Professional Documents
Culture Documents
Center for Theoretical Physics, Laboratory for Nuclear Science and Department of Physics,
Massachusetts Insfitute of Technology, Cambridge, Massachusetts 02139
1. INTRODUCTION
Attention has recently been drawn to the problem of defining the response of gauge
potentials to coordinate transformations and characterizing those configurations
that are invariant under specific coordinate changes. The reason that this classical
problem retains interest is the presence of gauge-invariance [I]. If one were discussing
a vector field A,(x) which did not undergo gauge transformations, then its response to
a coordinate transformation, whose infinitesimal form is
x” + X” = x’” + 8X”,
(1.1)
6.P = f”(X)
* This work is supported in part through funds provided by the U.S. Department of Energy
(DOE) under contract EY-76-C-02-3069.
257
0003-4916/80/080257-17$05.00/O
Copyright 0 1980 by Academic Press, Inc.
All rights of reproduction in any form reserved.
258 JACKIW AND MANTON
[Recall that the Lie derivative Lf of a general tensor field II”;:::: produces a tensor of
the same type.
The Lie derivative of products of tensors follows Leibnitz’ rule and contractions are
respected, so that the order of taking the Lie derivative and of contracting indices is
immaterial.] An invariant field is one that does not respond to the transformation;
i.e., its Lie derivative vanishes. However, gauge freedom modifies the above. The
concept of coordinate invariance must be extended to allow for non-invariance that
can be compensated by a gauge transformation. Thus an invariant gauge-potential
need not have vanishing Lie derivative; rather it must satisfy the following condition:
LA = f% + ~,(f*A,), (1.5)
This also shows that certain projections of an invariant field strength are [gauge-
covariant] derivatives of a scalar field [3]. Comparison of (1.3), (1.5), and (1.7) gives
the relationship between Y and W.
The composition law for the scalar fields W and Y follows directly by iterating (1.2)
and using (1.12). W satisfies [4]
L, w, - L, w, - [ w, , W,] - w, = 0, (1.13)
while from (1.8) we obtain the corresponding formula for !P, which also reflects the
composition law for gauge-covariant coordinate transformations (1.6) [2].
Moreover, (1.7) permits rewriting the first two terms on the right-hand side of (1.14a)
in terms of the same projection of F,, occurring on the left-hand side. This leaves
On the other hand, if the matter system interacts with an external, prescribed
gauge field, the invariance will be absent in the general case, since a generic external
field removes all symmetries. Consequently, there will no longer be any constants of
motion. However, in the special case when the external gauge fields themselves are
invariant under the infinitesimal coordinate transformation f, then the matter
dynamics will continue to be characterized by constants whose form will now be
Cf = Glatter+ Gern&lgaugefleld
. (1.16)
The second term on the right-hand side allows for a contribution to the total con-
served quantity from the external gauge field. We show here that such contributions
are indeed present; they are essentially determined by Y. In other words, for invariant
fields, the scalar field defined in (1.7) has the physical significance of carrying the
external gauge field contribution to the total constant of motion.
We shall make frequent use of Noether’s method for finding constants of motion.
Let us record how it works. Consider an action I depending on dynamical variables
x through a Lagrangian 2.
I = j d4x g11z9. (1.17)
[It is assumed that no derivatives of x higher than the first occur in A?.] Dynamical
equations of motion for x are determined by requiring that I be invariant against
arbitrary variations of x. Consider next a specific infinitesimal transformation x -+
x + 6x. If under this transformation one can demonstrate, without using equations
of motion, that the action changes into
(1.18)
i.e., apart from surface terms it is invariant, then the transformation is a symmetry
operation and the current P, given by
J* = w’2g) 6x - K”,
Wccx)
is conserved,
&Ja = 0. (1.19b)
C = 1 dr Jo. (1.20)
which could be expressed as a surface integral. Notice that (1.21) is true for any
Lagrangian with no explicit coordinate dependence. This result follows from the
properties of Lie derivatives quoted earlier. We just suppose that all tensorial quanti-
ties transform according to the Lie derivative, and that the Lagrangian 9 is a scalar
constructed from contractions of products of tensors.
Suppose now that the metric and associated geometric quantities are not dynamical.
This is the usual situation of a field theory in a curved space which is not a dynamical
theory of gravity. Furthermore, suppose that the gauge potential is an external
background field, so that only the other matter fields are dynamical. If there is some
combination of transformations which leave both the geometrical quantities and the
gauge-potential invariant, then we can obtain a conservation law for the matter
fields. The argument is simple. Applying this combination of transformation to the
matter, gauge, and geometrical fields, we know that the change in the action is (1.21).
But the gauge and geometrical fields did not change, so we could have applied the
transformations to the matter fields alone, still obtaining (1.21). In this way we find
a transformation of the matter fields which changes the action by a total divergence,
and Noether’s theorem guarantees a conservation law.
The above argument is a generalization, to theories with background gauge fields,
of the usual one relating conservation laws to invariances of the background geometry.
The group of invariances of both the geometry and the background gauge potential
is finite dimensional. In flat space, for example, a gauge-potential could be static and
rotationally invariant, so there would be a conserved energy and angular momentum,
but translation invariance would in general be broken and the conservation of momen-
tum would disappear.
The remainder of this paper is devoted to demonstraring and examplifying the
above. Jn Section II we discuss matter fields, coupled to dynamical gauge fields,
whereas in Section III the gauge fields will be taken to be external and fixed. In
Section IV the matter is taken to be a point particle. Here we encounter a classical
instance of our general theory: the non-kinematical addition to the angular momen-
tum of a charged particle moving in the field of an Abelian magnetic monopole.
262 JACKIW AND MANTON
A = -*p:, (2.6b)
AL:”
+fy:u= w:&,Y~
[The semicolon denotes coordinate covariant differentiation.]
CONSERVATION LAWS IN GAUGE THEORIES 263
Qggauge
=sd’x
?4g”2fa=%auge) (2.8)
Note that the symmetric, gauge-invariant, and traceless stress-energy tensor has
been constructed, because of our use of (2.7).
In the absence of other interactions the current and the stress-energy tensor are
covariantly conserved. The gauge-field contribution to the constant of motion (1.15) is
Now we couple matter to the gauge field. The only matter field theory that we shall
consider in detail is a fermionic, spin-4 Dirac model. The formalism in curved space
is complicated by the need to introduce a vierbein ez [5]. The Greek indices are
raised and lowered by the metric tensor g,,,; the Latin ones by the flat space metric
?lab . The two are related through the vierbein.
ab
guv = %beMev . (2.12)
P = ezy”. (2.13)
Before writing down an action we must still introduce the spin connection, CO:‘.
This is defined by
GM - wzbeb,) - (p t+ v) = 0. (2.14)
264 JACKIW AND MANTON
An explicit expression for WE’ in terms of the vierbein can be found, but we shah not
need this [5]. The action is now
(2.17a)
The action is also invariant under local Lorentz transformations which transform
only the spinors and the vierbein.
The action is, of course, gauge invariant, and finally, owing to the masslessness of the
Dirac fields, the action possesses Weyl symmetry, but to avoid further inessential
complications we shall henceforth ignore this.
The transformation laws of the vierbein, together with the detinition (2.14), imply
that the spin connection transforms like a gauge-potential for the local Lorentz
group, namely,
&%L = Lr% 9 (2.20)
&o, = --a,!2 + [w, ) sz]. (2.21)
Just as in the pure gauge case, when the curved space background is not dynamical,
the Fermi field dynamics will be invariant only under that combination of coordinate
and Lorentz transformations which leave the vierbein [and the metric] invariant. For
there to be such a combination, we need only assume that f is a Killing vector, so
that
L,g,, = 0 (2.22a)
CONSERVATION LAWS IN GAUGE THEORIES 265
or equivalently
.LY +.L = 0. (2.22b)
uniquely determines the matrix I&, , and this matrix does represent an infinitesimal
Lorentz transformation because of (2.12) and (2.22b). We denote this local Lorentz
transformation associated with f by s2,. Clearly, a combination of transformations
which leaves the vierbein invariant also leaves the spin connection invariant.
Having found a combination of transformations leaving the geometrical quantities
fixed, we can now construct a conservation law for the system. The transformation of
the Dirac field is
84 = faDa+ - &+h (2.24)
where D, is the gauge-covariant derivative. Just as in (1.6) and (2.7) we have supple-
mented the coordinate transformation with a gauge transformation. In this way we
obtain a gauge-invariant current
for the fermion. Note that the expected term involving Smatter vanishes when the
field equations are satisfied. The matter contribution to the constant of motion (1.15)
takes the form
CLatter = i s dr g1’2t,6T08,# + h.c. (2.26)
Obviously, when both the gauge field and the matter field undergo the coordinate
transformation, the combined action is invariant, and the combined current
We take f and Ln, to be the same quantities as before, that is, a coordinate and
local Lorentz transformation which together leave the metric and vierbein invariant.
Now we also assume that the gauge-potential has the invariance property
(3.1)
Recall that this equation has the interpretation that the (gauge-covariant) coordinate
transformation generated by f, followed by the further infinitesimal gauge trans-
formation generated by Y, , leaves A, invariant.
We know, from the general argument in Section I, how to find a conservation law
for the matter. We consider the transformation of the Dirac field
This is because the coordinate, local Lorentz, and gauge transformation acting on
the geometrical quantities and all the fields produces the change in the action (1.21),
but this particular combination of transformations leaves the geometry and gauge-
potential invariant, so the same change 81 results from the transformation of the
matter field alone. It follows that the conserved current is
In the transformation law (3.2), each term is separately gauge invariant. There is an
equivalent approach which is perhaps more canonical. In view of (1.8), we can
rewrite (3.2) as
44 = folad - Qf* - Wf# (3.9)
This is still gauge invariant as a whole, because of (1.11). Moreover (3.10) makes no
reference to the gauge-potential, but only to W, . This is interesting since the fields W,
which are determined by (1.13), depend only on the geometry of the symmetry group.
For given W’s, there is a large class of invariant gauge-potentials satisfying (1.3) [l],
but the constant of motion (3.10) is the same for each member of the class.
Now it might be argued that there is no such thing as a background gauge-poten-
tial, and that the conserved quantity we have obtained should be the same as that
obtained when regarding the gauge field as dynamical. In some sense this is the case,
but the demonstration requires some further assumptions, and a little care.
We assume first that the symmetric background gauge-potential satisfies the
vacuum-field equations. Then matter is introduced, and the matter source current is
assumed to be small, so that the matter behaves like a test charge. This source induces
a perturbation of the gauge field, and the dynamical gauge-potential can be written
A, = A: + a,, (3.11)
- 2 tr dr g1’2FoEBb!Pf . (3.12)
I
Integrating by parts, ignoring the surface term, and using the non-Abelian Gauss’ law
for the dynamical gauge field.
9ug1/2Fwo = p, (3.13)
268 JACKIW AND MANTON
show that the second term of (3.12) gives just the integral (3.8), so that if the first
term of (3.12) vanishes, then we shall have demonstrated the equivalence of the
dynamical gauge field and the external gauge-field contributions to the constant of
motion. By substituting (3.11) into the first term of (3.12) and expanding to first order
in a, , we obtain
The zeroth-order terms have disappeared because of the symmetry of the background
field (3.1). An integration by parts and use of -FL = [9:, @] casts the above into
the form
where all quantities, except a, , now refer to the background gauge-potential. Next
we use the symmetry equation for the background field to rewrite the covariant
derivative of !Pf in terms of components of the field tensor, giving
2 tr dr (-9i(g1’2Foafi) + -9,(g”2FoyfB)
s
Finally, the Killing equation satisfied by f, the Bianchi identity satisfied by Fm, , and
the field equation for the background field yield the desired result, that the above
vanishes.
We conclude that the contribution (3.8) to the constant of motion can be inter-
preted as being carried by the gauge field itself, and arises from the interaction of the
small induced uL1with the background potential A: , in the dynamical picture.
A useful generalization of this result is possible. Suppose the background gauge-
potential satisfies the field equations with a fixed background source, which itself is
symmetric. This will allow a much larger class of symmetric background gauge-
potentials. Again the constant of motion obtained by perturbing the gauge-potential
dynamically by the matter will be the same as that obtained by regarding only the
matter as dynamical. Both of these constants will however be different from before.
hence we take for the total action a sum of the gauge-field action [for the moment the
gauge field is assumed to be dynamical] and the particle action.
In the first term, describing free motion, ds is the proper interval and c a normaliza-
tion constant.
ds = (g,“(z) dz@dz”)1/2 = (guy(z) i’ti”)1/2 dh. (4.2a)
9J gl/2Ffqx)) = j”(x),
(4.3)
j”(x) = 1 dh svT a4(x - z).
rP - [.?A,, T] = 0. (4.4)
(4.5)
Equations (4.3)-(4.5) are the Wong equations; they may alternatively be derived by
considering action principles in which T is a dynamical quantity. These more compli-
cated approaches are not needed by us.
270 JACKIW AND MANTON
Coordinate transformations of zu with Killing vectors leave the kinetic term in-
variant. when the particle is massless, iup, = 0, conformal Killing vectors also lead
to invariance.] In order to study invariances of the interaction term in (4.1), we vary
separately zUaccording to 6zu = - f” [note the opposite sign from (1. l)], A,, according
to (1.6), but not T.
= -2 tr
I dX ad (f”A,T). (4.6)
The second equality follows from the first when (4.4) is used. The particle contribution
to the constant of motion follows from Noether’s theorem
We have here treated the gauge field as a dynamical variable; when it is an external
quantity whose prescribed form undergoes no variation, the change of the total
action, which is only the particle action, follows from (4.6):
For invariant gauge fields, we may use (1.7) and then (4.4) shows that (4.9a) is equal to
%particle= -2 tr
s
dX -$ (f”A, - Y,) T.
Just as in Section III, we can recognize, for spatial coordinate transformations, the
external gauge-field contribution to (4.10) as the “frozen” remnant of the dynamical
expression in (4.8). This follows from (3.8) when (4.3) is used for an evaluation of
gl/2p, in the parametrization X = x0.
The two terms in (4.10) are separately gauge invariant. Moreover, the first involves
the velocity of the particle, and so can be regarded as the mechanical contribution. It
is therefore natural in the matter-field case, (3.9, to regard as mechanical the gauge-
invariant matter contribution to the total constant of motion [even though, unlike the
corresponding term in (4. lo), the field analog contains the gauge-potential].
Another form for the constant of motion may be given. When it is noted from (4.1)
that the canonical momentum is
V. DISCUSSION
We have shown that the quantities W and ?P, which initially arise in a formal,
mathematical context as the infinitesimal gauge transformations which must supple-
ment coordinate changes in order to exhibit the invariance of symmetric gauge fields,
also carry the direct physical significance of contributing to conserved quantities
associated with the symmetry. This provides another example of the familiar identity
between constants of motion and generators of transformations, and in special cases
our results have been encountered previously. We now discuss the known examples in
the context of our general theory.
It can happen that the W’s satisfying the consistency condition (1.13) are position
independent, and the symmetry compensating gauge transformation is global. In
that case (3.10) shows that the total constant of motion can be written as
(5.la)
where
T = s dr g1j2p (5.lb)
is the total isospin [or its generalizations]. This situation obtains for spatial rotations,
where a possible form for W, is just nata. Here fi = efgkn*rk, nG$ = 1, so II is just the
595h7b2
272 JACKIW AND MANTON
axis of rotation, and ta {a = 1,2,3} are generators of the gauge group forming an
SU(2) subalgebra. In that case (5.1) is the familiar formula that total angular momen-
tum is composed of orbital, spin, and isospin parts. The Wu-Yang monopole [7],
the ‘t Hooft-Polyakov [8] and Julia-Zee [9] solutions, and the Witten ansatz [lo] are
all examples of spherically symmetric configurations with constant W. The one-
instanton Euclidean solution [l l]
-2ia,,x~
A,, = (5.2a)
x2 + x2 ’
cPy = (1/4i)(&Y - I?&), (52b)
(p = t-h 0, 2 = (iq I) (5.2~)
(5.3a)
(5.3b)
However, constant W is not the generic case. There exist field configurations which
are spherically symmetric yet for no gauge choice are the w’s constant. Examples
have been constructed which are essentially non-Abelian as well as Abelian. The
former are not widely known [3]. The latter have been appreciated in the context of
Dirac monopoles for a long time. The magnetic monopole with magnetic field
B = g(r/rs) (5.4)
satisfies the symmetry criterion (1.7) for rotational invariance. That equation here
reduces to
(n~r)hB=-VY (5.5a)
and from (5.4) we find
Y =gIl*P. (5.5b)
Also it is easy to check that W is not constant. When this is considered in the context
of charged particle mechanics, discussed in Section IV, we recognize the Poincare [13]
result that the angular momentum for the system has a nonkinematical contribution
J=rhp-gi+. (5.6)
Notice also that the field contribution, the second term in (5.6), is due to the inter-
action of the electric field of the electrically charged particle with the magnetic field
of the monopole; a result first discussed by Saha and Wilson [14]. This illustrates the
need to include, as we did in (3.1 l), in addition to the symmetric background gauge-
potential, the perturbation a, due to the back reaction of the charged matter. Indeed
CONSERVATION LAWS IN GAUGE THEORIES 213
we see that the discussion of Eq. (3.12) et seq. is just a generalization to arbitrary
transformations of the Saha-Wilson observation.
For a final example of local rotations, we recall that the instanton (5.2) is invariant
under O(5) rotations, but this symmetry is realized with a position-dependent W.
REFERENCES
1. Mathematicians studied symmetry properties of gauge fields over a quarter century ago; for
a discussion see S. KOBAYASHI AND K. NOMIZIJ, “Foundations of Differential Geometry,”
Interscience, New York, 1963. More recently physicists have looked at pieces of the problem.
See P. FORGACS AND N. MANTON (Comm. Math. Phys., in press) for a discussion and for refer-
ences to the relevant physics literature.
2. R. JACKIW, Phys. Rev. Lett. 41 (1979), 1635.
3. This observation is the starting point for various derivations of Higgs fields from Yang-Mills
fields; see N. MANTON, Nucl. Phys. B 158 (1979), 141.
4. In fact one really derives only the covariant gradient of (1.13), and the condition given in the
text is sufficient but not necessary. For an example that makes use of this distinction see L. BROWN
AND W. WEISBERGER, Nucl. Phys. B 157 (1979), 285.
5. An excellent summary is given by B. ZUMINO, in “Lectures on Elementary Particles and Quantum
Field Theory” (S. Deser, M. Grisaru, and H. Pendelton, Eds.), MIT Press, Cambridge, 1970.
6. S. WONG, Nuovo Cimento A 65 (1970), 689. This derivation is in flat space, but the generalization
to curved space is obvious.
7. T. T. Wu AND C. N. YANG, in “Properties of Matter under Unusual Conditions” (H. Mark and
S. Fernbach. Eds.), Interscience, New York, 1969.
8. G. ‘T HOOFT, Nucl. Phys. B 79 (1974), 276; A. POLYAKOV, Zh. Eksp. Teor. Fiz. Pis’ma Redaktsiyn
20 (1974), 430 [JETP Lett. 20 (1974), 1941.
9. B. JULIA AND A, ZEE, Phys. Rev. D 11 (1975), 2227.
10. E. WITTEN, Phys. Rev. Lett. 38 (1977), 121.
11. A. BELAVIN, A. POLYAKOV, A. SCHWARTZ, AND Y. TYUPKIN, Phys. Lett. B 59 (1975), 85.
12. R. JACKIW AND C. REBBI, Phys. Rev. D 14 (1976), 517.
13. H. POINCAR~, Compr. Rend. 123 (1896), 530.
14. M. SAHA, Indian J. Phys. 10 (1936), 145; H. WILSON, Phys. Rev. 75 (1949), 309.