Sei sulla pagina 1di 15

1191

Progress of Theoretical Physics, Vol. 42, No. 5, November 1969

Divergence of Axial-Vector Current


in the Gravitational Field

T oshiei KIMURA

Research Institute for Theoretical Physics, Hiroshima University


Takehara, Hiroshima-ken

(Received June 14, 1969)

To clarify the similarity and dissimilarity between the electromagnetic and gravitational
fields, the vacuum expectation value of the divergence of the axial-vector current is evaluated
in the external gravitational field. Within the framework of the perturbation approach in
the Minkowski space, the requirement of gauge invariance yields the finite extra term in
addition to the expected term. The additional term comes from the diagrams of the triangle
closed loop and the two-sided closed loop which is linked to the former to maintain gauge
invariance. Within the covariant formalism in the Riemannian space-time, it is also examined
whether the additional term can be inferred directly from the field equations by giving a
proper definition of the operator of the axial-vector current. To maintain gauge invariance
explicitly, the spinor fields in the current are separated infinitesimally and are parallel-transported
back. A difficulty lies in defining the correct current which gives the same additional term
as given by the perturbation treatment in the Minkowski space. The origin of the difficulty
is discussed in connection with the non-local nature of the gravity.

§I. Introduction
Recently Adler has shown that within perturbation theory the divergence
of the axial-vector current J 5'" in spinor electrodynamics contains the extra term
(a/4n)s'"v'A.pF'"vF'A.p in addition to the expected term 2mJ5. 1),*) This additional
term plays a role of resolving a discrepancy between perturbation theory and
PCAC hypothesis in the n°~2r decay in the 6-model_2)
The above problems of the appearance of the additional term seems to be the
revived one concerning the equivalence between the pseudoscalar and pseudovec-
tor couplings for the n°->2r problem discussed long ago. 3 ) The discrepancies
between formal and explicit calculations may be attributed to a consequence of
the incomplete mathematical framework of field theory where the local axial-
vector current operator J 5'" is represented by a bilinear form of singular field
operators at the same space-time point. As Schwinger showed, 4) a suitable limit-
ing process can lead to the formal equivalence between the pseudoscalar and
pseudovector couplings. However, the limiting process is not unique, and the
original Schwinger's method seems not necessarily correct because of gauge non-

*) In connection with the chiral transformations for hadron fields, Hiida, Ohnuki and Yamaguchi

have discussed the general form of the divergence of the axial-vector current in Prog. Theor. Phys.
Suppl. Extra Ntunber (1968), 337.
1192 T. Kimura

in variance.
In fact, Hagen has shown that the same result as that of Adler can be ob-
tained by taking the reasonable current operators in terms of a limiting procedure
which is explicitly gauge invariant. 5 ) In the case of massive spinors, the differ-
ence between Adler's and Schwinger's results comes from that of the vacuum
a
expectation value of #J51k, while J 5 gives the same result. That is, Adler's re-
sult does not contain the term of (a/ 4n) sp,v>..pFp,v F>..p in the vacuum expectation
a
value of #J51k, while Schwinger's result contains.
In the massless case, both methods of Adler and Schwinger show that the
divergence of the axial-vector current contains the additional term of the same
type. In this case, we have an interesting consequence that the axial-vector cur-
rent is not conserved, despite the fact that theory is invariant under r5 transfor-
mations. Of course, there may be an objection that we are ignorant of the
existence of a charged massless spinor particle.
The gravitational interaction is the most universal one irrespective of the
masses, charges and other attributes of matter fields, though it is very weak as
compared with the elementary particle interactions. Although there are several
similarities between the electromagnetic and gravitational fields, there are also
some dissimilarities which come from the algebraic property of gauge transfor-
mations. Furthermore, there is a question whether it is possible to introduce a
single Hilbert-space which can commonly be used for the description of a set of
quantum fields distributed over a wide range of the domain where the effect of
the gravity cannot be eliminated locally. 6 ) The non-local nature of the gravity
will be entangled with the incompleteness of the definition for the axial-vector
current in terms of a product of local operators, and there would appear a situa-
tion different from that of the electromagnetic case. It is therefore interesting to
see whether the additional terms appear in the divergences of the axial-vector
currents of the massless and massive spinor particles in the external gravitational
field. If they appear, it will be also interesting to see whether the terms are
expressed in terms of the Riemann-Christoffel curvature tensor which describes
the true gravitational field and can not be eliminated by taking any local coor-
dinate systems.
We shall first calculate the expectation values of J 51k and J 5 in the vacuum
of the spinor field within the approximation up to the order of tc 2 (tc: gravita-
tional constant) in the flat space. The behavior of the vacuum expectation value
for the energy-momentum tensor of quantized matter fields interacting with a
classical gravitational field was studied by Utiyama and De Witt.7) Infinities of
a new type appeared and they were removed by introducing counter terms into
Einstein's equation. It will be shown in § 2 that the vacuum expectation values
of J 5 # and J 5 can be determined with finite coefficients if we take account of the
gauge invariance (corresponding to the invariance under the general coordinate
transformations) and the non-linear property of the gravitational interactions (the
Divergence of Axial- Vector Current in the Gravitational Field 1193

guage-invariant regularization method). The divergence of the axial-vector cur-


rent leads to
(1·1)
where RpvA-p is Riemann-Christoffel curvature tensor. In the massless case, the
first term on the right-hand side should be omitted.
In § 3, in order to clarify the discrepancies among the results of Adler,
Schwinger and Hagen in electromagnetics (and gravitrodynamics) we extend
Schwinger's proper-time formalism so as to apply it to the present problem.
For this purpose, using the theory of manifestly general covariant Green's func-
tions,8) we evaluate higher order corrections directly in the coordinate space. It
is shown that there is a difficulty in giving a correct definition of the axial-
vector current in the external gravitational field. The last section is devoted
to a discussion of the relation between the ambiguities in the additional terms
and the non-local nature of the gravity.

§ 2. Perturbation calculation in the flat space

The equation of motion for a spinor particle with mass m 111 the gravita-
tional field with the metric tensor g pv is described by
(2·1)*)
where r"' (x) satisfies

and is expressed 111 terms of Vierbein h"' a(x) and ra matrix m the Minkowski
space as

The spin-affine connection T"' (x) is given by


T"'(x) = (1/S)hA-a(x) {a"'hA.fJ(x) -r;:;.hpfJ(x)} [ra, yfl], (2 ·2)
where r;p is the Christoffel symbols. In the above, the indices a and {3 are
used to indicate Bein indices.
We define the axial-vector current J 5 "' ( x) by
(2·3)**)

where
r5(x) = (1/24-v' -g)spvA.prp(x)rv(x)Y>,.(x)rp(x)
111 which s 0123 = 1. Similarly, the pseudoscalar current J 5 (x) 1s g1ven by

*) We use the unit fl=c=l. The Greek indices run from 0 to 3 and x 0 =t. The Minkowski
metric tensor is denoted by "IJ"'v: -noo=nll="IJ22="1J33=1.
'f'*) Here and in the following, we omit the normal ordering signs.
1194 T. Kimura

(2. 3')
By the equation of motion (2 ·1) and the relation

o"'rv(x) =T~~h(x)- [T"'(x), rv(x)],


the following equation

(2 ·4)
holds formally. As in the case of spinor electrodynamics discussed by Adler/)
it is expected that (2 · 4) does not hold if we take higher-order corrections into
account.
In order to see this discrepancy, we shall calculate the vacuum expectation
value of J 5 1" (x) and J 5 (x) by perturbation method in the flat space-time. The
Lagrangian density which yields (2 ·1) can be expanded in the flat space as 9)

L= -21 [-cfJr"'ov¢-ovcfJr"'¢]
- [ r;"'v+z-te(¢"'v-r;"'vc/h)
1

+ (1/8) /C2 {r;"'v (¢12 + 2¢2) - cp"'vcpl- cpi'P¢/} J+ ...

(2 ·5)

te¢""v = h/ (x) hva (x) - r;""v,


(2 ·6)
¢1 =¢"}.,A,= r;A,p¢A,p ' ¢2 = ¢A,p¢A,p •

We confine ourselves to the approximation up to the order of te 2• Owing


to the non-linearity of the gravitational interaction, in addition to the diagram
analogous to that of the electromagnetic case (see, Fig. 2), the Feynman diagram
shown in Fig. 1 contributes to the vacuum expectation value. The interaction
term responsible for this diagram is the last term of (2 · 5). The Feynman am-
plitude <L"' 5 ) for this diagram is written as

<L ") = -
1
- ---~---- ~-1-- A,IJ/3 (kl) A,pa (k2) (- ite 2) L/3 15 (kl k2) (2 ·7)
f"D (2nY (2kloY/2 (2k2oy;2 '1-' 'jl a PI" ' '

where L 13 al5pfL (kh /?2) is expressed as

Fig. 1. The diagram for the two-sided closed loop.


Divergence of Axial- Vector Current zn the Gravitational Field 1195

Fig. 2. The diagram for the triangle closed loop.

Lf3atJ PP = t {L1k/ + L2k2g} {f/a(J3$gtJ)PJ, + fJ p(J3C~tJ)a,J


+ {Lzkl(p + L4k2(p} f/a)(J3kl~ k2 71
cg 71 tJ)p + {Lukl(IJ + L6k2(1J} fJ J3)(ak/ k2 71 8~7JP)Jt' (2 •7')

where A<aBm = t {AaB 13 + A 13 Ba} and A<afJ p)(IJBm = t {Aar; p1JB13 + Apf/a<rB13 + Aar; p/3BIJ
+ Apr;a13 B 6 } . All the coefficients Lh L 2, .. ·, L 6 are represented by divergent in-
tegrals, and accordingly L 13 al!pJt has no definite meaning by itself. It should be
determined so as to make <L"' 5) gauge invariant by adding the term which comes
from another Feynman diagram (cf. Fig. 2) (i.e. the gauge invariant regulariza-
tion).
The lowest order interaction term
(2. 5')
contributes to the vacuum expectation value of J 5"' through the closed loop
"triangle diagram" shown in Fig. 2. The Feynman amplitude which comes
from the sum of the diagram shown in Fig. 2 and the corresponding diagram
in which ¢ 613 (k 1) and ¢~'a (k 2) are interchanged is written as

(2·8)

where

(2·9)

Here, it should be noted that the term containing r;"'v ¢">.. ">.. in (2 · 5') does not con-
tribute to (2 · 9).
In order to simplify the discussion, we shall impose the so-called coordinate
condition
(2 ·10)
Since the term r;""v¢">.. ">.. has no effect as mentioned above, the condition (2 ·10) nn-
plies that the terms containing k 16 , k 113 , k 2P and k 2a can be neglected after the
actual calculation of (2 · 9). Accordingly (2 · 9) can be written as
1196 T. Kimura

Tf3aripp, = t (A1k1,(= + A2k/) {r;a(f3e,(=ri)pp, + r; p(!3e,(=rJ)af'} + Askl(p/ja)(!3kl k21Je,(=7Jri)f'


+ A4k2(rJ/j mcakl/?,27Je,(=np)p, + (B1k/ + B2k2,(=) k2(!3e,(=rJ)(pf'kla)
+ Bskl(aklp)k2(!3k} k2 "lehri)p, + B4k2(!3k2rJ)kl(akl,(= k2 "le,(=np)f' (2 ·11)
in which Ai (i = 1, · · ·, 4), B 1 and B 2 are represented by divergent integrals, while
B 3 and B 4 are finite and given explicitly by

B 3 = - B 4 = 64n
2 11 dx
sl-.1;
dy
x2y2
'---------- ·---. (2 ·12)
o o [k 12y (1- y) + k/x (1- x) + 2 (k 1k2) xy + m 2]
We must remember that the terms L!3arJp" are left for further discussion. The
terms Li have the same forms as Ai, B 1 and B 2 (the terms L 2 and L 5 have no
effect by taking the condition (2 ·10) into account). In the following, we shall
consider that Li are included in Ai, B1 and B2.
Here, we determine Ai, B 1 and B 2 in terms of finite B 3 and B 4 by the follow-
ing requirement. The theory must be invariant under the following gauge trans-
formation:
(2 ·13)
which corresponds to the transformation under the general coordinate transfor-
mation. Therefore, T 13 a 6 p" satisfies
(2 ·14)
Equations (2·14) and (2·11) lead to
Ai = - (k1k2) Bi ,
(2 ·15)
B1 = - B2 = (k1k2) Bs •

Similarly, we can evaluate the Feynman amplitude through the diagram with
r"r5 replaced by 2mr5 in Fig. 2:
(2 ·16)

In the present case, T;a!Jp contains no divergent integrals and is given by


l il-x
4 1
r;arJp = - 4
S S d q dx
o
dy ------::---;;------:-:------:--~------c-:----~----=---;-::---=-:----~
[q + /?,1 2Y (1- y) + /<,2 2X (1- x) + 2 (k1k2) xy + m 2] 3
2

x Sp{[ir {q + /?,2X +k1 (1- y)}- m] [rciJ {2q + 2k2x+ !?1 (1- y)} mJ
x [ir{q+k2x+k1(1-y)} -m] [rcp{2q-!?2(1-2x) -2kly}a)]
(2 ·17)
Although this term is finite, it is not gauge invariant by itself. This is a differ-
ence between the electromagnetic and the gravitational cases. However, there
remams the term L 13 a!Jp which comes from the diagram with r"'r5 replaced by
2mr5 in Fig. l. By adding Lf3arJp we find the finite gauge invariant T 13 arJp
Divergence of Axial- Vector Current zn the Gravitational Field 1197

T 13 a6 p= T;a6p + L 13 a!J p
= !C1 {r;a<f3k/ kz"ssn6)p + r; p(/3k/ kz'1B!j 11 6)p} + C2k2<13k/ k2"ss6)(pr-kla) , (2 ·18)
where

From (2 ·11), (2 ·12), (2 ·15), (2 ·18) and (2 ·19) we have

(2. 20)

In the above derivation we put

kl 'J = 0 ' k2 2 = 0 ' (2. 21)


smce the differences from the above values can be transferred into the higher
order non-linear terms with respect to ¢pv· Equation (2 · 20) can be expressed
as
(2 ·22)
where Rr-v 6 r: is the Riemann-Christoffel curvature tensor and Is given by
(2. 23)
m the linear approximation. Equation (2 · 22) is nothing but the proper exten-
sion of the formal divergence equation (2 · 4).

§ 3. Formal derivation in a covariant form

As mentioned in § 1, Schwinger proposed the proper-time formalism in spinor


electrodynamics to exhibit the formal equivalence between the pseudoscalar and
pseudovector couplings. In this section, we shall formally extend his method for
our purpose, though the quantum theory and the general relativity are not con-
sidered to be unified.
For our purpose we must calculate the higher-order corrections in the gen-
eral covariant form in the coordinate space. To preserve the manifest covariance
throughout, we must rely upon the theory of covariant Green's function. 8 ) The
Green's function for a spinor particle in the gravitational field is defined by
V -g[y~'(x)G.r-(x, x') +mG(x, x')J = -o(x, x'), (3 ·1)
where G.r-=Car-+Tr-)G in which rr- is given by (2·2). The G(x,x') can be
written as
1198 T. Kimura

G(x, x') = [r'"(x)S.'"-mS(x, x')], (3·2)


where S (x, x') satisfies

V-g[S.'"'"(x, x') +iRS(x, x') -m 2S(x, x')] = -o(x, x'), (3·3)


where R is the curvature scalar. Following· DeWitt, we can expand S(x, x')
in the following form:

S(x, x)
, =-~~----I:;
Jlf2 ro

(4n-)2 n=o
an ( ------
2
8m
a) n i"'
o
-12 exp [ -z. ( m 2s -(J-) ds
s 2s
J
2
= -------~· I
iJ11 [ ----:-- + 2 { r- -log
1 1
2 +-log (2m 2(J + zc)
• }

rJ + zs
2
8n 2 2

X {l.2 (m I- a1) + !..__2 (m I- 2m


2
3
4 2
al + Za2) + ···}

- {l.m 2I+!_(!i_m 4
I-2m 2a1 +a 2) +~(!i_m 6I+ ... )}
2 2
2 4 ' 24 • 3 3

+ { ( 2~2-- -1- -2~4 + ... ) -- ~ (-~32_ + : : + ... } + .. ·l (3·4)

where the definitions of (J, ::1 112 and I are given in the Appendix. Here, the ar-
guments x, x' in (J (x, x'), an (x, x'), etc., are omitted. In the present case of
spmor particles the an's are determined from the recursion relations
a 0 =I(x, x'),
(3 ·5)
(J ·# an+l•f" +( n +1) an+l-
-A-lf2(Al/2
t:J t:J an
)
•#
'"+lR
4 an.

Since we are discussing whether (1·1) holds or not, it is sufficient to take


account of the terms which contain the derivatives of ¢'"v up to second order
and are of the order of tc 2• Therefore necessary coefficients are a 1 and a 2 •
From (3 · 5) and the limiting values (i.e. two points coincide) of (J (x, x'),
I(x, x'), etc., and their derivatives (which are given in Appendix), we have the
relation

.
11m 1 . a1.'" '" +-
a 2 = - 11111
1 R2 . (3·6)
x' ->.'V 2 x' ->.'V 288

From (3 · 5) and (3 · 6) the limiting values of a 1 and a 2 are given by

lim a1=l_R,
x'->:;;12

(3·7)

1 U[a;3}U[r3] Ra/3f"V Rr3#v >


+ 48
-
0 0
Divergence o/ Axial- Vector Current in the Gravitationa l Field 1199

where Sca 13 J = t [r a' r13].


From the quantum-field theoretical definition of G (x, x')
G (x, x') = - i(T (cjJ (x), (/) (x')) )e (x, x'), (3·8)

(Jl' (x)) and (J5(x)) are written as


(J5 (x)) = - i lim Sp {r 5(x) G (x, x')},
(3·9)
(J&"' (x)) =- i lim Sp {r"' (x) r5 (x) G (x, x')},
provided one takes the average of the forms obtained by letting x' approach x
from the future and from the past. The expectation value (J5) is readily evaluat-
ed from the trace property of fo matrix: r

(3 ·10)

In order to show the formal equivalence, we must calculate (J5 "')."' instead
of (J5"'). An appropriate care is necessary to define the proper (J5"')."'. we
stipulate it as
(J5"')."'= -i lim Sp{r"'(x')r5 (x')G."',(x', x") +G."',(x', x")r"'(x")r5 (x")}.
(3 ·11)

It 1s easily seen that (J5 "').p comes from the three parts
(J&"').p = (J&"'>:~ + (J&"'>:2 + (J5"'>:~' (3 ·12)

where

(3 ° 13)

The relation (3 · 6) being taken into account, (J5 "')~~ has no effect. Accordingly,
1 jl/2
<J5"'>:~ = - H~sn 2 Sp {r6a2}
(J&"')."' = (J&"'):~ + (J&"'):~ = -
2
(3 ·14)

Thus, it is formally shown that the equivalence 1s maintained by interchangin g


the order of the limiting process and the differentiatio n, and by taking a suitable
limiting process.
It should be noted that the above method is not unique and shows only that
1200 T. Kimura

the formal equivalence can be established by taking special cares. The two
methods carried out in § 2 and in the above give the same value (3 ·10) for
<J 5) . In the massive spinor case, the discrepancy sterns from <J5 ~').w The cal-
culation in § 2 yields no terms of the form of e"'v>opR"'"a/3R>op af3, while the above
method gives (3 ·14). In the following, it is shown that another definition for
<Jl)."' leads to a result different from the above.
As mentioned in § 1, Hagen has shown that Adler's result can be obtained
from field equations by using the gauge-invariant current 5)

We shall here see whether there exists an analogous procedure which leads to
the result of § 2. It can be suggested that the corresponding procedure is to
adopt the parallel-transport technique. That is, by starting from the formal axial-
vector current (2 · 3), we move the field operators <jJ and (/) infinitesimally to points
x" = x + ie and x' = x- !e, respectively, and then parallel-transport back to x
along the geodesic. In analogy to the path-dependent operators of Mandelstam/ 0)
the operators <jJ (x) and (/) (x) are replaced by (to the lowest order)

<jJ(x, +e) =</J(x") +t s~v"dy"Ava 13 (y)Sa 13 ¢(x'') + s~v"dy"Ava 13 (y) (y-x")afJ 13 ~;(x")

+ IG Sx:z:"dya¢af3 (y) fJ 13 <jJ (x"), (3 ·15)

(/i(x, -e) =(/J(x') -t(/J(x') SJ:x'dy"Sa 13 Avaf3(y) +fJ 13 (/i(x') s~1!'dy"Ava(3(y) (y-x')a

+ !GfJ 13 (/i (x') i:'dya¢a 13 (y), (3 ·15')

where A"'af3 Is defined by


T"' = tA"'a 13 Sa 13 = (1/8) A"'a 13 [ra, r13 ] (3 ·16)
in which T"' is given by (2 · 2). We shall consider the newly defined axial-vec-
tor and pseudoscalar currents*)

J5"'(x, e) =(/J(x, -e)r"'(x)r5(x)¢(x, +e),


(3 ·17)
J5(x, e) =(/J(x, -e)r5(x)¢(x, +e).
Since we are now concerned with term of the form e"'v>opR"'"a/3R>op a/3 111 the vacu-
um expectation value of <1
5"')."", we neglect the terms which do not contribute

*> It may be said that these operators are defined so as to possess the same transformation
properties under the gauge transformations and Vierbein rotations as the original formal ones. The
gauge transformations· mean the general coordinate transformations followed by the space-time
translation of the point (d. § 4).
Divergence of Axial- Vector Current in the Gravitational Field 1201

to this term. Accordingly we obtain the following equation with the aid of the
field equations:
(3 ·18)
with

(3 ·19)

where Rfk"a/3 = A"a/3.#- A#af3."'


The vacuum expectation value of A 5 in the limit x', x" ->x IS written as

<A5)=-.tiR#"a13 (x) lim o.""(x", x')Sp[ {r#(x)r5(x), Sa 13 }G(x", x')]. (3·20)


x", :t;"->x

The term which contributes to our additional term comes from the first term of
the expansion for S (x, x') given in (3 · 4). Accordingly,
1
<A5)=--R#va 13 (x) lim
0
·""(~;', ~') Sp[ {r#(x)r5(x), Sa 13 }r?.,(x)J.v,(x'', x')],
x',x"->x o (x , x )
2
32n
(3. 21)
where we have taken account of the fact that o.v" (x", x') is equal to (x"" - x"')
in the lowest order approximation. From (A· 9) and the defining equation of
I(x", x') given in (A·2), one gets
1. Ill I ( X",'X) -
.?c" - -O.p" ("
X , X - 1 ·P"("
')1"liD ]·P" ?c"-40 X , X ')R ?cprou<;'ro . (3·22)
a;"' x" ->X x"' x'' ->:JJ
In our case, it is necessary to define lim o.#o.vf o 2 in contrast to the electro-
magnetic case (where the relation F/ev?chFh = iYJ#"F?cpB?cp~nF~'l exists and there is
no need to define such a limit). In order to define it we perform the analytic
continuation to the Euclidean metric and average symmetrically over the direc-
tion of o.fk· This procedure leads to

<a.#o.v! o ) = iYJ #"a. A, a· A,I o = tYJ #"'


2
2
(3. 23)

After straightforward calculations we arrive at


V- g<A5) = (1/128n 2) R#" a/38p,v?cp R?cpaf3• (3. 24)
The above derivation shows that the definition of J~./}. which seems to be rea-
sonable brings about the additional term to the formal conservation equation.
The discrepancy of the coefficients appeared in (2 · 23) and (3 · 24) will be due
to the fact that we have not yet known the mathematically correct definition of
<J 5#).# in the external gravitational field (cf. § 4).
Finally, we mention the massless case. In the gravitational case, there IS
a troublesome problem stemmed from the difficulty that the expansion of Green's
function for the massless case is not valid strictly. This may be seen from (3 · 4)
2
which involves an expansion in inverse powers of m • More strictly, in order
that our expansion for the Green's function converges not only should m be
1202 T. Kimura

non-vanishing but also the gravitational field should be slowly varying over dis-
tances of the order of 1/m. If these conditions are not satisfied we are oblig-
ed to rely upon such a perturbation approach as carried out in § 2.

§ 4. Discussion

Within the framework of the perturbation approach in the Minkowski space


we have derived an additional term in the divergence of the axial-vector current
as shown in (2 · 23). There appears a situation similar to the electromagnetic
case discussed by Adler. This is due to the fact that there is a similarity be-
tween the gauge transformation property of the electromagnetic field and that of
the gravitational field (in the lowest order). As shown in § 2, however, there
appear situations which have no electromagnetic analogues. This is due to the
non-Abelian nature of the gauge transformation for the gravitational field. That
is, the diagram of the triangle closed loop should be linked to that of two-sided
closed loop such that the gauge invariance is maintained.
At first sight tbe appearance of the additional term (1/384n 2) c/"">.-P R~"vafJR>.-/ 13
seems to be strange. However, this may be comprehensible if the non-local
nature of the gravity is taken into account. Although the gravitational field
cannot be distinguished from inertial field by any experiment conducted on a
purely local basis, it can be distinguished by experiment carried out over an ex-
tended region, i.e. by experiment which measures field gradients. In the general
relativity the potential is the space-time metric, and the field gradients which is
the second derivatives of the metric can be expressed uniquely by the Riemann-
Christoffel curvature tensor which describes the true gravitational field. It would
therefore not be surprising even if, when higher-order quantum corrections to
matter fields are included, the Riemann-Christoffel tensors appear explicitly into
the dynamical equation. A similar situation was already found in the case of a
classical spinning particle (pole-dipole particle) for which the equation of motion
contains the curvature tensor and whose orbit differs from the geodesic. 11 )
The essential dissimilarity between the electromagnetic and gravitational
cases appears in the problem of defining the reasonable axial-vector or pseudo-
scalar currents in respective fields. To maintain the parallelism between the
electromagnetic and the gravitational fields, we give the name "gauge transfor-
mations" to the general coordinate transformations followed by a space-time
translation of the point by amount f (x). The theory of general relativity is in-
variant under the coordinate transformations and the translation of the point, and
therefore is invariant under the gauge transformations. The gauge transforma-
tions are more suitable for physical interpretation than the coordinate transfor-
mations, since under the former transformation the field variables are viewed
within a fixed frame (i.e. with the same numerical values of the coordinates).
For spinor field <jJ (x) whose coordinate transform is ¢' (x') = <jJ (x), the gauge
transform (jS (x) is given by
Divergence of Axial- Vector Current in the Gravitational Field 1203

If the theory is of translational invariance, both the invariance of the coordinate


and the gauge transformations lead to the same conclusion. In § 3, we have de-
fined the parallel-transport ¢ (x, e), given by (3 ·15), so as to make it invariant
under the gauge groups for the coordinate transformations and Vierbein rotations.
However, the definitions (3 ·17) for J5 # (x, c) and ] 5 (x, c) are path-dependent and
then depend on the preferred vector e#. Accordingly, ] 5# (x, e), etc., are not
necessarily translational invariant even in the limit c#-------?0. Furthermore, there
remains a problem that the rP-(x)y 5 (x) should be parallel-transported or not.
Though we do not parallel-transport in (3 ·17), there is no reason to do so. I£
we perform the parallel-transport for r# (x) r5 (x)' we would get a :result different
from that of § 3. In any event, the above complicated problems may be an or-
igin which causes discrepancies between the results of §§ 2 and ~~.
Finally, in connection with the parallel-transport construction for ] 5 # (x, c),
we briefly touch the problem of constructing the Hilbert-space in the general
relativity. It is readily seen that (3 ·15) can be expressed in terms of a unitary
transformation as (in the lowest order approximation)
(4 ·2)
where

ul = exp {- i;i I:"dyv Ava;S (y) Ma 18 }'


(4 ·3)
u2 = exp {- i!C Sxx"dya¢a;SP 18 }'

in which P 18 and Ma;S are the total energy-momentum four-vector and the angular
momentum tensor of the local system defined in the local coordinate systems at
x", and satisfy the well-known commutation relations. The above transformation
( 4 · 2) for the field operator can be interpreted as a unitary transformation for
the state vector ?Jf' in the Hilbert-space:

(4·4)
This can be expressed in a differential form as
8#?]!' (x) = i H-Apa;SMa;S + !CcpP- 18 P 18 } ?Jf' (x), (4·5)
which describes the connection between Hilbert spaces defined at a pair of ad-
jacent points. The above equation is nothing but the one proposed by Utiyama 6 )
a part from the second term on the right-hand side of ( 4 · 5) . Since ( 4 · 5) does
not satisfy the integrability condition, our state vector has a many-valued pro-
perty. It can be conjectured that this is the origin of difficulty when one at-
tempts to give a mathematically correct definition of the vacuum expectation
value for the divergence of the axial-vector currents.
1204 T. Kimura

Appendix

--Properties of two-point functions--

It is essential to introduce two-point tensors or two-point spinors (hi-tensors


or bi-spinors) in order to study covariant Green's function. A tWo-point tensor
or two-point spinor is a two-component function of two space-time points which
transforms according to the direct product of two tensor or spinor representa-
tions of the coordinate transformation and Vierbein groups. The bi-tensor which
plays a fundamental role in our covariant calculation is the bi-scalar of geodesic
interval o (x, x') defined by

o = tgt-t"o.p,o." = tga' wo. a'o. 13 ,, (A ·1)


which reduces to irJt-t" (xt-t- xt-t') (x"- x"') in Minkowski space. The geodesic
parallel displacement bi-spinor I(x, x') is defined by
a·t-t I.p, = o, a·t-t' I.p,' = o,
(A·2)
lim I= unit matrix,
x,;~.?J

and transforms like cjJ at x and like c/J' at x'. A determinant D of fundamental
importance in the theory of geodesic 1s giVen by
D = - det (Dp,a'), Dp,a' = - 0 (x, x') .p,a' (A·3)
which satisfies the relation
n- 1(Da·t-t) .p, = 4 . (A·4)
The bi-scalar L1 (x, x') defined by
L1 (x, x') = ( - g(x)) -1!2D(- g(x')) -1/2 (A·5)
plays an important role in Green's functions.
The Green's function S (x, x') given in (3 · 4) 1s defined by

S(x, x') = ( - g (x)) - 114 (- g (x')) - 114 1 00

ds<x, sjx', O)exp (- im 2s), (A·6)

where <x, sjx', 0) is the transition amplitude and satisfies the following Schrodinger
equation:

i~ <x, sjx', O)=<x, sjx', O).p,t-t+: R<x, sjx', 0). (A·7)

This equation is solved by the ansatz


co
<x, sj x', 0) = - i ( 4rc) - 2D 112s- 2ei< 2sft5) ~ an (x, x') sn. (A·8)
n=O

The recursiOn relations for an were given by (3 · 5).


We shall summarize the limiting values as x' ->x of two-point functions
which are necessary in our problem:
Divergence of Axial- Vector Current in the Gravitational Field 1205

0 =lim 6 =lim 6.~' =lim O'.~'v:>. =lim I.~'= lim L1~~ , 2

lim 6. f'V = g f'V , lim L1 = 1 , lim I= 1 ,

lim O'.v6t:p = ~ (Rvt:6p + Rvp6t:),

.
1lffi ,tl/2 -
.:J. /1V - -
1"lffi A-1/2-
.:J. /1V - -
1 R f'V '
6
• ,tl/2v _
1liD.:.~--- 1R (A·9)
.v f' of' '
6

1lln - 41 S [a(3] Ra 13 f'V


. I of'V- '

. I .v v f '--
1lffi - 61 S [a(3]
Ra,e v
f'V• '

. I •/.b f' v v- - 1 S [a(3] S [ro] Raf3 f'v Rrof'v ,


1liD S

References

1) S. L. Adler, Phys. Rev. 177 (1969), 2426.


2) J. S. Bell and R. Jackiw, Nuovo Cim. 60A (1969), 47.
3) H. Fukuda and Y. Miyamoto, Prog. Theor. Phys. 4 (1949), 347.
]. Steinberger, Phys. Rev. 76 (1949), 1180.
Z. Koba, N. Mugibayashi and S. Nakai, Prog. Theor. Phys. 6 (1951), :322.
4) J. Schwinger, Phys. Rev. 82 (1951), 664.
5) C. R. Hagen, Phys. Rev. 177 (1969), 2622.
See also, B. Zumino, Proceedings o; the To-pical Conference on Weak Interactions (Geneva,
1969), CERN 69-7, p. 361.
6) R. Utiyama, Prog. Theor. Phys. 33 (1965), 524.
7) R. Utiyama, Phys. Rev. 125 (1962), 1727.
R. Utiyama and B. S. DeWitt, J. Math. Phys. 3 (1962), 608.
8) B. S. DeWitt, Phys. Rev. 162 (1967), 1239.
See also, B. S. DeWitt, Dynamical Theory of Groups and Fields (Document on modern
physics, Gordon and Breach, 1965).
9) V. I. Ogievetskii and I. V. Polubarinov, Soviet Phys.-JETP 21 (1965), 1093.
T. Kimura, Prog. Theor. Phys. 36 (1966), 394.
10) S. Mandelstam, Ann. of Phys. 19 (1962), 25.
11) A. Papapetrou, Proc. Roy. Soc. A209 (1951), 248.

Potrebbero piacerti anche