PHYSICAL REVIEW D VOLUME 16, NUMBER 2 15 JULY 1977

Hamiltonian formulation and quantization of the spin-3/2 field*

G. SenjanoviC
Department of Physics, City College of the City University of New York, New York, New York 10031
(Received 25 February 1977)
We present a systematic Hamiltonian formulation and quantization of the massless spin-3/2 field theory
using the Hamiltonian methods of Dirac and Faddeev.

INTRODUCTION class and second-class constraints. We then dis-

cuss the choice of gauge and write the Hamiltonian
There exist now several one-loop renormaliz- path integral in several interesting gauges.
able supergravity theories,' in which the graviton Section IV i s devoted to the derivation of Dirac
is coupled to matter fields. It remains still to brackets in the transverse gauge a,$' = 0. This ef-
formulate a consistent renormalizable theory which fectively represents the operator canonical quanti-
contains all the physical particles coupled to the zation (in this gauge) of the spin-2 theory, since
graviton; nevertheless, even before such a theory the anticommutators a r e derived from these brack-
i s found, it i s clear that the one-loop renormaliz- e t s by the correspondence principle. We end Sec.
able supergravity theories hitherto invented a r e al- IV by counting the number of independent variables
ready perhaps the most elegant way out of the necessary to describe massless spin-: particles.
't Hooft-Veltman impasse.' Section V contains a brief summary and conclu-
In all such theories, a crucial role i s played by sions.
a massless spin-2 field, which, in addition to other
fields, i s coupled to the graviton. This has sig- I. ACTION FOR THE MASSLESS SPIN-; FIELD
naled a revival of interest in the quantization of
The massless spin-: field i s customarily de-
this field (both when f r e e or coupled to other
scribed by the action3
fields), a subject that has not received a signifi-
cant amount of attention for quite some time. A
recent paper3 deals with this question by applying
modern gauge-field quantization techniques-path
integrals and the Faddeev-Popov4 ansatz.
There i s , however, another path-integral quan- In Eq. (1.2) $, i s a complex anticommuting c -
tization method, developed by F a d d e e ~ which,~ is number spinor-vector field. Each vector compo-
(a) more rigorous and systemic, (b) explicitly nent 4, i s a four-component spinor. The field
canonical, and (c) s e r v e s a s a justification for the therefore has 16 complex components.
Faddeev-Popov ansatz. This method may be con- The action (1.1) i s invariant under the gauge
sidered a s a combination of functional integral transformation
techniques and Dirac's method6 of quantization of
constrained Hamiltonian systems. Wp(x)= a , ~ ( x ) , (1.3)
It i s this method that we would like to apply to the where ~ ( x i )s an arbitrary spinor field.
quantization of the spin-$ field. This i s the main In Eq. (1.2) the spinor vector field $. i s complex,
objective of the present paper. while the Dirac-Faddeev quantization procedure
We have organized the paper in the following requires real variables. To achieve a description
manner: in terms of real variables, we need only separate
In Sec. I, we review the Lagrangian for the mass-
l e s s spin-: field and i t s gauge invariance, paying
special attention to the case of the Majorana spinor where q,, a r e real (i = 1,2), and use the Majorana
and Majorana representation. representation
In Sec. II, we present a brief review of the meth-
ods of Dirac and Faddeev, with comments on what Y: = -Yp , (1.5)
to do when the degrees of freedom of the theory a r e where the asterisk denotes complex conjugation.
represented by anticommuting c numbers. We find then7
Section 111 i s where we approach the main objec-
tive: we derive all the Hamiltonian constraints for Y; = -YoYp Yo. (1.6)
the free spin-3 theory and classify them into first- Introducing A, = y,y,y,, and using A ; = -A,,
which follows from Eq. (1.6), one obtains for the
action (discarding surface contribution)
while some of 4's remain undetermined. Defining
the Hamiltonian

one can show, using Eq. (2.3) that it depends only

We s e e that the components qi completely de- on the q's andp's and does not contain any 9's. To
couple from each other and, moreover, their dy- account automatically for the constraints we intro-
namics a r e identical. Methodologically, this im- duce the total Hamiltonian
plies a simplification of the Hamiltonian formula-
tion and quantization. We need only discuss $, and
simply remember that the inclusion of $, i s trivial.
But in view of the fact that most of the application where Am i s a s e t of multipliers. H will generate
of the spin-2 quantization scheme i s in the realm correct equations of motion:
of extended supergravity theories which require
spin-; field to be Majorana spinor, we can simply
s e t $, equal to z e r o (note that in the Majorana rep-
resentation Majorana spinor can be taken a s a four-
component r e a l spinor). We therefore replace Eq.
(1.7) by This implies for a function g of q's andp's: g
= {g, H},where we have in mind the "naive" Pois-
son bracket.
remembering that +,
i s a r e a l spinor. In view of The constraints must remain z e r o a t all times,
our discussion, extension to the c a s e of complex which implies
field i s trivial.

11. REVIEW OF DIRAC'S AND FADDEEV'S METHOD: Sometimes, such equations s e r v e to determine
THE CASE OF ANTICOMMUTING VARIABLES
some ( o r all) of u's; often, however, they result
in new (so-called secondary) constraints
The review we present here will be brief to the
point of being sketchy and rudimentary; more com-
plete reviews and original articles will easily be One then continues this p r o c e s s until one a r r i v e s
found by the interested reader.' No proofs will be a t a stage when no m o r e secondary constraints a r e
given here. generated.
Consider a mechanical system with N degrees of At this point, the important classification of con-
freedom (extension to field theory i s straightfor- straints into first-class and second-class ones i s
ward) described by a Lagrangian introduced: If one labels all the constraints by the
L = L ( q 78 ) (2.1)
common name +, and {pa, p,) vanishes for all k
upon application of the constraints, then cp, i s
such that called a first-class constraint. A second-class
constraint i s simply a constraint which i s not f i r s t
det-
1 ai:qn 1 =O. class. Independent linear combinations of the con-
straints a r e just a s good constraints a s the origin-
A Lagrangian of this type i s called singular. De- a l ones; taking convenient linear combinations of
fining the conjugate momenta in the usual way the second-class constraints we can bring a s many
constraints a s possible into the f i r s t class.
With each first-class constraint we must then a s -
sociate a gauge ~ o n d i t i o n . ~
one i s struck by the immediate realization that Introduce a common name @, for first-class con-
(2.3) cannot in view of Eq. (2.2) be inverted for all straints cp,, gauge conditions x,, and second-class
the 4's; a modification of the standard passage to constraints 6,, @ , = (q,,X, ,6,) and define a matrix,
the Hamiltonian description i s therefore necessary. Ct, {@,,a,). Then define a Dirac bracket,
This was achieved by Dirac.' F r o m Eq. (2.3), by
eliminating the 4's, one obtains a certain number
of independent constraints Obviously,
-
16 H A M I L T O N I A N F O R M U L A T I O N A N D Q U A N T I Z A T I O N O F T H E ... 309

{*(,B}* = O (2.11) numbers), respectively, the Poisson brackets a r e

f o r an a r b i t r a r y B .
The quantization prescription i s the correspon-
dence principle: Replace {A, B}* by (l/i)[A, B],
where [A,B] i s the commutator of the two opera-
t o r s representing A and B in the quantum de-
scription. Equation (2.11) then implies [ai,B] = 0
s o that the constraints can now be consistently i m -
plemented a s operator identities cpi 8 0. where 3, and n, a r e coordinates and momenta,
When the mechanical ( o r field) v a r i a b l e s a r e respectively. The p r o p e r t i e s of these Poisson
fermionic, s o that they a r e classically anticom- brackets w e r e discussed in Ref. 9, where it w a s
muting c numbers, caution must be exercised in shown that their algebra i s a graded Lie algebra.
the definition of the naive Poisson bracket; a con- The path-integral quantization in the constrained
venient definition w a s recently achieved by Casal- Hamiltonian formulation w a s solved by Faddeev5
b ~ o n i .Using~ the notation E , and Oi f o r dynamical (in the special c a s e when the theory contains only
v a r i a b l e s which a r e even and odd elements of the f i r s t - c l a s s constraints) and generalized by Senja-
Grassmann algebra (algebra of anticommuting c novid1° and Yabuki." The path integral r e a d s

When the gauge conditions a r e in convolution, that is, when {X,, Xb} = 0 this becomes

The notation in Eqs. (2.13) and (2.14) i s some- We s t a r t now where we quit temporarily, i.e., a t
what symbolic; it m u s t either be supplemented by Eq. (1.8). The Lagrangian contains no time de-
appropriate boundary (i.e., initial and final) condi- rivatives a t Go, s o Eq. (2.3) implies f o r i t s con-
tions, in unique correspondence with the s t a t e s in jugate momentum
the m a t r i x elements displayed in these equations,
nc=O.
o r one must insert appropriate initial and final
wave functions corresponding to those state^.^ This i s a constraint. The momenta conjugate to
This completes the f o r m a l general apparatus, ljli a r e a l s o constrained
and we a r e now well equipped to apply it to theory
in question-the f r e e m a s s l e s s spin-; field.

111. THE HAMILTONIAN FORMULATION AND PATH- (subscript R denotes that the direction of the de-
INTEGRAL QUANTIZATION OF THE MASSLESS rivative i s f r o m right to left). The fundamental
SPIN-; THEORY Poisson brackets a r e given by
In what follows we shall d i s c u s s only the f r e e {n,a(% t ) , +,a(?, t)}=guv6a86(2-?) (3.1)
r e a l m a s s l e s s spin-$ field theory. The inclusion
of the electromagnetic interaction i s easy a s f a r a s {nba(XI,t ) , n U 8 ( ? , f)}= 0 , (3.2)
formal developments a r e concerned, but i s of little (P~.(% t ) , PVB(?, O F = 0. (3.3)
interest since such a theory p o s s e s s e s noncausal
The r e a d e r should be warned that these b r a c k e t s
wave propagation." The really interesting and in-
a r e s y m m e t r i c , a s is obvious f r o m the definition
t r i c a t e example i s the interaction of the spin-? of the Poisson bracket of the odd elements Of of the
field with the gravitational (vierbein) field, but
G r a s s m a n n algebra (formula 2.12).
t h e r e the only s e r i o u s complication is the quantiza-
F o r the constraints 0" nnR + E ~ ' ~ A we~ $ r~e c, o r d
tion s c h e m e f o r the gravity vierbein field itself and
the following for l a t e r use:
thus this i s a problem which in a s e n s e i s ortho-
gonal to ours. ( 2.
{eL(Z, t ) , o',($, t))= 2 ~ ' ~ ' ( ~ , ) , ~ -6 7) (3.4)
The next step i s clearly to find the Hamiltonian: Eq. (3.7) we have several interesting possibilities:

H. = i d 3 x K = =d3xcpui a $ ~ ~ u a , ( o . (3.5)
x =yivi (3.10)
called the Coulomb gauge by Das and reedm man,^
A secondary constraint results: transverse gauge
0=+,(3=
{no(x',t ) , H ) X = a'$, ,
gives and axial gauge
i l k aiAj$b=O
x=lji3.

No further secondary constraints a r e obtained, a s To write the path integral for S matrix elements,
the r e a d e r will readily verify. we use Eq. (2.14) since our gauge conditions a r e in
The following combination of constraints i s f i r s t convolution. since (det{0,0~})"~i s a number
class: ( 6 4 f i ) , the measure i s essentially given by
d e t i X n Xb}
, det~.
Thus, the measure in the path-integral expres-
To summarize, then, our theory has the follow- sion for the S matrix elements (2.14) i s essentially
ing first-class constraints: the s a m e a s for the Maxwell field [except, of
course, for the gauge condition (3.10)]; detM cor-
qJO=nO, (3'6) responding to gauge conditions (3. lo), (3. l l ) , and
cp = aknb - E i j k a i,$,~ , (3.7) (3.12) a r e , respectively:

and the following second-class constraints: detM=detlyiail ,

O ~ = ~ ~ + E ~ ' ~ A , + ~ . (3.8) detM = d e t v 2 ,

d e t ~ = d e t ( a , .l
The gauge condition to match the first-class con-
straint cpO i s For the sake of completeness, we will write down
the expression for the S-matrix elements in the
xO=qO, (3.9)
Lagrangian formulation; that i s we integrate over
whereas for the gauge condition that matches cp in n's. In the transverse gauge using (2.14) we have

( o u t i n ) = / ~ f m o ~ i o y i 0 b ( ~ 0 ) 6 ( n 0 ) 6- r("abkani ~bj + h ) 6 ( a b ( b ) ( d e ~ ) 6 ( n+bE " ' A , ~ )

Integrating over no, sowe obtain

(out1 in) = / ~ ~ ~ ~ ' b ( a , , ~ ' ) b ( da e~ w
n *a(nk
) + r . k ~ a ,exp
) i [ id3x(-cub+:~j)b - cbO'mqf~oailm)].
Integration over n,'s gives

Writing IV. DIRAC BRACKETS FOR THE

6(ab~'jb(:A,) = / 9 $ , det2 exp [(-2)(ab€"k~~j)(o]

The total number of constraints plus gauges i s
one can easily check the result:
seen to be rather large, and it would seem that
(out1 in) = la$,6 (a.@?(det~)r" , (3.13) this requires the inversion of a large matrix. F o r -
tunately, however, we can r e s o r t to an iterative
which i s completely analogous to the free Maxwell procedure which r e l i e s on the iterative property
field. of the Dirac bracket. This property can be de-
16
- H A M I L T O N I A N F O R M U L A T I O N A N D Q U A N T I Z A T I O N O F T H E ... 311

scribed in the following way: Take f i r s t any sub- So in our case, for Xa we take 8: of Eq. (3.8).
s e t of the a's {x,, {x)CG, d e t { ~ , , ~ ~ )and
#~} This is acceptable, since the matrix {oh, 8') has a
compute a preliminary bracket nonvanishing determinant and therefore i s inverti-
ble. Its inverse i s found to be (in block-matrix no-
{A,BIX={A,B)-{A,X,)(CX-')~~{X~,BI, (4.1) tation)
where (CX)abC{ x a , ~ b ) .
Then consider a second subset of the remaining
cp 's:
{Y,,{Y>c{@) - {XI, d e t { ~ c ,Y*)#O) 9

and compute a second-stage bracket For Y, we take the remaining subset Pi:
Pla=(Pa
where (CyIcd={yc, y d l X . of Eq. (3.7) and
Continue this process until all the a's a r e ex-
hausted. The claim is (we do not present the proof) Pza = X u
that the final result i s independent of the procedure of Eq. (3.11).
and, therefore, i s the same a s the Dirac bracket The inverse of {p,, pj) i s
computed in a single step.
Because of (2.11) {+,,B}* = 0 we have immediate-
ly {nO,B)* ={qC,B)*=O for a n y B . That means that
we need only compute {$,,qj}*, {qi, nj)* and Ini,
nj)* a s the fundamental brackets of the theory. F i r s t , we evaluate:

Using

{$ia(x')3qj0(31e= { + i a ( a , qj8(31 - Jd3xtd3~'{$ia(2), ~ ~ ( ~ ) } c ~ R?){0:(2),

~ , ~ o qjB(3)
- ~ ( ~ ~

=-~eij,~B-~~(;-?),
and similarly
{+ja(g),~ f i ( 3 ) =e -6a8ai6(:-3 7

{+ia(a,
p a ( 3 I e = a1cst,aB6(2-3 7

we readily obtain
i 1
{#i(@,4j(3)* = -cei,-'6(G-a+-aiaj76(Z-?)
4
-
- (a,akce.tr-l+ B ~ ~ ' c B u1- ~ ) ~ ~ ( x - ? ) . (4.5)

Because of constraints nR= - E ~ ' " * A , ~this

~ , imme- For the sake of completeness we list fundamental
diately produces Dirac brackets for a general complex field $ = $,
+i$,. From
{$,+I*= {vl,$l}* +z{VJ,, +2}*
1 1
- ; 1 ( a ~ ~ ~ ) € ~ ~ ' a ~ ~ 6 ( :(4.6)
-3. + ~ { $ Z , $ I } *- { # 2 , V J 2 1 *
we easily get
Similarly, we obtain
{$i(z), $ j ( a I *
1
{ n i ( a , n J ( a } *= + aiaj7))a(;- G) and similarly

{n,(z), n j ( 3 I * = O ,
-A4 ~ i j ' a . ( a ' 1~ ,a(;-3.
)~ (4.7)
also
 (3.9) and (3.11)], s o that the number of independent
degrees of freedom in the phase space i s four (we
suppress the coordinate dependence).

V. SUMMARY AND CONCLUSIONS

The reader should not worry about the factor $
in the above formula which can be most simply We have illustrated how the general method of
eliminated by just rescaling Lagrangian by a factor Dirac and Faddeev of the quantization of theories
-1 with constraints applies to the massless spin-$
2
For completeness we must show that our system field theory. The main result of our paper i s the
is Poincare' invariant. Now, in terms of Dirac derivation of fundamental Dirac brackets which
brackets Poincare'group algebra i s satisfied: correspond to the basic anticommutators in the
quantum version of the theory. We have also con-
firmed the result of Ref. 3 on the measure of S
matrix elements using the rigorous canonical ap-
proach. Our basic motivation, besides pure aca-
demic interest, comes from the fact that spin-;
field is one of basic constituents of supergravity
theories which a r e emerging a s strong candidates
(we do not present this tedious, but straightforward for renormalizable theories of matter and gravita-
calculation). This, together with the fact that Pp tion. In view of that it would be interesting to ex-
and M p U a r e constants of motion means that the tend our results to the supergravity theories, in
theory i s PoincarB covariant. particular, to find the measure in the path-integral
4,
We have also checked that = {+*,H) generates expressions for S matrix elements.
the correct equation of motion.
Before concluding this section we want to make ACKNOWLEDGMENTS
an important comment regarding the number of in-
dependent field variables of the theory. The total We thank Professor R. N. Mohapatra, Professor
number of degrees of freedom in the phase space B. Sakita, and S. Wadia for illuminating discus-
i s 32 real variables I),, and 8,~. On the other sions. Special thanks a r e due to P. ~enjanovic'for
hand, we have altogether 28 constraints [con- encouragement, advice, and careful reading of the
straints (3.6) through (3.8) and gauge conditions manuscript.

*Work supported by National Science Foundation under 5 ~ D.

. Faddeev, Teor. Mat. Fiz. 1, 3 (1969) [Theor.
Grant No. PHY75-07376 A01 and in part by the Re- Math. Phys. 1, 1 (1970)l.
search Foundation of the City University of New York 6 ~ A. . M. Dirac, Can. J. Math. 2, 129 (1950); Lectures
under Grant No. 11118. on Quantum Mechanics (Belfer Grad. School of Sci-
o orthe renormalizability of some supergravity theories ence, Yeshiva University, N.Y., 1964).
see, for example, M. T. Grisaru, P. van Nieuwenhui- - {v,
'our conventions are: g p u =diag (1, 1 , -1, - I ) , yU)
Zen, and J. M . Vermaseren, Phys. Rev. Lett. 37, 1662 =2gpul, Y,= ~ Y o Y ~ ~ Y ~ Y ~~, Y: =YOY,,YO.
~ ~ ~ = 1 ~
(1976). 'A. Hanson, T. Regge, and C. Teitelboim, Constrained
ere we have in mind the nonrenormalizability of grav- Hamiltonian Systems (Academia Nazionale dei Lincei,
itation coupled to matter fields; for detailed analysis Rome, 1976); P. SenjanoviE, Ph.D. thesis, 1975 (un-
and extensive list of references s e e P. van Nieuwen-
huizen. ITP Report No. ITP-SB-76-46 (unpublished).
3 ~ Das
. and D. Z . Freedman, Nucl. Phys.w, 227
'R. Casalbuoni, Nuovo Cimento
389 (1976).
=, =,
published); also Ann. Phys. (N.Y.) 100,227 (1976).
115 (1976);

(1976). The physics ofthe spin-3 field was first studied 'OP. SenjanoviE, Ref. 8.
by W. Rarita and J . Schwinger, Phys. Rev. 60, 61 "H. Yabuki, Kyoto University Report No. RIMS-183,
(1941). 1975 (unpublished).
4 ~ N.. Popov and L. D. Faddeev, Kiev Report No. ITP 1 2 ~Velo
. and D. Zwanziger, Phys. Rev. 186,1337
67-36, 1967 (unpublished); Translation NAL Report No. (1969).
THY-57, 1972 (unpublished).