You are on page 1of 160

a

r
X
i
v
:
h
e
p
-
t
h
/
9
5
1
0
1
4
0
v
3


1
4

M
a
r

1
9
9
6
Ministry of Higher and Special Secondary Education
of the USSR
Moscow State Lomonosov University
Physics Faculty
Division of Experimental and Theoretical Physics
Manuscript
1
UDK 530.12:531.51
Ivan Grigoryevich AVRAMIDI
Covariant Methods for the Calculation of
the Eective Action in Quantum Field Theory
and Investigation of HigherDerivative
Quantum Gravity
(01.04.02 Theoretical and Mathematical Physics)
DISSERTATION
for the Degree of
Candidate of Sciences
in Physics and Mathematics
Scientific Supervisor:
Professor V. R. Khalilov
Moscow 1986
1
Translated by the author on October 1995. Some misprints in the original text are
corrected. Available at @xxx.lanl.gov/hep-th/9510140
Contents
Introduction 2
1 Background eld method in quantum eld theory 11
1.1 Generating functional, Green functions and the eective action 11
1.2 Green functions of minimal dierential operators . . . . . . . . 17
1.3 Divergences, regularization, renormalization and the renorma-
lization group . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 Technique for the calculation of De Witt coecients and its
applications 25
2.1 Covariant expansions of eld variables in curved space . . . . 25
2.2 Structure elements of covariant expansions . . . . . . . . . . . 32
2.3 Technique for the calculation of De Witt coecients . . . . . . 40
2.4 De Witt coecients a
3
and a
4
at coinciding points . . . . . . . 44
2.5 Eective action of massive scalar, spinor and vector elds in
external gravitational eld . . . . . . . . . . . . . . . . . . . . 53
3 Partial summation of the semiclassical SchwingerDe Witt
expansion 58
3.1 Summation of asymptotic expansions . . . . . . . . . . . . . . 58
3.2 Covariant methods for the investigation of nonlocalities . . . . 61
3.3 Summation of the terms of rst order in external elds . . . . 65
3.4 Summation of the terms of second order in external elds . . . 70
3.5 Summation of the terms without covariant derivatives of ex-
ternal elds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
1
Contents 2
4 Higherderivative quantum gravity 88
4.1 Quantization of gauge eld theories. Unique eective action . 88
4.2 Oneloop divergences of higherderivative quantum gravity . . 96
4.3 Oshell oneloop divergences of the standard eective action
in arbitrary gauge and the divergences of the unique eective
action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.4 Renormalization group equations and the ultraviolet asympto-
tics of the coupling constants . . . . . . . . . . . . . . . . . . 115
4.5 Eective potential of higherderivative quantum gravity . . . . 123
5 Conclusion 141
Bibliography 143
Introduction
The classical theory of macroscopic gravitational phenomena, i.e., Einsteins
General Relativity (GR) [1, 2], cannot be treated as a complete selfconsistent
theory in view of a number of serious diculties that were not overcome since
the creation of GR [3].
This concerns, rst of all, the problem of spacetime singularities, which
are unavoidable in the solutions of the Einstein equations [14]. Close to these
singularities GR becomes incomplete as it cannot predict what is coming from
the singularity. In other words, the causal structure of spacetime breaks
down at the singularities [4].
Another serious problem of GR is the problem of the energy of the grav-
itational eld [5], which was critically analyzed, in particular, in the papers
of Logunov and collaborators [68]. In the papers [9, 10] a new relativistic
theory of gravitation (RTG) was proposed. In RTG the gravitational eld
is described by a spin2 tensor eld on a basic Minkowski spacetime. Such
an approach enables one to dene in the usual way the energymomentum
tensor of the gravitational eld and to obtain the usual conservation laws.
The curved spacetime in this approach is only an eective one that describes
the inuence of the gravitational eld on all the nongravitational matter.
Therein the identity (or geometrization) principle formulated in the papers
[10, 11] is embodied.
The diculties of the classical theory have motivated the need to con-
struct a quantum theory of gravitation. Also the recent progress towards
the unication of all nongravitational interactions [12] shows the need to
include gravitation in a general scheme of an unied quantum eld theory.
The rst problem in quantizing gravity is the construction of a covariant
perturbation theory [1321]. Einsteins theory of gravitation is a typical non
Abelian gauge theory with the dieomorphism group as a gauge group [14].
3
Introduction 4
The quantization of gauge theories faces the known diculty connected with
the presence of constraints [22, 23]. This problem was successfully solved in
the works of Feynman [13], De Witt [14] and Faddeev and Popov [15].
The most fruitful approach in quantum gravity is the background eld
method of De Witt [14, 2647]. This method is a generalization of the method
of generating functionals in quantum eld theory [4850] to the case of non
vanishing background eld. Both the gravitational and the matter elds can
have the background classical part.
The basic object in the background eld method is the eective action
functional. The eective action encodes, in principle, all the information
of the standard quantum eld theory. It determines the elements of the
diagrammatic technique of perturbation theory, i.e., the full onepoint prop-
agator and the full vertex functions, with regard to all quantum corrections,
and, hence, the perturbative Smatrix [14, 33, 45]. On the other hand, the
eective action gives at once the physical amplitudes in real external clas-
sical elds and describes all quantum eects in external elds [27, 28] (the
vacuum polarization of quantized elds, particle creation etc.) [5358]. The
eective action functional is the most appropriate tool for investigating the
structure of the physical vacuum in various models of quantum eld theory
with spontaneous symmetry breaking (Higgs vacuum, gluon condensation,
superconductivity) [24, 25, 51, 52].
The eective action makes it possible to take into account the back
reaction of the quantum processes on the classical background, i.e., to obtain
the eective equations for the background elds [33, 34, 45, 46, 5961]. In
this way, however, one runs into the diculty connected with the dependence
of the oshell eective action on the gauge and the parametrization of the
quantum eld. In the paper [34] a gaugeinvariant eective action (which
still depends parametrically on the gauge xing and the parametrization) was
constructed. The explicitly reparametrizationinvariant functional that does
not depend on the gaugexing (unique eective action) was constructed
in the papers [45, 46]. The unique eective action was studied in the
paper [47] in dierent models of quantum eld theory (including Einstein
gravity) and in the paper of the author and Barvinsky [173] in case of higher
derivative quantum gravity.
Thus, the calculation of the eective action is of high interest from the
point of view of the general formalism as well as for concrete applications.
The only practical method for the calculation of the eective action is the
Introduction 5
perturbative expansion in the number of loops [4850]. All the elds are
split in a background classical part and quantum perturbations propagating
on this background. The part of the classical action, which is quadratic in
quantum elds, determines the propagators of the quantum elds in external
elds, and higherorder terms reproduce the vertices of perturbation theory
[33].
At oneloop level, the contribution of the gravitational loop is of the same
order as the contributions of matter elds [62, 63]. At usual energies much
lower than the Planck energy (E hc
5
/G 10
19
GeV ) the contributions
of additional gravitational loops are highly suppressed. Therefore, a semi
classical concept applies when the quantum matter elds together with the
linearized perturbations of the gravitational eld interact with the external
background gravitational eld (and, probably, with the background matter
elds) [5356, 62, 63]. This approximation is known as oneloop quantum
gravity [33, 6466].
To evaluate the eective action it is necessary to nd, rst of all, the
Green functions of the quantum elds in the external classical elds of dif-
ferent nature. The Green functions in external elds were investigated by
a number of authors. Fock [67] proposed a method for solving the wave
equations in external electromagnetic elds by an integral transform in the
proper time parameter (fth parameter). Schwinger [68, 69] generalized
the propertime method and applied it to the calculation of the oneloop
eective action. De Witt [26, 28] reformulated the propertime method in
geometrical language and applied it to the case of external gravitational eld.
Analogous questions for the elliptic operators were investigated by mathe-
maticians [7076]. In the papers [77, 78] the standard SchwingerDe Witt
technique was generalized to the case of arbitrary dierential operators sat-
isfying the condition of causality.
The propertime method gives at once the Green functions in the neigh-
bourhood of the light cone. Therefore, it is the most suitable tool for in-
vestigation of the ultraviolet divergences (calculation of counterterms,
functions and anomalies). The most essential advantage of the propertime
method is that it is explicitly covariant and enables one to introduce various
covariant regularizations of divergent integrals. The most popular are the an-
alytic regularizations: dimensional regularization, function regularization
etc. [53, 54]. There are a lot of works along this line of investigation over
the last years [79106]. Although most of the papers restrict themselves to
Introduction 6
the oneloop approximation, the propertime method is applicable at higher
loops too. In the papers [37, 39, 40, 43, 44] it was applied to analyze twoloop
divergences in various models of quantum eld theory including Einsteins
quantum gravity.
Another important area, where the SchwingerDe Witt propertime me-
thod is successfully applied, is the vacuum polarization of massive quan-
tum elds by external background elds. When the Compton wave length
= h/mc, corresponding to the eld mass m, is much smaller than the char-
acteristic scale L of the background eld change, the propertime method
gives immediately the expansion of the eective action in a series in the
small parameter (/L)
2
[57, 58, 107]. The coecients of this expansion are
proportional to the socalled De Witt coecients and are local invariants,
constructed from the external elds and their covariant derivatives. In the
papers [98, 45] the general structure of the SchwingerDe Witt expansion of
the eective action was discussed. It was pointed out that there is a need to
go beyond the limits of the local expansion by the summation of the leading
derivatives of the external elds in this expansion. In the paper [45], based on
some additional assumptions concerning the convergence of the correspond-
ing series and integrals, the leading derivatives of the external elds were
summed up and a nonlocal expression for the oneloop eective action in
case of massless eld was obtained.
Thus, so far, eective and explicitly covariant methods for calculation of
the eective action in arbitrary background elds are absent. All the calcu-
lations performed so far concern either the local structures of the eective
action or some model specic background elds (constant elds, homogenous
spaces etc.) [53, 54].
That is why the development of general methods for covariant calculations
of the eective action, which is especially needed in the quantum theory
of gauge elds and gravity, is an actual and new area of research. The
papers of the author [171, 172] are devoted to the development of this line of
investigation. Therein an explicitly covariant and eective technique for the
calculation of De Witt coecients is elaborated. This technique is applicable
in the most general case of arbitrary external elds and spaces and can be
easily adopted to analytic calculations on computers. In these papers [171,
172] the renormalized oneloop eective action for massive scalar, spinor and
vector elds in external gravitational eld up to terms of order O(1/m
4
) is
calculated.
Introduction 7
In spite of impressive progress in oneloop quantum gravity, a complete
selfconsistent quantum theory of gravitation does not exist at present [108].
The diculties of quantum gravity are connected, in the rst line, with
the fact that there is not any consistent way to eliminate the ultraviolet
divergences arising in perturbation theory [109, 110]. It was found [111
117] that in the oneloop approximation the pure Einstein gravity is nite
on massshell (or renormalizable in case of nonvanishing cosmological con-
stant). However, twoloop Einstein gravity is no longer renormalizable on
shell [118]. On the other hand, the interaction with the matter elds also
leads to nonrenormalizability on massshell even in oneloop approximation
[119128].
Among various approaches to the problem of ultraviolet divergences in
quantum gravity (supergravity [129131], resummation [132134] etc. [109,
110]) an important place is occupied by the modication of the gravita-
tional Lagrangian by adding quadratic terms in the curvature of general
type (higherderivative theory of gravitation). This theory was investigated
by various authors both at the classical and at the quantum levels [135161].
The main argument against higherderivative quantum gravity is the
presence of ghosts in the linearized perturbation theory on at background,
that breaks down the unitarity of the theory [140151]. There were dierent
attempts to solve this problem by the summation of radiative corrections in
the propagator in the momentum representation [146149, 153155]. How-
ever, at present they cannot be regarded as convincing in view of causality
violation, which results from the unusual analytic properties of the Smatrix.
It seems likely that the problem of unitarity can be solved only beyond the
limits of perturbation theory [158].
Ultraviolet behavior of higherderivative quantum gravity was studied
in the papers [140157]. However, the oneloop counterterms were rst ob-
tained in the paper of Julve and Tonin [145]. The most detailed investigation
of the ultraviolet behaviour of higherderivative quantum gravity was car-
ried out in the papers of Fradkin and Tseytlin [153156]. In these papers,
an inconsistency was found in the calculations of Julve and Tonin. The one
loop counterterms were recalculated in higherderivative quantum gravity
of general type as well as in conformally invariant models and in conformal
supergravity [155, 156]. The main conclusion of the papers [153156] is that
higherderivative quantum gravity is asymptotically free in the physical re-
gion of coupling constants, which is characterized by the absence of tachyons
Introduction 8
on the at background. The presence of reasonably arbitrary matter does
not aect this conclusion.
Thus, the investigation of the ultraviolet behaviour of higherderivative
quantum gravity is an important and actual problem in the general program
of constructing a consistent quantum gravity. This problem is just the one
that the author was concern with in the papers [173, 174]. Therein the o
shell oneloop divergences of higherderivative quantum gravity in arbitrary
covariant gauge of the quantum eld were calculated. It was shown that
the results of previous authors contain a numerical error in the coecient of
the R
2
divergent term. The correction of this mistake radically changes the
asymptotic properties of the theory in the conformal sector. Although the
conclusion of [153156] about the asymptotic freedom in the tensor sector of
the theory remains true, the conformal sector exhibits just opposite zero
charge behavior in the physical region of coupling constants considered
in all previous papers [140155]. In the unphysical region of coupling
constants, which corresponds to the positive deniteness of the part of the
Euclidean action quadratic in curvature, the zerocharge singularities at
nite energies are absent.
This dissertation is devoted to further development of the covariant meth-
ods for calculation of the eective action in quantum eld theory and quan-
tum gravity, and to the investigation of the ultraviolet behaviour of higher
derivative quantum gravity.
In Chap. 1. the background eld method is presented. Sect. 1.1 contains
a short functional formulation of quantum eld theory in the form that is
convenient for subsequent discussion. In Sect. 1.2 the standard propertime
method with some extensions is presented in detail. Sect. 1.3 is concerned
with the questions connected with the problem of ultraviolet divergences,
regularization, renormalization and the renormalization group.
In Chap. 2 the explicitly covariant technique for the calculation of the
De Witt coecients is elaborated. In Sect. 2.1 the methods of covariant
expansions of arbitrary elds in curved space with arbitrary linear connec-
tion in the generalized covariant Taylor series and the Fourier integral are
formulated in the most general form. In Sect. 2.2 all the quantities that will
be needed later are calculated in form of covariant Taylor series. In Sect.
2.3, based on the method of covariant expansions, the covariant technique
for the calculation of the De Witt coecients in matrix terms is developed.
The corresponding diagrammatic formulation of this technique is given. The
Introduction 9
technique developed leads to the explicit calculation of De Witt coecients
as well as to the analysis of their general structure. The possibility to use
the corresponding analytic manipulations on computers is pointed out. In
Sect. 2.4 the calculation of the De Witt coecients a
3
and a
4
at coinciding
points is presented. In Sect. 2.5. the oneloop eective action for massive
scalar, spinor and vector elds in an external gravitational eld is calculated
up to terms of order 1/m
4
.
In Chap. 3 the general structure of the SchwingerDe Witt asymptotic
expansion is analyzed and the partial summation of various terms is carried
out. In Sect. 3.1 the method for summation of the asymptotic series due to
Borel (see, e.g., [167], sect. 11.4) is presented and its application to quantum
eld theory is discussed. In Sect. 3.2. the covariant methods for investi-
gations of the nonlocal structure of the eective action are developed. In
Sect. 3.3 the terms of rst order in the external eld in De Witt coecients
are calculated and their summation is carried out. The nonlocal expression
for the Green function at coinciding points, up to terms of second order in
external elds, is obtained. The massless case is considered too. It is shown
that in the conformally invariant case the Green function at coinciding points
is nite at rst order in external elds. In Sect. 3.4. the De Witt coecients
at second order in external elds are calculated. The summation of the terms
quadratic in external elds is carried out, and the explicitly covariant non
local expression for the oneloop eective action up to terms of third order in
external elds is obtained. All the formfactors, their ultraviolet asymptotics
and imaginary parts in the pseudoEuclidean region above the threshold are
obtained explicitly. The massless case in four and twodimensional spaces
is considered too. In Sect. 3.5 all terms without covariant derivatives of the
external elds in De Witt coecients, in the case of scalar eld, are picked
out. It is shown that in this case the asymptotic series of the covariantly
constant terms diverges. By making use of the Borel summation procedure
of the asymptotic series, the Borel sum of the corresponding semiclassical
series is calculated. An explicit expression for the oneloop eective action,
nonanalytic in the background elds, is obtained up to the terms with co-
variant derivatives of the external elds.
Chap. 4 is devoted to the investigation of higherderivative quantum
gravity. In Sect. 4.1 the standard procedure of quantizing the gauge the-
ories as well as the formulation of the unique eective action is presented.
In Sect. 4.2 the oneloop divergences of higherderivative quantum gravity
Introduction 10
with the help of the methods of the generalized SchwingerDe Witt technique
are calculated. The error in the coecient of the R
2
divergent term, due to
previous authors, is pointed out. In Sect. 4.3 the dependence of the diver-
gences of the eective action on the gauge of the quantum eld is analyzed.
The oshell divergences of the standard eective action in arbitrary covari-
ant gauge, and the divergences of the unique eective action, are calculated.
In Sect. 4.4 the corresponding renormalizationgroup equations are solved
and the ultraviolet asymptotics of the coupling constants are obtained. It
is shown that in the conformal sector of the theory there is no asymptotic
freedom in the physical region of the coupling constants. The presence of
the lowspin matter elds does not change this general conclusion: higher
derivative quantum gravity necessarily goes beyond the limits of the weak
conformal coupling at high energies. The physical interpretation of such ul-
traviolet behaviour is discussed. It is shown that the asymptotic freedom
both in tensor and conformal sectors is realized in the unphysical region
of coupling constants, which corresponds to the positivedenite Euclidean
action. In Sect. 4.5 the eective potential (i.e., the eective action on the de
Sitter background) in higherderivative quantum gravity is calculated. The
determinants of the second and fourthorder operators are calculated with
the help of the technique of the generalized function. The correctness of
the result for the R
2
divergence obtained in Sect. 4.2, as well as the cor-
rectness of the results for the arbitrary gauge and for the unique eective
action obtained in Sect. 4.3, are maintained. Both the eective potential in
arbitrary gauge and the unique eective potential are calculated. The unique
eective equations for the background eld, i.e., for the curvature of de Sit-
ter space, that do not depend on the gauge and the parametrization of the
quantum eld, are obtained. The rst quantum correction to the background
curvature caused by the quantum eects is found.
In the Conclusion the main results obtained in the dissertation are sum-
marized.
In the dissertation we use the unit system h = c = G = 1 (except for some
expressions where these quantities are contained explicitly) and the notation
and the sign conventions of the book [1].
Chapter 1
Background eld method in
quantum eld theory
1.1 Generating functional, Green functions
and the eective action
Let us consider an arbitrary eld (x) on a ndimensional spacetime given
by its contravariant components
A
(x) that transform with respect to some
(in general, reducible) representation of the group of general transformations
of the coordinates. The eld components
A
(x) can be of both bosonic and
fermionic nature. The fermion components are treated as anticommuting
Grassmanian variables [162], i.e.,

B
= (1)
AB

A
, (1.1)
where the indices in the exponent of the (1) are equal to 0 for bosonic
indices and to 1 for the fermionic ones.
For the construction of a local action functional S() one also needs a
metric of the conguration space E
AB
, i.e., a scalar product
(
1
,
2
) =
A
1
E
AB

B
2
, (1.2)
that enables one to dene the covariant elds components

A
=
B
E
BA
,
B
=
A
E
1 AB
, (1.3)
11
Chapter 1. Background field method 12
where E
1 AB
is the inverse matrix
E
1 AB
E
BC
=
A
C
, E
AC
E
1 CB
=
B
A
. (1.4)
The metric E
AB
must be nondegenerate both in bosebose and fermifermi
sectors and satisfy the symmetry conditions
E
AB
= (1)
A+B+AB
E
BA
, E
1 AB
= (1)
AB
E
1 BA
. (1.5)
In the case of gaugeinvariant eld theories we assume that the corresponding
ghost elds are included in the set of the elds
A
and the action S() is
modied by inclusion of the gauge xing and the ghost terms. To reduce the
writing we will follow, hereafter, the condensed notation of De Witt [26, 33]
and substitute the mixed set of indices (A, x), where x labels the spacetime
point, by one small Latin index i (A, x):
i

A
(x). The combined
summationintegration should be done over the repeated upper and lower
small Latin indices

1 i

i
2

_
d
n
x
1 A
(x)
A
2
(x). (1.6)
Now let us single out in the spacetime two causally connected in and
out regions, that lie in the past and in the future respectively relative to
the region, which is of interest from the dynamical standpoint. Let us dene
the vacuum states [in,vac> and [out,vac> in these regions and consider the
vacuumvacuum transition amplitude
out, vac[in, vac) exp
_
i
h
W(J)
_
(1.7)
in presence of some external classical sources J
i
vanishing in in and out
regions.
The amplitude (1.7) can be expressed in form of a formal functional in-
tegral [4850]
exp
_
i
h
W(J)
_
=
_
d/() exp
_
i
h
_
S() + J
i

i
_
_
, (1.8)
where /() is a measure functional, which should be determined by the
canonical quantization of the theory [20, 45]. The integration in (1.8) should
be taken over all elds satisfying the boundary conditions determined by
the vacuum states [in, vac > and [out, vac >. The functional W(J) is of
Chapter 1. Background field method 13
central interest in quantum eld theory. It is the generating functional for
the Schwinger averages
_

i
1

i
k
_
= exp
_

i
h
W(J)
_
_
h
i
_
k

k
L
J
i
1
J
i
k
exp
_
i
h
W(J)
_
, (1.9)
where
F())
out, vac[T(F()) [in, vac)
out, vac[in, vac)
, (1.10)

L
is the left functional derivative and T is the operator of chronological
ordering.
The rst derivative of the functional W(J) gives the mean eld (according
to the tradition we will call it the background eld)
_

i
_

i
(J) =

L
J
i
W(J), (1.11)
the second derivative determines the onepoint propagator
_

k
_
=
i

k
+
h
i
(
ik
, (1.12)
(
ik
(J) =

2
L
J
i
J
k
W(J),
and the higher derivatives give the manypoint Green functions
(
i
1
i
k
(J) =

k
L
J
i
1
J
i
k
W(J). (1.13)
The generating functional for the vertex functions, the eective action
(), is dened by the functional Legendre transform:
() = W(J) J
i

i
, (1.14)
where the sources are expressed in terms of the background elds, J = J(),
by inversion of the functional equation = (J), (1.11).
The rst derivative of the eective action gives the sources

i
()
,i
() = J
i
(), (1.15)
Chapter 1. Background field method 14
the second derivative determines the onepoint propagator

k
() T
ik
(), T
ik
= (1)
i+k+ik
T
ki
, (1.16)
T
ik
(
kn
=
n
i
,
where
R
is the right functional derivative,
n
i
=
B
A
(x, x

), and (x, x

) is
the deltafunction. The higher derivatives determine the vertex functions

i
1
i
k
() =

k
R

i
1

i
k
(). (1.17)
From the denition (1.14) and the equation (1.8) it is easy to obtain the
functional equation for the eective action
exp
_
i
h
()
_
=
_
d/() exp
_
i
h
_
S()
,i
()(
i

i
)
_
_
. (1.18)
Dierentiating the equation (1.15) with respect to the sources one can express
all the manypoint Green functions (1.13) in terms of the vertex functions
(1.17) and the onepoint propagator (1.12). If one uses the diagrammatic
technique, where the propagator is represented by a line and the vertex func-
tions by vertexes, then each dierentiation with respect to the sources adds
a new line in previous diagrams by all possible ways. Therefore, a many
point Green function is represented by all kinds of tree diagrams with a given
number of external lines.
Thus when using the eective action functional for the construction of
the Smatrix (when it exists) one needs only the tree diagrams, since all
quantum corrections determined by the loops are already included in the full
onepoint propagator and the full vertex functions. Therefore, the eective
equations (1.15),

,i
() = 0, (1.19)
(in absence of classical sources, J = 0) describe the dynamics of the back-
ground elds with regard to all quantum corrections.
The possibility to work directly with the eective action is an obvious
advantage. First, the eective action contains all the information needed to
construct the standard Smatrix [14, 29, 45]. Second, it gives the eective
equations (1.19) that enable one to take into account the inuence of the
Chapter 1. Background field method 15
quantum eects on the classical congurations of the background elds [59,
60].
In practice, the following diculty appear on this way. The background
elds, as well as all other Green functions, are not uniquely dened objects.
They depend on the parametrization of the quantum eld [45, 46]. Ac-
cordingly, the eective action is not unique too. It depends essentially on
the parametrization of the quantum eld o massshell, i.e., for background
elds that do not satisfy the classical equations of motion
S
,i
() = 0. (1.20)
On massshell, (1.20), the eective action is a well dened quantity and leads
to the correct Smatrix [14, 29, 30, 49].
A possible way to solve this diculty was recently proposed in the pa-
pers [45, 46], where an eective action functional was constructed, that is ex-
plicitly invariant with respect to local reparametrizations of quantum elds
(unique eective action). This was done by introducing a metric and a
connection in the conguration space. Therein, [45, 46], the unique eective
action for the gauge eld theories was constructed too. We will consider the
consequences of such a denition of the eective action in Chap. 5 when
investigating the higherderivative quantum gravity.
The formal scheme of quantum eld theory, described above, begins to
take on a concrete meaning in the framework of perturbation theory in the
number of loops [4850] (i.e., in the Planck constant h):
() = S() +

k1
h
k1

(k)
(). (1.21)
Substituting the expansion (1.21) in (1.18), shifting the integration variable
in the functional integral
i
=
i
+

hh
i
, expanding the action S() and
the measure /() in quantum elds h
i
and equating the coecients at
equal powers of h, we get the recurrence relations that uniquely dene all
the coecients
(k)
. All the functional integrals are Gaussian and can be
calculated in the standard way [24]. As the result the diagrammatic technique
for the eective action is reproduced. The elements of this technique are the
bare onepoint propagator, i.e., the Green function of the dierential operator

ik
() =

L

k
S(), (1.22)
Chapter 1. Background field method 16
and the local vertexes, determined by the classical action S() and the mea-
sure /().
In particular, the oneloop eective action has the form

(1)
() =
1
2i
log
sdet ()
/
2
()
, (1.23)
where
sdet = exp (str log ) (1.24)
is the functional Berezin superdeterminant [162], and
str F = (1
i
)F
i
i
=
_
d
n
x(1)
A
F
A
A
(x) (1.25)
is the functional supertrace.
The local functional measure /() can be taken in the form of the
superdeterminant of the metric of the conguration space
/= (sdet E
ik
())
1/2
, (1.26)
where
E
ik
() = E
AB
((x))(x, x

). (1.27)
In this case d/() is the volume element of the conguration space that
is invariant under the point transformations of the elds: (x) F((x)).
Using the multiplicativity of the superdeterminant [162], the oneloop eec-
tive action with the measure (1.26) can be rewritten in the form

(1)
() =
1
2i
log sdet

, (1.28)
with
sdet

i
k
= E
1 in

nk
. (1.29)
The local measure /() can be also chosen in such a way, that the leading
ultraviolet divergences in the theory, proportional to the deltafunction in
coinciding points (0), vanish [18, 20].
Chapter 1. Background field method 17
1.2 Green functions of minimal dierential
operators
The construction of Green functions of arbitrary dierential operators (1.22),
(1.29) can be reduced nally to the construction of the Green functions of
the minimal dierential operators of second order [78] that have the form

i
k
=
_

A
B
( m
2
) + Q
A
B
(x)
_
g
1/2
(x)(x, x

), (1.30)
where = g

is the covariant dAlambert operator,

is the covariant
derivative, dened by means of some background connection /

(x),

A
=

A
+/
A
B

B
, (1.31)
g

(x) is the metric of the background spacetime, g(x) = det g

(x), m is
the mass parameter of the quantum eld and Q
A
B
(x) is an arbitrary matrix.
The Green functions G
A
B
(x, x

) of the dierential operator (1.30) are two


point objects, which transform as the eld
A
(x) under the transformations
of coordinates at the point x, and as the current J
B
(x

) under the coordinate


transformations at the point x

. The indexes, belonging to the tangent space


at the point x

, are labelled with a prime.


We will construct solutions of the equation for the Green functions
_

A
C
( m
2
) + Q
A
C
_
G
C
B
(x, x

) =
A
B
g
1/2
(x)(x, x

), (1.32)
with appropriate boundary conditions, by means of the FockSchwinger
De Witt proper time method [6769, 26, 28] in form of a contour integral
over an auxiliary variable s,
G =
_
C
i ds exp(ism
2
)U(s), (1.33)
where the evolution function U(s) U
A
B
(s[x, x

) satises the equation

is
U(s) =
_

12+ Q
_
U(s),

1
A
B
, (1.34)
with the boundary condition
U
A
B
(s[x, x

C
=
A
B
g
1/2
(x)(x, x

), (1.35)
Chapter 1. Background field method 18
where C is the boundary of the contour C.
The evolution equation (1.34) is as dicult to solve exactly as the initial
equation (1.32). However, the representation of the Green functions in form
of the contour integrals over the proper time, (1.33), is more convenient to
use for the construction of the asymptotic expansion of the Green functions
in inverse powers of the mass and for the study of the behavior of the Green
functions and their derivatives on the light cone, x x

, as well as for
the regularization and renormalization of the divergent vacuum expectation
values of local variables (such as the energymomentum tensor, oneloop
eective action etc.).
Deforming the contour of integration, C, over s in (1.33) we can get
dierent Green functions for the same evolution function. To obtain the
causal Green function (Feynman propagator) one has to integrate over s
from 0 to and add an innitesimal negative imaginary part to the m
2
[26,
28]. It is this contour that we mean hereafter.
Let us single out in the evolution function a rapidly oscillating factor that
reproduces the initial condition (1.35) at s 0:
U(s) = i(4s)
n/2

1/2
exp
_


2is
_
T(s), (1.36)
where (x, x

) is half the square of the geodesic distance between the points


x and x

,
(x, x

) = g
1/2
(x) det (

(x, x

)) g
1/2
(x

) (1.37)
is the Van FleckMorette determinant, T T
A
B
(x, x

) is the parallel dis-


placement operator of the eld along the geodesic from the point x

to the
point x. We assume that there exists the only geodesic connecting the points
x and x

, the points x and x

being not conjugate, and suppose the twopoint


functions (x, x

), (x, x

) and T
A
B
(x, x

) to be singlevalued dierentiable
functions of the coordinates of the points x and x

. When the points x and


x

are close enough to each other this will be always the case [2, 26, 28].
The introduced transfer function (s)
A

B
(s[x, x

) transforms as a
scalar at the point x and as a matrix at the point x

(both its indexes are


primed). This function is regular in s at the point s = 0, i.e.,

B
(0[x, x

xx

=
A

B
(1.38)
Chapter 1. Background field method 19
independently on the way how x x

.
If one assumes that there are no boundary surfaces in spacetime (that
we will do hereafter), then the transfer function is analytic also in x close
to the point x = x

for any s, i.e., there exist nite coincidence limits of the


transfer function and its derivatives at x = x

that do not depend on the way


how x approaches x

.
Using the equations for the introduced functions [26, 28],
=
1
2

, (1.39)

T = 0, T
A
B
(x

, x

) =
A

B
, (1.40)

log
1/2
=
1
2
(n ), (1.41)
we obtain from (1.34) and (1.36) the transfer equation for the function (s):
_

is
+
1
is

_
(s) = T
1
_

1
1/2

1/2
+ Q
_
T(s). (1.42)
If one solves the transfer equation (1.42) in form of a power series in the
variable s
(s) =

k0
(is)
k
k!
a
k
, (1.43)
then from (1.38) and (1.42) one gets the recurrence relations for the De Witt
coecients a
k

a
0
= 0, a
0
A

B
(x

, x

) =
A

B
, (1.44)
_
1 +
1
k

_
a
k
= T
1
_

1
1/2

1/2
+ Q
_
Ta
k1
. (1.45)
From the equations (1.44) it is easy to nd the zeroth coecient
a
0
A

B
(x, x

) =
A

B
. (1.46)
The other coecients are calculated usually by dierentiating the relations
(1.45) and taking the coincidence limits [26, 28]. However such method of
calculations is very cumbersome and noneective. In this way only the coef-
cients a
1
and a
2
at coinciding points were calculated [26, 83, 84]. The same
coecients as well as the coecient a
3
at coinciding points were calculated
Chapter 1. Background field method 20
in the paper [76] by means of a completely dierent noncovariant method.
Thus, up to now a manifestly covariant method for calculation of the De Witt
coecients is absent.
In the Chap. 2 we develop a manifestly covariant and very convenient and
eective calculational technique that enables to calculate explicitly arbitrary
De Witt coecients a
k
as well as to analyze their general structure. Moreover,
the elaborated technique is very algorithmic and can be easily realized on
computers.
Let us stress that the expansion (1.43) is asymptotic and does not reect
possible nontrivial analytical properties of the transfer function, which are
very important when doing the contour integration in (1.33). The expansion
in the power series in proper time, (1.43), corresponds physically to the
expansion in the dimensionless parameter that is equal to the ratio of the
Compton wave length, = h/mc, to the characteristic scale of variation of
the external elds, L. That means that it corresponds to the expansion in
the Planck constant h in usual units [57, 58]. This is the usual semiclassical
approximation of quantum mechanics. This approximation is good enough
for the study of the light cone singularities of the Green functions, for the
regularization and renormalization of the divergent coincidence limits of the
Green functions and their derivatives in a spacetime point, as well as for the
calculation of the vacuum polarization of the massive elds in the case when
the Compton wave length is much smaller than the characteristic length
scale L, /L = h/(mcL) 1.
At the same time the expansion in powers of the proper time (1.43) does
not contain any information about all eects that depend nonanalytically
on the Planck constant h (such as the particle creation and the vacuum
polarization of massless elds) [28, 58]. Such eects can be described only
by summation of the asymptotic expansion (1.43). The exact summation in
general case is, obviously, impossible. One can, however, pick up the leading
terms in some approximation and sum them up in the rst line. Such partial
summation of the asymptotic (in general, divergent) series is possible only
by employing additional physical assumptions about the analytical structure
of the exact expression and corresponding analytical continuation.
In the Chap. 3 we will carry out the partial summation of the terms that
are linear and quadratic in external elds as well as the terms without the
covariant derivatives of the external elds.
Chapter 1. Background field method 21
1.3 Divergences, regularization, renormaliza-
tion and the renormalization group
A well known problem of quantum eld theory is the presence of the ultravi-
olet divergences that appear by practical calculations in perturbation theory.
They are exhibited by the divergence of many integrals (over coordinates
and momentums) because of the singular behavior of the Green functions at
small distances. The Green functions are, generally speaking, distributions,
i.e., linear functionals dened on smooth nite functions [48, 49]. There-
fore, numerous products of Green functions appeared in perturbation theory
cannot be dened correctly.
A consistent scheme for eliminating the ultraviolet divergences and ob-
taining nite results is the theory of renormalizations [48, 49], that can be
carried out consequently in renormalizable eld theories. First of all, one has
to introduce an intermediate regularization to give, in some way, the nite
values to the formal divergent expressions. Then one should single out the
divergent part and include the counterterms in the classical action that com-
pensate the corresponding divergences. In renormalizable eld theories one
introduces the counterterms that have the structure of the individual terms
of the classical action. They are interpreted in terms of renormalizations of
the elds, the masses and the coupling constants.
By the regularization some new parameters are introduced: a dimension-
less regularizing parameter r and a dimensionful renormalization parameter
. After subtracting the divergences and going to the limit r 0 the regu-
larizing parameter disappears, but the renormalization parameter remains
and enters the nite renormalized expressions. In renormalized quantum
eld theories the change of this parameter is compensated by the change of
the coupling constants of the renormalized action, g
i
(), that are dened at
the renormalization point characterized by the energy scale . The physical
quantities do not depend on the choice of the renormalization point where
the couplings are dened, i.e., they are renormalization invariant. The trans-
formation of the renormalization parameter and the compensating trans-
formations of the parameters of the renormalized action g
i
() form the group
of renormalization transformations [48, 163, 109]. The innitesimal form of
these transformations determines the dierential equations of the renormal-
ization group that are used for investigating the scaling properties (i.e., the
Chapter 1. Background field method 22
behavior under the homogeneous scale transformation) of the renormalized
coupling parameters g
i
(), the manypoint Green functions and other quan-
tities. In particular, the equations for renormalized coupling constants have
the form [109]

d
d
g
i
() =
i
( g()), (1.47)
where g
i
() =
d
i
g
i
() are the dimensionless coupling parameters (d
i
is the
dimension of the coupling g
i
), and
i
( g) are the GellMannLow functions.
Let us note, that among the parameters g
i
() there are also nonessential
couplings [109] (like the renormalization constants of the elds Z
r
()) that
are not invariant under the redenition of the elds. The equations for the
renormalization constants Z
r
() have more simple form [109]

d
d
Z
r
() =
r
( g())Z
r
(), (1.48)
where
r
( g) are the anomalous dimensions.
The physical quantities (such as the matrix elements of the Smatrix
on the massshell) do not depend on the details of the eld denition and,
therefore, on the nonessential couplings. On the other hand, the o mass
shell Green functions depend on all coupling constants including the non
essential ones. We will apply the renormalization group equations for the
investigation of the ultraviolet behavior of the higherderivative quantum
gravity in Chap. 4.
Let us illustrate the procedure of eliminating the ultraviolet divergences
by the example of the Green function of the minimal dierential operator
(1.30) at coinciding points, G(x, x), and the corresponding oneloop eective
action
(1)
, (1.28). Making use of the SchwingerDe Witt representation for
the Green function (1.33) we have
G(x, x) =

_
0
i ds i(4is)
n/2
exp(ism
2
)(s[x, x), (1.49)

(1)
=
1
2

_
0
ds
s
(4is)
n/2
exp(ism
2
)
_
d
n
xg
1/2
str (s[x, x). (1.50)
It is clear that in fourdimensional spacetime (n = 4) the integrals over
the proper time in (1.49) and (1.50) diverge at the lower limit. Therefore,
Chapter 1. Background field method 23
they should be regularized. To do this one can introduce in the proper time
integral a regularizing function (is
2
; r) that depends on the regularizing
parameter r and the renormalization parameter . In the limit r 0 the
regularizing function must tend to unity, and for r ,= 0 it must ensure the
convergence of the proper time integrals (i.e., it must approach zero su-
ciently rapidly at s 0 and be bounded at s by a polynomial). The
concrete form of the function does not matter. In practice, one uses the
cuto regularization, the PauliVillars one, the analytical one, the dimen-
sional one, the function regularization and others [54, 48, 49].
The dimensional regularization is one of the most convenient for the prac-
tical calculations (especially in massless and gauge theories) as well as for
general investigations [22, 50, 54, 111114]. The theory is formulated in
the space of arbitrary dimension n while the topology and the metric of the
additional n 4 dimensions can be arbitrary. To preserve the physical di-
mension of all quantities in the ndimensional spacetime it is necessary to
introduce the dimensionful parameter . All integrals are calculated in that
region of the complex plane of n where they converge. It is obvious that for
Re n < C, with some constant C, the integrals (1.49) and (1.50) converge
and dene analytic functions of the dimension n. The analytic continuation
of these functions to the neighborhood of the physical dimension leads to sin-
gularities in the point n = 4. After subtracting these singularities we obtain
analytical functions in the vicinity of the physical dimension, the value of this
function in the point n = 4 denes the nite value of the initial expression.
Let us make some remarks on the dimensional regularization. The ana-
lytical continuation of all the relations of the theory to the complex plane
of the dimension n is not singlevalued, since the values of a function of
complex variable in discrete integer values of the argument do not dene the
unique analytical function [165]. There is also an arbitrariness connected
with the subtraction of the divergences. Together with the poles in (n 4)
one can also subtract some nite terms (nonminimal renormalization). It is
also not necessary to take into account the dependence on the dimension of
some quantities (such as the volume element d
n
xg
1/2
(x), background elds,
curvatures etc.). On the other hand, one can specify the dependence of all
quantities on the additional n 4 coordinates in some special explicit way
and then calculate the integrals over the n 4 dimensions. This would lead
to an additional factor that will give, when expanding in n 4, additional
nite terms. This u uncertainty aects only the nite renormalization terms
Chapter 1. Background field method 24
that should be determined from the experiment.
Using the asymptotic expansion (1.43) we obtain in this way from (1.49)
and (1.50) the Green function at coinciding points and the oneloop eective
action in dimensional regularization
G(x, x) =
i
(4)
2
__
2
n 4
+C+ log
m
2
4
2
_
(m
2
a
1
(x, x)) m
2
_
+G
ren
(x, x), (1.51)
G
ren
(x, x) =
i
(4)
2

k2
a
k
(x, x)
k(k 1)m
2(k1)
, (1.52)

(1)
=
1
2(4)
2
_

1
2
_
2
n 4
+C+ log
m
2
4
2
_
(m
4
2m
2
A
1
+ A
2
)
+
3
4
m
4
A
0
m
2
A
1
_
+
(1)ren
, (1.53)

(1)ren
=
1
2(4)
2

k3
A
k
k(k 1)(k 2)m
2(k2)
, (1.54)
where C 0.577 is the Euler constant and
A
k

_
d
n
xg
1/2
str a
k
(x, x). (1.55)
Here all the coecients a
k
and A
k
are ndimensional. However in that part,
which is analytical in n 4, one can treat them as 4dimensional.
Chapter 2
Technique for the calculation of
De Witt coecients and its
applications
2.1 Covariant expansions of eld variables in
curved space
Let us single out a small regular region in the space, x a point x

in it,
and connect any other point x with the point x

by a geodesic x = x(),
x(0) = x

, where is an ane parameter.


The world function (x, x

) (in terminology of [2]), introduced in the


Chap. 1, has the form
(x, x

) =
1
2

2
x
2
(), (2.1)
where
x

() =
d
d
x

(). (2.2)
The rst derivatives of the function (x, x

) with respect to coordinates are


proportional to the tangent vectors to the geodesic at the points x and x

[2],

= x

(),

= x

(0), (2.3)
where

.
25
Chapter 2. Calculation of De Witt coefficients 26
Herefrom, it follows, in particular, the basic identity (1.39), that the function
(x, x

) satises,
(D 2) = 0, D

, (2.4)
the coincidence limits
[] = [

] = [

] = 0, (2.5)
[f(x, x

)] lim
xx

f(x, x

), (2.6)
and the relation between the tangent vectors

= g

, (2.7)
where g

(x, x

) is the parallel displacement operator of vectors along the


geodesic from the point x

to the point x. The nonprimed (primed) indices


are lowered and risen by the metric tensor in the point x (x

).
By dierentiating the basic identity (2.4) we obtain the relations
(D 1)

= 0,

, (2.8)
(D 1)

= 0,

, (2.9)
where

. (2.10)
Therefrom the coincidence limits follow
[

] = [

] =

, (2.11)
[
(
1
...

k
)

] = [
(
1
...

k
)

] = 0, (k 2). (2.12)
Let us consider a eld =
A
(x) and an ane connection /

= /
A
B
(x),
that denes the covariant derivative (1.31) and the commutator of covariant
derivatives,
[

] = 1

, (2.13)
1

+ [/

, /

]. (2.14)
Let us dene the parallel displacement operator of the eld along the
geodesic from the point x

to the point x, T = T
A
B
(x, x

), to be the solution
of the equation of parallel transport,
DT = 0, (2.15)
Chapter 2. Calculation of De Witt coefficients 27
with the initial condition
[T] = T(x, x) =

1. (2.16)
Herefrom one can obtain the coincidence limits
_

(
1
...

k
)
T
_
= 0, (k 1). (2.17)
In particular, when =

is a vector eld, and the connection /

is
the Christoel connection, the equations (2.15) and (2.16) dene the parallel
displacement operator of the vectors: T = g

(x, x

).
Let us transport the eld parallel along the geodesic to the point x

=
C

(x) = T
C

A
(x

, x)
A
(x) = T
1
, (2.18)
where T
1
= T
C

A
(x

, x) is the parallel displacement operator along the op-


posite path (from the point x to the point x

along the geodesic):


TT
1
=

1. (2.19)
The obtained object , (2.18), is a scalar under the coordinate transfor-
mations at the point x, since it does not have any nonprime indices. By
considering as a function of the ane parameter , let us expand it in the
Taylor series
=

k0
1
k!

k
_
d
k
d
k

_
=0
. (2.20)
Noting that d/d = x

and using the equation of the geodesic,


x

= 0, and the equations (2.3), (2.9) and (2.18), we obtain


= T

k0
(1)
k
k!

k
, (2.21)
where

k
=
_

(
1

k
)

_
. (2.22)
The equation (2.21) is the generalized covariant Taylor series for arbitrary
eld with arbitrary ane connection in a curved space.
Let us show that the series (2.21) is the expansion in a complete set of
eigenfunctions of the operator D, (2.4). The vectors

and

are the eigen-


functions of the operator D with the eigenvalues equal to 1 (see eqs. (2.8)
Chapter 2. Calculation of De Witt coefficients 28
and (2.9)). Therefore, one can construct the eigenfunctions with arbitrary
positive integer eigenvalues:
[0 > 1,
[n > [

1
...

n
>=
(1)
n
n!

n
, (n 1), (2.23)
D[n >= n[n > . (2.24)
We have
[n
1
> [n
k
>=
_
n
n
1
, . . . , n
k
_
[n >, (2.25)
where
_
n
n
1
, . . . , n
k
_
=
n!
n
1
! n
k
!
, n = n
1
+ + n
k
. (2.26)
Let us note that there exist more general eigenfunctions of the form
()
z
[n > with arbitrary eigenvalues (n + 2z)
D()
z
[n >= (n + 2z)()
z
[n > . (2.27)
However, for noninteger or negative z these functions are not analytic in
coordinates of the point x in the vicinity of the point x

. For positive integer


z they reduce to the linear combinations of the functions (2.23). Therefore,
we restrict ourselves to the functions (2.23) having in mind to study only
regular elds near the point x

.
Let us introduce the conjugate functions
< m[ <

m
[ = (1)
m
g

1
g
m

(
1

m)
(x, x

) (2.28)
and the scalar product
< m[n >=
_
d
n
x <

m
[

1
...

n
> . (2.29)
Using the coincidence limits (2.11) and (2.12) it is easy to prove that the set
of the eigenfunctions (2.23) and (2.28) is orthonormal
< m[n >=
mn
1
(n)
, 1
(n)

1
n

1
n
=

1
(
1

n
n)
. (2.30)
The introduced scalar product (2.29) reduces to the coincidence limit of the
symmetrized covariant derivatives,
< m[ >=
_

(
1

m)

_
. (2.31)
Chapter 2. Calculation of De Witt coefficients 29
Therefore, the covariant Taylor series (2.21) can be rewritten in a compact
form
[ >= T

n0
[n >< n[ > . (2.32)
Herefrom it follows the condition of completeness of the set of the eigenfunc-
tions (2.23) in the space of scalar functions regular in the vicinity of the point
x

1I =

n0
[n >< n[, (2.33)
or, more precisely,
(x, y) =

n0
1
n!

1
(x, x

n
(x, x

)g

1
(y, x

) g
n

n
(y, x

y
(
1

y
n)
(y, x

). (2.34)
Let us note that, since the parallel displacement operator T is an eigen-
function of the operator D with zero eigenvalue, (2.15), one can also introduce
a complete orthonormal set of isotopic eigenfunctions T[n > and < n[T
1
.
The complete set of eigenfunctions (2.23) can be employed to present an
arbitrary linear dierential operator F dened on the elds in the form
F =

m,n0
T[m >< m[T
1
FT[n >< n[T
1
, (2.35)
where
< m[T
1
FT[n >=
_

(
1

m)
T
1
FT
(1)
n
n!

n
_
(2.36)
are the matrix elements of the operator F (2.35). The matrix elements
(2.36) are expressed nally in terms of the coincidence limits of the deriva-
tives of the coecient functions of the operator F, the parallel displacement
operator T and the world function .
For calculation of the matrix elements of dierential operators (2.36) as
well as for constructing the covariant Fourier integral it is convenient to make
a change of the variables
x

= x

, x

), (2.37)
Chapter 2. Calculation of De Witt coefficients 30
i.e., to consider a function of the coordinates x

as the function of the vectors

(x, x

) and the coordinates x

.
The derivatives and the dierentials in old and new variables are con-
nected by the relations

,
dx

, d

dx

, (2.38)
where

= /x

= /

is dened in (2.10),

are the elements


of the inverse matrix,
=
1
, (2.39)
and is a matrix with elements

.
From the coincidence limits (2.11) it follows that for close points x and
x

det ,= 0, det ,= 0, (2.40)


and, therefore, the change of variables (2.37) is admissible.
The corresponding covariant derivatives are connected by analogous re-
lations

. (2.41)
From the denition of the matrices , (2.10), and , (2.39), one can get the
relations

]
= 0,

]
= 0, (2.42)

_
= 0, (2.43)
where is the Van FleckMorette determinant (1.37)
(x, x

) = g
1/2
(x

)g
1/2
(x) det() = g
1/2
(x

)g
1/2
(x) det()
1
= det( ) = (det( ))
1
= (det X)
1/2
, (2.44)
and , and X are matrices with elements

= g

= g

, (2.45)
X

. (2.46)
Chapter 2. Calculation of De Witt coefficients 31
Let us note that the conjugate eigenfunctions (2.28) of the operator D
can be expressed in terms of the operators

, (2.41),
< m[ <

m
[ =

m
)
(x, x

). (2.47)
Therefore, the coecients of the covariant Taylor series (2.21), (2.22), (2.31)
and (2.32) can be also written in terms of the operators

:
< m[ >= (1)
m
_

m
)

_
. (2.48)
The commutator of the operators

, when acting on the parallel displace-
ment operator T, has the form
[

]T =

1

T, (2.49)
where

T
1
1

T
=

+ [

/

,

/

], (2.50)

= T
1

T. (2.51)
The quantity (2.51) introduced here satises the equation (2.15),

= 0. (2.52)
On the other hand, when acting on the objects , (2.18), that do not
have nonprimed indices, the operators

commute with each other
[

] = 0. (2.53)
Thus the vectors

and the operators


, (2.41), play the role analogous


to that of usual coordinates and the operator of dierentiation in the tangent
space at the point x

. In particular,
[

] =

. (2.54)
Therefore, one can construct the covariant Fourier integral in the tangent
space at the point x

in the usual way using the variables

. So that for
the elds , (2.18), we have
(k) =
_
d
n
xg
1/2
(x)(x, x

) exp(ik

) (x),
Chapter 2. Calculation of De Witt coefficients 32
(x) =
_
d
n
k

(2)
n
g
1/2
(x

) exp(ik

) (k). (2.55)
Note, that the standard rule,

(x) =
_
d
n
k

(2)
n
g
1/2
(x

) exp(ik

)(ik

) (k), (2.56)
takes place and the covariant momentum representation of the function
has the form
(x, y) =
_
d
n
k

(2)
n
g
1/2
(x

)g
1/2
(x)(x, x

) exp
_
ik

(y, x

(x, x

)
__
.
(2.55)
2.2 Structure elements of covariant expan-
sions
Let us calculate the quantities , and X introduced in previous section.
By dierentiating the equations (2.8) and (2.9) and commuting the covariant
derivatives we get
D + (

1) +S = 0, (2.58)
D + (

1) = 0, (2.59)
where
=

, S = S

,

1 =

,
S

= R

. (2.60)
By solving the eq. (2.59) with respect to the matrix , substituting the
solution in (2.58) and taking into account (2.39) we obtain the linear equation
for the matrix (2.45)
_

1
_
D
2
+ D
_
+

S
_
= 0, (2.61)
where

S =

S

= g

, with the boundary condition, (2.39), (2.11),


[ ] = 1. (2.62)
Chapter 2. Calculation of De Witt coefficients 33
One can solve the equation (2.61) perturbatively, treating the matrix

S
as a perturbation. Supposing
=

1 +

, (2.63)
we obtain
_

1(D
2
+ D) +

S
_

=

S, [

] = 0. (2.64)
Herefrom we have formally

=
_

1(D
2
+ D) +

S
_
1

S =

k1
(1)
k+1
_
(D
2
+ D)
1

S
_
k

1. (2.65)
The formal expression (2.65) becomes meaningful in terms of the expansion
in the eigenfunctions of the operator D, (2.23). The inverse operator
(D
2
+ D)
1
=

n0
1
n(n + 1)
[n >< n[, (2.66)
when acting on the matrix

S, is well dened, since < 0[

S >= 0. Expanding
the matrix

S in the covariant Taylor matrix according to (2.21) and (2.32),

S =

n2
(1)
n
(n 2)!
K
(n)
, (2.67)
where K
(n)
is the matrix with entries K

(n)
K

(n)
= K

n
,
K

1
n
=
(
1

n2
R

n1
||n)
, (2.68)
we obtain
=

1 +

n2
(1)
n
n!

(n)
, (2.69)

(n)
=

n
=

1k[
n
2
]
(1)
k+1
(2k)!
_
n
2k
_

n
1
,,n
k
2
n
1
++n
k
=n
_
n 2k
n
1
2, , n
k
2
_
Chapter 2. Calculation of De Witt coefficients 34

K
(n
k
)
n(n + 1)

K
(n
k
1)
(n
1
+ + n
k1
)(n
1
+ + n
k1
+ 1)
(2.70)

K
(n
2
)
(n
1
+ n
2
)(n
1
+ n
2
+ 1)

K
(n
1
)
n
1
(n
1
+ 1)
, (2.71)
where
_
n
m
_
=
n!
m!(n m)!
. (2.70a)
Let us write down some rst coecients (2.70):

2
=
1
3
R

(
1
||
2
)
,

3
=
1
2

(
1
R

2
||
3
)
, (2.71)

4
=
3
5

(
1

2
R

3
||
4
)

1
5
R

(
1
||
2
R

3
||
4
)
.
Using the solution (2.69) one can nd all other quantities. The inverse matrix
=
1
has the form
=

1 +

n2
(1)
n
n!

(n)
, (2.72)

(n)
=

n
=

1k[
n
2
]

n
1
,,n
k
2
n
1
++n
k
=n
_
n
n
1
, n
k
_

(n
k
)

(n
1
)
. (2.73)
Some rst coecients (2.73) equal

2
=
1
3
R

(
1
||
2
)
,

3
=
1
2

(
1
R

2
||
3
)
, (2.74)

4
=
3
5

(
1

2
R

3
||
4
)

7
15
R

(
1
||
2
R

3
||
4
)
.
Chapter 2. Calculation of De Witt coefficients 35
Using (2.64) and (2.72) we nd the matrix X:
X

= g

n2
(1)
n
n!
X

(n)
, (2.75)
X

(n)
= X

n
= 2
(

)
(n)
+

2kn2
_
n
k
_

(nk)

)
(k)
. (2.76)
The lowest order coecients (2.76) read
X

2
=
2
3
R

(
1

2
)
,
X

3
=
(
1
R

2

3
)
, (2.77)
X

4
=
6
5

(
1

2
R

3

4
)
+
8
5
R

(
1
||
2
R

3

4
)
.
Finally, we nd the Van FleckMorette determinant (2.44):
= exp(2),
=

n2
(1)
n
n!

(n)
, (2.78)

(n)
=

n
=

1k[
n
2
]
1
2k

n
1
,,n
k
2
n
1
++n
k
=n
_
n
n
1
, , n
k
_
tr
_

(n
1
)

(n
k
)
_
. (2.79)
The rst coecients (2.79) equal

2
=
1
6
R

2
,

3
=
1
4

(
1
R

3
)
, (2.80)
Chapter 2. Calculation of De Witt coefficients 36

4
=
3
10

(
1

2
R

4
)
+
1
15
R
(
1
||
2
R

3

4
)
.
Let us calculate the quantity

/

, (2.51). By dierentiating the equation


(2.52) and using (2.50) we obtain
(D + 1)

/

, (2.81)
where

= g

T
1
1

. (2.82)
Herefrom we have

= (D + 1)
1

, (2.83)
where the inverse operator (D + 1)
1
is dened by
(D + 1)
1
=

n0
1
n + 1
[n >< n[.
Expanding the vector

L

, (2.82), in the covariant Taylor series (2.21), (2.32)

n1
(1)
n
(n 1)!
1

(n)
, (2.85)
where
1

(n)
= 1

n
,
1

1
n
=
(
1

n1
1

n)
, (2.86)
we obtain

n1
(1)
n
n!
/

(n)
, (2.87)
/

(n)
= /

n
=
n
n + 1
_
_
_
1

(n)

2kn1
_
n 1
k
_

(k)
1

(nk)
_
_
_
. (2.88)
In particular, the rst coecients (2.88) have the form
/

1
=
1
2
1

1
,
Chapter 2. Calculation of De Witt coefficients 37
/

2
=
2
3

(
1
1
||
2
)
, (2.89)
/

3
=
3
4

(
1

2
1
||
3
)

1
4
R
(
1
||
2
1

3
)
.
Thus we have obtained all the needed quantities in form of covariant Taylor
series: (2.69), (2.72), (2.75), (2.78) and (2.87).
In particular case, when all the derivatives of the curvature tensors can
be neglected,

= 0,

= 0, (2.90)
one can solve exactly the equations (2.61), (2.81) or, equivalently, sum up
all terms in the Taylor series, which do not contain the derivatives of the
curvature tensors. In this case the Taylor series (2.69) is a power series in
the matrix

S. It is an eigenmatrix of the operator D,
D

S = 2

S, (2.91)
and, therefore, when acting on the series (2.69), one can present the operator
D in the form
D = 2

S
d
d

S
, (2.92)
and treat the matrix

S as an usual scalar variable. Substituting (2.92) in
(2.61), we obtain an ordinary second order dierential equation
_
d
2
dt
2
+
2
t
d
dt
+

1
_
= 0, t

S, (2.93)
that has the general solution
= (

S)
1/2
_
C
1
sin

S + C
2
cos

S
_
, (2.94)
where C
1
and C
2
are the integration constants. Using the initial condition
(2.62), we have nally C
1
= 1, C
2
= 0, i.e.,
=
sin

S
. (2.95)
In any case this formal expression makes sense as the series (2.69). It is
certain to converge for close points x and x

.
Chapter 2. Calculation of De Witt coefficients 38
Using (2.95), (2.39) and (2.46) we nd easily the matrices and X
=

S
sin

S
,
X =

S
sin
2

S
, (2.96)
and the Van FleckMorette determinant , (2.44),
= det
_
_

S
sin

S
_
_
. (2.97)
The vector

/

, (2.87), in the case (2.90) is presented as the product of some


matrix H

, that does not depend on 1

, and the vector



L

, (2.82),

= H

. (2.98)
Substituting (2.98) in (2.81) and using the equation
D

L

=

L

, (2.99)
we get the equation for the matrix H
(D + 2)H

. (2.100)
Substituting (2.92) and (2.95) in (2.100), we obtain an ordinary rst order
dierential equation
_
t
d
dt
+

1 2
_
H =
sin t
t
, t

S. (2.101)
The solution of this equation has the form
H =

S
1
_

1 cos

S + C
3
_
, (2.102)
where C
3
is the integration constant. Using the initial condition [

/

] = 0,
we nd C
3
= 0, i.e., nally
H =

S
1
_

1 cos

S
_
,
Chapter 2. Calculation of De Witt coefficients 39

=
_

S
1
_

1 cos

S
__

. (2.103)
Let us apply the formulas (2.95)(2.97) and (2.103) to the case of the De
Sitter space with the curvature
R

= (

), = const. (2.104)
The matrix

S, (2.60), (2.61), and the vector

L

, (2.82), have in this case the


form

= 2

,

L

= 1

, (2.105)
where

2
,

2
=

, (2.106)

= 0,

being the projector on the hypersurface that is orthogonal to the vector

. Using (2.106) we obtain for an analytic function of the matrix



S
f(

S) =

k0
c
k

S
k
= f(0)(

) + f(2)

. (2.107)
Thus the formal expressions (2.95)(2.97) and (2.103) take the concrete form

2
( 1),


1
+

2
(
1
1), (2.108)
X

= g

2
+

2
(1
2
),
H

2
_
1
2

_
,
where
=
sin

2
, =
1 cos

2
2
.
Using eq. (2.105) one can neglect the longitudinal terms in the matrix H
(2.108) and obtain the quantity

/

= 1

. (2.109)
Chapter 2. Calculation of De Witt coefficients 40
In the case (2.90) it is transverse,

= 0. (2.110)
Let us stress that all formulas obtained in present section are valid for
spaces of any dimension and signature.
2.3 Technique for the calculation of De Witt
coecients
Let us apply the method of covariant expansions developed above to the cal-
culation of the De Witt coecients a
k
, i.e., the coecients of the asymptotic
expansion of the transfer function (1.43). Below we follow our paper [172].
From the recurrence relations (1.45) we obtain the formal solution
a
k
=
_
1 +
1
k
D
_
1
F
_
1 +
1
k 1
D
_
1
F (1 +D)
1
F, (2.111)
where the operator D is dened by (2.4) and the operator F has the form
F = T
1
(

1
1/2

1/2
+ Q)T. (2.112)
In the present section we develop a convenient covariant and eective
method that gives a practical meaning to the formal expression (2.111). It
will suce, in particular, to calculate the coincidence limits of the De Witt
coecients a
k
and their derivatives.
First of all, we suppose that there exist nite coincidence limits of the
De Witt coecients
[a
k
] lim
xx

a
k
(x, x

), (2.113)
that do not depend on the way how the points x and x

approach each
other, i.e., the De Witt coecients a
k
(x, x

) are analytical functions of the


coordinates of the point x near the point x

. Otherwise, i.e., if the limit


(2.113) is singular, the operator (1 +
1
k
D) does not have a singlevalued
inverse operator, since there exist the eigenfunctions of this operator with
eigenvalues equal to zero
D
n/2
[n >= 0,
_
1 +
1
k
D
_

(k+n)/2
[n >= 0. (2.114)
Chapter 2. Calculation of De Witt coefficients 41
Therefore, the zeroth coecient a
0
(x, x

) is dened in general up to an
arbitrary function f(

; x

), and the inverse operator (1 +


1
k
D)
1
is
dened up to an arbitrary function
k/2
f
k
(

; x

).
Using the covariant Taylor series (2.21) and (2.32) for the De Witt coef-
cients
a
k
=

n0
[n >< n[a
k
> (2.115)
and dening the inverse operator (1 +
1
k
D)
1
in form of the eigenfunctions
expansion (2.35),
_
1 +
1
k
D
_
1
=

n0
k
k + n
[n >< n[, (2.116)
we obtain from (2.111)
< n[a
k
>=

n
1
,,n
k1
0
k
k + n

k 1
k 1 +n
k1

1
1 +n
1
< n[F[n
k1
>< n
k1
[F[n
k2
> < n
1
[F[0 >, (2.117)
where
< m[F[n >=
_

(
1

m)
F
(1)
n
n!

n
_
(2.118)
are the matrix elements (2.36) of the operator F, (2.112).
Since the operator F, (2.112), is a dierential operator of second order,
its matrix elements < m[F[n >, (2.118), do not vanish only for n m + 2.
Therefore, the sum (2.117) always contains a nite number of terms, i.e., the
summation over n
i
is limited from above
n
1
0, n
i
n
i+1
+ 2, (i = 1, . . . , k 1; n
k
n). (2.119)
Thus we reduced the problem of calculation of the De Witt coecients
to the calculation of the matrix elements (2.118) of the operator F, (2.112).
For the calculation of the matrix elements (2.118) it is convenient to write
the operator F, (2.112), in terms of the operators

, (2.41). Using (2.112),


(2.41)(2.46) and (2.51) we obtain
F = T
1

1/2


1
X


1/2
T +

Q
Chapter 2. Calculation of De Witt coefficients 42
=
_

1(

) +

/

_
X

1(

) +

/

_
+

Q
=

1X

+ Y

+ Z, (2.120)
where

Q = T
1
QT,

log
1/2
, (2.121)
Y

=

1

+ 2X

, (2.122)
Z = X

_

/

_
+

_
X

+

/

__
+

Q. (2.123)
Now one can easily calculate the matrix elements (2.118). Using (2.54),
(2.5) and (2.48) we obtain from (2.118)
< m[F[m+ 2 >=

1
(
1
m

1
m
g

m+1

m+2
)
,
< m[F[m+ 1 >= 0,
< m[F[n > =
_
m
n
_

1
n
(
1
n
Z

n+1
m)

_
m
n 1
_

(
1

n1
(
1

n1
Y
n)
nm)
+
_
m
n 2
_

1
(
1

n2
(
1

n2
X

n1
n)

n1
m)
, (2.124)
where
Z

1
n
= (1)
n
_

n
)
Z
_
,
Y

1
n
= (1)
n
_

n
)
Y

_
,
X

1
n
= (1)
n
_

n
)
X

2
_
. (2.125)
In (2.124) it is meant that the binomial coecient
_
n
k
_
is equal to zero if
k < 0 or n < k.
Thus, to calculate the matrix elements (2.118) it is sucient to have the
coincidence limits of the symmetrized covariant derivatives (2.125) of the
coecient functions X

, Y

and Z, (2.122), (2.123), i.e., the coecients


of their Taylor expansions (2.48), that are expressed in terms of the Taylor
coecients of the quantities X

, (2.76),

/

, (2.88), and , (2.78), (2.79),


Chapter 2. Calculation of De Witt coefficients 43
found in the Sect. 2.2. From the dimensional arguments it is obvious that
for m = n the matrix elements < m[F[n >, (2.124), (2.125), are expressed in
terms of the curvature tensors R

, 1

and the matrix Q; for m = n + 1


in terms of the quantities R, 1 and Q; for m = n + 2 in terms
of the quantities of the form R
2
, R etc.
In the calculation of the De Witt coecients by means of the matrix
algorithm (2.117) a diagrammatic technique, i.e., a graphic method for
enumerating the dierent terms of the sum (2.117), turns out to be very
convenient and pictorial. The matrix elements < m[F[n >, (2.118), are
presented by some blocks with m lines coming in from the left and n lines
going out to the right (Fig. 1),
m
_
.
.
.

.
.
.
_
n
Fig. 1
and the product of the matrix elements < m[F[k >< k[F[n > by two
blocks connected by k intermediate lines (Fig. 2),
m
_
.
.
.

k
_
.
.
.

.
.
.
_
n
Fig. 2
that represents the contractions of corresponding tensor indices.
To obtain the coecient < n[a
k
>, (2.117), one should draw all possible
diagrams with k blocks connected in all possible ways by any number of in-
termediate lines. When doing this, one should keep in mind that the number
of the lines, going out of any block, cannot be greater than the number of
the lines, coming in, by more than two and by exactly one (see (2.124)).
Then one should sum up all diagrams with the weight determined for each
diagram by the number of intermediate lines from (2.117). Drawing of such
diagrams is of no diculties. Therefore, the main problem is reduced to the
calculation of some standard blocks.
Let us note that the elaborated technique does not depend at all on the
dimension and on the signature of the manifold and allows to obtain results
in most general case. It is also very algorithmic and can be easily adopted
for analytic computer calculations.
Chapter 2. Calculation of De Witt coefficients 44
2.4 De Witt coecients a
3
and a
4
at coincid-
ing points
Using the developed technique one can calculate the coincidence limits (2.113)
of the De Witt coecients [a
3
] and [a
4
]. The coecients [a
1
] and [a
2
], that
determine the oneloop divergences (1.51), (1.53), have long been calculated
by De Witt [26], Christensen [83, 84] and Gilkey [76]. The coecient [a
3
],
that describes the vacuum polarization of massive quantum elds in the low-
est nonvanishing approximation 1/m
2
, (1.54), was calculated in general
form rst by Gilkey [76]. The coecient [a
4
] in general form has not been
calculated until now.
The diagrams for the De Witt coecients [a
3
] and [a
4
] have the form,
(2.117),
[a
3
] =
j j j
+
1
3
j j j
+
2
4
j j j
+
2
4

1
2
j j j
+
2
4

1
3
j j j
+
2
4

1
5
j j j
,
(2.126)
[a
4
] =
j j j j
+
1
3
j j j j
+
2
4
j j j j
+
3
5
j j j j
+
1
5

2
4
j j j j
+
2
4

1
2
j j j j
+
3
5

2
3
j j j j
+
2
4

1
3
j j j j
+
3
5

1
3
j j j j
+
3
5

2
4
j j j j
+
3
5

2
6
j j j j
+
3
5

2
3

1
2
j j j j
+
3
5

2
4

1
2
j j j j
+
3
5

2
4

1
3
j j j j
Chapter 2. Calculation of De Witt coefficients 45
+
3
5

2
3

1
4
j j j j
+
3
5

2
6

1
2
j j j j
+
3
5

2
4

1
5
j j j j
+
3
5

2
6

1
3
j j j j
+
3
5

2
6

1
4
j j j j
+
3
5

2
6

1
5
j j j j
+
3
5

2
6

1
7
j j j j
. (2.127)
Substituting the matrix elements (2.124) in (2.126) and (2.127) we express
the coecients [a
3
] and [a
4
] in terms of the quantities X, Y and Z, (2.125),
[a
3
] = P
3
+
1
2
[P, Z
(2)
]
+
+
1
2
B

+
1
10
Z
(4)
+
3
, (2.128)
[a
4
] = P
4
+
3
5
_
P
2
, Z
(2)
_
+
+
4
5
PZ
(2)
P +
4
5
[P, B

]
+
+
2
5
B

PZ

2
5
B

+
1
3
Z
(2)
Z
(2)
+
2
5
B

Z
(2)
+
2
5
C

+
1
5
_
P, Z
(4)
_
+
+
4
15
T

+
1
35
Z
(6)
+
4
, (2.129)
where

3
=
1
6
U

1
Z

, (2.130)

4
=
4
15
PU

1
Z

+ U

1
_
3
10
Z

P +
1
10
PZ

+
3
10
Z

3
10
Y

1
5
Y

+
1
10
X

+
7
50
Z
(2)
_
+
1
5
U

2
Z

+
4
25
U

3
Z

. (2.131)
Chapter 2. Calculation of De Witt coefficients 46
Here the following notation is introduced:
P = [Z],
Z
(2)
= g

2
Z

2
, (2.132)
Z
(4)
= g

2
g

4
Z

4
,
Z
(6)
= g

2
g

4
g

6
Z

6
, (2.133)
Z
(2)
= g

2
Z

2
,
Z
(2)
= g

2
Z

2
, (2.134)
B

= Z

1
2
g

2
Y

2
, (2.135)
C

= Z

(2)

1
4
g

2
g

4
Y

4
, (2.136)
T

= Z

2
Y
()

2
+
1
4

1g

2
g

4
X

4
, (2.137)
U

1
= 2Y
()
+

1g

2
X

2
, (2.138)
U

2
= Y
()
+

1g

2
X
()

2
, (2.139)
U

3
= X
()
, (2.140)
and [A, B]
+
denotes the anticommutator of the matrices A and B.
Using the formulas (2.132)(2.140), (2.125), (2.122) and (2.123) and the
quantities X

,

/

and , (2.70)(2.89), and omitting cumbersome compu-


tations we obtain
P = Q +
1
6

1R, (2.141)
Y

= 1

+
1
3

1R

, (2.142)
U

1
= U

3
= 0, (2.143)
Z

P
1
3
J

, (2.144)
B

P +
1
3
J

, (2.145)
U

2
= 0, (2.146)
Chapter 2. Calculation of De Witt coefficients 47
Z
(2)
=
_
Q+
1
5

1R
_
+
1
30

1
_
R

_
+
1
2
1

, (2.147)
Z

= W


1
2

(
J
)
, (2.148)
T

= W

+
1
2

(
J
)
, (2.149)
Z
(2)
= V

+ G

, (2.150)
C

= V

, (2.151)
where
J

, (2.152)
W

=
(

)
_
Q+
3
20

1R
_
+

1
_
1
20
R


1
15
R

+
1
30
R

+
1
30
R

R


_
+
1
2
1
(
1

)
, (2.153)
V

= Q
(2)
+
1
2
[1

]
+

1
3
[J

, 1

]
+
+

1
_
1
5

R +
1
9
R

R +
1
15
R

1
15
R

_
, (2.154)
Q
(2)
= g

2
)
Q
=

Q+ [1

Q] +
1
3
[J

, Q] +
2
3
R

Q, (2.155)
G

=
1
5
J

2
15
[1

, J

]
1
10
[1

2
15
R

+
2
15
R

7
45
R

+
2
5

1
15

R1

. (2.156)
Chapter 2. Calculation of De Witt coefficients 48
The result for the coecient Z
(4)
has more complicated form
Z
(4)
= Q
(4)
+ 2[1

]
+
+
8
9
J

+
4
3

+61

+
10
3
R

1
_
3
14
2
R +
1
7
R

R
2
21
R

+
4
7
R

+
4
63

R
1
42

1
21

+
3
28

+
2
189
R

2
63
R

R


+
2
9
R

16
189
R


88
189
R


R


R


_
, (2.157)
where
Q
(4)
= g

2
g

(
1

4
)
Q
=
2
Q
1
2
[1

, [1

, Q]]
2
3
[J

Q]
+
2
3
R

Q +
1
3

Q. (2.158)
From the fact that the quantities U

1
, U

2
and U

3
, (2.138)(2.140),
are equal to zero, (2.143), (2.146), it follows that the quantities
3
and
4
,
(2.130), (2.131), are equal to zero too:

3
=
4
= 0. (2.159)
To calculate the De Witt coecient [a
3
], (2.128), it is sucient to have the
formulas listed above. This coecient is presented explicitly in the paper
Chapter 2. Calculation of De Witt coefficients 49
[76]. (Let us note, to avoid misunderstanding, that our De Witt coecients
a
k
, (1.43), dier from the coecients a
k
used by the other authors [26, 83,
84, 76] by a factor: a
k
= k! a
k
.)
Let us calculate the De Witt coecient A
3
, (1.55), that determines the
renormalized oneloop eective action (1.54) in the fourdimensional physical
spacetime in the lowest nonvanishing approximation 1/m
2
[57, 58]. By
integrating by parts and omitting the total derivatives we obtain from (2.128),
(2.141), (2.144), (2.145), (2.147) and (2.157)(2.159)
A
3
=
_
d
n
xg
1/2
str
_
P
3
+
1
30
P
_
R

+ R
_
+
1
2
P1

+
1
2
P P
1
10
J

+
1
30
_
21

2R

+ R

_
+

1
_

1
630
R R +
1
140
R

+
1
7560
_
64R

+48R

R


+ 6R

+17R

28R


R


R


_
__
. (2.160)
The formula (2.160) is valid for any dimension of the space and for any elds.
To obtain the explicit expression for the coecient [a
4
] one has to sub-
stitute (2.141)(2.159) in (2.129) as well as to calculate the quantity Z
(6)
,
(2.133). To write down this quantity in a compact way we dene the fol-
lowing tensors constructed from the covariant derivatives of the curvature
tensor:
I

1
n
=
(
1

n1
R

|| n)
,
K

1
n
=
(
1

n2
R

n1
n)
,
L

1
n
=
(
1

n1
R

n)
, (2.161)
M

1
n
=
(
1

n2
R

n1
n)
,
Chapter 2. Calculation of De Witt coefficients 50
1

1
n
=
(
1

n1
1

n)
,
and denote the contracted symmetrized covariant derivatives just by a num-
ber in the brackets (analogously to (2.132)(2.134). For example,
I

(2)
= g

2
I

2
,
K

(4)
= g

2
g

4
K

4
,
L

(4)
= g

2
g

4
L

4
, (2.162)
M
(8)
= g

2
g

8
M

8
,
1

(4)
= g

2
g

4
1

4
,
etc.
In terms of introduced quantities, (2.161), and the notation (2.162) the
quantity Z
(6)
, (2.133), takes the form
Z
(6)
= Z
M
(6)
+

1Z
S
(6)
. (2.163)
Here Z
M
(6)
is the matrix contribution
Z
M
(6)
= Q
(6)
+
5
2
[1

, 1
(4)
]
+
+
32
5
[1

, 1
(2)
]
+

8
5
[J

, 1
(4)
]
+
+
9
2
1

+
27
4
1

(2)
1
(2)
+
5
4
R

[1

, 1
(2)
]
+
+
5
2
R

[1

, 1

]
+
+
15
8
R

[1

, 1
(2)
]
+
+
44
15

[1

, J

]
+
+
22
5
K

[1

, 1

]
+
+
22
5
K
(2)
[1
,
, 1

]
+
+
64
45
R

16
15
R

, J

]
+
+
256
45
R
(||)
1

+
32
45
R

Chapter 2. Calculation of De Witt coefficients 51


+
_
6
5
K
(4)
+
17
40
R

+
17
60
R

_
1

+
_
24
5
K
(2)
+
17
40
R

+
17
40
R

+
17
30
R

+
17
60
R

R


+
51
80
R

_
1

, (2.164)
where
Q
(6)
= g

2
g

4
g

(
1

6
)
Q,
and Z
S
(6)
is the scalar contribution
Z
S
(6)
=
7
18
M
(8)
+ R

_
5
18
K
(6)

5
6
L
(4)
_
+
20
21
R

K
(4)

2
5

RL
(4)
+
_
5
3
K

(2)
+M

_
K
(4)

12
5
M

L
(2)
+
20
9
K

K
(2)
+
6
25
K

(4)
K
(4)

27
25
M

(2)
M
(2)
+
48
25
K

(2)
K
(2)

18
25
M

+
16
25
K

+
_
101
450
R

+
68
525
R

1
5
R

R


_
K
(4)
+
_

2
5
R

6
25
R

+
6
25
R

R


_
M
(2)
+
_

1
6
R

1
3
R

_
I
()(2)

1
3
R

+
_

2
5
R

R


R

_
L

+
_

6
5
R

+
2588
1575
R

+
4
25
R

R


Chapter 2. Calculation of De Witt coefficients 52
+
1048
1575
R

R


+
962
1575
R

_
K
(2)
+
1088
1575
R

+ R

4
45

RI
(2)
+
_
7
10
M

+
5
6
K
(2)
_
K

(2)

1
3
K

(2)
I
(2)

2
3
I

(2)
+
5
9
K


2
3
I

4
15
K

_
+ R

( )
__
7
5
K


9
10
K

16
15
I

_
M

+
_

7
10
K

(2)

9
10
K

(2)

4
15
I

(2)
_
M


8
45

RI


3
10

RK

(2)
+
20
21
K

(2)
K

(2)
+
40
21
K

K
(2)
+
40
21
K

K


_

7
450
R

+
1
90
R

R


R

+
817
6300
R

+
391
4725
R

R


R

2243
9450
R

1
75
R

R


R

16
4725
R

R


R


R

7
300
R

R


R

+
8
675
R

+
247
9450
R

32
4725
R


R


R


R


+
1
105
R

Chapter 2. Calculation of De Witt coefficients 53


+
32
4725
R

R


R


R

232
4725
R

R


+
299
4725
R

. (2.165)
Further transformations of the expression for the quantity Z
(6)
, (2.163)
(2.165), in general form appear to be pointless since they are very cumber-
some and there are very many independent invariants constructed from the
curvatures and their covariant derivatives. The set of invariants and the
form of the result should be chosen in accordance with the specic charac-
ter of the considered problem. That is why we present the result for the
De Witt coecient [a
4
] by the set of formulas (2.129), (2.141)(2.159) and
(2.161)(2.165).
2.5 Eective action of massive scalar, spinor
and vector elds in external gravitational
eld
Let us illustrate the elaborated methods for the calculation of the oneloop
eective action on the example of real scalar, spinor and vector massive
quantum matter elds on a classical gravitational background in the four
dimensional physical spacetime. In this section we follow our papers [171,
172].
The operator (1.22), (1.29) in this case has the form

=
_

_
+R+ m
2
, (j = 0),

+ m, (j = 1/2),
d + m
2
, (j = 1),
(2.166)
where m and j are the mass and the spin of the eld, is the coupling
constant of the scalar eld with the gravitational eld,

are the Dirac


matrices,
(

)
=

1g

, d is the exterior derivative and is the operator of


codierentiation on forms: ()
[
1

k
]
=

[
1

k
]
.
Chapter 2. Calculation of De Witt coefficients 54
The commutator of covariant derivatives (3.13) (2.14) has the form
1

=
_

_
0
1
4

.
R

(2.167)
Using the equations

=

1
_

1
4
R
_
,
(d + d)

= (

2
=
2
= 0, (2.168)
one can express the Green functions of the operator

, (2.166), in terms of
the Green functions of the minimal operator, (1.30),

1
=
_
Q + m
2
_
1
, (2.169)
where
=
_

_
1
m

,
1
1
m
2
d
Q =
_

_
R

1
4

1R .
R

(2.170)
Using the SchwingerDe Witt representation, (1.33),

(1)
=
1
2i
log sdet

=
1
2i

_
0
ds
s
sdet exp(is

) (2.171)
and the equations
tr exp(is

) = tr exp(is

),
exp(isd) = expis(d + d) exp(isd) + 1, (2.172)
tr exp(isd)

j=1
= tr exp(is )

j=0
,
Chapter 2. Calculation of De Witt coefficients 55
we obtain up to nonessential innite contributions (0),
log det(

+ m) =
1
2
log det
_

1
_
+
1
4
R + m
2
__
,
log det(d + m
2
) = log det
_

+ R

+ m
2

_
log det( +m
2
)

j=0
.
(2.173)
Thus we have reduced the functional determinants of the operators (2.166)
to the functional determinants of the minimal operators of the form (1.30).
By making use of the formulas (2.171)(2.173), (1.50) and (1.54) we obtain
the asymptotic expansion of the oneloop eective action in the inverse pow-
ers of the mass

(1)ren
=
1
2(4)
2

k3
B
k
k(k 1)(k 2)m
2(k2)
, (2.174)
where
B
k
=
_

_
A
k
1
2
A
k
A
k

(j=1)
A
k

(j=0,=0)
(2.175)
and the coecients A
k
are given by (1.55).
Let us stress that the coecients A
k
, (1.55), for the spinor eld contain
a factor (1) in addition to the usual trace over the spinor indices according
to the denition of the supertrace (1.25).
Substituting the matrices 1

, (2.167), and Q, (2.170), in (2.160) and


using (2.175) we obtain the rst coecient, B
3
, in the asymptotic expansion
(2.174)
B
3
=
_
d
4
xg
1/2
_
c
1
R R + c
2
R

+ c
3
R
3
+ c
4
RR

+c
5
RR

+ c
6
R

+ c
7
R

R


+c
8
R

+ c
9
R

+c
10
R


R


R


_
, (2.176)
Chapter 2. Calculation of De Witt coefficients 56
where the coecients c
i
are given in the Table 1.
Table 1
Scalar eld Spinor eld Vector eld
c
1
1
2

1
5
+
1
56

3
280

27
280
c
2
1
140
1
28
9
28
c
3
_
1
6

_
3
1
864

5
72
c
4

1
30
_
1
6

_

1
180
31
60
c
5
1
30
_
1
6

_

7
1440

1
10
c
6

8
945

25
756

52
63
c
7
2
315
47
1260

19
105
c
8
1
1260
19
1260
61
140
c
9
17
7560
29
7560

67
2520
c
10

1
270

1
108
1
18
The renormalized eective action (2.174) can be used to obtain the renor-
malized matrix elements of the energymomentum tensor of the quantum
matter elds in the external gravitational eld [57, 58]
_
out, vac[T

(x)[in, vac
_
ren
= h2g
1/2

(1)ren
g

(x)
+ O( h
2
). (2.177)
Such problems were intensively investigated last time [53, 54]. In partic-
ular, in the papers [57, 58] the vacuum polarization of the quantum elds in
the external gravitational eld of the black holes was investigated. In these
papers an expression for the renormalized oneloop eective action was ob-
tained that is similar to (2.174)(2.176) but does not take into account the
terms, that do not contribute to the eective vacuum energymomentum
tensor (2.177) when the background metric satises the vacuum Einstein
Chapter 2. Calculation of De Witt coefficients 57
equations, R

= 0. Our result (2.176) is valid, however, in general case of


arbitrary background space. Moreover, using the results of Sect. 2.4 for the
De Witt coecient [a
4
], one can calculate the coecient A
4
and, therefore,
the next term, B
4
in the asymptotic expansion of the eective action (2.174)
of order 1/m
4
. The technique for the calculation of the De Witt coecients
developed in this section is very algorithmic and can be realized on comput-
ers (all the needed information is contained in the rst three sections of the
present chapter). In this case one can also calculate the next terms of the
expansion (2.174).
However, the eective action functional
(1)
is, in general, essentially
nonlocal and contains an imaginary part. The asymptotic expansion in
the inverse powers of the mass (2.174) does not reect these properties. It
describes good the eective action only in weak gravitational elds (R m
2
).
In strong gravitational elds (R m
2
), as well as for the massless matter
elds, the asymptotic expansion (2.174) becomes meaningless. In this case it
is necessary either to sum up some leading (in some approximation) terms or
to use from the very beginning the nonlocal methods for the Green function
and the eective action.
Chapter 3
Partial summation of the
semiclassical
SchwingerDe Witt expansion
3.1 Summation of asymptotic expansions
The solution of the wave equation in external elds, (1.32), by means of the
proper time method, (1.33), turns out to be very convenient for investigation
of many general problems of the quantum eld theory, especially for the
analysis of the ultraviolet behavior of Green functions, regularization and
renormalization. However, in practical calculations of concrete eects one
fails to use the proper time method directly and one is forced to use model
noncovariant methods.
In order to use the advantages of the covariant proper time method it is
necessary to sum up the asymptotic series (1.43) for the evolution function.
In general case the exact summation is impossible. Therefore, one can try
to carry out the partial summation, i.e., to single out the leading (in some
approximation) terms and sum them up in the rst line. On the one hand one
can limit oneself to a given order in external elds and sum up all derivatives,
on the other hand one can neglect the derivatives and sum up all powers of
external elds.
In this way we come across a certain diculty. The point is that the
asymptotic series do not converge, in general. Therefore, in the paper [166] it
58
Chapter 3. Summation of SchwingerDe Witt expansion 59
is proposed to give up the SchwingerDe Witt representation (1.33), (1.49),
and treat it only as an auxiliary tool for the separation of the ultraviolet
divergences. It is stated there that the SchwingerDe Witt representation,
(1.33), exists for a small class of spaces only when the semiclassical
solution is exact.
However, the divergence of the asymptotic series (1.43) does not mean
at all that one must give up the SchwingerDe Witt representation (1.33).
The matter is, the transfer function (s) is not analytical at the point s = 0.
Therefore, it is natural that the direct summation of the power series in s,
(1.43), leads to divergences. In spite of this one can get a certain useful
information from the structure of the asymptotic (divergent) series.
Let us consider a physical quantity G() which is dened by an asymp-
totic expansion of the perturbation theory in a parameter
G() =

k0
c
k

k
. (3.1)
The convergence radius of the series (3.1) is given by the expression [165]
R =
_
lim
k
sup[c
k
[
1/k
_
1
. (3.2)
If R ,= 0 then in the disc [[ < R of the complex plane of the parameter
the series (3.1) converges and denes an analytical function. If the considered
physical quantity G() is taken to be analytical function, then outside the
convergence disc of the series (3.1), [[ R, it should be dened by analytical
continuation. The function

G() obtained in this way will be certain to
have singularities, the rst one lying on the circle [[ = R. The analytical
continuation through the boundary of the convergence disc is impossible if
all the points of the boundary (i.e., the circle [[ = R) are singular. In this
case the physical quantity G() appears to be meaningless for [[ R. If
R = 0, then the series (3.1) diverges for any , i.e., the function G() is not
analytic in the point = 0. In this case it is impossible to carry out the
summation and the analytical continuation. Nevertheless, one can gain an
impression of t he exact quantity G() by making use of the Borel procedure
for summation of asymptotic (in general, divergent) series [167].
The idea consists in the following. One constructs a new series with bet-
ter convergence properties which reproduces the initial series by an integral
Chapter 3. Summation of SchwingerDe Witt expansion 60
transform. Let us dene the Borel function
B(z) =

k0
c
k
(k + )
z
k
, (3.3)
where and are some complex numbers (Re , Re > 0).
The convergence radius

R of the series (3.3) equals

R =
_
_
lim
k
sup

c
k
(k + )

1/k
_
_
1
. (3.4)
Thus, when the coecients c
k
of the series (3.1) rise not faster than
exp(Mk log k), M = const, then one can always choose in such way,
Re M, that the convergence radius of the series (3.3) will be not equal
to zero

R ,= 0, i.e., the Borel function B(z) will be analytical at the point
z = 0. Outside the convergence disc, [z[

R, the Borel function is dened
by analytical continuation.
Let us dene

G() =
_
C
dt t
1
e
t
B(t

), (3.5)
where the integration contour C starts at the zero point and goes to innity in
the right halfplane (Re t +). The asymptotic expansion of the function

G() for 0 has the form (3.1). Therefore, the function



G(), (3.5),
(which is called the Borel sum of the series (3.1)) can be considered as the
true physical quantity G().
The analytical properties of the Borel function B(z) determine the con-
vergence properties of the initial series (3.1). So, if the initial series (3.1)
has a nite convergence radius R ,= 0, then from (3.4) it follows that the
Borel series (3.3) has an innite convergence radius

R = and, therefore,
the Borel function B(z) is an entire function (analytical in any nite part
of the complex plane). In this case the function

G(), (3.5), is equal to the
sum of the initial series (3.1) for [[ < R and determines its analytical con-
tinuation outside the convergence disc [[ R. If the Borel function B(z)
has singularities in the nite part of the complex plane, i.e.,

R < , then
from (3.4) it follows that the series (3.1) has a convergence radius equal to
zero R = 0, and, therefore, the physical quantity G() is not analytic at the
point = 0. At the same time there always exist a region in the complex
Chapter 3. Summation of SchwingerDe Witt expansion 61
plane of the variable where the Borel sum

G() is still well dened and can
be used for the analytical continuation to physical values of . In this way
dierent integration contours will give dierent functions

G(). In this case
one should choose the contour of integration from some additional physical
assumptions concerning the analytical properties of the exact function G().
3.2 Covariant methods for the investigation
of nonlocalities
The De Witt coecients a
k
have the background dimension L
2k
, where L
is the length unit. Therefore, the standard SchwingerDe Witt expansion,
(1.43), (1.52) and (1.54), is, in fact, an expansion in the background dimen-
sion [77, 78]. In a given order in the background dimension L
2k
both the
powers of the background elds, R
k
, as well as their derivatives,
2(k1)
R, are
taken into account. In order to investigate the nonlocalities it is convenient
to reconstruct the local SchwingerDe Witt expansion in such a way that the
expansion is carried out in the background elds but its derivatives are taken
into account exactly from the very beginning. Doing this one can preserve
the manifest covariance using the methods developed in the Chap. 2.
Let us introduce instead of the Green function G
A
B
(x, y) of the operator
(1.30), whose upper index belongs to the tangent space in the point x and
the lower one in the point y, a threepoint Green function (
A

B
(x, y[x

),
that depends on some additional xed point x

,
((x, y[x

) = T
1
(x, x

)
1/2
(x, x

)G(x, y)
1/2
(y, x

)T(y, x

). (3.6)
This Green function is scalar at the points x and y and a matrix at the point
x

. In the following we will not exhibit the dependence of all quantities on


the xed point x

.
The equation for the Green function ((x, y), (3.6), has the form (1.32)
_
F
x

1m
2
_
((x, y) =

1g
1/2
(x)
1
(x)(x, y), (3.7)
where F
x
is the operator (2.112) and

1 =
A
B
.
Let us single out in the operator F
x
the free part that is of zero order in
the background elds. Using (2.120) we have
F
x
=

1

x
+

F
x
, (3.8)
Chapter 3. Summation of SchwingerDe Witt expansion 62
where

x
= g

(x

, (3.9)

F
x
=

1

X

(x)

+ Y

(x)

+ Z(x), (3.10)

= X

(x) g

(x

), (3.11)
and the operators

and the quantities X

, Y

and Z are dened by the


formulas (2.41), (2.46), (2.122) and (2.123). The operator

F, (3.10), is of the
rst order in the background elds and can be considered as a perturbation.
By introducing the free Green function (
0
(x, y),
_

x
+ m
2
_
(
0
(x, y) =

1
1
(x)g
1/2
(x)(x, y), (3.12)
and writing the equation (3.7) in the integral form
((x, z) = (
0
(x, z) +
_
d
n
y g
1/2
(y)(y)(
0
(x, y)

F
y
((y, z), (3.13)
we obtain from (3.13) by means of direct iterations
((x, z) = (
0
(x, z) +

k1
_
d
n
y
1
g
1/2
(y
1
)(y
1
) d
n
y
k
g
1/2
(y
k
)(y
k
)
(
0
(x, y
1
)

F
y
1
(
0
(y
1
, y
2
)

F
y
k
(
0
(y
k
, z). (3.14)
Using the covariant Fourier integral (2.55) and the equations (2.56) and (2.57)
we obtain from (3.12) and (3.14) the momentum representations for the free
Green function
(
0
(x, y) =

1
_
d
n
k

(2)
n
g
1/2
(x

) exp
_
ik

(y)

(x)
__
1
m
2
+ k
2
, (3.15)
and for the full Green function
((x, y) =
_
d
n
k

(2)
n
g
1/2
(x

)
d
n
p

(2)
n
g
1/2
(x

)
exp
_
ip

(y) ik

(x)
_
((k, p), (3.16)
Chapter 3. Summation of SchwingerDe Witt expansion 63
where
((k, p) = (2)
n
g
1/2
(x

)
(k

)
m
2
+ k
2
+
(k, p)
(m
2
+ k
2
)(m
2
+ p
2
)
, (3.17)
(k, p) =

F(k, p) +

i1
_
d
n
q

1
(2)
n
g
1/2
(x

)
d
n
q

i
(2)
n
g
1/2
(x

F(k, q
1
)
1
m
2
+ q
2
1


F(q
i1
, q
i
)
1
m
2
+ q
2
i

F(q
i
, p), (3.18)

F(k, p) =

1

X

(k p)p

iY

(k p)p

+ Z(k p), (3.19)


k
2
g

(x

)k

,
and

X

(q), Y

(q) and Z(q) are the covariant Fourier components, (2.55),


of the coecients

X

, Y

and Z, (3.11), (2.46), (2.122), (2.123).


The formulas (3.15)(3.19) reproduce the covariant generalization of the
usual diagrammatic technique. Therefore, one can apply good elaborated
methods of the Feynman momentum integrals. The Fourier components of
the coecient functions

X

(q), Y

(q) and Z(q), can be expressed in terms


of the Fourier components of the external elds, R

, 1

and Q, using
the formulas obtained in Chap. 2. As usual [4850, 54] one should choose
the contour of integration over k
0
in the momentum integrals (3.15)(3.18).
Dierent ways of integration correspond to dierent Green functions. For
the causal (Feynman) Green function one should either assume k
2
k
2
i
or go to the Euclidean sector of the spacetime [4850].
Similarly, one can construct the kernels of any nonlocal operators of
general form, f( ), where f(z) is some function. In the zeroth approximation
in external elds we obtain
f( )(x, y) = T(x)
1/2
(x)

f( )(x, y)
1/2
(y)T
1
(y),

f( )(x, y) =
_
d
n
k

(2)
n
g
1/2
(x

) exp
_
ik

(y)

(x)
__
f(k
2
). (3.20)
An important method for the investigation of the nonlocalities is the
analysis of the De Witt coecients a
k
and the partial summation of the
asymptotic series (1.43). In this case one should limit oneself to some order
Chapter 3. Summation of SchwingerDe Witt expansion 64
in external elds and sum up all derivatives of external elds. In order to get
an eective expansion in external elds it is convenient to change a little the
diagrammatic technique for the De Witt coecients developed in Sect. 2.3.
Although all the terms in the sum (2.117) have equal background dimension,
L
n2k
, they are of dierent order in external elds. From the formula (2.117)
it is not seen immediately what order in external elds has each term of the
sum (2.117), i.e., a single diagram, since all the diagrams for the coecient
< n[a
k
> have k blocks. However, among these blocks, i.e., the matrix
elements < m[F[m+2 >, there are dimensionless blocks that do not have any
background dimension and are of zero order in external elds. These are the
blocks (matrix elements < m[F[m+2 >) with the number of outgoing lines
equal to the number of the incoming lines plus 2 (c.f. (2.124)). Therefore,
one can order all the diagrams for the De Witt coecients (i.e., the single
terms of the sum (2.117)) in the following way. The rst diagram contains
only one dimensionful block, all others being dimensionless. The second class
contains all diagrams with two dimensionful blocks, the third one three
etc. The last diagram contains k dimensionful blocks. To obtain the De Witt
coecients in the rst order in external elds it is sucient to restrict oneself
to the rst diagram. To get the De Witt coecients in the second order in
external elds it is sucient to restrict oneself to the rst diagram and the
set of diagrams with two dimensionful blocks etc. This method is completely
analogous to the separation of the free part of the operator F, (3.8). The
dimensionful matrix elements < m[F[n >, (with m n), of the operator F,
(2.120), are equal to the matrix elements of the operator

F, (3.10). When
calculating the matrix elements (2.124) and (2.125) one can also neglect the
terms that do not contribute in the given order in external elds.
After such reconstruction (and making use of (2.124)) the formula (2.117)
for the De Witt coecients < n[a
k
> takes the form
< n[a
k
>=

1Nk;

1i
1
<i
2
<<i
N1
k1;

n
i

1jN
_
2i
j
+n
j
1
i
j1
_
_
2i
j
+n
j
1
i
j
_
< n; k i
N1
1[F[n
N1
>< n
N1
; i
N1
i
N2
1[F[n
N2
>
< n
2
; i
2
i
1
1[F[n
1
>< n
1
; i
1
1[F[0 >, (3.21)
Chapter 3. Summation of SchwingerDe Witt expansion 65
where the following notation is introduced
< n; k[F[m > g

2
g

2k1

2k
<
1

n+2k
[F[
1

m
>, (3.22)
i
0
0, i
N
k, n
N
n,
and the summation over n
i
is carried out in such limits that all matrix el-
ements are dimensionful, (i.e., for each < n; k[F[m >, (3.22), there holds:
n + 2k m):
n
1
+2(i
1
1) 0, n
2
+2(i
2
i
1
1) n
1
, , n+2(ki
N1
1) n
N1
.
The formula (3.21) also enables to use the diagrammatic technique. How-
ever, for analysing the general structure of the De Witt coecients and for
the partial summation it is no longer eective. Therefore, one should use the
analytic expression (3.21).
3.3 Summation of the terms of rst order in
external elds
Let us calculate the coincidence limit of the De Witt coecients [a
k
] =<
0[a
k
> in the rst order in external elds. Using the formula (3.21) we
obtain
[a
k
] =
1
_
2k1
k
_
< 0; k 1[F[0 > +O(R
2
), (3.23)
where O(R
2
) denotes all omitted terms of second order in external elds.
Using (3.22), (2.121)(2.125) and (2.46) and the formulas of the Sect. 2.2
we obtain from (3.23) up to quadratic terms
[a
k
] =
k!(k 1)!
(2k 1)!
k1
_
Q+
k
2(2k + 1)

1R
_
+ O(R
2
), (k 1). (3.24)
The expression (3.24) can be used for the calculation of the transfer function
(s[x, x) in the rst order in external elds. Substituting (3.24) in (1.43),
we obtain
(s[x, x) =

1 +is
_
f
1
(is )Q +

1f
2
(is )R
_
+ O(R
2
), (3.25)
Chapter 3. Summation of SchwingerDe Witt expansion 66
where
f
1
(z) =

k0
k!
(2k + 1)!
z
k
, (3.26)
f
2
(z) =

k0
(k + 1)!(k + 1)
(2k + 3)!
z
k
. (3.27)
The power series (3.26) and (3.27) converge for any nite z and hence dene
entire functions. One can sum up the series of the type (3.26) and (3.27)
using the general formula
(k + l)!
(2k + 2l + 1)!
=
l!
(2l)!
1
_
0
d
2l
1
k!
_
1
2
4
_
k
, (3.28)
that is easily obtained from the denition of the Euler betafunction [168].
Substituting (3.28) in (3.26) and (3.27) and summing over k we obtain
the integral representations of the functions f
1
(z) and f
2
(z)
f
1
(z) =
1
_
0
d exp
_
1
4
_
1
2
_
z
_
, (3.29)
f
2
(z) =
1
_
0
d
1
4
(1
2
) exp
_
1
4
_
1
2
_
z
_
. (3.30)
The kernels of the nonlocal operators f
1
(is ), f
2
(is ) should be understood
in terms of covariant momentum expansions (3.20).
Using the obtained transfer function, (3.25), one can easily obtain the
Green function at coinciding points, G(x, x), in the rst order in external
elds. Substituting (3.25) in (1.49) and supposing Imm
2
< 0 (for the causal
Green function), we obtain after the integration over the proper time in the
ndimensional space
G(x, x) = i(4)
n/2
_

_
1
n
2
_

1m
n2
+
_
2
n
2
_
m
n4
_

F
1
_

4m
2
_
Q +

1

F
2
_

4m
2
_
R
_
_
+O(R
2
), (3.31)
Chapter 3. Summation of SchwingerDe Witt expansion 67
where (z) is the Euler gammafunction, and

F
1
(z) =
1
_
0
d
_
1 + (1
2
)z
_n
2
2
, (3.32)

F
2
(z) =
1
_
0
d
1
4
(1
2
)
_
1 + (1
2
)z
_n
2
2
. (3.33)
By expanding in the dimension n in the neighborhood of the point n = 4
and subtracting the pole 1/(n4) we obtain the renormalized Green function,
G
ren
(x, x), in the physical fourdimensional spacetime (1.51), (1.52), up to
terms of second order in external elds
G
ren
(x, x) =
i
(4)
2
_
F
1
_

4m
2
_
Q +

1F
2
_

4m
2
_
R
_
_
+ O(R
2
), (3.34)
where
F
1
(z) = 2 J(z), (3.35)
F
2
(z) =
1
18
_
5
3
z
_

1
6
_
1
1
2z
_
J(z), (3.36)
J(z) = 2(1 +z)
1
_
0
d
1 + (1
2
)z
. (3.37)
The formfactors F
1
(z), (3.35), and F
2
(z), (3.36), are normalized according
the conditions
F
1
(0) = F
2
(0) = 0. (3.38)
The integral (3.37) determines an analytical singlevalued function in the
complex plane z with a cut along the negative part of the real axis from 1
to :
J(z) = 2

1 +
1
z
log
_

z + 1 +

z
_
, [ arg(z + 1)[ < , (3.39)
J(x i) = 2

1 +
1
x
log
_

x 1 +

x
_
i

1 +
1
x
, (x < 1).
(3.40)
Chapter 3. Summation of SchwingerDe Witt expansion 68
Thus we obtained the nonlocal expression for the Green function at coin-
ciding points, (3.34). It reproduces the local SchwingerDe Witt expansion,
(1.52), up to quadratic terms in eternal elds by expanding in inverse powers
of the mass. The power series determining the formfactors F
1
(z), (3.35), and
F
2
(z), (3.36), converge in the region [z[ < 1. For z = 1 there is a threshold
singularity the branching point. Outside the circle [z[ 1 the formfactors
are dened by the analytic continuation. The boundary conditions for the
formfactors x uniquely the ambiguity in the Green function (3.34). For the
causal Green function (Imm
2
< 0) the lower bank of the cut is the physical
one. Therefore, with account of (3.40), the imaginary parts of the formfac-
tors (3.35) and (3.36) in the pseudoEuclidean region above the threshold
z = x i, x < 1, equal
ImF
1
(x i) =

1 +
1
x
, (3.41)
ImF
2
(x i) =

6
_
1
1
2x
_

1 +
1
x
. (3.42)
The ultraviolet asymptotics [z[ of the formfactors (3.35) and (3.36)
have the form
F
1
(z)

|z|
= log (4z) + 2 +O
_
1
z
log z
_
, (3.43)
F
2
(z)

|z|
=
1
6
_
log (4z) +
5
3
_
+ O
_
1
z
log z
_
. (3.44)
Let us consider the case of the massless eld, m
2
= 0. Taking the limit
m
2
0 in (3.31) for Re n > 2 we obtain the Green function of the massless
eld at coinciding points in the ndimensional space in the rst order in
external elds
G(x, x) = i(4)
n/2

_
2
n
2
_ _

_
n
2
1
__
2
(n 2)
( )
n
2
2
_
Q+
n 2
4(n 1)

1R
_
+ O(R
2
). (3.46)
Chapter 3. Summation of SchwingerDe Witt expansion 69
The formula (3.46) determines the analytic function of the dimension n
in the region 2 < Re n < 4. After the analytic continuation there appear
poles in the points n = 4, 6, 8, . . ., that reect the ultraviolet divergences,
and a pole in the point n = 2 reecting the infrared divergence. In the
twodimensional space eq. (3.46) gives
G(x, x) =
i
4
_
2
_
2
n 2
+C+ log

4
2
_
1
Q

1
1
R
_
+ O(R
2
). (3.47)
For even dimensions, n = 2N, (N 2), from (3.46) we obtain
G(x, x) =
i
(4)
N

(N 2)!
(2N 3)!
_

_
2
n 2N
+ (N 1)
2(2N 2) + log

4
2
_
N2
_
Q+
N 1
2(2N 1)

1R
_

1
2(2N 1)
2
N2
R
_
+ O(R
2
), (3.48)
where
(k) = C+

1lk1
1
l
. (3.49)
In particular, in the physical fourdimensional spacetime, n 4, we
have from (3.48)
G(x, x) =
i
(4)
2
_

_
2
n 4
+C2 + log

4
2
_
_
Q+
1
6

1R
_

1
18

1R
_
+ O(R
2
). (3.50)
The renormalized Green function of the massless eld can be obtained
by using the ultraviolet asymptotics of the formfactors (3.43) and (3.44) and
substituting the renormalization parameter instead of the mass, m
2

2
.
This reduces simply to a change of the normalization of the formfactors
(3.38).
Chapter 3. Summation of SchwingerDe Witt expansion 70
Let us stress that in the massless case the divergence of the coincidence
limit of the Green function in the rst order in external elds, (3.48), (3.50),
is proportional to the De Witt coecient [a
n
2
1
], (3.24). Therefore, in the
conformally invariant case [54, 53],
Q =
n 2
4(n 1)

1R, (3.51)
the linear part of the coecient [a
n
2
1
], (3.24), is equal to zero and the Green
function at coinciding points is nite in the rst order in external elds,
(3.47)(3.50). In odd dimensions, n = 2N + 1, the Green function, (3.31),
(3.46), is nite.
3.4 Summation of the terms of second order
in external elds
Let us calculate the coincidence limit of the De Witt coecients [a
k
] =<
0[a
k
> in the second order in external elds. From the formula (3.21) we
have up to cubic terms in external elds
[a
k
] =
1
_
2k1
k
_
< 0; k 1[F[0 >
+

1ik1;

0n
i
2(ki1)
_
2k1
i
_
_
2k1
k
__
2i+n
i
1
i
_
< 0; k i 1[F[n
i
>< n
i
; i 1[F[0 > +O(R
3
), (3.52)
where O(R
3
) denotes all omitted terms of third order in external elds.
The total number of terms, that are quadratic in external elds and con-
tain arbitrary number of derivatives, is innite. Therefore, we will also ne-
glect the total derivatives and the tracefree terms in (3.52) assuming to use
the results for the calculation of the coecients A
k
, (1.55), and the oneloop
eective action, (1.50), (1.54).
The number of the terms quadratic in external elds, in the coecients
A
k
is nite. Let us write the general form of the coecients A
k
up to the
Chapter 3. Summation of SchwingerDe Witt expansion 71
terms of the third order in external elds
A
k
=
_
d
n
xg
1/2
str
1
2
_

k
Q
k2
Q2
k
J

k3
J

+
k
Q
k2
R
+

1
_

k
R

k2
R

+
k
R
k2
R
_
+ O(R
3
)
_
, (3.53)
where J

.
All other quadratic invariants can be reduced to those written above up
to third order terms using the integration by parts and the Bianci identity.
For example,
_
dx
n
g
1/2
str 1

k
1

= 2
_
dx
n
g
1/2
str
_
J

k1
J

+ O(R
3
)
_
,
_
dx
n
g
1/2
R

k
R

=
_
dx
n
g
1/2
__
4R

k
R

R
k
R
_
+ O(R
3
)
_
.
(3.54)
Using (3.22), (2.121)(2.125) and (2.46) and the formulas of the Sect. 2.2
one can calculate the coecients
k
,
k
,
k
,
k
and
k
from (3.52) and (1.55).
Omitting the intermediate computations we present the result

k
=
k!(k 2)!
(2k 3)!
, (3.55)

k
=
k!(k 1)!
(2k 1)!
, (3.56)

k
= 2(k 1)
k!(k 1)!
(2k 1)!
, (3.57)

k
= 2
k!k!
(2k + 1)!
, (3.58)

k
= (k
2
k 1)
k!k!
(2k + 1)!
. (3.59)
Using the obtained coecients, (3.55)(3.59), we calculate the trace of
the transfer function (s), (1.43),
_
d
n
xg
1/2
str (s[x, x) =

k0
(is)
k
k!
A
k
. (3.60)
Chapter 3. Summation of SchwingerDe Witt expansion 72
Substituting (3.53) and (3.55)(3.59) in (3.60) we obtain
_
d
n
xg
1/2
str (s[x, x) =
_
d
n
xg
1/2
str
_

1 +is
_
Q +
1
6

1R
_
+
(is)
2
2
_
Qf
1
(is )Q + 2J

f
3
(is )
1

+ 2Qf
2
(is )R
+

1 (R

f
4
(is )R

+ Rf
5
(is )R)
_
+ O(R
3
)
_
, (3.61)
where
f
3
(z) =

k0
(k + 1)!
(2k + 3)!
z
k
, (3.62)
f
4
(z) =

k0
2
(k + 2)!
(2k + 5)!
z
k
, (3.63)
f
5
(z) =

k0
(k + 2)!
(2k + 5)!
(k
2
+ 3k + 1)z
k
, (3.64)
and f
1
(z) and f
2
(z) are given by the formulas (3.26), (3.27), (3.29) and (3.30).
The series (3.62)(3.64) converge for any nite z and dene in the same
manner as (3.26) and (3.27) entire functions. The summation of the series
(3.62) and (3.63) can be performed by means of the formula (3.28). Substi-
tuting (3.28) in (3.62) and (3.63) and summing over k we obtain the integral
representation of the functions f
3
(z) and f
4
(z):
f
3
(z) =
1
_
0
d
1
2

2
exp
_
1
4
(1
2
)z
_
, (3.65)
f
4
(z) =
1
_
0
d
1
6

4
exp
_
1
4
(1
2
)z
_
. (3.65)
Noting that from (3.64) it follows
f
5
(z) =
1
16
f
1
(z)
1
4
f
3
(z)
1
8
f
4
(z), (3.67)
Chapter 3. Summation of SchwingerDe Witt expansion 73
we obtain from the formulas (3.29), (3.65) and (3.66)
f
5
(z) =
1
_
0
d
1
8
_
1
2

2

1
6

4
_
exp
_
1
4
(1
2
)z
_
. (3.68)
Using (3.61) one can calculate the oneloop eective action up to cubic terms
in external elds. Substituting (3.61) in (1.50) and assuming Imm
2
< 0, after
integration over the proper time we obtain in the ndimensional space

(1)
=
1
2(4)
n/2
_
d
n
xg
1/2
str
_

n
2
_

1m
n
+
_
1
n
2
_
m
n2
_
Q +
1
6

1R
_
+
1
2

_
2
n
2
_
m
n4
_
Q

F
1
_

4m
2
_
Q
+2J

F
3
_

4m
2
_
J

+ 2Q

F
2
_

4m
2
_
R
+

1
_
R


F
4
_

4m
2
_
R

+ R

F
5
_

4m
2
_
R
_
_
+ O(R
3
)
_
, (3.69)
where

F
3
(z) =
1
_
0
d
1
2

2
_
1 + (1
2
)z
_n
2
2
, (3.70)

F
4
(z) =
1
_
0
d
1
6

4
_
1 + (1
2
)z
_n
2
2
, (3.71)

F
5
(z) =
1
_
0
d
1
8
_
1
2

2

1
6

4
_
_
1 + (1
2
)z
_n
2
2
, (3.72)
and

F
1
(z) and

F
2
(z) are given by the formulas (3.32) and (3.33).
Subtracting the pole in the dimension 1/(n 4), we obtain the renor-
malized eective action in the physical fourdimensional spacetime, (1.53),
(1.54), up to terms of third order in external elds

(1)ren
=
1
2(4)
2
_
d
4
xg
1/2
str
1
2
_
QF
1
_

4m
2
_
Q
Chapter 3. Summation of SchwingerDe Witt expansion 74
+2J

F
3
_

4m
2
_
J

+ 2Q

F
2
_

4m
2
_
R
+

1
_
R


F
4
_

4m
2
_
R

+ R

F
5
_

4m
2
_
R
_
_
+ O(R
3
)
_
,
(3.73)
where
F
3
(z) =
4
9
+
1
3z

1
6
_
1 +
1
z
_
J(z), (3.74)
F
4
(z) =
23
225
+
7
45z
+
1
15z
2

1
30
_
1 +
1
z
_
2
J(z), (3.75)
F
5
(z) =
1
900

37
360z

1
120z
2

1
60
_
1
3
z

1
4z
2
_
J(z), (3.76)
and F
1
(z), F
2
(z) and J(z) are given by the formulas (3.35)(3.37).
The formfactors F
3
(z), (3.74), F
4
(z), (3.75), and F
5
(z), (3.76), are nor-
malized by the conditions
F
3
(0) = F
4
(0) = F
5
(0) = 0. (3.77)
The normalization conditions of the formfactors (3.38) and (3.77) correspond
to the normalization of the eective action

(1)ren

m
2

= 0. (3.78)
Thus, by means of the partial summation of the local SchwingerDe Witt
expansion, (1.43), (3.60), we obtained a nonlocal expression for the eec-
tive action up to terms of third order in external elds, (3.73). Although the
power series, that dene the formfactors, i.e., the power series for the function
J(z), (3.37), converge only in the region [z[ < 1, z = 1 being the threshold
branch point, the expressions (3.35)(3.37) and (3.74)(3.76) are valid for any
z. That means that the proper time method automatically does the analyti-
cal continuation in the ultraviolet region [z[ . All the ambiguity, which
arises by the partial summation of the asymptotic expansion (3.60), reduces
to the arbitrariness in the boundary conditions for the formfactors. Specify-
ing the causal boundary conditions leads to the singlevalued expression for
Chapter 3. Summation of SchwingerDe Witt expansion 75
the eective action. Using the prescription m
2
m
2
i and the equation
(3.40) we obtain the imaginary parts of the formfactors (3.74)(3.76) in the
pseudoEuclidean region (z = x i) above the threshold (x < 1)
ImF
3
(x i) =

6
_
1 +
1
x
_

1 +
1
x
, (3.79)
ImF
4
(x i) =

30
_
1 +
1
x
_
2

1 +
1
x
, (3.80)
ImF
5
(x i) =

60
_
1
3
x

1
4x
2
_

1 +
1
x
. (3.81)
The imaginary parts of all formfactors, (3.41), (3.42), (3.79)(3.81), are
positive. This ensures the fulllment of the important condition
Im
(1)ren
> 0. (3.82)
The ultraviolet asymptotics [z[ of the formfactors (3.74)(3.76) have
the form
F
3
(z)

|z|
=
1
6
_
log (4z) +
8
3
_
+ O
_
1
z
log z
_
, (3.83)
F
4
(z)

|z|
=
1
30
_
log (4z) +
46
15
_
+ O
_
1
z
log z
_
, (3.84)
F
5
(z)

|z|
=
1
60
_
log (4z) +
1
15
_
+ O
_
1
z
log z
_
. (3.85)
Let us consider the case of the massless eld. Taking the limit m
2
0
in (3.69) in the region Re n > 2 we obtain the oneloop eective action for
the massless eld in the ndimensional space up to terms of third order in
external elds

(1)
=
1
2(4)
n/2
_
d
n
xg
1/2
str

_
2
n
2
_ _

_
n
2
1
__
2
(n 2)
_
1
2
Q( )
n
2
2
Q
+
1
2(n 1)
J

( )
n
2
3
J

+
n 2
4(n 1)
Q( )
n
2
2
R
+
1
4(n
2
1)

1
_
R

( )
n
2
2
R

+
1
8
(n
2
2n 4)R( )
n
2
2
R
_
Chapter 3. Summation of SchwingerDe Witt expansion 76
+O(R
3
)
_
. (3.86)
The formula (3.86) denes an analytical function of the dimension n in
the region 2 < Re n < 4. The analytical continuation leads to the poles in
the points n = 2, 4, 6, . . .. In odd dimensions the expression (3.86) is nite
and directly denes the eective action. In even dimensions the expression
(3.86) is proportional to the De Witt coecient A
n/2
, (3.53).
Separating the pole 1/(n 4) in (3.86) we obtain the eective action for
the massless eld in the physical fourdimensional spacetime up to terms
O(R
3
)

(1)
=
1
2(4)
2
_
d
n
xg
1/2
str
_
1
2
Q
_

2
n 4
C+ 2 log

4
2
_
Q
+
1
6
J

2
n 4
C+
8
3
log

4
2
_
J

+
1
6
Q
_

2
n 4
C+
5
3
log

4
2
_
R
+

1
_
1
60
R

2
n 4
C+
46
15
log

4
2
_
R

+
1
120
R
_

2
n 4
C+
1
15
log

4
2
_
R
_
+ O(R
3
)
_
. (3.87)
An analogous expression was obtained in the paper [45]. However, in that
paper the coecients
k
,
k
,
k
,
k
, and
k
, (3.55)(3.59), in the De Witt
coecient A
k
, (3.53), were not calculated. That is why in the paper [45]
it was assumed additionally that the power series in the proper time for
the functions f
i
(z), (3.26), (3.27) and (3.62)(3.64), converge as well as the
proper time integral (1.50) in the upper limit does. Besides, in [45] only di-
vergent and logarithmic, log( ), terms in (3.87) were calculated explicitly.
The complete result (3.87) together with the nite constants is obtained in
present work.
The renormalized eective action for the massless eld can be obtained
using the ultraviolet asymptotics of the formfactors (3.43), (3.44) and (3.83)
Chapter 3. Summation of SchwingerDe Witt expansion 77
(3.85) and substituting the renormalization parameter instead of the mass,
m
2

2
. This reduces just to a renormalization of the formfactors (3.38)
and (3.77). Although in the massless case the nite terms in (3.87) are
absorbed by the renormalization ambiguity, the nite terms in the ultravi-
olet asymptotics m
2
0 of the formfactors with xed normalization of the
eective action (3.78) are essential.
Let us consider the massless eld in the twodimensional space. In this
case there are no ultraviolet divergences; instead, the pole in the dimension
1/(n 2) in (3.86) reects the nonlocal infrared divergences

(1)
=
1
4(n 2)
_
d
n
xg
1/2
str
_
Q
1

Q + J

1
2
J

_
+
1
2(4)
_
d
2
xg
1/2
str
_
Q
_
log

4
2
+C
_
1

Q
+J

_
log

4
2
+C2
_
1
2
J

+ Q
1

R
+
1
12

1R
1

R + O(R
3
)
_
, (3.88)
where we made use of the fact that any twodimensional space satises iden-
tically the Einstein equations
R

=
1
2
g

R. (3.89)
For the scalar eld, J

= 0, in the conformally invariant case


(3.51), Q = 0, the eective action of the massless eld in the twodimensional
space is nite up to cubic terms in external elds, (3.88), and has the form

(1)
=
1
24(4)
_
d
2
xg
1/2
_
R
1

R + O(R
3
)
_
. (3.90)
On the other hand, any twodimensional space is conformally at. Therefore,
any functional of the metric g

is uniquely determined by the trace of the


functional derivative. Hence, the eective action of the massless conformally
invariant eld in the twodimensional space can be obtained by the integra-
tion of the conformal anomaly [45, 54, 95]. The exact answer has the form
Chapter 3. Summation of SchwingerDe Witt expansion 78
(3.90) without the third order terms O(R
3
), i.e., they vanish in the confor-
mally invariant case. When the conformal invariance is absent the infrared
divergences Q
1
Q and J

2
J

as well as the nite terms of higher orders


O(R
3
), (3.88), appear.
3.5 Summation of the terms without covari-
ant derivatives of external elds
In the Sects. 3.3 and 3.4 we summed up the linear and quadratic terms in
external elds in the asymptotic expansion of the transfer function (s) in
the powers of the proper time, (1.43). We showed that the corresponding
series (3.26), (3.27) and (3.62)(3.64) converge for any value of the proper
time and dene the entire functions (3.29), (3.30) and (3.65)(3.68).
In the present section we are going to sum up those terms in the asymp-
totic expansion of the transfer function at coinciding points, (1.43),
(s[x, x) =

k0
(is)
k
k!
[a
k
], (3.91)
that do not contain the covariant derivatives of the external eld. The sum-
mation of the linear and quadratic terms determines the highenergy asymp-
totics of the eective action ( R R
2
), whereas the terms without deriva-
tives determine its low-energy asymptotics ( R R
2
). We limit ourselves,
for simplicity, to the scalar case, i.e., we set 1

= 0.
The De Witt coecients in the coinciding points [a
k
] have the following
general form
[a
k
] =

0lk
Q
l
R
....
R
....
. .
kl
+O(R), (3.92)
where the sum contains a nite number of local terms constructed from
the curvature tensor and the potential term Q and O(R) denotes a nite
number of local terms constructed from the curvature tensor, the potential
term Q and their covariant derivatives, so that in the case of covariantly
constant curvature tensor and the potential term,

= 0,

Q = 0, (3.93)
these terms vanish, O(R) = 0.
Chapter 3. Summation of SchwingerDe Witt expansion 79
Thus, to nd the terms without covariant derivatives in the De Witt coef-
cients [a
k
], (3.92), it is sucient to restrict oneself to the constant potential
term, Q = const, and symmetric spaces, (3.93). Let us note that all sym-
metric spaces in two three and fourdimensional cases are characterized by
one dimensionful constant [164]. Therefore, the terms without derivatives in
(3.92) can be expressed only in terms of the scalar curvature R
[a
k
] =

0lk
c
k,l
R
kl
Q
l
+ O(R), (3.94)
where c
k,l
are numerical coecients.
On the other hand, from the denition of the De Witt coecients as the
coecients of the asymptotic expansion of the transfer function, (1.43), it is
easy to get the dependence of the De Witt coecients on the potential term
Q
[a
k
] =

0lk
_
k
l
_
c
kl
R
kl
Q
l
+ O(R), (3.95)
where
c
k
c
k,0
= R
k
[a
k
]

Q=0
are dimensionless De Witt coecients computed for Q = 0.
In order to calculate the coecients c
k
it is sucient to consider the case
of De Sitter space, (2.104), with the positive curvature
R

=
R
n(n 1)
_

_
, R = const > 0, (3.96)
since in other symmetric spaces, (3.93), they are the same as in the De Sitter
space.
The De Witt coecients in De Sitter space can be calculated in the lines
of the method developed in the Chap. 2. However, it is more convenient to
nd them by means of the asymptotic expansion of the Green function in the
coinciding points. It was obtained in the paper [80] in any ndimensional De
Sitter space
G(x, x) =
i
(4)
n/2
_
R
n(n 1)
_n
2
1

_
1
n
2
_

n
(), (3.97a)
Chapter 3. Summation of SchwingerDe Witt expansion 80

n
() =

_
n1
2
+ i
_

_
n1
2
i
_

_
1
2
+ i
_

_
1
2
i
_
, (3.97b)
where

2
=
n(n 1)
R
_
m
2
Q
n 1
4n
R
_
. (3.98)
The asymptotic expansion of the function (3.97b) in the inverse powers
of
2
has the form

n
() =

k0
d
k
(n)
2(
n
2
1k)
. (3.99)
The coecients d
k
(n) are determined by using the asymptotic expansion of
the gammafunction [168] from the relation

k0
d
k
(n)z
k
= exp
_
_
_

k1
(1)
k+1
k(2k + 1)
B
2k+1
_
n 1
2
_
z
k
_
_
_
, (3.100)
B
k
(x) being the Bernoulli polynomials [168], and have the form
d
0
(n) = 1,
d
k
(n) =

1lk
(1)
k+l
l!

k
1
,...,k
l
1
k
1
++k
l
=k
B
2k
1
+1
_
n1
2
_
k
1
(2k
1
+ 1)

B
2k
l
+1
_
n1
2
_
k
l
(2k
l
+ 1)
. (3.101)
On the other hand, using the SchwingerDe Witt presentation of the
Green function, (1.49), the proper time expansion of the transfer function,
(3.91), and the equation (3.95), we obtain the asymptotic SchwingerDe Witt
expansion for the Green function
G(x, x) =
i
(4)
n/2
_
R
n(n 1)
_n
2
1

k0

_
k + 1
n
2
_
k!
b
k
(n)
2(
n
2
1k)
,
(3.102)
where
b
k
(n) =
_
R
n(n 1)
_
k
[a
k
]

Q=
n1
4n
R
(3.103)
are the dimensionless De Witt coecients calculated for Q =
n1
4n
R.
Chapter 3. Summation of SchwingerDe Witt expansion 81
The total De Witt coecients for arbitrary Q are expressed in terms of
the coecients b
k
, (3.103), according to (3.95):
[a
k
] =

0lk
_
k
l
_
b
kl
(n)
_
R
n(n 1)
_
kl _
Q +
n 1
4n
R
_
l
+ O(R). (3.104)
Comparing (3.97), (3.99) and (3.102) we obtain the coecients b
k
(n)
b
k
(n) = (1)
k
k!

_
n
2
k
_

_
n
2
_
d
k
(n). (3.105)
To get the coecients b
k
(n) for integer values of the dimension n one has to
take the limit in (3.105). Thereby it is important to take into account the
dependence of the coecients d
k
(n), (3.101), on the dimension n. From the
denition of these coecients, (3.97b), (3.99), one can show that in integer
dimensions (n = 2, 3, . . .) they vanish for k [n/2]
d
k
(n) = 0,
_
n = 2, 3, 4 . . . ; k
_
n
2
__
. (3.106)
Let us consider rst the case of odd dimension (n = 3, 5, 7, . . .). In this
case the gammafunction in (3.105) does not have any poles and the formula
(3.105) immediately gives the coecients b
k
(n). From (3.106) it follows that
only rst (n 1)/2 coecients b
k
(n) do not vanish, i.e.,
b
k
(n) = 0,
_
n = 3, 5, 7, . . . ; k
n 1
2
_
. (3.107)
This is the consequence of the niteness of the Green function (3.97) in odd
dimension.
In even dimensions (n = 2, 4, 6, . . .) and for k
n
2
1 the expression
(3.105) is also singlevalued and immediately denes the coecients b
k
(n).
For k
n
2
there appear poles in the gammafunction that are suppressed by
the zeros of the coecients d
k
(n), (3.106). Using the denition of the coe-
cients d
k
(n), (3.100), (3.101), and the properties of the Bernoulli polynomials
[168], we obtain the coecients b
k
(n) in even dimension (n = 2, 4, 6, . . .) for
k n/2
b
k
(n) =
(1)
k
n
2
k!

_
n
2
_

_
k + 1
n
2
_

0l
n
2
1
(1)
l
k l
_
1 2
1+2l2k
_
B
2k2l
d
l
(n),
(3.108)
Chapter 3. Summation of SchwingerDe Witt expansion 82
_
n = 2, 4, 6, . . . ; k
n
2
_
,
where B
l
= B
l
(0) are the Bernoulli numbers and the coecients d
l
(n), (l
n/2 1), are calculated by means of the formula (3.101).
Let us list the dimensionless De Witt coecients b
k
(n), (3.103), in two
three and fourdimensional (physical) spacetime. Substituting n = 2 and
n = 4 in the formulas (3.101), (3.105) and (3.108) we obtain
b
k
(2) = (1)
k1
_
1 2
12k
_
B
2k
, (k 0), (3.109)
_

_
b
0
(4) = 1,
b
k
(4) = (1)
k
_
(k 1)
_
1 2
12k
_
B
2k

k
4
_
1 2
32k
_
B
2k2
_
, (k 1).
(3.110)
The expression for the coecients b
k
(n) in even dimension for k n/2,
(3.108), can be written in a more convenient and compact form. Using the
integral representation of the Bernoulli numbers [168]
B
2k
= (1)
k+1
4k
1 2
12k

_
0
dt
e
2t
+ 1
t
2k1
(3.111)
and the denition of the coecients d
k
(n), (3.97b), (3.99)(3.101), we obtain
from (3.108)
b
k
(n) =
4(1)
n
2
1
k!

_
n
2
_

_
k + 1
n
2
_

_
0
dt t
e
2t
+ 1

_
n1
2
+ it
_

_
n1
2
it
_

_
1
2
+ it
_

_
1
2
it
_
t
2kn
,
(3.112)
_
n = 2, 4, 6, . . . ; k
n
2
_
.
Thus we have obtained all the terms in the De Witt coecients [a
k
], that
do not contain covariant derivatives, (3.104), where the coecients b
k
(n) are
given by the formulas (3.105), (3.101) and (3.108)(3.112).
Using the obtained De Witt coecients, (3.104), one can calculate the
transfer function in the coinciding points, (3.91). Substituting (3.104) in
(3.91) and summing the powers of the potential term we obtain
(s[x, x) = exp
_
is
_
Q+
n 1
4n
R
__

_
isR
n(n 1)
_
+ O(R), (3.113)
Chapter 3. Summation of SchwingerDe Witt expansion 83
where
(z) =

k0
z
k
k!
b
k
(n). (3.114)
Let us divide the series (3.114) in two parts
(z) =
1
(z) +
2
(z), (3.115)
where

1
(z) =

0k[
n
2
]1
z
k
k!
b
k
(n), (3.116)

2
(z) =

k[
n
2
]
z
k
k!
b
k
(n). (3.117)
The rst part
1
(z), (3.116), is a polynomial. Using the expression (3.105)
for coecients b
k
and the denition of the coecients d
k
, (3.97), (3.99), one
can write it in the integral form

1
(z) = 2
(z)
n
2
1

_
n
2
_

_
0
dt t exp (t
2
)

_
n1
2
+ i
t

z
_

_
n1
2
i
t

z
_

_
1
2
+ i
t

z
_

_
1
2
i
t

z
_
. (3.118)
The second part
2
(z), (3.117), is an asymptotic series. In odd dimension
all coecients of this series are equal to zero, (3.107). Therefore, in this case

2
(z) = 0 up to an arbitrary nonanalytical function in the vicinity like
exp(const/z).
From the obtained expressions (3.108)(3.112) one can get the asymp-
totics of the coecients b
k
(n) as k in even dimensions
b
k
(n)

k
= 4
(1)
n
2
1

_
n
2
_
k!(2k 1)!

_
k + 1
n
2
_
(2)
2k
, (n = 2, 4, 6, . . .). (3.119)
Herefrom it is immediately seen that in this case the asymptotic series (3.117)
diverges for any z ,= 0, i.e., its convergence radius is equal to zero and
the point z = 0 is a singular point of the function
2
(z). This makes the
asymptotic series of the terms without covariant derivatives of external elds
very dierent from the corresponding series of linear and quadratic terms in
Chapter 3. Summation of SchwingerDe Witt expansion 84
external elds, (3.26), (3.27), (3.62)(3.64), which converge for any z and
dene entire functions, (3.29), (3.30), (3.65)(3.68).
Based on the divergence of the asymptotic series (3.117), it is concluded
in the paper [166] that the SchwingerDe Witt representation of the Green
function is meaningless. However, the asymptotic divergent series (3.117)
can be used to obtain quite certain idea about the function
2
(z). To do
this one has to make use of the methods for summation of asymptotic series
discussed in Sect. 3.1.
Let us dene the function
2
(z) according to the formulas (3.5) and (3.3):

2
(z) =

_
0
dy y
1
e
y
B(zy

), (3.120)
where
B(z) =

k
n
2
z
k
k!
b
k
(k + )
, (3.121)
and and are some complex numbers, such that Re
_

n
2
+
_
> 0. The
series (3.121) for the Borel function B(z) converges in case Re > 1 for
any z and in case Re = 1 for [z[ <
2
. This can be seen easily using the
formulas (3.4) and (3.119). Let us substitute the integral representation of
the coecients b
k
, (3.112), in (3.121) and change the order of integration and
summation. We get
B(z) = 4
(z)
n
2

_
n
2
_

_
0
dt t
e
2t
+ 1

_
n1
2
+ it
_

_
n1
2
it
_

_
1
2
+ it
_

_
1
2
it
_
H(zt
2
), (3.122)
where
H(z) =

k0
z
k
k!
_
k + +
n
2
_
. (3.123)
The series (3.123) converges for any z and denes an entire function. For
example, for = 1, = 1
n
2
it reduces to the Bessel function of zeroth
order
H(z)

=1, =1
n
2
= J
0
_
2

z
_
. (3.124)
One can show that the integral (3.122) converges for sure on the negative
part of the real axis, Re z < 0, Imz = 0. Therefore, for such z it gives the
Chapter 3. Summation of SchwingerDe Witt expansion 85
analytical continuation of the series (3.121). The whole Borel function B(z)
on the whole complex plane can be obtained by the analytic continuation of
the integral (3.122).
Let us substitute now the expression (3.122) for the Borel function in
the integral (3.120) and change the order of integration over t and over y.
Integrating over y and summing over k we obtain

2
(z) = 4
(z)
n
2

_
n
2
_

_
0
dt t
e
2t
+ 1

_
n1
2
+ it
_

_
n1
2
it
_

_
1
2
+ it
_

_
1
2
it
_
exp(zt
2
). (3.125)
This integral converges in the region Re z 0. In the other part of the
complex plane the function
2
(z) is dened by analytical continuation. In
this way we obtain a singlevalued analytical function in the complex plane
with a cut along the positive part of the real axis from 0 to . Thus the
point z = 0 is a singular (branch) point of the function
2
(z). It is this fact
that makes the power series over z, (3.117), to diverge for any z ,= 0.
In the region Re z < 0 one can obtain analogous expression for the total
function (z), (3.115). Changing the integration variable t t

z in
(3.118) and adding it to (3.125) we obtain
(z) = 2
(z)
n
2

_
n
2
_

_
0
dt t tanh(t)

_
n1
2
+ it
_

_
n1
2
it
_

_
1
2
+ it
_

_
1
2
it
_
exp(zt
2
). (3.126)
Thus, we summed up the divergent asymptotic series of the terms without
covariant derivatives of the external eld in the transfer function, (3.113),
(3.114), (3.126).
Using the obtained transfer function, (3.113), (3.118), (3.126), one can
calculate the oneloop eective action, (1.50), up to terms with covariant
derivatives O(R). In order to obtain the renormalized eective action
(1)ren
with the normalization condition (3.78) it is sucient to subtract from the
transfer function (s), (3.113), the potentially divergent terms and integrate
over the proper time. As the result we obtain in twodimensional space

(1)ren

n=2
=
1
2(4)
_
d
2
xg
1/2
_
_
m
2
Q
1
6
R
_
log
m
2
Q
1
8
R
m
2
+Q +
1
8
R 2R
1
_
2
m
2
Q
R

1
4
_
+ O(R)
_
, (3.127)
Chapter 3. Summation of SchwingerDe Witt expansion 86
and in physical fourdimensional spacetime

(1)ren

n=4
=
1
2(4)
2
_
d
4
xg
1/2
_

1
2
_
m
4
2m
2
_
Q+
1
6
R
_
+Q
2
+
1
3
QR +
29
1080
R
2
_
log
m
2
Q
9
48
R
m
2

1
2
m
2
_
Q+
9
48
R
_
+
5
4
Q
2
+
43
96
QR
+
41
1024
R
2
+
1
36
R
2

2
_
12
m
2
Q
R

9
4
_
+ O(R)
_
,
(3.128)
where

1
(z) =

_
0
dt t
e
2t
+ 1
log
_
1
t
2
z
_
, (3.129)

2
(z) =

_
0
dt t
e
2t
+ 1
_
t
2
+
1
4
_
log
_
1
t
2
z
_
. (3.130)
In odd dimensions the eective action is nite. Substituting (3.118),
(3.113) in (1.50) and integrating over the proper time we obtain, for example,
in threedimensional space

(1)

n=3
=
1
3(4)
_
d
3
xg
1/2
_
_
m
2
Q
1
6
R
_
3/2
+ O(R)
_
. (3.131)
In higher dimensions the terms without covariant derivatives of the cur-
vature tensor, (3.92), no longer reduce only to powers of the scalar curvature
R, (3.94), they also contain invariants of the Ricci and Weyl tensors.
Thus via the partial summation of the covariantly constant terms of
the SchwingerDe Witt asymptotic expansion (an expansion in background
dimension) we obtained a nonanalytical (in the background eld expres-
sion) for the eective action up to terms with covariant derivatives, (3.127)
(3.131). Although the corresponding asymptotic series diverge, the expres-
sions (3.129) and (3.130) are dened in the whole complex plane z with the
Chapter 3. Summation of SchwingerDe Witt expansion 87
cut along the positive part of the real axis. There appears a natural arbi-
trariness connected with the possibility to choose the dierent banks of the
cut.
Chapter 4
Higherderivative quantum
gravity
4.1 Quantization of gauge eld theories. Uni-
que eective action
Let M =
i
be the conguration space of a boson gauge eld and S()
be a classical action functional that is invariant with respect to local gauge
transformations

i
= R
i

()

, (4.1)
forming the gauge group G. Here

are the group parameters and R


i

()
are the local generators of the gauge transformations that form a closed Lie
algebra
[

IR

,

IR

] = C

IR

, (4.2)
where

IR

R
i

()

i
(4.3)
and C

are the structure constants of the gauge group satisfying the Jacobi
identity
C

[
C

]
= 0. (4.4)
The classical equations of motion determined by the action functional S()
have the form

i
() = 0, (4.5)
88
Chapter 4. Higherderivative quantum gravity 89
where
i
S
,i
is the extremal of the action. The equation (4.5) denes the
mass shell in the quantum perturbation theory.
The equations (4.5) are not independent. The gauge invariance of the
action leads to the rst type constraints between the dynamical variables
that are expressed through the Noether identities

IR

S = R
i

i
= 0. (4.6)
The physical dynamical variables are the group orbits (the classes of gauge
equivalent eld congurations), and the physical conguration space is the
space of orbits / = M/G. To have coordinates on the orbit space one
has to put some suuplementary gauge conditions (a set of constraints) that
isolate in the space M a subspace /

, which intersects each orbit only in one


point. Each orbit is represented then by the point in which it intersects the
given subspace /

. If one reparametrizes the initial conguration space by


the new variables, M = I
A
,

, where I
A
are the physical gauge-invariant
variables enumerating the orbits, /= I
A
, and

are group variables that


enumerate the points on each orbit, G =

, then the gauge conditions


can be written in the form

() =

, (4.6a)
where

are some constants, i.e.,


,i
= 0. Thus we obtain the coordinates
on the physical space /

= I
A
,

.
The group variables

transform under the action of the gauge group


analogously to (4.1)

= Q

()

, (4.7)
where
Q

() =
,i
()R
i

(), (det Q ,= 0), (4.8)


are the generators of gauge transformations of group variables. They form a
representation of the Lie algebra of the gauge group
[

I Q

,

I Q

] = C

I Q

, (4.9)
where

I Q

.
Chapter 4. Higherderivative quantum gravity 90
All physical gaugeinvariant quantities (in particular, the action S())
are expressed only in terms of the gaugeinvariant variables I
A
and do not
depend on the group variables

S() =

S(I
A
()). (4.10)
The action

S(I
A
) is an usual nongaugeinvariant action. Therefore one can
quantize the theory in the variables I
A
, and then go back to the initial eld
variables
i
. The functional integral for the standard eective action takes
the form [45]
exp
_
i
h
()
_
=
_
d/()(

()

) det Q()
exp
_
i
h
_
S() (
i

i
)
,i
()
_
_
, (4.11)
where /() is a local measure. The measure /() is gauge invariant up
to the terms R
i
,i
and C

that are proportional to derivatives of the


function at coinciding points (0) and that should vanish by the regular-
ization. The exact form of the measure /() must be determined by the
canonical quantization of the theory. In most practically important cases
/() = 1 + (0)(. . .) [45]. Therefore, below we will simply omit the mea-
sure /().
Since the eective action (4.11) must not depend on the arbitrary con-
stants

, one can integrate over them with a Gaussian weight. As result we


get
exp
_
i
h
()
_
=
_
ddet Q()(det H)
1/2
exp
_
i
h
_
S()
1
2

()H

() (
i

i
)
,i
()
__
,
(4.12)
where H

is a nondegenerate matrix (det H ,= 0), that does not depend


on quantum eld (H

/
i
= 0). The determinants of the operators Q and
H in (4.12) can be also represented as result of integration over the anti
commuting variables, so called Faddeev-Popov [15, 14] and Nielsen-Kallosh
[31, 32] ghosts.
Chapter 4. Higherderivative quantum gravity 91
Using the equation (4.12) we nd the oneloop eective action
= S + h
(1)
+ O( h
2
), (4.13)

(1)
=
1
2i
log
det
det H(det F)
2
, (4.14)
where

ik
= S
,ik
+
i
H

k
,
i

,i
()

=
, (4.15)
F = Q()

=
. (4.16)
On the mass shell,
,i
= 0, the standard eective action, (4.12), and
therefore the Smatrix too, does not depend neither on the gauge (i.e., on
the choice of arbitrary functions

and H

) nor on the parametrization


of the quantum eld [34, 35, 45]. However, o mass shell the background
eld as well as the Green functions and the eective action crucially depend
both on the gauge xing and on the parametrization of the quantum eld.
Moreover, the eective action is not, generally speaking, a gaugeinvariant
functional of the background eld because the usual procedure of the gauge
xing of the quantum eld automatically xes the gauge of the background
eld.
The gauge invariance of the eective action o mass shell can be preserved
by using the De Witts background eld gauge. In this gauge the functions

() and the matrix H

depend parametrically on the background eld


[14, 34]

() =

(, ),

(, ) = 0, H

= H

(),
and are covariant with respect to simultaneous gauge transformations of the
quantum and background elds, i.e., they form the adjoint representation
of the gauge group

IR

()

(, ) +

IR

()

(, ) + C

(, ) = 0,

IR

()H

() C

() C

() = 0. (4.17)
In the background eld gauge method one xes the gauge of the quantum
eld but not the gauge of the background eld. In this case the standard
eective action (4.12) will be gauge invariant functional of the background
Chapter 4. Higherderivative quantum gravity 92
eld provided the generators of gauge transformations R
i

() are linear in
the elds, i.e., R
i
,kn
() = 0 [34].
However, the oshell eective action still depends parametrically on the
choice of the gauge (i.e., on the functions

, and H
,
), as well as on the
parametrization of the quantum eld . As a matter of fact this is the same
problem, since the change of the gauge is in essence a reparametrization of
the physical space of orbits /

[45]. Consequently we do not have any unique


eective action o mass shell.
The oshell eective action can give much more information about quan-
tum processes than just the Smatrix. It should give the eective equations
for the background eld with all quantum corrections,
,i
() = 0.
A possible solution of the problem of constructing the oshell eective
action was proposed by Vilkovisky [45, 46] both for usual and the gauge eld
theories.
The condition of the reparametrization invariance of the functional S()
means that S() should be a scalar on the conguration space M which is
treated as a manifold with coordinates
i
. The non-covariance of the eective
action () can be traced to the source term (
i

i
)
,i
() in (4.12) as
the dierence of coordinates (
i

i
) is not a geometric object. One can
achieve covariance if one replaces this dierence by a twopoint function,

i
(, ), that transforms like a vector with respect to transformations of the
background eld and like a scalar with respect to transformations of the
quantum eld. This quantity can be constructed by introducing a symmetric
gaugeinvariant connection
i
mn
() on M and by identifying
i
(, ) with
the tangent vector at the point to the geodesic connecting the points
and . More precisely,
i
(, ) is dened to be the solution of the equation

i
=
i
, (4.18)
where

i
=

i
+
i
km
()
m
, (4.19)
with the boundary condition

=
= 0. (4.20)
Thus we obtain a unique eective action

() according to Vilkovisky
exp
_
i
h

()
_
=
_
d(det H())
1/2
det Q(, )
Chapter 4. Higherderivative quantum gravity 93
exp
_
i
h
_
S()
1
2

(, )H

()

(, ) +
i
(, )

,i
()
__
,
(4.21)
Since the integrand in (4.21) is a scalar on M, the equation (4.21) denes a
reparametrizationinvariant eective action. Let us note, that such a modi-
cation corresponds to the denition of the background eld in a manifestly
reparametrizationinvariant way,
<
i
(, ) >= 0, (4.22)
instead of the usual denition
i
=<
i
>.
The construction of the perturbation theory is performed by the change
of the variables in the functional integral,
i

i
(, ), and by expanding
all the functionals in the covariant Taylor series
S() =

k0
(1)
k
k!

i
1

i
k
_

(i
1

i
k
)
S()
_

=
. (4.23)
The diagrammatic technique for the usual eective action results from the
substitution of the covariant functional derivatives
i
instead of the usual
ones in the expression for the standard eective action (up to the terms
(0) that are caused by the Jacobian of the change of variables).
In particular, in the oneloop approximation, (4.13)(4.16),

(1)
=
1
2i
log
det

det H(det F)
2
, (4.24)
where

ik
=
i

k
S +
i
H

k
. (4.25)
To construct the connection on the physical conguration space /

let
us introduce, rst of all, a non-degenerate gaugeinvariant metric E
ik
() in
the initial conguration space M that satises the Killing equations
T
m
R
n
+T
n
R
m
= 0, (4.26)
where R
m
= E
mk
R
k

and T
m
means the covariant derivative with Christoel
connection of the metric E
ik
_
i
jk
_
=
1
2
E
1im
(E
mj,k
+ E
mk,j
E
jk,m
). (4.27)
Chapter 4. Higherderivative quantum gravity 94
The metric E
ik
() must ensure the non-degeneracy of the matrix
N

= R
i

E
ik
R
k

, (det N ,= 0). (4.28)


This enables one to dene the De Witt projector [14, 33]

m
i
=
m
i
R
m

i
,
2

, (4.29)
where
B

n
= N
1
R
i

E
in
. (4.30)
The projector

projects every orbit to one point,

m
i
R
i

= 0 (4.31)
and is orthogonal to the generators R
k

in the metric E
ik
R
k

E
km

m
i
= 0. (4.32)
Therefore, the subspace

M is the space of the orbits.


The natural physical conditions lead to the following form of the connec-
tion on the conguration space [45, 46]

i
mn
=
_
i
mn
_
+ T
i
mn
, (4.33)
where
T
i
mn
= 2B

(m
T
n)
R
i

+ B

(m
B

n)
R
k

T
k
R
i

. (4.34)
The main property of the connection (4.33) is

n
R
i

R
i

. (4.35)
It means that the transformation of the quantity
i
under the gauge transfor-
mations of both the quantum and background eld is proportional to R
i

(),
and, therefore, ensures the gauge invariance of the term
i

,i
in (4.21). This
leads then to the reparametrization invariance and the independence of the
unique eective action

() on gauge xing o mass shell.
On the other hand, it is obvious that on mass shell the unique eective
action coincides with the standard one,

()

onshell
= ()

onshell
,
Chapter 4. Higherderivative quantum gravity 95
and leads to the standard Smatrix [45].
To calculate the unique eective action one can choose any gauge. The
result will be the same. It is convenient to use the orthogonal gauge, i.e.,
simply put the nonphysical group variables equal to zero,

(, ) = R
i

()E
ik
()
k
(, ) = 0, (4.36)
i.e.,
i
=
i

, where
i

i
n

n
. The non-metric part of the connection
T
i
mn
satises the equation

m
k
T
i
mn

n
j
= 0. (4.37)
Using this equation one can show that it does not contribute to the quantity

. Therefore in the orthogonal gauge (4.36) the quantity T


i
mn
in the con-
nection (4.33) can be omitted. As a result we obtain for the unique eective
action the equation
exp
_
i
h

()
_
=
_
d
_
R
i
()
i
(, )
_
det

Q(, )
exp
_
i
h
_
S() +
i
(, )

,i
()
_
_
, (4.38)
where

(, ) = R
i
()R
i

()
The change of variables
i
(, ) and covariant expansions of all
functionals of the form
S() =

k0
(1)
k
k!

i
1

i
k
_
T
(i
1
T
i
k
)
S()
_

=
(4.39)
lead (up to terms (0)) to the standard perturbation theory with simple
replacement of usual functional derivatives by the covariant ones with the
Christoel connection, T
i
. In particular, in the oneloop approximation we
have

(1)
=
1
2i
log
det

(det N)
2
, (4.40)
where

ik
= T
i
T
k
S = S
,ik
+
_
j
ik
_
S
,j
(4.41)
Chapter 4. Higherderivative quantum gravity 96
4.2 Oneloop divergences of higherderiva-
tive quantum gravity
The theory of gravity with higher derivatives as well as the Einstein gravity
and any other metric theory of gravity is a nonAbelian gauge theory with
the group of dieomorphisms G of the spacetime as the gauge group. The
complete conguration space M is the space of all pseudoRiemannian met-
rics on the spacetime, and the physical conguration space, the space of
orbits /, is the space of geometries on the spacetime.
We will parametrize the gravitational eld by the metric tensor of the
spacetime

i
= g

(x), i (, x). (4.42)


The parameters of the gauge transformation are the components of the in-
nitesimal vector of the coordinate transformation
x

(x),

(x), (, x). (4.43)


The local generators of the gauge transformations in the parametrization
(4.42) are linear in the elds, i.e., R
i
,kn
= 0, and have the form
R
i

= 2
(
g
)
(x, y), i (, x), (, y). (4.44)
Here and below, when writing the kernels of the dierential operators, all the
derivatives act on the rst argument of the function.
The local metric tensor of the conguration space, that satises the
Killing equations (4.26), has the form
E
ik
= g
1/2
E
,
(x, y), i (, x), k (, y),
E
,
= g
(
g
)

1
4
(1 +)g

, (4.45)
where ,= 0 is a numerical parameter. In the fourdimensional pseudo
Euclidean spacetime it is equal to the determinant of the 10 10 matrix:
= det
_
g
1/2
E
,
_
. (4.46)
Chapter 4. Higherderivative quantum gravity 97
The Christoel symbols of the metric E
ik
have the form
_
i
jk
_
=
_
,

_
(x, y)(x, z),
i (, x), j (, y), k (, z),
_
,

_
=
(
(
g
)(

)
)
+
1
4
_
g

+ g

+
1
E
,
g

_
,

)
. (4.47)
The matrix N

(), (4.28), is a second order dierential operator of the form


N

() = 2g
1/2
_
g

+
1
2
( 1)

_
(x, y). (4.48)
The condition that the operator N

, (4.48), should be nondegenerate in


the perturbation theory on the at background, det N[
R=0
,= 0, imposes a
constraint on the parameter of the metric, ,= 3.
Let us write the classical action with quadratic terms in the curvature of
general type
S() =
_
d
4
xg
1/2
_
R

+
1
2f
2
C
2

1
6
2
R
2

1
k
2
R + 2

k
4
_
, (4.49)
where R


1
4

is the topological term,

is the
antisymmetric Levi-Civita tensor, C
2
C

is the square of the


Weyl tensor, is the topological coupling constant, f
2
the Weyl one,
2

the conformal one, k


2
the Einstein one, and = k
2
is the dimensionless
cosmological constant. Here and below we omit the surface terms R that
do not contribute neither to the equations of motions nor to the Smatrix.
The extremal of the classical action, (4.5), and the Noether identities,
(4.6), have the form

i
() =

S
g

= g
1/2
_
1
k
2
_
R

1
2
g

R + g

_
+
1
f
2
_
2
3
(1 +)R(R

1
4
g

R) +
1
2
g

2R

R


+
1
3
(1 2)

R R

+
1
6
(1 + 4)g

R
__
, (4.50)
Chapter 4. Higherderivative quantum gravity 98

= 0, (4.51)
where f
2
/(2
2
).
Let us calculate the oneloop divergences of the standard eective action
(4.14). From (4.14) we have up to the terms (0)

div
(1)
=
1
2i
_
log det [
div
log det H[
div
2 log det F[
div
_
. (4.52)
The second variation of the action (4.49) has the form
S
,ik
=
_

F
(0)


F
(2)

+

F
(3)


F
(3)
+

F
(4)
_
g
1/2
(x, y), (4.53)
where

F
(0)
,

F
(2)
,

F
(3)
and

F
(4)
mean the tensor 1010 matrices (

F
(0)
= F
,
(0)
, etc.). They satisfy the following symmetry relations

F
T
(2)
=

F
(2)
,

F
T
(3)
=

F
(3)
,

F
T
(4)
=

F
(4)
,

F
(2)
=

F
(2)
, (4.54)
with the symbol T meaning the transposition, and have the form
F
,
(0)
=
1
2f
2
_
g

_
g
(
g
)

1 + 4
3
g

_
+
1 + 4
3
_
g

+ g

_
+
2
3
(1 2)

2
(

g
)(

_
,
F
,
(2)
=
1
2f
2
_
2R

_
g
(
g
)

1
2
g

_
+ 4g

R
(||)
+2
(
(
R
)(

)
)
+ 2
_
g

(
(
R
)
)
+ g

(
(
R
)
)
_
Chapter 4. Higherderivative quantum gravity 99
4
_
R
(
(
g
)(

)
)
+ R
(
(
g
)(

)
)
_
+
4
3
(1 +)
_
R

_
+ R

__
+
_
m
2
2
+
2
3
(1 +)R
_
_
g

g
(
g
)
+ g

+ 2
(
(
g
)(

)
)
_
_
,
F
,
(3)
=
1
2f
2
_
1
2
_
g

(
R
)

(
R
)

_
+
1
3
(1 2)
_
g

_
+ g
((

)
R
)

g
((

)
R
)

+
1
12
(1 + 4)
_
g


)
R g


)
R
_
+
1
6
(1 2)
_

g
)(

)
R
(

g
)(

)
R
_
+
2
3
(2 )
_


)
R


)
R

_
+
3
2
_

(
R
)(


(
R
)(

_
_
,
F
,
(4)
=
1
2f
2
_

1
2
_
g

+ g

_
+ 4R
( )

R

_
g
(
g
)

1
2
g

_
_
R

1
3
(1 +)R
2
+
1
3
(1 + 4) R m
2
2
(R 2)
_
Chapter 4. Higherderivative quantum gravity 100

3
2
R

_
g

R


+ g

R


_
+ R
(
R
)
+
_
m
2
2
+
2
3
(1 +)R
_
_
R

+ R

3g
((
R
))
R
(||)
_
+6g
((
R
) )

R

+
1
2
_
R
( )(

R
)
+ R
( )(

R
)
_

4
3
(1 +)R

_
g

R + g

R
_
+2 R
(||)

4
3
(1 2)g
((

)
R
+2 R
((
g
))
+
2
3
(1 +)
_

)
R

+
(

)
R

_
_
, (4.55)
where m
2
2
= f
2
/k
2
.
Next let us choose the most general linear covariant De Witt gauge con-
dition [33]

(, ) = R
k

()E
ki
()h
i
, (4.56)
where h
i
=
i

i
. In usual notation this condition reads

= 2g
1/2

_
h

1
4
(1 +)

h
_
, (4.57)
where raising and lowering the indices as well as the covariant derivative
are dened by means of the background metric g

and h h

. The ghost
operator F, (4.8), (4.16), in the gauge (4.56) is equal to the operator N,
(4.48),
F

= N

= 2g
1/2
_
g

+
1
2
( 1)

_
(x, y). (4.57a)
Chapter 4. Higherderivative quantum gravity 101
It is obvious that for the operator , (4.15), to be nondegenerate in the
at spacetime it is necessary to choose the matrix H as a second order
dierential operator
H

=
1
4
2
g
1/2
g

+ R

+ P

(x, y), (4.58)


where and are numerical constants, P

is an arbitrary symmetric tensor,


e.g.
P

= p
1
R

+ g

_
p
2
R + p
3
1
k
2
_
, (4.59)
p
1
, p
2
and p
3
being some arbitrary numerical constants. Such form of the op-
erator H does not increase the order of the operator , (4.15), and preserves
its locality. Thus we obtain a very wide sixparameter (, , , p
1
, p
2
, p
3
)
class of gauges. In particular, the harmonic De DonderFockLandau gauge,

_
h

1
2

h
_
= 0, (4.60)
corresponds to = 1 and = 0. For = 0 the dependence on other
parameters disappear.
It is most convenient to choose the minimal gauge

2
=
2
0
f
2
, =
0

1
3
(1 2), =
0

3
1 +
, (4.61)
which makes the operator , (4.15), diagonal in leading derivatives, i.e., it
takes the form

ik
=
1
2f
2
_

E
(0)
2
+

+

V

+

P
_
(x, y), (4.62)
where

E
(0)
= E
,
( =
0
),

D

= D
,

= V
,

,

P = P
,
,
are some tensor matrices that satisfy the same symmetry conditions, (4.54).
They have the following form
D
,

=
1
2f
2
_
2R

_
g
(
g
)

1
2
_
1 +
p
1
16
_
1 + 4
1 +
_
2
_
g

_
Chapter 4. Higherderivative quantum gravity 102
+2
_
1 +
p
1
4
_
1 + 4
1 +
__
_
g

R
(
(

)
)
+ g

R
(
(

)
)
_
+ 4g

R
(||)
2p
1

(
(
R
)(

)
)
4
_

(
(
g
)(
R
)
)
+
(
(
g
)(
R
)
)
_
+
4
3
(1 +)
_
R

_
+ R

__
+
_
m
2
2
+
2
3
(1 +)R
_
_
g

g
(
g
)
+ g

+ 2
(
(
g
)(

)
)
_
+
_
p
2
R + p
3
1
k
2
_
_
2
(
(
g
)(

)
)
+
1
2

1 + 4
1 +
_
g

+ g

1
8
_
1 + 4
1 +
_
2
g

__
,
V
,

=
1
2f
2
_
_
1
2
+
p
1
8

1 + 4
1 +
_
_
g

(
R
)

(
R
)

_
+
1
3
(1 2)
_
g

_
+g
((

)
R
)

g
((

)
R
)

+
1
12
(1 + 4)
_
1
3
4

p
1
1 +
_
_
g


)
R g


)
R
_
+
_
1 2
6

p
2
2
_
_

g
)(

)
R
(

g
)(

)
R
_
+
2
3
(2 )
_


)
R


)
R

_
+
_
2 +
p
1
2
_
_

(
R
)(


(
R
)(

_
_
,
P
,
=
1
2f
2
_

1
2
_
1 +
p
1
4

1 + 4
1 +
_
_
g

+ g

_
Chapter 4. Higherderivative quantum gravity 103

_
g
(
g
)

1
2
g

_
_
R

1
3
(1 +)R
2
+
1
3
(1 + 4) R m
2
2
(R 2)
_

3
2
_
1
p
1
12

1 + 4
1 +
_
R

_
g

R


+ g

R


_
+(2 +p
1
)R
(
R
)

1
2
p
1
_
R
(
R
)( )

+ R
(
R
)( )

_
+6g
((
R
) )

R

4
3
(1 +)R

+ 4R
( )

R
( )

+
_
m
2
2
+
2
3
(1 +)R
_
_
R

+ R

3g
((
R
))
R
(||)
_
+
_
p
2
R + p
3
1
k
2
_
_
g
((
R
))
R
(||)
_
+ 2 R
(||)

_
g

R + g

R
_
+ 2 R
((
g
))
+
2
3
(1 +)
_
g

+
(

)
R

+
(

)
R

4
3
(1 2)g
((

)
R
_
. (4.63)
The divergences of the determinants of the operators H, (4.58), F, (4.57a),
and , (4.62), can be calculated by means of the algorithms for the non-
minimal vector operator of second order and the minimal tensor operator
of forth order. These algorithms were obtained rst in [152154] and con-
rmed in [77] by using the generalized SchwingerDe Witt technique. In the
Chapter 4. Higherderivative quantum gravity 104
dimensional regularization up to the terms R they have the form
log det

+ R

+ P

div
= i
2
(n 4)(4)
2
_
d
4
xg
1/2
_

8
45
R

+
7
60
C
2
+
1
36
R
2
+
1
6
( + 6)R

1
12
( + 2)RP +
1
48

2
P
2
+
1
24
(
2
+ 6 + 12)P

_
, (4.64)
where


1
, P P

,
and
log det
_

E
2
+

+

V

+

P
_

div
= i
2
(n 4)(4)
2
_
d
4
xg
1/2
tr
_

1
_

1
180
R

+
1
60
C
2
+
1
36
R
2
_
+
1
6


E
1

P +
1
12
R

E
1

D
1
6
R


E
1

D

+
1
48

E
1

D

E
1

D +
1
24

E
1

D


E
1

D

_
, (4.65)
where

D

D

E,

D

,

V

and

P are the tensor matrices (

E = E
AB
,

E
1
= E
1
AB
,

D

=
D
AB
etc.),

1 =
A
B
,

1

= 1
A
B
is the commutator of covariant derivatives
of the tensor eld
[

]h
A
= 1
A
B
h
B
, (4.66)
tr means the matrix trace and n is the dimension of the spacetime.
Chapter 4. Higherderivative quantum gravity 105
In our case, h
A
= h

,
1
A
B
= 1

,
= 2
(
(
R
)
)
. (4.67)
Using eqs. (4.64) and (4.65) we obtain in the minimal gauge (4.61)
log det H

div
= i
2
(n 4)(4)
2
_
d
4
xg
1/2
_

8
45
R

+
7
60
C
2
+
1
36
R
2
+
13 + 10
12(1 +)
R

5 + 2
24(1 +)
RP
+
1
192
_
1 2
1 +
_
2
P
2
+
28
2
+ 80 + 61
96( + 1)
2
P

_
,
(4.68)
log det F

div
= i
2
(n 4)(4)
2
_
d
4
xg
1/2
_

1
540
(20
2
+ 100 + 41)R

+
1
135
(5
2
+ 25 + 2)C
2
+
1
81
(5
2
+ 16 + 20)R
2
_
,
(4.69)
log det

div
= i
2
(n 4)(4)
2
_
d
4
xg
1/2
_

1
54
(4
2
+ 20 + 253)R

+
1
54
(4
2
+ 20 + 367)C
2
+
1
162
(200
2
+ 334 + 107)R
2
+
1
6
_
40 26
3

_
f
2
k
2
R
+
1
k
4
_
4
3
(14f
2
+
2
) +
1
2
(5f
4
+
4
)
_

13 + 10
12(1 +)
R

Chapter 4. Higherderivative quantum gravity 106


+
5 + 2
24(1 +)
RP
1
192
_
1 2
1 +
_
2
P
2

28
2
+ 80 + 61
96( + 1)
2
P

_
. (4.70)
Substituting the obtained expressions (4.68)(4.70) in (4.52) we get the
divergences of the standard oneloop eective action o mass shell

div
(1)
=
1
(n 4)(4)
2
_
d
4
xg
1/2
_

1
R

+
2
C
2
+
3
R
2
+
4
1
k
4
+
1
k
2
(R 4)
_
, (4.71)
where

1
=
196
45
,

2
=
133
20
, (4.72)

3
=
5
18
f
4

4
+
5
6
f
2

2
+
5
36
, (4.73)

4
=
1
2
(5f
4
+
4
) +
2
3

_
10
f
4

2
+ 15f
2

2
_
, (4.74)
=
5
3
f
4

2

13
6
f
2

1
2

2
. (4.75)
Therefrom it is immediately seen that the gauge xing tensor, P

, (4.59)
does not enter the result. In the next section we will calculate the divergences
of the eective action in arbitrary gauge and will show that the tensor P

does not contribute in the divergences in general case too. If one puts P

= 0
then the divergences of the operator H do not depend on the gauge xing
parameters at all.
Our result for divergences, (4.71)(4.75), does not coincide with the re-
sults of the papers [152155] in the coecient
3
, (4.73). Namely, the last
term in (4.73) is equal to 5/36 instead of the incorrect value 1/36 obtained
in [152155]. We will check our result, (4.73), by means of completely inde-
pendent computation on the De Sitter background in Sect. 4.5.
Chapter 4. Higherderivative quantum gravity 107
4.3 Oshell oneloop divergences of the stan-
dard eective action in arbitrary gauge
and the divergences of the unique eec-
tive action
Let us study now the dependence of the obtained result for the divergences of
the standard eective action, (4.71)(4.75), on the choice of the gauge. Let
us consider the variation of the oneloop eective action (4.1) with respect
to variation of the gauge condition (i.e., the functions

and H

(1)
=
1
i
_
_

1 ik

k
H

R
i

F
1
_

i
+
1
2
_

1 ik

k
H
1

_
H

_
. (4.76)
Using the Ward identities

1 ik

k
H

R
i

F
1
=
1 ik

j
R
j
,k
F
1
, (4.77)

1 ik

k
H
1

= H
1

F
1

j
R
j
,k

_
R
k

1 kn

m
R
m
,n
_
F
1
H
1

, (4.78)
that follow from the Noether identity, (4.6), one can derive from (4.76)

(1)
=
1
i
_

1 ik

j
R
j
,k
F
1

i
+
1
2
F
1

j
R
j
,k
_
R
k

1 kn

m
R
m
,n
_
F
1
(H
1

)
_
. (4.79)
Herefrom, it follows, in particular, that the oneloop eective action on mass
shell, = 0, (4.5), does not depend on the gauge,

onshell
= 0. (4.80)
Chapter 4. Higherderivative quantum gravity 108
Since the eective action on the mass shell is well dened, it is analytical
in background elds in the neighborhood of the mass shell (4.5). Therefore,
it can be expanded in powers of the extremal [45]. As the extremal has the
background dimension (in our case, (4.50), equal to four in mass units), this
expansion will be, in fact, an expansion in the background dimension. It is
obvious, that to calculate the divergences of the eective action it is sucient
to limit oneself to the terms of background dimension not greater than four.
Thus one can obtain the divergences by taking into account only linear terms
in the extremal. Moreover, from the dimensional grounds it follows that only
the trace of the extremal (4.50),
g

= g
1/2
_
1
k
2
(R 4)
1

2
R
_
, (4.81)
contributes to the divergences. Therefore, a dependence of the divergent
part of the eective action on the gauge parameters appears only in the
coecient, (4.75). The other, coecients, (4.72)(4.74), do not depend on
the gauge.
So, from (4.79) we obtain the variation of the oneloop eective action
with respect to the variation of the gauge

div
(1)
=
1
i
_

j
R
j
,k

1 ki
F
1

1
2

j
R
j
,k
R
k

F
1
F
1
(H
1

)
_

div
. (4.82)
Herefrom one can obtain the divergences of the eective action in any
gauge. To do this one has, rst, to x some form of the gauge condition
with arbitrary parameters, then, to calculate the divergences of the eective
action for some convenient choice of the gauge parameters and, nally, to
integrate the equation (4.82) over the gauge parameters.
Let us restrict ourselves to the covariant De Witt gauge, (4.56), with ar-
bitrary gauge parameters , and P

. Since the coecient at the variation


H
1
in (4.82) does not depend on the matrix H (i.e., on the gauge parame-
ters , and P

), one can integrate over the operator H immediately. Thus


we obtain the divergences of the eective action in arbitrary gauge

div
(1)
(, , , P) =
div
(1)
(
0
,
0
,
0
) +
1
i
_

_

0
d U
div
1
()
Chapter 4. Higherderivative quantum gravity 109

1
2
_
U
div
2
(, , , P) U
div
2
(,
0
,
0
, P)
_
_
, (4.83)
where
div
(1)
(
0
,
0
,
0
) is the divergent part of the eective action in the min-
imal gauge, (4.71)(4.75),
U
1
() =
j
R
j
,k

1 ki
(,
0
,
0
, P)F
1
()R
n

in
, (4.84)
E

in
=
d
d
E
in
=
1
4
g
1/2
g

(x, y),
U
2
(, , , P) =
j
R
j
,k
R
k

F
1
()F
1
()H
1

(, , P)

div
. (4.85)
To calculate the quantities U
1
, (4.84), and U
2
, (4.85), one has to nd
the ghost propagators F
1
() and H
1

(, , P) for arbitrary , , and


P and the gravitational propagator
1
(,
0
,
0
) for arbitrary parameter
and minimal values of other parameters,
0
and
0
. The whole background
dimension that causes the divergences is contained in the extremal
i
. So,
when calculating the divergences of the quantities U
1
, (4.84), and U
2
, (4.85),
one can take all propagators to be free, i.e., the background quantities (the
spacetime curvature, the commutator of covariant derivatives etc.) and the
mass terms can be neglected. This is so because together with any dimension-
ful terms there appear automatically a Green function
1
and the whole
term becomes nite. Therefore, in particular, the gauge xing tensor P

does not contribute in the divergences of the eective action at all. In the
minimal gauge, (4.61), we have shown this by explicit calculation in previous
section.
Using the explicit forms of the operators F(), (4.57a), H(, ), (4.58),
and (,
0
,
0
), (4.15), (4.53), we nd the free Green functions of these
operators
F
1
() =
1
2
_
g

+
1
3

_
2
g
1/2
(x, y), (4.86)
H
1

(, ) = 4
2
_
g

_
2
g
1/2
(x, y), (4.87)
Chapter 4. Higherderivative quantum gravity 110

1 ik
(,
0
,
0
) = 2f
2
_
E
1
(0),
2
+
2

+
1
(g

+ g

)
_
4
g
1/2
(x, y), (4.88)
where
E
1
(0),
E
1
,

=
0
= g
(
g
)

1 + 4
12
g

, (4.89)

1
=
1
3
_
+ 1 +
3
3
_
, (4.90)

2
=
4
3( + 1)
_
+ 1 +
3
3
_
2
. (4.91)
Let us note that in the minimal gauge, (4.61),
1
=
2
= 0 and, therefore,

1 ik
(0)
= 2f
2
E
1
(0),
2
g
1/2
(x, y).
Substituting the free propagators, (4.86)(4.88), in (4.84) and (4.85) we
obtain the divergences of the quantities U
1
and U
2
U
div
1
=
f
2
3
_
d
4
x

_
(
2
+ 6
1
2
1
0
)

3
g
1/2
(x, y)

div
y=x

1 +
3
_
g

2
g
1/2
(x, y)

div
y=x
_
, (4.92)
U
div
2
= 2
2
_
3 +
4
( 3)
2
(1 )
_
_
d
4
x

3
g
1/2
(x, y)

div
y=x
.
(4.93)
Using the divergences of the coincidence limits of the Green functions and
their derivatives in the dimensional regularization
2
g
1/2
(x, y)

div
y=x
= i
2
(n 4)(4)
2
,

3
g
1/2
(x, y)

div
y=x
= i
2
(n 4)(4)
2
1
4
g

, (4.94)
Chapter 4. Higherderivative quantum gravity 111
and the explicit form of the extremal

, (4.50), we obtain
U
div
1
= i
1
(n 4)(4)
2
6
2
( 3)
2
_
1 +
2
( + 1)( 3)
_
_
d
4
x g
1/2
1
k
2
(R 4),
U
div
2
= i
1
(n 4)(4)
2

2
_
3 +
4
( 3)
2
(1 )
_
_
d
4
x g
1/2
1
k
2
(R 4).
(4.95)
Substituting these expressions in (4.83) and integrating over we obtain
nally

div
(1)
(, , , P) =
div
(1)
(
0
,
0
,
0
)

1
(n 4)(4)
2
_
d
4
x g
1/2
(, , )
1
k
2
(R 4),
where
(, , ) =
13
6
f
2
+
4
3

3
2

2
2
( 3)
2
(1 )
+
6
2
( 2)
( 3)
2
. (4.96)
Thus the oshell divergences of the eective action in arbitrary gauge
have the same form, (4.71), where the coecients
1
,
2
,
3
and
4
do not
depend on the gauge and are given by the expressions (4.72)(4.74), and the
coecient reads
(, , ) =
5
3
f
4

2
+
5
6

3
2

2
2
( 3)
2
(1 )
+
6
2
( 2)
( 3)
2
. (4.97)
In particular, in the harmonic gauge, (4.60), ( = 1 and = 0), we have
(1, 0, ) =
5
3
f
4

2

2
3

2
. (4.98)
The dependence of the divergences on the parametrization of the quan-
tum eld also exhibits only in the coecient. Rather than to study this
dependence, let us calculate the divergences of the unique eective action

,
(4.24), that does not depend neither on the gauge nor on the parametrization
of the quantum eld.
Chapter 4. Higherderivative quantum gravity 112
From (4.24) we have

div
(1)
=
1
2i
_
log det

[
div
log det H[
div
2 log det F[
div
_
. (4.99)
The unique eective action

, (4.24), (4.99), diers from the usual one, (4.14),
(4.52), only by the operator

, (4.25). It is obtained from the operator ,
(4.15), by substituting the covariant functional derivatives instead of the
usual ones:
S
,ik

i

k
S = T
i
T
k
S T
j
ik

j
= S
,ik

j
ik

j
,

ik
=

loc
ik
+ T
j
ik

j
=
ik
+
j
ik

j
, (4.100)
where

loc
ik
= T
i
T
k
S +
i
H

k
. (4.101)
Since the nonmetric part of the connection T
j
ik
, (4.34), is nonlocal, the
operator

, (4.100), is an integrodierential one. The calculation of the
determinants of such operators oers a serious problem. However, as the
nonlocal part of the operator

, (4.100), is proportional to the extremal
i
,
it exhibits only o mass shell. Therefore, the calculation of the determinant
of the operator

, (4.100), can be based on the expansion in the nonlocal
part, T
j
ik

j
. To calculate the divergences it is again sucient to limit oneself
only to linear terms
log det

div
= log det

loc

div
+

1 mn
loc
T
i
mn

div
. (4.102)
To calculate this expression one can choose any gauge, because the answer
for the unique eective action does not depend on the gauge. Let us choose
the De Witt gauge, (4.56),

i
= R
k

E
ki
, F

= N

. (4.103)
Using the Ward identities for the Green function of the operator

loc
,
(4.101), in De Witt gauge we get (up to terms proportional to the extremal)
B

1 ik
loc
= N
1
H
1

N
1
R
k

+ O(),
B

1 ik
loc
B

k
= N
1
H
1

N
1
+ O(). (4.104)
Chapter 4. Higherderivative quantum gravity 113
Using the explicit form of T
i
mn
, (4.34), in (4.102) and the eqs. (4.104) we
obtain
log det

div
= log det

loc

div
U
div
3
,
where
U
3
=
j
T
k
R
j

R
k

N
1
H
1

N
1
. (4.105)
Finally, one has to x the operator H (i.e., the parameters and ) and to
determine the parameter of the metric of the conguration space, , (4.45),
(4.46).
In the paper [45] some conditions on the metric E
ik
were formulated,
that make it possible to x the parameter . First, the metric E
ik
must be
contained in the term with highest derivatives in the action S(). Second, the
operator N

, (4.28), (4.48), must be nondegenerate within the perturbation


theory. To nd the metric E
ik
, (4.45), i.e., the parameter , (4.46), one
should consider the second variation of the action on the physical quantum
elds, h

h, that satisfy the De Witt gauge conditions, R


i
h
i

= 0, and
identify the metric with the matrix E in the highest derivatives
h
i

(S
,ik
)h
k

=
1
2f
2
_
d
4
x h

g
1/2
E
,
()
2
h

+ terms with the curvature. (4.106)


This condition leads to a quadratic equation for that has two solutions

1
= 3

+ 1
,
2
= 3. (4.107)
As we already noted above, the value =
2
= 3 is unacceptable, since
the operator N, (4.48), in this case is degenerate on the at background.
Therefore, we nd nally
= 3

+ 1
. (4.108)
Let us note that this value of coincides with the minimal one, =
0
,
(4.61). Thus, if we choose the matrix H in the same form, (4.58), with the
minimal parameters, =
0
and =
0
, (4.61), then the operator

loc
becomes a minimal operator of the form (4.62)

loc
ik
=
ik
+
_
j
ik
_

j
, (4.109)
Chapter 4. Higherderivative quantum gravity 114
where
ik
is given by the expressions (4.62) and (4.63).
The divergences of the determinant of the operator

loc
, (4.109), can be
calculated either by direct application of the algorithm (4.65) or by means
of the expansion in the extremal
log det

loc

div
= log det
loc

div
+ U
div
4
, (4.110)
where
U
4
=
1 mn
_
i
mn
_

i
.
Using the formulas (4.99), (4.105), (4.110) and (4.52) we obtain the diver-
gences of the unique eective action

div
(1)
=
div
(1)
(
0
,
0
,
0
) +
1
2i
(U
div
3
U
div
4
), (4.111)
where
div
(1)
(
0
,
0
,
0
) are the divergences of the eective action in the minimal
gauge, (4.71)(4.75). The quantities U
div
3
and U
div
4
are calculated by using
the free propagators, (4.86)-(4.88), in the minimal gauge, (4.61),
U
div
3
= U
div
2
(
0
,
0
,
0
) 4f
2
_
d
4
x

_
,

_
g

+
1
3
(1 2)

_
4
g
1/2
(x, y)

div
y=x
,
U
div
4
=
_
d
4
x

_
,

_
2f
2
E
1
,
2
g
1/2
(x, y)

div
y=x
, (4.112)
where U
div
2
(
0
,
0
,
0
) is given by the formula (4.95) in the minimal gauge
(4.61). Using the divergences of the coincidence limits of the derivatives
of the Green functions, (4.94), and the Christoel connection, (4.47), and
substituting the minimal values of the parameters , and , (4.61), we
obtain
U
div
3
= i
1
(n 4)(4)
2
4
3
(2
2
+ f
2
)
_
d
4
x g
1/2
1
k
2
(R 4),
U
div
4
= i
1
(n 4)(4)
2
6(
2
f
2
)
_
d
4
x g
1/2
1
k
2
(R 4). (4.113)
Chapter 4. Higherderivative quantum gravity 115
Thus, the oshell divergences of the oneloop eective action,

(1)
, have
the standard form (4.71), where the coecients are determined by the
same expressions (4.72)(4.74) and the coecient, (4.75), has an extra
contribution due to the quantities U
div
3
and U
div
4
, (4.113), in (4.111). It has
the form
= (
0
,
0
,
0
) +
1
3
(11f
2
5
2
) =
5
3
f
4

2
+
3
2
f
2

13
6

2
, (4.114)
where (
0
,
0
,
0
) is given by (4.75).
4.4 Renormalization group equations and the
ultraviolet asymptotics of the coupling
constants
The structure of the divergences of the eective action, (4.71), indicates that
the higherderivative quantum gravity is renormalizable o mass shell. Thus
one can apply the renormalization group methods to study the highenergy
behavior of the eective (running) coupling constants [48, 49, 163, 109]. The
dimensionless constants , f
2
,
2
and are the essential coupling constants
[109] but the Einstein dimensionful constant k
2
is nonessential because its
variation reduces to a reparametrization of the quantum eld, i.e., up to total
derivatives
S
k
2

onshell
= 0. (4.115)
Using the oneloop divergences of the eective action, (4.71), we obtain
in the standard way the renormalization group equations for the coupling
constants of the renormalized eective action [48, 109]
d
dt
=
1
,
d
dt
f
2
= 2
2
f
4
, (4.116)
d
dt

2
= 6
3

4
=
5
3
f
4
+ 5f
2

2
+
5
6

4
, (4.117)
d
dt
=
1
2

4
=
1
4
(
4
+ 5f
4
) +
1
3

_
10
f
4

2
+ 15f
2

2
_
, (4.118)
Chapter 4. Higherderivative quantum gravity 116
d
dt
k
2
= k
2
, (4.119)
where t = (4)
2
log(/
0
), is the renormalization parameter and
0
is a
xed energy scale.
The ultraviolet behavior of the essential coupling constants (t), f
2
(t),

2
(t) and (t) as t is determined by the coecients (4.72)(4.74). They
play the role of generalized GellMannLow functions, (1.47), and do not
depend neither on the gauge condition nor on the parametrization of the
quantum eld. The nonessential coupling constant k
2
(t) is, in fact, simply
a eld renormalization constant. Thus the coecient, (4.75), (4.97) and
(4.114), play in (4.119) the role of the anomalous dimension (1.48). Corre-
spondingly, the ultraviolet behavior of the constant k
2
(t) depends essentially
both on the gauge and the parametrization of the quantum eld. It is obvious
that one can choose the gauge condition in such a way that the coecient ,
(4.97), is equal to zero, = 0. In this case the Einstein coupling constant is
not renormalized at all, i.e., k
2
(t) = k
2
(0) = const.
The solution of the eqs. (4.116) is trivial and reads
(t) = (0) +
1
t,
f
2
(t) =
f
2
(0)
1 + 2
2
f
2
(0)t
. (4.120)
Noting that
1
< 0 and
2
> 0, (4.72), we nd the following. First,
the topological coupling constant (t) becomes negative in the ultraviolet
limit (t ) and its absolute value grows logarithmically regardless of
the initial value (0). Second, the Weyl coupling constant f
2
(t) is either
asymptotically free (at f
2
> 0) or has a zerocharge singularity (at f
2
< 0).
We limit ourselves to the rst case, f
2
> 0, since, on the one hand, this
condition ensures the stability of the at background under the spin2 tensor
excitations, and, on the other hand, it leads to a positive contribution of the
Weyl term to the Euclidean action (4.49).
The solution of the equation (4.117) can be written in the form

2
(t) =
c
1
f
2p
(t) c
2
f
2p

f
2p
(t) f
2p

f
2
(t), (4.121)
Chapter 4. Higherderivative quantum gravity 117
where
c
1,2
=
1
50
(549

296401)
_

_
0.091
21.87
, (4.122)
p =

296401
399
1.36, f
2p



2
(0) c
1
f
2
(0)

2
(0) c
2
f
2
(0)
f
2p
(0). (4.123)
There are also two special solutions

2
1,2
(t) = c
1,2
f
2
(t), (4.124)
that correspond to the values f
2p

= 0, in (4.121). These solutions are


asymptotically free but only
2
2
(t) is stable in the ultraviolet limit.
The behavior of the conformal coupling constant
2
(t) depends essentially
on its initial value
2
(0). In the case
2
(0) > c
1
f
2
(0) we have f
2p
(0) >
f
2p

> 0 and, therefore, the function


2
(t), (4.121), has a typical zerocharge
singularity at a nite scale t = t

determined from f
2p
(t

) = f
2p

2
(t)

tt
= c
3
f
2(p+1)

f
2p
(t) f
2p

+ O(1), (4.125)
where
c
3
= c
1
c
2
=

296401
25
21.78. (4.126)
In the opposite case,
2
(0) < c
1
f
2
(0), the function
2
(t), (4.121), does
not have any singularities and is asymptotically free,

2
(t)

t
= c
2
f
2
(t) c
3
f
2p

f
2(p+1)
(t) + O
_
f
2(1+2p)
_
. (4.127)
Thus, to the contrary to the conclusions of the papers [152155], we
nd that in the region
2
> 0 there are no asymptotically free solutions.
The asymptotic freedom for the conformal coupling constant
2
(t) can be
achieved only in the negative region (0) < 0, (4.127).
The exact solution of the equation for the dimensionless cosmological
constant, (4.118), has the form
(t) = (t)
_
_
_

1
(0)(0) +
t
_
0
dA()
1
()
_
_
_
, (4.128)
Chapter 4. Higherderivative quantum gravity 118
where
A() =
1
4
_
5f
4
(t) +
4
(t)
_
,
(t) =

c
1
f
2p
(t) c
2
f
2p

f
2p
(t) f
2p

2/5

f
2
(t)

q
, (4.129)
q =
2
665
(241 +

296401) 0.913.
The ultraviolet behavior of the cosmological constant (t), (4.128), cru-
cially depends on the initial values of both the conformal coupling constant,

2
(0), and the cosmological constant itself, (0). In the region
2
(0) >
c
1
f
2
(0) the solution (4.128) has a zerocharge pole at the same scale t

,
similarly to the conformal coupling constant
2
(t), (4.125),
(t)

tt
=
3
14
c
3
f
2(1+2p)

f
2p
(t) f
2p

+ O(1). (4.130)
In the opposite case,
2
(0) < c
1
f
2
(0), the function (t), (4.128), grows in the
ultraviolet limit
(t)

t
= c
4
f
2q
(t) + O(f
2
). (4.131)
The sign of the constant c
4
in (4.131) depends on the initial value (0), i.e.,
c
4
> 0 for (0) >
2
(0) and c
4
< 0 for (0) >
2
(0), where

2
(0) = (0)

_
0
dA()
1
(). (4.132)
In the special case (0) =
2
(0) the constant c
4
is equal to zero (c
4
= 0) and
the solution (4.128) takes the form
(t) =
2
(t) = (t)

_
t
dA()
1
(). (4.133)
The special solution (4.133) is asymptotically free in the ultraviolet limit

2
(t)

t
= c
5
f
2
(t) + O
_
f
2(1+p)
_
, (4.134)
where
c
5
=
5
266

5 +c
2
2
q + 1
4.75. (4.135)
Chapter 4. Higherderivative quantum gravity 119
However, the special solution (4.133) is unstable because of the presence of
growing mode (4.131). Besides, it exists only in the negative region < 0.
In the positive region > 0 the cosmological constant is not asymptotically
free, (4.131).
Our conclusions about the asymptotic behavior of the cosmological con-
stant (t) also dier essentially from the results of the papers [152155] where
the asymptotic freedom for the cosmological constant in the region > 0 and

2
> 0 for any initial values of (0) was established.
Let us discuss the inuence of arbitrary lowspin matter (except for spin
3/2 elds) interacting with the quadratic gravity (4.49) on the ultraviolet
behavior of the theory. The system of renormalization group equations in
presence of matter involves the equations (4.116)(4.119) with the total
functions

i,tot
=
i
+
i,mat
, (4.136)
where
i,mat
is the contribution of matter elds in the gravitational diver-
gences of the eective action, (4.71), and the equations for the masses and
the matter coupling constants. The values of the rst three coecients at
the terms quadratic in the curvature have the form [5456]

1,mat
=
1
360
_
62N
(0)
1
+ 63N
1
+ 11N
1/2
+ N
0
_
,

2,mat
=
1
120
_
12N
(0)
1
+ 13N
1
+ 6N
1/2
+ N
0
_
, (4.137)

3,mat
=
1
72
_
N
1
+ (1 6)
2
N
0
_
,
where N
s
is the number of the elds with spin s, N
(0)
1
is the number of
massless vector elds, is the coupling constant of scalar elds with the
gravitational eld. In the formula (4.137) the spinor elds are taken to be
twocomponent. The coecients (4.137) possess important general proper-
ties

1,mat
< 0,
2,mat
> 0,
3,mat
> 0. (4.138)
The gravitational functions (4.72)(4.74) obtained in previous sections
have analogous properties for f
2
> 0 and
2
> 0. Therefore, the total
functions, (4.136), also satisfy the conditions (4.138) for f
2
> 0 and
2
> 0.
The properties (4.138) are most important for the study of the ultraviolet
Chapter 4. Higherderivative quantum gravity 120
asymptotics of the topological coupling constant (t), the Weyl coupling con-
stant f
2
(t) and the conformal one
2
(t).
The solution of the renormalization group equations for the topological
and Weyl coupling constants in the presence of matter have the same form
(4.120) with the substitution
tot
. Thus the presence of matter does not
change qualitatively the ultraviolet asymptotics of these constants: the cou-
pling (t) becomes negative and grows logarithmically and the Weyl coupling
constant is asymptotically free at f
2
> 0.
The renormalization group equation for the conformal coupling constant

2
(t) in the presence of the matter takes the form
d
dt

2
=
5
3
f
4
+ 5f
2

2
+
1
12
_
10 +N
1
+ (1 6)
2
N
0
_

4
. (4.139)
Therefrom one can show that at
2
> 0 the coupling constant
2
(t) has a
zerocharge singularity at a nite scale.
The other properties of the theory (in particular, the behavior of the
constant
2
(t) at
2
< 0) depend essentially on the particular form of the
matter model. However, the strong conformal coupling,
2
1, at
2
> 0
leads to singularities in the cosmological constant as well as in all coupling
constants of matter elds.
Thus, we conclude that the higherderivative quantum gravity interacting
with any lowspin matter necessarily goes out of the limits of weak conformal
coupling at high energies in the case
2
> 0. This conclusion is also opposite
to the results of the papers [152155] where the asymptotic freedom of the
higherderivative quantum gravity in the region
2
> 0 in the presence of
rather arbitrary matter was established.
Let us also nd the ultraviolet behavior of the nonessential Einstein
coupling constant k
2
(t). The solution of the equation (4.119) has the form
k
2
(t) = k
2
(0) exp
_
_
_
t
_
0
d()
_
_
_
. (4.140)
The explicit expression depends on the form of the function and, hence, on
the gauge condition and the parametrization of the quantum eld, (4.97). We
cite the result for two cases: for the standard eective action in the minimal
gauge and the standard parametrization (4.75) and for the unique eective
Chapter 4. Higherderivative quantum gravity 121
action (4.114). In both cases the solution (4.140) has the form
k
2
(t) = k
2
(0)
(t)
(0)
, (4.141)
where
(t) =

c
1
f
2p
(t) c
2
f
2p

f
2p
(t) f
2p

f
2
(t)

r
,
s =
_

_
13
5
3
5
, r =
_

_
3
665
(269 +

296401) 3.67
2
1995
(437 + 2

296401) 0.653
. (4.142)
Here and below the upper values correspond to the unique eective action and
the lower values correspond to the standard eective action in the minimal
gauge and the standard parametrization.
Therefrom it is immediately seen that the Einstein coupling constant k
2
(t)
grows in the ultraviolet limit (t )
k
2
(t)

t
= c
6
f
2r
(t) + O
_
f
2(pr)
_
. (4.143)
Let us note that the ultraviolet behavior of the dimensionful cosmological
constant, (t) = (t)/k
2
(t), is essentially dierent in the case of the unique
eective action and in the standard case
(t)

t
= c
7
f
2
+ O
_
f
2(+p)
_
, (4.144)
where
= r q
_

_
2.76
0.26
.
In the rst case (t) rapidly approaches zero t
2.76
and in the second case
it grows as t
0.26
.
It is well known that the functional formulation of the quantum eld
theory assumes the Euclidean action to be positive denite [49, 50, 64]. Oth-
erwise, (what happens, for example, in the conformal sector of the Einstein
gravity), one must resort to the complexication of the conguration space
to achieve the divergence of the functional integral [64, 115, 116].
Chapter 4. Higherderivative quantum gravity 122
The Euclidean action of the higherderivative theory of gravity diers
only by sign from the action (4.49) we are considering. It is positive denite
in the case
> 0, (4.145)

2
< 0, (4.146)
f
2
> 0, >
3
4

2
. (4.147)
It is not necessary to impose the condition (4.145) if one restricts oneself
to a xed topology. However, when including in the functional integral of
quantum gravity the topologically nontrivial metrics with large Euler char-
acteristic, the violation of the condition (4.145) leads to the exponential
growth of their weight and, therefore, to a foamlike structure of the space-
time at microscales [64]. It is this situation that occurs in the ultraviolet
limit, when (t) , (4.120).
The condition (4.146) is usually held to be nonphysical [140157]. The
point is, the conformal coupling constant
2
plays the role of the dimension-
less square of the mass of the conformal mode on the at background. In
the case
2
< 0 the conformal mode becomes tachyonic and leads to the
instability of the at space (i.e., oscillations of the static potential, unstable
solutions etc. [144]).
As we showed above, the higherderivative quantum gravity in the region

2
> 0 has unsatisfactory zerocharge behavior in the conformal sector,
(4.125), (4.130). In the region of strong conformal coupling (
2
1) one
cannot make anything denitive conclusions on the basis of the perturba-
tive calculations. However, on the qualitative level it seems likely that the
singularity in the coupling constants
2
(t) and (t) can be interpreted as a
reconstruction of the ground state of the theory, i.e., the conformal mode
freezes and a condensate is formed.
We nd the arguments against the nonphysical condition (4.146) to
be not strong enough. First, the higherderivative quantum gravity, strictly
speaking, cannot be treated as a physical theory within the limits of per-
turbation theory because of the presence of the ghost states in the tensor
sector that violate the unitarity of the theory [143155, 157]. This is not
surprising in an asymptotically free theory (that always takes place in the
tensor sector), since, generally speaking, the true physical asymptotic states
have nothing to do with the excitations in the perturbation theory [158].
Chapter 4. Higherderivative quantum gravity 123
Second, the correspondence with the macroscopic gravitation is a rather ne
problem that needs a special investigation of the lowenergy limit of the high
derivative quantum gravity. Third, the cosmological constant is always not
asymptotically free. This means, presumably, that the expansion around the
at space in the high energy limit is not valid anymore. Hence, the solution
of the unitarity problem based on this expansion by summing the radiation
corrections and analyzing the position of the poles of the propagator in mo-
mentum representation is not valid too. In this case the at background
cannot present the ground state of the theory any longer.
From this standpoint, in high energy region the higherderivative quan-
tum gravity with positive denite Euclidean action, i.e., with an extra con-
dition

2
< c
1
f
2
0.091f
2
, (4.148)
seems to be more intriguing. Such theory has an unique stable ground state
that minimizes the functional of the classical Euclidean action. It is asymp-
totically free both in the tensor and conformal sectors. Besides, instead of
the contradictory zerocharge behavior the cosmological constant just grows
logarithmically at high energies.
Let us stress once more the main conclusion of the present section. Not-
withstanding the fact that the higherderivative quantum gravity is asymp-
totically free in the tensor sector of the theory with the natural condition
f
2
> 0, that ensures the stability of the at space under the tensor per-
turbations, the condition of the conformal stability of the at background,

2
> 0, is incompatible with the asymptotic freedom in the conformal sector.
Thus, the at background cannot present the ground state of the theory in
the ultraviolet limit. The problem with the conformal mode does not appear
in the conformally invariant models [97, 98]. Therefore, they are asymptoti-
cally free [152156]; however, the appearance of the R
2
divergences at higher
loops leads to their nonrenormalizability [161].
4.5 Eective potential of higherderivative quan-
tum gravity
Up to now the background eld (i.e., the spacetime metric) was held to be
arbitrary. To construct the Smatrix one needs the background elds to be
Chapter 4. Higherderivative quantum gravity 124
the solutions of the classical equations of motion, (4.5), i.e., to lie on mass
shell. It is obvious that the at space does not solve the equations of motion
(4.5) and (4.50) for nonvanishing cosmological constant. The most simple
and maximal symmetric solution of the equations of motion (4.5) and (4.50)
is the De Sitter space
R

=
1
12
(

)R, (4.149)
R

=
1
4
g

R, R = const,
with the condition
R = 4. (4.150)
On the other hand, in quantum gravity De Sitter background, (4.149),
plays the role of covariantly constant eld strength in gauge theories,

= 0. Therefore, the eective action on De Sitter background determines, in


fact, the eective potential of the higherderivative quantum gravity. Since
in this case the background eld is characterized only by one constant R,
the eective potential is an usual function of one variable. The simplicity
of De Sitter background makes it possible to calculate the oneloop eective
potential exactly.
In the particular case of De Sitter background one can also check our
result for the R
2
divergence of the oneloop eective action in general case,
(4.71), i.e., the coecient
3
, (4.73), that diers from the results of [152155]
and radically changes the ultraviolet behavior of the theory in the conformal
sector (see Sect. 4.4).
For the practical calculation of the eective potential we go to the Eu-
clidean sector of the spacetime. Let the spacetime be a compact four
dimensional sphere S
4
with the volume
V =
_
d
4
xg
1/2
= 24
_
4
R
_
2
, (R > 0), (4.151)
and the Euler characteristic
=
1
32
2
_
d
4
xg
1/2
R

= 2. (4.152)
In present section we will always use the Euclidean action that diers only by
sign from the pseudoEuclidean one, (4.49). All the formulas of the previous
Chapter 4. Higherderivative quantum gravity 125
sections remain valid by changing the sign of the action S, the extremal

i
= S
,i
and the eective action .
On De Sitter background (4.149) the classical Euclidean action takes the
form
S(R) = 24(4)
2
_
1
6
_

1

2
_

1
x
+ 2
1
x
2
_
, (4.153)
where x Rk
2
and = k
2
. For > 0 it has a minimum on the mass shell,
R = 4, (x = 4), (4.150), that reads
S
onshell
= (4)
2
_
4
_

1

2
_

_
. (4.154)
Our aim is to obtain the eective value of the De Sitter curvature R from
the full eective equations
(R)
R
= 0. (4.155)
Several problems appear on this way: the dependence of the eective ac-
tion and, therefore, the eective equations on the gauge condition and the
parametrization of the quantum eld, validity of the oneloop approximation
etc. [61].
First of all, we make a change of eld variables h

=

h

+
1
4
g

+ 2
(

)
, (4.156)
= h , h = g

, (4.157)

+
1
2

, (4.158)
where the new variables,

h

and

, satisfy the dierential constraints

= 0,

h

= 0, (4.159)

= 0. (4.160)
In the following we will call the initial eld variables h

, without any restric-


tions imposed on, the unconstrained elds and the elds

h

and

, which
satisfy the dierential conditions (4.159) and (4.160), the constrained ones.
Chapter 4. Higherderivative quantum gravity 126
When the unconstrained eld is transformed under the gauge transfor-
mations with parameters

, (4.43), (4.44),
h

= 2
(

)
,

, (4.161)
the constrained elds transform in the following way

= 0, = 0, (4.162)

, = 2. (4.163)
Therefore, the transverse traceless tensor eld

h

and the conformal eld


are the physical gaugeinvariant components of the eld h

, whereas

and are pure gauge nonphysical ones.


Let us write down the second variation of the Euclidean action on De
Sitter background
S
2
(g + h)
1
2
h
i
S
,ik
h
k
=
_
d
4
x g
1/2
_
1
4f
2

2
_
m
2
2
+
f
2
+
2
3
2
R
_

2
_
R
6
_
+
1
2
m
2
2
(R 4)
_

3
32
2
_

0
(m
2
0
)
0
_

R
3
_
+
1
3
m
2
0
(R 4)
_

2
3
m
2
0
(R 4)
0
(0)
2
3
m
2
0
(R 4)
0
(0)
0
_

R
2
_

_
+
1
4k
2
(R 4)

1
_

R
4
_

_
, (4.164)
where
m
2
2
=
f
2
k
2
, m
2
0
=

2
k
2
,
and

s
(X) = +X (4.165)
are the constrained dierential operators acting on the constrained elds of
spin s = 0, 1, 2.
Chapter 4. Higherderivative quantum gravity 127
When going to the mass shell, (4.150), the dependence of S
2
, (4.164), on
the nonphysical elds

and disappears and (4.164) takes the form


S
2
(g + h)

onshell
=
_
d
4
xg
1/2
_
1
4f
2

2
_
m
2
2
+
4
3
(f
2
+
2
)

2

_

2
_
2
3

3
32
2

0
(m
2
0
)
0
_

4
3

_
. (4.166)
Herefrom it follows, in particular, the gauge invariance of the second varia-
tion, (4.164), on the mass shell, (4.150),
S
2
(g + h)

onshell
= 0. (4.167)
The eigenvalues
n
of the constrained dAlambert operator
s
(0) =
and their multiplicities d
n
are [61]

n
=
2

n
,

n
= n
2
+ 3n s,
2
=
R
12
,
d
n
=
1
6
(2s + 1)(n + 1 s)(n + 2 +s)(2n +3), n = s, s +1, . . . . (4.168)
Thus the condition of stability of the De Sitter background (4.149),
S
2
(g + h)

onshell
> 0, (4.169)
imposes the following restrictions (for > 0): in the tensor sector
f
2
> 0,
1
f
2

1
3
2
<
1
4
, (4.170)
and in the conformal sector

2
< 0,
1
4
<
1
3
2
. (4.171)
However, even in the case when these conditions are fullled there are still left
ve zero modes of the operator
0
_

4
3

_
in the conformal sector, (4.166).
Along these conformal directions,
1
, in the conguration space the Euclidean
action does not grow
S
2
(g +
1
)

onshell
= 0. (4.172)
Chapter 4. Higherderivative quantum gravity 128
This means that, in fact, the De Sitter background (4.149) does not give
the absolute minimum of the positive denite Euclidean action. This can be
veried by calculating the next terms in the expansion of S(g+
1
). However,
in the oneloop approximation these terms do not matter.
To calculate the eective action one has to nd the Jacobian of the change
of variables (4.156)(4.158). Using the simple equations
_
d
4
xg
1/2
h
2

=
_
d
4
xg
1/2
_

h
2
+ 2

1
_

R
4
_

+
3
4

0
(0)
0
_

R
3
_
+
1
4
h
2
_
, (4.173)
_
d
4
xg
1/2

=
_
d
4
xg
1/2
_

2
+
1
4

0
(0)
_
, (4.173)
we obtain
dh

= d

d d (det J
2
)
1/2
,
d

= d

d (det J
1
)
1/2
, (4.174)
where
J
2
=
1
_

R
4
_

0
_

R
3
_

0
(0),
J
1
=
0
(0).
Let us calculate the eective action (4.12) in De Witt gauge (4.57). The
ghost operator F, (4.16), in this gauge equals the operator N, (4.48). On De
Sitter background (4.149) it has the form
F

= N

= 2g
1/2
_
g

_

R
4
_
+
1
2
( 1)

_
(x, y). (4.175)
The operator of averaging over the gauges H, (4.58), and the gauge xing
term have the form
H

=
1
4
2
g
1/2
_
g

_
+
R + P
4
_
+

_
(x, y), (4.176)
Chapter 4. Higherderivative quantum gravity 129
S
gauge

1
2

=
_
d
4
xg
1/2
1
2
2
_

1
_
R + P
4
_

2
1
_

R
4
_

+
1
16
(1 )
_

0
(P

)
0
(0)
2( 3)
0
(P

)
0
_
R
3
_

0
(0)
+( 3)
2

0
(P

)
2
0
_
R
3
_

0
(0)
__
, (4.177)
where
P = g

= (p
1
+ 4p
2
)R + 4p
3
1
k
2
, P


P
1
.
Using the equation
_
d
4
xg
1/2

( +X) +

=
_
d
4
xg
1/2
_

1
(X)

+
1
4

0
_
X
R
4
1
_

0
(0)
_
, (4.178)
we nd the determinants of the ghost operators (4.175) and (4.176)
det F = det
1
_

R
4
_
det
0
_
R
3
_
,
det H = det
1
_
R + P
4
_
det
0
(P

) . (4.179)
For the operators F and H to be positive denite we assume < 1 and
< 3.
Thus using the determinants of the ghost operators, (4.179), and the
Jacobian of the change of variables, (4.174), and integrating exp(S
2
S
gauge
)
we obtain the oneloop eective action o mass shell in De Witt gauge (4.57)
with arbitrary gauge parameters , , and P

(1)
= I
(2)
+ I
(1)
+ I
(0)
, (4.180)
Chapter 4. Higherderivative quantum gravity 130
where
I
(2)
=
1
2
log det
_

2
_
R
6
_

2
_
m
2
2
+
f
2
+
2
3
2
R
_
+
1
2
m
2
2
(R 4)
_
, (4.181)
I
(1)
=
1
2
log
det
_

1
_

R
4
_

1
_
R+P
4
_
+
1
2

2
k
2
(R 4)
_
_
det
1
_

R
4
__
2
det
1
_
R+P
4
_ , (4.182)
I
(0)
=
1
2
log
det
_

2
0
_
R
3
_

0
(P

)
0
(m
2
0
) +T
_

R
4
_
+(
_

R
4
_
2
_
_
det
0
_
R
3
__
2
det
0
(P

)
,
(4.183)
T = 4m
2
0

2
3
( 3)
2

0
_
R

2
3
_

0
(P

8
2
k
2
(1 )( 3)
2

0
_
m
2
0
_

0
_

R
2
_
, (4.184)
( = 16

2

2
k
4
(1 )( 3)
2
.
The quantity I
(2)
describes the contribution of two tensor elds, I
(1)
gives
the contribution of the vector ghost and I
(0)
is the contribution of the scalar
conformal eld o mass shell. The contribution of the tensor elds I
(2)
,
(4.181), does not depend on the gauge, the contribution of the vector ghost
I
(1)
, (4.182), depends on the parameters and P and the contribution of the
scalar eld I
(0)
, (4.183), (4.184), depends on all gauge parameters , ,
and P.
The expressions for I
(1)
and I
(0)
, (4.182), (4.183), are simplied in some
particular gauges
I
(1)

=0
=
1
2
log det
1
_

R
4
_
, (4.185)
I
(0)

=0
=0
= I
(0)

=
=0
=
1
2
log
det
_

0
_

R
3
_

0
(m
2
0
) +
1
3
m
2
0
(R 4)
_
det
0
_

R
3
_
,
(4.186)
Chapter 4. Higherderivative quantum gravity 131
I
(0)

=0
=1
= I
(0)

=
=1
=
1
2
log
det
_

0
_

R
2
_

0
(m
2
0
) +
1
2
m
2
0
(R 4)
_
det
0
_

R
2
_
,
(4.187)
I
(0)

=
=
1
2
log
det [
0
(0)
0
(m
2
0
) m
2
0
(R 4)]
det
0
(0)
, (4.188)
Let us also calculate the gaugeindependent and reparametrizationinva-
riant unique eective action, (4.21), on De Sitter background in the orthog-
onal gauge, (4.36), (4.38). In the oneloop approximation, (4.40), it diers
from the standard eective action in De Witt gauge, (4.14), (4.56), (4.58),
with = 0 only by an extra term in the operator

, (4.41), due to the
Christoel connection of the conguration space, (4.27), (4.47). Therefore,
the unique eective action, (4.38), (4.40), can be obtained from the standard
one, (4.12), (4.14), in De Witt gauge, (4.56), (4.58), for = 0 by substituting
the operator T
i
T
k
S, (4.41), for the operator S
,ik
, i.e., by the replacing the
quadratic part of the action S
2
(g + h), (4.164), by

S
2
(g + h) = S
2
(g + h)
1
2
h
i
_
j
ik
_

j
h
k
,
h
i
_
j
ik
_

j
h
k
=
1
4
(
1
1)
1
k
2
(R 4)
_
d
4
xg
1/2
_
h
2


1
4
h
4
_
=
1
4
(
1
1)
1
k
2
(R 4)
_
d
4
xg
1/2
_

h
2
+2

1
_

R
4
_

+
3
4

0
(0)
0
_

R
3
_

_
, (4.189)
where is the parameter of the conguration space metric that is given,
according to the paper [45], by the formula (4.108). Let us note that in
the Einstein gravity one obtains for the parameter the value = 1 [45,
46]. Therefore, the additional contribution of the connection (4.189) vanishes
and the unique eective action on De Sitter background coincides with the
standard one computed in De Witt gauge, (4.56), (4.58), with = 0 and
= 1 (i.e., in the harmonic De DonderFockLandau gauge (4.60)).
Taking into account the Jacobian of the change of variables (4.174) and
the ghosts determinant (4.179) the functional measure in the constrained
Chapter 4. Higherderivative quantum gravity 132
variables takes the form
dh

(R
i
h
i
) det N = d

dd (

_
( 3)
0
_
R
3
_

_
det
0
_
R
3
_

_
det
1
_

R
4
_
det
0
_

R
3
__
1/2
. (4.190)
Integrating exp(

S
2
) we obtain the oneloop unique eective action

(1)
=

I
(2)
+

I
(1)
+

I
(0)
, (4.191)
where

I
(2)
=
1
2
log det
_

2
_
R
6
_

2
_
m
2
2
+
f
2
+
2
3
2
R
_
+
1
2
m
2
2
(R 4)
_
,
(4.192)

I
(1)
=
1
2
log det
1
_

R
4
_
, (4.193)

I
(0)
=
1
2
log
det
_

0
_
R
3
_

0
(m
2
0
)
1
3
m
2
0
(R 4)
_
det
0
_
R
3
_
. (4.194)
The eqs. (4.191)(4.194) for = 1 do coincide indeed with the standard
eective action in the gauge = 0, = 1, (4.180), (4.181), (4.185), (4.187).
However, to obtain the unique eective action in our case one has to put in
(4.191)(4.194) = 3f
2
/(f
2
+ 2
2
), (4.108).
On the mass shell (4.150) the dependence on the gauge disappears and
we have
I
onshell
(2)
=
1
2
log det
2
_
2
3

_
+
1
2
log det
2
_
m
2
2
+
4
3

(f
2
+
2
)

2

_
,
(4.195)
I
onshell
(1)
=
1
2
log det
1
() , (4.196)
I
onshell
(0)
=
1
2
log det
0
(m
2
0
). (4.197)
Chapter 4. Higherderivative quantum gravity 133
Herefrom one sees immediately the spectrum of the physical excitations
of the theory: one massive tensor eld of spin 2 (5 degrees of freedom), one
massive scalar eld (1 degree of freedom) and the Einstein graviton, i.e., the
massless tensor eld of spin 2 (5 3 = 2 degrees of freedom). Altogether
the higher-derivative quantum gravity (4.49) has 5 + 2 + 1 = 8 degrees of
freedom.
To calculate the functional determinants of the dierential operators we
will use the technique of the generalized function [55, 61, 64, 79, 80, 115,
116]. Let us dene the function by the functional trace of the complex
power of the dierential operator of order 2k,
(k)
,

s
_
p;
(k)
/
2k
_
tr
_

(k)
/
2k
_
p
, (4.198)
where

(k)
= P
k
( ), (4.199)
P
k
(x) is a polynomial of order k, is a dimensionful mass parameter and s
denotes the spin of the eld, the operator
(k)
is acting on.
For Re p > 2/k the function is determined by the convergent series
over the eigenvalues (4.168)

s
_
p;
(k)
/
2k
_
=

n
d
n
_
P
k
(
n
)/
2k
_
p
, (4.200)
where the summation runs over all modes of the dAlambert operator, (4.168),
with positive multiplicities, d
n
> 0, including the negative and zero modes
of the operator
(k)
. The zero modes give an innite constant that should
be simply subtracted, whereas the negative modes lead to an imaginary part
indicating to the instability [61]. For Re p 2/k the analytical continuation
of (4.200) denes a meromorphic function with poles on the real axis. It is
important, that the function is analytic at the point p = 0. Therefore, one
can dene the nite values of the total number of modes of the operator
(k)
(taking each mode k times) and its functional determinant
k tr 1 = B(
(k)
), (4.201)
log det
_

(k)
/
2k
_
=

s
(0), (4.202)
where
B(
(k)
) = k
s
(0),
Chapter 4. Higherderivative quantum gravity 134

s
(p) =
d
dp

s
(p). (4.203)
Under the change of the scale parameter the functional determinant
behaves as follows

s
_
0;
(k)
/
2k
_
= B(
(k)
) log

2

2
+

s
_
0;
(k)
/
2k
_
. (4.204)
Using the spectrum of the dAlambert operator (4.168) we rewrite (4.200)
in the form

s
_
p;
(k)
/
2k
_
=
2s + 1
3

s+
3
2
, =1
(
2
j
2
)
_
P
k
_

2
_

9
4
s
___
p
,
(4.205)
where

2
=
R
12
, j = s +
1
2
.
The sum (4.205) can be calculated for Re p >
2
k
by means of the Abel
Plan summation formula [168]

1
2
f() =

1
2
k
1
2
f() +

_
k+
dt f(t)
+

_
0
dt
e
2t
+ 1
[if(k + it) if(k + + it)] , (4.206)
where the integer k should be chosen in such a way that for Re > k the
function f() is analytic, i.e., it does not have any poles. The innitesimal
parameter > 0 shows the way how to get around the poles (if any) at
Re = k. The formula (4.206) is valid for the functions f() that fall o
suciently rapidly at the innity:
f()

||
[[
q
, Re q > 1. (4.207)
When applying the formula (4.206) to (4.205) the second integral in (4.206)
gives an analytic function of the variable p. All the poles of the function
are contained in the rst integral. By using the analytical continuation and
integrating by parts one can calculate both (0) and

(0).
Chapter 4. Higherderivative quantum gravity 135
As a result we obtain for the operator of second order,

s
(X) = +X, (4.208)
and for the operator of forth order,

(2)
s
(X, Y ) =
2
2X +Y, (4.209)
the nite values of the total number of modes (4.201) and the determinant
(4.202)
B(
s
(X)) =
2s + 1
12
_
(b
2
+ j
2
)
2

2
3
j
2
+
1
30
_
, (4.210)
B
_

(2)
s
(X, Y )
_
= 2
2s + 1
12
_
(b
2
+ j
2
)
2

2
3
j
2
+
1
30
a
2
_
, (4.211)

s
(0;
s
(X)/
2
) =
2s + 1
3
F
(0)
s
(

X), (4.212)

s
(0;
(2)
s
(X, Y )/
4
) =
2s + 1
3
F
(2)
s
(

X,

Y ), (4.213)
where
b
2
=

X s
9
4
,

X =
X

2
,
a
2
=

Y

X
2
,

Y =
Y

4
,
and
F
(0)
s
(

X) =
1
4
b
2
(b
2
+ 2j
2
) log b
2
+
1
2
j
2
b
2
+
3
8
b
4
+2

_
0
dt t
e
2t
+ 1
(t
2
+ j
2
) log [b
2
t
2
[
+

1
2
s
1
2
(
2
j
2
) log(
2
+ b
2
), (4.214)
F
(2)
s
(

X,

Y ) =
1
4
(a
2
b
4
2j
2
b
2
) log(b
4
+ a
2
)
Chapter 4. Higherderivative quantum gravity 136
a(b
2
+ j
2
)
_
arctan
_
b
2
a
_


2
_

1
4
a
2
+
3
4
b
4
+ j
2
b
2
+2

_
0
dt t
e
2t
+ 1
(t
2
+ j
2
) log
_
(b
2
t
2
)
2
+ a
2
_
+

1
2
s
1
2
(
2
j
2
) log
_
(
2
+ b
2
)
2
+ a
2
_
. (4.215)
The introduced functions, (4.214) and (4.215), are related by the equation
F
(0)
s
(

X) + F
(0)
s
(

Y ) = F
(2)
s
_

X +

Y
2
;

X

Y
_
. (4.216)
In complete analogy one can obtain the functional determinants of the
operators of higher orders and even nonlocal, i.e., integrodierential, op-
erators.
Using the technique of the generalized function and separating the de-
pendence on the renormalization parameter we get the oneloop eective
action, (4.180)(4.184),

(1)
=
1
2
B
tot
log
R
12
2
+
(1) ren
, (4.217)
where

(1) ren
=
(1)

2
=
2
=
R
12
. (4.218)
For the study of the eective action one should calculate, rst of all,
the coecient B
tot
. To calculate the contributions of the tensor, (4.181),
and vector, (4.182), elds in B
tot
it suces to use the formulas (4.210) and
(4.211) for the operators of the second and the forth order. Although the
contribution of the scalar eld in arbitrary gauge, (4.183), (4.184), contains
an operator of eighth order, it is not needed to calculate the coecient B
for the operator of the eighth order. Noting that on mass shell, (4.150), the
contribution of the scalar eld (4.197) contains only a second order operator,
one can expand the contribution of the scalar eld o mass shell in the
extremal, i.e., in (R 4), limiting oneself only to linear terms.
Chapter 4. Higherderivative quantum gravity 137
One should note, that the dierential change of the variables (4.156)
(4.158) brings some new zero modes that were not present in the non
constrained operators. Therefore, when calculating the total number of
modes (i.e., the coecient B
tot
) one should subtract the number of zero
modes of the Jacobian of the change of variables (4.174):
B
tot
=

i
B(
i
) ^(J). (4.219)
Using the number of zero modes of the operators entering the Jacobians,
(4.174),
^(
0
(0)) = 1, ^
_

0
_

R
3
__
= 5, ^
_

1
_

R
4
__
= 10,
(4.220)
we obtain
^(J) = 2 15 1 = 29. (4.221)
Thus we obtain the coecient B
tot
B
tot
=
20
3
f
4

4
+ 20
f
2

2

634
45
+ 24
1
x
+16
_
3
4
(
4
+ 5f
4
) +
_
10
f
4

2
+ 15f
2

2
6
__
1
x
2
,(4.222)
where x = Rk
2
and the coecient is given by the formula (4.97).
The unique eective action (4.191) has the same form, (4.217), with the
coecient

B
tot
of the form (4.222) but with the change , where is
given by the formula (4.114).
On the other hand, the coecient B
tot
can be obtained from the gen-
eral expression for the divergences of the eective action (4.71) on De Sitter
background (4.149),
B
tot
= 4
1
+ 24
3
+ 24
1
x
+ 16
_
3
2

4
6
_
1
x
2
, (4.223)
where the coecients
1
,
3
and
4
are given by the formulas (4.72)(4.74),
and the coecient is given by the formula (4.97) for standard eective ac-
tion in arbitrary gauge and by the expression (4.114) for the unique eective
action.
Chapter 4. Higherderivative quantum gravity 138
Comparing the expressions (4.222) and (4.223) we convince ourselves that
our result for the coecient
3
, (4.73), that diers from the results of other
authors [152155] (see Sect. 4.2), and our results for the divergences of the
eective action in arbitrary gauge, (4.97), and for the divergences of the
unique eective action, (4.114), are correct.
On mass shell (4.150) the coecient (4.222) does not depend on the gauge
and we have a singlevalued expression
B
onshell
tot
=
20
3
f
4

4
+ 20
f
2

2

634
45
+
_
10
f
4

2
+ 15f
2

2
_
1

+
3
4
(
4
+ 5f
4
)
1

2
. (4.224)
Let us also calculate the nite part of the eective action (4.218). Since
it depends essentially on the gauge, (4.180)(4.184), we limit ourselves to the
case of the unique eective action, (4.191)(4.194). Using the results (4.202)
and (4.212)(4.215) we obtain

(1)ren
=
1
6
_
5F
(2)
2
(Z
1
, Z
2
) 3F
(0)
1
(3)
+F
(2)
0
(Z
3
, Z
4
) F
(0)
0
_
2
f
2

2
4
__
, (4.225)
where
Z
1
= 6f
2
1
x
+ 2
f
2

2
+ 3,
Z
2
= 96(f
2
+ 2
2
)
1
x
2
+ 48(f
2
+
2
)
1
x
+ 8
f
2

2
+ 8, (4.226)
Z
3
= 6
2
1
x

f
2

2
2,
Z
4
= 96(f
2
+ 2
2
)
1
x
2
,
F
(k)
s
are the functions introduced above, (4.214), (4.215), and x = Rk
2
.
Chapter 4. Higherderivative quantum gravity 139
On mass shell (4.150) the eective action does not depend on the gauge
and has the form, (4.13), (4.154), (4.195)-(4.197), (4.217),

onshell
= (4)
2
_
4
_

1

2
_

_
+ h
_
1
2
B
onshell
tot
log

3
2
k
2
+
onshell
(1)ren
_
+ O( h
2
), (4.227)
where

onshell
(1)ren
=
1
6
_
5F
(0)
2
_
3
f
2

+ 4
f
2

2
+ 4
_
+5F
(0)
2
(2) 3F
(0)
1
(3) +F
(0)
0
_
3

__
. (4.228)
The expression (4.227) gives the vacuum action on De Sitter background
with quantum corrections. It is real, since the operators in (4.195)(4.197) do
not have any negative modes provided the conditions (4.170) and (4.171) are
fullled. Although the operator
0
(m
2
0
) has one negative mode, = const,
subject to the condition (4.171), it is nonphysical, since it is just the zero
mode of the Jacobian of the change of variables (4.174). All other modes of
the operator
0
(m
2
0
) are positive subject to the condition (4.171) in spite of
the fact that m
2
0
< 0. Depending on the value of De Sitter curvature R o
mass shell there can appear negative modes leading to an imaginary part of
the eective action.
Dierentiating the eective action, (4.13), (4.153), (4.217), (4.222), (4.225),
we obtain the eective equation for the background eld, i.e., the curvature
of De Sitter space,
1
hk
2

R
=
1
h
24(4)
2

x 4
x
3
+
1
2x
B
tot
(x)
+
1
2
B

tot
(x) log
x
12
2
k
2
+

(1)ren
(x) + O( h) = 0, (4.229)
where
B

tot
(x) =
B
tot
x
=
32
x
3
_
3
4
(
4
+ 5f
4
)
Chapter 4. Higherderivative quantum gravity 140
+
_
10
f
4

2
+ 15f
2

2
6
__
24
1
x
2
,(4.230)

(1)ren
(x) =

(1)ren
x
, x = Rk
2
. (4.231)
The perturbative solution of the eective equation (4.229) has the form
Rk
2
= 4 h

2
3(4)
2
_
B
onshell
tot
+ 4B

tot
(4) log

3
2
k
2
+8

(1)ren
(4)
_
+O( h
2
). (4.232)
It gives the corrected value of the curvature of De Sitter space with regard
to the quantum eects.
Perturbation theory near this solution is applicable for ,= 0 in the region
f
2

2
1. For 1, i.e., when is of Planck mass order 1/k
2
, the
contributions of higher loops are essential and the perturbation theory is not
adequate anymore.
Apart from the perturbative solution (4.232) the equation (4.229) can
also have non-perturbative ones. In the special case = 0 non-perturbative
solution R ,= 0 means the spontaneous creation of De Sitter space from the
at space due to quantum gravitational uctuations. Therefore, it seems
quite possible that De Sitter space, needed in the inational cosmological
scenarios of the evolution of the Universe [170], has quantumgravitational
origin [169]. However, almost any nonperturbative solution has the order
Rk
2
1 and, therefore, is inapplicable in the oneloop approximation.
Chapter 5
Conclusion
The following main results are obtained in the present dissertation.
1. The methods for the covariant expansions of arbitrary elds in a curved
space with arbitrary connection in generalized Taylor series and the
Fourier integral in most general form are formulated.
2. A manifestly covariant technique for the calculation of De Witt coe-
cients on the basis of the method of covariant expansions is elaborated.
The corresponding diagrammatic formulation of this technique is given.
3. The De Witt coecients a
3
and a
4
at coinciding points are calculated.
4. The renormalized oneloop eective action for the massive scalar, spinor
and vector elds in an external gravitational eld up to the terms of
order 1/m
4
is calculated.
5. Covariant methods for studying the nonlocal structure of the eective
action are developed.
6. The terms of rst order in background elds in De Witt coecients are
calculated. The summation of these terms is carried out and a non
local covariant expression for the Green function at coinciding points
up to terms of second order in background elds is obtained. It is shown
that in the conformally invariant case the Green function at coinciding
points is nite in the rst order in the background elds.
141
Chapter 5. Conclusion 142
7. The terms of second order in background elds in De Witt coecients
are calculated. The summation of these terms is carried out and a man-
ifestly covariant nonlocal expression for the oneloop eective action
up to the terms of third order in background elds is obtained. All
formfactors, their asymptotics and imaginary parts (for standard def-
inition of the asymptotic regions, ground states and causal boundary
conditions) are calculated. A nite eective action in the conformally
invariant case of massless scalar eld in twodimensional space is ob-
tained.
8. The covariantly constant terms in De Witt coecients for the case of
scalar eld are calculated. It is shown that the corresponding Schwinger
De Witt series diverges. The Borel summation of the covariantly con-
stant terms is carried out and an explicit nonanalytic expression in
the background elds for the oneloop eective action up to the terms
with covariant derivatives of the background elds is obtained.
9. The oshell oneloop divergences of the eective action in arbitrary
covariant gauge as well as those of the unique eective action in higher
derivative quantum gravity are calculated.
10. The ultraviolet asymptotics of the coupling constants of the higher
derivative quantum gravity are found. It is shown that in the physical
region of the coupling constants, that is characterized by the absence
of the tachyons on the at background, the conformal sector has zero
charge behavior. Therefore, the higherderivative quantum gravity
at higher energies goes beyond the limits of weak conformal coupling.
This conclusion does not depend on the presence of the matter elds
of low spins. In other words, the condition of the conformal stability of
the at background, which is held usually as physical, is incompatible
with the asymptotic freedom in the conformal sector. Therefore, the
at background cannot present the ground state of the theory in the
ultraviolet region.
11. It is shown, that the theory of gravity with a quadratic in the curva-
ture and positive denite Euclidean action possesses a stable nonat
ground state and is asymptotically free both in the tensor sector and
the conformal one. A physical interpretation of the nontrivial ground
Chapter 5. Conclusion 143
state as a condensate of conformal excitations, that is formed as a result
of a phase transition, is proposed.
12. The eective potential, i.e., the oshell oneloop eective action in
arbitrary covariant gauge, and the unique eective action in the higher
derivative quantum gravity on De Sitter background, is calculated. The
determinants of the operators of second and forth orders are obtained
by means of the generalized function.
13. The gauge and parametrizationindependent unique eective equa-
tions for the background eld, i.e., for the curvature of De Sitter space,
are obtained. The perturbative solution of the eective equations, that
gives the corrected value of the curvature of De Sitter background space
due to quantum eects, is found.
In conclusion, I would like to thank my scientic supervisor, Prof. Dr.
V. R. Khalilov, for the general guidance, allround help and many
pieces of good advice on all stages of my work. I am also much obliged
to A. O. Barvinsky for many valuable discussions and constant interest
in the work, to all members of the Department of Theoretical Physics,
Physics Faculty, of Moscow State University for the attention to my
work and friendly support, and to the chiefs of the scientic seminars,
where the results obtained in this dissertation were presented, Associate
Member of the USSR Academy of Sciences E. S. Fradkin, Academician
M. A. Markov and Prof. Dr. V. N. Ponomarev, as well as to the
participants of these seminars for fruitful discussions.
Bibliography
[1] C. W. Misner, K. S. Thorne and J. A. Wheeler, Gravitation, (San
Francisco: Freeman, 1973).
[2] J. L. Synge, Relativity: The general theory, (Amsterdam: North-
Holland, 1960).
[3] D. D. Ivanenko and G. A. Sardanashvili, Gravitation, (Kiev: Naukova
Dumka, 1985), 198 pp.
[4] S. Hawking and J. Ellis, The largescale structure of spacetime, (Cam-
bridge: Cambridge Univ. Press, 1973).
[5] L. D. Faddeev, The energy problem in the theory of gravitation, Uspekhi
Fiz. Nauk, 1982, vol. 136, pp. 435457.
[6] A. A. Logunov and V. N. Folomeshkin, Does the radiation of the grav-
itational waves change the energy of a source in Einstein theory of
gravitation?, Teoret. Mat. Fiz., 1977, vol. 33, No 2, pp. 174184.
[7] V. I. Denisov and A. A. Logunov, Does the gravitational radiation exist
in general relativity?, Teoret. Mat. Fiz., 1980, vol. 43, No 2, pp. 187
201.
[8] V. I. Denisov and A. A. Logunov, Does the general relativity have the
classical Newtonian limit?, Teoret. Mat. Fiz., 1980, vol. 45, No 3, pp.
291301.
[9] A. A. Logunov and M. A. Mestvirishvili, Relativistic theory of gravita-
tion, Teoret. Mat. Fiz., 1984, vol. 61, pp. 322346.
144
Bibliography 145
[10] A. A. Logunov and M. A. Mestvirishvili, The bases of the relativistic
theory of gravitation, (Moscow: Moscow State Univ., 1985), 204 pp.
[11] A. A. Logunov, Lectures on the theory of relativity and gravitation. A
modern analysis of the problem, (Moscow: Moscow State Univ., 1985),
258 pp.
[12] H. Georgi, Unied theory of elementary particles, Uspekhi Fiz. Nauk,
1982, vol. 136, pp. 286316.
[13] R. P. Feynman, Quantum theory of gravitation, Acta Phys. Pol., 1963,
vol. 24, No 6, pp. 697722.
[14] B. S. De Witt, Quantum theory of gravity II. The manifestly covariant
theory, Phys. Rev., 1967, vol. 162, No 5, pp. 11951238.
[15] L. D. Faddeev and V. N. Popov, Feynman diagrams for the YangMills
eld, Phys. Lett. B, 1967, vol. 25, No 1, pp. 2930.
[16] S. Mandelstam, Feynman rules for the gravitational elds from the
coordinate-independent eldtheoretic formalism, Phys. Rev., 1968, vol.
175, No 5, pp. 16041623.
[17] E. S. Fradkin and I. V. Tjutin, Smatrix for YangMills and gravita-
tional elds, Phys. Rev. D, 1970, vol. 25, No 12, pp. 28412857.
[18] E. S. Fradkin and G. A. Vilkovisky, Smatrix for gravitational eld II.
Local measure. General relations. Elements of renormalization theory,
Phys. Rev. D, 1973, vol. 8, pp. 42414285.
[19] L. D. Faddeev and V. N. Popov, Covariant quantization of the gravi-
tational eld, Uspekhi Fiz. Nauk, 1973, vol. 111, No 3, pp. 427450.
[20] E. S. Fradkin and G. A. Vilkovisky, Quantization of relativistic sys-
tems with constraints. Equivalence of canonical and covariant formal-
ism in quantum theory of gravitational eld, Preprint Th2332, CERN,
Geneva, 1977, 53 pp.
[21] B. L. Voronov and I. V. Tjutin, On renormalization of the Einstein
gravity, Yadernaya Fiz., 1981, vol. 33, No 6, pp. 17101722.
Bibliography 146
[22] A. A. Slavnov and L. D. Faddeev, Introduction to quantum theory of
gauge elds, (Moscow: Nauka, 1978), 240 pp.
[23] V. N. Ponomarev, A. O. Barvinsky and Yu. N. Obukhov, Geometro
dynamical methods and the gauge approach in the theory of gravita-
tional interactions, (Moscow: Energoatomizdat, 1984), 167 pp.
[24] V. N. Popov, Functional integrals in quantum eld theory and statistical
physics, (Moscow: Atomizdat, 1976), 256 pp.
[25] Quantum theory of gauge elds, Ed. N. P. Konopleva, (Moscow: Mir,
1977), 432 pp.
[26] B. S. De Witt, Dynamical theory of groups and elds, (New York:
Gordon and Breach, 1965), 230 pp.
[27] B. S. De Witt, Quantum theory of gravity III. The application of the
covariant theory, Phys. Rev., 1967, vol. 162, No 5, pp. 12391256.
[28] B. S. De Witt, Quantum eld theory in curved spacetime, Phys. Rep.
C, 1975, vol. 19, pp. 296357.
[29] R. E. Kallosh, Renormalization in nonAbelian gauge theories, Nucl.
Phys. B, 1974, vol. 78, No 2, pp. 293312.
[30] M. T. Grisaru, P. van Nieuwenhuizen and C. C. Wu, Background
eld method versus normal eld theory in explicit examples: Oneloop
divergences in the Smatrix and Greens functions for YangMills and
gravitational eld, Phys. Rev. D, 1975, vol. 12, No 10, pp. 32033213.
[31] N. K. Nielsen, Ghost counting in supergravity, Nucl. Phys. B, 1978,
vol.140, No 3, pp. 499509.
[32] R. E. Kallosh, Modied Feynman rules in supergravity, Nucl. Phys. B,
1978, vol. 141, No 1,2, pp. 141152.
[33] B. S. De Witt, Quantum gravity: New synthesis, in: General relativity,
Eds. S. Hawking and W. Israel, (Cambridge: Cambridge Univ. Press.
1979)
Bibliography 147
[34] B. S. De Witt, Gauge invariant eective action, in: Quantum gravity
II, Second Oxford Symp. 1980, Eds. C. J. Isham, R. Penrose and D.
W. Sciama, (Oxford: Oxford Univ. Press, 1981), pp. 449487.
[35] D. G. Boulware, Gauge dependence of the eective action, Phys. Rev.
D, 1981, vol. 23, No 2, pp. 389396.
[36] L. F. Abbot, The background eld method beyond one loop, Nucl. Phys.
B, 1981, vol. 185, No 1, pp. 189203.
[37] I. Jack and H. Osborn, Twoloop background eld calculations for ar-
bitrary background elds, Nucl. Phys. B, 1982, vol. 207, No 3, pp. 474
504.
[38] S. Ichinose and M. Omote, Renormalization using the background eld
method, Nucl. Phys. B, 1982, vol. 203, No 2, pp. 221267.
[39] C. Lee, Propertime renormalization of multiloop amplitudes in the
background eld method (I).
4
theory, Nucl. Phys. B, 1982, vol. 207,
No 1, pp. 157188.
[40] D. M. Capper and A. MacLean, The background eld method at two
loops. A general gauge YangMills calculation, Nucl. Phys. B, 1982,
vol. 203, No 3, pp. 413422.
[41] I. Jack and H. Osborn, Background eld calculations in curved space
time (I). General formalism and application to scalar elds, Nucl. Phys.
B, 1984, vol. 234, No 2, pp. 331364.
[42] I. Jack, Background eld calculations in curved spacetime (II). Appli-
cation to a pure gauge theory, Nucl. Phys. B, 1984, vol. 234, No 2, pp.
365378.
[43] A. O. Barvinsky and G. A. Vilkovisky, The generalized Schwinger
De Witt technique and unique eective action in quantum gravity, in:
Quantum gravity, Proc. IIIrd Sem. Quantum Gravity, Moscow 1984,
Eds. M. A. Markov, V. A. Berezin and V. P. Frolov, (Singapore: World
Sci. Publ., 1985), pp. 141160.
Bibliography 148
[44] D. M. Capper, J. J. Dulwich and R. M. Medrano, The background eld
method for quantum gravity at two loops, Nucl. Phys. B, 1985, vol. 254,
No 3,4, pp. 737746.
[45] G. A. Vilkovisky, The Gospel according to De Witt, in: Quantum Grav-
ity, Ed. S. Christensen, (Bristol: Hilger, 1983), pp. 169209.
[46] G. A. Vilkovisky, The unique eective action in quantum eld theory,
Nucl. Phys. B, 1984, vol. 234, pp. 125137.
[47] E. S. Fradkin and A. A. Tseytlin, On the new denition of oshell
eective action, Nucl. Phys. B, 1984, vol. 234, No 2, pp. 509523.
[48] N. N. Bogolyubov and D. V. Shirkov, Introduction to the theory of
quantized elds, (Moscow: Nauka, 1976), 479 pp.
[49] C. Itzykson and J.B. Zuber, Quantum eld theory, (New York:
McGrawHill, 1980).
[50] P. Ramond, Field theory: A modern primer, (Massachusetts: Ben-
jamin, Reading, 1981).
[51] O. Y. Karmanov, Eective action and spontaneous symmetry breaking
in curved spacetime, in: Problems of the theory of gravitation and
elementary particles, No 13, Ed. K. P. Stanyukovich, (Moscow: Ener-
goatomizdat, 1982), pp. 4657.
[52] A. A. Grib, Problems of the noninvariance of the vacuum in quantum
eld theory, (Moscow: Atomizdat, 1978), 127 pp.
[53] A. A. Grib, S. G. Mamaev and V. M. Mostepanenko, Quantum eects
in strong external elds, (Moscow: Atomizdat, 1980), 295 pp.
[54] N. D. Birrel and P. C. W. Davies, Quantum elds in curved space,
(Cambridge: Cambridge Univ. Press, 1982)
[55] G. M. Gibbons, Quantum eld theory in curved spacetime, in: General
relativity, Eds. S. W. Hawking and W. Israel, (Cambridge: Cambridge
Univ. Press, 1979), pp. 639679.
Bibliography 149
[56] L. Parker, Aspects of quantum eld theory in curved spacetime, in
Recent Developments in Gravitation, Proc. NATO Advanced Institute,
Cargese 1978, Eds. Maurice Levy and S. Deser (Plenum Publishing,
1979).
[57] V. P. Frolov, Vacuum polarization near the black holes, in: Quantum
theory of gravity, Proc. IInd Sem. Quantum Gravity, Moscow 1981,
(Moscow: Inst. Nuclear Research, 1983), pp. 176187.
[58] V. P. Frolov, Physical eects in the gravitational eld of the black holes,
Proc. P. N. Lebedev Physical Inst., 1986, vol. 169, pp. 3131.
[59] V. P. Frolov and G. A. Vilkovisky, Spherically symmetric collapse in
quantum gravity, Phys. Lett. B, 1981, vol. 106, No 4, pp. 307313.
[60] G. A. Vilkovisky and V. P. Frolov, Spherically symmetric collapse in
quantum gravity, in: Quantum theory of gravity, Proc. IInd Sem. Quan-
tum Gravity, Moscow 1981, (Moscow: Inst. Nuclear Research, 1983),
pp. 155169.
[61] E. S. Fradkin and A. A. Tseytlin, Oneloop eective potential in gauge
O(4) supergravity and the problem of the term, Nucl. Phys. B, 1984,
vol. 234, No 2, pp. 472508.
[62] M. J. Du, Inconsistency of quantum eld theory in curved spacetime,
in: Quantum gravity II, Second Oxford Symp. 1980, Eds. C. J. Isham,
R. Penrose and D. W. Sciama, (Oxford: Oxford Univ. Press, 1981),
pp. 81105.
[63] T. W. B. Kibble, Is semiclassical gravity theory viable?, in: Quantum
gravity II, Second Oxford Symp. 1980, Eds. C. J. Isham, R. Penrose
and D. W. Sciama, (Oxford: Oxford Univ. Press, 1981), pp. 6380.
[64] S. Hawking, Path integrals applied to quantum gravity, in: General
relativity, Eds. S. W. Hawking and W. Israel, (Cambridge: Cambridge
Univ. Press, 1979)
[65] M. J. Du, Covariant quantization, in: Quantum gravity, Oxford Symp.
1974, Eds. C. J. Isham, R. Penrose and D. W. Sciama, (Oxford: Oxford
Univ. Press, 1975), pp. 78135.
Bibliography 150
[66] S. Deser, Quantum gravitation: Trees, loops and renormalization, in:
Quantum gravity, Oxford Symp. 1974, Eds. C. J. Isham, R. Penrose
and D. W. Sciama, (Oxford: Oxford Univ. Press, 1975), pp. 136173.
[67] V. A. Fock, The proper time in classical and quantum mechanics,
Izvestiya of USSR Academy of Sciences, Physics, 1937, No 4,5, pp.
551568.
[68] J. S. Schwinger, On gauge invariance and vacuum polarization, Phys.
Rev., 1951, vol. 82, No 5, pp. 664679.
[69] J. S. Schwinger, The theory of quantized elds V., Phys. Rev., 1954,
vol. 93, No 3, pp. 615628.
[70] S. Minakshisundaram, Eigenfunctions on Riemann manifolds, J. Indian
Math. Soc., 1953, vol. 17, pp. 158165
[71] H. P. McKean and I. Singer, Curvature and the eigenvalues of the Lapla-
cian, J. Di. Geom., 1967, vol. 1, pp. 4369.
[72] R. T. Seeley, Complex powers of an elliptic operator, Proc. Symp. Pure
Math., 1967, vol. 10, pp. 288307.
[73] T. Sakai, On eigenvalues of Laplacian and curvature of Riemannian
manifold, Tohoku Math. J., 1971, vol. 23, pp. 589603.
[74] M. F. Atiyah, R. Bott and V. K. Patodi, On the heat equation and the
index theorem, Invent. Math., 1973, vol. 19, pp. 279330.
[75] N. E. Hurt, Geometric quantization in action: Applications of har-
monic analysis in quantum statistical mechanics and quantum eld the-
ory, (Dordrecht: Reidel, 1983).
[76] P. B. Gilkey, The spectral geometry of Riemannian manifold, J. Di.
Geom., 1975, vol. 10, pp. 601618.
[77] A. O. Barvinsky and G. A. Vilkovisky, The generalized Schwinger
De Witt technique and the unique eective action in quantum gravity,
Phys. Lett. B, 1983, vol. 131, No 4,5,6, pp. 313318.
Bibliography 151
[78] A. O. Barvinsky and G. A. Vilkovisky, The generalized Schwinger
De Witt technique in gauge theories and quantum gravity, Phys. Rep.
C, 1985, vol. 119, No 1, pp. 174.
[79] S. W. Hawking, Zeta function regularization of path integrals in curved
spacetime, Comm. Math. Phys., 1977, vol. 55, pp. 133148.
[80] J. S. Dowker and R. Critchley, Eective Lagrangian and energy
momentum tensor in de Sitter space, Phys. Rev. D, 1976, vol. 13, No
12, pp. 32243232.
[81] J. S. Dowker and R. Critchley, Stresstensor conformal anomaly for
scalar, spinor and vector elds, Phys. Rev. D, 1977, vol. 16, No 12, pp.
33903394.
[82] J. S. Dowker, Single loop divergences in six dimensions, J. Phys. A:
Gen. Phys., 1978, vol. 10, pp. 6369.
[83] S. M. Christensen, Vacuum expectation of the stress tensor in an arbi-
trary curved background: the covariant point separation method, Phys.
Rev. D, 1976, vol. 14, No 10, pp. 24902501.
[84] S. M. Christensen, Regularization, renormalization and covariant
geodesic point separation, Phys. Rev. D, 1978, vol. 17, No 4, pp. 946
963.
[85] S. M. Christensen and M. J. Du, Axial and conformal anomalies for
arbitrary spin in gravity and supergravity, Phys. Lett. B, 1978, vol. 76,
No 5, pp. 571574.
[86] S. M. Christensen and M. J. Du, New gravitational index theorems
and super theorems, Nucl. Phys. B, 1979, vol. 154, pp. 301342.
[87] S. L. Adler, J. Lieberman and Y. J. Ng, Regularization of the stress
energy tensor and scalar particles propagating in general background
metric, Ann. Phys. (USA), 1977, vol. 106, pp. 279321.
[88] L. S. Brown and J. P. Cassidy, Stresstensor trace anomaly in a grav-
itational metric: General theory, Maxwell elds, Phys. Rev. D, 1977,
vol. 15, pp. 28102829.
Bibliography 152
[89] L. S. Brown, Stresstensor trace anomaly in a gravitational metric:
Scalar elds, Phys. Rev. D, 1977, vol. 15, No 6, pp. 14691483.
[90] D. M. Capper and M. J. Du, Trace anomalies in dimensional regular-
ization, Nuovo Cim. A, 1974, vol. 23, pp. 173183.
[91] S. Deser, M. J. Du and C. J. Isham, Nonlocal conformal anomalies,
Nucl. Phys. B, 1976, vol. 111, pp. 4555.
[92] M. J. Du, Observations of conformal anomalies, Nucl. Phys. B, 1977,
vol. 125, No 2, pp. 334348.
[93] D. M. Capper and M. J. Du, Conformal anomalies and renormaliz-
ability problem in quantum gravity, Phys. Lett. A, 1975, vol. 53, No 5,
pp. 361362.
[94] N. K. Nielsen, H. Romer and B. Schroer, Anomalous currents in curved
space, Nucl. Phys. B, 1978, vol. 136, No 3, pp. 475492.
[95] R. J. Reigert, A nonlocal action for the trace anomaly, Phys. Lett. B,
1984, vol. 134, No 1, pp. 5660.
[96] F. Englert, C. Trun and R. Gastmans, Conformal invariance in quan-
tum gravity, Nucl. Phys. B, 1976, vol. 117, pp. 407432.
[97] E. S. Fradkin and G. A. Vilkovisky, Conformal o massshell ex-
tensions and elimination of conformal anomalies in quantum gravity,
Phys. Lett. B, 1978, vol. 73, No 2, pp. 209213.
[98] E. S. Fradkin and G. A. Vilkovisky, Conformal invariance and asymp-
totic freedom in quantum gravity, Phys. Lett. B, 1978, vol. 77, No 3,
pp. 262266.
[99] E. S. Fradkin and G. A. Vilkovisky, On renormalization of quantum
eld theory in curved spacetime, Preprint, Inst. Theoretical Physics,
Univ. Bern, Bern, 1976, 144 pp.
[100] N. D. Birrel and I. G. Taylor, Analysis of interacting quantum eld
theory in curved spacetime, J. Math. Phys. (USA), 1980, vol. 21, No
7, pp. 17401760.
Bibliography 153
[101] T. S. Bunch, BPHZ renormalization of eld theory in curved spacetime,
Ann. Phys. (USA), 1981, vol. 131, No 1, pp. 118148.
[102] L. S. Brown and J. C. Collins, Dimensional renormalization of scalar
eld theory in curved spacetime, Ann. Phys. (USA), 1980, vol. 130, No
1, pp. 215248.
[103] S. J. Hathrell, Trace anomalies and
4
theory in curved space, Ann.
Phys. (USA), 1982, vol. 149, No 1, pp. 136197.
[104] S. J. Hathrell, Trace anomalies and QED in curved space, Ann. Phys.
(USA), 1982, vol. 142, No 1, pp. 3663.
[105] M. D. Freeman, Renormalization of nonAbelian gauge theories in
curved spacetime, Ann. Phys. (USA), 1984, vol. 153, No 2, pp. 339
366.
[106] D. J. Toms, Renormalization of interacting scalar eld theory in curved
spacetime, Phys. Rev. D, 1982, vol. 26, No 10, pp. 27132729.
[107] A. I. Vainshtein, V. I. Zakharov, V. A. Novikov and M. A. Shifman,
Calculations in external elds in QCD: The operator approach, Yader-
naya Fiz. 1984, vol. 39, No 1, pp. 124137.
[108] C. J. Isham, Quantum gravity an overview, in: Quantum gravity II,
Second Oxford Symp. 1980, Eds. C. J. Isham, R. Penrose and D. W.
Sciama, (Oxford: Oxford Univ. Press, 1981), pp. 162.
[109] S. Weinberg, Ultraviolet divergences in quantum gravity theories, in:
General relativity, Eds. S. W. Hawking and W. Israel, (Cambridge:
Cambridge Univ. Press, 1979)
[110] M. Du, Ultraviolet divergences in extended supergravity theories, in:
Supergravity 1981, Eds. S. Ferrara and J. Taylor, (Cambridge: Cam-
bridge Univ. Press, 1982), p. 197.
[111] D. M. Capper, G. Leibrandt and M. R. Medrano, Calculation of the
graviton selfenergy using dimensional regularization, Phys. Rev. D,
1973, vol. 8, No 12, pp. 43204331.
Bibliography 154
[112] G. tHooft and M. Veltman, Oneloop divergences in the theory of grav-
itation, Ann. Inst. H. Poincare, 1974, vol. 20, No 1, pp. 6994.
[113] R. E. Kallosh, O. V. Tarasov and I. V. Tjutin, Oneloop niteness of
quantum gravity o mass shell, Nucl. Phys. B, 1978, vol. 137, No 1,2,
pp. 145163.
[114] D. M. Capper, A general gauge graviton loop calculation, J. Phys. A:
Gen. Phys., 1980, vol. 13, No 1, pp. 199213.
[115] G. W. Gibbons, S. W. Hawking and M. J. Perry, Path integral and the
indeniteness of the gravitational action, Nucl. Phys. B, 1978, vol. 138,
No 1, pp. 141150.
[116] S. M. Christensen and M. J. Du, Quantizing gravity with a cosmolog-
ical constant, Nucl. Phys. B, 1980, vol. 170, [FS1], No 3, pp. 480506.
[117] M. J. Du, The cosmological constant in quantum gravity and super-
gravity, in: Quantum gravity II, Second Oxford Symp. 1980, Eds. C.
J. Isham, R. Penrose and D. W. Sciama, (Oxford: Oxford Univ. Press,
1981), pp. 488500.
[118] M. H. Goro and A. Sagnotti, The ultraviolet behavior of Einstein
gravity, Nucl. Phys. B, 1986, vol. 266, pp. 709736.
[119] D. M. Capper and M. J. Du, Oneloop neutrino contribution to the
graviton propagator, Nucl. Phys. B, 1974, vol. 82, No 1, pp. 147154.
[120] D. M. Capper, M. J. Du and L. Halpern, Photon corrections to the
graviton propagator, Phys. Rev. D, 1974, vol. 10, No 2, pp. 461467.
[121] S. Deser and P. van Nieuwenhuizen, Oneloop divergences of quantized
EinsteinMaxwell elds, Phys. Rev. D, 1974, vol. 10, No 2, pp. 401410.
[122] S. Deser and P. van Nieuwenhuizen, Nonrenormalizability of the quan-
tized EinsteinMaxwell system, Phys. Rev. Lett., 1974, vol. 32, pp.
245247.
[123] S. Deser, H. S. Tsao and P. van Nieuwenhuizen, Oneloop divergences
of the EinsteinYangMills system, Phys. Rev. D, 1974, vol. 10, No 10,
pp. 33373342.
Bibliography 155
[124] M. NouriMoghadam and J. G. Taylor, Oneloop divergences for
Einsteincharged meson system, Proc. Roy. Soc. London, 1975, vol.
344, No 1636, pp. 8799.
[125] S. Deser and P. van Nieuwenhuizen, Nonrenormalizability of the quan-
tized DiracEinstein system, Phys. Rev. D, 1974, vol. 10, No 2, pp.
411420.
[126] E. Sezgin and P. van Nieuwenhuizen, Renormalizability properties of
antisymmetric tensor elds coupled to gravity, Phys. Rev. D, 1980, vol.
22, No 2, pp. 301307.
[127] A. van Proeyen, Gravitational divergences of the electromagnetic inter-
actions of massive vector particles, Nucl. Phys. B, 1980, vol. 174, No
1, pp. 189206.
[128] A. O. Barvinsky and G. A. Vilkovisky, Divergences and anomalies
for coupled gravitational and Majorana spin1/2 elds, Nucl. Phys. B,
1981, vol. 191, No 1, pp. 237259.
[129] P. van Nieuwenhuizen, Supergravity, Phys. Rep., 1981, vol. 68, No 4,
pp. 189398.
[130] Geometric ideas in physics, Ed. Y. I. Manin, (Moscow: Mir, 1983), 240
pp.
[131] Supergravity 1981, Eds. S. Ferrara and J. Taylor, (Cambridge: Cam-
bridge Univ. Press, 1982).
[132] B. S. De Witt, Gravity: A universal regulator, Phys. Rev. Lett., 1964,
vol. 13, No 3, pp. 114118.
[133] I. B. Khriplovich, Gravitation and nite renormalizations in quantum
electrodynamics, Yadernaya Fiz., 1966, vol. 3, No 3, pp. 575581.
[134] A. Salam, Calculation of the renormalization constants, in: Quantum
gravity and topology, Ed. D. D. Ivanenko, (Moscow: Mir, 1973), pp.
180215.
[135] H. Weyl, Raum, Zeit, Materie, (Berlin: Springer, 1923), 333 pp.
Bibliography 156
[136] A. Eddington, The mathematical theory of relativity, (Cambridge:
Cambridge Univ. Press, 1924), 311 pp.
[137] R. Utiyama and B. S. De Witt, Renormalization of a classical gravi-
tational eld interacting with quantized matter elds, J. Math. Phys.
(USA), 1962, vol. 3, No 4, pp. 608618.
[138] T. V. Ruzmaikina and A. A. Ruzmaikin, Quadratic corrections to La-
grangian density of the gravitational eld and singularity, Zhurnal Ex-
periment. Teoret. Fiz., 1969, vol. 57, No 8, pp. 680685.
[139] V. L. Ginzburg, D. A. Kirzhnits and A. A. Lyubushin, On the role
of quantum uctuations of the gravitational eld in general relativity,
Zhurnal Experiment. Teoret. Fiz., 1971, vol. 60, No 2, pp. 451459.
[140] M. NouriMoghadam and J. G. Taylor, Ghost elimination in quantum
gravity, J. Phys. A: Gen. Phys. , 1976, vol. 9, No 1, pp. 5971.
[141] M. NouriMoghadam and J. G. Taylor, Ghost killing in quantized
EinsteinMaxwell theory, J. Phys. A: Gen. Phys. , 1976, vol. 9, No
1, pp. 7376.
[142] D. F. Neville, Gravity Lagrangian with ghostfree curvaturesquared
terms, Phys. Rev. D, 1978, vol. 18, No 10, pp. 35353541.
[143] K. S. Stelle, Renormalization of higherderivative quantum gravity,
Phys. Rev. D, 1977, vol. 16, No 4, pp. 953969.
[144] K. S. Stelle, Classical gravity with higher derivatives, Gen. Rel. Grav.,
1978, vol. 9, No 4, pp. 353371.
[145] J. Julve and M. Tonin, Quantum gravity with higher derivative terms,
Nuovo Cim. B, 1978, vol. 46, No 1, pp. 137152.
[146] E. Tomboulis, Renormalizability and asymptotic freedom in quantum
gravity, Phys. Lett. B, 1980, vol. 97, No 1, pp. 7780.
[147] E. Tomboulis, 1/Nexpansion and renormalization in quantum gravity,
Phys. Lett. B, 1977, vol. 70, No 3, pp. 361364.
Bibliography 157
[148] A. Salam and J. Strathdee, Remarks on highenergy stability and renor-
malizability of gravity theory, Phys. Rev. D, 1978, vol. 18, No 12, pp.
44804485.
[149] B. Hasslacher and E. Mottola, Asymptotically free quantum gravity and
black holes, Phys. Lett. B, 1981, vol. 99, No 3, pp. 221224.
[150] S. L. Adler, Einstein gravity as a symmetry breaking eect in quantum
eld theory, Rev. Mod. Phys., 1982, vol. 54, No 3, pp. 729766.
[151] S. C. Lee and P. van Nieuwenhuizen, Counting of states in higher
derivative eld theories, Phys. Rev. D, 1982, vol. 26, No 4, pp. 934937.
[152] E. S. Fradkin and A. A. Tseytlin, Higher derivative quantum gravity:
Oneloop counterterms and asymptotic freedom, Preprint No 70, P. N.
Lebedev Physical Inst., Moscow, 1981, 59 pp.
[153] E. S. Fradkin and A. A. Tseytlin, Renormalizable asymptotically free
quantum theory of gravity, Phys. Lett. B, 1981, vol. 104, No 5, pp.
377381.
[154] E. S. Fradkin and A. A. Tseytlin, Renormalizable asymptotically free
quantum theory of gravity, Nucl. Phys. B, 1982, vol. 201, No 3, pp.
469491.
[155] E. S. Fradkin and A. A. Tseytlin, Asymptotically free renormalizable
theory of gravity and supergravity, in: Quantum gravity, Proc. IInd
Sem. Quantum Gravity, Moscow 1981, (Moscow: Inst. Nuclear Re-
search, 1983), pp. 1827.
[156] E. S. Fradkin and A. A. Tseytlin, Asymptotic freedom in extended con-
formal supergravities, Phys. Lett. B, 1982, vol. 110, No 2, pp. 117122.
[157] N. H. Barth and S. M. Christensen, Quantizing fourthorder gravity
theories: The functional integral, Phys. Rev. D, 1983, vol. 28, No 8,
pp. 18761893.
[158] E. Tomboulis, Unitarity in higherderivative quantum gravity, Phys.
Rev. Lett., 1984, vol. 52, No 14, pp. 11731176.
Bibliography 158
[159] B. L. Voronov and I. V. Tjutin, On the renormalization of R
2
gravity,
Yadernaya Fiz., 1984, vol. 39, No 4, pp. 9981010.
[160] D. Neville, Conformal divergences and spacetime foam in R+R
2
theory,
Phys. Rev. D, 1982, vol. 26, No 10, pp. 26382644.
[161] E. S. Fradkin and A. A. Tseytlin, Conformal anomaly in Weyl theory
and anomaly free superconformal theories, Phys. Lett. B, 1984, vol.
134, No 3,4, pp. 187193.
[162] F. A. Berezin, Introduction to algebra and analysis with anticommuting
variables, (Moscow: Moscow State Univ. 1983), 208 pp.
[163] A. A. Vladimirov and D. V. Shirkov, Renormalization group and ul-
traviolet asymptotics, Uspekhi Fiz. Nauk, 1979, vol. 129, No 3, pp.
407441.
[164] D. Kramer, H. Stephani, M. MacCallum and E. Herlt, Exact solutions
of Einstein equations, (Moscow: Energoatomizdat, 1982), 416 pp.
[165] A. G. Sveshnikov and A. N. Tikhonov, Theory of functions of complex
variable, (Moscow: Nauka, 1970), 304 pp.
[166] M. Brown, Solutions of the wave equation in curved spacetime: Non
existence of the De Witt integral in De Sitter spacetime, Class. Quant.
Grav., 1985, vol. 2, No 4, pp. 535538.
[167] R. Rajaraman, Solitons and Instantons in quantum eld theory,
(Moscow: Mir, 1985), 416 pp.
[168] A. Erdelyi, W. Magnus, F. Oberhettinger and F. G. Tricomi, Higher
Transcendental Functions, vol. I, (New York: McGrawHill, 1953).
[169] L. P. Grishchuk and Ya. B. Zeldovich, Complete cosmological theories,
in: Quantum gravity, Proc. IInd Sem. Quantum Gravity, Moscow 1981,
(Moscow: Inst. Nuclear Research, 1982), pp. 3948.
[170] A. D. Linde, Inating Universe, Uspekhi Fiz. Nauk, 1984, vol. 144, No
2, pp. 117214.
Bibliography 159
[171] I. G. Avramidi, Oneloop approximation in quantum theory of scalar,
spinor and vector elds in an external gravitational eld, in: Modern
theoretical and experimental problems of general relativity and gravi-
tation, Proc. VIth Soviet Gravitational Conf. Moscow 1984, Ed. V.
N. Ponomarev, (Moscow: Moscow State Pedagogical Inst., 1984), pp.
199201.
[172] I. G. Avramidi, Background eld method in quantum theory, Deposited
at VINITI, No 151285 Dep., (Moscow: Moscow State Univ., 1984),
41 pp.
[173] I. G. Avramidi and A. O. Barvinsky, Asymptotic freedom in higher
derivative quantum gravity, Phys. Lett. B, 1985, vol. 159, No 4,5,6, pp.
269274.
[174] I. G. Avramidi, Asymptotic behavior of gravity theory with higher
derivatives, Yadernaya Fiz., 1986, vol. 44, No 1(7), pp. 255263.

You might also like