrose limit of the SU(2)- and SL(2, R)-branes
In this appendix we discuss the Penrose limit of the symmetric branes in S 3 and in
AdS3 using coordinate systems adapted to their world-volume. For S 3 we use spherical
499
coordinates

ds 2 = k dt 2 + d 2 + sin2 d 2 + sin2 d 2 ,
H = 2k sin2 sin .
(E.1)
The symmetric branes sit at n = n/k. The integer n, 0 < n < k, parameterizes a uniform
world-volume flux F = n/2 sin , which stabilizes the brane [73]. When n = 0 or n = k,
the brane world-volume degenerates to a point. In order to describe the S1-branes we make
the following change of variables
t=
x + x
+
,
2
k
x + x
,
2
k
= .
k
(E.2)
This leads in the limit k to the NappiWitten wave in Rosen coordinates. We can
easily see that the flux on the brane world-volume becomes
1
F B + 2F = sin x + d d,
(E.3)
2
as expected. Moreover we can exploit the relation between the brane location and the
spin of the SU(2) representations
j =
(2j + 1)
,
k+2
k
j = 0, . . . , ,
2
(E.4)
to derive an analogous relation between the labels of the H4 conjugacy classes (u, ) and
the quantum numbers of the H4 representations. If we scale the spin of SU(2) in such a
way as to obtain a discrete representation Vp, [35]
k
j = p ,
2
we obtain
p > 0,
u = 2(p + n),
(E.5)
2 = (2 2p 1).
(E.6)
For the D2 we set

t=
u
,
2
u 2v
,
2
k
= + ,
2
k
= + ,
2
k
(E.7)
and take the limit k , which leads to the NappiWitten wave in Brinkman coordinates.
In this case we focus on S 2 -branes very close to the equator of S 3 and scale the SU(2) spin
as
k
j= + k .
(E.8)
4
2
As expected, the twisted-branes are in one-to-one correspondence with the representations
0 and V 0
invariant under the action of the external automorphism , Vs,0
s,1/2 . Note that in the
first case, the null geodesic used to take the limit intersects the brane world-volume while
in the second case it is contained within the brane world-volume.
500
The limit of the AdS2 -branes is better described using the following coordinate system
for AdS3 S 1

ds 2 = k d 2 + k cosh2 d2 cosh2 d 2 + k dx 2 ,
H = 2k cosh2 cosh .
(E.9)
The AdS2 -branes are surfaces with constant and a world-volume flux F = k
2 cosh .
The Penrose limit is
=
u 2v
+
,
2
k
x=
u
,
2
= ,
k
= ,
k
(E.10)
with k . In the process, the AdS2 -brane at constant (with Neumann boundary
condition along S 1 ) becomes the D2-brane at constant , with a null world-volume flux
Fu =
2 , as expected. Similarly, the limit of the H2 -branes is more easily described if we
use hyperbolic coordinates for AdS3 writing

ds 2 = k d 2 + k sin2 d2 + sinh2 d 2 + k dx 2 ,
H = 2k sin2 sinh ,
(E.11)
with [, ], 0. The H2 -branes are surfaces with constant and a world-volume
k
flux F = 2
sinh . In the limit k the change of coordinates
=
x +
,
2
x=
x + 2x
,
2
k
= ,
k
= ,
(E.12)
leads to the NappiWitten wave in Rosen coordinates (here 2 = y12 + y22 ). The H2 -branes
with Dirichlet boundary conditions along S 1 become the S1-branes with the flux given in
(2.13).
Appendix F. The DBI approach

In this appendix we will study the more general class of rotationally invariant solutions
found in Section 8. We consider B = 0. It is convenient to distinguish the following cases:
(I) |B + E| < 2. In this case the brane embedding can be written as

|B|
4AB
2
u=
(F.1)
log r + r +
+ u0 ,
4 (B + E)2
4 (B + E)2

4 BE E 2
4AB
2
v = v0 + 2A
log r + r +
(4 (B + E)2 )3/2
4 (B + E)2

B +E
4AB

r r2 +
.
4 (B + E)2
2 4 (B + E)2
(F.2)
(II) |B + E| = 2. Here the embedding simplifies to

B
u=
r + u0 ,
4A
501
(F.3)
B 3Ar r 3
.
A
3B
(III) |B + E| > 2. In this cases we obtain a trigonometric embedding

r (B + E)2 4
|B|
u = u0 +
arcsin
,
4AB
(B + E)2 4
v = v0 +
|B|
(F.4)
(F.5)
4AB
r r 2 +
v = v0
(B + E)2 4
2 (B + E)2 4

r (B + E)2 4
2A(E 2 + BE 4)
+
arcsin
.
((B + E)2 4)3/2

4AB
(F.6)
Solving for r and substituting we obtain that the points on the brane are on the curve

E 2 + BE 4
(u u0 )
(B + E)2 4 (v v0 ) 2A
(B + E)2 4

2(u u0 )
1
|AB 3 | sin (B + E)2 4
=
.
2
|B|
(F.7)
Using the embedding equations (8.1.9) and (8.1.10) we can calculate the induced metric
as

2
4 E2 2 2 2
2
2
2 4E
2
2
r u dr + r d = r
du + d
d s =
B2
B2
2
(F.8)
while the antisymmetric tensor is Br = 2ur. The induced two-dimensional curvature is

R u + ru .
(F.9)
The induced metric is flat when A = 0, when the solution is

u=
|B|
4 (B
+ E)2
log r + u0 ,
B +E
v = v0
r 2.
2
2 4 (B + E)
(F.10)
Our symmetric branes are a special case of the flat branes with B = 0.
The open string metric is the induced metric rescaled by (det g + B)/ det g. We find

4B 2 u2 + (4 E 2 )r 2 u 2 4 E 2 2 2 2
2
2
2
dsopen
=
r
u
dr
+
r
d
(4 E 2 )r 2 u 2
B2

4 E2 2
4B 2 u2
2
2
du + d .
= r +
(F.11)
(4 E 2 )u 2
B2
502
For the symmetric branes this is again flat.

The critical electric field case E = 2 is a bit special and we will discuss it here, separately. The gauge field equation implies in this case
r 2 u 2 + 2u v = 1
(F.12)
while the others

2r 2 u = B|2ur + F |,

A
r 2 u + v = 1 + 2 |2ur + F |.
r
The previous equations can be massaged into

1
B
B
4

(u u0 ) ,
r = R sin
u =
4A 1 B4 r 2
B
(F.13)
(F.14)
4A

R
r2
r
B 2
r 1 2 R arcsin
v = v0 +
.
2
R
R
B(B 4)
(F.15)
This
class of solutions describes an embedding with a compact support 0 < r < R with
R=
4A
B4 .
The induced metric here is degenerate
d s 2 = r 2 d 2 .
(F.16)
These are the null branes mentioned (but not analyzed in detail) in the main body of this
paper.
F.1. S1 fluctuations
Expanding around the classical solution u , v , F satisfying (8.1.8)(8.1.10)
u u + u,
v v + v,
F F + F
we obtain to quadratic order

L = L + L2 + O u3 , v 3 , F 3 ,
L2 =

1
1 2u v r 2 u2 r 2 (2ur + F )2
3
2L

+ 2r 2 (2u r + F ) v + r 2 u u + u v (2ur + F )

L2 u2 v 2 + r 2 + v 2 u 2 + 2 1 u v u v

r 4 u2 v 2 r 4 r 2 + v 2 + (2u r + F )2 u 2

2r 2 (2u r + F )2 + r 2 1 u v u v .
From Eqs. (8.1.8)(8.1.10)
(F.17)
(F.18)
(F.19)
L =
2 2
r u ,
|B|
(2u r + F )2 =
1 2u v r 2 u2 =
503
4 E2 2 2
r u ,
B2
E2 2 2
r u
B2
(F.20)
follow and we can rewrite L2 as

E[(2A Er2 )u + Bv ] 2
(4 E 2 ) 4AB + (4 (B + E)2 )r 2
F + 2ur +
L2 =
16r 2
4 E2
4(A2 AEr2 + r 4 )u 2 + B 2 v 2 + 2(2AB [(E(B + E)) 4]r 2 )u v

4r 2 4AB + (4 (B + E)2 )r 2

4AB + 4 (B + E)2 r 2
4(A2 AEr2 + r 4 )u 2 + B 2 v 2 + 2(2AB [(E(B + E)) 4]r 2 )u v

.
4(4 E 2 )r 2
(F.21)
Specializing to the symmetric solutions A = B = 0

u = constant,
Er2
v v0
,
2 4 E2
Er
2u r + F =
4 E2
2r
L =
,
4 E2
(F.22)
we obtain

4 E2
16r 2 2
4 E 2 (2ur + F )2 + 2E 2 r 2 u (2ur + F ) +
L2 =
u + 8u v
16r
4 E2

+ 4 + E 2 r 4 u 2 + 8r 2 u v .
(F.23)
We redefine
V =v+
2r 2
u,
4 E2
A A +
E2r 2u
4 E2
(F.24)
and rewrite the action (after performing an integration by parts) as
2

4 E2
8ur
r 2 u2
2
2
.
4E
L2 =
F+
+ 8 u V + r u V + 32
16r
4 E2
4 E2
(F.25)
It is obvious from the Lagrangian above that u satisfies the flat two-dimensional Laplace
equation
2u = 0
1
1
(ru ) + 2 u = 0.
r
r
(F.26)
504
The solution to the equation for the gauge fluctuation is

F + 2ur
E2 2
2C
r
r u =
4 E2
(4 E 2 )
(F.27)
with C a constant.
Finally the u equation reads
1
1
4
2V = (rV ) + 2 V =
(2u + C).
r
r
4 E2
(F.28)
Thus the V field is a free field subject to a source linear in u and C.

The regular solution of (F.26) is u = u0 constant. On the other hand, in order that (F.28)
has a regular solution we must have C = 2u0 and then V =
V0 constant.
In
the
critical
case,
we
take
E
2
and
rescale
F
F
/
4 E 2 , (u, v) (u, v)
2
4 E to obtain

2
4AB + (4 B 2 )r 2
S2 (E = 2) =
F + 2ur + 2 2 A r 2 u + Bv
2
16r

2
4AB + (4 B 2 )r 2
2 A r 2 u + Bv .
(F.29)
2
4r
This system is degenerate. The solution to its equation of motion is
F + 2ur =
4Cr 2
4AB + (4 B 2 )r 2

10Cr 2
2 A r 2 u + Bv =
.
4AB + (4 B 2 )r 2
(F.30)
Appendix G. Bulk one-point couplings from the DBI action

We can compute the coupling to the bulk metric from
S
SDBI
,
G
(G.1)

1
+ B + F ) (G
+ B + F ) + (G
B F ) x x .
=
det(G
4
(G.2)
For the D2-branes at y = y0 we obtain the following coupling

1
S uv =
,
2
2 1 fuv

2
1 fuv
xx
S =
2
S vv =
2 + x 2 + 2y f
fux
0 ux

,
2
2 1 fuv
S vx =
fuv (fux y0 )

,
2
2 1 fuv
(G.3)
all others being zero. In summary,
0
1
S = 2 1fuv2
0
0
At the symmetric point
0 1
1 1 x 2
S=
0
0
2
0
0
For the D1-branes
2
2 1fuv
2 +x 2 +2y f
fux
0 ux
2
1fuv
fuv (fux y0 )
2
2 1fuv
0
0
0
1
0
0
fuv (fux y0 )
2
2 1fuv
2
1fuv
2
505
(G.4)
0
0
.
0
0
(G.5)
B 2 4 E2
u x x .
r r x r x +
4u
4B
For the symmetric configurations we obtain
SD1 =
(G.6)

r
E2r 3
E
vv
vr
rr
=
,
SD1
= r 2 ,
SD1
= 4 E2 ,
SD1
4
4
4 4 E2
2
4E
.
SD1
=
4r
The coupling to the antisymmetric tensor is given by
(G.7)
SDBI
,
B
A =

1
B F ) x x .
+ B + F ) (G
+ B + F ) (G
det(G
4
(G.8)
(G.9)
By direct calculation for the D1 case we obtain that the only non-zero components are
E
E
E
Av = v ,
Ar = .
u,
4
4
4
For the particular case of symmetric D1-branes we have
Au =
E2r
E
Av =
,
Ar = .
2
4
4 4E
For the D2-branes we obtain
fuv 2
0
0
0
1fuv
0 0
fuv
fux y0
0
0
1
1
0 0
2
2

1fuv
1fuv
A=
limf 0 2 0 y0
fux y0
2
0
0
0
2
1fuv
0 0
0
0
0
0
(G.10)
Au = 0,
(G.11)
0
y0
0
0
0
0
.
0
0
(G.12)
506
The one-point coupling to the dilaton is given by

SDBI
+ B + F ).
F
= det(G
We obtain
FD1 =
2r 2
2r 2 u
=
B
(4 (B + E)2 )r 2 + 4AB

2
FD2 = 1 fuv
symmetric
(G.13)
symmetric
2r
4 E2
1.
(G.14)
(G.15)
Appendix H. Some useful series and integrals
xz
Ln (x)zn
n=0

n=0
e z1
=
.
(1 z)1+
(H.1)

zn
Ln (x) = ez (xz) 2 J 2 xz .
(n + + 1)

2
dx ex x +1 Ln x 2 J (xy)
= 21 n1 ( )n y e

(H.2)
dx ex x Ln (x)Lm (x) = n,m
y4
Ln

y 2
.
4( )
( + n + 1)
,
n!
(H.3)
(H.4)

dx xJm (sx)Jm (tx) =
(s t)
,
s
(H.5)
dx e x x Ln (x)Lm (x)
(m + n + + 1) ( )n ( )m
F (m, n, m n , ),
m!n!
n+m++1
(H.6)
( )
.
( )( )
(H.7)
where
=
2 2
2
x Ln x J (xy)
dx x +1 ex L
m
= (1)
m+n
507
1 y4
(2)
y e
m+n
Lm
2

y2
+mn y
Ln
.
4
4
(H.8)
2
2 /2
x +1 ex Ln x 2 J (xy) dx
0
y
y
e 4
=
(n + 1 + /2)
n!
(2)+1

n
(1)l (n l + 1/2)(l + 1/2)
2 2l
y 2
L2l
. (H.9)
(l + 1 + /2)(n l)!
2(2 )
l=0

n=0
n!
2 xyz
Ln (x)Ln (y)zn (xyz) 2 z(x+y)
=
e 1z I
.
(n + + 1)
1z
1z
j l
j l
c 1 Ll
l+nj
(b1 c1 )c2
l+nj
Lj
(H.10)
(b2 c2 )
j =0

= ec1 b2 (c1 + c2 )n Lnl (b1 + b2 )(c1 + c2 ) .
dr re
a r2
r2
2
qsm+n
qm
Lm
(H.11)

2
ar 2
q+nms (a b)r
ns br
Ls
Lm+kn
2
2
2
(1)s+nk

(k + q)!(k + m)! (a b)nkm bs

a
=
F k, s, q k,
m!s!k!(k + m n)!
b
a k+q+1

a
.
F k, m k + n, m k,
ab

du e
u2 +
iu(q1 q2 )
(H.12)

i(q1 + q2 ) nm nm 2 (q1 + q2 )2
u+
u +
Lm
2
2
(q1 q2 )2
e 4
= i n+m m+n1 Hn (q1 )Hm (q2 ).
2 2 m!

0
(H.13)
1 b
ei sin a
J (ax)eibx dx =
,
a 2 b2
a > b.
(H.14)
508

e
x 2
1
J (x) dx =
2
2
8
e
,
I/2

0
Re() > 0, > 0, Re() > 1.

(H.15)
1
ds cos(bs)J0 (as) =
,
2
a b2
x dx eax J0 (xy) =
2
1 y2
e 4a .
2a
a > b.
(H.16)
(H.17)
References
[1] R. Gven, Plane waves in effective field theories of superstrings, Phys. Lett. B 191 (1987) 275.
[2] D. Amati, C. Klimcik, Strings in A shock wave background and generation of curved geometry from flat
space string theory, Phys. Lett. B 210 (1988) 92;
D. Amati, C. Klimcik, Nonperturbative computation of the Weyl anomaly for a class of nontrivial backgrounds, Phys. Lett. B 219 (1989) 443.
[3] H.J. de Vega, N. Sanchez, Particle scattering at the Planck scale and the AichelburgSexl geometry, Nucl.
Phys. B 317 (1989) 731.
[4] G.T. Horowitz, A.R. Steif, Spacetime singularities in string theory, Phys. Rev. Lett. 64 (1990) 260;
G.T. Horowitz, A.R. Steif, Strings in strong gravitational fields, Phys. Rev. D 42 (1990) 1950.
[5] A.R. Steif, Nonperturbative time dependent classical string solutions for the closed bosonic string, Phys.
Rev. D 42 (1990) 2150.
[6] A.A. Tseytlin, A class of finite two-dimensional sigma models and string vacua, Phys. Lett. B 288 (1992)
279, hep-th/9205058;
A.A. Tseytlin, String vacuum backgrounds with covariantly constant null Killing vector and 2-d quantum
gravity, Nucl. Phys. B 390 (1993) 153, hep-th/9209023;
A.A. Tseytlin, Finite sigma models and exact string solutions with Minkowski signature metric, Phys. Rev.
D 47 (1993) 3421, hep-th/9211061.
[7] C.R. Nappi, E. Witten, A WZW model based on a nonsemisimple group, Phys. Rev. Lett. 71 (1993) 3751.
[8] R. Gueven, Plane wave limits and T-duality, Phys. Lett. B 482 (2000) 255, hep-th/0005061.
[9] M. Blau, J. Figueroa-OFarrill, C. Hull, G. Papadopoulos, Penrose limits and maximal supersymmetry,
Class. Quantum Grav. 19 (2002) L87, hep-th/0201081;
M. Blau, J. Figueroa-OFarrill, C. Hull, G. Papadopoulos, A new maximally supersymmetric background of
IIB superstring theory, JHEP 0201 (2002) 047, hep-th/0110242.
[10] M. Blau, J. Figueroa-OFarrill, G. Papadopoulos, Penrose limits, supergravity and brane dynamics, Class.
Quantum Grav. 19 (2002) 4753, hep-th/0202111.
[11] D. Berenstein, J.M. Maldacena, H. Nastase, Strings in flat space and pp waves from N = 4 super-Yang
Mills, JHEP 0204 (2002) 013, hep-th/0202021.
[12] R.R. Metsaev, Type IIB GreenSchwarz superstring in plane wave RamondRamond background, Nucl.
Phys. B 625 (2002) 70, hep-th/0112044.
[13] R.R. Metsaev, A.A. Tseytlin, Exactly solvable model of superstring in plane wave RamondRamond background, hep-th/0202109.
[14] A. Pankiewicz, Strings in plane wave backgrounds, Fortschr. Phys. 51 (2003) 1139, hep-th/0307027.
[15] J.C. Plefka, Lectures on the plane-wave string/gauge theory duality, Fortschr. Phys. 52 (2004) 264, hepth/0307101.
509
[16] D. Sadri, M.M. Sheikh-Jabbari, The plane-wave/super-YangMills duality, hep-th/0310119.

[17] A.A. Tseytlin, Spinning strings and AdS/CFT duality, hep-th/0311139.
[18] S.R. Das, C. Gomez, S.J. Rey, Penrose limit, spontaneous symmetry breaking and holography in pp-wave
background, Phys. Rev. D 66 (2002) 046002, hep-th/0203164.
[19] E. Kiritsis, B. Pioline, Strings in homogeneous gravitational waves and null holography, JHEP 0208 (2002)
048, hep-th/0204004.
[20] R.G. Leigh, K. Okuyama, M. Rozali, pp-waves and holography, Phys. Rev. D 66 (2002) 046004, hepth/0204026.
[21] D. Berenstein, H. Nastase, On lightcone string field theory from super-YangMills and holography, hepth/0205048.
[22] M. Asano, Y. Sekino, T. Yoneya, pp-wave holography for Dp-brane backgrounds, Nucl. Phys. B 678 (2004)
197, hep-th/0308024.
[23] K. Sfetsos, Gauged WZW models and non-Abelian duality, Phys. Rev. D 50 (1994) 2784, hep-th/9402031.
[24] A.A. Kehagias, P.A.A. Meessen, Exact string background from a WZW model based on the Heisenberg
group, Phys. Lett. B 331 (1994) 77.
[25] G. Papadopoulos, J.G. Russo, A.A. Tseytlin, Solvable model of strings in a time-dependent plane-wave
background, Class. Quantum Grav. 20 (2003) 969, hep-th/0211289.
[26] M. Blau, M. OLoughlin, G. Papadopoulos, A.A. Tseytlin, Solvable models of strings in homogeneous plane
wave backgrounds, Nucl. Phys. B 673 (2003) 57, hep-th/0304198.
[27] H. Dorn, H.J. Otto, Two and three point functions in Liouville theory, Nucl. Phys. B 429 (1994) 375, hepth/9403141.
[28] A.B. Zamolodchikov, A.B. Zamolodchikov, Structure constants and conformal bootstrap in Liouville field
theory, Nucl. Phys. B 477 (1996) 577, hep-th/9506136.
[29] J. Teschner, On the Liouville three point function, Phys. Lett. B 363 (1995) 65, hep-th/9507109.
[30] J. Teschner, On structure constants and fusion rules in the SL(2, C)/SU(2) WZNW model, Nucl. Phys.
B 546 (1999) 390, hep-th/9712256;
J. Teschner, Operator product expansion and factorization in the H3+ WZNW model, Nucl. Phys. B 571
(2000) 555, hep-th/9906215.
[31] J.M. Maldacena, H. Ooguri, Strings in AdS3 and SL(2, R) WZW model. I, J. Math. Phys. 42 (2001) 2929.
[32] E. Witten, On string theory and black holes, Phys. Rev. D 44 (1991) 314.
[33] R. Dijkgraaf, H. Verlinde, E. Verlinde, String propagation in a black hole geometry, Nucl. Phys. B 371
(1992) 269.
[34] K. Becker, M. Becker, Interactions in the SL(2, R)/U (1) black hole background, Nucl. Phys. B 418 (1994)
206, hep-th/9310046.
[35] G. DAppollonio, E. Kiritsis, String interactions in gravitational wave backgrounds, Nucl. Phys. B 674
(2003) 80, hep-th/0305081.
[36] M. Bianchi, G. DAppollonio, E. Kiritsis, O. Zapata, String amplitudes in the Hpp-wave limit of AdS3 S 3 ,
JHEP 0404 (2004) 074, hep-th/0402004.
[37] D.I. Olive, E. Rabinovici, A. Schwimmer, A class of string backgrounds as a semiclassical limit of WZW
models, Phys. Lett. B 321 (1994) 361, hep-th/9311081.
[38] E. Kiritsis, C. Kounnas, String propagation in gravitational wave backgrounds, Phys. Lett. B 320 (1994)
264;
E. Kiritsis, C. Kounnas, Phys. Lett. B 325 (1994) 536, Addendum.
[39] E. Kiritsis, C. Kounnas, D. Lst, Superstring gravitational wave backgrounds with spacetime supersymmetry, Phys. Lett. B 331 (1994) 321.
[40] B. Pioline, M. Berkooz, Strings in an electric field, and the Milne universe, JCAP 0311 (2003) 007, hepth/0307280.
[41] M. Berkooz, B. Durin, B. Pioline, D. Reichmann, Closed strings in Misner space: stringy fuzziness with a
twist, JCAP 0410 (2004) 002, hep-th/0407216.
[42] V. Fateev, A.B. Zamolodchikov, A.B. Zamolodchikov, Boundary Liouville field theory. I: boundary state
and boundary two-point function, hep-th/0001012.
[43] A.B. Zamolodchikov, A.B. Zamolodchikov, Liouville field theory on a pseudosphere, hep-th/0101152.
[44] A. Giveon, D. Kutasov, A. Schwimmer, Comments on D-branes in AdS3 , Nucl. Phys. B 615 (2001) 133,
hep-th/0106005.
510
[45] P. Lee, H. Ooguri, J.W. Park, Boundary states for AdS2 -branes in AdS3 , Nucl. Phys. B 632 (2002) 283.
[46] B. Ponsot, V. Schomerus, J. Teschner, Branes in the Euclidean AdS3 , JHEP 0202 (2002) 016, hepth/0112198.
[47] S. Ribault, V. Schomerus, Branes in the 2-D black hole, JHEP 0402 (2004) 019, hep-th/0310024.
[48] A. Fotopoulos, V. Niarchos, N. Prezas, D-branes and extended characters in SL(2, R)/U (1), hepth/0406017.
[49] K. Hosomichi, Bulk-boundary propagator in Liouville theory on a disc, JHEP 0111 (2001) 044, hepth/0108093.
[50] B. Ponsot, Liouville theory on the pseudosphere: bulk-boundary structure constant, Phys. Lett. B 588 (2004)
105, hep-th/0309211.
[51] B. Ponsot, J. Teschner, Boundary Liouville field theory: boundary three point function, Nucl. Phys. B 622
(2002) 309, hep-th/0110244.
[52] O. Bergman, M.R. Gaberdiel, M.B. Green, D-brane interactions in type IIB plane-wave background,
JHEP 0303 (2003) 002, hep-th/0205183.
[53] M.R. Gaberdiel, M.B. Green, The D-instanton and other supersymmetric D-branes in IIB plane-wave string
theory, Ann. Phys. 307 (2003) 147, hep-th/0211122;
M.R. Gaberdiel, M.B. Green, Fortschr. Phys. 51 (2003) 713.
[54] M.R. Gaberdiel, M.B. Green, S. Schafer-Nameki, A. Sinha, Oblique and curved D-branes in IIB plane-wave
string theory, JHEP 0310 (2003) 052, hep-th/0306056.
[55] K.S. Cha, B.H. Lee, H.S. Yang, A complete classification of D-branes in type IIB plane wave background,
JHEP 0403 (2004) 058, hep-th/0310177.
[56] Y. Hikida, S. Yamaguchi, D-branes in pp-waves and massive theories on worldsheet with boundary,
JHEP 0301 (2003) 072, hep-th/0210262.
[57] K. Skenderis, M. Taylor, Branes in AdS and pp-wave spacetimes, JHEP 0206 (2002) 025, hep-th/0204054;
K. Skenderis, M. Taylor, Open strings in the plane wave background. I. Quantization and symmetries, Nucl.
Phys. B 665 (2003) 3, hep-th/0211011;
K. Skenderis, M. Taylor, Open strings in the plane wave background. II. Superalgebras and spectra,
JHEP 0307 (2003) 006, hep-th/0212184;
K. Skenderis, M. Taylor, An overview of branes in the plane wave background, Class. Quantum Grav. 20
(2003) S567, hep-th/0301221.
[58] D.Z. Freedman, K. Skenderis, M. Taylor, Worldvolume supersymmetries for branes in plane waves, Phys.
Rev. D 68 (2003) 106001, hep-th/0306046;
D.Z. Freedman, K. Skenderis, M. Taylor, Properties of branes in curved spacetimes, JHEP 0402 (2004) 030,
hep-th/0311079.
[59] O.J. Ganor, U. Varadarajan, Nonlocal effects on D-branes in plane-wave backgrounds, JHEP 0211 (2002)
051, hep-th/0210035.
[60] S. Stanciu, A.A. Tseytlin, D-branes in curved spacetime: NappiWitten background, JHEP 9806 (1998) 010.
[61] J.M. Figueroa-OFarrill, S. Stanciu, More D-branes in the NappiWitten background, JHEP 0001 (2000)
024;
J.M. Figueroa-OFarrill, S. Stanciu, Penrose limits of Lie branes and a NappiWitten braneworld, JHEP 0306
(2003) 025.
[62] H. Takayanagi, T. Takayanagi, Open strings in exactly solvable model of curved spacetime and pp-wave
limit, JHEP 0205 (2002) 012, hep-th/0204234.
[63] Y. Hikida, H. Takayanagi, T. Takayanagi, Boundary states for D-branes with traveling waves, JHEP 0304
(2003) 032, hep-th/0303214.
[64] N. Ohta, K.L. Panigrahi, S. Siwach, Intersecting branes in pp-wave spacetime, Nucl. Phys. B 674 (2003)
306, hep-th/0306186.
[65] C.P. Bachas, M.R. Gaberdiel, World-sheet duality for D-branes with travelling waves, JHEP 0403 (2004)
015, hep-th/0310017.
[66] S.F. Hassan, R.R. Nayak, K.L. Panigrahi, D-branes in the NS5 near-horizon pp-wave background, hepth/0312224.
[67] M. Berkooz, M.R. Douglas, R.G. Leigh, Branes intersecting at angles, Nucl. Phys. B 480 (1996) 265, hepth/9606139.
[68] M. Gutperle, A. Strominger, Spacelike branes, JHEP 0204 (2002) 018.
511
[69] A. Abouelsaood, C.G. Callan, C.R. Nappi, S.A. Yost, Open strings in background gauge fields, Nucl. Phys.
B 280 (1987) 599.
[70] A.Y. Alekseev, V. Schomerus, D-branes in the WZW model, Phys. Rev. D 60 (1999) 061901.
[71] G. Felder, J. Frohlich, J. Fuchs, C. Schweigert, The geometry of WZW branes, J. Geom. Phys. 34 (2000)
162.
[72] A.Y. Alekseev, A. Recknagel, V. Schomerus, Non-commutative world-volume geometries: branes on SU(2)
and fuzzy spheres, JHEP 9909 (1999) 023.
[73] C. Bachas, M.R. Douglas, C. Schweigert, Flux stabilization of D-branes, JHEP 0005 (2000) 048.
[74] C. Bachas, M. Petropoulos, Anti-de-Sitter D-branes, JHEP 0102 (2001) 025, hep-th/0012234.
[75] G. Sarkissian, M. Zamaklar, Diagonal D-branes in product spaces and their Penrose limits, JHEP 0403
(2004) 005, hep-th/0308174.
[76] P. Lee, H. Ooguri, J.W. Park, J. Tannenhauser, Open strings on AdS2 branes, Nucl. Phys. B 610 (2001) 3.
[77] J.L. Cardy, Boundary conditions, fusion rules and the verlinde formula, Nucl. Phys. B 324 (1989) 581.
[78] J.L. Cardy, D.C. Lewellen, Bulk and boundary operators in conformal field theory, Phys. Lett. B 259 (1991)
274.
[79] D.C. Lewellen, Sewing constraints for conformal field theories on surfaces with boundaries, Nucl. Phys.
B 372 (1992) 654.
[80] G. Pradisi, A. Sagnotti, Y.S. Stanev, Completeness conditions for boundary operators in 2D conformal field
theory, Phys. Lett. B 381 (1996) 97.
[81] I. Runkel, Boundary structure constants for the A-series Virasoro minimal models, Nucl. Phys. B 549 (1999)
563.
[82] R.E. Behrend, P.A. Pearce, V.B. Petkova, J.B. Zuber, Boundary conditions in rational conformal field theories, Nucl. Phys. B 570 (2000) 525;
R.E. Behrend, P.A. Pearce, V.B. Petkova, J.B. Zuber, Nucl. Phys. B 579 (2000) 707.
[83] H. Sonoda, Sewing conformal field theories I and II, Nucl. Phys. B 311 (1988) 401.
[84] A.Y. Alekseev, S. Fredenhagen, T. Quella, V. Schomerus, Non-commutative gauge theory of twisted Dbranes, Nucl. Phys. B 646 (2002) 127.
[85] N. Seiberg, L. Susskind, N. Toumbas, Strings in background electric field, space/time noncommutativity and
a new noncritical string theory, JHEP 0006 (2000) 021, hep-th/0005040.
[86] R. Gopakumar, S. Minwalla, N. Seiberg, A. Strominger, OM theory in diverse dimensions, JHEP 0008
(2000) 008, hep-th/0006062.
[87] O. Aharony, J. Gomis, T. Mehen, On theories with light-like noncommutativity, JHEP 0009 (2000) 023,
hep-th/0006236.
[88] G.W. Moore, N. Seiberg, Classical and quantum conformal field theory, Commun. Math. Phys. 123 (1989)
177;
G.W. Moore, N. Seiberg, Lectures on Rcft, RU-89-32, Presented at Trieste Spring School, 1989.
[89] N. Seiberg, E. Witten, String theory and noncommutative geometry, JHEP 9909 (1999) 032.
[90] D. Cremades, L.E. Ibanez, F. Marchesano, Yukawa couplings in intersecting D-brane models, JHEP 0307
(2003) 038;
D. Cremades, L.E. Ibanez, F. Marchesano, Computing Yukawa couplings from magnetized extra dimensions,
hep-th/0404229.
[91] M. Cvetic, I. Papadimitriou, Conformal field theory couplings for intersecting D-branes on orientifolds,
Phys. Rev. D 68 (2003) 046001.
[92] D. Lust, P. Mayr, R. Richter, S. Stieberger, Scattering of gauge, matter, and moduli fields from intersecting
branes, hep-th/0404134.
[93] E. Gava, K.S. Narain, M.H. Sarmadi, On the bound states of p- and (p + 2)-branes, Nucl. Phys. B 504
(1997) 214, hep-th/9704006.
[94] J.L. Karczmarek, H. Liu, J. Maldacena, A. Strominger, UV finite brane decay, JHEP 0311 (2003) 042,
hep-th/0306132.
[95] T. Eguchi, Y. Sugawara, Modular bootstrap for boundary N = 2 Liouville theory, JHEP 0401 (2004) 025,
hep-th/0311141.
[96] D. Israel, A. Pakman, J. Troost, Extended SL(2, R)/U (1) characters, or modular properties of a simple
non-rational conformal field theory, JHEP 0404 (2004) 045, hep-th/0402085.
512
[97] J.G. Russo, A.A. Tseytlin, Constant magnetic field in closed string theory: an exactly solvable model, Nucl.
Phys. B 448 (1995) 293, hep-th/9411099.
[98] Y. Hikida, Boundary states in the NappiWitten model, hep-th/0409185.
[99] E. Kiritsis, Introduction to Superstring Theory, Leuven Univ. Press, Leuven, 1998, hep-th/9709062.
[100] J.M. Maldacena, G.W. Moore, N. Seiberg, Geometrical interpretation of D-branes in gauged WZW models,
JHEP 0107 (2001) 046, hep-th/0105038.
[101] M.R. Gaberdiel, Fusion rules and logarithmic representations of a WZW model at fractional level, Nucl.
Phys. B 618 (2001) 407, hep-th/0105046.
Continuum limit of the integrable sl(2/1) 33

superspin chain
Fabian H.L. Essler a , Holger Frahm b , Hubert Saleur c,d
a The Rudolf Peierls Centre for Theoretical Physics, University of Oxford, 1Keble Road, Oxford OX1 3NP, UK
b Institut fr Theoretische Physik, Universitt Hannover, 30167 Hannover, Germany
c Service de Physique Thorique, CEA Saclay, 91191 Gif-sur-Yvette, France
d Department of Physics and Astronomy, University of Southern California, Los Angeles, CA 90089, USA
Received 12 January 2005; accepted 17 January 2005
Abstract
By a combination of analytical and numerical techniques, we analyze the continuum limit of the
integrable 3 3 3 3 sl(2/1) superspin chain. We discover profoundly new features, including
a continuous spectrum of conformal weights, whose numerical evidence is infinite degeneracies of
the scaled gaps in the thermodynamic limit. This indicates that the corresponding conformal field
theory has a non compact target space (even though our lattice model involves only finite-dimensional
representations). We argue that our results are compatible with this theory being the level k = 1,
SU(2/1) WZW model (whose precise definition requires some care). In doing so, we establish
several new results for this model. With regard to potential applications to the spin quantum Hall
effect, we conclude that the continuum limit of the 3 3 3 3 sl(2/1) integrable superspin
chain is not the same as (and is in fact very different from) the continuum limit of the corresponding
chain with two-superspin interactions only, which is known to be a model for the spin quantum Hall
effect. The study of possible RG flows between the two theories is left for further study.
PACS: 05.50; 71.27.+a; 11.10.-z
E-mail address: saleur@physics1.usc.edu (H. Saleur).

doi:10.1016/j.nuclphysb.2005.01.021
514
F.H.L. Essler et al. / Nuclear Physics B 712 [FS] (2005) 513572
1. Introduction
The sigma model approach [1] to phase transitions in noninteracting disordered systems
provides a convenient and appealing description of the physics at hand. The best known
example is the transition between plateaux in the integer quantum Hall effect (class A),
which can be described using a U (2n)/U (n) U (n) sigma model in the limit n 0.
More recently, the equivalent spin quantum Hall effect in d-wave superconductors has
been considered (class C [2]), and described with the SP(2n)/U (n) sigma model, in the
limit n 0 [3].
In both cases, the existence of a delocalization transition is associated with the masslessness in the IR of the sigma model at topological angle = . Identification of the
corresponding conformal field theory would lead to a presumably exact determination of
critical exponents which have been studied in great details in numerical works [4,5] as well
as in experiments.
Unfortunately, the determination of the IR fixed points in these problems has proven
extremely difficult. The canonical example of what might happen is the case n = 1 of
either replica model, which coincides with the O(3) sigma model at = . In this case,
there is a wealth of evidence [6] that the IR fixed point is the SU(2) WessZumino model
at level k = 1. In the flow, the field is promoted from being coset valued to group valued,
a highly nonperturbative feature which might have been hard to believe, were it not for the
existence of exact solutions [7].
The most reasonable way to understand what happens in the disordered system is first
to trade the replica limit for a super group [8,9], U (1, 1/2)/U (1/1) U (1/1) in the
quantum Hall effect, and OSP(2n/2n)/U (n|n) in the spin quantum Hall effect. Once the
corresponding sigma models have been satisfactorily defined [10], it is then tempting to
conjecture, by analogy with the O(3) example above, what the IR fixed point might be.
In this way, Zirnbauer [11] arrived at a proposal for the quantum Hall effect that is related
with the WZW model (at level k = 1) on PSU(1, 1/2). Roughly, this is obtained by observing that the base of the supersymmetric target space is H 2 S 2 , and then guessing
that this gets promoted to H 3 S 3 in the IR. One then tries to build the minimal proposal in agreement with this and known results on the exponents. It is important that for
the PSU(1, 1/2) WZW model the kinetic term is an exactly marginal deformation [12,13].
The final proposal of Zirnbauer corresponds to a particular value of this kinetic term, and
is not the WZW model. Hence, the symmetry is a global symmetry GL GR instead of
a local one. Unfortunately it is not clear whether this proposal gives the correct IR fixed
point. In particular, it is rather unpleasant to have to select a particular value for an exactly
marginal coupling constant if one hopes to describe what seems to be genuine universal
physics.
A less abstract direction of attack comes from network models [14]. These can be shown
to correspond, in the proper anisotropic limit to quantum spin chains with super group
symmetries [15]. Ideally, the problem would then be solved if one were able to apply in
this situation the formidable arsenal developed over the years in the study of quantum spin
chains. Unfortunately, there are difficulties here as well. For the ordinary as well as the spin
quantum Hall effect, the Hilbert space of the chain is of type V V V V , while
In the ordinary
the Hamiltonian is the invariant Casimir, the proper generalization of S . S.
515
quantum Hall effect, V and V are infinite-dimensional, while in the spin quantum Hall
effect, V and V are the two conjugate three-dimensional representations.
In either case, the Hamiltonian is not integrable. More precisely, the Hamiltonian does
not seem to derive from a family of commuting transfer matrices obtained via a standard
quantum group approach. In the spin quantum Hall case, the Hamiltonian can however be
diagonalized [16] (more precisely, the zeroes of the characteristic polynomial can be found,
as the Hamiltonian itself is not fully diagonalizable) using known results about the XXZ
chain and representation theory of the Temperley Lieb algebra [17]. No such approach
seems possible in the ordinary quantum Hall spin chain.
Since one is after the universality class of the IR fixed point, it is natural to wonder [18]
whether the Hamiltonian might be substituted by an integrable version without affecting the
exponents. Examples are known where such a trick would not work: for instance, the higher
odd spin integrable SU(2) spin chains flow to level k = 2s WZW models, while the chains
with Heisenberg type couplings flow to level k = 1. But examples are also known where
the trick does work. Indeed, with unitary groups and finite-dimensional representations,
group symmetry on the lattice plus criticality implies that the continuum limit is a WZW
model, the stablest of all being k = 1 [19]. Therefore, a spin chain with SU(n) symmetry,
if critical, will generically be in the same universality class as the integrable Sutherland
chain, that is the level one SU(n) WZW model. Criticality is however not guaranteed in
general. For instance, if one takes SU(n) and the alternance of n and n representations, the
chain whose Hamiltonian is the Casimir has a gap, while the integrable one is gapless, and
in the universality class of the WZW model [20].
Our purpose in this paper is to explore the matter further by concentrating on the spin
quantum Hall effect where many things are known exactly. In a nutshell, the question we
want to explore is whether the integrable 3 3 3 3 sl(2/1) spin chain is massless
and if so, in the same universality class as the nonintegrable 3 3 3 3 Heisenberg
chain, which is known to describe the physics of the spin quantum Hall effect.
A positive answer would raise hopes for an integrable approach to the ordinary quantum
Hall effect. Unfortunately, we will see that the answer is a resounding no. The reason
for this is the proliferation of possible fixed points that appear when ordinary compact
symmetries are replaced by super group, maybe noncompact, symmetries, and sheds light
on what should be directions of further research.
We will present evidence from various sources that the integrable alternating 3, 3 chain
is in the universality class of the SU(2/1) WZW model at level 1 [21]. This model has also
been called OSP(2/2)2 in the literature [22], and has appeared previously in the study of
Dirac fermions in a random gauge potential [23]. In contrast, it is already known from [17]
that the nonintegrable alternating 3, 3 Heisenberg chain does not scale to a WZW model,
but to a new kind of theory, where the currents have logarithmic OPEs, and the symmetry
is not of KacMoody type [24].
It is important to stress here that the WZW model is a very different theory from the
one [17] associated with the Heisenberg Hamiltonian; in particular, it has a continuous
spectrum of critical exponents, which is not bounded from below. Trying to infer generic
physical properties from the integrable chain would be a considerable mistake in this case,
and presumably in the ordinary quantum Hall case as well.
516
The paper is organized as follows. In Section 2, we gather information, obtained from

the literature as well as our own calculations, about the WZW model. In Sections 3 and 4,
we present the results of a numerical and analytical study of the Bethe ansatz equations of
the integrable spin chain, comparing our results with the expectations for the WZW model.
Section 5 contains elements of extension to the case of higher level. Section 6 goes back to
the spin quantum Hall problem. Technical details are provided in Appendices AD.
2. First considerations on the WZW model

2.1. Algebraic generalities
We start with a short review of the base sl(2/1) algebra. It contains a sub sl(2) with
generators J0 , J03 , an extra U (1) generator B0 , and two pairs of fermionic generators
V0 and W0 . Its representation theory is complicated, although this is one of the best
understood cases in the super Lie algebra literature [2528]. Typical representations are
characterized by a pair of numbers traditionally called b, j (and denoted [b, j ] in what
follows). Here, j is a sl(2) spin quantum number and takes values 0, 1/2, 1, . . . , whereas
b can be any complex number. Typicality requires b = j . Typical representations are
irreducible. Their dimension is 8j , and they decompose into sl(2) representations with
spin j, j 1/2, j 1/2, j 1 with, respectively, b numbers b, b + 1/2, b 1/2, b (for
j = 1/2, the value j 1 is omitted). We introduce the notation jb to denote an sl(2)
representation of spin j for which the other u(1) generator takes the constant value b, so
b+1/2
b1/2
the former decomposition reads jb j 1/2 j 1/2 jb1 . The Casimir is diagonal in
these representations, and proportional to j 2 b2 .
A particularly important representation in the WZW model will be the four-dimensional
[b = 0, j = 1/2] which is represented graphically in Fig. 1. It is the defining representation
for osp(2/2).
Fig. 1. Graphical representation of the representation [0, 1].
517
Fig. 2. Graphical representation of the representation [0, 1/2, 1/2, 0].
There are many more representations. Simple atypical (that is, irreducible atypical)
occur when b = j , and have dimension 4j + 1. Examples are the three-dimensional representations [1/2, 1/2] and [1/2, 1/2] that we will use to define the Hilbert space for
j
j +1/2
the quantum spin chain. Representations [j, j ] contain only multiplets j and j 1/2 , so
1/2
[1/2, 1/2] contains 1/2 and 01 . Similarly, representations [j, j ] contain only multiplets
j
j 1/2
1/2
with j and j 1/2 , so [1/2, 1/2] contains 1/2 and 01 . While typical representations have vanishing superdimension, simple atypical ones have superdimension equal to
plus or minus one.
The other atypical representations are not simple: they are indecomposable1 but partly
reducible, and form what is often called blocks. The most important example is the one
appearing in the tensor product [0, 1/2] [0, 1/2] which decomposes as the adjoint [0, 1]
and another eight-dimensional representation usually denoted as [0, 1/2, 1/2, 0]. This
representation is the semi direct sum of [1/2, 1/2], [1/2, 1/2], and of two [0, 0] represen1/2
1/2
tations. Its su(2) content is 1/2 , 1/2 , 01 , 01 , and 00 twice. A graphical representation
is given in Fig. 2. Note that this representation has vanishing superdimension.
It is useful to represent such an indecomposable representation by its quiver diagram
[29]. Denote by j the representation [j, j ] and by j the representation [j, j ] (these
simple atypical representations can be obtained by maximum supersymmetrization of the
fundamental or its conjugate2j boxes in a Young diagram representation). Then the eight
1 We may use indecomposables as short hand for non simple indecomposables, while in the mathematics
literature, indecomposable encompasses irreducible.
518
dimensional indecomposable can be represented as
1/2

1/2
(2.1)

0
Here, the bottom part of the diagram means that the representation contains a singlet as
an invariant subspace. There is then two submodules of dimension three which are not
invariant but are (and then, isomorphic to 1/2 and 1/2) modulo the singlet, and finally,
on top, a submodule of dimension one which is invariant modulo all the others. A generic
element of the algebra, written in the basis made of the top, middle and bottom components
of the quiver in this order, will have the form
0
0
0
0
a 1/2
0
0
1/2
=
(2.2)
.
1/2
b
0
1/2
0
c
d
e
0
Here, the are 3 3 matrices, a and b are 3 1, d and e are 1 3, and c is 1 1.
More generally, replacing 0 by j , and 1/2, 1/2 by j 1/2, j + 1/2, or the same
with the conjugate of all these representations, gives the quiver diagram for indecomposables of dimension 8 (for j = 0) and 16j + 4 (j > 0). These representations (denoted
[j, j 1/2, j + 1/2, j ]) and their conjugates are the only indecomposables whose quiver
has a loop. They are projective representations (see the appendix for some algebraic
reminders). All the other indecomposables2 are not projective, and associated with onedimensional quivers. They will be discussed soon.
We now turn to the current algebra. First, a note about conventions. We chose normalizations such that the SU(2/1) WZW at level k contains a sub SU(2) current algebra at level
k in the standard convention for the latter. Other conventions would denote this model as
the OSP(2/2) level 2k model (the two superalgebras are isomorphic).
The basic OPEs read:
k
2
+ J 3,
z2 z
J
,
J 3 (z)J (0) =
z
k
k
J 3 (z)J 3 (0) = 2 ,
B(z)B(0) = 2 ,
2z
2z
V+
W+
,
J + (z)W (0) =
,
J + (z)V (0) =
z
z
J + (z)J (0) =
2 We restrict here to cases where the Cartan subalgebra is diagonalizable, which should be the relevant one for
our problem.
519
V
W
,
J (z)W + (0) =
,
z
z
V
W
J 3 (z)V (0) =
,
J 3 (z)W (0) =
,
2z
2z
V pm
W
,
B(z)W (0) =
,
B(z)V (0) =
2z
2z
J+
J
W + (z)V + (0) =
,
W (z)V (0) =
,
z
z
k
B J3
,
V + (z)W (0) = 2 +
z
z
k
B J 3
V (z)W + (0) = 2 +
(2.3)
z
z
and the generators of the base algebra obtained by taking the zero modes of the currents.
The Sugawara construction then leads to vanishing central charge independently of the
level k, as the adjoint has vanishing superdimension. The conformal weights for affine

representations [b,
j ] based on irreducible typical representations [b, j ] read
J (z)V + (0) =
j 2 b2
.
(2.4)
k+1
Affine highest weight representations built on simple atypical representations have conformal weight equal to zero. Finally, for those built on indecomposable atypical representations, L0 is not diagonalizable.
h=
2.2. The case k = 1 and a free field representation

The simplest example, and the one most likely to be encountered as the continuum limit
of a lattice model, is the case k = 1. It is well known [21,22] that this theory admits a free
field representation, based on a pair of symplectic fermions 1 , 2 , and a pair of bosons
, . The boson is compact and has the usual metric, the boson has a metric of the
opposite sign. The propagators are

1 (z, z )2 (w, w)
= ln |z w|2 ,

(z)(w) = ln(z w),

(z) (w) = ln(z w).
(2.5)
The central charge of the model is 2 + 1 + 1 = 0, as required for the SU(2/1) WZW
models.
The currents admit the following free field representation:
J+ = e
2i
,
i 2i
J = e
,
1
J 3 = i,
2
520
1
B = i .
2
The fermionic currents consist of two sl(2) doublets with b = 1/2:
V =e
1 i(+ )
2
W = e
(2.6)
1 ,
1 i( )
2
2 .
(2.7)
The WZW model has never (to our knowledge) been completely understood, even for
k = 1: questions such as the operator content or the operator algebra only admit partial
answers, at best. Of course, it is tempting to speculate that the operator content derives
somehow from the sub SU(2) part, which for k = 1 contains only the affine representations
at j = 0, j = 1/2. The j = 1/2 case corresponds, for the fundamental representation of
osp(2/2) which has b = 0, to a conformal weight h = 18 , and one thus expects to have,
for k = 1, at least the representations with h = 0 and h = 18 . But it is easy to argue that
the operator content is considerably richer, and in particular exhibits logarithmic features.
Indeed, consider the tensor product of the [0, 1/2] representation (the multiplet for the
fundamental field of the theory) with itself;
[0, 1/2] [0, 1/2] = [0, 1] + [0, 1/2, 1/2, 0].
(2.8)
The right-hand side does not contain the true singlet [0, 0], which means, jumping from
tensor product in the base algebra to fusion product in the conformal field theory [22],
that the identity field must have logarithmic partners, and fields of dimension h = 0 be
organized in several representations of sl(2/1).
To gain further insight on this question, we can study the OPE associated with (2.8)
by using the free field representation. The doublet in the [0, 1/2] representation can be
represented by [22]
v = e
(2.9)
where is the = twist field in the symplectic fermion theory, with dimension 1/8,
which, added to +1/4, gives the desired 1/8. The two singlets meanwhile can be represented as [22]
1
2
w = e
(2.10)
where
are the = 1/2 and = 3/2 twist fields in the symplectic fermion theory.
Their dimension is 3/8, which, added to 1/4, gives the desired 1/8. The fields in the
adjoint representation are the currents, whose expression was given earlier. Finally, for the
fields in the indecomposable representation, they are given by
1 ,

ei 2
1 2
1 1
2 ,
ei

2
2 2 ,
1
(2.11)
where the fields are arranged exactly as in the quiver for [0, 1/2, 1/2, 0] given above. All
these fields have vanishing conformal dimension. The superdimension of [0, 1/2, 1/2, 0]
521
vanishes. On the other hand, it is known (see below) that the Witten index of the theory is
equal to one. This means that in the level k = 1 WZW model there must be at least the fields
in (2.11) and a true identity field, which is not embedded in a bigger indecomposable
representation. The free field representations is thus not complete as is, since two fields
1 must be introduced.
In fact, it is possible to build (many) other indecomposable representations associated
with fields of vanishing conformal weight. The simplest are obtained as representations
containing either the field 1 or the field 2 . For instance, the fields
2 ,
(+ )
ei
2
(2.12)
0
1/2
form a representation one can represent by the diagram . Similarly, the fields
1 ,
( )
ei
2
1/2
(2.13)
0
can be represented by the diagram . The module at the end of the arrow is invariant, and quotienting by this module amounts to erasing the dot, that is the remaining
module is then invariant. For more general such diagrams, the dots will represent atypical
representations of the type [j, j ] discussed earlier.
These structures do not come alone, but are always embeddable in larger structures (in
mathematical terms, they are not projective representations). This is easily seen in the

free field representation. For instance, introducing the field ei 2( ) 2 and acting with
1
1/2
the generators gives rise to . where the [1, 1] submodule is constituted by

the fields
ei
e
2( )
(+ )
2 ,
2 1 ,
( )
ei
2 1 ,
i 2(+ )

2
2 ,
2 .
(2.14)
In turn, these fields can be connected to bigger representations, and one finds that the
conformal fields of vanishing weight built on top of a single fermion 1 or 2 can be
arranged into the semi-infinite quiver
p
p+ 12
p+1
1/2
(2.15)
or its mirror image. The highest weight (that is, the field annihilated by the zero modes of
ni
( )
ni
( )
all raising operators) of n/2 is the field e 2

2 for n even, e 2
for n odd, and
similarly for +n/2.
Of course, since we are at level k = 1, these highest weights for n > 1 are not affine
highest weights, that is, their OPE with the SU(2) currents gives poles of degree higher
than one. For instance for n = 2,
J (z)ei
2( )
2 (w) =
1
i 2(z) i 2( )
:e
e
2 (w):.
(z w)2
(2.16)
Therefore, applying modes with positiveindices will produce fields with negative dimen
sion. In this example we get the field ei 2 2 of dimension 1. This field belongs to the
522
representation [b = 3/2, j = 1/2], for which ei/ 2(3 ) 2 is an affine highest weight.
The pattern generalizes, and one finds that for any n > 1, the fields at weight h = 0 in the
semi-infinite quiver for node n/2 are affine descendents of the field

i

exp ( (2n 1) ) 2 n1 2
(2.17)
2
which is the highest weight of [n + 1/2, 1/2], with h = n(n 1)/2. Concatenating
[1/2, 1/2] and [0, 0] in a four-dimensional representation we will still denote (abusively)
by [1/2, 1/2], the set of conformal fields associated with the semi-infinite quiver seems
to be made of all affine highest weight representations with base [b = n + 1/2, j = 1/2],
n = 1, . . . , . Similarly for the mirror image, we get all the representations [b = n
1/2, 1/2], n = 1, . . . , .
Note that as we increase the value of n we get larger and larger negative dimensions,
and the semi-infinite quiver corresponds to a theory whose spectrum is not bounded from
below. This feature is rather characteristic of supergroup WZW models. It does not seem
to make physical applications impossible however, as we will discuss below.
The algebraic structure of the representations of the affine algebra associated with the
semi-infinite quivers is very interesting of course, and deserves much further study, beyond
the scope of this paper. We shall refer to them, whenever necessary, as infinite blocks
representations of the current algebra.
Note that in a physical theory, left and right sectors have to be combined. While we have
treated the fermions so far as chiral objects, they are not, and when completing with the
left moving parts of the boson operators, the left moving part of the fermions should also
be included.
Let us now backtrack a bit and wonder what of this algebraic zoo will appear in the
WZW model. First, observe that the free field representation of the currents uses only the
derivatives of the fermions, which are well-defined conformal (chiral) fields. One might
have hoped that maybe a theory can be defined that does not use the fermion fields themselves (and thus is based only on the small algebra [30]). However, consideration of the
four point functions and the KnizhnikZamolodchikov equation seem to exclude this possibility, as the identity clearly must have logarithmic partners, and the fermions themselves
are necessary to reproduce this feature in the free field representations [22]. One could
then have hoped that fields of dimension zero are reduced to the true singlet and the eightdimensional indecomposable [0, 1/2, 1/2, 0]. This seems also incorrect. For instance, it
is known that the fusion of the twist fields with dimension h = 1/8 and h = 3/8 in the
symplectic fermions theory does produce the fermionic fields themselves [30]. Also, we
shall see later that modular transformations of characters suggests the presence of more
fields of weight zero. If so, then the nonprojective indecomposables must appear, and then
there is no reason to restrict them. We in fact speculate that the infinite blocks defined previously appear in the WZW model at level one. This in turn implies appearance of arbitrarily
large negative conformal weights.
Indecomposable representations of sl(2/1) lead to affine highest weight representations
with vanishing conformal weights. To have a nonvanishing conformal weight for an affine
highest weight representation requires dealing with a typical, irreducible representation
of sl(2/1). There, the presence of a sub SU(2) at level one presumably allows only spins
523
j = 0, 1/2 to occur. So a minimal guess for the WZW model would be to add to the infinite
blocks just discussed, the affine highest weight representation based on [0, j = 1/2]. But
we have already seen appearance of the representations [n + 1/2, 1/2] in relation with
the infinite blocks. Thus it seems to us that the spectrum of the WZW model should at least
include

[0,
0] [0,
0]

[n +
1/2, 1/2] [n
1/2, 1/2].
(2.18)
n=
2.3. Relevance to lattice models?

The appearance of arbitrarily large negative conformal weights is related with the fact
that the invariant metric on the supergroup is not positive definite, and thus the functional
integral ill defineda feature directly at the origin of the wrong sign in the propagator
of the free field . This might lead one to think that the conformal field theory we have
tried to define through the free field representation of the SU(2/1) current algebra cannot
in any case appear as the continuum limit of a lattice model, for which (see below) one can
easily establish the existence of a unique ground state. But things are more subtle. It might
be for instance that the arbitrarily large negative conformal weights3 correspond to nonnormalizable states, and thus are absent from the spectrum and the partition function. This
would be the situation for a lattice model whose continuum limit is a noncompact boson
[32]. It might also be that the arbitrarily large negative conformal weights do correspond
to normalizable states, but that the correspondence between the spectrum and the partition
function is more complicated than for unitary conformal field theories say, and proceeds
through some sort of analytic continuation. This seems to be the case for the system for
instance [33], where instead of the naive characters, the modular invariant partition function is made of character functions, which encode the arbitrarily large negative conformal
weights in a subtle way. We will argue that this is the situation in the SU(2/1) model as
well. In other words, the spectrum deduced from the lattice model may be related to the
spectrum of the conformal field theory only through some sort of analytical continuation.
We will discuss below a very naive version of this continuation.
Finally, and getting a bit ahead of ourselves, one might recall that Zirnbauer (see [34]
for a thorough discussion) has suggested to replace the target space from (in this case)
SU(2/1) with its indefinite metric by a non compact target space with a positive metrica
Riemannian symmetric superspace. The resulting functional integral would then be well
defined, and the spectrum bounded from below. It is not clear to us how one would study
this model in practice. But it is tempting to conjecture that its spectrum would coincide
with the analytical continuation of the naive spectrum deduced from the algebraic study of
the current algebra, that is the spectrum observed in the lattice model. If so, the question of
what is the continuum limit of the lattice model may have, in a way, two answers. This
obviously requires more thinking.
3 Note that these are not necessarily excluded on physical grounds. It might well be that in some systems,
these arbitrarily large negative conformal weights do describe physical observables, like powers of the wave
function [31].
524
To proceed, we now turn to characters and their modular transformations. This will give
us a handle on the mysterious analytical continuation at play, and allow us to make contact
with the conjecture (2.18).
3. Some results about the WZW model from characters

3.1. Integrable characters
3.1.1. Ramond sector
At level k = 1, four admissible characters [35] have been identified. In the Ramond
sector (periodic fermions on the cylinder) one has the field j = 1/2, b = 0 with character

2
R,I
h=1/8
(3.1)
(, , ) = q 1/8 z1/2 F R (, , )
q 2a +a z2a z2a1 .
aZ
3
Here we use the definition = Tr q L0 c/24 e2i J0 e2iB0 , J03 and B0 being the zero
modes of the J 3 field (the third component of the spin) and the B field. Normalizations are
such that for the highest weight state in the h = 1/8 R sector, J03 = j = 1/2, B0 = b = 0.
We have set z = e2i and = e2i . Finally,
F R (q, z, ) =
(1+z1/2 1/2 q n )(1+z1/2 1/2 q n1 )(1+z1/2 1/2 q n )(1+z1/2 1/2 q n1 )

.
(1q n )2 (1zq n )(1z1 q n1 )
(3.2)
n=1
It is convenient to rewrite this character as [35]

R,I
(, , ) =
h=1/8

1
0 (, )1/2 (, ) + 1/2 (, )0 (, ) ,
( )
(3.3)
where the s in (3.3) are characters of the SU(2)1 WZW model, and are given by the
following well-known expressions [36]
0 (, ) =
1 a2 a
q z ,
( )
aZ
1 (a+1/2)2 a+1/2
1/2 (, ) =
q
z
( )
(3.4)
aZ
(note that these characters are even functions of , as expected by symmetry of the J03
spectrum). and we have set z = e2i ; similarly we set = e2i . The specialized Ramond
character = = 0 for the h = 1/8 field reads therefore

2
R,I
h=1/8
( ) = 0 1/2 ( ) = 4q 1/8 1 + 5q + 22q 2 + ,
(3.5)
where we recognize the overall multiplicity 4, coinciding with the dimension of the representation [0, 1/2]. The latter decomposes as one spin 1/2 and two spin 0 SU(2) representations.
525
The Ramond character for the identity field is considerably more complicated [35]:
R,I V
= F R (, , )
h=0
q 2a z2a
1 q 2a z1
aZ
(1 + q a z1/2 1/2 )(1 + q a z1/2 1/2 )
(3.6)
The specialized character turns out to be finite, despite the divergence of F R :

2

a

1 + qn 4
q 2a +a
R,I V
2a 2 1 q
h=0 =
aq
8
1 16
1 qn
1 + qa
(1 + q a )2
a=1
a=1
= 1 + 8q + 24q + .
2
(3.7)
We thus see that in the Ramond sector, all conformal weights are of the form h = 0 mod integers or h = 18 mod integers. Setting = 1 gives us Tr(1)F (a supertrace) and evaluates
sR,I
sR,I V
what is called the super character; we check that h=1/8
( ) = 0 while h=0
( ) = 1.
It is possible to write the specialized character in a more compact form using the function [37]
q a 1/8
1
.
( )3 ( )
1 + q a1/2
2
h3 ( ) =
(3.8)
aZ
One finds then

R,I V
( ) =
h=0
2
02 1/2
2
02 + 1/2
( ) + 20 1/2 h3 ( ).
(3.9)
3.1.2. NeveuSchwarz sector

In the NS sector, the fermions have antiperiodic boundary conditions, and the SU(2/1)
symmetry is broken down to a sub SU(2) U (1). The lowest energy state in this sector
has negative conformal dimension, leading to an effective central charge equal to ceff = 3.
The corresponding character is
NS,I
h=1/8
(, , ) =

1
1/2 (, )1/2 (, ) + 0 (, )0 (, ) .
( )
The specialized characters reads

NS,I
h=1/8
( ) = q 1/8 1 + 4q 1/2 +
(3.10)
(3.11)
so half integer gaps appear in this sector.

The other character is again more complicated
NS,I V
(, , )
h=1/4
= q 1/4 z1/2 F NS (, , )

1 zq 2a+1
2
,
q 2a +2a z2a
(1 + q a+1/2 z1/2 1/2 )(1 + q a+1/2 z1/2 1/2 )

aZ
(3.12)
526
where
F NS =
(1+z1/2 1/2 q n1/2 )(1+z1/2 1/2 q n1/2 )(1+z1/2 1/2 q n1/2 )(1+z1/2 1/2 q n1/2 )
.
(1q n )2 (1zq n )(1z1 q n1 )
n=1
(3.13)
The specialized character is again finite, despite the divergence of

NS,I V
h=1/4
( ) = h3 ( ) 0 ( )2 + 1/2 ( )2
F NS :
(3.14)
NS,I
and like h=1/8
, exhibits half integer gaps.
3.2. The remaining operator content

It is crucial to realize that the characters we wrote down all have complicated or unusual
modular transformations. A possible strategy to proceed is then to try to supplement these
characters with others in order to obtain a finite-dimensional representation of the modular
group, and maybe modular invariants.
3.2.1. A continuum above h = 1/8 in the R sector
sNS,I
Consider for instance the NeveuSchwarz charac3.2.1.1. Transformation of h=1/8
ters. They correspond to fermions being antiperiodic in both directions. Alternatively, this
corresponds to an antiperiodic chain along the space direction, while one takes a trace
along the (imaginary) time direction. Imagine instead taking a supertrace along the imaginary time direction, or setting = 1 in the expression of the characters, giving rise again
to a supercharacter. The specialized supercharacter for h = 1/8 reads
sNS,I
( ) =
h=1/8
2
02 1/2
( ).
(3.15)
Recall now the modular transformation of the basic SU(2)1 WZW characters
1

1
0
0
2
2
(1/ ) =
( ).
1
1/2
1/2
1
(3.16)
Meanwhile the function obeys (1/ ) = i ( ). So the only way to write the
modular transform of this SU(2/1) (super)character as a sum of powers of q is to introduce
an integral representation of the factor
1
d ei
(3.17)
and thus

sNS,I
(1/ ) = 20 ( )1/2 ( )
h=1/8
q /2
R,I
= h=1/8
( )

d q
2 /2
(3.18)
527
Now observe that under modular transformation the boundary conditions which determine
the super NS character turn into boundary conditions where the fermions are periodic
in the space direction, and one takes a trace in the imaginary time direction. Hence the
right-hand side of (3.18) should appear in the generating function of the spectrum in the
sl(2/1) symmetric sector of our chain if the associated conformal field theory contains the
NeveuSchwarz characters discussed previously. This means there should be a continuous
component starting at the h = 1/8 value.
To find out about this continuous component, we can keep track of the B0 and J03 numbers by keeping the parameters , :

3
sNS,I
(, , ) Tr (1)F q L0 zJ0 B0
h=1/8

1
0 (, )0 (, ) 1/2 (, )1/2 (, ) .
=
(3.19)
( )
The modular transformation of general SU(2) characters reads
1

1
0
0
2
2
i 2 /2
(1/, / ) = e
(, )
(3.20)
1
1/2
1/2
1
and thus
sNS,I
h=1/8
(1/, /, / )
ei(
2 + 2 )/2

0 (, )1/2 (, ) + 1/2 (, )0 (, ) .
(3.21)
i ( )
Due to the indefinite metric for the SU(2/1) supergroup, the overall factor in modular
2
2
2
2
transformations should however not be ei( + )/2 , but rather ei( )/2 (since the
2
Casimir is proportional to j 2 b2 ). We are then left with an overall factor ei / whose
interpretation is slightly delicate. What we propose is to write
1
i 2 /
1
=
1
db ei b e2ib .
2
(3.22)
The phase factor is ambiguous and depends on a choice of cut. As for the integral, is
divergent in the physical situation where Im > 0. If we assume however that this integral
is defined by some sort of analytical continuation from the case Im < 0, we see that
we can formally interpret the leftover 2 exponential in our modular transform (3.21) as
resulting from an integral over fields with conformal dimension and B0 charges:
1
h = b2 ,
(3.23)
B0 = b, < b < .
2
Moreover, the right-hand side of the modular transform is, for each value of b, of the form

1 i b2 2ib
e
0 (, )1/2 (, ) + 1/2 (, )0 (, )
(3.24)
e
( )
which is exactly the character for the affine representation with base [b, 1/2] (integrable
character in class I in the language of [35]). It thus seems that the model should contain in
fact, not only the representation [0, 1/2], but the continuum [b, 1/2] with b real. This is at
528
odds with the first section where we speculated that [n + 1/2, 1/2] only appeared. See
below for a discussion of this issue.
While the conformal weights associated with those representations become arbitrarily
large and negative, the analysis of the characters and modular transforms suggests that the
finite size spectrum is given by a sort of analytical continuation, whose result, by the mechanism discussed above, is to make the continuum appear above the 1/8, not below. Note
also that while the B0 numbers in the chiral and antichiral sector appear continuous, the
lattice analysis anyway only identifies the combination L + R, which is made of integers.
sNS,I V
We can perform the same analysis for the specialized
3.2.1.2. Transformation of h=1/4
supercharacter for h = 1/4:

0 1/2
sNS,I V
2
.
( ) = 2 2
+ h3 02 1/2
h=1/4
(3.25)
2
0 + 1/2
The function h3 obeys

1
h3 (1/ ) = h3 ( ) +
q /2
.
2 cosh
(3.26)
Thus
sNS,I V
h=1/4
(1/ )
2 2
1/2
0
1
= 2
( ) + 20 1/2 ( ) h3 ( ) +
2
( )
1/2 + 0

2
q /2
d
,
2 cosh
(3.27)
where we again obtain a continuous spectrum starting at h = 18 . This time however, instead
of a flat measure over all the reals, one has a measure proportional to 1/ cosh . Note we
can write

2
q /2
sNS,I V
R,I V
R,I
.
h=1/4 (1/ ) = h=0 ( ) + h=1/8 d
(3.28)
2 cosh
To find out about the spectrum of B0 and J03 values is considerably more complicated,
and the tired reader may want to jump to formula (3.44). The following calculations rely
on the paper [38]; one starts by writing
sNS,I V
(, , )
h=1/4
= e
i/2 i
sNS

+

K4 ,
,
(, , ) K4 ,
,
,
2
2
2
2
(3.29)
where
F sNS =
(1z1/2 1/2 q n1/2 )(1z1/2 1/2 q n1/2 )(1z1/2 1/2 q n1/2 )(1z1/2 1/2 q n1/2 )
.
(1q n )2 (1zq n )(1z1 q n1 )
n=1
(3.30)
529
It is useful to rewrite this prefactor as

F sNS =
+1
+1
10 (, +
)10 (,
)
2
2
.
1/8
3
11 (, )q
(q)
(3.31)
The K are Appell functions

Kl (, , ) =
eim2 l +2iml
1 e2i( ++m )
(3.32)
mZ
and the are usual theta functions

1 + z1 q m
1 + zq m
1 qm ,
10 (q, z) =
m0
m1
m1
m0
m1
m1

11 (q, z) =
1 z1 q m
1 zq m
1 qm .
Modular transformations of the theta functions are as follows:
1
2 i
11 (1/, / ) = i i ei+i (1/2) + 4 11 (, ),
(1/2)2
10 (1/, / ) = i ei 10 (, + 1/2 /2).
(3.33)
(3.34)
As for the Appell functions it is much more complicated. One finds [38]
Kl (1/, /, / )
= eil(
+
2 2 )/
l1

Kl (, , )
eil( +a/ l)
2 /
(l, l a )00 (l, l + a ),
(3.35)
a=0
where 00 (q, z) = 10 (q, zq 1/2 ) and

i
1
x 2 sinh(x i (1 + 2/ ))
(, ) =
.
dx e
2
2 i
sinh(x i )
(3.36)
Putting everything together gives

sNS,I V
h=1/4
(1/, /, / )
1 i( 2 2 )/2 2i/ i/
e
e
e
10 (, ( + )/2)10 (, ( )/2)
11 (, )3 ( )

K4 (1/, ( 1)/2, /2 ) K4 (1/, ( 1)/2, /2 ) .
(3.37)
530
Using the general Appell transform formulas gives

K4 1/, ( 1)/2, /2

2
2
= ei( )/ ei e2i/ K4 , ( 1)/2, /2
+
l1

ei( 1+a/2)
2 /

(4, 2 a ) 4, 2 + (2 + a) .
(3.38)
a=0
We recognize the expected global factor ei(
2 2 )/2
. Also, observe that
10 (q, x 1 y 1 )10 (q, x 1 y)

10 (q, xy)10 (q, xy 1 )
=
11 (q, x)
11 (q, x 1 )
(3.39)
while
10 (, ( + )/2)10 ( ( )/2)
.
(3.40)
11 (, )ei/4 3 ( )
We also observe that K4 (, , ) = K4 (, 1, ). Therefore, the contribution to (3.37)
coming from the K4 term in the transform (3.38) gives
F R (, , ) =
First term = ei(
2 2 )/2
R,I V
h=0
(, , ).
(3.41)
The second term reads, after making the function explicit,

F R (, , )iei(
2 + 2 )/2
3

2
(1)a (4, 2 a )q a /8 za/2
a=1
2m2 +ma
2m

z z2ma .
(3.42)
We recognize the prefactor of the function as the formal extension of the Ramond character formula, initially valid only for j = 1/2, b = 0, h = 1/8 to j = a, b = 0, h = a 2 /8,
R,I
h=a
2 /8 . However, setting

2
2
q 2m +ma z2m z2ma
Xa q a /8 za/2
(3.43)
m
it is easy to show that Xa = Xa4 = Xa . Therefore, X0 = X2 = 0, and X1 = X3 .

Taking advantage of these symmetries leads us to a considerably simplified expression
sNS,I V
(1/, /, / )
h=1/4
= ei(
2 2 )/2
R,I V
h=0
(, , )
cosh(2x i / )
(3.44)
.
2 cosh(x i )
As a check we can set = 0 in this expression. Using the invariance of h3 ( ) + h3 (1/ )
we see that

2
1
q /2
(3.45)
d
f ( )
( )
2 cosh
R,I
+ h=1/8
(, , )ei(
2 + 2 )/2
dx ex
is invariant, and thus

2
2
ex
q /2
dx
= d
2 cosh
2 cosh x i
531
(3.46)
which allows us to recover expression (3.28).

We now go back to the more general expression. Crucial is the fact, as mentioned earlier,
2
that the exponential has the
term with positive sign instead of a negative one. We thus
write in the integral x = y i/ and thus formally obtain

2
i i 2 /
i
ei(+y) /
iy 2 / cosh 2y/
e
=
.
(3.47)
dy e
dy
cosh iy
cos y
Using the formal integral representation of the exponential we write this as

2i(+y)b
1
2e
2
db dy ei b
= db e2ib ei b
.
cos y
cos b
(3.48)
Finally, we thus rewrite (3.44) as

sNS,I V
h=1/4
(1/, /, / )
= ei(
2 2 )/2
R,I V
h=0
(, , )
2
2
R,I
+ h=1/8
(, , )ei( )/2
db e2ib ei b
1
.
cos b
(3.49)
This corresponds again to representations [b, 1/2] but this time with a measure that is not
flat, but proportional to 1/ cos b. The interpretation is discussed below.
NS,I
NS,I V
3.2.1.3. Transformation of h=1/8
and h=1/4
Finally, we can also consider the ordinary NS character. We know that the boundary conditions corresponding to it character are
invariant under the modular transformation. But using the expression
NS,I
( ) =
h=1/8
2
02 + 1/2
( )
(3.50)
we see that the modular transform has a continuous component starting at the ground state
of the sector, namely h = 1/8:

2

q /2
1
NS,I
NS,I
2
h=1/8
(3.51)
h=1/8
( ) d
(1/ ) =
( ) = 02 + 1/2
.
i
Similarly

NS,I V
NS,I V
NS,I
(1/ ) = h=1/4
( ) + h=1/8
( )
h=1/4
q /2
d
.
2 cosh
(3.52)
A final note: in the supersymmetric literature (the model has some relations to N = 4
theories [37]), the subset with discrete quantum numbers is referred to as massless, the one
with continuous quantum numbers as massive.
532
3.3. Interpretation
Comparing the free field analysis of the WZW model with the analysis of the modular
transformations of basic characters suggests that the spectrum of arbitrarily large negative
dimensions admits regularized expressions for characters, and thus partition function and
spectra of lattice Hamiltonians. For these, there is a single ground state, and the negative
dimensions are folded back into a continuous spectrum of positive dimensions.
The slightly surprising thing is that we find in this way indications of a continuous spectrum based on [b, 1/2] while we initially expected a discrete one [n + 1/2, 1/2]. There
are stronger reasons for a discrete spectrum of b than the ones given in the first sections.
For instance the free field representation of the generators selects a particular radius for the
noncompact boson, and vertex operators corresponding to charges continuous charges b
will in general be nonlocal with respect to these currents! Of course it might be in fact that
the free field representation we have used is good only for a subset of fields, and does not
allow one to explore the full operator content. But it could also be that the continuous spectrum we read from the modular transformations is an artifact from the (rather uncontrolled)
regularization at work. A hint in this direction comes from the integral

2
q /2
2 cosh
which, after continuation to the imaginary axis, has poles precisely at the values b =
n + 12 . It seems difficult to conclude on this issue without more work on the conformal
field theory side.
4. Analysis of the lattice model

The next step is to extract as much information about the lattice model as possible and
compare it with our proposal for the continuum limit. Although we have mostly talked
about a quantum spin chain so far, integrability allows one to consider it as part of a larger
family of commuting transfer matrices related to an integrable vertex model.
4.1. Definition of the integrable vertex model
The vertex model is constructed from R-matrices acting on tensor products of spaces
carrying the fundamental representation 3 of the graded Lie algebra sl(2/1) and its dual 3.
The R-matrices satisfying YangBaxter equations (YBEs) have been constructed in [39
41] (see also [42]) and read
2
2
R3 3 (v) = 1 P,
R33 (v) = 1 P,
v
v
2
2
R33 (v) = 1 + O,
(4.1)
R33
(v) = 1 + O.
v
v
Here P and O are the graded permutation and monoid operators of sl(2/1), respectively
(note that the action of the monoid O on 3 3 differs from that on 3 3). As a consequence
of O2 = O we have R33
(u 1)R33 (u 1) = 1.
533
From these R-matrices we construct two families of monodromy matrices (labeled by

the spectral parameter v). The monodromy matrices act on the tensor product of a quan L (the Hilbert space of the superspin chains considered below), and a
tum space (3 3)
three-dimensional matrix space carrying a representation 3 (3)

(1)
(1)
(2)
(2)
(L)
(L)
T3 (v) R33 (v)R33 (v 1 )R33 (v)R33 (v 1 ) R33 (v)R33 (v 1 ),

(1)
(1)
(2)
(2)
(L)
(L)
T3 (v) R33
(v 1 + )R3 3 (v)R33
(v 1 + )R3 3 (v) R33
(v 1 + )R3 3 (v).
(4.2)
Here the parameter can be chosen freely without affecting the essential properties (symmetries and locality of the interactions) of the models considered in the following. Variation
of allows one to tune the coupling between the two sublattices of 3 and 3 spins, respectively. Later we will focus on the case of = 0 which corresponds to the most symmetric
coupling. As a consequence of the YBE
R33 (u v) T3 (u) T3 (v 1 ) = T3 (v 1 ) T3 (u) R33 (u v)
the corresponding transfer matrices

(v) = str0 T (v) , = 3, 3,
(4.3)
(4.4)
commute with each other for arbitrary values of the spectral parameter v. Here str0 denotes
the supertrace over the matrix space.
By staggering the transfer matrices (4.4) one can construct the integrable vertex model
shown in Fig. 3. Arrows pointing upwards or to the right (downwards or to the left) indicate
representation of sl(2/1).
that the corresponding link carries the fundamental 3 (the dual 3)
The vertex weights corresponding to the various arrow configurations are shown in Fig. 4.
Fig. 3. The vertex model.
Fig. 4. The vertex weights. Arrows pointing upwards or to the right (downwards or to the left) indicate that the
representation of sl(2/1).
corresponding link carries the fundamental 3 (the dual 3)
534
The transfer matrix of this model includes two layers in the vertical direction and is equal
to (v) = 3 (v)3 (v).
The vertex model constructed in this way was suggested as a description of the spin
quantum Hall effect in Ref. [41]. The notations of [41] are obtained by setting v = 2x
and = 2u 1, where x and u are quantities defined in [41].
The Hamiltonian of an integrable supersymmetric spin chain with local (nearest and
next-nearest neighbor) interactions only is obtained by taking the logarithmic derivative of
the two-layer transfer matrix at v = 0 4

3 (v)3 (v)
H = v ln
,
a(v)2L
v=0
(4.5)
where a(v) = 1 2/v.5 The Hamiltonian (4.5) is not hermitian and its eigenvalues are
in general complex. However, for small systems we find that the ground state and lowlying excited states have real eigenvalues. We assume that this continues to hold true in
the thermodynamic limit. Under this assumption we find that in the thermodynamic limit
low lying excited states can be understood in terms of a conformal field theory and we
can extract the central charge and scaling dimensions from the finite-size energies for the
ground state and low-lying excited states.
The
Hamiltonian (4.5) contains a piece proportional to the sl(2/1) invariant product
[43] j Sj . Sj +1 where Sj are sl(2/1) generators on site j , and the contraction defines
the invariant form. In addition to this Heisenberg coupling however, our Hamiltonian (4.5)
contains more complicated terms involving three neighboring sites.
The eigenvalues and eigenvectors of the transfer matrices were constructed by Links
and Foerster in [39] by means of he graded quantum inverse scattering method [44,45].
As usual different choices of reference state lead to different forms for the Bethe ansatz
equations [4554]. In the following we will analyze the Bethe ansatz equations for the
choice of grading [010] which read

uj + i
uj i
L
=
+ i + i
+ i i
M

uj + i
,
uj i
j = 1, . . . , N,
=1
L
=
N

uk + i
,
uk i
= 1, . . . , M.
(4.6)
k=1
Eqs. (4.6) coincide in fact with the BAE describing the spectrum of the so-called quantum
transfer matrix of the 1D supersymmetric tJ modelsee Appendix D.
Each solution of (4.6) parametrizes an eigenstate with quantum numbers B = (N
M)/2 and J 3 = L (M + N )/2 (we use the notation B B0 + B 0 and J 3 = J03 + J03 ).
4 A similar Hamiltonian was considered in [39] from the point of view of introducing impurities into a t J
model.
5 As usual the overall scale of the Hamiltonian is at our disposal.
535
The eigenvalues of the transfer matrices (4.4) are [39]

N
M

L

L
(v) = a(v)
a(1 + iuk v) + a(1 + v)
a(v i )
k=1
N

a(1 + iuk v)
k=1
=1
M

a(v i ),
=1
M
N

L

L
(v)
= a(v)
a(i + 1 v) + a(1 v)
a(v + iuk )
=1
M

k=1
a(i + 1 v)
=1
N

a(v + iuk ).
(4.7)
k=1
The eigenvalues of the Hamiltonian (4.5) are obtained by taking the appropriate logarithmic derivatives of the eigenvalues of the transfer matrix

(v)(v)

E {uk }, { } = v ln

2L
a(v)
v=0
=

2
2
.
2
u + 1 =1 ( + i)2 + 1
k=1 k
M
Similarly the eigenvalues of the momentum operator are obtained are

i
(v)(v)

P {uk }, { } = ln
2
a(v)2L v=0

N
M

+ i + i
uk + i
+i
.
=i
ln
ln
uk i
+ i i
k=1
(4.8)
(4.9)
=1
In the definition of the momentum we have taken into account that the underlying translational symmetry is by two sites of the lattice.
4.2. The ground state of the lattice model
In order to gain some insight in the structure of the spectrum of the Hamiltonian (4.5) we
have diagonalized the transfer matrix and the Hamiltonian explicitly on small lattices. For
L = 1 and 2 (i.e., 2 and 4 sites) this was done analytically and for L = 3, 4, 5 (corresponding to 6,8,10 sites, respectively) numerically. The results of this analysis are summarized
in Appendix B.2.
Perhaps the most striking result of this analysis is that the space of true singlets (i.e.,
states annihilated by all the generators of the algebra, which are not the image of any other
state under the action of one of these generators) appears to be one-dimensional. This is
not entirely obvious: although the superdimension of the tensor product is always equal to
one, it is a priori conceivable that there are more singlets, and that the superdimension one
536
results from cancellations. But the result can be established rigorously (see Appendix A).
Note however that one can always combine an invariant (i.e., annihilated by all generators)
state that is not the image of any other state with another invariant state, and still get a
state that is not the image of any other state. In that respect, although the singlet appears
only once in the decomposition of the spectrum into representations of sl(2/1), the exact
formula for this singlet depends on the Hamiltonian.
Based on our results for small systems (see Appendix B.2) we conjecture the following
form for the eigenvalues of the transfer matrices 3 (v), 3 (v) of this singlet state

(v 2)(v + 1 ) L
(v 2)(v + 1 + ) L
,
1 (v) =
.
1 (v) =
(4.10)
v(v 1 )
v(v 1 + )
Interestingly, there exists no nondegenerate solution of the BAE (4.6) giving rise to these
eigenvalues. Applying a twist in the boundary conditions for small systems and studying
the evolution of BA roots as this twist goes to zero we were able to verify, however, that the
singlet is in fact described by the degenerate solution uk i, 0, k, = 1, . . . , L,
which clearly gives 1 (v) from the general form (4.7). This finding coincides with the observation in [55] regarding the distribution of BA-roots for the ground state of the quantum
transfer matrix for the tJ model in the thermodynamic limit. The eigenvalue of the full
transfer matrix corresponding to the singlet state is

(v 2)2 (v + 1 )(v + 1 + ) L
.
(v) = 1 (v)1 (v) =
(4.11)
v 2 (v 1 )(v 1 + )
For = 0 this state is the ground state of the Hamiltonian with energy
E0 = 4L.
(4.12)
The energy of this singlet is proportional to L without any finite size corrections which
implies that the central charge of the conformal field theory arising in the scaling limit of
the lattice model is zero
c = 0.
(4.13)
We now turn to a discussion of Bethe ansatz results on the low-lying excitations of the
lattice system. We present a rather short discussion here, and refer to Appendix B for the
details.
4.3. Excitations
Based on an analysis of the analytical properties of the bare scattering phase shifts appearing on the right-hand sides of the Bethe ansatz equations (4.6) we are able to classify
the possible root configurations, so-called strings, arising in their solution (see Appendix B.1). We identify the root configurations giving rise to low energy excited states by
considering small systems. We find that many low-lying states are built from two types
of complexes involving one spectral parameter u and one spectral parameter each (the
n = 1 strange string configurations, see (B.3) and (B.4))
type : u() = x
i
+ ,
2
() = x +
+
,
2
537
i
i
+ + ,
(+) = x + +
.
(4.14)
2
2
A particular class of low-lying excited states is obtained by combining N+ strange
strings of type + with N strange strings of type
()
i
()
, j = 1, . . . , N ,
uj = xj + j, = j
2
(+)
i
(+)
, k = 1, . . . , N+ .
uk = yk + + k,+ = k
(4.15)
2
Our numerical results indicate that the corrections to the ideal string solutions
are of order 1/L. Each of the two complexes (4.14) constitutes an acceptable, non-selfconjugate, solution of the Bethe ansatz equations (4.6). Such solutions have been encountered in other studies of nonunitary Hamiltonians, like the sl(3) integrable spin chain [56].
To make contact to the analysis of the WZW model we have to determine the spectrum
of scaling dimensions for the field theory describing the scaling limit of the lattice model.
Conformal invariance implies that the dimensions can be extracted from a finite size scaling analysis of the low-lying energies of the latter. Before evaluating these energies in
large, finite systems it is very useful to consider an infinite volume first. In the thermodynamic limit we can neglect the deviations from the ideal string solutions as they do not
contribute to the leading O(1) behavior of the energies. Then, using (4.15) in (4.6) and
multiply the equations for the components of a strange string we arrive at the following set
of equations involving only the real centres of the strange strings

N

N+

xj + 3i2 xj + 2i L
xj xk + 2i xj yk + i 2
=
,
xj xk 2i
xj yk i
xj 3i2 xj 2i
k=1
k=1
type +: u(+) = x +
j = 1, . . . , N+ ,
yj +
yj
3i
2
3i
2
yj +
yj
i
2
i
2
L
=
N

N+

yj xk + i 2 yj yk + 2i
,
yj xk i
yj yk 2i
k=1
k=1
j = 1, . . . , N .
(4.16)
The algebraic equations (4.16) can be turned into coupled integral equations for root densities in the limit L , N , NL = n fixed (see Appendix B.3)
h
+ (x) + +
(x) + a4 + |x + 2a2 |x = a1 (x) + a3 (x),
h
(y) + a4 |y + 2a2 + |y = a1 (y) + a3 (y).
(y) +
Here (x) are the root densities of the two types of strings (4.15),
ing densities of holes in these distributions,
2n
1
,
2
4u + n2
and denotes a convolution, i.e.,
an (u) =

an f |x =
dx an (x x )f (x ).
(4.17)
h
are the correspond-
(4.18)
(4.19)
538
The energy per site of such states in the thermodynamic limit is obtained from (4.8)

1
lim EN+ ,N = 2
(4.20)
dx (x) a1 (u) + a3 (u) .
L L
=
The eigenstates of lowest energy within this class are obtained by filling the states with
negative energy. Such states correspond to distributions with (x) = 0 (h (x) = 0) for
|x| > A (|x| < A ). They are characterized by the total density of type- strange
strings
A
N
=
L
dx (x).
(4.21)
For the low-lying multiplets identified for small systems in Appendix B.2.4 we have N+ +
N L. Due to the symmetry of the equations we can restrict our attention to the case
N+ N . Within the description in terms of string densities these states correspond to
= 1 independent
the choice of A+ = : it is straightforward to see in this case that N+ +N
L
of A .
We can now proceed and apply standard Bethe ansatz methods to extract the finite size
energy gaps E of these low-lying states with respect to the completely filled Fermi seas
of type- strange strings (corresponding to A = ). Note that to extract the scaling
dimensions from this spectrum according to the finite size scaling predictions of conformal
field theory the gaps have to be measured with respect to the true ground state discussed
in Section 4.2. This offset between the lowest excitation considered in this section and the
ground state energy E0 is not accessible to the analytical methods applied below and will
be determined by numerical solution of the Bethe equations.
4.3.1. Excitations with N+ = N
The particular subset of solutions discussed above with N+ = N needs to discussed
separately. Let us first consider the thermodynamic limit. Within the integral equation
approach the lowest energy state of this type is described by A+ = A A and the
densities of + type and type strings are found to coincide. Introducing + (x) =
(x) (x) one is left with a single integral equation instead of (4.17). For states with
N+ = N L/2 such that
N 1
=
L L
2
this integral equation reads
lim

(x) +

dx 2 a2 (x x ) + a4 (x x ) (x ) = a1 (x) + a3 (x),
(4.22)
(4.23)
which is solved by Fourier transformation giving

(x) =
1
.
2 cosh x
(4.24)
539
Substituting the solution of the integral equation into (4.20) we then find that the energy
for states with NL = 12
EN+ ,N
= 4.
L
L
lim
(4.25)
All states (of the type we are considering) with NL+ = NL 12 are thus degenerate with the
singlet ground state in the thermodynamic limit.
Let us now turn to solutions of the Bethe ansatz equations in the finite volume. Solving these equations for finite systems we observe that an important role is played by root
configurations in which each strange-string of type + becomes degenerate with a
strange-string configuration of type . After suitable relabeling of the indices this implies yj = xj for j = 1, . . . , N+ in (4.15). In such a configuration the spectral parameters
of the two levels of the Bethe ansatz equations (4.6) coincide
()
uj
(+)
= j
(+)
uj
()
= j
(4.26)
We will analyze such solutions to the Bethe ansatz equations next.

4.4. The TakhtajanBabujian subset
The results of our numerical solution indicate that such degeneratesolutions are a
subset of an even larger class of solutions to the Bethe ansatz equations (4.6) in sectors
with same number of roots on both levels, i.e., N = M. These solutions are obtained by
setting all spectral parameters uj on the first level equal to the spectral parameters on the
second level j , i.e., by requiring that
uj j ,
j = 1, . . . , N.
The roots then have to satisfy

n
uj + i L uj uk + i
=
.
uj i
uj uk i
(4.27)
(4.28)
k=1
This system of equations is identical with the Bethe equations for the spin-1 antiferromagnetic TakhtajanBabujian (TB) chain with L lattice sites and antiperiodic boundary
conditions [57,58]

1
Sj . Sj +1 (Sj . Sj +1 )2 .
4
L
H TB =
(4.29)
j =1
Here Sj are sl(2) spin-1 operators. The degenerate strange string solutions (4.15) with
N+ = N and (4.26) discussed above become the 2-strings u
j = uj i/2 in the
TakhtajanBabujian-classification. We note that the normalization of the Hamiltonian
(4.29) is such that the gaps in our system will be twice the gaps in the TB chain. Alcaraz
and Martins have carried out a finite-size scaling analysis of the anisotropic TB chain with
general twisted boundary conditions [58] and identified the operator content of the underlying conformal field theory. We summarize some of their results in Appendix C.
540
Boundary conditions must be handled with extreme care. This is especially true in theories with super group symmetries: since one deals with graded tensor products, periodic
and antiperiodic boundary conditions for fermions do not necessarily translate into periodic
and antiperiodic boundary conditions for the spin operators in the chain.
4.4.1. Periodic boundary conditions
The sector we are considering here, N = M, is simple in that respect: in that case,
since the total spin J 3 is integer, we are exploring only bosonic states. Therefore, periodic
boundary conditions for the fermions translate into antiperiodic boundary conditions for
the spins of the TakhtajanBabujian chain (for a detailed discussion of boundary conditions
in chains with supergroup symmetries, see [59]).
To proceed, we now recall some results about that chain [58] (see also Appendix C).
The best way to encode the part of the spectrum that comes from the TakhtajanBabujian
chain is to use the generating function of levelsthe conformal partition function [60]. Let
us introduce the quantities
2
1
n + (m + /) ,
8
2
1
nm = n (m + /) .
(4.30)
8
We further introduce the partition functions for the Ising model with twisted boundary
conditions

Ising 2 Ising 2 1 |3 | |4 |
Ising

,
(0, 0) = 0
+ 1/2
=
Z
+
2 ||
||
Ising 2 1 |2 |
Z Ising (0, 1) = Z Ising (1, 0) = 1/16 =
,
2 ||

1 |3 | |4 |
Ising Ising
Ising Ising
Ising
.
Z
(4.31)
(1, 1) = 0 1/2 + 0 1/2 =
2 ||
||
nm =
The generating function of the levels in the TakhtajanBabujian chain is then

TB
Zeven
=
1
1

Ising
Zr,s
r=0 s=0
q nm 1/24 q nm 1/24
(4.32)
m=r+2Z
n=s+2Z
and coincides with the partition function of the SU(2) WZW model at level two [60].
We notice that, in the equivalence between the TakhtajanBabujian spectrum and a subset of our sl(2/1) integrable spin chain spectrum, the TakhtajanBabujian model has an
repreeven number of sites if the sl(2/1) model has an even number L of 3 (and thus 3)
sentations. Expression (4.32) is relevant to this case only. One can also consider the case
of this number being odd. Then the generating function of levels reads
TB
=
Zodd
1
1

r=0 s=0
Ising
Zr,s
q nm 1/24 q nm 1/24 .
(4.33)
m=r+2Z+1,n=s+2Z+1
In the case = , however, expressions (4.32) and (4.33) can be shown to coincide!
541
Let us now extract some conformal weights for the sl(2/1) chain from these expressions.
To do so, one has to
Double the gaps, because the normalizations of the Hamiltonians are such that the
energies of the sl(2/1) chain are twice the energies of the TB chain.
2
Shift the energy differences by a constant 2L
(after the rescaling). This shift in necessary, because the finite-size spectrum in the TB chain with antiperiodic boundary
conditions is calculated with respect to the ground state with L even and periodic
boundary conditions [58]. The energy difference between the true ground state and the
2
TB ground state for even L and periodic boundary conditions is equal to 2L
.
For r, s = 0, we get from (4.32) (that is L even) dimensions in the c = 0 theory which
read
n2 (m + 1)2 1
+
,
2
2
4
h h = n(m + 1), n, m 2Z.
h + h =
(4.34)
Hence we have the possibility n = 0, m = 0, m = 2, giving rise to h = h = 18 . All other

scalar operators have h = h = 18 + integer. The lowest nonscalar operator corresponds to
n = 2, m = 2 with h = 17/8 = 1/8 + 2, h = 1/8.
For r = 1 and s = 0
n2 (m + 1)2
h + h =
+
,
2
2
h h = n(m + 1), n 2Z, m 2Z + 1.
(4.35)
Here all the dimensions are integers. For r = 0 and s = 1,

n2 (m + 1)2
h + h =
+
,
2
2
h h = n(m + 1), n 2Z + 1, m 2Z
(4.36)
with again only integer dimensions. Finally, the case r = s = 1 leads to

n2 (m + 1)2 3
+
+ ,
2
2
4
h h = n(m + 1) + 1, n 2Z + 1, m 2Z + 1.
h + h =
(4.37)
The case n = 1, m = 1 corresponds to h = 9/8, h = 1/8.

In fact, the generating function of the sl(2/1) dimensions deduced from the TB spectrum
can be written after a few manipulations as the sum of two simple contributions
TB subset
Z1/8
= 2q
1/8

1 + q n 1 + q 2n (q q)
n=1
(4.38)
542
and
Z0TB subset =

1 + q n+1 1 + q 2n+1 (q q).
(4.39)
n=0
Of course these two contributions bear resemblance to the two Ramond characters discussed in the first section. However, it is important to understand that we cannot deduce
from these the degeneracies of the sl(2/1) chain. Rather, all we can deduce form the argument is that the dimensions appearing in the expansion of these two objects appear also
in the sl(2/1) spectrum, with a multiplicity at least equal to the multiplicity it has in this
expansion. We denote such a relationship by
sl(2/1) chain
TB subset
Z1/8
Z1/8
sl(2/1) chain
Z0TB subset Z0
.
(4.40)
It can be checked that this is compatible with the spectrum of the chain containing at least
the modulus square of the integrable characters, that is
R,I 2
TB subset
,
Z1/8
h=1/8

R,I V 2
Z0TB subset h=0
(4.41)
.
We can make some more detailed guess about which states on right are represented
TB subset contains two fields with
by a term on the left in (4.40), (4.41). For instance, Z1/8
dimension h = h = 1/8, and we know that these fields must have B = 0. These fields must
also have vanishing SU(2) spin, since they are obtained by taking the n number of the
TakhtajanBabujian chain equal to zero. This means that their SU(2) spin in the sl(2/1)
chain must also vanish, and thus, in the continuum limit, they must be the symmetric and
antisymmetric combinations of J03 = 1/2, J03 = 1/2 and J03 = 1/2, J03 = 1/2. The
fields with h = 1/8 and J03 = J03 = 0 are, in particular, absent in the subspectrum obtained
from the TakhtajanBabujian chain. But we know they are there in the full spectrum.
Similarly, Z0TB subset contains a single field with h = h = 1; once again it must have
B = J 3 = 0. On the right-hand side, we have 8 8 fields with such weights, obtained by
tensoring the adjoint with itself. It is highly likely that the field on the left comes from the
part of this tensor product where the J03 = 0, B0 = 0 field for right movers is tensored with
the J03 = 0, B 0 = 0 field for left movers.
irreIn Tables 1 and 2 we summarize the lowest scaling dimensions (that is, x = h + h,
spective of whether h = h or not) in the TB subset and their sl(2/1) quantum numbers for
Table 1
Smallest scaling dimensions and sl(2/1) quantum numbers in the TB subset for L odd
(B, S) = (0, 0)
(B, S) = (0, 1)
(B, S) = (0, 2)
(B, S) = (0, 3)
m = 1
1, 9
4 4
17 , 25
4 4
m=0
5, 5
4 4
13 , 13
4 4
m=1
9 , 17
4 4
25 , 33
4 4
543
Table 2
Smallest scaling dimensions and sl(2/1) quantum numbers in the TB subset for L even
(B, S) = (0, 0)
(B, S) = (0, 1)
(B, S) = (0, 2)
(B, S) = (0, 3)
m = 1
5, 5
4 4
21 , 21
4 4
m=0
1, 9
4 4
9 , 17
4 4
13 , 13
4 4
29 , 29
4 4
m=1
L even and L odd. We note that these scaling dimensions occur with the same multiplicities
as in the odd L case, but the sl(2/1) quantum numbers are different.
4.4.2. Antiperiodic boundary conditions and the NeveuSchwarz sector
We can also consider the case of antiperiodic boundary conditions for the spin chain
in the sector N = M. This should correspond to antiperiodic boundary conditions for the
fermions, i.e., the NS sector. Note that the sl(2/1) symmetry is then broken. In Appendix B.6 we summarize some results of a numerical solution of the Bethe ansatz equations
with antiperiodic boundary conditions in the sl(2/1) chain for small L.
The same argument we used above in the periodic case shows that the sl(2/1) chain
with antiperiodic boundary conditions has a particular sector that is related to the spectrum of the TakhtajanBabujian chain with periodic boundary conditions, i.e., = 0. This
time, the generating function of the sl(2/1) chain levels can be written as the sum of two
contributions

3
TB subset
1 + q 2n (q q)
= q 1/4
Z1/4
(4.42)
1
and
TB subset
= 2q 1/8 (q q).
Z1/8
(4.43)
Now the correspondence between R and NS weights is as follows

1
J03,R ,
4
1
h NS = h R + + J03,R .
(4.44)
4
chain with weight h = h
R = 1/8 should get, reWe thus see that the two fields in Z1/8
R
spectively, hNS = hNS = 1/8, and hNS = hNS = 7/8 = 1/8 + 1. The field in the 1/8
multiplet whose dimension becomes hNS = h NS = 3/8 = 1/8 + 1/2 is the one with
J03,R = J03,R = 0, and it is not present in the TakhtajanBabujian spectrum. Thus the form
chain is compatible with the SU(2/1) WZW hypothesis, in the NS sector.
of Z1/8
Similarly, the singlet in Z0TB subset has h = J03 = 0 and thus acquires, under spectral flow,
a weight hNS = h NS = 1/4. As for the multiplet with hR = h R = 1, under spectral flow, it
acquires hNS = h NS = 5/4. The field with hR = h R = 1 and J03,R = J03,R = 1/2 which
acquires under spectral flow hNS = h NS = 3/4 = 1/4 + 1/2 is absent from the Takhtajan
Babujian part of the spectrum.
hNS = hR +
544
It is also satisfactory to see that, within the TakhtajanBabujian part of the spectrum,
we have
1
hNS = (n + m)2 =
4
1
h NS = (n m)2 =
4
and therefore
1
(n + m + 1)2
4
1
(n m 1)2
4
1 1
(n + m) hR +
4 2
1 1
+ (n m) hR +
4 2
1
J03,R = (n + m + 1),
2
1
3,R
J0 = (n m 1).
2
1
J03,R ,
4
1
+ J03,R ,
4
(4.45)
(4.46)
Observe in particular that J03,R + J03,R = n, the scalar J 3 quantum number. All these relations are compatible with our WZW conjecture for the continuum limit of the sl(2/1)
chain.
4.5. Strange string excitations with N+ = N
The integral equations (4.17) for strange string excitations with N+ = N can be analyzed by Fourier transformation. This allows one to express the Fourier transforms of
h () in terms of the Fourier transforms of the root densities ().
the hole densities
Inverting the resulting matrix equation, we obtain

1
e||
1 + e2|| h
h
(),
+ () +
2
2 cosh(/2) 4 sinh
2 sinh2
1 + e2|| h
1
e||
h
() =
(4.47)
()
+
().
2 cosh(/2) 4 sinh2
2 sinh2

Here we have defined the Fourier transform by f() = dx eix f (x) and a n () =
en||/2 . The right-hand side of (4.47) is most conveniently written in matrix form using

e||
cosh()
1
.
R()
=
(4.48)
1
cosh()
2 sinh2
+ () =
The matrix R is the Fourier transform of the resolvent R = (1 K)1 of the (matrix)
integral operator in our original equations (4.17) with kernel

a4 (x y) 2a2 (x y)
.
K(x y) =
(4.49)
2a2 (x y) a4 (x y)
Starting from (4.17) (or, equivalently, (4.47)) and (4.20) we can analyze the finite size
spectrum of these excitations by standard methods. Cases similarly to the one considered
here with complete symmetry between the excitations in the + type and type sectors
have been studied in Refs. [61,62]. There the following general formula for the finite size
spectral gaps E of low lying excitations over a filled Fermi sea with several components
in a relativistic invariant model (i.e., with a single Fermi velocity) was derived. In our case
545
it takes the form

L
1
(4.50)
E = (N )T R 1 (0)N + D T R(0)D.
2v
4
Here v = is the characteristic
of the system and is calculated in Appendix B.5,

velocity
+
where
N is the change in the number of roots of
N is a 2-component vector N
N
+
type compared to the ground state. Similarly, D is the 2-component vector D
D where
D is the number of solutions of type backscattered from the left to the right of the
Fermi sea.
In our case there is a slight additional subtlety here coming from the fact that the ground
state is made of complex solutions (4.15): in such a situation it is known thateven in the
usual cases such as sl(2)on top of the Gaussian spectrum, deduced from our general
formula (4.50), the dimensions of other operators may appear, as well as selection rules.
At the same time it has been observed in these cases that most of the basic qualitative
conclusions drawn from an analysis based on the string picture are valid and provide the
basis for a quantitative description when amended for example by results from a numerical
computation of these gaps (see, e.g., [58,63,64]).
Applying (4.50) blindly to our case requires a bit of care as the matrix R has divergent
elements. This rather unusual feature can be traced back to the divergence of the Fourier
transform of the kernel describing the scattering between complexes of type + and
for N+ = N . We can regularize these elements by giving a small value

1 + + 2 2 /3 1 2 /6
1
1 = 2 2 . (4.51)
R()
2
,
R(0)
2 2
2
1 2 /6 1 + + 2 2 /3
From this it follows that
L
1
1
E = (N+ + N )2 + 0 (N+ N )2 + (D+ + D )2
2v
2
8
+ (D+ D )2 .
(4.52)
A divergence of the kernels similar to the one leading to (4.52) here has been encountered
in the study of the osp(2/2) chain (that is, based on the four-dimensional typical representation [0, 1/2]) and the study of the antiferromagnetic Potts model. In both cases, the
divergence of the kernels (in Fourier space) at = 0 has been interpreted [32] as a manifestation of a continuous spectrumwhich is equivalent to the infinite degeneracy of the
finite size gaps between states with N+ + N = const in (4.52).
We emphasize that this result is obtained for O(1) deviations in N compared to the
state with equal numbers of the + and type strings (4.23). In other words, the result
holds for finite values of N , D.
In fact, the state with the largest possible asymmetry between these distributions, i.e.,
N+ = L, N = 0, can be analyzed in the thermodynamic limit giving
1 cosh 2 x
,
(x) 0
+ (x) =
(4.53)
2 cosh x
for the densities. The energy of this state is EN+ ,0 = 2L(1 + ln 2) and the state is clearly
highly excited. In summary, we expect on the basis of the above analysis that there is a
546
macroscopic degeneracy of the finite size spectrum identified in the TakhtajanBabujian

subset.
As we have previously mentioned, these results ought to be taken cum grano salis as
they have been obtained within the framework of the string hypothesis. The latter neglects
the finite size deviations in (4.15), which also contribute to order O(L1 ) to the gaps.
Furthermore, in order to extract conformal dimensions from the finite size spectrum we
need to determine the difference between the ground state energy E0 (see Section 4.2) to
that of the state N = 0 = D to order 1/L which is beyond the scope of the analytical
methods applied here. In light of these complications and in order to verify the observation
from (4.52) that each gap is infinitely degenerated (or equivalently that the spectrum has
a continuous component added to the discrete part identified via the TakhtajanBabujian
mapping), we have carried out an extensive numerical analysis of the Bethe ansatz equations (4.6) for large, finite L. Due to the presence of strong logarithmic corrections (i.e.,
corrections of order 1/(L ln L) to (4.50) we had to study systems with L up to 5000, i.e.,
10 000 lattice sites.
We summarize the results of this finite-size scaling analysis for several low-lying
[0, 1] octet states (N+ + N = L 1) as well as the octet and the indecomposable
[0, 1/2, 1/2, 0] representations (N+ = N = L/2) in the TakhtajanBabujian sector in
Appendix B.4. The scaling of their energies expected from conformal invariance is

L
E(N+ , N ) E0 h + h = x,
(4.54)
L
2v
where x scaling dimension. We plot the left-hand side of Eq. (4.54) as a function of the
lattice length for several different excited states in Fig. 5.
Fig. 5. Corrections to scaling of the energies of the low-lying states studied in Appendix B.4. The dashed lines
indicate the rational function extrapolation L .
547
Fig. 6. Finite-size scaling of low-lying energy levels for the system with antiperiodic boundary conditions. E0 is
the ground state in the sector with periodic boundary conditions (see Section 4.2).
Within rational function extrapolation of our numerical data we find that the energies of
the states considered all become degenerate to order 1/L in the thermodynamic limit. The
scaling dimension obtained from this extrapolation of (4.54) is
1
x= .
4
(4.55)
As has been discussed above, we have to extend the analysis of the finite size gaps to
the case of antiperiodic boundary conditions for a proper identification for the continuum
limit. For this we have studied states build from the strings (4.15) in some detail which
again are the relevant configurations for the lowest excitations. The details of the numerical analysis are given in the appendix, here we just present the main results which give
some evidence that again there are many states degenerate with the ground state (in the
TakhtajanBabujian sector) in the thermodynamic (L ) limit. All of these states have
N+ + N = L, but different values of N = N+ N . Their scaling dimensions extrapolate to x = 1/4 relative to the ground state of the chain with periodic boundary
conditions. The corrections to this scaling for some states are shown in Fig. 6.
4.6. Summary
As a summary, we represent on Fig. 7 the lowest gaps we have observed in the R and
NS sectors. Values in parentheses is are expected from the study of integrable characters
of the WZW model, but we have not been able to investigate them numerically.
548
Fig. 7. The spectra from the lattice model in the R and NS sector, in agreement with the WZW prediction. The
thick lines represent a continuum of critical exponents (extending to infinity). Values in parenthesis were expected
but beyond our reach.
5. Higher level
It is interesting to consider the same problem with larger representationsmore precisely, the fully supersymmetrized atypical representations of type [j, j ] (s = 2j boxes
in the Young diagram representation). In this case, the Bethe equations allow for a sector
reproducing twice the gaps of the spin s TakhtajanBabujian chain. For the latter, the gaps
are obtained once again by combining the gaps from a free boson and disorder operators
from a Z2s parafermionic theory. The gaps from the free boson sector read now
2
1
n + (m + s/) ,
8s
2
1
n (m + s/) .
nm =
8s
nm =
(5.1)
The sl(2/1) symmetric case (R sector) corresponds to = . The leading gaps with respect to the c = 0 theory read then
h + h =
(m + s)2
s
r(2s r)
+
,
2s
2(s + 1) 2s(s + 1)
h = h
(5.2)
with the constraint that m = r mod 2s. The choice m = r 2s gives then
(s r)2
,
h = h =
4(s + 1)
r = 0, . . . , s 1.
(5.3)
This obviously reproduces the conformal weights in the SU(2/1) WZW model for representations [0, j ] with j = 0, 1/2, . . . , s1
2 at level s.
Meanwhile in the Neveu Schwarz sector, ceff =

other formula
h = h =
m2
(m s + r)(m + s r)
+
.
4(s + 1)
4s(s + 1)
6s
s+1 ,
549
and for the weights we get the
(5.4)
The case m = r 2s now gives

(2s r + 1)2 1
h = h =
.
4(s + 1)
4
(5.5)
This can be written as

(s r)2
s s r
h = h =
+ +
4(s + 1) 4
2
(5.6)
matching the formula for the spectral flow in the SU(2/1) level s WZW theory. Hence
central charges and TB subset match the WZW prediction in this case as well, and it is
likely that a more thorough study would find similar agreement for the continuum parts of
the spectrum too.
6. Conclusions and speculations

To conclude, from the information at our disposal using the TakhtajanBabujian part of
the spectrum, analytical study of the Bethe ansatz, and extensive numerical calculations,
we have obtained consistency with the hypothesis that the continuum limit is the SU(2/1)
level one WZW model. This consistency includes the set of all dimensions in the R sector,
and only a subset of dimensions in the NS sector. It is highly indicative that our conjecture
is correct that the effective central charge in the NS sector is c = 3. As for the part of the NS
spectrum whose weights are of the form hNS = 18 + 12 + integers hNS = 14 + 12 + integers,
we have argued they are situated outside of the TakhtajanBabujian part of the spectrum,
and thus we have nothing to say about them based on this chain of arguments. It would of
course be a crucial further test to extract these exponents from the Bethe ansatz solution
(this might be difficult as there are sitting inside the continuum). We note that it would be
extremely useful to compare more thoroughly than we have done the group content of the
lattice and field theory models: this however may be a very nontrivial exercise, in particular
because fine structures of the representations (such as indecomposability) may not evolve
straightforwardly in the continuum limit process [65].
One of the main difficulties we have encountered is the lack of understanding of the
WZW theory itself. In particular, the free field representation as well as algebraic arguments point to the existence of a spectrum of dimensions unbounded from below, and
maybe continuous. While this is somewhat expected in view of the unboundedness of the
action in the field theory, it is not totally clear how this should be related to the spectrum
and partition function of a lattice model. We have suggested that some sort of analytical
continuation of the characters is at work, folding back the spectrum of arbitrarily large negative dimensions into a continuum of positive dimensions. This is strongly supported by
calculations involving modular transformations of characters, and gives rise to predictions
550
in remarkable agreement with lattice calculations. A point that remains unclear is whether
this continuation is a formal operation, or has the meaning of changing the target space to a
Riemannian symmetric superspace as argued in [34]. Clearly more work is needed to clarify these questions, which we believe are crucial to our understanding of CFT description
of the transition between plateaux.
Finally, we come back to our original question concerning the universality of 3 33

3 sl(2/1) spin chains. It is easy to check that the analysis in [17] for the Heisenberg
model (with S . S nearest neighbor coupling only) in the R sector, gives a totally different
spectrum. First, multiplicities are all finite. Then, the following scalar operators
(3P + 1)2 1
4M 2 1
(6.1)
,
h = h =
24
24
appear, as well as many other nonscalar operators whose chiral weights generically read
(modulo integers)
h = h =
(3P /N + 2M)2 1
(6.2)
, N |M.
24
Except for the lowest weights, there is no indication that any of these weights appear in
the spectrum of the integrable sl(2/1) chain. Moreover, imposing antiperiodic boundary
conditions for the fermions (i.e., the NS sector) gives results which are totally different
from the ones of the integrable spin chain; in particular,
instead of being three, the effective
central charge is an irrational number c = 1 + 92 ln2 3+2 5 1.84! In fact, pretty much the
only features common to the two systems are the vanishing of the central charge in the R
sector, and the lowest gap h = h = 18 extremely ubiquitous features of superalgebra field
theories.
Since the integrable model requires fine tuning of nearest neighbor and next to nearest
neighbor couplings, we expect that it describes the least stable of the two fixed points, as
evidenced by the higher value of its effective central charge in the NS sector.
It seems desirable to investigate the complete phase diagram of the model, a task that
could be made easier by using the loop representation. We hope to get back to this question
later. A property that could prove useful in this analysis is that the ground state of the
Heisenberg chain, like the ground state of the integrable spin chain, is a true singlet, with
trivial value of the energy. It does not seem to be the same state however.
h=
Acknowledgements
We are grateful to G. Goetz, V. Gurarie, A. Klmper, J. Links, A.W.W. Ludwig, M. Martins, N. Read, V. Schomerus, M. Shiroishi, A. Taormina, J. Teschner and A.M. Tsvelik
for helpful discussions. H.S. is especially thankful to J. Germoni for his wonderful crash
courses on super algebra representation theory. F.H.L.E. thanks the Department of Mathematics at the University of Brisbane, where part of this work was carried out, for hospitality.
This work was supported in part by the Deutsche Forschungsgemeinschaft under grants
Fr 737/2 and Fr 737/3 (H.F.), the DOE (H.S.), the EPSRC under grant GR/R83712/01
(F.E.), the Humboldt Foundation (H.S.), and the network EUCLID (TMR, network contact
HPRN-CT-2002-00325).
551
Appendix A. Some algebraic considerations

L : for
Numerical study indicates that there is a unique singlet in the product (3 3)
2
decomposes as
instance, the tensor product (3 3)
2 = [0, 0] + 4[0, 1] + [0, 2] + [1/2, 3/2]
(3 3)
+ [1/2, 3/2] + [0, 1/2, 1/2, 0].
(A.1)
One has to be careful in defining what one means by singlet in this problem, as there are
invariant states (that is, states annihilated by all the generators of the algebra) within some
of the indecomposable representations too. We will call true singlet a representation isomorphic to the trivial representation, and which is not a quotient of a larger indecomposable
L , it appears as an invariant state which does
representation. When decomposing (3 3)
not lie in the image of the raising or lowering generators of the algebra.
It is known from representation theory of sl(2/1) [25] that in the tensor product we are
interested in, there can only appear typical representations, simple (irreducible) atypical
of the type [j, j ] and nonsimple (indecomposable) atypical, representations. The latter
have vanishing superdimension, as do the typical representations. It is obvious that the
superdimension of our tensor product is one but one could imagine that this one does not
come from the existence of a single [0, 0] representation. For instance, it could arise as
the result of a cancellation between several [0, 0] reps (each of superdimension one, as it
is easy to see that all [0, 0] states have to be bosonic in our system) and several simple
atypical representations with negative superdimension. It is easy to exclude this possibility
if one recognizes that the typical representations as well as the indecomposable atypical
appearing in our tensor product are projective, while simple (irreducible) atypical are not
(recall here [29] that a representation is projective if it cannot appear as a quotient of a
bigger, indecomposable block). In particular, [0, 0] is not projective. A well-known result
of algebra [29,66,67] says that
projective anything = projective.
(A.2)
But consider 3 3 = [0, 0] + [0, 1]. The adjoint appearing here is typical and thus projective. Hence, tensor products [0, 1]L decompose onto projectives only, that is typical
representations and nonsimple atypical. No simple atypical ever can appear in these tensor
products. Thus, [0, 0] can appear only once, as the result of the tensor product of [0, 0]
with itself, which is our initial claim.
Note that the result could also be proven by using the fact that the tensor product of any
indecomposable representation (either irreducible like a simple typical, or reducible like a
nonsimple atypical) of vanishing superdimension decomposes as a sum of representations
each with vanishing superdimension themselves [68].
Notice that a similar result is known to take place in the SU(2)q product of an even number of fundamental (two-dimensional) representations when q 3 = 1, so the spin j = 1
representation has vanishing q-dimension. In this case, there is a unique singlet representation, which coincides with the space Ker S + / Im S + [69,70].
2 , there are two invariant (that is,
Of course, if we get back for instance to (3 3)
annihilated by all the generators in the algebra) vectors, one in [0, 0] and one in the indecomposable. The vector in [0, 0] is a true singlet, which means simply that it is not
552
in the image of the generators. Obviously, a generic combination of the two invariant
vectors will still have this property. This means that the way [0, 0] will be expressed in
terms of the basis vectors in 3 and 3 will in fact depend on the Hamiltonian (one could
imagine selecting the good true singlet by requiring that it be orthonormal to the image of the generators in the subspace of vanishing U (1) charge. However this will not
work, as the invariant state in the indecomposable has zero norm square). Remarkable expressions for such states have been obtained recently (albeit in a different language) in
the context of the RazumovStroganov conjecture [71]. We leave this point to further
study.
Appendix B. Details of the analysis of the lattice model

B.1. String hypothesis
The standard way to classify the solutions of (4.6) is to consider configurations of spectral parameters that sit on poles of the bare scattering phase shifts (the right-hand sides in
(4.6)). A simple calculation yields the following types of strings
(1) Reals
Unpaired, purely real spectral parameters uj and .
(2) Wide strings (see Fig. 8)
Type-1: composites containing n 1 s and n us (n > 1)
(n,1)
+ i(n + 1 2k),
u,k = u(n,1)
(n,1)
,j
= u(n,1)
+ i(n 2j ),
k = 1, . . . , n,
j = 1, . . . , n 1, u(n,1)
R.
(B.1)
Type-2: composites containing n s and n 1 us (n > 1)

(n,2)
,k = (n,2) + i(n + 1 2k),

(n,2)
u,j
= (n,2)
+ i(n 2j ),
k = 1, . . . , n,
j = 1, . . . , n 1, (n,2) R.
(B.2)
Fig. 8. Wide strings of lengths three and two, respectively. The circles/triangles denote the positions of the us/ s
involved in the string.
553
Fig. 9. (a) The two types of n = 1 strange strings; (b) A + type strange string with n = 2.
(3) Strange strings (see Fig. 9)

Composites containing n s and n us (n 1)

1
(n,)
(n)
u,k = u + i n + 1 2k
, k = 1, . . . , n,
2

1
(n,)
= u(n)
i
n
+
1
2k
, u R,
,k
2
or

1
,
= u(n)
i
n
+
1
2k
2

1
(n,+)
,
,k = u(n)
+
i
n
+
1
2k
2
(n,+)
u,k
(B.3)
k = 1, . . . , n,
u R.
(B.4)
We note that solutions of this type of solution is quite different from usual string solutions
in that the set of roots on one level of the Bethe equations is not invariant under complex
conjugation. The other solutions discussed above (reals and wide strings) are invariant
under this operation. This is somewhat similar to what was found recently in [56]. It is important to note that although strange strings are not invariant under complex conjugation,
the corresponding energy (4.8) in the case = 0 is still real.
(4) Narrow strings
Composites containing n s and n us (n > 1)
i
(n,n)
u,k = u(n,n)
+ (n + 1 2k), k = 1, . . . , n,
2
i
(n,n)
(n,n)
R.
,j = u + (n + 1 2j ), j = 1, . . . , n, u(n,n)
(B.5)
Narrow strings may be thought of as special cases of strange strings or wide strings in the
following sense.
Combining a +-strange string of length n and centre u(n) with a -strange string
of length n and centre u(n) we obtain a narrow string of even length 2n. This is shown
for n = 1 in Fig. 10(a).
554
Fig. 10. (a) Combining a pair of + and strange strings of length 1 gives a narrow string of length 2;
(b) Combining a type-1 wide string of length 2 with a type-2 wide string of length 2 gives a narrow string of
length 3.
Combining a type-1 wide string of length n and centre u(n,1) with a type-2 wide string
of length n and centre u(n,1) we obtain a narrow string of length 2n 1. This is shown
for the case n = 2 in Fig. 10(b).
It is clear from our discussion that narrow strings are not fundamental but are merely
degenerate cases of strange string solutions of the Bethe ansatz equations.
We have verified explicitly by numerical solution of the Bethe ansatz equation (4.6),
mainly for the case = 0, that solutions of the above types exist at least for small stringlengths n. On the other hand it is entirely possible that the above classification is not
complete.
However, it is important to note that if
(a) our classification is complete and
(b) in the thermodynamic limit solutions of the Bethe ansatz equations (4.6) with the
same values of J 3 per site and B per site as the ground state tend towards ideal string
solutions (as is often the case in Bethe ansatz solvable models),
then all energies are real for = 0. Given the string hypothesis (B.1)(B.4) the usual way
of proceeding in a Bethe ansatz solvable model is to construct the thermodynamic Bethe
ansatz (TBA) equations at finite temperature and then use them to determine the distribution of roots in the zero temperature ground state. This applicability of this procedure to
our case is questionable as it is not clear whether the minimization of the free energy functional is a meaningful procedure for the case of a non-hermitian Hamiltonian. Disregarding
this issue one finds that the standard integer method of classifying solutions of the Bethe
ansatz equations does not seem to work. In particular, we find that in some cases the numbers of solutions of a particular type for fixed L, N and M are found to differ from what
the integer method predicts. We do not have a good explanation for this fact at present and
leave this issue for future studies. Irrespective of this, the Bethe ansatz does not provide a
complete set of states in any case. This follows from the peculiar representation theory of
superalgebras. It was shown in [39] that for the model at hand all eigenstates obtained from
the Bethe ansatz are highest weight states of sl(2/1). For integrable models exhibiting a
555
classical Lie-algebra symmetry a complete set of states can be obtained from these highestweight states by acting with all possible combinations of lowering operators [72]. In the
present case there are states that belong to reducible but indecomposable representations,
for which the highest weight state is not a cyclic one [25]. As a consequence, acting on
the Bethe ansatz states with lowering operators is not sufficient to obtain a complete set of
states. This complication is absent in other integrable, supersymmetric models, for which
complete sets of states have been obtained by the aforementioned procedure [50,7376].
Given these complications we will follow a different, less ambitious approach: by studying finite systems we will try to identify the ground state and some low-lying excited
states of the Hamiltonian (4.5) in terms of distributions of roots of the Bethe ansatz equations (4.6).
B.2. Small finite systems
For L = 1, 2 the transfer matrix 3 (v) can be diagonalized directly.
B.2.1. L = 1
The tensor product 3 3 is fully reducible and decomposes into a singlet [0] and the
typical octet [b, j ] = [0, 1]
3 3 = [0, 0] + [0, 1].
(B.6)
The corresponding eigenvalues of 3 (v) are readily obtained

2
2
+
,
v v1
2
1 2 1
+
.
1 (v) = 1
+1v +1v1
Similarly we find for the eigenvalues of 3 (v)
8 (v) = 1
(B.7)
2
2
+
,
v v1+
2
+1 2 +1
1 (v) = 1
(B.8)
+
.
1v 1v1+
Obviously, 8 (v), 8 (v) are the eigenvalues on the pseudo vacuum (M = 0 = N ) while
1 (v), 1 (v) can be obtained from (4.7) in the sector with (M, N ) = (1, 1) for u1 = ,
1 = 0 which clearly solve the BAE (4.6). We note that the singlet eigenvalue satisfies the
functional equation

(v + 1 )(v 3 ) L
1 (v)1 (v 1 ) =
(B.9)
.
(v 1 )2
8 (v) = 1
Furthermore, for = 0, we find 1 (v)|=0 = (v 2)(v + 1)/v(v 1) = 1 (v)|=0 . The

eigenvalues of the Hamiltonian are
E1 =
4
,
1
E8 =
2
.
1
(B.10)
556
B.2.2. L = 2
The tensor product of representations appearing in the L = 2 (4 site) system can be
decomposed as
2 = [0, 0] + 4[0, 1] + [0, 2] + [1/2, 3/2] + [1/2, 3/2] + [0, 1/2, 1/2, 0].
(3 3)
(B.11)
Here, the atypical representation [0, 1/2, 1/2, 0] is an eight-dimensional reducible but
indecomposable representation which is the semi direct sum of [1/2, 1/2], [1/2, 1/2],
and two [0, 0] representations [25]) (see Section 2 of the main text). Note that this block is
the same indecomposable block as the one which appears in the product [0, 1/2] [0, 1/2].
Diagonalization of the transfer matrix yields the eigenvalues of the 4-site system. In
Table 3 we present their values together with the labels of the sectors (M, N ) in which they
appear in the BA and the corresponding irreducible representation of the algebra sl(2/1)
in the decomposition (B.11).
Table 3
Spectrum of the transfer matrices for L = 2
[B, S]
Eigenvalue of 3 (u), 3 (u)
Bethe roots
(0, 0)
[0, 2]
vacuum
(0,1)
[+ 12 , 32 ]
4
4
16 (v) = 1 v4 + 42 + v1
+
v
(v1)2
4
4
16 (v) = 1 v4 + 42 + v1+
+
v
(v1+)2
(+)
4
4
12 (v) = 1 v4 42 + v1
+
v
(v1)2
(+)
4
4
12 (v) = 1 v4 + 42 + v1+
v
(v1+)2
(1, 0)
[ 12 , 32 ]
()
4
4
12 (v) = 1 v4 + 42 + v1
v
(v1)2
()
4
4
12 (v) = 1 v4 42 + v1+
+
2
1 = i
(1, 1)
[0, 1]2
(M, N )
(1,2)
(v1+)
u1 = i[
(, v)
(1,2)
8 (, v)
(1, 1)
[0, 1]2
(3,4)
(, v)
(1, 2)
(2, 1)
(2, 2)
(2, 2)
[0, 12 , 12 , 0]
[0, 12 , 12 , 0]
[0]
4
x (v) = 1 v4 42 + v1
v
4
x (v) = 1 v4 42 + v1+
v
4
x (v) = 1 v4 42 + v1
v
4
x (v) = 1 v4 42 + v1+
v
4
x (v) = 1 v4 42 + v1
v
2 1]
1 = u1 + i
2
u1 = i[ 3 +3 ]
(3,4)
8 (, v)
[0, 12 , 12 , 0]
u1 = 0
1 = i u1
4
(v1)2
4
(v1+)2
4
(v1)2
4
(v1+)2
4
(v1)2
4
4
x (v) = 1 v4 42 + v1+
v
(v1+)2
2

1 (v) = (v2)(v+1)
v(v1)
(v2)(v+1+) 2
1 (v) =
v(v1+)
u1,2 = i[

1 + 2 ]
1 = i
u1 = 0

1,2 = i 1 + 2
1
1 = i + +i
2
u1,2 =
1 1( 1 + 1 )2
1
2
1
1
+
1
557
Here the eigenvalues corresponding to the four (typical) octets [0, 1] are

4
1 4
4
(1,2)
8 (, v) = 1
+
2
1+v
+1v
1+v1

4
1
,
+ 1 (v 1 )2

4
+1 4
4
(1,2)
+
8 (, v) = 1
2
1v
1v
1v1+

4
+1
,
1 (v 1 + )2
2 + + 4 2 2 + 3 4 2 2 + 3 4
(3,4)
8 (, v) = 1
v
1+
(1 + )2
v2
2
2 2 + 3
4
4
+ + 4 2 2 + 3
+
,
2
v1
1+
(1 + )
(v 1 )2
(3,4)
(3,4)
8 (, v) = 8 (, v).
(B.12)
Let us consider the [0, 12 , 12 , 0] multiplet in more detail. There are altogether three
highest-weight states, i.e., states annihilated by the sl(2/1) raising generators. Their quantum numbers are B = 12 , J 3 = 12 and B = J 3 = 0, respectively. From the table above
we see that the Bethe ansatz yields three states in this multiplet, which by [39] are highest
weight states. The Bethe ansatz states with (M, N ) = (1, 2) and (M, N ) = (2, 1) are readily identified as B = J 3 = 12 , B = J 3 = 12 highest-weight states. The solution to the
Bethe ansatz equations with (M, N ) = (2, 2), leading to the eigenvalue x , is very peculiar to say the least: it contains a free parameter! For any value of 2 the roots given above
satisfy the BAE (4.6) and give the eigenvalue x (v) of the transfer matrix! We identify this
last highest-weight state with the invariant singlet state of [0, 12 , 12 , 0]. The second singlet
state in the multiplet is a cyclic state, i.e., all states in the representation can be obtained
from it by acting with symmetry generators. However, it appears that it is not possible
to obtain this state from the Bethe ansatz! In other words, the Bethe ansatz is manifestly
incomplete due to the peculiar representation theory of superalgebras.
For the eigenvalue 1 (v) corresponding to the unique regular singlet in the tensor
product there exists no nondegenerate solution of the BAE (4.6). Applying a twist in the
boundary conditions and studying the evolution of BA roots as this twist goes to zero we
were able to verify that the singlet is in fact described by the singular solution uk i,
0 which clearly gives 1 (v) from the general form (4.7). This finding coincides with
the observation in [55] about the distribution of BA-roots for the ground state of the quantum transfer matrix for the tJ model in the thermodynamic limit. For finite L the authors
of [55] find this degeneracy to be lifted which is a consequence of the fact that their system is subject to a phase shift when compared to ours (see Appendix D). Note that this
phase shift also breaks the supersymmetry of the system. Furthermore it is not obvious
that the two states actually are mapped onto one another when the phase shift is changed
adiabatically: the distribution of roots of the function D in [55] seems to be very different.
558
We note further that 1 (v) is just the square of the eigenvalue obtained for the singlet
in the 2-site system (B.7). Hence, 1 (v) is again the solution of the functional equation
(B.9). This observation lead us to formulate the following conjecture for the ground state
eigenvalue of the transfer matrix

(v 2)(v + 1 ) L
1 (v) =
(B.13)
.
v(v 1 )
For the L = 2 system the eigenvalues of the Hamiltonian are (4.8):
8
(1,2)
()
,
Ex = 4,
E8
= 2,
E12 = 2,
E16 = 0,
2 1
28 86 + 564 + 402 + 42 3 + 2 (86 + 284 + 202 + 24)

(3,4)
E8
=
.
(2 1)3 (2 3 + 2 )2
E0 =
(3,4)
For = 0 we find that E8 |=0 = 6, i.e., the eigenvalues are equally spaced integers.
We note that the conjecture (B.13) implies that the ground state E0 = 4L/(2 1) is
strictly extensive without any finite size corrections. For a conformally invariant system this
implies that the central charge of the underlying Virasoro algebra is c = 0 as expected
for the class of models considered here.
B.2.3. L = 3, 4, 5
By numerically diagonalizing the transfer matrix for system sizes up to L = 5 we find
that in general there is exactly one eigenstate with an eigenvalue given by

(v 2)2 (v + 1 )(v + 1 + ) L
.
(v) = 1 (v)1 (v) =
(B.14)
v 2 (v 1 )(v 1 + )
This is in agreement with the conjecture (B.13). For = 0 this state is the ground state of
the Hamiltonian with energy
E0 = 4L.
(B.15)
From now on we set = 0. The first few low-lying multiplets for L = 3 are shown in
Table 4. We see that the first excited state is an octet with energy 11.1231. Then there
are two degenerate octets with energy E = 9.60555. Finally there are two degenerate
[0, 12 , 12 , 0] multiplets. We only list one Bethe ansatz solution for each of these representations. Like in the two-site case L = 2, the invariant singlets of [0, 12 , 12 , 0] multiplets
correspond to one-parameter families of solutions to the Bethe ansatz equations.
In Table 5 we list the low-lying sl(2/1) multiplets for L = 4. The first excited states are
two octets with energy E = 15.0484. Next there is a single [0, 12 , 12 , 0] multiplet with
energy E = 14.1677. Clearly the degeneracies for L = 4 are different from L = 3. This
suggests that there might be a difference in structure of the low-lying excitation spectra for
odd and even L.
Table 6 summarizes the lowlying states for L = 5. The first excited multiplet is an octet,
then there are two degenerate octets, followed by two degenerate [0, 12 , 12 , 0] multiplets.
The structure of degeneracies coincides with the L = 3 case.
559
Table 4
Low-lying multiplets for L = 3
[B, S]
[0, 1]
11.1231
P /2
BA roots
[0, 1]
9.60555
[0, 1]
9.60555
[0, 12 , 12 , 0]
9.00000
[0, 12 , 12 , 0]
9.00000
[0, 1]
6.5
[0, 1]
6.5
[0, 1]
6.5
[0, 1]
6.5
u1 = 1 = 0.530i,
u2 = 2 = 0.530i
u1 = 1 = 0.254 0.608i,
u1 = 1 = 0.254 + 0.608i,
u2 = 2 = 0.254 0.608i
u2 = 2 = 0.254 + 0.608i
1 = 2 = 0.144 0.382i,
u1 = 1.427i
u2 = u3 = 0.163 + 0.713i
1 = 2 = 0.144 + 0.382i,
u1 = 1.427i
u2 = u3 = 0.163 0.713i
u2 = 2 = 1.104 + 0.175i
u1 = 1 = 0.238 0.547i,
u1 = 1 = 0.238 + 0.547i,
u2 = 2 = 1.104 0.175i
u2 = 2 = 1.104 0.175i
u1 = 1 = 0.238 + 0.547i,
u2 = 2 = 1.104 + 0.175i
u1 = 1 = 0.238 0.547i,
[0, 1]
6.23607
u1 = 1 = 0.535 0.309i,
[0, 1]
6.23607
u1 = 1 = 0.535 0.309i,
u2 = 2 = 0.535 + 0.309i
u2 = 2 = 0.535 + 0.309i
0
0
0
0
1/3
1/3
1/3
1/3
1/3
1/3
Table 5
[B, S]
BA roots
[0, 1]
15.0484
u1 = 1 = 0.227 + 0.554i,
u2 = 2 = 0.227 + 0.554i
u1 = 1 = 0.227 0.554i,
u2 = 2 = 0.227 0.554i
u3 = 3 = 0.50948i
[0, 1]
15.0484
u3 = 3 = 0.509i
[0, 12 , 12 , 0]
14.1677
u1 = 1 = 0.051 + 0.521i,
u2 = 2 = 0.871 + 0.700i
u3 = 3 = 0.051 0.521i,
u4 = 4 = 0.871 0.700i
u2 = 2 = 0.426 0.654i
u2 = 2 = 0.426 + 0.654i
[0, 1]
13.0749
u1 = 1 = 0.426 0.654i,
u3 = 3 = 0.596i
[0, 1]
13.0749
u1 = 1 = 0.426 + 0.654i,
u3 = 3 = 0.596i
B.2.4. Root configurations for a class of low-lying states

Our results for small systems indicate that the lowest-lying excitations of the model are
in the octet sector, i.e., belong to [B, S] = [0, 1], or form indecomposable [0, 12 , 12 , 0]
multiplets. Within the Bethe ansatz for a system of length L the octets are parametrized
by L 1 roots uj and while the indecomposables are found for example in the sector
B = 0 = J 3 with L roots on both levels. Within our classification of solutions by means of
560
Table 6
[B, S]
[0, 1]
19.4230
[0, 1]
BA roots
18.66651
u1 = 1 = +0.200 + 0.523i,
u2 = 2 = +0.200 0.523i
u3 = 3 = 0.200 + 0.523i,
u4 = 4 = 0.200 0.523i
u1 = 1 = +0.390 0.593i,
u3 = 3 = 0.559i
u1 = 1 = 0.390 + 0.593i,
u3 = 3 = +0.559i
u2 = 2 = 0.390 0.593i,
[0, 1]
18.66651
u2 = 2 = +0.390 + 0.593i,
[0, 12 , 12 , 0]
18.28188
u2 = 2 = 0.113 0.543i,
u3 = 3 = 0.263 0.551i
18.28188
u4 = 4 = 0.503i
(A)
u1 = 1 = 1.105 + 0.671i,
[0, 12 , 12 , 0]
u4 = 4 = +0.503i
u4 = 4 = 0.835 0.743i
u5 = 5 = 0.02364 + 0.50539i
(A) with uj uj ,
the string hypothesis from Appendix B.1 the corresponding roots form certain sequences
of strange strings of length n = 1 (or their degenerate realization as 22 narrow strings
(B.5)). We denote the number of these + and type strange strings by N+ and N ,
respectively. We constrain ourselves to the case N+ N ; states with N > N+ are obtained by exchanging us and s and are by construction degenerate. We denote the energy
levels of the states by EN+ ,N .
4.4.1 Even L
A low-energy state with B = 0 = J 3 belonging to a [0, 12 , 12 , 0] multiplet is obtained
by choosing N+ = N = L2 and considering a realization of this configuration as a collection of 22 narrow strings. This implies that after a suitable relabeling of roots we
have uj = j , for j = 1, . . . , L. Using this degeneracy the Bethe ansatz equations (4.6)
can be rewritten as

uj + i
uj i
L
=
N

uj uk + i
,
uj uk i
j = 1, . . . , N,
(B.16)
k=1
with N = L. These equations are identical to the Bethe ansatz equations of the SU(2)
symmetric TakhtajanBabujian spin-1 chain [57] for a lattice of L sites and antiperiodic boundary conditions. The operator content of an anisotropic version of this model
(with twisted rather than antiperiodic boundary conditions) was determined by Alcaraz
and Martins [58]. We show in Appendix C how our results relate to theirs.
The lowest lying excitation has N+ = L2 and N = L2 1 and is the highest weight
state of a [0, 1] multiplet. The distribution of strings is symmetric around the imaginary
561
Fig. 11. Strange string solutions of the Bethe ansatz equations (4.6) in the octet sector for a system of size L = 10
with different N = N+ N : circles (triangles) denote uj ( ). (a) N = 1, (b) N = 3, (c) N = 5, (d)
and (e) N = 7, (f) N+ = 9, N = 0.
axis, i.e.,
!
"
{uj | j = 1, . . . , L 1} uj | j = 1, . . . , L 1 ,
(B.17)
and similarly for {j | j = 1, . . . , L 1}. The root distribution for this state for L = 10
is shown in Fig. 11(a).
Other [0, 1] octet states are obtained by taking N+ = L2 + n and N = L2 1 n, n =
1, 2, . . . , i.e., by replacing n type strange strings by + type strange strings. The
corresponding root distributions are shown in Fig. 11(b)(f). We note that the lowest
energy state for given N corresponds to a root distribution that is symmetric around
the imaginary axis. Higher energy states correspond to asymmetric distributions. An
example for L = 10 is shown in Fig. 11(e).
4.4.2 Odd L
The lowest lying state has N+ = N = L1
2 and is the highest weight state of a [0, 1]
multiplet. The sets {uj | j = 1, . . . , L 1} and {j | j = 1, . . . , L 1} are closed under
complex conjugation. Hence, after a suitable relabeling of roots we have j = uj , for
j = 1, . . . , L 1. Similarly as for the B = 0 = J 3 state for even L this state can be
562
considered as a collection of 22 narrow strings and the Bethe ansatz equations can
be cast in the form (B.16) with N = L 1.
A sequence of low energy [0, 1] octet highest weight states is obtained by taking N+ =
L1
L1
2 + n, N = 2 n, n = 1, 2, . . . .
In the following section we carry out a finite-size scaling analysis for the states we have
just described.
B.3. Low-lying excitations in the thermodynamic limit
In this appendix we derive the coupled integral equations (4.17) from the algebraic
equations (4.16). We start by taking the logarithm of (4.16)

(xj ) +
xj
3

=

N+
N
2Ij
1 xj xk
2 xj yk
+
+
,
L
L
4
L
2
k=1
k=1
j = 1, . . . , N+ ,

N+
N
2Jj
yj
2 yj xk
1 yj yk
=
+
+
,
(yj ) +
3
L
L
2
L
4
k=1
k=1
j = 1, . . . , N ,
(B.18)
where (x) = 2 arctan(2x) and Ij (Jj ) are integer or half integer numbers characterizing
the state uniquely. Following Yang and Yang we introduce counting functions z (v) by

N+
N
1 x xk
2 x yk
x
z+ (x) = (x) +
3
L
4
L
2
k=1
k=1
k=1
k=1

N
N+
1 y yk
2 y xk
y
.
z (x) = (y) +
3
L
4
L
2
(B.19)
By definition, the counting function evaluated at a root of Eqs. (B.18) yields 2 times the
corresponding integer numbers divided by L, e.g., z+ (xj ) = 2Ij /L. Taking the derivative
of (B.19) and then L we obtain the integral equations (4.17) for the densities of roots
h by using that
and holes
dz (x)
h
.
(x) + (x) =
2
(B.20)
dx
To proceed further, we restrict ourselves to states with symmetric distributions of the integers, e.g.,
1
Ij = (N+ 1) + j, j = 1, . . . , N+
2
and similarly for the Jk . For these distributions we have
h
(x) = 0,
for |x| < A ,
(B.21)
(B.22)
563
where A are fixed by the conditions z (A ) = N /L. The integral equations for the root
densities may then be simplified to
A+
+ (x) +
A
dx a4 (x x )+ (x ) + 2
A+
A
(y) +
dy a2 (x y ) (y ) = a1 (x) + a3 (x),
A

A+
dy a4 (y y ) (y ) + 2
dy a2 (y x )+ (x ) = a1 (y) + a3 (y),
A+
(B.23)
where
2n
1
.
4u2 + n2
The energy per site of such states in the thermodynamic limit is obtained from (4.8)
$
# N+
N
1
1
1
lim EN+ ,N =2 lim
a1 (xk ) + a3 (xk ) +
a1 (yk ) + a3 (yk ) ,
L L
L L
L
an (u) =
k=1
=2
A

dx (x) a1 (u) + a3 (u) .
k=1
(B.24)
=A
B.4. Low-lying excitations for large, finite lattice sizes

B.4.1. [0, 12 , 12 , 0] states in the sector {j } {uj } for even L
The considerations in the previous section are straightforwardly applied to the lowenergy states with [B, J 3 ] = [0, 0] in the indecomposable [0, 12 , 12 , 0] multiplets for
even L. For these we have to consider solutions satisfying
j = uj ,
j = 1, . . . , L,
(B.25)
which consist of degenerate type- strange strings. In the thermodynamic limit the corresponding root distributions functions are given by (4.24) and hence we may use the
counting function

z(x) = 2 arctan ex
(B.26)
2
to obtain initial values (4.15) for a numerical solution of the Bethe ansatz equations (4.6):
using

Ij
1
1
+
xj = yj = ln tan
(B.27)
L
4
with
1
Ij = (L + 2) + j,
4
j = 1, . . . ,
L
2
(B.28)
564
as starting values a small number of iterations is sufficient to obtain a numerical solution

of the Bethe ansatz equations for lattices with more than 10 000 sites or L 5000. We
present our results on the finite size energies of an indecomposable in Table 7. The scaling
behavior as a function of 1/ log(L) is shown in Fig. 5.
B.4.2. Octet states in the sector {j } {uj } for odd L
The same approach can be used to obtain the finite size energies of the lowest octet
states of the odd L systems: these states are characterized by (L 1)/2 degenerate strange
strings of each type + and . For the numerical solution of the Bethe ansatz equations
we use again (B.27) but with
1
Ij = (L + 1) + j,
4
j = 1, . . . ,
L1
.
2
(B.29)
Table 7
Finite size results for the energy of the low-lying zero-momentum indecomposable consisting of two-strings with
{uj } = {j }
L
L(E L , L E0 )
EL,L
2 2
8
10
128
256
2 2
31.1255373481719104
39.3093986078541491
511.95141708922984
1023.9761543904651
6.9957012146247166
6.9060139214585092
6.2186125785810873
6.1044760409276932
512
2047.9882636363875
6.0090181695995852
1024
4095.9942108167784
5.9281236189417541
2048
4096
8191.9971392992984
16383.998584343625
5.8587150368839502
5.7985285085172748
4.95(2) (extrapolated)
Table 8
Finite size results for the energy of the low-lying zero-momentum octet consisting of two-strings with {uj } =
{j }. Rational function extrapolation has been used to obtain the L value
L
E L1 , L1
2
5
7
161
19.4230243905225
27.5669616952430
643.976352749708
L(E L1 , L1 E0 )
2
2.88487804738772
3.03126813329929
3.80720729705104
321
1283.98786068577
3.89671986852727
641
2563.99379470064
3.97759688996348
1281
5123.99684487686
4.04171273971125
2561
10243.9983994774
4.09893837823756
5121
20483.9991898007
4.14903019789338
565
Table 8 summarizes results for the finite-size energies of the lowest-lying state in the octet
sector. The momentum of this state is zero. For the extrapolation to L = we assume
the corrections of LE to vanish as a function of 1/ log(L). The scaling of the energies
together with the result of the rational function extrapolation is shown in Fig. 5.
B.4.3. More octet states built from strange strings
Octet states from non-degenerate strange string solutions exist for both even and odd
system lengths and are described by Bethe ansatz equations with 2(L 1) rapidities uj and
arranged in N+ (N ) strange strings of type + () with N+ + N = L 1. We have
solved the Bethe ansatz equations (4.6) for such states numerically for (N = N+ N ):
L even: N = 1, 3, 5, 7;
L odd: N = 2 (N = 0 corresponds to the state studied in Appendix B.4.2).
Starting values for the numerical solution of the Bethe ansatz equations for are obtained
from counting functions similar to (B.26) with proper symmetric choice of the quantum
numbers Ij , Jk . The numerical finite size spectrum of these states is given in Table 9. The
energy differences to the ground state as a function of 1/ log L are plotted in Fig. 5.
Based on our results we conjecture that for any fixed N
lim L(E 1 (L1+N ), 1 (L1N ) E0 ) 4.93,
(B.30)
i.e., we expect a macroscopic degeneracy of states in the finite size spectrum to order 1/L
of the super spin chain!
B.5. The Fermi velocity
In order to determine scaling dimensions we need to know the value of an important
parameter, the Fermi velocity vF . Indeed, the Hamiltonian of our system is determined up
to a scale only, and from conformal field theory we only know, for instance, that the ground
state energy E0 of the Hamiltonian on a lattice of L sites scales for L like

E0
cvF
= e
(B.31)
+ o L2 ,
2
L
6L
where e is the ground state energy density and c is the central charge of the CFT. The
determination of vF is straightforward is many systems solvable by Bethe ansatz, but not so
here. We suggest that (we used this value in the main text to deal directly with a properly
normalized Hamiltonian)
v = .
(B.32)
Let us briefly list the evidence supporting (B.32).

(1) We can determine the dispersion of excited states over the lowest octet discussed
above, that are described by the integer method (not all states are, as pointed out above).
It follows from (4.8) and (4.9) that the bare energy and bare momentum of a 22 narrow
566
Table 9
Finite size energies for the low-lying octets consisting of strange strings for L even and odd, respectively
L
2 2
L(E L , L 1 E0 )
479.96564417058130
959.98261219919220
1919.9912162303681
3839.9955678047386
7679.9977658894386
4.1226995302440628
4.1730721938711213
4.2162094233208336
4.2549074509588536
4.2894922778941691
E L , L 1
120
240
480
960
1920
15359.998874847095
3840
2 2
4.3205871549434960
E L+1 , L3
2
483.95681608386576
9.6397875255777501
1923.9895168083938
3843.9948148688195
7683.9974299991591
121
241
481
961
1921
15363.998724015133
3841
L(E L+1 , L3 E0 )
2
5.2252538522429859
5.1206335762225308
5.0424151625950344
4.9829110644573120
4.9369716154324124
4.9010578736833850
E L +1, L 2
2
511.94458966164291
1023.9737491964416
128
256
2047.9874366076488
4095.9939402783793
8191.9970594424680
16383.998566214848
512
1024
2048
4096
L(E L +1, L 2 E0 )
2
7.0925233097077580
6.7202057109388988
6.4324568838346750
6.2051549395546317
6.0222618254764871
5.8727839789301150
string are given by

0 (u) = 4 a1 (u) + a3 (u) ,

p0 (u) = 2 (u) + (u/3) ,
(B.33)
(B.34)
where u denotes the real centre of the string. Given the integral equation (4.23) for the root
density of the octet state, we may define a dressed energy (u) and a dressed momentum
p(u) as usual by

(u) = 0 (u)

du 2a2 (u u ) + a4 (u u ) (u ),
p(u) = p0 (u)

du 2a2 (u u ) + a4 (u u ) p(u ).
(B.35)
(B.36)
567
These equations are easily solved by Fourier transform

(u) =
2
,
cosh(u)

p(u) = 2 arctan sinh(u) .
(B.37)
We may eliminate the spectral parameter and obtain an expression of the dressed energy in
terms of the dressed momentum directly
(p) = 2 cos(p/2).
(B.38)
The contribution of a 22 string with centre u to the ground state energy and momentum is
then given by (u) and p(u) respectively. Making a hole excitation by removing a 22
string with centre u from the ground state would change the energy and momentum by
Eh = (u),
Ph = p(u),
(B.39)
and hence
Eh (Ph ) = 2 cos(Ph /2).
(B.40)
Linearizing this dispersion around the Fermi points Ph = kF = gives the Fermi
velocity

dEh (Ph )
= .
v=
(B.41)
dPh Ph =
(2) Following [77] we can determine v from functional equations for the largest eigenvalue (u) of the transfer matrix by comparing them to crossing equations for relativistic
S-matrices. From (4.11) with = 0 we obtain
(u) = (1 u),
(B.42)
which again implies that v = .

(3) All of our finite-size spectra for small L = 5, 7, 9, 11 systems are consistent with
v = as are the extrapolated values for very large systems.
Now the scaling dimensions and conformal spins of the excitations considered above can
be determined from the finite size spectra of the Hamiltonian and the momentum operator
L
x = h + h =
(E + 4L),
2 2
L
P.
s = h h =
2
(B.43)
B.6. Antiperiodic boundary conditions

As discussed in the main text, antiperiodic boundary conditions play a crucial role in
the identification of the continuum limit (even though they might not play any role in the
physical network model, for instance). We have studied in some detail the sector M = N .
The Hamiltonian is defined by taking the trace in (4.4) instead of the supertrace; observe
568
Table 10
L=3
L=4
L=5
L=6
E = 13.5335
E = 17.4031
E = 20.8890
E = 24.8774
u1 = 0.3510 0.6258i
u2 = u1
u1 = 0.2958 0.5513i
u2 = u1
u1 = 0.7564 + 0.8124i
u2 = u1
u1 = 0.4385 + 0.5483i
u2 = u1
u3 = 0.5219i
u3 = 0.2958 0.5513i
u4 = u3
u3 = 0.1694 + 0.5980i
u4 = u3
u5 = 0.5013i
u3 = 0.4385 + 0.5483i
u4 = u3
u5 = 0.5207i
u6 = u5
that the boundary conditions break the sl(2/1) symmetry, but respect the bosonic spinsu(2). The Bethe ansatz equations become

uj + i
uj i
+ i
i
L
=
N

uj + i
,
uj i
j = 1, . . . , N,
N

uk + i
,
uk i
= 1, . . . , N.
=1
L
=
(B.44)
k=1
The sector {uj } { } now reduces to the TakhtajanBabujian chain with periodic boundary conditions. By diagonalizing the transfer matrix for small systems, we find that the
ground state lies in the TB sector. For example for L = 3, 4, 5, 6 we have (j uj ) (Table 10).
The lowest-lying excited states above the TB ground state are formed by solutions of
the Bethe ansatz equations (B.44) involving strange strings. In Fig. 6 we show the finitesize scaling of the TB ground state as well as two of these excited states. The numerical
results indicate that both states become degenerate with the TB ground state in the L
limit. The scaling dimensions extrapolate to 14 relative to the ground state with periodic
boundary conditions. We conjecture that there again is an infinite degeneracy in the thermodynamic limit.
Appendix C. Relation to the spin-1 TakhtajanBabujian model

Alcaraz and Martins [58] carried out a finite-size scaling analysis of the anisotropic
spin-1 TakhtajanBabujian (TB) chain based on the following equations

sinh( [uj + i])

sinh( [uj i])
L
= exp(i)
Ln

k=1
sinh( [uj uk + i])

,
sinh( [uj uk i])
j = 1, . . . , L n.
E=
Ln
1
(sin 2 )2
,
2
cos 2 cosh 2 uj
j =1
(C.1)
P=
Ln

2 arctan tanh( uj ) cot( ) .
569
(C.2)
j =1
The twist angle was taken to vary in the interval 0 < . Eqs. (C.1), (C.2) reduce to
the equations of interest here if we take the limits , 0 and rescale E by a factor
of 4 and P by a factor of 2. The finite size spectrum in the TB chain in the limit ,
0 is of the form [58]

2
(x + M + M ) + o L1 , M, M = 0, 1, 2, . . . ,
L

2
PM,M
(C.3)
(s + M M ) + o L1 .
=
L
Here E0 ( = 0, L) is the energy of the ground state for zero twist-angle = 0 and even L.
It is equal to
EM,M
= E0 ( = 0, L) +

2
+ o L1 .
(C.4)
8L
After the rescaling we then obtain the finite-size energies and momenta in the sl(2/1) chain

2 2
1

+ o L1 , M, M = 0, 1, 2, . . . ,
2x
+
2M
+
2M
EM,M
=
4L
+
L
4

2
PM,M
(C.5)
(2s + 2M 2M ) + o L1 .
=
L
Here we have used that the Fermi velocity in the sl(2/1) chain is v = (B.32). This shows
that the scaling dimensions in the TB model and the sl(2/1) chain are related by
E0 ( = 0, L) = L
a doubling of the scaling dimensions in the sl(2/1) chain as compared to the TB model,
combined with a shift by 14 ;
a doubling of the conformal spins.
Appendix D. Relation to the quantum transfer matrix of the supersymmetric tJ

model
The vertex model considered in this paper also appears in the quantum transfer matrix (QTM) approach to the thermodynamics of the integrable supersymmetric tJ chain:
shifting the rapidities used in Ref. [55] (JKS) as
1
1
i
vk k + iu,
(D.1)
wj uj + iu + ,
2
2
2
we obtain from (JKS.18) the BAE in our notation (4.6) with twisted boundary conditions

uj + i
uj + i N/2
= ei
,
uj i
uj i
570
+ i + i
+ i i
N/2
= ei
uk + i
k
uk i
(D.2)
when we set the free parameter in our model to = 4u + 1. The twist entering (D.2)
is related to the chemical potential of the tJ model as exp(i) = exp() (
is the inverse temperature). Note that the QTM deals with the thermodynamic limit of
the quantum chain, the system size N = 2L appearing in (D.2) is the so-called Trotter
number appearing in the expression of the partition function by means of the QTM. The
temperature dependence in the QTM approach enters via u = 4/N .
The eigenvalues of the QTM (JKS.15/6) can be identified with the eigenvalue (4.7) of
the transfer matrix carrying the representation 3 in the matrix space

L

JKS (v) = v 2 + u2 2i(v + iu)
(D.3)
provided that we choose the twist properly, i.e., exp() = 1.
References
[1] D.E. Khmelnitskii, JETP Lett. 38 (1983) 454;
H. Levine, S.B. Libby, A.M.M. Pruisken, Nucl. Phys. B 240 (1984) 30, 49, 71;
A.M.M. Pruisken, Phys. Rev. Lett. 61 (1988) 1297;
A.M.M. Pruisken, Nucl. Phys. B 235 (1984) 277.
[2] A. Altland, M. Zirnbauer, Phys. Rev. B 55 (1997) 1142.
[3] T. Senthil, J.B. Marston, M.P.A. Fisher, Phys. Rev. B 60 (1999) 4245, cond-mat/9902062.
[4] M. Janssen, M. Metzler, M. Zirnbauer, Phys. Rev. B 59 (1999) 15836.
[5] F. Evers, A. Mildenberger, A.D. Mirlin, Phys. Rev. B 64 (2001) 241303.
[6] I. Affleck, D. Haldane, Phys. Rev. B 36 (1987) 5291.
[7] A. Zamolodchikov, Al. Zamolodchikov, Nucl. Phys. B 379 (1992) 602.
[8] K.B. Efetov, Adv. Phys. 32 (1983) 53.
[9] H.A. Weidenmller, Nucl. Phys. B 290 (1987) 87.
[10] M. Zirnbauer, J. Math. Phys. 37 (1996) 4986.
[11] M. Zirnbauer, Conformal field theory of the integer quantum Hall plateau transition, hep-th/9905054.
[12] N. Berkovits, C. Vafa, E. Witten, JHEP 9903 (1999) 018.
[13] M. Bershadsky, S. Zhukov, A. Vaintrob, Nucl. Phys. B 559 (1999) 205.
[14] J.T. Chalker, P.D. Coddington, J. Phys. C 24 (1988) 2665;
For a recent review see B. Kramer, T. Ohtsuki, S. Kettemann, cond-mat/0409625.
[15] M. Zirnbauer, Ann. Phys. 3 (1994) 513;
D.H. Lee, Phys. Rev. B 50 (1994) 10788;
J. Kondev, J.B. Marston, Nucl. Phys. B 497 (1997) 639.
[16] I.A. Gruzberg, A.W.W. Ludwig, N. Read, Phys. Rev. Lett. 82 (1999) 4524, cond-mat/9902063.
[17] N. Read, H. Saleur, Nucl. Phys. B 613 (2001) 409.
[18] R. Gade, J. Phys. A 31 (1998) 4909.
[19] I. Affleck, in: Fields, Strings and Critical Phenomena, Les Houches, 1988.
[20] I. Affleck, Nucl. Phys. B 305 (1988) 582.
[21] P. Goddard, D. Olive, G. Waterson, Commun. Math. Phys. 112 (1987) 591.
[22] A.W.W. Ludwig, A free field representation of the OSP(2/2) current algebra at level k = 2 and Dirac
fermions in a random SU(2) gauge potential, cond-mat/0012189.
[23] D. Bernard, Conformal field theory applied to 2D disordered systems, hep-th/9509137;
M.J. Bhaseen, J.S. Caux, I. Kogan, A.M. Tsvelik, Nucl. Phys. B 618 (2001) 465.
[24] N. Read, H. Saleur, in preparation.
[25] M. Marcu, J. Math. Phys. 21 (1980) 1277, 1284.
571
[26] D. Leites, Acad. Nauk CCCP (1983) 764 (in English).

[27] J. Germoni, J. Algebra 209 (1998) 367.
[28] J. Germoni, Reprsentations indcomposables des superalgbres de Lie spciales linaires, Publications de
lInstitut de Recherche Mathmatique Avance, Universit de Strasbourg.
[29] D.J. Benson, Representations and Cohomology, Cambridge Studies in Advanced Mathematics, vol. 30,
Cambridge Univ. Press, Cambridge, 1998.
[30] H.G. Kausch, Nucl. Phys. B 583 (2000) 513;
M. Gaberdiel, H.G. Kausch, Nucl. Phys. B 538 (1999) 631.
[31] C. de C. Chamon, C. Mudry, X.G. Wen, Phys. Rev. B 53 (1996) R7638.
[32] J. Jacobsen, H. Saleur, in preparation.
[33] F. Lesage, P. Mathieu, J. Rasmussen, H. Saleur, Nucl. Phys. B 647 (2002) 363.
[34] M. Bocquet, D. Serban, M. Zirnbauer, Nucl. Phys. B 578 (2000) 628.
[35] P. Bowcock, M. Hayes, A. Taormina, Nucl. Phys. B 510 (1998) 739.
[36] P. di Francesco, P. Mathieu, D. Senechal, Conformal Field Theory, Springer.
[37] T. Eguchi, A. Taormina, Phys. Lett. B 200 (1988) 315;
T. Eguchi, A. Taormina, Phys. Lett. B 210 (1988) 125.
[38] A.M. Semikhatov, A. Taormina, I. Yu Tipunin, math.QA/0311314.
[39] J. Links, A. Foerster, J. Phys. A 32 (1999) 147, cond-mat/9806129.
[40] J. Abad, M. Ros, J. Phys. A 32 (1999) 3535, cond-mat/9806106.
[41] R.M. Gade, J. Phys. A 32 (1999) 7071.
[42] S. Derkachov, D. Karakhanyan, R. Kirschner, Nucl. Phys. B 583 (2000) 691.
[43] M. Scheunert, W. Nahm, V. Rittenberg, J. Math. Phys. 18 (1977) 146.
[44] P.P. Kulish, E.K. Sklyanin, J. Sov. Math. 19 (1982) 1596.
[45] P.P. Kulish, J. Sov. Math. 35 (1985) 2648.
[46] C.K. Lai, J. Math. Phys. 15 (1974) 1675.
[47] B. Sutherland, Phys. Rev. B 12 (1975) 3795.
[48] P. Schlottmann, Phys. Rev. B 36 (1987) 5177.
[49] F.H.L. Essler, V.E. Korepin, Phys. Rev. B 46 (1992) 9147.
[50] A. Foerster, M. Karowski, Nucl. Phys. B 396 (1993) 611.
[51] F.H.L. Essler, V.E. Korepin, K. Schoutens, Int. J. Mod. Phys. B 8 (1994) 3205.
[52] F.H.L. Essler, V.E. Korepin, Int. J. Mod. Phys. B 8 (1994) 3243.
[53] M.P. Pfannmller, H. Frahm, Nucl. Phys. B 479 (1996) 575;
M.P. Pfannmller, H. Frahm, J. Phys. A 30 (1997) L543;
H. Frahm, M.P. Pfannmller, Zap. Nauchn. Semin. S.-Peterburg. Otdel. Mat. Inst. Steklov. (POMI) 251
(1998) 94.
[54] H. Frahm, Nucl. Phys. B 559 (1999) 613636.
[55] G. Jttner, A. Klmper, J. Suzuki, Nucl. Phys. B 487 (1997) 650, cond-mat/9611058.
[56] H. Saleur, B. Wehefritz-Kaufmann, Phys. Lett. B 481 (2000) 419.
[57] L. Takhtajan, Phys. Lett. A 87 (1982) 479;
H. Babujian, Nucl. Phys. B 215 (1983) 317.
[58] F.C. Alcaraz, M.J. Martins, J. Phys. A 23 (1990) 1439;
F.C. Alcaraz, M.J. Martins, J. Phys. A22 (1989) 1829.
[59] H. Saleur, B. Wehefritz-Kaufmann, Nucl. Phys. B 663 (2003) 443.
[60] P. di Francesco, H. Saleur, J.B. Zuber, Nucl. Phys. B 300 (1988) 393.
[61] J. Suzuki, J. Phys. A 21 (1988) L1175.
[62] H. de Vega, E. Lopes, Nucl. Phys. B 362 (1991) 261.
[63] H. Frahm, N.-C. Yu, J. Phys. A 23 (1990) 2115.
[64] H. Frahm, N.-C. Yu, M. Fowler, Nucl. Phys. B 336 (1990) 396.
[65] F. Woynarovich, On the symmetry of excitations in SU(2) Bethe ansatz systems, cond-mat/9812415.
[66] P. Martin, Potts Models, World Scientific.
[67] C.W. Curtis, I. Reiner, Representation Theory of Finite Groups and Associative Algebras, Wiley
Interscience, 1962.
[68] O. Mathieu, Adv. Stud. Pure Math. 26 (2000) 145.
[69] V. Pasquier, H. Saleur, Nucl. Phys. B 330 (1990) 523.
572
[70] V. Chari, A. Pressley, A Guide to Quantum Groups, Cambridge.

[71] A.V. Razumov, Yu.G. Stroganov, J. Phys. A 34 (2001) 3185.
[72] L.D. Faddeev, L. Takhtajan, J. Sov. Math. 24 (1984);
A.N. Kirillov, J. Sov. Math. 34 (1985) 2298;
F.H.L. Essler, V.E. Korepin, K. Schoutens, Phys. Rev. Lett. 67 (1991) 3848.
[73] F.H.L. Essler, V.E. Korepin, K. Schoutens, Int. J. Mod. Phys. B 8 (1994) 3205;
K. Schoutens, Nucl. Phys. B 413 (1994) 675.
[74] G. Bedrftig, H. Frahm, J. Phys. A 28 (1995) 4453.
[75] G. Bedrftig, F.H.L. Essler, H. Frahm, Phys. Rev. Lett. 77 (1996) 5098;
A. Foerster, J. Links, A.P. Tonel, Nucl. Phys. B 552 (1999) 707.
[76] H. Frahm, M.P. Pfannmller, A.M. Tsvelik, Phys. Rev. Lett. 81 (1998) 2116.
[77] M.J. Martins, B. Nienhuis, R. Rietman, Phys. Rev. Lett. 81 (1998) 504.
First principle approach to correlation functions

of spin-1/2 Heisenberg chain: fourth-neighbor
correlators
H.E. Boos a,1 , M. Shiroishi b , M. Takahashi b
a Physics Department, University of Wuppertal, D-42097 Wuppertal, Germany
b Institute for Solid State Physics, University of Tokyo, Kashiwanoha 5-1-5, Kashiwa, Chiba 277-8581, Japan
Received 6 October 2004; accepted 25 January 2005
Abstract
We show how correlation functions of the spin-1/2 Heisenberg chain without magnetic field in
the anti-ferromagnetic ground state can be explicitly calculated using information contained in the
quantum KnizhnikZamolodchikov equation [qKZ]. We find several fundamental relations which
the inhomogeneous correlations should fulfill. On the other hand, it turns out that these relations
can fix the form of the correlations uniquely. Actually, applying this idea, we have obtained all the
correlation functions on five sites. Particularly by taking the homogeneous limit, we have got the
analytic form of the fourth-neighbor pair correlator Sjz Sjz+4 .
PACS: 05.50.+q; 75.10.Jm; 75.50.Ee
E-mail addresses: boos@physik.uni-wuppertal.de (H.E. Boos), siroisi@issp.u-tokyo.ac.jp (M. Shiroishi),

mtaka@issp.u-tokyo.ac.jp (M. Takahashi).
1 On leave of absence from the Institute for High Energy Physics, Protvino 142281, Russia.
doi:10.1016/j.nuclphysb.2005.01.041
574
H.E. Boos et al. / Nuclear Physics B 712 [FS] (2005) 573599
1. Introduction
The anti-ferromagnetic spin-1/2 Heisenberg XXX chain [1]
H=
L

x x

y y
Sj Sj +1 + Sj Sj +1 + Sjz Sjz+1 ,
(1.1)
j =1
is one of the most fundamental models in the study of quantum magnetism in low dimensions (see, for instance, the book [2]).
The Hamiltonian (1.1) was diagonalized by Bethe in 1931 [3] and the ground state in
the thermodynamic limit was investigated by Hulthn in 1938 [4]. The physical content
was clarified by Faddeev and Takhtajan [6] using the algebraic Bethe ansatz formulated by
Faddeev, Sklyanin and Takhtajan (see the review [5]). Exact integrability of the Heisenberg
model is intimately connected with the YangBaxter equation [7].
The Heisenberg model also appears as an effective Hamiltonian of more complicated
systems such as the one-dimensional Hubbard model at half-filling with a large coupling
constant U .
Since discovering of the model a lot of physical quantities have been calculated exactly
which afford significant and reliable data for the experiments as well as numerical simulations. However, most of them are the bulk quantities and we have obtained little results on
the correlation functions. Especially significant are the (static) spinspin pair correlators
y
Sjz Sjz+k (or equivalently Sj+ Sj+k /2; Sj = Sjx iSj ), for which only the first and the
second neighbor (k = 1, 2) have been known for a long time:

1
1
ln 2 0.14771572685,
Sjz Sjz+1 =
(1.2)
12 3
z z
1
4
3
Sj Sj +2 =
(1.3)
ln 2 + (3) 0.06067976996.
12 3
4
Here (s) is the Riemann zeta function and denotes the ground state expectation
value. Note that (1.2) was derived directly from the ground state energy in the thermodynamic limit by Hulthn, while (1.3) was obtained by Takahashi in 1977 via the strong
coupling expansion for the ground state energy of the half-filled Hubbard model [8] (see
also another derivation by Dittrich and Inozemtsev [9]). Unfortunately we cannot generalize these methods so as to calculate further correlations Sjz Sjz+k k3 .
On the other hand, in 1990s, correlation functions were studied systematically for the
Heisenberg XXZ model. It was mainly developed by Kyoto group (Jimbo, Miwa, Miki,
Nakayashiki, [14], see also the book [15]). Utilizing the representation theory of the quan2 ), they first derived multiple integral representation of arbitrary
tum affine algebra Uq (sl
correlators for the massive XXZ anti-ferromagnet. In 1996 Jimbo and Miwa extended the
above results to the XXX chain and the massless XXZ chain [17]. The manifest integral
formula for special case of correlation functions, Emptiness Formation Probability (EFP)
of the XXX model was earlier obtained by Korepin, Izergin, Essler and Uglov in [16].
Strictly speaking the formulas in the massless region were a conjecture since they were obtained as solutions to the quantum KnizhnikZamolodchikov equations (qKZ equations)
with a certain assumption. However, the same integral formulas have been reproduced by
575
Kitanine, Maillet and Terras, [20,21] in the framework of the quantum inverse scattering
method. So we have now at hand the multiple integral representations for any arbitrary
correlation functions for the Heisenberg XXZ chain.
In 2001, Boos and Korepin invented a method to evaluate these multiple integrals for
the Heisenberg XXX chain [22,23]. They have shown that the integrand can be reduced
to a certain canonical form so that the integrations can be performed. They applied the
method to the case of the emptiness formation probability. There has been also suggested
conjecture that the correlation functions for the XXX model may be expressed in terms of
the Riemanns zeta function at odd arguments with rational coefficients. For instance, the
results for P (4) and P (5) look as follows:

P (4) 1/2 + Sjz 1/2 + Sjz+1 1/2 + Sjz+2 1/2 + Sjz+3
173
11
51
55
1
2 ln 2 +
(3)
ln 2 (3) (3)2 (5)
5
60
6
80
24
85
ln 2 (5),
+
24

P (5) 1/2 + Sjz 1/2 + Sjz+1 1/2 + Sjz+2 1/2 + Sjz+3 1/2 + Sjz+4
=
(1.4)
281
45
489
6775
1 10
ln 2 +
(3)
ln 2 (3)
(3)2
(5)
6
3
24
2
16
192
425
12125
1225
6223
ln 2 (5)
(3) (5)
(5)2 +
(7)
+
6
64
256
256
11515
42777
(1.5)
ln 2 (7) +
(3) (7).
64
512
The last result was obtained in [28].
More recently the method was applied to calculate other correlation functions of the
Heisenberg chain (1.1). Here let us list the results in [29] where all the independent correlation functions on four lattice sites are obtained:
z z
37
14
3
125
1
3 ln 2 + (3)
ln 2 (3) (3)2
(5)
Sj Sj +3 =
12
6
3
2
24
25
ln 2 (5)
+
3
= 0.05024862725 . . . ,
=
Sjx Sjx+1 Sjz+2 Sjz+3
1
91
1
3
35
1
+
ln 2
(3) + ln 2 (3) + (3)2 + (5)
240 12
240
6
80
96
5
ln 2 (5),
x z 24x

Sj Sj +1 Sj +2 Sjz+3
=
1
77
5
3
65
1
ln 2 +
(3)
ln 2 (3) (3)2 (5)
240 6
120
12
20
96
5
+ ln 2 (5),
6
576
Sjx Sjz+1 Sjz+2 Sjx+3
1
169
5
3
65
1
ln 2 +
(3)
ln 2 (3) (3)2 (5)
240 4
240
12
20
96
5
+ ln 2 (5).
(1.6)
6
Note that the other correlation functions among four lattice sites are expressed as a linear
combination of the correlations above. For example we have
z z

Sj Sj +1 Sjz+2 Sjz+3

= Sjx Sjx+1 Sjz+2 Sjz+3 + Sjx Sjz+1 Sjx+2 Sjz+3 + Sjx Sjz+1 Sjz+2 Sjx+3
=
1
29
2
21
35
95
1
ln 2 + (3) ln 2 (3) (3)2 (5) +
ln 2 (5),
80 3
30
3
80
96
24
1
1

3
1
+ Sjz Sjz+1 + Sjz Sjz+2 + Sjz Sjz+3 + Sjz Sjz+1 Sjz+2 Sjz+3 ,
P (4) =
(1.7)
16 4
2
4
=
etc.
In papers [3032] the above results were generalized to the XXZ Heisenberg chain
both in massless and in massive regime.
Generalizing the results (1.6) to the correlation functions on five lattice sites is a natural
next problem. In principle, it is possible to calculate them similarly from the multiple
integrals as was actually done for P (5). It, however, will take a tremendous amount of
work to complete them.
It is worthy to note that all the above results have one specific feature, namely, they
all are expressed in terms of one-dimensional integrals. In series of papers [2527] this
phenomenon was explained by means of the above mentioned connection of correlation
functions in the inhomogeneous case with the quantum KnizhnikZamolodchikov equation, more precisely, by means of a duality between solutions to the qKZ equation on level
4 and 0. It was also argued that this fact has a deep mathematical origin based on theory
of deformed hyper-elliptic integrals, symplectic group and special type of cohomologies
investigated earlier by Nakayashiki and Smirnov [18,19]. One of basic outcomes of the
above scheme is a general ansatz for correlation functions in which only one transcendental
function participates with some rational functions as coefficients. Some special additional
properties derived from the qKZ equation allowed to fix P (n) until n = 6.
In recent paper [33] the properties of the qKZ equation have been further investigated
and the above ansatz has been proved by means of some special recursion relations which
linearly connect correlation functions at n, n 1 and n 2. Coefficients in these relations
appeared to be related to special transfer matrices over an auxiliary space of fractional
dimension. The solution to these relations has also been found in the paper [33] which
gives an explicit result for rational coefficients before the transcendental functions staying
in the ansatz for correlation functions. In spite of the fact that this formula is explicit for
any number of sites n, it is still hard to use it for getting manifest results for concrete n.
In this paper we generalize the scheme suggested in paper [25] to the case of an arbitrary
correlation function. We use a set of relations for correlation functions that follow from
the qKZ equation. We optionally call them first principle relations. The claim is that
577
these relations together with the above mentioned ansatz completely fix rational functions
which participate in this ansatz. We find the whole set of correlation functions in case of
n = 5. After taking homogeneous limit we come to the result in terms of the Riemann zeta
function in full agreement with the above conjecture, which now should be considered as
proved. The first principle relations as well as results for correlation functions should be
definitely equivalent to those given by the formula (3.22) in the paper [33]. But it is still to
show such an equivalence explicitly.
2. Quantum KnizhnikZamolodchikov equations and fundamental relations for

correlation functions
In this section we would like to describe an algebraic scheme which allows to calculate
any correlation function for the XXX model. In fact we generalize here the method applied
in paper [24] to the particular case of the EFP.
Let R() be the R-matrix for the XXX model
+ i 0
0
0
R0 () 0
i
0
R() =
(2.1)
,
0
i
0
+ i
0
0
0 + i
where
R0 () =
)( 12
( 2i
2i )
.
( 2i
)( 12 + 2i
)
Sometimes we shall also use notation
1 0
0 0
i
z
0 z+i
0
z+i
,
R(z)
=
0
z
i
0
0
z+i
z+i
(2.2)
where it is implied that = z.

The corner stone of our scheme is the quantum KnizhnikZamolodchikov equation
[qKZ] [1013]. It is a linear system of equations for a vector function gn . In particular,
the qKZ equation on level 4 looks as follows
gn (1 , . . . , j +1 , j , . . . , 2n )1 ,...,j +1 ,j ,...,2n
,
= R(j j +1 )j ,j +1 gn (1 , . . . , j , j +1 , . . . , 2n )1 ,...,j ,j +1 ,...,2n ,

j
j +1
(2.3)
gn (1 , . . . , 2n1 , 2n + 2i)1 ,...,2n1 ,2n

= gn (2n , 1 , . . . , 2n1 )2n ,1 ,...,2n1 .
(2.4)
It was suggested by the Kyoto school [14] that the correlation functions are connected with
solutions to the qKZ equation. In particular, for the massless case Jimbo and Miwa [17]
found that one should take a special solution of the qKZ on level 4 which satisfies one
578
additional constraint
gn (1 , . . . , j 1 , j , j i, j +2 , . . . , 2n )1 ,...,j 1 ,j ,j +1 ,j +2 ,...,2n
= j j ,j +1 gn1 (1 , . . . , j 1 , j +2 , . . . , 2n )1 ,...,j 1 ,j +2 ,...,2n .
(2.5)
Solutions to the above Eqs. (2.4), (2.5) are meromorphic functions with possible singularities at the points
(j k ) = l,
l Z\0
but regular at (j k ) = .

Due to Jimbo and Miwa, any correlation function

,...,

P11,...,nn E11 Enn
is given by the following formula
,...,
P11,...,nn (z1 , . . . , zn ) = gn (1 , . . . , n , n + i, . . . , 1 + i)1 ,...,n ,n ,..., , (2.6)

1
where as above j = zj .
Let us list the first principle relations for the correlation functions which follow from
the qKZ equation:
translational invariance
,...,
,...,
P11,...,nn (z1 + x, . . . , zn + x) = P11,...,nn (z1 , . . . , zn );
(2.7)
transposition, negating and reverse order relations

,...,
,...,
,...,n
n
1
P11,...,nn (z1 , . . . , zn ) = P1,...,
(z1 , . . . , zn ) = P1 ,...,n (z1 , . . . , zn )
1
,...,
= Pnn,...,11 (zn , . . . , z1 );
(2.8)
intertwining relation

...
, ...
+1 j
R j j +1 (zj zj +1 )P...jj+1
,j ... (. . . zj +1 , zj . . .)
j j +1
... ,
...

+1
j j +1
= P...jj,jj+1
... (. . . zj , zj +1 . . .)Rj j +1 (zj zj +1 );
(2.9)
first recurrent relation

, ,...,
, ,...,n
(z, z3 , . . . , zn ),
1 3 ,...,n
, ,...,
1 , ,...,
P11,22,...,nn (z + i, z, z3 , . . . , zn ) = 1 ,2 1 2 P2 ,3
n
3
P11,22,...,nn (z i, z, z3 , . . . , zn ) = 1 ,2 1 2 P2 ,3 ,...,
n (z, z3 , . . . , zn );
(2.10)
second recurrent relation

,..., ,...,n
lim P11,...,jj,...,n (z1 , . . . , zj , . . . , zn )
zj
1 ,..., ,...,n
= j ,j P 1,..., j,...,n (z1 , . . . , z j , . . . , zn );
j
2 1
(2.11)
identity relation

,...,
P11,...,nn (z1 , . . . , zn ) =
n
1 ,...,

i i = i i
579
,...,
P11,...,nn (z1 , . . . , zn )
1 ,...,n

i i = i i
+,...,+
,...,
= P+,...,+
(z1 , . . . , zn ) = P,...,
(z1 , . . . , zn );
(2.12)
reduction relation
+, ,...,
, ,...,
,...,
P+,22,...,nn (z1 , z2 , . . . , zn ) + P,22,...,nn (z1 , z2 , . . . , zn ) = P22,...,nn (z2 , . . . , zn ),

,...,
,...,
,+
P11,,...,n1
(z , z2 , . . . , zn ) + P11,,...,n1
(z , z2 , . . . , zn )
n1 ,+ 1
n1 , 1
,...,
n1
= P11,...,n1
(z1 , . . . , zn1 ).
(2.13)
The relations (2.7), (2.8) and (2.13) were discussed by Jimbo and Miwa in the paper [17].
The intertwining relation (2.9) is the direct corollary of the qKZ equations. The proof of the
first and the second recurrent relations (2.10), (2.11) can be performed in the same lines as
it was described in the paper [24]. In particular, the relation (2.11) was completely proved
in the paper [33]. The identity relation (2.12) is a direct consequence of the sl2 symmetry
of the solution to the qKZ equation.
In papers [25,26], a general ansatz for the correlation functions was suggested. Then it
was proved2 in [33]
,...,
P11,...,nn (z1 , . . . , zn )
n
[2]
,...,
A11 ,...,nn (k1 , . . . , k2m |z1 , . . . , zn )
m=0 1k1
=
=k2m n
G(zk1 zk2 ) G(zk2m1 zk2m ),
(2.14)
where
G(x) = 2

x2 + 1
(1)k k 2
x + k2
(2.15)
k=1
,...,
and A11 ,...,nn (k1 , . . . , k2m |z1 , . . . , zn ) are rational functions with known denominator and
polynomial numerator of a known degree in variables z1 , . . . , zn , namely,
,...,
A11 ,...,nn (k1 , . . . , k2m |z1 , . . . , zn ) =
where

i<j (zi
,...,
Q11 ,...,nn (k1 , . . . , k2m |z1 , . . . , zn )

,

i<j (zi zj )
(2.16)
zj ) means the product in which the differences with {i, j } = {k2l , k2l+1 }
,...,
and i, j
/ K should be excluded, K is the set k1 , . . . , k2m . Q11 ,...,nn (k1 , . . . , k2m |z1 , . . . , zn )
2 The function () used in [33] is related to the function G by () = G(i) + 1/2.
580
are polynomials of the variables z which have the same degree in each variable as in the
denominator and also the same degree of homogeneity. Some useful properties of the function G(x) are listed below [24],
G(x) = G(x),
1
G() = ,
2
G(x i) = (x) + (x)G(x),
(2.17)
(2.18)
(2.19)
where
x 2i
x(x 2i)
,
(x) = 2
.
x i
x +1
Moreover G(x) is a generating function of the alternating zeta functions
(x) =
(2.20)

G(x) = 2 1 + x 2
(1)k x 2k a (2k + 1),
(2.21)
k=0
with
a (s)

(1)n1
n=1
ns

= 1 21s (s).
(2.22)
Conjectural formula for the very first term when m = 0 looks as follows
,...,
,...,
A11 ,...,nn (0|z1 , . . . , zn ) = Q11 ,...,nn (0|z1 , . . . , zn ) =
1
,
(n + 1) nr
(2.23)
where r is the number of down (or up) spins in the set 1 , . . . , n (or 1 , . . . , n ).
Our claim is that the first principle relations (2.7)(2.13) together with the ansatz
(2.14) and conjecture (2.23) completely defines all unknown coefficients in polynomials
,...,
Q11 ,...,nn (k1 , . . . , k2m |z1 , . . . , zn ) and hence the whole set of correlation functions.
Here let us list the explicit form of the inhomogeneous correlation functions for n = 1
and n = 2.
1
P++ (z1 ) = P (z1 ) = ,
2
1 1
++
P++ (z1 , z2 ) = P (z1 , z2 ) = + G(z1 z2 ),

3 6
1
1
+
+
(z1 , z2 ) = P+
(z1 , z2 ) = G(z1 z2 ),
P+
6 6
1 1
+
+
P+ (z1 , z2 ) = P+ (z1 , z2 ) = + G(z1 z2 ).
(2.24)
6 3
One can easily confirm these correlation functions satisfy the first principle relations.
In the subsequent sections, we denote the inhomogeneous EFP as

++...+
...
(z1 , z2 , . . . , zn ) = P...
(z1 , z2 , . . . , zn ) .
Pn (z1 , z2 , . . . , zn ) P++...+
(2.25)
++
Especially we have P2 (z1 , z2 ) = P++
(z1 , z2 ). The explicit form of the above inhomogeneous EFPs up to n = 6 was obtained in [24].
581
3. Inhomogeneous correlation functions for n = 3

Here we consider the inhomogeneous correlation functions for n = 3 and the corresponding first principle relations. This case was considered as an example in the paper
[33]. But we reproduce it here in order to make our approach more illustrative. To describe

the explicit expressions of P112233 (z1 , z2 , z3 ), we introduce some notations:

G12
G13
G23
Gij G(zij ),
G123
+
+
,
z13 z23 z12 z32 z21 z31

1
1
1
1
1
1
A
G12 +
G13 +
G23 ,
G123 =
(3.1)
z13 z23
z23 z12
z31 z21
where we have also introduced an abbreviation zij = zi zj . With these notations, the EFP
+++
(z1 , z2 , z3 ) is given by [24]
P3 (z1 , z2 , z3 ) P+++
P3 (z1 , z2 , z3 ) =
1
1
1
+ (G12 + G13 + G23 ) G123 .
4 12
12
(3.2)

Next we consider the inhomogeneous correlation functions P112233 (z1 , z2 , z3 ) in the

sector 1 + 2 + 3 = 1 + 2 + 3 = 1. For convenience, we first exhibit the final results
in Table 1. There each row corresponds to the subscript 1 , 2 , 3 , while each column corresponds to the superscript 1 , 2 , 3 .
Below we show the strategy to calculate these correlation functions from our first
principle relations. Assume we have already obtained the EFP, P3 (z1 , z2 , z3 ) and all the
correlations for n = 2. As is shown in Table 1, there are 9 correlation functions in the sector 1 + 2 + 3 = 1 + 2 + 3 = 1. However, we find, they are actually connected through
the intertwining relation (2.9) as follows,
z23 + i ++
i ++
P++ (z1 , z3 , z2 )
P
(z1 , z2 , z3 ),
z23
z23 ++
z21 + i ++
i ++
++
(z1 , z2 , z3 ) =
P++ (z2 , z1 , z3 )
P
(z1 , z2 , z3 ),
P++
z21
z21 ++
++
(z1 , z2 , z3 )
P++

i ++
++
++
= P++
P
(z2 , z1 , z3 ) +
(z2 , z1 , z3 ) P++
(z1 , z2 , z3 ) ,
z21 ++
++
(z1 , z2 , z3 ) =
P++
Table 1
Inhomogeneous correlation functions for n = 3 in the sector 1 + 2 + 3 = 1

P112233
++
++
++
++
1
1
12 + 12 (G23 G12 G13 )
1
+ 12 G123
1 + 1 G i GA
12
6 12
12 123
1
i
1
A
12 + 6 G12 + 12 G123
1
1
1
12 + 6 G13 6 G123
i
A
12 G123
1 + 1 G + i GA
12
6 23
12 123
++
++
1
1
1
12 + 6 G13 6 G123
i
A
+ 12 G123
1
1
12 + 12 (G13 G23 G12 )
1
12 G123
1 + 1 G i GA
12
6 23
12 123
1
1
12 + 12 (G12 G13 G23 )
1
+ 12 G123
582

++
P++
(z1 , z2 , z3 ) =
++
P++
(z1 , z2 , z3 )
z32 + i ++
i ++
P++ (z1 , z3 , z2 )
P
(z1 , z2 , z3 ),
z32
z32 ++

i ++
++
P++ (z2 , z1 , z3 ) P++
(z1 , z2 , z3 ) ,
z21
+
i ++
z
i ++
12
++
P++
(z1 , z2 , z3 ) =
P++ (z2 , z1 , z3 )
P
(z1 , z2 , z3 ),
z12
z12 ++
++
(z1 , z2 , z3 )
P++

i ++
++
++
= P++
P
(z1 , z3 , z2 ) +
(z1 , z3 , z2 ) P++
(z1 , z2 , z3 ) .
z32 ++
++
= P++
(z2 , z1 , z3 ) +
(3.3)
++
These relations imply we need to calculate only two correlations, for example, P++
(z1 ,
++
z2 , z3 ) and P++ (z1 , z2 , z3 ). The remaining correlation functions are obtained immedi++
(z1 , z2 , z3 ) can be derived easily from
ately from (3.3). Moreover the diagonal one, P++
the reduction relation (2.13),
++
(z1 , z2 , z3 )
P++
= P2 (z2 , z3 ) P3 (z1 , z2 , z3 )

1
1
1
1 1
+ (G12 + G13 + G23 ) G123
= + G23
3 6
4 12
12
1
1
1
=
+ (G23 G12 G13 ) + G123 .
12 12
12
(3.4)
++
Therefore our task reduces to calculate one off-diagonal correlation P++
(z1 , z2 , z3 ).
Note that there are different ways to connect the correlations from (3.3). They, however,
are consistent with each other due to the YangBaxter relation for R ij (z).
++
(z1 , z2 , z3 ). First, we find
Look at other first principle relations and apply them to P++
the reverse order relation gives
++
++
(z1 , z2 , z3 ) = P++
(z3 , z2 , z1 ),
P++
(3.5)
++
which means P++
(z1 , z2 , z3 ) is symmetric with respect to the exchange z1 z3 . Next,
++
++
(z1 , z2 , z3 ) = P++
(z1 , z2 , z3 ) together with (3.3),
the transposition relation P++
(3.4) and (3.5) gives an equation
++
(z1 , z2 , z3 )
P++
z12 + i z32 + i ++
i z31 + i ++
=
P++ (z1 , z2 , z3 )
P++ (z2 , z1 , z3 )
z12
z32
z12 z31
i z13 + i ++
1
P++ (z1 , z3 , z2 )
P2 (z1 , z3 )
z32 z13
z12 z32
i z12 + i
i z32 + i
P2 (z1 , z2 ) +
P2 (z2 , z3 ).
+
z31 z32
z13 z12
(3.6)
++
The recurrent relations for P++
(z1 , z2 , z3 ) are given by
++
(z + i, z, z3 ) = 0,
P++
++
+
P++
(z i, z, z3 ) = P+
(z, z3 ),
(3.7)
583
++
++
lim P++
(z1 , z2 , z3 ) = lim P++
(z1 , z2 , z3 ) = 0,
z1
z3
1 +
++
(z1 , z2 , z3 ) = P+
(z1 , z3 ).
lim P++
z2
2
(3.8)
The identity relation gives a relation,

z32 + i ++
z23 + i ++
P++ (z1 , z2 , z3 ) +
P++ (z1 , z3 , z2 )
z32
z23
= 2P3 (z1 , z2 , z3 ) P2 (z2 , z3 ),
(3.9)
where we have used the relations (3.3) and (3.4). Eqs. (3.5)(3.9) with the translational
invariance
++
++
P++
(z1 + x, z2 + x, z3 + x) = P++
(z1 , z2 , z3 )
(3.10)
++
are our first principle relations reduced to P++
(z1 , z2 , z3 ).
++
(z1 , z2 , z3 ) in the form
According to the ansatz (2.14) and (2.23), let us assume P++
++
P++
(z1 , z2 , z3 ) =
c11 + c12 z13 + c13 z23 + c14 z13 z23

1
+
G12
12
z13 z23
c21 + c22 z12 + c23 z32 + c24 z12 z32
+
G13
z12 z32
c31 + c32 z21 + c33 z31 + c34 z21 z31
+
G23 ,
z21 z31
(3.11)
where cij are the coefficients which will be determined by the relations (3.5)(3.9). Note
that from the symmetry relation (3.5), we at once have several relations
c31 = c11 ,
c32 = c13 ,
c33 = c12 ,
c34 = c14 ,
c23 = c22 .
(3.12)
Relations (3.6)(3.9) generate equations for the coefficients cij . Although they are an
overdetermined system of the equations for cij , one can figure out it has a unique solution
1
c11 = c21 = c31 = ,
6
c14 = c34 = 0,
c12 = c13 = c22 = c23 = c32 = c33 =
1
c24 = .
6
++
Namely the correlation function P++
(z1 , z2 , z3 ) is determined as

1
1
1
1
i
1
++
G12
+
(z1 , z2 , z3 ) =
+
+
P++
12 6
z13 z23 2
z23 z13

1
1
i 1
1
G13
1
+
+
6
z12 z32 2 z32 z12

1
1
i 1
1
G23
+
+
6
z21 z31 2 z31 z21
1
1
1
i
+ G13 G123 + GA
=
.
12 6
6
12 123
i
,
12
(3.13)
(3.14)
584
Other correlation functions are obtained from the intertwining relations (3.3). In this
way we could calculate all the inhomogeneous correlation functions as shown in Table 1.
The correlation functions in the sector 1 + 2 + 3 = 1 + 2 + 3 = 1 are simply obtained by the negation relations,

, ,
P112233 (z1 , z2 , z3 ) = P11,22,33 (z1 , z2 , z3 ).
(3.15)
Now let us think of an inhomogeneous generalization of the next-nearest neighbor correlation function,
z z
S1 S3 (z1 , z2 , z3 )
1
++
++
= P3 (z1 , z2 , z3 ) P++
(z1 , z2 , z3 ) + P++
(z1 , z2 , z3 )
2

++
(z1 , z2 , z3 )
P++
1
= P2 (z1 , z2 ) P2 (z2 , z3 ) + 2P3 (z1 , z2 , z3 )
4
1
1
1
=
+ G13 G123
12 6
6

1
1
1 G(z12 ) G(z13 ) G(z23 )
.
+ G(z13 )
=
+
+
(3.16)
12 6
6 z13 z23
z12 z32
z21 z31
Expanding G-function as

G(z) = 2 1 + z2 a (1) z2 a (3) + ,
(3.17)
and substituting it into (3.16), one can take the homogeneous limit z1 , z2 , z3 0,

1
1
a (1) a (1) a (3)
lim S1z S3z (z1 , z2 , z3 ) =
zi 0
12 3
1
4
3
=
ln 2 + (3),
12 3
4
which reproduces the expression (1.3).
(3.18)
4. Inhomogeneous correlation functions for n = 4

Next we consider the inhomogeneous correlation functions for n = 4, P11223344 (z1 , z2 ,

++++
(z1 , z2 , z3 , z4 ) was calculated in [24],
z3 , z4 ). The EFP P4 (z1 , z2 , z3 , z4 ) = P++++
P4 (z1 , z2 , z3 , z4 )
1
(0)
= + A(0)
4,1 (z1 , z2 , z3 , z4 )G(z12 ) + A4,1 (z1 , z3 , z2 , z4 )G(z13 )
5
(0)
(0)
+ A4,1 (z1 , z4 , z3 , z2 )G(z14 ) + A4,1 (z3 , z2 , z1 , z4 )G(z23 )
(0)
(0)
+ A4,1 (z4 , z2 , z3 , z1 )G(z24 ) + A4,1 (z4 , z3 , z2 , z1 )G(z34 )

(0)
585
(0)
+ A4,2 (z1 , z2 , z3 , z4 )G(z12 )G(z34 ) + A4,2 (z1 , z3 , z2 , z4 )G(z13 )G(z24 )

(0)
+ A4,2 (z1 , z4 , z3 , z2 )G(z14 )G(z23 ),

with
(0)
(0)
A4,i (z1 , z2 , z3 , z4 ) =
Q4,i (z1 , z2 , z3 , z4 )
z13 z14 z23 z24
(i = 1, 2),
(0)
Q4,2 (z1 , z2 , z3 , z4 )

3
2 2
3
3
1
2
(z13 z24 1)(z14 z23 1) +
z12 +
z34
+
,
+
=
36
5
2
2
2
2

1 2
(0)
(0)
Q4,1 (z1 , z2 , z3 , z4 ) = 2Q4,2 (z1 , z2 , z3 , z4 )
z12 + 4 z34
+1 .
60
(4.1)
Let us think of the correlation functions in the sector 1 + 2 + 3 + 4 = 1 + 2

+ 3 + 4 = 2. There are 16 correlations in this sector. They, however, are connected via the
intertwining relations similarly in the case of n = 3. In fact, if one can calculate a single
+++
correlation, for example, P+++
(z1 , z2 , z3 , z4 ), other correlation functions are obtained
with the help of the intertwining relations and the reduction relation
+++
(z1 , z2 , z3 , z4 ) = P3 (z2 , z3 , z4 ) P4 (z1 , z2 , z3 , z4 ).
P+++
(4.2)
++
P++
(z1 ,
Therefore we can invoke the same strategy as was done for the calculation of
z2 , z3 ) in the previous section. Namely, we reduce the first principle relations to those
+++
(z1 , z2 , z3 , z4 ) and solve them. Below we list the equations
for a single functions P+++
corresponding to (3.5)(3.9) for n = 3:
symmetry relations
+++
+++
+++
(z1 , z2 , z3 , z4 ) = P+++
(z1 , z3 , z2 , z4 ) = P+++
(z4 , z3 , z2 , z1 );
P+++
transposition relation
+++
(z1 , z2 , z3 , z4 )
P+++
(z12 + i)(z13 + i)(z42 + i)(z43 + i) +++
=
P+++ (z1 , z2 , z3 , z4 )
z12 z13 z42 z43
i(z12 + i)(z14 + i)(z32 + i) +++
P+++ (z1 , z2 , z4 , z3 )
z12 z14 z32 z43

i(z13 + i)(z14 + i)(z23 + i) +++
P+++ (z1 , z3 , z4 , z2 )
z13 z14 z23 z42

i(z43 + i)(z41 + i)(z23 + i) +++
P+++ (z2 , z1 , z3 , z4 )
z43 z41 z23 z12

i(z42 + i)(z41 + i)(z32 + i) +++
P+++ (z3 , z1 , z2 , z4 )
z42 z41 z32 z13

1 + i(z12 + z43 ) + z12 z34 + z13 z24 +++
P+++ (z3 , z1 , z2 , z4 )
+
z12 z13 z24 z34
(4.3)
586
i(z42 + i)(z43 + i)
i(z12 + i)(z13 + i)
P3 (z1 , z2 , z3 ) +
P3 (z2 , z3 , z4 )
z42 z43 z41
z12 z13 z14
z23 + i
z32 + i
P3 (z1 , z2 , z4 ) +
P3 (z2 , z3 , z4 );
+
z32 z43 z31
z23 z42 z21
(4.4)
first recurrent relations

+++
lim P+++
(z1 , z2 , z3 , z4 ) = 0,
z1
1 ++
+++
lim P+++
(z1 , z2 , z3 , z4 ) = P++
(z1 , z3 , z4 );
2
z2
(4.5)
second recurrent relation

+++
P+++
(z + i, z, z3 , z4 ) = 0,
+++
++
P+++
(z i, z, z3 , z4 ) = P++
(z, z3 , z4 );
(4.6)
identity relation
z42 + i z43 + i +++
z23 + i z24 + i +++
P+++ (z1 , z2 , z3 , z4 ) +
P+++ (z1 , z3 , z4 , z2 )
z42
z43
z23
z24
z32 + i z34 + i +++
P+++ (z1 , z2 , z4 , z3 )
+
z32
z34
= 2P4 (z1 , z2 , z3 , z4 ) P3 (z2 , z3 , z4 );
(4.7)
translational invariance
+++
+++
P+++
(z1 + x, z2 + x, z3 + x, z4 + x) = P+++
(z1 , z2 , z3 , z4 ).
(4.8)
+++
(z1 , z2 , z3 , z4 ) may be written in the form
According to the ansatz (2.14), (2.23), P+++
+++
P+++
(z1 , z2 , z3 , z4 )
1
(1)
+ A(1)
=
4,1 (z1 , z2 , z3 , z4 )G(z12 ) + A4,1 (z1 , z3 , z2 , z4 )G(z13 )
20
(1)
+ A(1)
4,2 (z1 , z4 , z2 , z3 )G(z14 ) + A4,3 (z3 , z2 , z1 , z4 )G(z23 )
(1)
(1)
+ A4,1 (z4 , z2 , z3 , z1 )G(z24 ) + A4,1 (z4 , z3 , z2 , z1 )G(z34 )

(1)
(1)

(1)
+ A4,5 (z1 , z4 , z3 , z2 )G(z14 )G(z23 ),

(1)
(1)
A4,i (z1 , z2 , z3 , z4 ) =
Q4,i (z1 , z2 , z3 , z4 )
z13 z14 z23 z24
(i = 1, . . . , 5),
(4.9)
where we have taken the symmetry relations (4.3) into account. We can assume the poly(1)
nomials Q4,i (z1 , z2 , z3 , z4 ) as

j j j j
(1)
C4,i (j1 , j2 , j3 , j4 )z11 z22 z33 z44 .
Q4,i (z1 , z2 , z3 , z4 ) =
(4.10)
0j1 ,j2 ,j3 ,j4 2
j1 +j2 +j3 +j4 4
587
Substituting (4.9) and (4.10) into (4.4)(4.8), we can extract the equations for coefficients
C4,i (j1 , j2 , j3 , j4 ) and solve them. Actually we have made a Mathematica program to perform the calculation. The result is
(1)
Q4,1 (z1 , z2 , z3 , z4 )
=
1
1 2
1
1
1
19
7
+ z2 z2
z z2 z2 + z2
90 80 12 12 13 144 14 36 23 48 24 30 34
1 2 2
1 2 2
1 2 2
z z +
z z +
z z
144 13 24 144 14 23 720 12 34

1
1
1 2
1
1 2
2
+ i z13 z24 + z12
z34 z12 z34
z13
z24
9
36
72
18
36

5
1 2
2
+ z13 z24 z14 z23 ,
72
24
(1)
Q4,2 (z1 , z2 , z3 , z4 )
=
59 2
1 2
1 2
1 2
1 2
4 2
19
+
z12 z13
z23
z14
z24
+ z34
90 360
12
12
12
12
45
1 2 2
1 2 2
17 2 2
z z
+ z13 z24 + z14 z23
18
18
360 12 34

1
1
1 2
1
1 2
1
2
2
+ i z13 + z24 z13 z24 z13 z24 z14 z23 z14 z23 ,
9
9
18
18
18
18
(1)
Q4,3 (z1 , z2 , z3 , z4 )
=
29 2
1 2
1 2
1 2
1 2
17 2
1 2 2
19
+
z12 z13
z34 + z13
z23
z14
z24
+
z24
90 360
48
48
48
48
360
72
1 2 2
1 2 2
z z
+ z14
z23
72
180 12 34

1
1
1 2
1
1 2
1
2
2
,
+ i z13 z24 + z13
z24 + z13 z24
+ z14
z23 + z14 z23
36
36
72
72
72
72
(1)
Q4,4 (z1 , z2 , z3 , z4 )
=
1 2
1 2
1 2
1 2
1 2
1 2
7
+ z12
z13
z14
z23
z24
+ z34
45 15
24
72
72
24
15
1 2 2
1 2 2
1 2 2
z z z z + z z
+
360 12 34 72 13 24 72 14 23

1
1
1 2
1
1 2
1
2
2
,
z13 z24 + z12
+i
z34 z12 z34
z13
z24 + z13 z24
18
18
36
36
18
18
(1)
Q4,5 (z1 , z2 , z3 , z4 )
=
17 2
1
1
1
1
17 2
7
+
z z2 z2 z2 z2 +
z
45 180 12 24 13 24 14 24 23 24 24 180 34
588
1 2 2
1 2 2
1 2 2
z12 z34 + z13
z24 + z14
z23
90
36
36

1
1
1 2
1
1 2
1
2
2
.
z13 + z24 z13
+i
z24 z13 z24
z14
z23 z14 z23
18
18
36
36
36
36
(4.11)
Unfortunately we could not find a further compact way to express the results above.
Other correlation functions in this sector are calculated by the intertwining relations. As
it is not possible to present all the explicit results here, we instead show the formal expressions in Table 2. Here C(z1 , z2 , z3 , z4 ) and Z(z1 , z2 , z3 , z4 ) are respectively the real
+++
and the imaginary part of the correlation function P+++
(z1 , z2 , z3 , z4 ) (4.11). Similarly,
A(z1 , z2 , z3 , z4 ), B(z1 , z2 , z3 , z4 ), D(z1 , z2 , z3 , z4 ), E(z1 , z2 , z3 , z4 ) and X(z1 , z2 , z3 , z4 ),
Y (z1 , z2 , z3 , z4 ) are respectively the real and the imaginary parts of other correlation functions. Using the intertwining relations, they are successively calculated from
C(z1 , z2 , z3 , z4 ) and Z(z1 , z2 , z3 , z4 ),

1
Z(z1 , z2 , z3 , z4 ) Z(z1 , z2 , z4 , z3 ) ,
z34

1
Y (z1 , z2 , z3 , z4 ) = Z(z1 , z2 , z4 , z3 ) +
C(z1 , z2 , z4 , z3 ) C(z1 , z2 , z3 , z4 ) ,
z34

1
A(z1 , z2 , z3 , z4 ) = B(z1 , z3 , z2 , z4 ) +
Y (z1 , z2 , z3 , z4 ) Y (z1 , z3 , z2 , z4 ) ,
z23

1
X(z1 , z2 , z3 , z4 ) = Y (z1 , z3 , z2 , z4 ) +
B(z1 , z3 , z2 , z4 ) B(z1 , z2 , z3 , z4 ) ,
z23
D(z1 , z2 , z3 , z4 )

1
= P3 (z1 , z3 , z4 ) P4 (z2 , z1 , z3 , z4 ) +
X(z1 , z2 , z3 , z4 ) X(z2 , z1 , z3 , z4 ) ,
z21
E(z1 , z2 , z3 , z4 )

1
= B(z2 , z1 , z3 , z4 ) +
(4.12)
Y (z1 , z2 , z3 , z4 ) Y (z2 , z1 , z3 , z4 ) .
z21
B(z1 , z2 , z3 , z4 ) = C(z1 , z2 , z4 , z3 ) +
+++
+++
Note that the diagonal correlation functions P+++
(z1 , z2 , z3 , z4 ) and P+++
(z1 , z2 ,
z3 , z4 ) have no imaginary parts. Note also that we have found the imaginary part of
+++
(z1 , z2 , z3 , z4 ) is simply expressed in terms of the function GA
P+++
123 introduced
Table 2
Inhomogeneous correlation functions for n = 4 in the sector 1 + 2 + 3 + 4 = 2

P 1 2 3 4
1 2 3 4
+++
+++
+++
+++
+++
P3 (z2 , z3 , z4 )
P4 (z1 , z2 , z3 , z4 )
A(z1 , z2 , z3 , z4 )
+ iX(z1 , z2 , z3 , z4 )
A(z1 , z2 , z3 , z4 )
iX(z1 , z2 , z3 , z4 )
D(z1 , z2 , z3 , z4 )
C(z1 , z2 , z3 , z4 )
iZ(z1 , z2 , z3 , z4 )
B(z4 , z3 , z2 , z1 )
iY (z4 , z3 , z2 , z1 )
+++
B(z1 , z2 , z3 , z4 )
+ iY (z1 , z2 , z3 , z4 )
E(z1 , z2 , z3 , z4 )
i (GA + GA )
24
123
234
B(z1 , z2 , z3 , z4 )
iY (z1 , z2 , z3 , z4 )
E(z1 , z2 , z3 , z4 )
i (GA + GA )
+ 24
123
234
D(z4 , z3 , z2 , z1 )
+++
C(z1 , z2 , z3 , z4 )
+ iZ(z1 , z2 , z3 , z4 )
B(z4 , z3 , z2 , z1 )
+ iY (z4 , z3 , z2 , z1 )
A(z4 , z3 , z2 , z1 )
+ iX(z4 , z3 , z2 , z1 )
P3 (z1 , z2 , z3 )
P4 (z1 , z2 , z3 , z4 )
+++
A(z4 , z3 , z2 , z1 )
iX(z4 , z3 , z2 , z1 )
589
Table 3
Inhomogeneous correlation functions for n = 4 in the sector 1 + 2 + 3 + 4 = 0

P11223344
++
+ +
+ +
+ +
+ +
+ +
++
+ +
+ +
+ +
+ +
+ +
C1 ({zi })
C2 ({zi })
C3 ({zi })
C4 ({zi })
C5 ({zi })
C6 ({zi })
C2 ({zi })
C7 ({zi })
C8 ({zi })
C9 ({zi })
C10 ({zi })
C5 ({zi })
C3 ({zi })
C8 ({zi })
C11 ({zi })
C12 ({zi })
C9 ({zi })
C4 ({zi })
C4 ({zi })
C9 ({zi })
C12 ({zi })
C11 ({zi })
C8 ({zi })
C3 ({zi })
C5 ({zi })
C10 ({zi })
C9 ({zi })
C8 ({zi })
C7 ({zi })
C2 ({zi })
C6 ({zi })
C5 ({zi })
C4 ({zi })
C3 ({zi })
C2 ({zi })
C1 ({zi })
in (3.1). These properties are reflected in some symmetry relations of the functions
A(z1 , z2 , z3 , z4 ), . . . , Z(z1 , z2 , z3 , z4 ). For example we have
A(z1 , z2 , z3 , z4 ) = A(z2 , z1 , z3 , z4 ), E(z1 , z2 , z3 , z4 ) = E(z1 , z3 , z2 , z4 ),

1 A
A
A
G123 + GA
Y (z1 , z2 , z3 , z4 ) = Y (z2 , z1 , z3 , z4 ) +
234 G213 G134 .
24
We also remark the following relations hold from the identity relations
(4.13)
A(z1 , z2 , z3 , z4 ) + B(z1 , z1 , z3 , z4 ) + C(z1 , z2 , z3 , z4 )

= 2P4 (z1 , z2 , z3 , z4 ) P3 (z2 , z3 , z4 ),
A(z1 , z2 , z3 , z4 ) + D(z1 , z2 , z3 , z4 ) + E(z1 , z2 , z3 , z4 ) + B(z4 , z3 , z2 , z1 )
= P4 (z1 , z2 , z3 , z4 ),

1 A
G123 + GA
234 = 0,
24
X(z1 , z2 , z3 , z4 ) + Y (z1 , z2 , z3 , z4 ) + Z(z1 , z2 , z3 , z4 ) = 0.
X(z1 , z2 , z3 , z4 ) Y (z4 , z3 , z2 , z1 )
(4.14)
Unfortunately we have not been able to afford a nicer compact way to describe the functions A(z1 , z2 , z3 , z4 ), . . . , Z(z1 , z2 , z3 , z4 ), from which the relations such as (4.13), (4.14)
follow naturally.
Next let us consider the sector 1 + 2 + 3 + 4 = 1 + 2 + 3 + 4 = 0. There are 36
correlation functions in this sector. But according to (2.8), we need to know only 12 correlation functions C1 (z1, z2 , z3 , z4 ), . . . , C12 (z1 , z2 , z3 , z4 ) as shown in Table 3.
In the present case, they can be derived from those we have obtained already as follows:
C1 (z1 , z2 , z3 , z4 )
++
+++
= P++
(z2 , z3 , z4 ) P+++
(z1 , z2 , z3 , z4 )
1
1
1
+ (G34 G23 G24 ) + G234 D(z1 , z2 , z3 , z4 ),
=
12 12
12
C2 (z1 , z2 , z3 , z4 )
++
+++
= P++
(z2 , z3 , z4 ) P+++
(z1 , z2 , z3 , z4 )

1
i A
1
+ G12 E(z1 , z2 , z3 , z4 ) +
G GA
=
234 ,
12 6
24 123
590
C3 (z1 , z2 , z3 , z4 )
++
+++
(z2 , z3 , z4 ) P+++
(z1 , z2 , z3 , z4 )
= P++

1
1
1
1 A
+ G24 G234 B(z4 , z3 , z2 , z1 ) i Y (z4 , z3 , z2 , z1 ) G234 ,
=
12 6
6
12
C4 (z1 , z2 , z3 , z4 )
= C3 (z4 , z3 , z2 , z1 )

1
1
1
1
,
+ G13 G123 B(z1 , z2 , z3 , z4 ) + i Y (z1 , z2 , z3 , z4 ) GA
=
12 6
6
12 123

i
i
C3 (z2 , z1 , z3 , z4 )
C5 (z1 , z2 , z3 , z4 ) = 1 +
C3 (z1 , z2 , z3 , z4 ),
z12
z12
C6 (z1 , z2 , z3 , z4 ) = P4 (z1 , z2 , z3 , z4 )
5
Cj (z1 , z2 , z3 , z4 ),
j =1
C7 (z1 , z2 , z3 , z4 )
+
+
= P+
(z1 , z2 , z3 ) P+
(z1 , z2 , z3 , z4 )
++
+++
(z1 , z2 , z3 ) P+++
(z1 , z2 , z3 , z4 )
= P++
1
1
1
+ (G13 G23 G12 ) G123 D(z1 , z2 , z3 , z4 ),
12 12
12
C8 (z1 , z2 , z3 , z4 )
=
++
+++
(z2 , z3 , z4 ) P+++
(z1 , z2 , z3 , z4 )
= P++

1
1
1
,
+ G34 A(z4 , z3 , z2 , z1 ) i X(z4 , z3 , z2 , z1 ) + GA
=
12 6
12 234
C9 (z1 , z2 , z3 , z4 )
= C8 (z4 , z3 , z2 , z1 )

1
1 A
1
+ G12 A(z1 , z2 , z3 , z4 ) + i X(z1 , z2 , z3 , z4 ) + G123 ,
=
12 6
12
C10 (z1 , z2 , z3 , z4 )

i
C5 (z1 , z3 , z2 , z4 ) C5 (z1 , z2 , z3 , z4 ) ,
= C6 (z1 , z3 , z2 , z4 ) +
z32
C11 (z1 , z2 , z3 , z4 )
++
+++
(z1 , z2 , z3 ) P+++
(z1 , z2 , z3 , z4 )
= P++
1
1
1
= (G12 + G13 + G24 + G34 ) + (G123 + G234 ) + P4 (z1 , z2 , z3 , z4 ),
6 12
12
C12 (z1 , z2 , z3 , z4 )

i
C9 (z1 , z2 , z4 , z3 ) C9 (z1 , z2 , z3 , z4 ) .
= C10 (z1 , z2 , z4 , z3 ) +
(4.15)
z43
591
++
Among the correlation functions above, let us pick up a diagonal one P++
(z1 , z2 ,
z3 , z4 ), since it has a relatively simple explicit form.
++
P++
(z1 , z2 , z3 , z4 )

1
1
(2)
(2)
+ A4,1 (z1 , z2 , z3 , z4 )G(z12 ) + A4,1 (z1 , z3 , z2 , z4 )
G(z13 )
30
12

1
1
(2)
G(z
G(z23 )
+ A(2)
(z
,
z
,
z
,
z
)
)
+
A
(z
,
z
,
z
,
z
)
14
4,1 1 4 3 2
4,1 3 2 1 4
12
12

1
(2)
(2)
G(z24 ) + A4,1 (z4 , z3 , z2 , z1 )G(z34 )
+ A4,1 (z4 , z2 , z3 , z1 )
12
(2)
(2)

(2)
+ A4,2 (z1 , z4 , z3 , z2 )G(z14 )G(z23 ),

with
(2)
(2)
A4,i (z1 , z2 , z3 , z4 ) =
Q4,i (z1 , z2 , z3 , z4 )
z13 z14 z23 z24
(i = 1, 2),
Q(2)
4,2 (z1 , z2 , z3 , z4 )

5
1
2 2
3
3
2
(z13 z24 + 1)(z14 z23 + 1) +
z12 +
z34
,
=
+
36
5
2
2
2
2

1 2
(2)
(2)
Q4,1 (z1 , z2 , z3 , z4 ) = 2Q4,2 (z1 , z2 , z3 , z4 )
z12 1 z34
+1 .
60
(4.16)
The correlation functions in the remaining sector 1 + 2 + 3 + 4 = 1 + 2 + 3 + 4 =
2 are obtained from those in 1 + 2 + 3 + 4 = 1 + 2 + 3 + 4 = 2 by the negation
relation

, , ,
P11223344 (z1 , z2 , z3 , z4 ) = P11,22,33,44 (z1 , z2 , z3 , z4 ).
(4.17)
In this way we can get all the inhomogeneous correlation functions for n = 4.

Now we shortly consider the diagonal elements of the correlation functions P11223344
(z1 , z2 , z3 , z4 ). We claim they can be expressed only in terms of the EFPs, P2 (z1 , z2 ),
++
P3 (z1 , z2 , z3 ), P4 (z1 , z2 , z3 , z4 ) and P++
(z1 , z2 , z3 , z4 ). For example, we have
+++
P+++
(z1 , z2 , z3 , z4 ) = P3 (z2 , z3 , z4 ) P4 (z1 , z2 , z3 , z4 ),
+++
++
P+++
(z1 , z2 , z3 , z4 ) = P2 (z3 , z4 ) P3 (z2 , z3 , z4 ) P++
(z1 , z2 , z3 , z4 ),
++
P++
(z1 , z2 , z3 , z4 )
1
= P2 (z1 , z2 ) P2 (z2 , z3 ) P2 (z3 , z4 ) + P3 (z1 , z2 , z3 ) + P3 (z2 , z3 , z4 )
2
++
(z1 , z2 , z3 , z4 ),
+ P++
++
P++
(z1 , z2 , z3 , z4 )
= P2 (z2 , z3 ) P3 (z1 , z2 , z3 ) P3 (z2 , z3 , z4 ) + P4 (z1 , z2 , z3 , z4 ).
(4.18)
592
Then, especially, the inhomogeneous generalization of the third-neighbor correlator S1z S4z
is represented
z z
S1 S4 (z1 , z2 , z3 , z4 )
1 ++++
+++
+++
= P++++
(z1 , z2 , z3 , z4 ) P+++
(z1 , z2 , z3 , z4 ) + P+++
(z1 , z2 , z3 , z4 )
4
++
+++
++
P++
(z1 , z2 , z3 , z4 ) + P+++
(z1 , z2 , z3 , z4 ) P++
(z1 , z2 , z3 , z4 )
++
+
+++
+ P++
(z1 , z2 , z3 , z4 ) P+
(z1 , z2 , z3 , z4 ) P+++
(z1 , z2 , z3 , z4 )
++
++
+
+ P++
(z1 , z2 , z3 , z4 ) P++
(z1 , z2 , z3 , z4 ) + P+
(z1 , z2 , z3 , z4 )
++
+
+
P++
(z1 , z2 , z3 , z4 ) + P+
(z1 , z2 , z3 , z4 ) P+
(z1 , z2 , z3 , z4 )

+ P (z1 , z2 , z3 , z4 )
1
= + P2 (z1 , z2 ) + P2 (z2 , z3 ) + P2 (z3 , z4 )
4

2 P3 (z1 , z2 , z3 ) + P3 (z2 , z3 , z4 )

++
+ 2 P4 (z1 , z2 , z3 , z4 ) P++
(z1 , z2 , z3 , z4 )
1
1
1
+ G14 + (G123 + G234 ) + A(3)
=
4,1 (z1 , z2 , z3 , z4 )G12
12 6
6
(3)
(3)
+ A4,1 (z1 , z3 , z2 , z4 )G13 + A4,1 (z1 , z4 , z3 , z2 )G14
(3)
+ A(3)
4,1 (z3 , z2 , z1 , z4 )G23 + A4,1 (z4 , z2 , z3 , z1 )G24
(3)
(3)
+ A4,1 (z4 , z3 , z2 , z1 )G34 + A4,2 (z1 , z2 , z3 , z4 )G12 G34

(3)
(3)
+ A4,2 (z1 , z3 , z2 , z4 )G13 G24 + A4,2 (z1 , z4 , z3 , z2 )G14 G23 ,

where
(3)
A4,2 (z1 , z2 , z3 , z4 )

(0)
(3)
= 2 A4,2 (z1 , z2 , z3 , z4 ) A4,2 (z1 , z2 , z3 , z4 )

2
1
1
1
,
=
9 z13 z14 z23 z24 z14 z23 z13 z24
(3)
A4,1 (z1 , z2 , z3 , z4 )

(0)
(3)
= 2 A4,1 (z1 , z2 , z3 , z4 ) A4,1 (z1 , z2 , z3 , z4 )

5
1
1
1
3
3
=
.
18 z13 z14 z23 z24 z14 z23 z13 z24 z14 z24 z13 z23
By taking the homogeneous limit zi 0, we recover the result (1.6)

lim S1z S4z (z1 , z2 , z3 , z4 )
zi 0

1
1
a (1) + 2 a (1) a (3)
12 3
(4.19)
593

14
92
50
56
8
+ a (1) + a (3) a (5) a (1)a (3) a (3)2
3
9
9
9
3

80
+ a (1)a (5)
9
74
50
56
1
3a (1) + a (3) a (5) a (1)a (3)
=
12
9
9
9
8
80
2
a (3) + a (1)a (5)
3
9
37
125
14
3
1
25
3 ln 2 + (3)
(5)
ln 2 (3) (3)2 +
ln 2 (5).
=
12
6
24
3
2
3
(4.20)
Other correlation functions in (1.6) are reproduced in a similar way.
5. Correlation functions in the case n = 5
Following the scheme in the previous sections we could further obtain all the correlation
functions for the XXX chain on five lattice sites. In fact we have calculated two inhomo++++
+++
geneous correlation functions P++++
(z1 , z2 , z3 , z4 , z5 ) and P+++
(z1 , z2 , z3 , z4 , z5 )
from the first principle relations. Other correlation functions are derived by the intertwining relations, etc., as before. The calculations have been performed by the heavy use of
Mathematica and the obtained results are too complicated to be described thoroughly
in this paper. Therefore here we shall omit to describe them. We, however, give explicit
form of the inhomogeneous generalization of the fourth-neighbor correlation function
S1z S5z (z1 , z2 , z3 , z4 , z5 ).
z z
S1 S5 (z1 , z2 , z3 , z4 , z5 )
1
1 5 P1122334455 (z1 , z2 , z3 , z4 , z5 )
4 =
j
1
+ A5,1 (z1 , z2 , z3 , z4 , z5 )G12 + A5,1 (z1 , z3 , z2 , z4 , z5 )G13
=
12
+ A5,1 (z1 , z4 , z3 , z2 , z5 )G14 + A5,1 (z5 , z2 , z3 , z4 , z1 )G25
+ A5,1 (z5 , z3 , z2 , z4 , z1 )G35 + A5,1 (z5 , z4 , z3 , z2 , z1 )G45
+ A5,2 (z2 , z3 , z1 , z4 , z5 )G23 + A5,2 (z2 , z4 , z1 , z3 , z5 )G24
+ A5,2 (z3 , z4 , z1 , z2 , z5 )G34 + A5,3 (z1 , z5 , z2 , z3 , z4 )G15
+ A5,4 (z1 , z2 , z3 , z4 , z5 )G12 G34 + A5,4 (z1 , z3 , z2 , z4 , z5 )G13 G24
594

+ A5,5 (z1 , z5 , z2 , z4 , z3 )G15 G24 ,
where
A5,i (z1 , z2 , z3 , z4 , z5 ) =
Q5,i (z1 ,z2 ,z3 ,z4 ,z5 )
(i = 1, 2, 3),
(i = 4, 5)
z13 z14 z15 z23 z24 z25

Q5,i (z1 ,z2 ,z3 ,z4 ,z5 )
z13 z14 z15 z23 z24 z25 z34 z35
with
Q5,1 (z1 , z2 , z3 , z4 , z5 )
1
(z15 + z25 )(z13 + z24 )(2 z13 z14 z14 z23 ) 6(z13 z14 2)(z23 z24 2)
=
36

2
2
+ z12
(z13 z23 + z14 z24 16) + 10z34
2 ,
Q5,2 (z1 , z2 , z3 , z4 , z5 )
= Q5,1 (z1 , z2 , z3 , z4 , z5 ) +

1
z13 z23 5z14 z24 + z45 (z14 + z24 ) 5 ,
18
Q5,3 (z1 , z2 , z3 , z4 , z5 )

1
z15 z25 3z13 z14 z23 z24
18

2
4(z13 z24 + z14 z23 ) 3z34
+5 ,
= Q5,1 (z1 , z2 , z3 , z4 , z5 ) +
Q5,4 (z1 , z2 , z3 , z4 , z5 )

1
2(z13 z24 + z14 z23 )z15 z25 z13 z23 z24 z45 z14 z23 z24 z35 z13 z14 z23 z45
=
18
3 2

2
2 2
2
2
2
2
z13 + z14
z13 z14 z24 z35 + z12
z34 + 6 z12
+ z34
+ z23
+ z24
2

2
2
2
2
+ 12 ,
z15
+ z25
+ z35
+ z45
Q5,5 (z1 , z2 , z3 , z4 , z5 )
1
= Q5,4 (z1 , z2 , z3 , z4 , z5 ) + (2 z13 z24 z14 z23 )z15 z25 z35 z45 .
(5.1)
9
By taking the homogeneous limit, we obtain the fourth-neighbor correlation function for
the homogeneous Heisenberg model as follows:

z z
Sj Sj +4 = lim S1z S5z (z1 , z2 , z3 , z4 , z5 )
zi 0
16
290
1172
700
1
a (1) +
a (3) 72a (1) a (3)
a (3)2
a (5)
12
3
9
9
9
220
400
455
4640
a (1) a (5)
a (3) a (5)
a (5)2 +
a (7)
+
9
9
3
9
3920
a (1) a (7) + 280a (3) a (7)
595
16
145
293
1
875
ln 2 +
(3) 54 ln 2 (3)
(3)2
(5)
12
3
6
4
12
+
1450
275
1875
3185
ln 2 (5)
(3) (5)
(5)2 +
(7)
3
16
16
64
1715
6615
ln 2 (7) +
(3) (7) = 0.034652776982 . . . .
4
32
(5.2)
This is one of the main new results of this paper. Furthermore we have obtained all the other
correlation functions for n = 5. We present below the homogeneous limit of independent
ones

=
1
517
25
203
1
1525
+
ln 2
(3) +
ln 2 (3) +
(3)2 +
(5)
240 12
480
12
80
384
215
5
125
735
ln 2 (5) + (3) (5) +
(5)2
(7)
12
16
32
256
1029
441
ln 2 (7)
(3) (7)
64
64
= 0.0098924350847 . . . ,

Sjx Sjz+1 Sjx+2 Sjz+4
(5.3)
1
301
9
1079
1
975
ln 2 +
(3) ln 2 (3)
(3)2
(5)
240 4
160
2
160
128
+
365
185
1375
735
ln 2 (5)
(3) (5)
(5)2 +
(7)
8
128
128
128
1323
4851
ln 2 (7) +
(3) (7)
32
256
= 0.0027889733995 . . . ,

(5.4)
5
569
61
1109
1
775
ln 2 +
(3)
ln 2 (3)
(3)2
(5)
240 12
240
12
160
96
+
140
185
1375
735
ln 2 (5)
(3) (5)
(5)2 +
(7)
3
128
128
128
1323
4851
ln 2 (7) +
(3) (7)
32
256
= 0.00505666089481 . . . ,
(5.5)
596
1
4445
1
419
35
663
+ ln 2
(3) +
ln 2 (3) +
(3)2 +
(5)
240 4
160
6
80
384
85
1625
2303
115
ln 2 (5) + (3) (5) +
(5)2
(7)
2
64
128
256
833
5733
+
ln 2 (7)
(3) (7)
16
256
= 0.01857662093837 . . . ,

x z
Sj Sj +1 Sjx+3 Sjz+4
=
(5.6)
1
2285
5
1883
28
551
ln 2 +
(3)
ln 2 (3)
(3)2
(5)
240 12
480
3
40
128
165
1375
3577
1135
ln 2 (5)
(3) (5)
(5)2 +
(7)
+
12
64
64
256
4851
343
ln 2 (7) +
(3) (7)
4
128
= 0.00108936897291 . . . ,
x z

Sj Sj +1 Sjz+3 Sjx+4
=
1
497
28
551
1
1155
ln 2 +
(3)
ln 2 (3)
(3)2
(5)
240 2
120
3
40
64
165
1375
1135
3577
ln 2 (5)
(3) (5)
(5)2 +
(7)
+
12
64
64
256
4851
343
ln 2 (7) +
(3) (7)
4
128
= 0.001573361370145 . . . .
(5.7)
(5.8)
As a confirmation of our results, we have evaluated the numerical values of the correlation
functions for the finite system size N up to 32 (Table 4). We have further applied an extrapolation to this data such as c + c1 /N 2 + c2 /N 4 + c3 /N 6 , and obtained the estimated
values c in the limit N . One can clearly observe our analytical results coincide
with the extrapolated values with extremely high accuracy (more than 4 digits).
Table 4
Numerical values of correlation functions for finite systems
N = 26
Sjz Sjz+4
N = 28
N = 30
N = 32
c (N = )
0.0357233
0.0355713
0.0354497
0.0353508
0.0346535
0.0099654
0.0099553
0.0099471
0.0099405
0.0098925
0.0028572
0.0028477
0.0028400
0.0028337
0.0027890
0.0052334
0.0052083
0.0051882
0.0051719
0.0050564
0.0185725
0.0185731
0.0185737
0.0185741
0.0185766
0.0011377
0.0011309
0.0011254
0.0011209
0.0010893
0.0016526
0.0016413
0.0016323
0.0016250
0.0015732
597
We make several comments on our results. First, the two other diagonal correlation
functions are given as a sum of non-diagonal ones similarly in the case of four lattice sites,
z z


1
7
127
15
891
375
ln 2 +
(3)
ln 2 (3)
(3)2
(5)
80 12
40
2
80
32
165
1125
595
2205
ln 2 (5)
(3) (5)
(5)2 +
(7)
+
8
64
64
256
4263
3969
ln 2 (7) +
(3) (7)
64
128
= 0.0121601225799 . . . ,
z z


=
2335
1
2
1307
77
1541
ln 2 +
(3)
ln 2 (3)
(3)2
(5)
80 3
240
6
80
96
245
3875
395
4851
ln 2 (5)
(3) (5)
(5)2 +
(7)
+
3
64
128
256
1911
13671
ln 2 (7) +
(3) (7)
16
256
= 0.0190606133356 . . . .
(5.9)
(5.10)
The expression for P (5) (1.5) is reproduced as

3
1
1

1
1
+ Sjz Sjz+1 + Sjz Sjz+2 + Sjz Sjz+3 + Sjz Sjz+4
32 2
8
4
8

z z
1
+ Sj Sj +1 Sjz+2 Sjz+3 + Sjz Sjz+1 Sjz+3 Sjz+4 + Sjz Sjz+1 Sjz+3 Sjz+4
2
1 10
6775
281
45
489
=
ln 2 +
(3)
ln 2 (3)
(3)2
(5)
6
3
24
2
16
192
425
12125
1225
6223
ln 2 (5)
(3) (5)
(5)2 +
(7)
+
6
64
256
256
11515
42777
ln 2 (7) +
(3) (7)
64
512
= 2.0117259 . . . 106 .
P (5) =
In fact, arbitrary correlation functions on five lattice sites are similarly calculated.
Finally we notice that the next-nearest chiral correlator simplifies drastically as

(Sj Sj +1 ) (Sj +3 Sj +4 )

= 6 Sjx Sjz+1 Sjx+3 Sjz+4 Sjx Sjz+1 Sjz+3 Sjx+4
598
21
75
1
ln 2 (3) + (5)
2
16
64
= 0.015976382058 . . . ,
(5.11)
as well as a similar correlator

(Sj Sj +1 ) (Sj +2 Sj +4 )

7
9
25
47
175
(3) + ln 2 (3) + (3)2 +
(5)
ln 2 (5)
16
2
8
64
4
= 0.04707380576617 . . . .
= ln 2
(5.12)
Recall that a similar simplification occurs in the case of nearest chiral correlator [34]

(Sj Sj +1 ) (Sj +2 Sj +3 )

=
1
3
ln 2 (3).
2
8
(5.13)
6. Summary
In this paper, we calculate correlation functions of the Heisenberg chain without magnetic field in the anti-ferromagnetic ground state using several fundamental functional
relations followed from the quantum KnizhnikZamolodchikov equations. This is the generalization of the method proposed for the emptiness formation probability. In fact we have
calculated all the correlation functions on five lattice sites. Especially we could get the analytic formula for the fourth-neighbor correlation Sjz Sjz+4 . In principle we can continue
to calculate Sjz Sjz+k k5 , which will be reported in the further publications.
Acknowledgements
Authors are grateful to F. Ghmann, M. Jimbo, A. Klmper, V. Korepin, S. Lukyanov,
T. Miwa, K. Sakai, J. Sato, F. Smirnov, Y. Takeyama and Z. Tsuboi for useful discussions. H.B. would like to thank to ISSP of Tokyo University where work on this paper
was partially done. This work is in part supported by Grant-in-Aid for the Scientific Research (B) No. 14340099. M.S. is also supported by Grant-in-Aid for Young Scientists
(B) No. 14740228. H.B. was supported by INTAS grant #00-00561 and by the RFFI grant
#04-01-00352.
References
[1] W. Heisenberg, Z. Phys. 49 (910) (1928) 619.
599
[2] M. Takahashi, Thermodynamics of One-Dimensional Solvable Models, Cambridge Univ. Press, Cambridge,
1999.
[3] H. Bethe, Z. Phys. 76 (1931) 205.
[4] L. Hulthn, Ark. Mat. Astron. Fys. A 26 (1939) 1.
[5] L. Faddeev, How algebraic Bethe ansatz works for integrable models, in: A. Connes, K. Gawedzki, J. ZinnJustin (Eds.), Les Houches Session LXIV, 1995, Quantum Symmetries, Elsevier, Amsterdam, 1998, pp. 149
219.
[6] L. Faddeev, L. Takhtajan, Spectrum scattering of excitations in the one-dimensional isotropic Heisenberg
model, J. Sov. Math. 24 (1984) 241267.
[7] R. Baxter, Exactly Solved Models in Statistical Mechanics, Academic Press, New York, 1982.
[8] M. Takahashi, J. Phys. C: Solid State Phys. 10 (1977) 1289.
[9] J. Dittrich, V.I. Inozemtsev, J. Phys. A 30 (1997) L623.
[10] V.G. Knizhnik, A.B. Zamolodchikov, Nucl. Phys. B 247 (1984) 83.
[11] F.A. Smirnov, Form Factors in Completely Integrable Models of Quantum Field Theory, Adv. Series in
Math. Phys., vol. 14, World Scientific, Singapore, 1992.
[12] F.A. Smirnov, Int. J. Math. Phys. A 7 (Suppl. 1B) (1992) 813.
[13] I. Frenkel, N. Reshetikhin, Commun. Math. Phys. 146 (1992) 1.
[14] M. Jimbo, K. Miki, T. Miwa, A. Nakayashiki, Phys. Lett. A 168 (1992) 256.
[15] M. Jimbo, T. Miwa, Algebraic Analysis of Solvable Lattice Models, CBMS Regional Conference Series in
Mathematics, vol. 85, American Mathematical Society, Providence, 1994.
[16] V. Korepin, A. Izergin, F. Essler, D. Uglov, Phys. Lett. A 190 (1994) 182.
[17] M. Jimbo, T. Miwa, J. Phys. A 29 (1996) 2923.
[18] A. Nakayashiki, F.A. Smirnov, Commun. Math. Phys. 217 (2001) 623.
[19] A. Nakayashiki, On the Cohomology of Theta Divisor of Hyperelliptic Jacobian, in: M. Guest, et al. (Eds.),
Integrable Systems, Topology and Physics, in: Contemporary Mathematics, vol. 309, American Mathematical Society, Providence, 2002.
[20] N. Kitanine, J.M. Maillet, V. Terras, Nucl. Phys. B 554 (1999) 647.
[21] N. Kitanine, J.M. Maillet, V. Terras, Nucl. Phys. B 567 (2000) 554.
[22] H.E. Boos, V.E. Korepin, J. Phys. A 34 (2001) 5311.
[23] H.E. Boos, V.E. Korepin, Evaluation of integrals representing correlators in XXX Heisenberg spin chain, in:
MathPhys Odyssey 2001, Birkhuser, Basel, 2001, p. 65.
[24] H.E. Boos, V.E. Korepin, F.A. Smirnov, Nucl. Phys. B 658 (2003) 417.
[25] H.E. Boos, V.E. Korepin, F.A. Smirnov, J. Phys. A: Math. Gen. 37 (2004) 323.
[26] H.E. Boos, V.E. Korepin, F.A. Smirnov, New formulae for solutions of quantum KnizhnikZamolodchikov
equations on level 4 and correlation functions, hep-th/0305135, Moscow Math. J., in press.
[27] H.E. Boos, V.E. Korepin, F.A. Smirnov, Connecting lattice and relativistic models via conformal field theory,
math-ph/0311020, Prog. Math., in press.
[28] H.E. Boos, V.E. Korepin, Y. Nishiyama, M. Shiroishi, J. Phys. A: Math. Gen. 35 (2002) 4443.
[29] K. Sakai, M. Shiroishi, Y. Nishiyama, M. Takahashi, Phys. Rev. E 67 (2003) 065101.
[30] G. Kato, M. Shiroishi, M. Takahashi, K. Sakai, J. Phys. A: Math. Gen. 36 (2003) L337.
[31] M. Takahashi, G. Kato, M. Shiroishi, J. Phys. Soc. Jpn. 73 (2004) 245.
[32] G. Kato, M. Shiroishi, M. Takahashi, K. Sakai, J. Phys. A: Math. Gen. 37 (2004) 5097.
[33] H. Boos, M. Jimbo, T. Miwa, F. Smirnov, Y. Takeyama, A recursion formula for the correlation functions of
an inhomogeneous XXX model, hep-th/0405044.
[34] N. Muramoto, M. Takahashi, J. Phys. Soc. Jpn. 68 (1999) 2098.
Master equation for spinspin correlation functions

of the XXZ chain
N. Kitanine a,1 , J.M. Maillet b , N.A. Slavnov c , V. Terras d
a LPTM, UMR 8089 du CNRS, Universit de Cergy-Pontoise, France
b Laboratoire de Physique, UMR 5672 du CNRS, ENS Lyon, France
c Steklov Mathematical Institute, Moscow, Russia
d LPMT, UMR 5825 du CNRS, Montpellier, France
Abstract
We derive a new representation for spinspin correlation functions of the finite XXZ spin- 12
Heisenberg chain in terms of a single multiple integral, that we call the master equation. Evaluation of this master equation gives rise on the one hand to the previously obtained multiple integral
formulas for the spinspin correlation functions and on the other hand to their expansion in terms of
the form factors of the local spin operators. Hence, it provides a direct analytic link between these
two representations of the correlation functions and a complete re-summation of the corresponding
series. The master equation method also allows one to obtain multiple integral representations for
dynamical correlation functions.
PACS: 71.45.G; 75.10.Jm; 11.30.Na; 03.65.Fd
E-mail addresses: kitanine@ptm.u-cergy.fr (N. Kitanine), maillet@ens-lyon.fr (J.M. Maillet),

nslavnov@mi.ras.ru (N.A. Slavnov), terras@lpm.univ-montp2.fr (V. Terras).
1 On leave of absence from Steklov Institute at St. Petersburg, Russia.
doi:10.1016/j.nuclphysb.2005.01.050
N. Kitanine et al. / Nuclear Physics B 712 [FS] (2005) 600622
601
1. Introduction
One of the central question in the theory of quantum integrable models [14] is the
exact computation of their correlation functions. Apart from few cases, like free fermions
[510] or conformal field theories [11], this problem is still far from its complete solution.
In particular, the computation of manageable expressions for two point functions of local
operators and their asymptotic behavior at large distance is a central open problem. If one
considers the case T = 0, such a problem reduces to the calculation of the average value in
the ground state | of the product of two local operator 1 , 2 :
g12 = |1 2 |.
(1.1)
There are basically two main strategies to evaluate such a function:

(i) to compute the action of local operators on the ground state 1 2 | = |
and then to
calculate the resulting scalar product g12 = |;
(ii) to insert a sum over a complete set of states |i (for instance, a complete set of
eigenstates of the Hamiltonian) between the local operators 1 and 2 and to obtain
the representation for the correlation function as a sum over one-point matrix elements
(form factor type expansion [1214])

g12 =
(1.2)
|1 |i i |2 |.
i
The aim of this paper is to give a direct, analytic relation between the approaches (i)
and (ii) in the case of the XXZ spin- 12 finite chain. For the sake of simplicity we mainly
consider the case of zero-temperature and equal-time z correlation function. It will be
clear however that our method is more general. In particular, it allows one to compute
dynamical correlation functions. The corresponding results will be described in a sequel to
the present paper.
We consider the periodical XXZ spin- 12 Heisenberg chain in an external magnetic field.
The Hamiltonian of this model is given by [15]
H=
M

x x
y y
z
1 hSz ,
m m+1 + m m+1 + mz m+1
(1.3)
m=1
where
Sz =
M
1 z
m ,
2
[H, Sz ] = 0.
(1.4)
m=1
x,y,z
Here is the anisotropy parameter, h the external classical magnetic field, m

are the
spin operators (in the spin- 12 representation), associated with each site of the chain m, and

2
M is even. The quantum space of states is H = M
m=1 Hm , where Hm C is called the
x,y,z
act as the corresponding Pauli matrices
local quantum space at site m. The operators m
in the space Hm and as the identity operator elsewhere.
602
The method to compute eigenstates and energy levels (Bethe ansatz) of the Hamiltonian
(1.3) was proposed by Bethe in 1931 in [16] and developed later in [1719]. The algebraic
version of the Bethe ansatz was created in the framework of the quantum inverse scattering
method by Faddeev and his school [4,20,21]. However, the knowledge of the correlation
functions of the XXZ chain has for a long time been restricted to the free fermion point
= 0.
In the case of the XXZ chain the approach (i) leads to multiple integral representations for the correlation functions. In the thermodynamic limit, at zero temperature and
for zero magnetic field, such representations were obtained from the q-vertex operator approach (also using corner transfer matrix technique) in the massive regime > 1 in 1992
[22] and conjectured in 1996 [23] for the massless regime 1 < 1 (see also [24]).
A proof of these results together with their extension to non-zero magnetic field was given
in 1999 [25,26] for both regimes using algebraic Bethe ansatz and the actual resolution of
the so-called quantum inverse scattering problem [25,27]. Using these results, the spontaneous magnetization was obtained in [28]. For the case of the XXX model ( = 1) this
type of representation was later studied in [29,30]. Integral representations for spinspin
correlation functions were derived in [31,32]. Recently the generalization of this result for
finite temperature was given in [33].
In the framework of the approach (ii) integral representations for the form factors of the
XXZ chain in the thermodynamic limit were obtained in [22,24,3437]. Determinant representations for the form factors of the finite chain were computed in [25,28]. An effective
summation of the form factor series in the case of free fermions was done in [38,39].
To explain our method we will study the two point correlation function of the third
z
. Following the papers [31,39,46] we use for this purpose a
components of spin 1z m+1
special generating function Q1,m (see (4.1), (4.2)). On the way to relate the previously
obtained multiple integral representation of the spinspin correlation function to its form
factor type expansion we derive what we call the master equation (see (4.4)) for this generating function. It is given as a single multiple integral of Cauchy type over a certain contour
in CN

Q1,m = dz1 dzN F {z} .
(1.5)
The integrand F ({z}) is a function of the N variables z1 , . . . , zN . It is mainly given in

terms of the eigenvalues of the twisted transfer matrix (see (3.1)) and of the functions
Y defining the corresponding twisted Bethe equations (see (3.5), (3.6)). It is a periodical function of each argument zj , vanishing at zj . Therefore there are two ways
to evaluate the integral (1.5): either to compute the residues in the poles inside , or to
compute the residues in the poles within strips of the width i outside .
z
in terms of
The first way leads to a representation of the correlation function 1z m+1
the previously obtained [31] m-multiple integrals. Evaluation of the integral (1.5) in terms
of the poles outside gives us the form factor type expansion of the correlation function
(i.e., an expansion in terms of matrix elements of z between the ground state and all
excited states).
We would like to stress that the objects entering the above master equation for the
finite XXZ chain are quite generic in the context of quantum integrable models solvable
603
by the algebraic Bethe ansatz. Therefore we conjecture that similar formula holds true in
more general situations, in particular for the field theory models. For these models the
approach (i) usually leads to short distance expansions for the correlation functions, while
the approach (ii) gives their long distance expansion. The problem of putting into explicit
(analytic) correspondence these two regimes has been the subject of several works along
the last fifteen years (see, e.g., [4043]), remaining however up to now an open problem.
We hope that the method presented here could ultimately shed some new light on these
topics.
The paper is organized as follows. In the next section we give the main notations and definitions of the XXZ model using the framework of the algebraic Bethe ansatz. In Section 3
the twisted transfer matrix T is introduced. We describe the properties of this operator and
related objects, like twisted Bethe equations, their solutions and the scalar product formulas
for the eigenstates of T with arbitrary vectors. In Section 4 we derive the master equation.
The last two sections are devoted to the above mentioned two evaluations of the multiple contour integral defining this master equation. In Section 5 we reproduce from it the
multiple integral representation for the z correlation function obtained in [31]. This representation is given for the finite chain and in the thermodynamic limit both for the massless
and massive regimes. In Section 6 we obtain the form factor type expansion for the z
correlation function. In the conclusion we give several comments on the key points of our
methods. We also discuss several perspectives of the obtained result and announce the multiple integral representation for the dynamical correlation function of the third components
of spin, which in fact is the subject of a forthcoming publication [49].
2. Algebraic Bethe ansatz for the XXZ chain

In the framework of the algebraic Bethe ansatz the Hamiltonian (1.3) can be obtained
from the monodromy matrix T (), that in its turn is completely defined by the R-matrix.
The R-matrix of the XXZ chain acts in the space C2 C2 and is equal to2
sinh( + )
0
0
0
0
sinh sinh
0
, cosh = .
R() =
(2.1)
0
sinh sinh
0
0
0
0
sinh( + )
It is a solution of the YangBaxter equation. Identifying one of the two vector spaces of
the R-matrix with the quantum space Hm , we obtain the quantum L-operator at the site m
Lm () = R0m ( /2).
(2.2)
Here R0m acts in C2 Hm . Then the monodromy matrix T () is constructed as an ordered

product of the L-operators with respect to all the sites of the chain

A() B()
T () =
= LM () L2 ()L1 ().
(2.3)
C() D()
2 Note that we use here a different normalization for the R-matrix compared to [31].
604
The operator T () acts in V H, where V C2 is usually called auxiliary space of T ().

The Hamiltonian (1.3) at h = 0 is related to T () by a trace identity

dT () 1
+ const.
H = 2 sinh
(2.4)
T ()
d
=/2
Here
T () = tr T () = A() + D().
(2.5)
Later on we shall consider the inhomogeneous XXZ model, for which

Lm () = Lm (, m ) = R0m ( m ),
T () = LM (, M ) L2 (, 2 )L1 (, 1 ),
(2.6)
where m are arbitrary complex numbers attached to each lattice site that are called inhomogeneity parameters. In the homogeneous limit m = /2 and we come back to the
original model.
The commutation relations between the entries of the monodromy matrix are defined
by the YangBaxter quadratic relations,
R12 (1 2 )T1 (1 )T2 (2 ) = T2 (2 )T1 (1 )R12 (1 2 ).
(2.7)
Eq. (2.7) holds in the space V1 V2 H (where Vj C2 ). The matrix Tj () acts in a

non-trivial way in the space Vj H, while the R-matrix is non-trivial in V1 V2 .
In the framework of the algebraic Bethe ansatz an arbitrary quantum state can be obtained from the states generated by the action of the operators B() on the reference state
|0 with all spins up,
| =
N

B(j )|0,
N = 0, 1, . . . , M.
(2.8)
j =1
The eigenstates of the transfer matrix T () can be constructed in the form (2.8), where the
parameters j satisfy the system of Bethe equations
a(j )
N

sinh(k j + ) = d(j )
k=1
k=j
N

sinh(k j ),
j = 1, . . . , N,
k=1
k=j
(2.9)
and a(), d() are the eigenvalues of the operators A() and D() on the reference state.
In the normalization (2.1), (2.2), we have
a() =
M

sinh( a + ),
(2.10)
sinh( a ).
(2.11)
a=1
d() =
M

a=1
605
Remark. Generally speaking the system of equations (2.9) is neither a necessary nor a sufficient condition for the vector (2.8) to be an eigenstate of the transfer matrix. For example,
the state (2.8) with all spins down (N = M) is the eigenstate of T for generic {}. On the
other hand the system (2.9) possesses solutions, which do not correspond to any eigenstate
of T . We discuss all these questions in more details in Section 3 and Appendix A.
The eigenvalue (|{}) of the operator T () corresponding to an eigenstate of the
form (2.8) is
N
N

sinh(k + )
sinh( k + )
+ d()
.
|{} = a()
sinh(k )
sinh( k )
k=1
(2.12)
k=1
The dual states can be constructed similarly to (2.8) via the operators C()
| = 0|
N

C(j ),
N = 0, 1, . . . , M.
(2.13)
j =1
Here 0| = |0+ and the dual eigenstates of T () are given in the form (2.13), where
the parameters j satisfy the same system of equations (2.9). Generically | = | due
If, however, the state (2.8) is the ground state of the
to the involution C() = B ().
N
Hamiltonian (1.3), then | = c | with c = 1 for > 1 and c = 1 for 1 < 1,
respectively.
In the end of this section we give the explicit representations of the local spin operators
in terms of the entries of the monodromy matrix. Such a representation is given by the
solution of the quantum inverse scattering problem [25,27]
j =
j
1
1

T (k ) tr T (j )
T (k ).
k=1
(2.14)
k=1
Here in the r.h.s. acts in the auxiliary space of T (), while j in the l.h.s. acts in the
local quantum space Hj . The formulas of the quantum inverse scattering problem permit us
to embed the problem of calculation of the correlation functions of the local spin operators
into the algebraic Bethe ansatz.
3. Twisted transfer matrix and scalar products

In this section we introduce new objects and notations, which will be used through
all the paper. Starting from this section we consider only the subspace H(M/2N ) of the
quantum space H with fixed (but arbitrary) number of spins down N .
First, we introduce the twisted transfer matrix T ()
T () = A() + D(),
(3.1)
where is a complex parameter. We denote also as | ({}) and ({})| the eigenstates (respectively the dual eigenstates) of the operator T () in the subspace H(M/2N ) .
606
The corresponding eigenvalue is denoted as (|{}). The same notations without a subscript correspond to the case = 1 considered in the previous section.
The eigenstates of T () (and their dual states) have the form
N

{} =
B(j )|0,
N

{} = 0|
C(j ),
j =1
(3.2)
j =1
with an eigenvalue
N
N

sinh(k + )
sinh( k + )
|{} = a()
+ d()
,
sinh(k )
sinh( k )
k=1
(3.3)
k=1
where the parameters {} satisfy the system of twisted Bethe equations

a(j )
N

sinh(k j + ) = d(j )
k=1
k=j
N

sinh(k j ),
j = 1, . . . , N.
k=1
k=j
(3.4)
It is also convenient to introduce the function
N

Y |{} =
sinh(k ) |{}
k=1
= a()
N

sinh(k + ) + d()
k=1
N

sinh(k ).
(3.5)
k=1
In terms of this function the system (3.4) reads

Y j |{} = 0, j = 1, . . . , N.
(3.6)
Just like in the case = 1, for generic not all the solutions of the system (3.6) correspond to eigenvectors of T (). A brief sketch of the properties of the solutions of (3.6) is
given in Appendix A (for details we refer the reader to the original paper [44]). Here we
merely recall some definitions and one of the main results of [44].
Definition 3.1. A solution {} of the system (3.6) is called admissible, if
d(j )
N

sinh(j k + ) = 0,
j = 1, . . . , N,
(3.7)
k=1
k=j
and unadmissible otherwise. A solution is called off-diagonal if the parameters {} are

pair-wise distinct and diagonal otherwise. A solution is called degenerated, if the Jacobian
of the system (3.6) vanishes, and non-degenerated otherwise. A solution is called trivial, if
the corresponding state | ({}) is the zero vector.
607
One of the main result of [44] is that, for generic and { }, the eigenstates of T ()
corresponding to the admissible off-diagonal solutions of (3.6) form a basis in the space
H(M/2N ) . Generically this basis is not normalized and non-orthogonal, although it is orthogonal to the dual basis. We would like to mention that for some specific choice of
and { } certain unadmissible solutions may also contribute to the basis of the eigenstates
in H(M/2N ) .
Consider now the scalar products of eigenstates (3.2) and arbitrary states of the
form (2.8). The explicit results for such scalar products at = 1 were obtained in [25,
45]. Applying the methods used in these papers one can easily prove
Proposition 3.1. Let {} satisfy the system (3.6), {} be generic complex numbers. Then
0|
N

N

C(j ) {} = {}
B(j )|0
j =1
j =1
N

a=1

d(a ) XN1 {}, {} det

k |{} .
N
j
(3.8)
Here XN ({}, {}) is the Cauchy determinant composed of the parameters {} and {}

XN {}, {} = det
N
1
sinh(k j )
N
a>b sinh(a b ) sinh(b a )

.
N
a,b=1 sinh(b a )
(3.9)
To make these formulas more explicit we introduce for arbitrary positive integers n and
n (n n ) and arbitrary sets of variables 1 , . . . , n , 1 , . . . , n and 1 , . . . , n , such that
{} {}, the following n n matrix ({}, {}|{})

n

sinh(a k + )
( )j k {}, {}|{} = a(k ) t (j , k )
a=1

d(k ) t (k , j )
n

sinh(a k ),
(3.10)
a=1
with
t (, ) =
sinh
.
sinh( ) sinh( + )
Then Eq. (3.8) reads
(3.11)
608
0|
N

C(j ) {}
j =1
N

B(j )|0
= {}
j =1
N
= N
a>b sinh(a
a=1 d(a )
b ) sinh(b a )

det {}, {}|{} .
N
(3.12)
Taking the limit {} {} in (3.12) we obtain the square of the norm of the eigenstate
(recall that generically ({})| = | ({}) ) [47,48]
N

d(a )

det {}, {}|{} .
{} {} = N a=1
(3.13)
N
a,b=1 sinh(a b )
a=b
This equation can also be written in terms of a Jacobian using

Y j |{} ,
det {}, {}|{} = det
N
N
k
where {} satisfy the system (3.6).
(3.14)
4. Master equation for z correlation function

In this paper we consider the generating functional for the correlation function of the
z
. Following the papers [39,46] we define for any comthird components of spins 1z m+1
plex number the operator Q1,m as3

m

1+ 1
+
nz .
Q1,m =
(4.1)
2
2
n=1
The generating functional is equal to the expectation value

({})|Q1,m |({})
Q1,m =
,
({})|({})
(4.2)
where |({}) is an eigenstate of T () in the subspace H(M/2N ) . Taking the second

lattice derivative of Q1,m and the second derivative with respect to at = 1 we
extract the two-point correlation function of the third components of local spins:

2
1
z
z
1 1 1 m+1 = 2 Q1,m+1 Q1,m Q2,m+1 + Q2,m .

(4.3)
2
=1
The main result of this paper is an integral representation for the generating functional (4.2).
3 Setting = e one has Q = exp( m (1 z )) and, hence, this is exactly the operator considered
n
n=1
1,m
2
in [31].
609
Theorem 4.1. Let the inhomogeneities { } be generic and the set {} be an admissible offdiagonal solution of the system (2.9). Then there exists 0 > 0 such, that for || < 0 the
expectation value of the operator Q1,m in the inhomogeneous finite XXZ chain is given
by the multiple contour integral

N
m
N

dzj
(a |{z})
1
2
sinh (a zb )
Q1,m =
N!
2i
(a |{})
{ } {} j =1
a,b=1
detN
N
(j |{z})
zk
a=1
detN
a=1 Y (za |{z}) detN
(zk |{})
j
Y (k |{}) .
(4.4)
The integration contour is such that the only singularities of the integrand (4.4) within
{ } {} which contribute to the integral are the points { } and {}.
We call (4.4) the master equation.
Remark 1. The master equation (4.4) gives the expectation value Q1,m with respect to
an arbitrary eigenstate |({}) of T . In particular one can chose {} such that in the
homogeneous limit a /2 the corresponding eigenstate |({}) goes to the ground
state of the XXZ Hamiltonian.
Remark 2. The master equation (4.4) is an integral representation of the expectation value
Q1,m , which holds true at least for || small enough. On the other hand this expectation
value is evidently a polynomial in of degree m. Therefore the representation (4.4) can
be easily continued in from any vicinity of the origin to the whole complex plane. This
does not mean, however, that one can set to be an arbitrary specific value directly in the
integrand of (4.4).
Proof of Theorem 4.1. In order to compute the expectation value of Q1,m , we express
this operator in terms of the twisted transfer matrix. Due to (2.14) one has
Q1,m =
m

T (a )
a=1
m

T 1 (b ).
(4.5)
b=1

Then we can use the explicit formula for the multiple action of m
a=1 T (xa ) for an arbitrary
N
set of complex numbers {x} on an arbitrary state 0| j =1 C(j ) [31]
0|
N

C(j )
m

j =1
a=1
min
(m,N )
n=0
0|
T (xa )

Rn {x+ }, {x }, {+ }, { }
{}={+ }{ }
{x}={x+ }{x }
|+ |=|+ |=n
a+
C(xa )
C(b ),
(4.6)
610
where the coefficient Rn ({x+ }, {x }, {+ }, { }) is given by

Rn =
sinh(b a )
a>b
a,b+
sinh(xb xa )

1
sinh(b a )
a+ b
a<b
a,b+

xa |{x+ } { } det {x+ }, {+ }|{x+ } { } .
n
(4.7)
Here ({x+ }, {+ }|{x+ } { }) is given by expressions (3.10) in which the sets {},
{}, {} have to be replaced by {x+ }, {+ } and {x+ } { }, respectively.
The summation in (4.6) is taken with respect to partitions of the sets {} and {x} into
disjoint subsets {} = {+ } { } and {x} = {x+ } {x }, such that the number of
elements in the subsets {+ } and {x+ } coincides and is equal to n.
In the paper [31] we directly applied Eq. (4.6) for the computation of Q1,m , specifying xj = j for j = 1, . . . , m. The peculiarity of the inhomogeneity parameters is that
d(j ) = 0, what simplifies the formulas. In the present paper, however, we are going to
keep the parameters {x} arbitrary, i.e., we shall consider the expectation value of the operator
m
m

T (xa )
T 1 (xb ),
Qm , {x} =
a=1
(4.8)
b=1
where the parameters x1 , . . . , xm are generic complex numbers.

The normalized expectation value of Qm (, {x}) on the finite lattice with respect to an
arbitrary eigenstate of the transfer matrix |({}) has the form
m

1 (x )|({})

({})| m
b
a=1 T (xa ) b=1 T
Qm , {x} =
(4.9)
.
({})|({})
The action to the right of the product of T 1 (xb ) produces merely a numerical factor:
m

m

T 1 (xb ) {} =
1 xb |{} {} .
b=1
(4.10)
b=1
Acting to the left with the product of T (xa ) by means of (4.6) with m < N , and using the
expression (3.12) of the scalar product at = 1, we obtain
m

Qm , {x} =

n {x+ }, {x }, {+ }, { } ,
n=0 {}={+ }{ }
{x}={x+ }{x }
|+ |=|+ |=n
with
n =
1
m

xb |{}
sinh(a xb )
sinh(xa xb )
b=1
a+
a=b
(4.11)
611

Y xa |{x+ } { }
detN ({}, {x+ } { }|{})

.
det {x+ }, {+ }|{x+ } { }
n
detN ({}, {+ } { }|{})
(4.12)
Here Y and without subscript are equal, respectively, to Y and at = 1. The
elements in the sets {x+ } { } and {+ } { } are ordered accordingly.
Like in [31], we can now perform a re-summation over the partitions of the set {x} by
introducing contour integrals over auxiliary variables z1 , . . . , zn :

Qm , {x}

m
m
n

dzj
1
1
=
sinh (b xa )
n!
2i
n=0 {}={+ }{ } a=1 b
|+ |=n
n
m

a=1 b=1
{x} j =1
m
n

Y (xa |{z} { })
1
1
sinh(za xb )
(xa |{})
Y (za |{z} { })
a=1
a=1

detN ({}, {z} { }|{})
det {z}, {+ }|{z} { }
,
n
detN ({}, {+ } { }|{})
(4.13)
where the closed contour {x} surrounds the
x1 , . . . , xm and do not contain any
other pole of the integrand, i.e., zeros in Cn of the functions Y (za |{z} { }). Since at
this stage of the computations x1 , . . . , xm are generic complex numbers, we can always
choose them separated from the zeros of Y (za |{z} { }) for any subset { }.
Thus, the sum over partitions of the set {x} in (4.11) is replaced with a contour integral.
In paper [31] we replaced in the thermodynamical limit the sum over partitions of the set
{} with the integrals over the support of the spectral density of the ground state. Here
we treat this sum in a different way. Namely, we perform a second re-summation over the
partitions of the set {} in (4.13) also in terms of a contour integral. Indeed, one can easily
see that

Res
det {z1 , . . . , zN }, {+ } { }|{z1 , . . . , zN }
{zn+1 ,...,zN }={ } N

=
Y a |{z1 , . . . , zn } { }
points4

det {z1 , . . . , zn }, {+ }|{z1 , . . . , zn } { } .
n
(4.14)
This enables us to express the generating functional Qm (, {x}) for the finite XXZ chain
as a single multiple integral:
4 More precisely, {x} is the boundary of a set of polydisks D (r) in Cn . Namely, {x} = m D
a
a=1 a (r),
n
where Da (r) = {z C : |zk xa | = r, k = 1, . . . , n}. The integration contour { } {} in (4.4) should be

understood in a similar manner in CN .
612

Qm , {x}
1
=
N!
m
N
N

dzj (xa |{z})
1
2i
(xa |{})
Y (za |{z})
{x} {} j =1
a=1
a=1

detN ({}, {z}|{})
,
det {z}, {}|{z}
N
detN ({}, {}|{})
(4.15)
where the closed contour {x} {} surrounds the points x1 , . . . , xm and 1 , . . . , N

(spectral parameters corresponding to given eigenstate of T ()) and no other poles of the
integrand.
Finally, in order to reproduce the expectation value Q1,m , we should set the generic
parameters {x} in (4.15) to be equal to the inhomogeneities 1 , . . . , m . This definitely can
be done if the system

Y zj |{z} = 0,
j = 1, . . . , N
(4.16)
has no solution inside the integration contour { } {}. Thus, we need to analyze the
properties of the solutions of the system of the twisted Bethe equations. Hereby, since
Qm (, {x}) is a polynomial in , it is sufficient to determine this polynomial in an open
ball around some specific value, for example, for || small enough.
The detailed analysis of the system (4.16) was done in the paper [44]. We give some
of the basic statements of this paper in Appendix A. Here we present only the results
necessary for writing and evaluating the integral (4.15).
Lemma 4.1. Let the inhomogeneities 1 , . . . , m be generic, 1 , . . . , N be an admissible
off-diagonal solution of (2.9) and || be small enough. Then all admissible off-diagonal
solutions of the system (4.16) are separated from the points { } and {}.
Corollary 4.1. There exists a contour { } {} such that the points { } and {} are
inside this contour, while all admissible off-diagonal solutions of the system (4.16) are
outside { } {}.
Lemma 4.2. Let the contour { } {} satisfy the conditions of the Corollary 4.1, and
xk k . Then:
(1) the only poles inside the contour { } {} which provide non-vanishing contribution to the integral (4.15) are in { } {};
(2) the only poles outside the contour { } {} which provide non-vanishing contribution to the integral (4.15) are the admissible off-diagonal solutions of the system (4.16).
The proof of these lemmas is given in Appendix A using the results of [44].
Lemma 4.2 guarantees that for xk = k there are no any other poles of the integrand
(4.15) inside the contour { } {} and contributing to the integral except zj = k and
613
zj = k . Thus, setting xk = k and using d(k ) = 0 we arrive at

1
Q1,m =
N!

N m
N

dzj sinh(a b ) sinh(za b + )
2i
sinh(za b ) sinh(a b + )
{ } {} j =1
a=1 b=1
detN ({z}, {}|{z}) detN ({}, {z}|{})

.
N
a=1 Y (za |{z}) detN ({}, {}|{})
(4.17)
It remains to express the determinants of the matrices and in terms of Jacobians

using (3.8)(3.14), and we obtain the master equation (4.4). 2
5. Multiple integral representation

Lemma 4.2 permits us to evaluate the integral (4.4) in two different ways: either to
compute the residues of the integrand inside the contour { } {}, or to compute the
residues outside this contour. In this section we consider the first way, which immediately
leads us to the representation for Q1,m obtained in [31].
For this purpose it is more convenient to use (4.17), which in fact is equivalent to (4.4).
This multiple integral can be presented as

N

{ } {} j =1
dzj =
N

n=0

n
n
CN
{ } j =1

dzj
N
n
dzj .
(5.1)
{} j =1
Hereby, since the number of the poles inside { } is m and the integrand vanishes as soon
as zj = zk , the sum in (5.1) is actually restricted to n m. Evaluating N n integrals in
the points { } we arrive at (4.13) with xk = k , and hence d(xk ) = 0. Moreover, we can
also set d(zk ) = 0, since the remaining integrals surround only the poles at zk = j and
after evaluation of these integrals the functions d(zk ) vanish. Finally, using the fact that the
set {} satisfies the system (2.9) we obtain

n
m

dzj
1
b U (b |{z}, { }, { })
Q1,m =
n +
n!
2i
b=1 U (zb |{z}, { }, { })
n=0 {}={+ }{ }
|+ |=n
{ } j =1

detN ({}, {z} { }|{})|d(z)=0
det M {+ }, {z}
,
n
detN ({}, {+ } { }|{})
(5.2)
where

U |{z}, { }, { }
= a()
n

a=1
and
sinh(za + )

a
sinh(a + )
m

a=1
sinh( a )
,
sinh( a + )
(5.3)
614

(M )j k {}, {z}
= t (zk , j ) + t (j , zk )
n

sinh(a j + ) sinh(j za + )
.
sinh(j a + ) sinh(za j + )
(5.4)
a=1
Observe that, unlike in the master equation, the integrand in the r.h.s. of (5.2) is a polynomial in , since enters only the determinant of the matrix M (5.4). Therefore one can
set to be an arbitrary complex number directly in the integral representation (5.2).
This representation also allows one to proceed, if necessary, to the homogeneous limit
simply by setting k /2. Finally, Eq. (5.2) is convenient for taking the thermodynamic
limit N, M N/M = const. We refer the reader for the details to the paper [31]
and here simply recall the final result for the ground state expectation value of Q1,m
in the thermodynamic limit and in an external magnetic field in the massive and massless
regimes:
m

Q1,m =
n=0
1
(n!)2

n
n
m

dzj
sinh(b a ) sinh(zb a + )
d n
2i
sinh(zb a ) sinh(b a + )
{ } j =1
b=1 a=1
n
n

sinh(a zb + ) sinh(zb a + )
sinh(a b + ) sinh(za zb + )
a=1 b=1

det M {}|{z} det (j , zk ) .
n
(5.5)
Here the function (, z) is the inhomogeneous spectral density of the ground state,
having the support on the contour C as defined in [31].
6. Form factor expansion

Evaluating the integral (4.4) by the residues outside the integration contour we arrive at
the expansion over form factors for Q1,m . Recall that due to Lemma 4.2, the only poles
outside the contour { } {} which contribute to the integral (4.4), are the admissible
off-diagonal solutions of (4.16). Hence
N
m

(a |{})
Q1,m = (1)N
sinh2 (a b )
(a |{})
{} a,b=1
a=1

k |{})
detN (
detN
j
k

,
k |{})
k |{})
detN Y (
detN Y (
j
j
(j |{})
(6.1)
where the sum is taken with respect to all admissible off-diagonal solutions 1 , . . . , N
of the system (4.16). Due to Theorem A.1 the Jacobian matrix Y (k |{})/j is nondegenerated.
615
Using the formula (3.8) we identify each Jacobian in (6.1) with the corresponding scalar
products, leading to
m

(a |{}) ({})| ({}) ({})|({})
Q1,m =
.
(a |{}) ({})| ({}) ({})|({})
(6.2)
{} a=1
It remains to use that the state | ({}) is the eigenstate of T ( ) with the eigenvalue
( |{}) and the state |({}) is the eigenstate of T ( ) with the eigenvalue ( |{}).
Thus, Eq. (6.2) can be written in the form

m
1 ( )|({})
({})| m
b
b=1 T (b )| ({}) ({})| b=1 T
.
Q1,m =
({})| ({}) ({})|({})
{}
(6.3)
Observe that we did not use the statement (b) of Theorem A.1 on the completeness of the
set | ({}). The sum over eigenstates of T appears automatically as the result of the
evaluation of the multiple integral (4.4) by the residues outside the integration contour.
Taking the second lattice derivative of (6.1) and then differentiating twice with respect
to at = 1 (see Appendix B) we obtain the form factor type expansion directly for the
correlation function of the third components of the spin
z
({})|1z |({}) ({})|m+1
|({})
z z
z
1z m+1
,
= 1 m+1 +
({})|({}) ({})|({})
{}={}
(6.4)
where the form factors of z are given by (B.11). Observe that here the summation is
taken with respect to the eigenstates of the operator T , but not T . In other words, in the
homogeneous limit, this is the sum over the excited states of the Hamiltonian.
Of course, Eq. (6.4) might be obtained by means of inserting the complete set of the
eigenstates of the Hamiltonian between local spin operators. Such a representation, however, would be quite formal without a precise description of the set of the states on which
the summation has to be performed. On the one hand it is clear that the sum should be
taken with respect to solutions of the system (2.9), but on the other hand it is also clear that
not any solution of this system corresponds to an eigenstate of the Hamiltonian. Note that
one cannot simply restrict the sum in (6.4) to the admissible off-diagonal solutions, since
in the homogeneous limit certain unadmissible solutions of the system (2.9) give rise to the
basis of the eigenstates.
The master equation approach gives the way to overcome these difficulties. Namely, if
we use the approach (1.2), we can insert the set of the eigenstates of the twisted transfer
matrix between local operators. Then the summation in (6.4) is restricted to the admissible
off-diagonal solutions of the system (3.6). The unadmissible solutions corresponding to the
eigenstates of the Hamiltonian appear in this case only as the limit of admissible solutions
at 1. Since for || small enough all admissible solutions are in the vicinities of the
points { }, the sum over excited states can be written as a multiple integral with respect
to a contour surrounding these points. This allows one to obtain the master type equation
directly for the spinspin correlation functions, starting form their form factor expansions.
616
7. Conclusion
The main result of this paper is the master equation (4.4). We have shown that this
equation draws a link between multiple integral representations and form factor expansions
for the correlation functions.
The master equation sheds a new light on the role of the auxiliary contour integrals in the
representations for the correlation functions derived in [31]. Originally they appeared as a
result of a resummation of the elementary blocks obtained in [22,23,26]. In the framework
of the master equation approach the same auxiliary contour integrals are equivalent to the
sum over excited states.
Using the master equation method one can derive multiple integral representations for
other correlation functions of the XXZ model, including many-point correlators. Indeed,
an arbitrary correlation function in the finite chain can be reduced to the multiple sum
over a complete set of the eigenstates of the twisted transfer matrix. On the other hand
the explicit formulas for the form factors of the local spin operators obtained in [25] can
be easily generalized for the case when one or both states are eigenstates of the twisted
transfer matrix. Presenting this sum as a contour integral of the type (4.4) and evaluating
this integral by the residues inside this contour we can express an arbitrary correlation
function of the XXZ model as a multiple integral of the form (5.5).
Moreover, we conjecture that representations similar to (4.4) should exist for the correlation functions of other models solvable by the algebraic Bethe ansatz. This conjecture
is based on the fact that the integrand of (4.4) depends mostly on the eigenvalues of the
twisted transfer matrix, which is a typical object for the algebraic Bethe ansatz. As we
have mentioned already in introduction, it would be very desirable to obtain an analog of
this master equation in the case of the field theory models. For these models such an integral representation could give an analytic link between short distance and long distance
expansions of the correlation functions.
Finally, the master equation method opens a way to obtain multiple integral representations for time-dependent correlation functions. This subject will be considered in our
forthcoming publication [49]. Here we only announce a generalization of the representation (5.5) for the time-dependent case in the massive and massless regimes

2

z
z
2
,
(t) = 2 1z (0) 1 + 2Dm
Q
(m,
t)
1 (0)m+1
(7.1)

2
=1
2 means the second lattice derivative and
where Dm
Q (m, t)

1
n
=
d
(n!)2
n=0
{ 2 }
n
n

dzj sinh(a zb + ) sinh(zb a + )
2i
sinh(a b + ) sinh(za zb + )
j =1
a,b=1
n

eit (E(zb )E(b ))+im(p(zb )p(b )) det M {}|{z}
b=1

det Rn j , zk |{}, {z} .
(7.2)
617
Here M is given by (5.4). The functions E() and p() are the bare one-particle energy
and momentum

sinh( 2 )
2 sinh2
,
p()
=
i
log
.
E() =
(7.3)
sinh( + 2 ) sinh( 2 )
sinh( + 2 )
The contour {/2} surrounds the points /2 and does not contain any other singularities of the integrand. The function Rn (, z|{}, {z}), as a function of z and other arguments
fixed, has a cut between the points z = /2 and z = /2. Therefore it is defined differently in the vicinities of these points

(, z), z /2,

Rn , z|{}, {z} =
b +) sinh(zb z+)
1(, z + ) nb=1 sinh(z
sinh(b z+) sinh(zzb +) , z /2,
(7.4)
where (, z) is the inhomogeneous spectral density of the ground state.
Observe that the time and the distance dependence of the generating function (7.2) are
associated to the bare energy E() and momentum p(), making Eq. (7.2) very suggestive.
In the limit t = 0 the integrals over {/2} vanish, and we arrive at the time-independent
representation (5.5).
Acknowledgements
J.M.M., N.S. and V.T. are supported by CNRS. N.K., J.M.M., V.T. are supported by
the European network EUCLID-HPRNC-CT-2002-00325. J.M.M. and N.S. are supported
by INTAS-03-51-3350. N.S. is supported by the FrenchRussian Exchange Program, the
Program of RAS Mathematical Methods of the Nonlinear Dynamics, RFBR-02-01-00484,
Scientific Schools 2052.2003.1. N.K., N.S. and V.T. would like to thank the Theoretical
Physics group of the Laboratory of Physics at ENS Lyon for hospitality, which makes this
collaboration possible.
Appendix A. Solutions of the twisted Bethe equations
In this appendix we prove Lemmas 4.1, 4.2 using the results of [44]. Consider the system
of equations (4.16)

Y zj |{z} = 0, j = 1, . . . , N.
(A.1)
Without loss of generality we can identify two solutions of this system {z} and {z } if they
are equal modulo i .
It was proved in [44] that
Theorem A.1 [44]. Let and inhomogeneities { } be generic. Then
(a) All admissible off-diagonal solutions of the system (A.1) are non-degenerated.
(b) The set of the states (3.2) corresponding to the admissible off-diagonal solutions form
a basis in the subspace H(M/2N ) .
618
The strategy of [44] was to consider the limit of the system (A.1) at 0 and then
to deform the solutions {z(0)} to the case = 0. Then the statement (a) follows from the
implicit function theorem. Indeed, the solutions of the system (A.1) at = 0 are evident.
In particular all admissible solutions have form zj (0) = pj . One can easily check that
the Jacobian matrix Y (zj |{z})/zk at = 0 for admissible solutions is a diagonal matrix
with non-vanishing entries. Hence, in this case the solution {z()} is a holomorphic deformation of the solution {z(0)}, and therefore for || small enough all admissible solutions
are in the vicinities of { }. On the other hand it is clear that any admissible solution at
= 1 is separated from the admissible solutions at = 1. Thus, admissible solutions are
separated from the points { } and {}. This proves Lemma 4.1.
In order to prove Lemma 4.2 we use the properties of the matrix ({}, {}|{}) (3.10)
formulated in the following two lemmas. Let the parameters 1 , . . . , n and 1 , . . . , n be
generic complex numbers, be arbitrary complex.
Lemma A.1. Suppose there exist a , b {}, such that a = p , b = p , where p
is one of the inhomogeneity parameters. Then detn ({}, {}|{}) = 0.
Proof. We have d(a ) = a(b ) = 0 and t (b , j ) = t (j , a ). Thus, the columns ( )j a
and ( )j b are proportional to each other, hence detn = 0. 2
Lemma A.2. Let q 2 = e2 be a root of unity, i.e., = iQ/P , where Q < P are positive
integers. Suppose there exists {a1 , . . . , aP } {}, such that
sinh(aj +1 aj + ) = 0,
with aP +1 a1 .
(A.2)
Then detn ({}, {}|{}) = 0.

Proof. It is easy to see that
P

j =1
t (, aj ) =
P
t (aj , ) = 0,
(A.3)
j =1
since both these sums are i -periodical holomorphic functions of vanishing at

. Hence the matrix ({}, {}, {}) contains linearly depended lines and its determinant vanishes. 2
Note that if the set {} contains several subsets of the type (A.2), then the order of the
zero of detn ({}, {}|{}) is not less than the number of such subsets.
If q 2 is not root of unity, then unadmissible solutions of the system (A.1) contain a
pair za = p and zb = p . Off-diagonal unadmissible solutions are non-degenerated
at = 0 [44]. Hence, due to Lemma A.1 they do not give a contribution to the integrals
(4.17), (4.15). The diagonal solutions {z()} of (A.1) obtained as a deformation of diagonal
solutions {z(0)} at = 0 preserve their multiplicity [44]. Hence, they also do not give any
contribution to the integrals (4.17), (4.15) due to vanishing of detN and detN as soon
as two or more z coincide.
If q 2 is a root of unity, then unadmissible solutions may contain strings {zaj } satisfying
the condition (A.2). Such solutions are not isolated since the value of za1 is not fixed. We
619
have seen however that due to Lemma A.2 the determinant of the matrix ({z}, {}|{z})
vanishes exactly on these string type solutions. Hence, these unadmissible solutions of the
system (A.1) do not contribute to the integrals (4.17), (4.15).
Thus, the only solutions of the system (A.1) which can give non-vanishing contribution to the integrals (4.17), (4.15) are admissible off-diagonal solutions, which are in the
vicinities of { } for || small enough. This proves Lemma 4.2.
Appendix B. The z form factor
The explicit formulas for form factors of local spin operators in the finite XXZ chain
were obtained in [25]. Here we propose a slightly modified method to derive the form
factor of the operator z .
Consider a matrix element of the operator Q1,m between an eigenstate |({}) of the
transfer matrix T () (for example, ground state) and an eigenstate ({})| of the operator T (). Using (4.5) we immediately obtain
m

(a |{()})
() Q1,m {} =
() {} .
(a |{})
(B.1)
a=1
Here we have indicated explicitly that the parameters {} depend on , for they are solutions of the system (3.6). On the other hand it is clear that

Q1,m+1 Q1,m {}
()
=1

1
z
{} .
= (1) 1 m+1
(B.2)
2
Thus, we have

1
z
{}
(1) 1 m+1
2

m
(a |{()})
(m+1 |{()})
=
(a |{})
(m+1 |{})
a=1
N
a=1 d(a ())
N
a>b sinh(a b ) sinh(b () a ())

det () , {} () .
(B.3)
=1
In order to evaluate explicitly the derivative over in (B.3) one should distinguish two
cases: {(1)} = {} and {(1)} = {}. In the first case (m+1 |{()}) (m+1 |{}) as
1, therefore

z

)|({})
({})|(1 m+1
(m+1 |{()})
=2
1
(B.4)
({})|({})
(m+1 |{})
=1

N

dk ()
= 2
t (k , m+1 ).
(B.5)
d =1
k=1
620
The derivatives dk ()/d can be found from (3.4) via

N
N

Y (j |{}) dk ()
+ d(j )
sinh(a j ) = 0.
k
d =1
k=1
(B.6)
a=1
In the second case we can compute explicitly the derivative of det . Indeed, consider
the N -dimensional vector-column v with the components
vk =
N

N

sinh k () a
sinh1 k () a () .
a=1
(B.7)
a=1
a=k
If {(1)} = {}, then this vector at = 1 has at least one non-zero component, say vN = 0.
Then multiplying the kth column of ( )j k by vk /vN and adding the first (N 1) columns
to the last one, we obtain
( )j N +
N
1

k=1

1

vk
1
( )j k =
Y j |{} =
a(j )
sinh(a j + )
vN
vN
vN
N
a=1
(B.8)
(see Appendix B of [31] for the proof). Thus, the last column is proportional to 1 ,
hence, taking the derivative over of det one has to differentiate only this column.
This gives

det () , {}| ()
= det (1) , {}| (1) ,
(B.9)
N
N
=1
where

j k {}, {}|{} =
()
( )j k ({}, {}|{})|=1 , k = 1, . . . , N 1,

v1N a(j ) N
a=1 sinh(a j + ), k = N.
(B.10)
Thus, for {} = {} we obtain

z
{} m+1 {}

m

(m+1 |{})
(a |{})
=2
1
(a |{})
(m+1 |{})
a=1
N

a=1 d(a )
det {}, {}|{} .
N
N
a>b sinh(a b ) sinh(b a )
(B.11)
References
[1] R.J. Baxter, Exactly Solved Models in Statistical Mechanics, Academic Press, London, 1982.
[2] L.D. Faddeev, in: J.B. Zuber, R. Stora (Eds.), Recent Advances in Field Theory and Statistical Mechanics,
Les Houches 1982, Elsevier, Amsterdam, 1984, p. 561.
621
[3] M. Gaudin, La Fonction dOnde de Bethe, Masson, Paris, 1983.

[4] V.E. Korepin, N.M. Bogoliubov, A.G. Izergin, Quantum Inverse Scattering Method and Correlation Functions, Cambridge Univ. Press, Cambridge, 1993.
[5] L. Onsager, Phys. Rev. 65 (1944) 117.
[6] E. Lieb, T. Shultz, D. Mattis, Ann. Phys. 16 (1961) 407.
[7] B.M. McCoy, Phys. Rev. 173 (1968) 531.
[8] T.T. Wu, B.M. McCoy, C.A. Tracy, E. Barouch, Phys. Rev. B 13 (1976) 316.
[9] B.M. McCoy, C.A. Tracy, T.T. Wu, Phys. Rev. Lett. 38 (1977) 793.
[10] M. Sato, T. Miwa, M. Jimbo, Publ. Res. Int. Math. Sci. 14 (1978) 223;
M. Sato, T. Miwa, M. Jimbo, Publ. Res. Int. Math. Sci. 15 (1979) 201;
M. Sato, T. Miwa, M. Jimbo, Publ. Res. Int. Math. Sci. 16 (1980) 531.
[11] A.A. Belavin, A.M. Polyakov, A.B. Zamolodchikov, Nucl. Phys. B 241 (1984) 333.
[12] M. Karowski, P. Weisz, Nucl. Phys. B 139 (1978) 455.
[13] F.A. Smirnov, Form Factors in Completely Integrable Models of Quantum Field Theory, Advanced Series
in Mathematical Physics, vol. 14, World Scientific, Singapore, 1992.
[14] H. Babudjan, A. Fring, M. Karowski, A. Zapletal, Nucl. Phys. B 538 (1999) 535.
[15] W. Heisenberg, Z. Phys. 49 (1928) 619.
[16] H. Bethe, Z. Phys. 71 (1931) 205.
[17] R. Orbach, Phys. Rev. 112 (1958) 309.
[18] L.R. Walker, Phys. Rev. 116 (1959) 1089.
[19] C.N. Yang, C.P. Yang, Phys. Rev. 150 (1966) 321;
C.N. Yang, C.P. Yang, Phys. Rev. 150 (1966) 327.
[20] L.D. Faddeev, E.K. Sklyanin, L.A. Takhtajan, Theor. Math. Phys. 40 (1980) 688.
[21] L.A. Takhtajan, L.D. Faddeev, Russian Math. Surveys 34 (1979) 11.
[22] M. Jimbo, K. Miki, T. Miwa, A. Nakayashiki, Phys. Lett. A 168 (1992) 256.
[23] M. Jimbo, T. Miwa, J. Phys. A: Math. Gen. 29 (1996) 2923.
[24] M. Jimbo, T. Miwa, Algebraic Analysis of Solvable Lattice Models, American Mathematical Society, 1995.
[25] N. Kitanine, J.M. Maillet, V. Terras, Nucl. Phys. B 554 (1999) 647, math-ph/9807020.
[26] N. Kitanine, J.M. Maillet, V. Terras, Nucl. Phys. B 567 (2000) 554, math-ph/9907019.
[27] J.M. Maillet, V. Terras, Nucl. Phys. B 575 (2000) 627, hep-th/9911030.
[28] A.G. Izergin, N. Kitanine, J.M. Maillet, V. Terras, Nucl. Phys. B 554 (1999) 679.
[29] H. Boos, V. Korepin, F. Smirnov, J. Phys. A 37 (2004) 323.
[30] H. Boos, M. Jimbo, T. Miwa, F. Smirnov, Y. Takeyama, A recursion formula for the correlation functions of
an inhomogeneous XXX model, hep-th/0405044.
[31] N. Kitanine, J.M. Maillet, N.A. Slavnov, V. Terras, Nucl. Phys. B 641 (2002) 487, hep-th/0201045.
[32] N. Kitanine, J.M. Maillet, N.A. Slavnov, V. Terras, Nucl. Phys. B 642 (2002) 433, hep-th/0203169.
[33] F. Ghmann, A. Klumper, A. Seel, Integral representations for the correlation functions of the XXZ chain
at finite temperature, hep-th/0405089.
[34] K. Miki, Phys. Lett. A 186 (1994) 217.
[35] M. Jimbo, T. Kojima, T. Miwa, Y.H. Quano, J. Phys. A 27 (1994) 3267.
[36] T. Kojima, F. Miki, Y.H. Quano, J. Phys. A 28 (1995) 3479.
[37] Y.H. Quano, J. Phys. A 31 (1998) 1791.
[38] O. Babelon, D. Bernard, Phys. Lett. B 288 (1992) 113.
[39] F. Colomo, A.G. Izergin, V.E. Korepin, V. Tognetti, Theor. Math. Phys. 94 (1993) 11.
[40] Al.B. Zamolodchikov, Nucl. Phys. B 348 (1991) 619.
[41] Al.B. Zamolodchikov, Int. J. Mod. Phys. A 10 (1995) 1125.
[42] H. Babudjan, M. Karowski, Towards the construction of Wightman functions of integrable quantum field
theories, hep-th/0301088.
[43] A.A. Belavin, A.V. Belavin, A.V. Litvinov, Y.P. Pugai, Al.B. Zamolodchikov, Nucl. Phys. B 676 (2004) 587.
[44] V. Tarasov, A. Varchenko, Int. Math. Res. Notices 13 (1996).
[45] N.A. Slavnov, Theor. Math. Phys. 79 (1989) 502.
[46] A.G. Izergin, V.E. Korepin, Commun. Math. Phys. 99 (1985) 271.
622
[47] M. Gaudin, B.M. McCoy, T.T. Wu, Phys. Rev. D 23 (1981) 417.
[48] V.E. Korepin, Commun. Math. Phys. 86 (1982) 391.
[49] N. Kitanine, J.M. Maillet, N.A. Slavnov, V. Terras, Dynamical correlation functions of the XXZ spin- 12
chain, hep-th/0407108, Nucl. Phys. B, in press.
CUMULATIVE AUTHOR INDEX B711B712
Akemann, G.
Aulakh, C.S.
B712 (2005) 287

B711 (2005) 275
Bak, D.
Bedford, J.
Bernreuther, W.
Bonciani, R.
Boos, H.E.
Brandhuber, A.
Buchbinder, E.I.
Buchbinder, I.L.
Buchmller, W.
B712 (2005) 115

B712 (2005) 59
B712 (2005) 229
B712 (2005) 229
B712 (2005) 573
B712 (2005) 59
B711 (2005) 314
B711 (2005) 367
B712 (2005) 139
Chang, J.-F.
Choi, S.Y.
B712 (2005) 347

B711 (2005) 83
DAppollonio, G.
de Vega, H.J.
Dobashi, S.
Dobashi, S.
B712 (2005) 433

B711 (2005) 604
B711 (2005) 3
B711 (2005) 54
Eden, B.
Essler, F.H.L.
B712 (2005) 157

B712 (2005) 513
Frahm, H.
B712 (2005) 513
Gehrmann, T.
Girdhar, A.
Giuliano, D.
B712 (2005) 229

B711 (2005) 275
B711 (2005) 480
Hamaguchi, K.
Heinesch, R.
Hung, P.Q.
Hyakutake, Y.
B712 (2005) 139

B712 (2005) 229
B712 (2005) 325
B712 (2005) 115
0550-3213/2005 Published by Elsevier B.V.

doi:10.1016/S0550-3213(05)00222-1
Janssen, B.
Janssen, B.
Jarczak, C.
B711 (2005) 392

B712 (2005) 371
B712 (2005) 157
Kawai, H.
Kim, S.
Kimura, T.
Kiritsis, E.
Kitanine, N.
Kniehl, B.A.
Krs, B.
Kramer, G.
Krykhtin, V.A.
Kuroki, T.
B711 (2005) 253

B712 (2005) 115
B711 (2005) 163
B712 (2005) 433
B712 (2005) 600
B711 (2005) 345
B711 (2005) 112
B711 (2005) 345
B711 (2005) 367
B711 (2005) 253
Lebedev, O.
Leineweber, T.
Lozano, Y.
Lozano, Y.
B712 (2005) 139

B712 (2005) 229
B711 (2005) 392
B712 (2005) 371
Maillet, J.M.
Maniatis, M.
Marchesano, F.
Martins, M.J.
Masiero, A.
Mastrolia, P.
Matsuo, Y.
Melo, C.S.
Miller, D.J.
B712 (2005) 600

B711 (2005) 345
B712 (2005) 20
B711 (2005) 565
B712 (2005) 86
B712 (2005) 229
B711 (2005) 253
B711 (2005) 565
B711 (2005) 83
Nath, P.
Nirschl, M.
Nitta, M.
B711 (2005) 112

B711 (2005) 409
B711 (2005) 133
Ohta, N.
Osborn, H.
Osborn, J.C.
B712 (2005) 115

B711 (2005) 409
B712 (2005) 287
624
Palomares-Ruiz, S.
Pashnev, A.
Petcov, S.T.
Pilo, L.
Pinnow, H.A.
Polychronakos, A.P.
Profumo, S.
B712 (2005) 392

B711 (2005) 367
B712 (2005) 392
B712 (2005) 3
B711 (2005) 530
B711 (2005) 505
B712 (2005) 86
Slavnov, N.A.
Sodano, P.
Soddu, A.
Sokatchev, E.
Spence, B.
Splittorff, K.
Spradlin, M.
B712 (2005) 600

B711 (2005) 480
B712 (2005) 325
B712 (2005) 157
B712 (2005) 59
B712 (2005) 287
B711 (2005) 199
Quirs, M.
B712 (2005)
Ratz, M.
Remiddi, E.
Ribeiro, G.A.P.
Riccioni, F.
Riotto, A.
Rodrguez-Gmez, D.
Rodrguez-Gmez, D.
B712 (2005) 139

B712 (2005) 229
B711 (2005) 565
B711 (2005) 231
B712 (2005) 3
B711 (2005) 392
B712 (2005) 371
Takahashi, M.
Terras, V.
Tran, N.-K.
Travaglini, G.
B712 (2005) 573

B712 (2005) 600
B712 (2005) 325
B712 (2005) 59
Ullio, P.
B712 (2005) 86
Verbaarschot, J.J.M.
Volovich, A.
von Gersdorff, G.
B712 (2005) 287

B711 (2005) 199
B712 (2005) 3
Saharian, A.A.
Saleur, H.
Snchez, N.G.
Sanz, V.
Shen, Y.-G.
Shiroishi, M.
Shiu, G.
Shore, G.M.
B712 (2005) 196

B712 (2005) 513
B711 (2005) 604
B712 (2005) 3
B712 (2005) 347
B712 (2005) 573
B712 (2005) 20
B712 (2005) 411
Wang, L.-T.
Wiese, K.J.
B712 (2005) 20
B711 (2005) 530
Yoneya, T.
Yoneya, T.
B711 (2005) 3
B711 (2005) 54
Zerwas, P.M.
B711 (2005) 83

Nucl - Phys.B v.712

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nucl - Phys.B v.712

Uploaded by

Copyright:

Available Formats

Nuclear Physics B 712 (2005) 319

Supersymmetry from boundary conditions

E-08193 Bellaterra (Barcelona), Spain

Received 18 November 2004; accepted 5 January 2005

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

2. Fermions in the interval

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

These equations only have trivial solutions (are over-constrained) unless

where cf1,2 are complex parameters or, setting zf = (cf1 /cf2 )

Physically inequivalent BCs span a complex projective space CP 1 homeomorphic to the

where the second equality holds due to condition (2.9).

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

As a consequence of the transformation (2.13) the SS parameter manifests itself only in

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

M11 = (q1 + iq2 ),

M22 = (q1 iq2 )

Consider now f = , using an SU(2) transformation we can always rotate v in the

3. Super-YangMills action in the interval

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

To compensate for this we add the brane action

which transforms into

Now the supersymmetry variation at each boundary is proportional to (1 + i 5 T (f ) )(yf ).

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

4. Fermions in the orbifold

2 (x, y) = 2 (x, y),

1 (x, R y) = 1 (x, R + y),

where = 1. The second condition is often replaced by demanding periodicity ( = +1)

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

fixed point at y = yf (f = 0, ) they behave as12

where 0 = (y),  = (R y) and the sign function (y) is defined as

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

+ 2 N11 1 + N12 2 (y) = 0.

assumption of only 1 smooth is inconsistent unless N22 = N12 = 0 or N11 = N12 = 0.

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

Similar expressions at are obtained by doing the before-mentioned changes. Now we

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

which gives the solution

where the covariant derivative is now given by

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

G. von Gersdorff et al. / Nuclear Physics B 712 (2005) 319

P. Horava, Phys. Rev. D 54 (1996) 7561, hep-th/9608019.

Nuclear Physics B 712 (2005) 2058

Model building and phenomenology of flux-induced

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

theories from D3-branes we need to consider a compact manifold M6 containing singular

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

where labels the D3-brane worldvolume dimensions, (3) representing D = 4 gauge

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

yielding a chiral spectrum which, in the case of Z4 is

In the case at hand we have

where 2 ,3 = 21 ,3 , etc. Notice that the case k = 0, corresponding to untwisted tadpoles,

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

defined if both conditions

F. Marchesano et al. / Nuclear Physics B 712 (2005) 2058

Now the supersymmetry variation at each boundary is proportional to (1 + i 5 T (f ) )(yf ).

where 0 = (y), = (R y) and the sign function (y) is defined as

2 m2 would require us to adjust (1.72 + 3.5) 0 in