PHYSICAL REVIEW D VOLUME 15, NUMBER 4 15 FEBRUARY 1977

Functional measure for quantum gravity

Michio Kaku*
Department of Physics, City College of the City University of New York, h'ew York, New York 10031

P. SenjanoviCt
Department of Physics. University of California at Berkeley, Berkeley, California 94720
(Received 9 November 1976)
We derive the functional measure for quantum gravity by reducing to independent degrees of freedom in the
light-cone gauge. We use the recently developed functional techniques devised to handle field theories with
second-class constraints in order to analyze quantum gravity quantized along null planes, where it has been
shown that all eight dependent components of the metric may be explicitly functionally eliminated, leaving
only two unconstrained variables. Using our result, we confirm the result of Fradkin and Vilkovisky for the
functional measure for quantum gravity, though we disagree with several other authors who have different
measures.

INTRODUCTION The presence of $O(x) may seem irksome to a

In the literature there i s considerable confusion
over the precise functional measure for quantum
gravity. Previous results, based on vague invari-
ance arguments, have all been discredited. Re-
cent advances in the quantization of field theories
with gauge invariancesl have made possible the
conclusive determination of the elusive functional
measure for quantum gravity. The key to solving
the problemz lies in observing that in the full!,
reduced, unconstrained, noncovariant Hamiltonian
formulation, where all dependent variables have
been explicitly integrated over, the functional
measure i s equal to unity. Working backwards, it
i s then possible to determine the functional mea-
s u r e that must be added into the constrained La-
grangian formalism (which i s covariant and con-
tains many dependent variables) in order to r e -
produce the correct measure (unity) for the r e -
duced Hamiltonian system with only independent
components. This procedure, first carried out
by Faddeev and PopovYZgave the following result
for the measure:
[In Eq. (1)gaX i s the inverse of the metric tensor
g,v, and g = detg,,. 1
A quite independent study of the problem was
made by Fradkin and Vilkovisky , 3 and in their
paper disagreement was found with the result of
Faddeev and Popov. Fradkin and Vilkovisky ob-
tain
lvf,, = [g''z(x)g"O(x) dg"*(x)]
0s X
= [<31z(x)g00(x) dgaX(x)].
USA
naive reader (the measure looks noncovariant),
but a s Fradkin and Vilkovisky carefully demon-
strate, one cannot judge by appearances alone.
We refer the reader to their paper for details.
In this paper, we shall present a new investi-
gation of the problem of functional measure,
which i s an application (and extension) of one of
the recent studies of quantum gravity (reduced to
independent variables) in the light-cone gauge.l
The advantage of the light-cone gauge i s that all
eight redundant components of the gravitational
metric can be explicitly functionally eliminated
from the Lagrangian, s o that we a r e left with an
unconstrained noncovariant Hamiltonian, where we
know that the functional measure i s one. In the
usual Coulomb gauge for gravity all redundant com-
ponents cannot be explicitly eliminated from the
functional integral. Seven redundant components
of the metric tensor can be completely eliminated,
but the last redundant component requires com-
plex, formal iterative procedures. In the light-
cone gauge, however, all eight redundant com-
ponents can be explicitly eliminated.
We shall close the introduction with an impor-
tant comment concerning Eq. (2). As the reader
will note by inspecting the paper by Fradkin and
V i l k ~ v i s k y Eq.
, ~ (2) holds only if one accepts a
particular determination of certain ambiguous
t e r m s of the type [see their Eqs. (2.8), (2.9),
(2. l o ) , the discussion following those, and also
the remark following their Eq. (2.21))
which consists in writing
[a,aUDF(x - yi],.,= [a,a. /& c ~ ~ . ( ~ - ~ )
(2) ]x-y9
(3)
integrating f i r s t over ko in the integral in Eq. ( 4 ) ,
taking the derivatives, and tlzer~taking the limit
x - y. In the resulting expression only the t e r m
proportional to 6'4'(0) is i ~ o l a t e d . ~
If, however, one chooses the determination of
Eq. (3) which consists in writing
(where q, is a certain timelike vector of n o r m 1
and 5 ; a r e t h r e e orthonormal v e c t o r s orthogonal
to q u ) and integrating f i r s t o v e r k , in Eq. ( 3 ) , then
the tensor s t r u c t u r e multiplying 6'4'(0) in Eq. (3)
i s q,qv. The special c a s e q, = 6,, corresponds
to the previous determination.
This m o r e general determination has the con-
sequence of replacing Eq. (2) by
What can b e said about the light-cone-type reg-
ularization of ( 3 ) , which consists in integrating
f i r s t over k - = (kO+ k 3 ) / f l ? T h i s i s clearly the
one we a r e interested in, since we shall study the
functional m e a s u r e f o r gravity quantized on the
light cone, in which type of quantization such a
regularization i s a consequence of the standard ex-
pression f o r D,(x) ( s e e Ref. 5 ) ,
In Eq. ( 6 ) ) p .x stands f o r p'x-+.b-x+ = qx-+ (3/2q)x+.
The r e a d e r will easily check that a t e r m propor-
tional to 6'4'(0) (with this type of regularization)
will be found only when
Repeating the arguments of Fradkin and Vilkovis-
ky ( f o r t h i s type of regularization), one finds in-
stead of Eqs. (2) and (5)
In Eq. (9), the left-hand s i d e is the S-matrix
element f o r transition f r o m the s t a t e li) containing
I
only physical gravitons to a s t a t e f ) of the s a m e
nature.
On the right-hand s i d e a ( x ) and p(x) a r e two
field variables corresponding to two unconstrained
d e g r e e s of freedom of a physical graviton, and
T h e m e a s u r e thus depends on the type of regu-
larization adopted f o r the ambiguous t e r m s , but
of course quaiitun1 g ~ n v i t ytheovy does not, since
the contributions of the m e a s u r e simply cancel
other contributions in the perturbation expansion
proportional to 6'4'(0) a s long a s the regularization
procedure is uniform, a s demonstrated by Fradkin
and Vilkovisky.
I. REDUCTIOK OF QUANTUM GRAVITY IN THE LICHT-
CONE GAUGE TO INDEPENDENT FIELD VARIABLES
In a recent study by one of the authors (M.K.),4
the method of path integrals was used to effect
( a f t e r a 2 t 2 decomposition of the m e t r i c tensor)
a complete elimination of the eight dependent com-
ponents, A formulation of quantum gravity quan-
tized on the null planes in t e r m s of only two un-
constrained Lagrangian d e g r e e s of freedom w a s
thus obtained.
F o r this purpose, we use the light-cone gauge
condition
-
In Eq. (8), k = det(gir) ( i ,j = 1 , 2 ) , 1 is an a r b i t r a r y
constant, and of a l l the 6,+ only 6-+ i s nonvanishing
(and equal t o one).
It was a l s o demonstrated that this gauge i s
ghost-free. In functional integral language, this
simply means that the Faddeev-Popov determinant
corresponding to this choice of gauge is a constant
independent of the field variables.
L e t u s emphasize that the gauge selection c o r -
responding to Eq. (8) differs f r o m most o t h e r s
known in the l i t e r a t u r e on quantum gravity, in that
it implies a radically different quantization
scheme, one along the s u r f a c e s of constant "null-
plane time" x+= (xO+ x1)/d2 r a t h e r than s u r f a c e s
of constant xO.
The final result of this investigation can be d e -
scribed by
& ( a , P ) is the corresponding Lagrangian, which
we shall display shortly. ~ /i sfthe i coupling
constant of gravity.
In the intermediate expression O f and d i a r e the
field wave functionals f o r the s t a t e s If)
and li);
n,b, i s the 6 functional corresponding t o the gauge
choice (8), A,, is the corresponding Faddeev-
15
- FUNCTIONAL MEASURE FOR QUANTUM GRAVITY
Popov determinant, and M i s the functional mea-
s u r e discussed by Faddeev and Popov and Fradkin
and Vilkovisky. It i s left unspecified in Eq. (9).
By displaying an independent derivation of the
factor in curly brackets we shall be able to deter-
mine dl.
The form of $ ( a , P ) in Eq. (9) i s the following:
-
p) = ik1I2(eija+a-eij)
K'C(CY, + ( 2 l +$)k1I2a+a-
lnk
+ ~ k i [ e i ~ ( a , a m e i j e-Z2(aiejk)(akeii)ejl]
m readers who have sometimes encountered the
+ 2eif(aikS1')ajki1' + $ k - i + l e i j d ~ i d ~ ~ ,
(10)
-
,pfj = kl+leij-1 1 a l(kic'ma-emi)+ik1i2(a-e1m)aielm
a-
- k'/2aia-lnk+?k1i2a,lnk a, lnk -~k1128ia.lnk]
A dimensional factor K' (where ~ / fi si the gravi-
tational coupling constant), usually set equal to
unity by an appropriate choice of units, has been
reintroduced for future convenience.
In Eq. (10) and (11), eif a r e given in t e r m s of a
and p by
where
and, finally, k i s expressed in t e r m s of e i j and
e i j in the following fashion:
@f and @ i a r e final and initial field wave function-
als; { , ) denotes a standard equal-time Poisson
bracket, en(;) a r e the second-class constaints,
Q,'S a r e the s e t of fields appearing in the theory,
II, a r e their conjugate momenta, X is the Hamil-
tonian, and the square root of the functional de-
terminant i s precisely the above-mentioned func-
tional measure.
11. THE HAMILTONIAN PATH INTEGRAL FOR
CONSTRAINED SYSTEMS WITH SECOND-
CLASS CONSTRAINTS
While the Lagrangian exhibited in Eq. (10) con-
tains only two independent field variables, a ( x )
and / 3 ( x ) , we shall see shortly that the Hamiltonian
formulation corresponding to that Lagrangian con-
tains constraints. This may seem bizarre to those
statement that "constraints in the Hamiltonian
formalism correspond to a gauge invariance of
the theory" [and in the transition to the reduced-
variable description the original gauge invariance
of quantum gravity has disappeared (after all, the
gauge has been fixed before that transition)].
However, the above statement i s only partially
true-it applies only to the so-called first-class
constraints, which a r e those whose standard Pois-
son brackets with each other vanish upon applica-
tion of the constraint e q ~ a t i o n s . ~
Second-class constraints6 (which can be defined
a s those that a r e not first-class) can also appear
in certain field theories, and they do not c o r r e s -
pond to a gauge invariance.
A notable example a r e field theories rewritten
in t e r m s of null-plane variables. All these con-
tain second-class constraint^.^
The functional measure for constrained Hamil-
tonian systems containing only first-class con-
straints was found by Faddeev.' For the case of
systems containing both first- and second-class
constraints, the measure was evaluated (extending
Faddeev's method) by one of the present authors
(P. S.) ,' and independently by Yabuki.lo We shall
need only a special case of their result, namely
the one holding for theories containing only sec-
ond-class constraints, which can be formulated
a s follows:
111. FUNCTIONAL MEASURE FOR GRAVITY IN THE LIGHT-
CONE G.4UGE REDUCED TO INDEPENDENT VARlABLES
We a r e now well equipped to proceed with the
calculation of the appropriate functional measure
corresponding to the Lagrangian (10).
The f i r s t step will consist in rewriting the
relevant portions of that Lagrangian directly in
t e r m s of the variables a,P. We shall call a = q,
and p = q,.
It i s a simple exercise to show that
K , S (=~~)~ k ~ ~ " l l , , ( a + ~ , ) ( a - ~ , )
+ (21+ $)k1/2a+a-lnk+ ~ ( q,)
where
In Eq. (16), V(q) does not contain "time" deriva-
tives (i.e., derivatives with respect to x') of the
q's, and 1z i s still given by the chain of equations
(14), with the last equation in the chain now r e -
written a s
For the purpose of transition to the Hamiltonian
formalism, the system will be taken to "evolve"
along the direction of varying x+, i.e., the role of
the time variable will be played by the null-plane
time x+.
One can now transcribe the action correspond-
ing to Eq. (16) a s
+(21+i)(a+qs) j d z
- 6(a- l q ~ ~ - , ~ + , ; tinuous
1%
6qs(x-,x+,2 )
The functional derivative in Eq. (19) i s an ordi-
nary local derivative a s f a r a s the dependence of
q, on x+ and ;;i s concerned. It i s a genuine func-
tional derivative only a s f a r a s the dependence of
q, on x' goes. E.g.,
To go over to the Hamiltonian formalism, one
must derive the conjugate momenta. One obvious-
ly finds
Equations (20) represent constraints, since they
do not contain "time" derivatives.
Our constraints a r e therefore
(16)
One then derives the null-plane Hamiltonian:
In the second step, the constraints (21) were used.
These a r e no secondary constraints generated
from the constraints appearing in Eq. (21) by ap-
plication of the equations of motion. The proof
i s simple. According to Dirac: the equation of
motion for a quantity f i s given by
The symbol { , ) denotes standard Poisson
brackets, X, a r e a s e t of a priori arbitrary)
multipliers, and 8, a r e the constraints. In our
case the contraints a r e labeled by a s e t of con-
lndices x', ;; [we a r e dealing with a field
(i.e., continzrum) theory at fixed x+], and by a dis-
crete index 1.11 (1n = 1,2).
In our case, therefore, Eq. (24) reads (if one
takes f to be the constraints themselves)
+ /dy-d~A,(x+, Y-, XQ.(X+, , Y - , ?)I.
x-,i ) om(x+,
(25)
An explicit evaluation in the r e s t of this section
will show that for our theory
and thus Eq. (25) simply determines the multi-
pliers and generates no new constraints. Equa-
tion (26) i s also Dirac's criterion: which, when
fullfilled, implies that all the constraints O,(x)
(and any linear combination of those) a r e second
class.
After some rather straightforward algebra,
one finds, using Eqs. (21) and (22),
15
- FUNCTIONAL MEASURE FOR QUANTUM GRAVITY
Using standard properties of determinants, we can see that
Primes in Eq. (29) denote derivatives with respect
to x-.
Let us now go back to Eq. (7). In the light-cone
gauge: using g = kZT+land g+'= kmi, one finds for
M in this equation
On the other hand, the measure of Faddeev and
Popov [Eq. (Z)] gives in this gauge
Equation (28), combined with the use of
yields
This must equal the measure in Eq. (9), which
produces the equation
We have been able to calculate S only by ex-
panding it in powers of K [up to (and including)
terms of order K ' ] and have found
To this order, this agrees with Eq. (30) and dis-
agrees with Eq. (31) [note that k = 1+ O ( K ~ ) ]and
,
therefore settles the controversy in favor of Frad-
kin and Vilkovisky.
To spare the reader from the unenlightening in-
tricacies of the calculation which includes t e r m s
of order K ~ ,we shall limit our presentation to the
discussion of the evaluation of S up to (and includ-
ing) t e r m s of order K ~ . The (cumbersome) evalua-
tion of the O ( K ~terms
) was accomplished by a
straightforward extension of the techniques used
to perform the calculations of the t e r m s of order
K ~ .
Throughout, we shall use
This is the standard determination made in light-
cone quantization (see, e.g., the paper by Corn-
1024 MICHIO KAKU AND P. SENJANOVIC -
15

wall and Jackiw," especially the formula a t the det(1 + K'Q, + K4QZ)= exp T r In( 1+ KZQ1 + K4QJ
end of their Sec. 11).
= exp T~(K'Q,+ K ~ Q-' B K ~ Q ~. ' )
Extracting an infinite constant factor det(a/ax-)
we obtain for S, to O(K'), In Eq. (36),

det( 1 + K'Q) = I + K Z TrQ . (36)

In the calculation to O ( K ~ )we

, used and

6(a- lnk),, 6k1/2(z') d(a-lnh),

+ f
2K
(21+ i)
/dl'( --
6qrn(x) 6ql(y) 6 q 1 ( ~ ~ )6qm(x)
obtained from Eq. (29) by keeping only O(K') t e r m s and relabeling, e.g., z - - z.
Note that Q, itself is of order KO.
Observing that

we find

where we have used the antisymmetry of (l/a-),, But in view of Eq. (35),
which follows from Eq. (35). To order K', we
have
- 6
6ql(y)
, 1 6 lnk(z)
(2)=-
s o that TrQ vanishes. This completes our pro-

= K 2 La~ L r(~)zy2q;(y)].
a- (41)
mised presentation of the proof that

Using Eq. (41) and similar identities for 6 lnk(x)/

6q,(v) and 6(a- 1nk12),/6qrn(x),we find t c order K'.

T ~ =Q/dxdy ($) {GyX q;(31)] Iu:(x) SUMMARY

+ 2(2l+ i) /dxdydz (k),& [($) ZY

'q:(y)]
We have presented a method of calculating
the functional measure for quantum gravity r e -

"[(k),, .
duced to independent variables in the light-cone
(42) gauge, studied recently by Aragone and Chela
x ax ~;(x)] Flores, Kaku, and Scherk and S c h w a r ~ .This ~
A partial integration in y in both t e r m s and an method i s based on recent results by one of the
integration in z in the second term produces the present authors (P.S.)' (independently found also
expression by ~ a b u k i " ) for the functional measure in the
Hamiltonian-type path integral for an arbitrary
T r g = - j d x q:(x)ql(x) (&) ' XX
Hamiltonian system containing second-class con-
straints.
Our result was then used to extract information
about the functional measure for quantum gravity
 FUNCTIONAL MEASURE FOR QUANTUM GRAVITY
in i t s standard formulation. The resulting expres-
sion was found to agree with the result of Fradkin
and V i l k ~ v i s k y . It
~ therefore disagrees with the
result of Faddeev and P o p o ~ . ~
For objections to the result of Faddeev and Pop-
ov. a s well a s a discussion of the relevance of the
*Work supported in p a r t by the NSF under Grant No.
MPS 75-07376 and i n p a r t by the R e s e a r c h Foundation
of CUNY under Grant No. 11136.
t ~ e s e a r c hsponsored by the National Sclence Foundation
under Grant No. PHY 74-08175-A02.
'L. D. Faddeev and V. N. Popov, Kiev Report No. IT-67-
36 (unpublished).
'L. D. Faddeev and V. N. Popov, Usp. Fiz. Nauk 111,
427 (1973) [Sov. Phys.-Usp. 16,777 (1974)l .
3 ~ S. . Fradkin and G. A. Vilkovisky, Phys. Rev. D 8 ,
4241 (1973).
4 ~ Kaku,. Nucl. Phys. E,99 (1975). Other r e c e n t
s t u d i e s a r e C. Aragone and J. Chela F l o r e s , Nuovo
Cimento E ,225 (1975); J. Scherk and J. H. Schwarz,
Gen. Relativ. G r a v i t . fi, 537 (1975).
1025
problem of measure, we refer the reader to Ref.
3.
ACKNOWLEDGMENTS
We acknowledge helpful discussions with Profes-
s o r S. Deser and Dr. A. Jevicki.
9. B. Kogut and D. E. Soper, Phys. Rev. D L , 2901
(1970).
6 ~ A. . M. Dirac, Can. J. Math. 2, 147 (1950); P. A. M.
Dirac, Lectures on Quantum hlechanics (Academic,
New York, 1965).
7 ~ Banyai
. and L. Mezincescu, Phys. Rev. D j , 417
(1973); Rev. Roum. Phys. 18,1035 (1973).
'L. D. Faddeev, T e o r . Mat. Fiz. 1, 3 (1969) [ T h e o r .
Math. P h y s . 1, 1 (1970)l.
'P. SenjanoviE, Ph.D. t h e s i s , City University of New
York, 1975 (unpublished); P . SenjanoviE, Ann. P h y s .
(N.Y.) 100,227 (1976).
'OH. Yabuki, Report No. RIMS-183, 1975 (unpublished).
"J. M. Cornwall and R. Jackiw, Phys. Rev. D 4 , 367
(1971).