T H E P R O C E E D I N G S OF T H E PHYSICAL SOCIETY

Section A

VOL. 65, PART 3

1March 1952

No. 387 A

Quantization of Einsteins Gravitational Field: Linear Approximation

BY SURAJ N. GUPTA
Cavendish Laboratory, Cambridge

*
*

Communicated by L. Rosenfeld; MS. recezved 3rd Octobev 1951

The approximate linear form of Einsteins gravitational field is quantized by using an indefinite metric. It is shown that only two types of gravitons can be observed, n the presence of interaction. The though many more can exist in virtual states i observable gravitons are shown to be particles of spin 2. Using the interaction representation, the interaction of the gravitational field with the matter field is briefly discussed.

&LSTRACT.

1. INTRODUCTION

present investigation is undertaken to carry out the quantization of Einsteins gravitational field, and to investigate the interaction of gravitational quanta or gravitons with other elementary particles. For simplicity we shall first quantize the approximate linear form of Einsteins field, while the exact treatment of the non-linear gravitational field will be given in a subsequent paper. Some work on the quantization of the gravitational field has been carried out in earlier days by Rosenfeld (1930), but since then very little progress has been made in this direction. For the quantization of the approximate linear gravitational field we shall follow the same treatment as has been applied to the radiation field in an earlier paper (Gupta 1950, to be referred to as A). Thus we shall use an indefinite metric for the components of the gravitational field with a negative commutator, and then the gravitons corresponding to these components will be made unobservable by means of supplementary conditions. In this way we shall find that only two types of gravitons can be observed, though many more can exist in virtual states. Since the gravitational field has a large number of components, the present treatment is necessarily more involved than the treatment for the radiation field, but we shall not come across any new difficulty. It must be observed that the supplementary conditions play a vital role in our treatment. For, if there were no supplementary conditions, the present theory would involve observable states with negative probabilities, which would be physically meaningless, It appears, therefore, that the role of the supplementary conditions (or the coordinate conditions) is more fundamental than was originally intended by Einstein. * Now I.C.I. Research Fellow at the Unlverslw of Manchester

HE

PROC PHYS. SOC. LXV, 3-A

I1

I 62

Suraj N . Gupta
L I N E A R A P P R O X I M A T I O N F O R THE G R A V I T A T I O N A L F I E L D

5 2.

The fundamental equation for Einstein's gravitational field is given by RFv- . p g p v = - ~ K ~ T P , ......(1) where the symbols gp", Rpv,and R have the usual meaning, K is a constant, and TPv is the energy momentum tensor of the 'matter' field, where matter includes everything except the gravitational field. Since the covariant divergence of the left-hand side of (1) is known to vanish, it follows that the covariant divergence of T W " will also vanish, whence one obtains
.,...( 2)

where amp= Tab( -g)lI2 is the energy momentum tensor density for the matter field. Also. in order to express the conservation of energy and momentum, we write ......( 3) axy (a; $1 = 0 ;

here'f"' is the energy momentum pseudo-tensor density for the gravitational field, which satisfies the equation

... (4)
Following Einstein (1918), we now obtain the linear approximation for the above field equations by putting g,v = % + K S , 9 . ......( 5 ) where E,," are the constant Minkowskian values for g,,,, and the second and higher powers of K are to be neglected. Using the usual flat space-time tensor notation*, we can then write eqns. (1) and (4)as

. .. (7)
I

According to Einstein, we can also choose coordinate conditions or supplementary conditions given bv so that ( 6 ) reduces to It is convenient to put

0 2 h p v -$apv02h;,i.=~Tp,,.

. * . * . .(9)

hpy = y p v -+~,,VYi.i, and thu. we can write (9), (8) and (7) as
O2YpV=KT,,"#

......(10)

.....I(1l) .....I(12)

* We take the space-tlme coordmates as xll x2, x l , zt.

Quantization

oj

the Gravitational Field

163

It follows from the above equations that

In particular, we obtain for the Hamiltonian density

The eqns. ( l l ) , (12) and (13), describing a linear gravitational field, are so simple that it may seem surprising that they have to be regarded only as an approximation to a more involved theory. I t should, therefore, be noted that the supplementary conditions (12) are compatible with the field eqn. (11) only in an approximate sense, for (11) and (12) give respectively

which agree only in the first approximation. On the other hand, we shall show in the subsequent paper that for the exact non-linear gravitational field the supplementary conditions are exactly compatible with the field equations.

S3

QUANTIZATION OF T H E GRAVITATIONAL FIELD

Lagrangian density for the linear gravitational field may be taken as

where y,,,, and y are to be treated as independent variables. Th,is gives in the usual way for the field equations and the Hamiltonian density . . . . . . (17) OY = 0, 0 2 Y p y= 0,

In order that (18) may agree with (15), we shall ensure by means of a supplementary condition, to be discussed in the next section, that the expectation value of y is equal to that of ypp. But for the purpose of quantization we regard y as an independent field quantity, so that we may easily split the contributions of the various components in the Hamiltonian (18). Now we have to obtain the commutation relations for y p v and y , keeping in view the fact that Y ~ , , = Y , , ~ For ~ . this we note that the canonical conjugate of yI1 is given by 71 --=-aL 1 8711 . . . . . , (19) 11- a(ayll/at) 2 at * Therefore the commutation relation [yl1(r, t ) , Tl1(r, t)]=ia(r-r) gives [yll(r, t ) , (ayll/at) (r, t ) ]=2i8(r - r) ; or, using Schwingers notation (Schwinger 1948), we have , * . * * (20) [Yll(X),Yll(X)I = 2 i W - XL

On the other hand, since y12 = y21, we get

....
SO

I .

(21)

that the commutation relation in this case is given by [YlZ(X), YlZ(X)l - x).

=w

.... (22)
*
11-2

164

Suraj N . Gupta
a8

According to (20) and (22), the commutation relations for yIIy may be written
[ Y J ~ ) ,YA,(~)I =i(S$v,+ Sp,SvJD(x- x ) Further, the canonical conjugate of y is given by

......(23)

and hence we obtain

[y(x),y(x)] = -4iD(x-x). ......(24) Following A, we now regard yPyand y as self-adjoint quantities, and expand them as y p y= (2k)-1/2{aP,( k) * r - k t ) + ut PV (k) r-kt)}, ......(25)
k
k

=x2(2k)-1/2{a(k)

ez(k.r-kt)+

at(k) e-l(k*r--%t)},

......(26)

where a dagger denotes an adjoint. Substituting the above Fourier expansions in (23) and (24) respectively, we get
(27) [apv(k), aX,(k)I = S $ v e + Sp$vI. [a(k), ~ ( k ) = ] - 1. (28) Again, substituting (25) and (26) in (18), using the commutation relations (27) and (28), and omitting the zero-point energy, we obtain for the Hamiltonian of the gravitational field

......

......

J Hdv =~k{ju~v(k)upy(k) -at(k)a(k)).

azz( k), a33( k)

......(29)

For the present purpose it is convenient to replace the operators a,,(k), and aoo(k) by ail( k), aL2(k), a&( k) and aAo(k),which are given by

Xciording to (27) and (30), the commutation relations for the new operators are
[aldk), 4 (k)I = 1, [a&(k), aL6(k)I = 1, [a;&), 4 (k)l= 1, bJO(k), .66(k)1= 1, and the Hamiltonian (29) may be written as

......(31)

Hdv=~k{a1Z(k)a12(k)f aidk)a23(k) + a~l(k)a31(k)

- alo(k)a,o(k) - aSo(k)a,o(k)

- a$o(k)a,,(k)

+ all(k)al,(k) + @;m42(k)+ ai;(k)a;,(k)

+ aAJ(k)aCxk)- .t(k)a(k)>,

......(32)

where we have used the fact that aJk) = avF(k). Thus we have, in all, eleven types of gravitons corresponding to the eleven independent components of ypyand y . We shall refer to the gravitons corresponding to the operator a,#) as the a,,-gravitons, and so on. According to (27) and (28), the commutators involving a,,(k) and a(k) have a negative sign. Therefore, as explained in A, we have to use an indefinite metric for the components ytoand y of the gravitational field, while the remaining components are to be treated in the usual way. Thus, all the operators @12(k),

Quantization of the Gravitational Field

165

%o(k), %o(k), a30(k), &(k), ai&), aho(k) and a(k) appear as absorption operators, while their adjoints are all emission operators. It is to be noted that though now all the gravitons have positive energies, some states will occur with negative probabilities. For, due to the use of an indefinite metric, we have to normalize any state Y as
* .(33) where n,,, n2,, n30 and n are the numbers of the a,,-, azo-,uao- and a-gravitons respectively in the state Y. This occurrence of negative probabilities, however, does not raise any difficulty of physical interpretation, for we shall show below that for all real states we have n , , =n20=n30=n=0.
* . . I

y t y q - l)n,o+Aso+%80t%

94.

THE S U P P L E M E N T A R Y C O N D I T I O N S

I n the absence of interaction the supplementary conditions for the gravitational field may be taken as

(34)

and

(yL:)-y(+))Y =0,

..... .(35)

where yht and y(+)are the positive frequency parts of yap and y respectively. It follows from (34), (35) and their adjoint equations that (aypy/8xp) =0, (yMp) = ( y ) . Thus the supplementary conditions (34) agree with (12) in the classical limit, while (35) ensures that the Hamiltonian (18) is equivalent to (15). Substituting the Fourier expansions (25) in (34), and choosing the x3 axis along k, we obtain {~3i(k)-aoi(k)}Y=O, .....-( 36~) {~3z(k)-~o,(k)}Y=O, ......(36b) {a33(k)--a,3(k)}Y={1/2a~3(k)-a,,(k)}F=0, ......(36c) { ~ ~ o ( k ) - ~ ~ ~ ( k ) } = { ~ ~ ~ ( k =O. ) - ~ 2 ~......( ~ ~ ( k36d) )}Y

As shown in A, a solution of (36 a) consists of a normalized state containing np a3,- and a,,-gravitons, and a superposition of a series of redundant states with arbitrary coefficients and zero normalization. These redundant states, however, do not contribute to any observable effect, Moreover, since the normalization is conserved in course of time, a redundant state will always remain redundant. Thus for all practical purposes we may ignore such states, and regard the aal- and a,,-gravitons as entirely absent in a non-interacting gravitational field. Similarly, (36b) implies the absence of u3,- and a,,-gravitons in the absence of interaction, The situation regarding eqns. (36 c) and (36 d) is slightly simpler, for both of these equations involve the operator a,,(k). It can be easily seen that the only states satisfying (36c) and (36 d) are those which do not contain any ~ 0 3 - , a&- and a&gravitons. We now consider the supplementary condition (35). Substituting the Fourier expansions (25) and (26) in (39, we get (37) {a,,(k) - 2a(k)}Y =O. Using (30), (36c) and (36d), we obtain from (37) (38) {&(k) - a(k)}Y?= O s

......

......

Again, as explained above in the case of eqn. (36a), (38) implies the absence of U& and a-gravitons in a pure gravitational field.

I 66

Suraj N . Gupta

We have seen that the supplementary conditions (34) and (35) enable us to eliminate nine types of gravitons, so that for a pure gravitational field we may take
n2, = n31= nol = no2= no3 = n22= nys=no, = n = 0,
I f f

. . . . . .(39)

where nZ3 is the number of a,,-gravitons, and so on. Therefore, in a pure gravitational field, only aI2- and a;,-gravitons can exist. It follows that the state of vacuum for the gravitational field may be defined as that containing no uI2- and &-gravitons, because the other types of gravitons will be excluded by the supplementary conditions. Moreover, since according to our treatment all the components of yk:' and y(+) contain the absorption operators, the state of vacuum will satisfy the relations
y y r o

=o,

y'+'Yo =o.

.. . . . . (40)

$ 5 . SPIN O F THE GRAVITONS

I n the last section it was shown that only two types of gravitons can exist in a pure gravitational field. We shall now discuss the spin of these gravitons. From the Lagrangian density (16), one can obtain the canonical energy momentum tensor tllyfor the gravitational field as
a

This energy momentum tensor, however, has to be modified according to the treatment of Belinfante (see Wentzel 1949). For this we consider an infinitesimal Lorentz transformation ~X,=~W,,,X, wlth 80,,,,= -&U,,,,, . . . . . .(42) and find a tensor f p , e given by

Then the modified energy momentum tensor 0," will be i a

efiv=tpv-

2 G(fpv,Q+fw,

v+fev,g)*

. . .. . .(44)

and the angular momentum density is given by

M p , e =x $ u e - x v e p e
=X p t v e

-x v t p , +

fpv,

..
,,

I .

(45)

Thus we obtain for the components of the angular momentum

P,k=

-i\

M,k,4

dv= Ptb+ Plh,

...

(46)

with

Ptb= -i

dV(X,Tk,-Xkq4),

......(47)

Quantization of the Gravitational Field

167

where P Z ' is the orbital angular momentum, and Plk is the contribution due to the spin. Substituting the Fourier expansions (25) in (48), we get

et(k'*r-k'O-ai,(k') e-t(k' r-h't)l t (a,,(k) e2(k"-kt)+ut '?(k) e-z(k.r-kt) 1 r - k ' t ) - aj,(k') e-z(k'. r-h't)}] x { a,,(kf)

x {a',(k')

= =
k

f [at,(k)aE,(k) -a$(k)a,,(k) - a,,(k)aL(k) + aX,(k)a,,(k)l

G&(k)%q(k) - at,(k)a,,(k)1.

......(49)

Since the real gravitons correspond to the operators and a m = S{a,,(k) - 4 k ) h we have to consider only that part of Plk which involves the operators a,@), a,,(k), a,&) and their adjoints. These operators occur only in the component Pi2, and their contribution is given by

p:2 = E i[all(k)all(k)+ 42(k)@l,(k)- all(k)azl(k) - 42(k)a,z(k)l

k

=X 2i[af2(k)all(k) - a{f(k)a,Z(k)]. k

...... (50)

In order to separate the contributions of the two independent components in the above equation, we introduce the operators 1 1 a-(k)= -{&(k)+ki&)}, ..(SI) u+(k)= --qj{~&(k)-~&(k)}, 42 where, according to (27) and (31), ,we have [a+(k), aL(k)l =E+), .t_(k)I= 1, ......( 52) [a+(k), aL(k)] =O. In terms of these new operators we obtain Pi2 = E 2 [ ~ y ( k ) ~ + (k U) + (k)a-(k)l

....

= E 2[n+(k) - ~ ( k ) ] ,
k

......(53)

where n,(k) and n-(k) give the numbers of the gravitons corresponding ZO the operators a+(k) and a-(k). This shows that the gravitons are particles of spin 2, and they have two independent spin states with the spin axis parallel or antiparallel t o the direction of motion of the gravitons.
6. TREATMENT OF INTERACTION

It is evident from (11) that the interaction of the gravitational field with the matter field may be taken into account by modifying the Lagrangian density (16) as

which gives for the field equations 02~pv=~TNn 0'~ =~Tpp*

*.

.,..(55)

I 68

&raj N . Gupta

I n the presence of interaction, the supplementary conditions (34) and (35) have also to be modified. Following A, we may state the supplementary conditions in the Heisenberg representation as

rg]+Y=o,

[yflfl-y]+Y=O,

... . . .(56)

where [ ]+ denotes the positive frequency part. These conditions are, of course, identical with (34) and (35) in the absence of interaction. But the necessity of writing them in the form (56) arises from the fact that in the Heisenberg representation yflyand y cannot be split into the positive and the negative frequency parts in the presence of interaction. However, such a splitting is still possible for ay,,,/ax,, and yflp-y. For, according to (55), ayPy/ax, and ypp- y still satisfy the wave equation for plane waves :

where we have taken into account the approximation used in $2. When we pass over from the Heisenberg representation to the interaction representation, the required supplementary conditions may be obtained in the same way as in the case of the radiation field. Following the treatment given for the radiation field (Gupta 1951)*, we easily find in the interaction representation

(yp-y(*))Y=o,

where D(+)(x-x) is the positive frequency part of D ( x - x ) . It should be observed that, due to the occurrence of the T,,-term in (57), those gravitons which could not exist in a pure gravitational field can now appear in virtual states in the presence of interaction. For practical purposes it is most convenient to use the invariant perturbation theory based on the interaction representation. For this we require, in addition to the commutation relations (23) and (24), the vacuum expectation values of the anticommutators. Using the vacuum conditions (40), we easily find, by following Schwinger (1949),
({Y,Y(x), Y&)I
>a = (4A18*p G w w x - X), ({y(x), = -4D(1)(x - 2 ) .

. . . . . (58)
*

It should be noted that the use of an indefinite metric does not make any difference to the calculation of the vacuum expectation values, for the indefinite metric operator 7, defined in A, is just equal to unity for the state of vacuum, which does not contain any gravitons.

4 7.

LORENTZ-INVARIANCE

I n the present paper we have treated the components yo% of yp,, by means of an indefinite metric, while the other components have been treated in the usual way. Thus, though the expectation value of yfly is always real, we have to regard y,,, as a complex tensor with yo, antiHermitian and the other components Hermitian. I n this way the Lorentz-invariance of our treatment is not qyite

* The letter, referred to above, contalns an error. There the relatlon precedlng eqn. (9) should read [8D(8-d)/8tIt -t, = - S(X, -21) S(X, --x,)~(x, - 8 8 ) .

Quantization o j the Gravitational Field

169

obvious. However, the Lorentz-invariance of the present theory can be easily established by splitting up yPy into the Hermitian and the antiHermitian components in a more general way, as we have done in A. For this we take yPv to be of the form y p y = Njl)N!l)n(ll) + N(1)N!z)A(12) + NjI)N!J)A(lB) - N(1)N!O)n(lo) +N L z ) N ! l ) A ( 2 + 1 ) N(B)N!Z)A(22) + NfJN(3)R(23) - N(2)N(O)A(20) P V + j$Tj3)N!l)n(W + Nj3)N!2)A(32) + N(3)N!3)A(38) - N(3)N!O)A(30) - Njo)N(I)A(O1) - N(O)N!2)A(02) - N(o)NL3)A(03) + N(O)N!O)A(OO), .. . . .(59) , ,

where, as in A, N f ) , N r ) , Nj3)and NF) are a set of orthogonal unit four-vectors, Njl), NL3)being space-like, and N,? time-like. Further, A(lI), A(12), , . , are scalars, of which A(1o), A(zo), A(30), A(01), A(o2) and A(0a)are.a&iHermitian, while the rest are Hermitian. Thus, according to (59), we have split ypvinto sixteen components in a formally Lorentz-invariant way. We can now follow the entire treatment of the present paper with two differences: Firstly, instead of yPv, we have to expand the A's into Fourier expansions of the form (25). Secondly, instead of taking the x3 axis along k, we have to choose, as in A, Nf), Nj2),N;")in such a way that kPNjl'= kPNj2'= 0, kPN/13) = - K PN OP) . . . . (60) The treatment for y , of course, remains unchanged, because y is a scalar independent of yP,. I n this way the vacuum conditions (40) and all the other results remain unchanged. Since (59) is a tensor equation, all our results will now be formally Lorentz-invariant. Still, in order to establish the relativistic invariance of our treatment one further point remains to be clarified. According to (59), if we choose a frame of reference in which NF)=(O, 0, 0, i), i.e. the time-axis coincides with NLO), the components yo%of yPv will be antiHermitian while the other components will be Hermitian. Since this property of yPy is not preserved in all frames of reference, our treatment has singled out those frames of reference in which the time-axis coincides with Njo). Thus, in the present theory Nj0) appears as a preferred direction. However, this does not matter, for NF) can be chosen in an arbitrary way, and the results of physical interest are independent of its direction. In fact, after the vacuum conditions (40) have been established, NF) never occurs explicitly in the interaction representation.

.. .

REFERENCES

EINSTEIN, A.,1918,Berlzner Berzchte, p. 154. GUPTA, S. N., 1950, Proc. Phys. Soc. A,63,681 ; 1951, Ibzd., 64,850. ROSENFELD, L ,1930,Z .Phys., 65,589. SCHWINGER, J., 1948,Phys. Rev., 74, 1439;1949,Ibtd., 75, 651. WENTZEL, G., 1949, Quantum Theory of Fzelds (New York , Intersclence), p. 217.