ORO 3992 328

D e c e m b e r 1977

GAUGE THEORY OF GRAVITY AND SUPERGRAVITY

ON A GROUP MANIFOLD

BY

Y u v a l N e 'e m a n and T u l l i o Re g g e
 ..'it n.i» i; . 1 i • i» i * i • . - t. i t . , i; : * i tv .

• i ! r-t I i : * 11>; *,i ; *7- 1• ’ '■'];« “ l'i!!'■1.i (>\

>i" t• , tin • i •• : r «’u: I i .ii;-. f

r•, r. ■ *1 .i:t -; m;,.1,• , • .. i: •"i< •;t . !).< i•n.i" i*.*■

■’ - ; • ’ . < i- •, • t ii j i .i 1 r if ' ill !'

' .;• ' i ‘ 'i'1 I. i1" *!<. I1 i]•

’ - in > :m >1 «• ’ t «. t j i>: 1 i . i ’< 1 c i i

•: r e , ' t • in • r ; nit • <i " A ' ’<* ■' ' ! * i * > »»n .

• ... ’‘• '“ tr.il • * i- iT • i i' *" i s: :1 : ‘r.J

t • .f : » «i i, :i t • , n v f o t n ; U : o r : , ! ;h \rihn ! < 1 ’ <. } ' . . !, j'l/u .

' r. • *: : ‘ fe i ^ t u !•. t ht v , • ‘ i . >i! *' • .m bin'll, .M v. 1 . i ia .

it ...... -i u-1"

jt -v’:i i•» t i -\>" i-1 *\ . In ih»»; »»■ji

* " u iT\ :i!’ \ i i\"V. *<
.v u *’.i*>'iv
 1 2

X. Introduction 2. Differential Geometry and Lie Groups

In recent years, the intriguing properties of Graded This is a vast subject of which we can only reproduce a
Lie AlgebTas (GLA) have generated considerable interest.1"4^ few key definitions and results.^ We also do not attempt
In particular, it is widely felt that these structures may to generalize systematically the theory to GLA’s® ^ but shall
provide us with a new principle, unifying Gravity with some provide a few examples of interest.
of the fundamental interactions of Particle Theory. Such Let M be a differentiable manifold with covering Ua and
theories (and any gauge theory involving the action of a let x„
a be coordinates on Uo , A p-form on ITci is an expression
gauge group with a non-trivial action on space-time) are of the kind,
most naturally formulated by using the concept of (graded or
ungraded) Grassmann algebras of forms in the context of the n » Z dx*1 „ dx 2 ~ ~ <lx P EA?*«-A (xa5 (2.1)
o a a » x p
theory of Lie groups. IA> - (Ax....Ap)
Much of the ground work for this kind of theory has
been anticipated by E. Cartan and is commonplace in the cur­ where (a,b represent the gradings of A,B when the variables

rent mathematical literature. However, it is not as broadly are elements of a Grassmann algebra)

known as it deserves among physicists. Part of our discussion dx* dxB

a a - -(-l)ab
v ' dx®
a „ dx*
a
is accordingly concerned with the problem of casting unified
theories in the language of forms. In order to make the and the are differentiable functions of the xQ . On Uan U g
paper as self-sufficient as possible, we rederive known we identify the forms na and ng provided
material of the theory of Lie groups, thus also displaying Bi B2 Bd
A. 3x. A- 9Xa A ax.*1 fa1
the consistency features of the method. Finally, existing n - E Z dx 1 - S r * •” - dxcP - S - (V xa »
“ (AHB> “ jT T j/ 2 Ap P
theories are analyzed and discussed in the light of the geo­ a a a (2>2)
metrical formalism. In this case, the set of all forms na defines a unique
form n on M. In the sequel, unless specified, we drop the
a dependence.
The /> product of differentials can be extended by linearity
in an obvious way to generic forms, and we have
 3 4

„Pa .Cqb - ( - n P ^ ^ b . npa (2.3) It can be checked that

where npa, are p,q forms respectively, and a,b are their dX*v » A*dv
GIA gradings. (2.7)
The differential dnp of a p-form is a (p*l)-fom defined X*(v . o) - A*v * X*a
recursively by the properties:

The d,« operations are thus coordinate independent and are

d(dx ) » 0 dQ(x) - £ dxA preserved through maps. This property makes them ideal in
A 3x
(2.4) treating covariant theories. We now apply the algorithms of
d(npa « 5qb) - dnp a .e(,b ♦ (-l)P np a .d£<,b forms to Lie groups. (We refer the reader to references 2, 6, 7
for the effects of grading as in (2.4).)
For any form, we always have d(dn) = 0. We have adopted here Let G be a Lie group (or a formal group in the Berezin-Kats
(2.2, 2.4) the convention7^ of putting the differential (or the sense®^) parametrized by x^***xc . We use in the sequel elements
I c l C ( M.
contravariant index) to the left of the derivative (or the co­ x,y,z*-*€G whose parameters are x •••x , y •••y , etc. The
variant index), when summing over ''dummy" indices. Note that element xA ■ 0 denoted by e is the identity in G. The product
the operator d behaves as a p* 1, a « 0 element in this product on G is a map C-G 4 G, and we write
Grassmann manifold of differential forms.
Given manifolds M and N, a map X, M * N, and a form vp on z * xy instead of z « A(x,y) ,
N, the "pull back" X*vP is a p-form on M, defined by the follow­
ing procedure. satisfying the associativity condition
If vP is given by

t (xy) = (tx)y y t .x .y
A A
V " {A}dy 1 *’**“ ^ P % * * • * (y) : y€N (Z-5)
In coordinates we have for the product element z ,
and the map X is realized by y « *(x), then we have,

zA -- AA (x1***xc ,y1**»yc) (2.8)

B B
X*v - E d x 1 Zlg- . ••• „dx p N u ) (X(x)) (2.6)
{B) 3x 1 9x P
 5 6

yielding the differential forms, Consider now the (c* 1-matrix) 1-forms

dzA - d*M V(x,y)* ♦ dyN W(x,y)£ (2.9) id » dx W(x"1,x)

( 2 . 11 )
In the sequel we use the c * 1 row matrices IT - dx V(x,x’1)

dz ■ (dz1*«*dzc) , dx ■ {dx1***dxc) etc. Using (2.10) it is easily checked that

and consider Vjj, wjj as c < c matrices. Briefly, then l*(a)w * w , r*(a)it ■ n ( 2 . 12 )

dz • dxV * dyW u and ir are therefore called Cartan's left and right invariant
forms, respectively, or more briefly the L.I. and R.I. forms.
Associativity implies t(xy) “ (tx)y. This equation can By the properties of d and „, any form obtained from u(n)
be differentiated in all sets of variables, and the results through these operations is also L.I. (R.I.). Therefore, if
compared. One gets: we expand

V(t,x)V(tx,y) - V(t,xy) (a) du) (2.13)

K be- “H
W(t,x)V(tx,y) » V(x,y)W(t,xy) (b) (2.10)
the C*BE - -C*EB must be L.r. o-forms (that is functions)
W(tx,y) - W(x,y)W(t,xy) (c) and are therefore constants, the structure constants of the
Lie algebra of G. Similarly one has
As a consequence of (2.10) we have V(x,e) * K(e,x) • 1.
The product on G gives two kinds of natural maps on G, the (2.14)
- K be
left and right translations:
Equations (2.13), (2.14) are referred to as the Cartan-
G -UaL c where 1(a) (x) • ax
Maurer equations. Under right translations we have:
G r(a^» G where r(a)(x) « xa
 7 8

r*(a)(o « dx V(x,a)W(a"1x"1 ,xa) In (2.19), in , u are the Cartan L.I. forms written in
y *
(2.15) the variables x,y respectively. Also, x translates y on the
■ dx W(x'1,x)W(a'1,e)V(a’1,a) » uad(a_1) left and does not appear in w ; y translates x on the right
and appears in ad(y"1) according to (2.15). By inserting (2.19)
where
in (2.1J) we find an identity relating 2-forms. Separate
ad(a_1) - W(a*1,e)V(a'1,a) - K{a':,e)V'a(e.a*1) (2.16) identification of the coefficients of dx « dx, d x A d y , dy a dy
produces:
(In deriving (2.IS) and (2.16), use has been made of (2.10).)
Similarly, CBFDad(y-l)gA - CABEad(y-l)'Bad(y-1)‘E
( 2 . 20)

l*(a)n • nV(e,a‘1)W(a,a'1) • itad(a) (2.17) d(ad(y'1)gA - C^g^adfy'1) ^

By using (2.10) one further finds that: These identities are very useful in establishing the
gauge covariance of field equations.
ad(a)ad(x) ■ ad(ax) (2.18) The map a ♦ ad(a'1)T » cd(a) is also a representation,
the "coadjoint" representation of G, which is equivalent to
T
so that ad(a) is a c * c representation of G, the adjoint repre­ the adjoint in a number of interesting cases ( stands for
sentation. The matrix ad(a) is tied to the structure constants the transposed matrix). Tangent vectors transform according
by useful identities. to the coadjoint representation.
Consider the map G x G 4 G. From the previous comments 1-forms are essentially synonymous to covariant vector
it is clear that A*u, A*ir also satisfy the Cartan-Maurer fields used by physicists. Contravariant fields appear in
equations. If we use x,y respectively as coordinates on the the mathematical literature as "tangent vectors" and are
first (second) copy of G, we can write xy ■ A(x,y) and, written in the form:

A*u « <i)„ ♦ <i)_ad(y'*) (2.19) uM - t t (2.21)

7 * 3x
 9 10

M
Given a 1-form Q ■ dx q^, we can define the scalar product [Da ,Db] ■ CEa b De (2.25)

which shows that the form a finite Lie algebra, i.e., by

Given two tangent vectors, T, U we have a natural bracket definition the Lie algebra of G. In this role, the ClAg
operation: reappear in a form more familiar to the physicist. However,
(2.25) is fully equivalent in content to the Cartan-Maurer
[T.UJ - I > 2* - U* (2.22) equations. From (2.23), (2.24) it is also clear that the
L 3x 3x J3x
are invariant under left-translations.
defining a Lie-algebra structure on tangent vectors. If we One can similarly introduce a right-invariant vector
write field SB such that nA (Sg) » 6AB , and

uA - (SA .SB1 « -CEa b Se (2.26)

M The geometrical interpretation of DA is clear when con­

and introduce 0>A through
sidering the identity W(x"1,x)W(x,e) • W(e,x) » 1. From (2.11)

or i ' V ■ «$. (2-23) and the second equation in (2.23) we have,

Ife can then define tangent vectors DA through da - ^ (2-z7)

But now if <l>(x) is a generic function on G, the expression

°«• " • “>

Clearly, uA(Dg) » fifB dyADAv,(x) ■= dyAWAM (x,e) -\*(x)

(Note that we use A, B, ... for "Latin" or anholonomic indices,
and N, N, ... for "Greek" or holonomic indices.] » dyA ^ * *(X)
3y 3x"
After some algebraic manipulations one finds by using
(2.23), (2.15): dxM -24, * <50/ (2.28)
3x
 XX 12

represents the infinitesimal change 6i|i of 1J1 under an infini­ 3. Examples of Lie Groups of Physical Interest
tesimal translation to the right in the argument of i(i. In this
The (Euclidean) Poincar£ group P is defined as the set of
sense, the DA generate right translations. Similarly, the SA
pairs x ■ (E,x), where S £ S0(4), x £ j R ? . The product rule
generate left translations. Quite naturally then both the
is, setting t = (0,t):
L.I. of 0^ and the R.I. of can be expressed as

(z,h = (se,Et*J) - (s,x)(0,t) (3.1)

t DA * S B^ s 0
(2.29)

Thus, the index A,B etc. of Chapter 2 runs in the range

1...10. Because of the peculiar structure of P we use a * 1...4
and the pairs (a,b) with a > b, instead of A. In this way,
the differential form of (3.1), corresponding to (2.9) reads,

dZ - d=0 ♦ EdO
(3.2)
dz » d=t ♦ dx ♦ Edt

The forms u are obtained as in (2.11) by setting d=, dx - 0

in (3.2) and then carrying out the replacements (0,t) ♦ (E,x)
and (H,x) •* (S.x)*1. The result is

SI - H'1dH
(3.3)
o « E'^dx

n is thus a skew-symmetric matrix with elements u8**

 13 14

Componentwise Me have,

(3.7)
uab . =cad=cb _ca 3
3xc
(3.4)
Clearly
oa - =cadxc
t!(Da) - o,(Dab) - 0
The Cartan-Maurer equations (2.13) read:
(3.8)
d<1>ab ♦ uac . <ocb - 0
(3.S) "fC V ■ «fa
doa ♦ u»ac » oC » 0

For the R.I. vector fields we find,

Similarly, the L.I. forms are (putting do, dt » 0, and replac­
ing (e,t) by (H.x)'1) a 3 b _ -af 3 _bf 3
x — r~ - x
ab sir 3x
(3.9)
IT ■ dss 9 p * dx - TTX (3.6)
3x '

Tangent vectors appear as operators, linear in — and A general field can be considered as an element of the
i|i ( x )
35 3x M
and tangent to the defining variety of SO(4), with the PoincarS group, with its x dependence mentioned explicitly,
normalization, and its dependence on 2 implicit and fixing its "intrinsic"
spin (except that we have to replace the Poincarfi group P by
=ab-cb .ac
its covering group?)^ ^ the usual convention (3.1) in de­
fining the PoincarS group product that picks left-translations
A convenient set of L.I. vectors is: (i.e., S^) for coordinates and fields: (A,a)(E,x) »
(A=,Ax ♦ a) which e.g. for = = 0, x = xu and an infinitesimal
P transformation A = 1 ♦ g^uv^, a » av yields (O.x^) *
(0,xu * e[uvlxv * au). Had we used right translation, we would
 15 16

have found (0,xp) - (0.x11). The entire picture is This product is associative provided:
thus R.I., and it is natural that the Hilbert space considera­
tions should lead to the Sg) set of (3.9), with "total" u'1CH)Tau(=) - Sabyb (3.11)
angular momentum Sab including both that linear action on
field components ("spin") represented by the =faf — part, or in brief, u ‘1 (H)yu(H) « Ey
fa 3
and a complementary "orbital” piece x l — rr .
3x
On the other hand, the procedure for "gauging” a symmetry Upon differentiation of (3.10) we get:
implies the use of (coordinate-invariant) forms, which thus
4
have to be the L.I. forms w (and their orthonormal L.I. dZ « dE0 ♦ Ed0
algebra 0A). The L.I. set fDab»Da) is naturally
factorized (up to a Lorentz-transformation on the D#) and dz * dEt ♦ dx ♦ jd?YU(E)e ♦ |£ydu(E)e
(3.12)
will thus appear in the infinitesimal treatment of gauging, ♦ Edt ♦ jiyufEjde
with only Dab giving rise to a gauge-invariance of the Lagrangian.
The action of Dfl merges with general-coordinate transformations dc * du(E)8 + d£ ♦ u(E)de
on x space, and a modified 6a can indeed be identified with
the latter.10 Repeating the procedure leading to u we obtain
The formalism can be extended to supergroups. A most
interesting example is provided by the Graded Poincartf Group S3 - E dE
(GP) ,11)12)13) Consider first a Grassmann algebra A « A + © A "
where A+, A' are respectively the even and odd terms in A. We o * E^fdx - jSydC) , (3.13)
can then form a set GP of triples x ■ (H,x,E) with xa e A+,
5° € A", E e SO(4). Ke consider x as an SO(4) vector, £ as a
Majorana spinor. We define a product:
The corresponding R.I. forms are
z • (Z,z,C) - xt - (HO.Ht ♦ x ♦ ^ -CHJ.3 , u (E)e ♦ 5) (3.10)

with t « ( e . t . e )
 17

H - dEE'1 GP thus becomes a generalization of the Euclidean group P on

/ ? 4. Here the metric
p « dx - dSS’1* - |d|T5 - J|Yitaboabe (3.14)
ds2 - dxa ® dxa » oa ® oa (3.17)
_ _ j, l a b abr
if * d £ • 5* Of €
is obviously left invariant on P and therefore P-invariant on

In deriving these results we used the identities (see App.) A*. Similarly, we have o n X ? ^ 4 , using (3.13)

|*abaab - du(H)u’1(E) ds2 • oa * oa + kw x in (3.18)

(S.15)
|<Dab(jab - u'1(S)du(=) which is also GP-invariant. Notice however that on odd super­
forms the product7^ „(<3) is commutative (anticonunutative) and

Supersymmetric theories are frequently constructed15^ on even superforms „(®) is anticommutative (commutative) accord­
in "superspace" ft*?*, Elements of form equivalence ing to the usual convention (2.3). In this way we retrieve the
classes of GP under right-multiplication by S0(4) elements. well-known metric*7'1®^ of the superspace "vacuum." Also of
Therefore, great interest are the tangent vectors. The L.I. tangent vec­
tors are given by:
(S.x.C) = (SS.x.e)

and # 4 / 4 is conveniently parametrized by (x,C) only. CP

multiplication on the left by (©,t,9) therefore acts naturally
^ 4 / 4 and we have:
on

x ’ » 0x + t + 9y u (0)C
(3.16)

I' - u(6)5 + 9
 19 20

The R.I. tangent vectors are: 4. Curvature and covariant derivatives

a 3 b 3 „af 9 _bf 3 — ab 3 Let C be endowed with a set pA of 1-forms (a "c-bein" or

ab * — ft ’ x — » * = ' - -rsr * 5° r=
3x 3x "fundamental forms") generalizing ojA , but not subjected to the
A
3 Maurer-Cartan equation. We define curvature R as:
(3.20)
3xa
ra = dpA - 2<:*b e p B - p E f4-1)

Hence, wA has vanishing curvature. The covariant derivative

A
of a set of p-forms n is given by*
By neglecting the terms in SA (spin) we obtain the infini­
tesimal generators of motions onX?*^*. The D components
(Dn)A « dnA - C^•BE
b c p B - nE (4.2a)
are related to the "covariant derivatives” of Salam and
Strathdee.2) The Maurer-Cartan equations appear as:
It is convenient to introduce a similar definition of a co­

, ab ♦ uac A (itcb 8 0 variant derivative of a "contravariant" set nA ,

du

(Dn)A * drtA + cEBApB ^ "H (4.2b)

(3.21)

A A
, + 1 aboab ~ (u * 0
. If we vary p in (4.1) by the infinitesimal 6p , we find,
dm

«RA = (D«p)A (4.3)

In deriving (3.19) to (3.21), care must be taken to keep a
consistent ordering of Fermi variables in order to avoid con
Ke also have
fusing results.
(D2n)A « (D(Dn))A * -C^b e RB « n E (4.4)

(DR)A « 0 (Bianchi identities) (4.5)

d ( A A) « (Dn)A - CA + (-l)pnA ~ (D5)A (4.6)

•We write D ® for G only when the nature of the group has to
be explicitly stated.
 22
21
Then

We next consider two copies of G, (G ;G ) with coordinates

X y (Dr(A)n)A ■ (Dri)Bad(A‘1{x))BA
(x;y), respectively, and a map A: G ! . Let in general
(4.10b)
“x ,px etc* (“y »Py» etc.) be forms on Gx (Gy), respectively. (Dr(A)t)A » (DC)Bcd(X'1(x))®A “ ad(X(x))A(D5)B
To any such map A Me define a gauge transformation T(A) as

follows: (see (2.19)) Equations (4.8), (4.10) display the covariant nature of D and

curvature.
r(X)pA - X*uA ♦ P ^ d C X ' ^ x ) ) ^ (4.7) Let H e G be a subgroup of G of dimension c'. Let M « G/ii
be the corresponding homogeneous manifold. In the sequel, wc

(Recall that the dual map X* is defined in (2.6) and ad(a) in often refer to the case G * P, H * S0(4), M = ^ , or also

(2.16).) G - GP, H « S0(4), M ’ ^ 4/4-

It is often convenient to rewrite (4.7) for infinitesimal G can be considered as a principal l-undle with fiber H,

gauge transformations. In this case, we suppose that A(x) is base space M. In the above quoted case, this bundle is trivial,
given by AM (x), where XM are infinitesimals. Using (2.11) and but already for G = SO(3), H = SO(2), M = one finds a non­

(2.20) we then find that the infinitesimal change is given by trivial Hopf fibering, used in dealing with Dirac monopoles.

«pA • DXA (4.8) Similar situations will arise in Gravity and Supergravity, and

where the r.h.s. is defined by (4.2a). it is thus better to use from the start the general formal

Using (4.1) and (2.20) one verifies that the curvature machinery developed for non-trivial bundles. In the following

T(X)r £ calculated from f(X)pA is given by: we treat the conventional case of Bose-type manifolds and refer
the reader to references 6-7 for the (l:ermi) Graded case, al­
though we shall later develop in detail GP, with a graded mani­
T(X)R* = R5ad(X‘1(x))jA (4.9)
fold.

Similarly, taking Let Ua be a covering of M, p the projection G 5 M.

p"*(Ua) is parametrized by elements (y,qn ,<0, where y € M,

r ( X ) nA « nBad(A‘ 1( x ) ) j A qa c ii. h'e give transition functions ^gCy) 6 H on Ua A ug ,

such that on Ua O Ug O U^, *ag<'BY * then identify
' (4.10a)
r(X)?A * " ad(X(x))A B cB
elements,
(y ,q i ) = (v,qg ,61
 23a
23
where q^ - and y e Uq D Ug. In this way, all sets
A
p *(UQ) are "glued" together into G. On Ua we have a map xa! for a e H. Correspondingly we split the set of forms w into

sets ujj. Up such that,

p'^v ~ - 3- ~ G

Xa(y.qo.a) - qa uH (#J - 0 . - 0

Given the L.I. form <jA on G, xa defines a form on P*^(^a).

These sets are acted upon separately by the two representa­
For each a we suppose given the form,
tions in which ad(a) splits. Clearly we have also

PAa * (4-U )
Mf,&] c &
where tB is a 1-form on M. If the set pA has to define a
a o If furthermore M is a symmetric manifold.
A
unique form o on G independent of a, we must have the matching
If weak-reducibility holds, the set of forms p^, with equations
conditions,
(4.11)-(4.12) satisfies the conditions for a Cartan connection

on G as a principle bundle. In this case the H-components of

■ ♦§,/ * ’% ad(O s A (4*12)
A ra represent proper curvature, and the F-components are known
If (4.12) holds, p is called a factOTized c-bein on the pair
as the torsion. If [ - 0, then the curvature of pH as a
(G,H). Let $,Jjfbe the Lie algebras of G,H respectively. We
Cartan connection coincides with the G-curvature (4.1). All
take a vector subspace >f of *§ such that % -SC* $f,
these conditions are met when G ■ P, H * S0(4), M ■ , which
n Each element a € H induces a coadjoint transforma­
can be identified as a structure with ordinary Riemannian
tion on which leaves invariant,
geometry. In this case then (4.11) reduces to

cd(a)J^cJ(f ab ,_-l._.ab . cd_db_ca

p = (= d=) + T = s

(4.13)
If however t? is such that cd(a)^ C */( as veil, we say a _ca c
p * 2 T
that M * G/H is a weakly reductive homogeneous manifold. This _ab _ „db_ca „cd
Clearly, then R * = = k
happens in particular if H isr compact, semi-simple and con­
(4.14)
nected, or even discrete. If weak-reducibility holds, cd(a)

and ad(a) each reduce completely into two representations of H

where 24 25

/ C Ci - dTCd ♦ TCi* . Tad

than to a factor G/H, we should certainly move in that direc­
(4.15)
tion.
A C - dTC ♦ TCa . Ta
However, present theories with torsion do seem to have
As we saw in (4.9),
some drawback in this respect, since torsion does not propa­
ra ad(q(;1) -A (4 9a)
gate. While ordinary curvature does not vanish outside of
In a Riemannian manifold, the condition /?a « 0 is imposed, matter, torsion does.22^ This disparity probably has its
and the metric tensor is given by roots in the factorization hypothesis, which breaks from the
very beginning the symmetry of the Poincartf .croup, since
gjivdyW ® dyV “ £ T° ® tc (4.16) it factorizes out in a trivial way the Lorentz group dependence.
(We have used S0(4) in our calculation for simplicity, but all
t c is then identified with the conventional vierbein, and results obviously carry for S0(l,3) with minor modifications.)
& c * 0 allows one to calculate xca from t* and dta through If we do not assume factorization, the dependence of the c-bein
a procedure which is essentially the known formula for the on the Lorentz variables must be dictated by the field equa­
Christoffel symbols in terms of the metric tensor. Once tab tions and boundary conditions. Chapter S contains a discussion
is known, the curvature * ab calculated from (4.15) yields the on how to attempt this task and on the difficulties encountered.
conventional Riemann tensor. We now review the Poincard gauge transformations, before
Some generalized forms of conventional relativity have and after factorization. Eq. (4.7) is our fundamental start­
been proposed, where torsion does not necessarily vanish, ing point, since it realizes a transformation in which the
following the original ideas--long forgotten--of E. Cartan.19^ 20'2, "y" copy of the group (the transformation parameters) are func­
These proposals, although far from any physical application22^ tions y = *(x) of the "x" copy variables (now used as the back­
are nevertheless intriguing in that they Temove in part the ground manifold, an extension of space-time). For the P-gauge
disparity of treatment between "internal" and "external” com- funct ions
A oA
ponents of the p , si multiplets and deal effectively with a
Poincarfi curvature rather than with a Lorentz curvature only, U: Uab(E,xl, tta(E,x)
as used by Einstein. To the extent that we might be tempted
to give a more fundamental meaning to the group G itself rather
 27
26

Notice the similarity with (4.13), the factorization equation.

we find the transformations,
The action of an infinitesimal eab is the same as in (4.18),

(o’)ab « (r(U)p)ab = (U*1)acdUcb + (U‘1)acpcdUdb «Tab = d£ab + TadEdb - EaCTCb = J)(H)eab

(4.17)
(4.20)
(p’)a - (r(U)p)a - ( U - V c{pc + duc ♦ pcdud)

h’e thus do not appear to "gauge the translations," i.e., the

For infinitesimal transformations, ea! » Uab - 4ab, ea « ua , we
ua and ea gauge functions are not used.
get,
A gauge theory expressed in terms of forms is automatically

. ac cb ac cb _ n ab covariant, that is, invariant under the Group of General

op ■ dc ♦ p e - C D * De
Coordinate Transformations (GCTG). An interesting further
(4.18)
adaptation was developed originally by Hehl et al.22^ and by
«pa - dea ♦ paccc - EacpC ■ Dca
von der Heyde,10^ in the pursuance of the program begun by
Utiyama, Sciama, Kibble9^ and aiming at a derivation of General
Note that D is here the Poincare covariant derivative,
Relativity from gauge-like principles. The new equations are
and includes the cac action for pa. Holonomically, it entails
useful in connecting the GCTG action to that of the gauge group
derivatives and summations over all ten variables and
(H C. G in our above treatment). An infinitesimal element of
differentials. However, we show in (5.12) that the full M
GCTG is given by a tangent vector e on G. We then have,
Poincare gauge is too strong as a symmetry of the Einstein
Ldgrangian. Instead, factorization of the Eab variables ac­
a*M - cM (4.21)
cording to (4.11)-(4.K) reduces the gauge gToup to H = SO(4)
and
SpM = 6(<3xM om B)
(T*)ab - (U'1)acdUcb ♦ (U'1)aCTCdUdb
(4.19) B
3p,
dx N StM 'B . j„M„N opM
(t -)3 * U f l)aCTC 3x

M 3pMB 1 (4.22)
dX
TFl
 28
29

where n,m are the gradings of the indices N, M for the case of
a graded manifold, and where we have defined an Anholonoraic The commutators of the Dg no longer close to form the Lie

coordinate variation algebra Jtf of G. In order to identify the necessary modifica­

tions, we start from the definition of curvature (4.1) in
A
eB - eMpjjB (4.23) terms of the anholonomic components of R :

The last line can be rewritten as

RA - lpB „pERB -A (4.28)

«pB - deB - 2(e,dpB) (4.24) thus yielding

since dpA - | pB .pE(CABE ♦ RBEA) = 0 (4.29)

This entails the (Graded Lie bracket) commutation relations,

D
and the scalar product (e,dp ) is a 1-forra obtained by con- f V V - (C^BE * RBEA>'DA C4-30)
M M B
trading e with the second factor dx in dp . Using the
B
G-covariant derivative De as in (4.2a) we find, In case of factorization under a subgroup H C G, some important
simplifications occur. In particular, I>A ■ for A « {#},
«pB - DeB + CBDEpD . eNp^E - 2(e,dpB)
where { ) denotes the range. Indeed, for F <

- DeB + (e, - 2dpB ♦ C®d e p D . pE)

CF(Da) « uF(DA) * 0

6pB - De B - 2(e ,RB) » De® - pEeFREpB (4.26)

because t B in (4.11) contains only "horizontal" (i.e. M * G/H)
We T'efer to (4.26) as the Anholonomized General Coordinate
differentials, orthogonal to Dft. Similarly, for
Transformation (AGCT). Whenever RB « 0, the AGCT coincides
with an ordinary gauge transformation (4.8).
p B(Da) = u B(Da) - sba
Yet a third way of introducing Sector fields on G consist
in using the natural extension Dg of the set Dg given in
Therefore, quite trivially, the commutation relations of the
(2.27). We define the 6g as in (2.23-24),
Da , A £ (A^) are the same as those of DA> In addition, the

PA (6b ) = (4.27)
 31
30

commutator [DB»6 E 1 for B € ( ^ ) is equal to the "flat" case Dc

c - ad(q)'ET*w ^I-i- - 12 A €2 W t„AS,)
Al * C u
m * C e {& ]
for all E, since RjjA » 0 whenever {itO'f’E. Indeed, fol­ (4.34)
lowing (4.9a), RA has no vertical components since is Dc - Dc C 6 {#}
constructed on M > G/H only.
The factorized c-bein for the pair (G,H) can be written where is the R.I. vector field on H. The quantity in the
(4.11) as brackets corresponds to the conventional covariant derivative
3(H)
li
pA - <uA ♦ tBad(q'1)3A A €(/),-* satisfies (4.27)
J (4.31)
PA - TBau(q"1)gA A € {J) P A (Db ) - 6 A b V A,B

•* M
where for simplicity, we neglect the index a and we write We now apply the vector fields DA to x , as a specific
for Xj“A . As stated, for Afi{>^). We now compute realization of (4.21). This corresponds to
explicitly Dp for F €{3**}. We use coordinates zu on M * G/H.
This being the case, if •M . T-M "B . ,-B
A A * A 6A

T -lJ *B , (using 4.23). Applying these fields to (4.26), we obtain

TC Tu C
(4.52)
Tc - .d(q)‘V «a p B • -p DCc % a * r da B> (4-3S)

so that which generalizes the global group transformation, and exhibits

pA lTc) - «?c A ’C (4.33) the role of the &A in generating when "gauged" ordinary gauge
transformations (4.8) for \ € {At) and AGCT for A € ( ? ) as in
This suggests the form (4.26). Note however that the DA do not close to form a Graded
Lie Algebra in a strict sense, unless A € iM)-
 32*
32

Specifying to P, we have the AGCT (4.39)

Dab ’ Dab

6pab - Deab - 2(e,Rab) ba ■ EbV (8v - \ TCd(x)SW )

(4.36)
«pa - Dea - 2(e,Ra)
} S(H) . =cf 3 . =d£ 3 (4.40)
cd -

which should be compared with (4.18).

« a■
However, as explained above, (4.2a) ensures that in the
factorized case as given by (4.13) we have with the commutators

(4.41a)
R(cd)(e’f)tab) “ R(cd)(ef)a ' R(cd)e(ab) " R(cd)e* * 0 (4-37) lDab'Dcd] “ 6adDbc * 4bcDad ' 4acDbd ' 5bdDac

so that (D is still D ^ ) , (4.41b)

fDab-5d ’ «bc5a ' 6ac6b

«pab - Deab - PCedR;;ab (4.38a) (4.41c)

i v v - Rabcd°cd * r;;c“c

6p - Dea - 2(e,R ) (4.38b) For GP we have for (4.26) before factorization,

so that eab of S0(4) produces an ordinary gauge, whereas the Spab - deab ♦ pa t .ctb - ptb .eat - 2(£,Rab)
translations ea are indeed replaced by an Anholonomized coordi­
nate transformation gauge. The corresponding Lie algebra Dcab - 2(e ,Rab) (4.42a)
generators Dgj, and vector fields Da in the bundle of tangent
frames are, Spa - dea ♦ pat . eT - p* . cat ♦ PYae - 2(E,Ra)

» Dca - 2(E,Ra) (4.42b)

 32c
32b

6p « de ♦ j(paboab)-c - joabp * e ab - 2(e,R) One last remark about factorization: had we chosen
A A
H * G, then factorization means trivially p ■ w , that is,

* De - 2(e,R) (4.42c) empty space and a global symmetry only. If instead H ■ e,

factorization imposes no conditions on p \ All D^ commutators
generate curvature terms, which is another way of visualizing
and after factorization of S0(4) (with D ■ D^GP^),
the difficulties of providing for an invariant action, as
discussed in Chapter S.
«p,b - De*b - PCedR*jab - PCea Rca(ab (4.43a)

«p* - Dea - 2(c,Ra) (4.43b)

«p - De - 2(c,R) (4.43c)

Aside froa (4.41a), (4.41b), we have for the most general

factorized case the commutators

[Dtb.DJ - (o,b)D (4.44a)

I V f y " “. ^ ^ c d * R.;C“c + *.b5 ^ - 44b)

W V V ” RiaCdDcd * R« C»c * C Dr (4‘44c)

(note that though B in D^ is an "upper-spinor" index as in tB,

and has to be lowered by Cag, we write it as a subindex like a
or (ab)).

< V V “ (A B5a * RaBCdDcd ♦ RaBCfic * K ? hy (4‘44d)

 33 34

5. Lagrangians and Field Equations IIIb) In addition to Ilia), we request factorization of the
c-beins (reducing the "objects" to fields on the G/H
It is often argued in much o£ the recent work on unified
variables) and vanishing of torsion in the G/H manifold.
theories that the final goal should be a purely "geometrical"
Alternatively, factorization may appear as the result of Ilia.
theory. The definition of "geometrical" is at the moment not
All theories of current interest meet I. Classical Gen­
clearly stated, and correspondingly there is room for disagree­
eral Relativity satisfies I, Ila, Illb and does not involve
ment as to whether many of the proposed theories really fit
lib. From the results of Ch. 4, conditions I, Illb are ob­
the criteria for geometry. In an attempt to narrow the dis­
vious. Ila follows from the (Trautman) action,
cussion we introduce in the sequel a number of interesting
geometric criteria.

6A ‘ "c * 'h * * ‘ T° ‘ ,4 ‘ab cd csa>

which coincides with Einstein's Lagrangian. (Again, we denote

I) The "objects" or fields in the theory are c-beins pA on a
anholonomic indices by Latin lower case indices, using lower
convenient Lie Group (or Supergroup). Note that this
case Greek indices for the space variables.) Variation of
generally implies a self-sourced theory with no explicit flb *
(5.1) proceeds as follows: We vary first * , thus obtaining
"matter" fields; alternatively, the geometric criteria
characterize only the "generalized vacuum" limit of the
©A - f 6#ab . xC „ Td tabcd = / D ( « t ) ab . Tc . Td Eabcd (5.2)
theory.
-ve4
Ila) A Lagrangian density is formed from the c-beins by using where D(6r)ab can be interpreted either as a G-covariant (i.e.,
the operations D and ~ only, plus contraction over indices. Poincare P-covariant) or as an H-(“ Euclidean "Lorentz" S0(4))-
covariant differentiation (the latter is the "anholonomic"
lib) (Enlarged definition) in addition to Ila), we allow the
conventional one in General R e l a t i v i t y . I n the last
use of the dual (*) of a form.
event, integration by parts, and the antisymmetry in (a ♦-♦b)

Ilia) We consider a preferred subgroup H C G and construct a yields the equation:

theory which is gauge-invariant under H, but not under the
full gauged G. Jsvab - Dtc « A - abcd » 0 (S.3)
 35 36

Now by definition, the H-covariant derivative D tc is just Setting dxp * dxv A dxA - euvAadva M.v.X.a = 1...4
the torsion f t c . It follows,
a ab . t8 Tb g P°
/\ • *p - a A • *jiv
J*6tab « /{c * Tdeabcd - 0 (5.4)

(5.6) can be written as

Since 6tab is arbitrary, we get

/ ? P°..,.Ta*q tb«TflEahfrdxM
x*.«yv *a ** abfc - dxV ' dxA “ 0
/ t C « T**eabcd “ 0 Va.b
using
(5.5)
HC . Td - ft* A TC a b £ *6
T.pT .oT.AEabfc ‘ TC ppoA
*

where A - /?
Now (S.5) implies / { c * 0. Indeed, setting
*d
and deleting the factors dva , t c we get,

/tfC * “?abTa - rb • "ith u^ab '

a*° en 1c,,vXa - 0 (5.7)
A •"UV BpoA v
and defining ucab » eefabufef, (5.S) implies
or

ucab » uacb , but also ucab « -i;cba

* 0 <s-8>

A few manipulations ShM that ucab = 0 » u^ab> and therefore which are Einstein's equations for empty space.
/fc - Rc - 0. This method is clearly equivalent to Palatini's in that
The variation in ta then yields the vanishing of the torsion follows from the variational
principle (and of course from factorization).

( * “b - T^ eabfc " 0 V*= (5.6) Tentatively, it would appear possible to assume (5.1)
with nonfactorized c-beins (and with Rab instead of ^ ab).
One is then faced with two options, neither of which appears
entirely satisfactory.
 37 38 •

Since there is no factorization, there is no priviledged Ra * pb - Rb « pa “ 0

role of S0(4) (i.e., H) in P, and it would appear desirable (5.11)
to extend the integration to the entire group space (G) rather
R 3 b * e C e a b c d “ 0

than to G/H only. The action then takes the form:

resembling the factorized expressions (5.S)-(5.6). Here,though lack
• A - I Rab * pc „ Pdeabcd . v (5.9) of factorization inhibits direct deduction, heuristic arguments do
■'P
yield Ra = 0- Interestingly enough, (5.11) which follows
where v is a 6-form including the dH differentials (at least from the variation of the fields pab, pa also suffices to
in the factorizable case). None of the obvious forms tried guarantee that the variation in^l4 vanish trivially. In fact,
so far for v seems to reproduce Einstein's equations. A second the differential of the Lagrangian density
option is to consider
d,
p abed

8 A “ l?»Rab * °C “ P<leabcd (5'10)

where m * is a submanifold of G or rather a section over G, vanishes as a consequence of (5.11).

when this is taken as a bundle. This seems to be unnatural, This fact is not so strange. Any variation of Pi* can be

for was in the original intention the base space G/H. compensated by a change of coordinates in G; the embedding
Furthermore, the very presence o f # 4 in the variational functions are thus unchanged, whereas pa , pab do change.
principle makes it a dynamical variable, thus also subject This of course is taken care of by the field equations (5.11).
to variation. This implies the added complication that the The option (5.10) thus appears to contain an automatic answer
equations describing the embedding of/^f4 in G should include to the threat of "degeometrization" to which we alluded.
arbitrary functions which must be considered as fields, in Returning to the further interpretation of the equations (5.11)
which case we would have strayed out of our conditions (I) of the unfactorized set, we remind ourselves that the factorized
for a "geometric" theory.
forms T a b , ta constructed through the gauges (4.11) satisfied Ra * 0
If, however, we decide to overlook these difficulties
and equations (5.5). They thus also satisfy (5.11), since p ,o ,
and try (5.10), we find field equations,
Ra , Rab can be reconstructed for the factorized case through (4.13) -

(4.14). Factorization thus supplies a solution for the unfactorized

 38 b

38 c

field equations (S.11). We can now make the converse conjecture, The form (5.1) of the action brings us to another argu­
namely that equations (5.11) uniquely yield the solution p A <- t A via ment. Clearly, (5.1) is invariant under SO(4) (i.e., H) gauge

(4.11), apart from a generic coordinate transformation on G. If transformations, but not under proper local Poincare gauge

this be true, factorization would be a result of (5.11). This would transformations (that is including translations). It is
instructive to see how this happens. By specializing to a
be consistent with our previous deduction from heuristic arguments
translation-gauge transformation as it appears in (4.IS), we find,
that (S.11) implies Ra • 0, just as can be directly deduced from

factorization. The same would now be true of (5.8), i.e. the

Einstein equation. The theory, as extended here to forms on all

of G, but with the action integral restricted to an arbitrary 4-dimen­

sional submanifold, would still have the same physical content as

Einstein's theory. T(X)pab = pab

Actually, since the choice ? K 4 has turned out to be ir­ (5.12)

relevant, we can choose 3 / I 4, i.e., identify the submani­ T(A)pa - dua + p* + pabub « pa ♦ DU8
fold with "space-time." This then sets our Lagrangian (5.10)
to be equal to the factorized (5.1), as a boundary condition:

where ua is the local translation and D is the SO(4) covariant

PA ( * 4) - tA
lerivative. Therefore, Rab remains unchanged but pa does get
nodified. The change in Ra can be computed directly
with pA (G) derivable from pA (/^4) through a transformation
which is just the inverse of our factorization gauge (4.11).
Conceivably, the unfactorized (5.10) may lead to new solutions
(with torsion) in the presence of matter fields. However,
in the present case, extending the forms to the entire P manifold,

could be legarded just as a way of writing down simultaneously all

gauge-related solutions (with appropriate coordinate transformations

automatically completing the procedure).

 39
40

MacDowell and Mansouri have recently23^ proposed a curious

alternative to the customary Einstein Lagrangian. Their basic
approach is to consider a G = SO(5), H * S0(4) theory (actually

T(X)R* - DT(X)pa • Dpa + D2us - Ra + Rabub (S.13) S0(3.2) and S0(3.1) respectively in the physical case) and a
factor space S4 - G/H. One therefore has curvatures,

Thus, a torsionless c-bein acquires torsion under the

Rab - dpab ♦ pac . pcb - -Rba (5.14)
action of a translation gauge. To no one's surprise, both the
with a,b * 1...5
Lagrangian and the field equation break the translational gauge
In order to compare this formalism with the previous one,
symmetry. We have seen in (4.24) how the invariance under gen­
eral coordinate transformations (GCTG) in G/H can be represented we split the range by calling p8^ « epa, where e is a constant.

as »»" A S C T " translation gauge. Such a gauge corresponds to We then have

a translation generator obeying a new algebra with curvatures

as structure-functions, instead of structure constants. The Rab - dpab ♦ pac „ Pc b - e2pa „ Pb

a,b = 1...4 (5.15)

Lagrangian obeys that AGCT gauge-invariance, instead of
„a _ , a ac c
(S.12). However, the curvatures themselves in (4.24)-(4.30) are R « dp ♦ p <. p

determined by the Lagrangian through the equations of motion.

What is then the criterion by which we select a subgroup In the e ■* 0 limit, (5.15) reduces to (4.15). This limit cor­

H (in this case S0(4)), whose symmetry is not broken, out of responds to the Inonu-Wigner contraction of S0(5) into P. In

the whole group G? Clearly, some part of the symmetry must ref.23^, however, the action integral is chosen just to be:

be broken. A Lagrangian built out of curvatures only, and

invariant under G is by definition a characteristic form and
-*J* a,b = 1...4 (5.16)
yields only topological information, but no field equation.
(r « R ^ S b c d
'-L J»ab
2 e .
E J

Applying (5.15) we get,

 41 42

C*BG(E>C®DE(E> + C?Rn(c)C?Pr(E) ♦ C*UK(c)C?rJ,(c ) - 0 (5.18)

- j»r/«“ - '‘Sbcd * - »\*««
(5.17)
- e2Jpa „ pb . pc „ Pd ea b c d - Aj ♦ A2 ♦ A3 Differentiating, we find
where
E*1’ - dpab ♦ Pac „ Pcb - le— C®DE(e) * C*BG<eJ ~ °T + cycl- * 0 I5-19)

is the "Poincarfi" curvature. Now A^ is a topological invariant Now the 2-form

and yields no contribution to the field equation. A2 is the
B G
classical Einstein Lagrangian, Aj is a cosmological term. In 3e U> A u>
e*0
the limit e ♦ 0, one retrieves the conventional theory, and c 1
A A
is essentially the radius of the de Sitter space, i.e., the has vanishing covariant derivative at c * 0 (with P « « )

vacuum solution when cosmological terms are present. as a consequence of (5.19). We have indeed,
In a way, the MacDowell-Mansouri formulation is more
aesthetic in that only curvature components are used, and the (Dp)A « duA - c* uB * UC * \ C®deo.D . uE . u
search for a Lagrangian is narrowed down to the choice of a
jC^ ®C, ED E
1 *BG r G B D
suitable S0(5)-breaking term. • 2 - f e - *DE “ <l) /\ i •BG ie-
A different way of looking at the issue is the following.
(i)D /s a)E „ uG • 0
aC-BG -B -A sl-DE A
Can we anticipate the form of a Poincarfi-breaking Lagrangian ie ‘DE *BG it
(5.20)
that can be derived by contraction from an Ansatz involving
only curvature terms? In order to answer this question it We call any form nA with vanishing covariant derivative
pays off to look at the procedure of contraction of a group * a "pseudo-curvature” form. Clearly
in terms of the Cartan-Maurer equations. Contraction implies
that one can choose the structure constants cfgE (e) as func­ cA • ^ bd *8 - “D (! . 21 )
tions of a parameter e, and that the resulting groups G£ are
all isomorphic except at e » 0, where the contracted group is a pseudo-curvature corresponding to a trivial contraction
arises. We therefore have the Jacobi identities: of G where CAB„(e) - f(e)CABD, and f(0) - 1. Therefore, nA is
 43 44

remlly an equivalence class modulo cA (see the comnent at the (Dv)ab - dvab *• w b t . vat - Uat . vbt ♦ J 5 . va
end of this chapter). These classes already appear in the (S.22b)

Chevalley cohomology on Lie Algebras.245 The existence of a (Dv) - dv + uat * vt

non-trivial class is related to a non-trivial contraction pro­
cedure. Indeed for P we have the non-trivial class uab » &>a ,><i>b , If we find either p or vA as a polynomial in u , we
P* “ 0, which corresponds to the contraction of S0(5) into P. can replace wA by pA in this polynomial. In this case, it is
On the contrary, SO(S) itself has no such class. no longer true that (Du)A » 0, ( D v ) a « 0 but rather those
£
Thus, on SO(S) we have the MacDowell-Mansouri procedure derivatives become linear expressions in the curvature R .
but no pseudo-curvature, while on P we have a pseudo-curvature This provides an interesting argument when building Lagrangian
but the MacDovell-Mansouri procedure fails: Ra^RC(^eabcd is densities. If we have (as in Relativity) an action of the
not only Lorentz-invariant, but Poincartf-(gauge)invariant as kind
well, and transforms away into a surface integral. The presence
of pseudo-curvature is therefore a warning that it is possible A -JVvA (S. 23a)
A
to construct a Lagrangian density directly with i> without
"decontracting" first. Alternatively, pA is the infinitesimal then variation in pA gives a contribution of the kind (sym­
difference between contracted and decontracted definitions of bolically)
curvature. We should also note that we are interested more gen-
A --
erally in all forms u or vA possessing vanishing covariant SA » /{(D< P ) \ + RA «PB} = 0
derivatives.
On P we then have, M Integrating the first term by parts we obtain,

.ab , ab at tb tb at -JspA(Dv)a * J«pARB = 0

(Du) « dp ♦ « - V - « a u
(5.22a)

a
-(Dv)A 4 R — J * 0 (5.23b)
and ip
 46

A A
Equation (5.23) is identically satisfied for p * w , are also pseudo-curvatures. This allows the construction of
p A A A
since R = 0 and (Dv)A is linear in R . Therefore, p » u an algebra of pseudo-curvatures,of interest in tryi«e to define
can be assumed as "empty space." a suitable Lagrangian formalism.
The above discussion would lead us to the conclusion that One final comment with respect to the possible use of the
if G is semi-simple, then the MacDowell-Mansouri procedure trivial-contraction cA ps-c'irvature (S.21). To apply it to an
seems to be ideally suited for the construction of a Lagrangian. action such as (S.23a), we first have to lower the index
If on the other hand G is the result of a contraction, it has
in a natural way a pseudo-curvature form vA (we can take
?a " Sa b '8
vab “ eabcdpC * p^ ’ va “ which can be used in forming a
Lagrangian. The G-symmetry in this Lagrangian is broken be­ where we have used the group metric (in a semisimple group, its
cause pseudo-curvature "remembers" which subgroup H in G is inverse gAB also exists). If we now try the action
not affected by contraction (e.g., SO(4), in going from SO(S)
to P). The way in which this breaking occurs is dictated by
the structure of the group G itself, rather than by some ad-hoc
/'
»
choice of H. It should be emphasized that only the 2-form p , we discover that this is again a topological invariant, and
with (Dp)A -0 can be related directly to the contraction procedure. yields no equation of motion, for either semisimple of contracted
*
A generic ps-form (ps--pseudo-curvature) with (DC) or (Dv)A “ 0 G. This is because 5A carries no information about a specific
is still useful in building Lagrangians but its geometrical in­ H C G, no symmetry breaking.
terpretation may be less straightforward, involving the use of For our further discussion of supergravity, we include
A
various group invariants besides the group metric. here the GP invariant derivatives of multiplets U and
A A
Finally, we remark that if p , c , vA , etc., are pseudo­
curvatures, then (Dp)ab - dpab ♦ pa t . p tb - Pt b .Uat

(Dy) a * dpa ♦ Pat * Pl - P* - Pat ♦ PYaP (5.25)

Dp - dp ♦ |(paV b) . P - ^oabP - Pab

 47 48

I
I
( D O ab ■ dC.b ’ pta - H b * Ptb * 5ta ♦ |pT(aab)T(CC) > 6. Supersynunetric Theories and the Factorization nf
Subgroups ' ------ — — -------------

A consistent theory of supergravity has been formulated"

CD«a ■ d‘a * Pt a ^ b (S.26)
in a second -quantized formalism, by finding a set of local
graded infinitesimal transformations on a set of fields in­
Dc - dt ♦PYaCa - |pabccrab J
cluding gravitation, and leaving a Lagrangian invariant'(after
imposing a constraint equivalent to the vanishing of a general­
ized torsion). Various attempts have been made to findla geo­
metrical formulation: a presentation of the theory in the
context of a Palatini-treatment of Einstein's Relativity,"®^7^
’SI
an attempt to derive it as a contraction-limit in superspace,'
a partial description using a superspace with constant tor­
sion in its physical vacuum16^ (see also ref.7^’15^), and a
generalization of the MacDowell-Mansouri treatment of gravity.
Other works have pursued the program for a generalized supcr-
symmetry group (including internal degrees of freedom) and
for the super-conforaal group.
In this article, we are interested in the supergroup CP
and associated subgroups, i.e., the problem of picking II C G.
A) The first candidate is H * S0(4). In this case,
GP/S0(4) where is the set of pairs (x,£) in a
Grassmann algebra A, where xa 6 A*, iJ 6 A (see (3.16)1,
i.e., xa is a commuting U a an anticommuting) variable. Notice
however that strictly speaking, each xa contains mor<? informa­
tion than a real number. A itself is a real space of very high
dimensionality, allowed ly higher than^*4, and containing all
 49 50

B) A second choice is H 5 S0(4)© A = **AR. This is a

particular subgroup whose elements are the pairs

nilpotent elements of the form A* ® A* etc. We refer to this

space as "superspace." Theories using (=,0,£r )qp * e ®a r
( 6 . 1)
this form of H should have complete local Lorentz gauge in­
variance and complete covariance under general coordinate ! u * i Y 5K R - o
transformations x' * f(x,i), (' » g(x,(). In this class of
theories we have the super-Riemannian (V^ 4 in a graded exten­ It is therefore a semi-direct product. A mirror image

sion of the Schouten notation22^) Arnowitt-Nath approach*4^ and GAl is obtained by considering

the alternative torsion (U^ superspace of Wess-Zumino.*6^

(H,5l), | d - iY5K L “ 0 ( 6 . 2)
Theories based on this framework have non-trivial computational
tasks to overcome and for this reason still have some loose
ends in their final shape. In particular, the way in which instead of (6.1).

ordinary General Relativity is or should be contained in these We have GAR GA^ “ S0(4). This choice for H seems to have

theories has not been elucidated. In the 4 version16^ the been overlooked. We remark however that GP is a contraction

language of forms is used but only field equations and no of 0Sp(l,4), which contains 0Sp(l,2£) as a subsupergroup.

Lagrangian are given. In the V4 ^ theory, 141‘ the standard The contraction of OSp(l,2£) leads to GAR. Those groups ap­
' 291 pear in the MacDowell-Mansouri version of supergravity,23^
Einstein tensor calculus is used in the graded version, ’ up
to the actual writing of a scalar curvature R. The resulting but it seems that their Lagrangian is not invariant under

field equations have not yet been investigated in detail (even local GAR gauge transformations. Moreover, their forms are'

the listing of all fields has not been completed). A major written o n ^ 4 (or possibly our l a r g e r ^ 4 since the same

stumbling block remains the need to use the Berezin integration formalism would apply), and not on GP/GPR

formalism over the odd variables, a procedure which selects The elements of ty 4 ’ 2 are pairs,

the term of the highest degree in those variables, with a re­

sulting increase in the volume of computations. U,CL), j(l - iysH L - 0 , x&jfl (6.3)

with an "analytic" 2-component CL Majorana spinor projection.

 51 52

C) We might also select the sub(graded)group H = GP^, those of the Lorentz generators S.^
defined as the set of triples
a1J e A'

x * (=,x,5)
but the translations Sj have nilpotent coefficients in GPj,

where now 5 € A ‘ but x e A* * A", that is x £ A+ but is a nil-

potent element. Obviously this is still a supergroup and also a* e A* © A" C A* , € GPj (6.5a)

GP1 C GP, but GPj is a proper sub(super)group of GP. By ob­

vious considerations we have GP/GPj , where this is now a as against either nilpotent or real ones in GP,
purely ordinary manifold on the .reals. This is thus the fac­
torization appropriate for a (Euclidean) space-time theory of super- a1 e A* , a1Sj e GP (6.5b)
gravity. We shall detail it at this stage as a heuristic model.
The GPj factorization has no counterpart in ordinary Lie Similarly, the space-time vector forms o taXe A*(A © A )

groups since both GP^ and GP have apparently the same Lie values on the Lie algebra DGp(DGp ). fl is always A* valued and

algebra. This seems a very interesting possibility for we ui has A* values on the D of either group.
would write forms on ordinary space, but the theory would It is quite conceivable to write an A+ valued form o as
still have local supersymmetric GPj invariance. The differ­ a sum:
ence between GP^ and GP lies only in the choice of the coeffi­
cient ring of the element of the Lie algebra.
The generic element where or is the real-valued, and Oj is the GP^ nilpotent-valued
part. We can thus also rewrite for GP the Cartan-Maurer equa­
al*S.j + a1Si + aS (6.4) tions (3.21) in the form:
duab ♦ uat „ u>tb = 0 ]
will have the coefficients of the spinors S do* + uat - o*i = 0 j
r T > (6.6)
. 1 , abab, , „ r K ’
du) ♦ j(u> o Ju) * 0
a,a e A . a ♦ u ) at .. O t 1- a n j
d O j j ♦ y u jy (i» = 0 J
 S3 54

The GPj equations are obtained by putting

0 , o » 0 ^^, o e GP^ . (6.7) Rabpfp*eabfe + 4iKY5p V p (6.9)

Thus (3.21) can be used foi oth GP and GP^ provided we restrict = RabPfPC£abfe + 2iRYspaYap - 2ipYsP3YaR
the values of o suitably to nilpotent elements for GP^. In a ♦ surface terms

physical situation with curvature, these forms are replaced by

A
a set p with the sane scher jf values. Factorization implies,
We shall develop the GPj choice in more detail. It is
pab = AtaTtsAsb + AtadAtb
particularly interesting if one looks for pseudo-curvature
forms.
We have no proof that the choice in the following is pa = Ata(Dal + t1 + ■cyta - |’'tYtD.a) ( 6 . 10)

unique, but it does lead to a new "geometric'’ derivation of

previously known results. Using (3.21) it can be checked that p - U(A,-1 ) (t ♦ Da)

the multiplet cA given by

where D denotes the G. R. derivative

5ab ” tabcdM
Da » da * iiaboaba
( 6 . 11 )
6 8
( . )
Da1 - da1 ♦ rta . a1

Ca * -4ipY8paY5 or c° » -4iysYapaP
(Tts,Tt,T) are forms o n ^ 4 , and (A,a,a) £ GPj.
has vanishing covariant derivative. We propose therefore the We set here
Lagrangian density ui8b ■ AtadAtb ]
o = A"1 (da - jcTyda) ( (6.12)
ui = u(A’*)da J
 55 56

which satisfies (3.21), so that (6.10) indeed realizes (4.31). The final result appears to be rather complicated:
The corresponding factorization of the curvatures is given by

-. i „ . ..i, . i J C V
Rab - Ata/ T V b
♦ 2i[v/^r5Yfa) ♦ (aysrf/f ) J/lf - 4 i ^ r 5 r£/f a{

Ra . Ata(/ltbab , * t + ^ t o . lsYt ^ e f 0efa) ♦ y ^ ab(«Y£5a)Cnreo)eabfe

(6.13) - y ^ ab(“YfS“ ) l“YeDo)eabfe

R - u(A'1)(yf ♦ ^ abaabo)
- iy^Y5(aYfDa)YfT + iTY5 (avfDa)

R « (/{ - j 5T4 aboab)u(A) ♦ 2 i / ^ Y j ( t y ^ oi) Y ^ D o - 2iDaYj(T'Yfa ) Y f^

The factorized /£A are constructed from (t ^ . t *,!) using ♦ j ^ ab(TYsYfDa)5:YsYeaeabfe (6.15)
A t£ t
the same formulae yielding R from (p ,p ,p):
The values of the action have to be real, i.e., in A .

>£ab - dtab -
♦ t-ab „ t
-tb We constrain

/ £ e A‘ , / V <S= (A')2 (6 .16)

/ t a • dxa ♦ Tat -T* ♦ (6.14)
i.e. , no real contribution to the factorized generalized tor­

/t ■ dT ♦ j-(Tab0ab )T • Dt
sion This is a generalization of the

y?f ■ 0
We can now insert (6.13), (6.10) in (6.9). We denote by Lp
result of C. R. Indeed, (6.16) implies that the only real
the Lagrangian density constructed as in (6.9), and by LT the
1~ a
allowed contribution to the G. R. torsion be given by " 2 TY T
same with pA ■* tA . There is no Ata, dAta dependence in iT at
all. It is therefore possible to choose Ata ■> 6ta, dAta « C. as can be seen in (6.14). This can be considered as a boundary
condition taken from the Kibble-Sciama version22^ of G. R.
which fixes the contribution to (real) torsion so that it be
 57 58

proportional to the spin of the spinor fields present (here The relation to conventional supergravity is by no means
the Ty field with J = 3/2). Since the original theory is trivial. It is of course obvious that all forms can be taken
postulated in GP, we have also allowed residual (nilpotent) to have values in [ A]0 « reals, A , [A ]2 (* nilpotent even
contributions in addition in (6.16). Both conditions in variables) only,because higher order terms do not contribute
(6.16) are satisfied by (6.13) provided o £A", a £ A" * A" to the action (an [A*]3 term in L must be paired to at least
as required by GP^. another A’ term since L is even, and this vanishes by (6.17)),
Since the action is given by a d4x integration, we have Moreover, variations must be carried over consistently with
to define such an integral for functions with (nilpotent) (6.16). This we do by adding a term
values in A © A . These are now products of either the
j Aa/?a , Aa e. A* 0 A' (6.20)
a(x), a(x) gauge functions or of * x A' valued forms,
to I, where Aa is a Lagrange multiplier. Upon variation of
l„-J*i(x)...*2 N(x)d4x (6>17J t, t3 , Tab we find,

We choose to define Y a T *fL ■ 0 (mod. [A']3) (6.21a)

IN - 0 , for N > 2 (6.18)

y£abrfeabfe - 2 i 7 ^ SYe T ♦ 2DA* - 0 (mod. [A']4) (6.21b)

This choice ensures in (6.15) that

^ fTecabfe » -Aatb + A V (mod. [A ]4) (6.21c)
J d 4x U p - lT) = 0 (6.19)
Equations (6.2l) are GPj invariant in this (6.16) "weak"
and Lt as given by (6.15) with p ■* t is GPj gauge-supersymmetric sense. Within this definition we use (6.10) and (6.13) as
in a natural way. This end product which is achieved without the gauge transformations, viewing p * Tx (A,aj,a)T, R »
further assumptions by our choice of cA shows a formal Tx(A,a,a)/f. Also, since (6.21) is obviously S8(4) local
similarity with the Supergravity theory of Freedman, van gauge invariant, we may set Aab * 5aD. The remaining gauge
Nieuwenhuizen and Ferrara.25)26)23) functions ab (x), a(x) with x e ^ . 4 , are of order [A J2
 59 60

and A respectively. Within this choice, the only change in is a compensating variation for/C.a. That this i" indeed so
(6.21) which is not obviously of order [A']3 is in (6.21a) can be checked through a somewhat laborious procedure in which

the identities and equations

but
a ef 1/ a ef ef 1, a ef ef a, /^ abTb = ' ^ “‘’^ a b f e = 0 mod- tA ,Z (6.25)
y cr - y (r o + o r ) ♦ 2" ( r » - °r )

(but not mod. [A']4) are repeatedly used. Clearly, Aa in

l(iY5£aefV ♦ ieV - 6*V)
(6.21) can be deduced from 3 through some linear duality
yielding
operation, which shows that the relation between these two
4 - JiY5Yt«M f t A #£**« 4 * aV x a« .
sets of forms is involu tory. If Aa ■ 0, then^.3 " 0, and
(6.21) reduces to conventional supergravity.
The £irst term is of order fA~13 because of (6.21b), and the
second is of order [A*]3 because of the fiianchi identity

However, this identification is formal, since the latter

(D/C)* = ^ taTa ♦ >€ytr « 0 (6.22) theory25^26^23^ does not restrict the (odd-grading) spinor
gauge functions by condition:; such as (6.17).
Thus (6.21a) is weakly GPj gauge invariant. The gauge
covariance of (6.21b,c) requires a change in Aa. If according
to (6.13) and (6.10),

^ ‘‘a1 ♦ - \SymA stasta

/t • . K ♦ \ A a\ a\
2 (6.23)
t' * x + Da
xb ' « xb ♦ Dab + iYba - iqfybDa
then

A,e - A* ♦ j ^ 8bafeabfe ‘ |-(aoabYsYe<»)/Cab (6.24)

 61 62

7. Supergravity . Therefore, solving the equations of motion (7.2-7.4)

on GP is somewhat tantamount to an exponentiation of all
Conventional Supergravity rests instead on an action with
ordinary and supersymmetric transformations, just as solving
a Lagrangian density defined as a 4-form, integrated on
the unfactorized Einsroin cauations CS.1I) amounts to carry­
a submanifold of GP, in analogy to unfactorized General
ing out all possible SO(4) gauge transformations. Here,
Relativity (5.10).
however, our Lagrangian has no supursymmetric gauge trans­
formation in the standard sense defined by (4.7), and there
" • 1> is no simplifying factorization of the odd Grassmann variables.
The integration of supersynunetries is therefore just as diffi­
On a generic , the variational equations are25,26,27^
cult as integrating the equations of motion. In this sense,
the C“ should be regarded as Fermionic "times.”
(7.2)
This conclusion is strengthened by a close analysis of
the commutation relations of in GP, in view of equations
RabPfeabfe ' 2iRW " 0 (?'3)
(7.2-7.4) which imply the vanishing of all except a handful
of the anholonomic components Rg£A .
YapaR » o (7.4)
First, (7.2) asserts that

The equations of motion impose severe restrictions on the

Rgg8 = 0 -VB.E (7.5a)
components Before we derive these results, it is quite
interesting to analyze the relationship between a generic
We next observe that (7.4) implies the vanishing of the
solution to equations (7.2-7.4) and its restriction to "
3-form yapaR. Contracting (7.4, with the tangent vectors
particular CGP.
(Dc ,Do,Dg), and using (4.27)
In analogy to the case of /I* C P, we claim that the entire
physics is already contained in any single 4^ , and that the
25) pa (Do ) - 0
conventional supersymmetry transformations ' of Supergravity
relate the fields on 9H* to the fields on any other subvariety
 63 64

ve obtain (with yc acting on the unwritten spinor index of R) R(ab)c = R (ab)ct ' R (ab)(cd) ° 0 (7.5d)

We note that only RC(j survives.

Turning to (7.3), contracting with (Dc ,Da ,DB) and using
It follows that (7.5c) yields,

RoB * 0 (7.5b) RaS eabce = 0

We Hay contract (7.4) next with (DC ,D£,D0 ) and obtain, so that

R;-ab = 0 (7.Se)

Taking c + { and multiplying by yCY^ yields, Similarly,

* Rfa + A c * 0 R (cd)(eV)a b “ R(cd)'aa b = 0 f7-5^

and for f f c f s. and contraction of (7-3 ) with (Dm »Dp.Dst) leads to

-*Sr m “ ^ Rf« “ ^ Crcc - *#Rso " 0 R(s‘t)mab = 0 (7-Sg)

so that We thus find that ail curvature components invol ving an

S0(4) index (cd) vanish, as would be expected from an S0(4)-
R = 0 (7.5c) factorized theory. The Hc<* variables appear factorized in
sa
the expressions for the curvature, at least. To that extent,
This proof holds if we replace the spinor index a in Da with we may regard this fact as consistent with the hypothesis of
a "vertical" index A »: (ab). Thus,
 65 66

factorization of S0(4) appearing as a result of the equations (though the summation over M,N is now reduced to vector-
of notion. Indeed, we surmise as for General Relativity vector and vector-spinor MN sets only). This explains the
that this does occur here too. Curvature is in any occurrence of in ref. 30) as distinct from "Rai,cd'"
case restricted to /{4/4 = GP/H, i.e. to superspace. The It is the fiabcd of ref. 30) which coincides with the present
commutation relations of the DA are thus given by (4.41a-b) RC{j3b, while Ra|jC(j of ref. 30) is an auxiliary quantity of
and (4.44a-d). no physical and geometrical content.
Contraction of (7.3) with C6]n,Bp,6a) leads to In matching these components with those of ref. 30),
care should be taken in setting up a correspondence between

C^abpe - Rapabeabme ' = 0 (7'6) our geometrical operators (and in general any vector field
on GP) and the set of second-quantized operators^(x), f t /(x ) ,
(the dot on the first B in the last term denotes a "lower- 5Tab(x) of ref. 30). The latter can be written as a GP multi-
spinor") which leads to a non-trivial relation between the plet ji^(x) with the commutation relations,
spinor-curvature and the spinor-components of the ordinary
curvature Rab. Relation (7.6) is the same as equation (11c)
“ «(*.*') ( c % E + r be A}^ a (^ (7‘8)
of ref. 30),’ in which R_is
mp
denoted FT mpand R*‘ab
own
is essen-
tially I the somewhat unwieldy expression given for the in analogy to our (4.41a,b; 4.44a-d) set. Note that with the
variation of the Relativity connection field, in the original vanishing of components in (7.S), (4.44b-d) now become,
papers on Supergravity.25'275 That this is indeed can
be seen from (4.43a). f5a A l - RabCdDed * Rab5 (7-9a>
Finally, we need to compute

W V V ■ Ra«CdDcd (7-9b)
Rcd = ZAd 6c *Vw (7’7)
“ tYa)aB5a (7.9c)
A possible source of confusion arises in trying to cal­
culate anholonomic components by using the inverses of the and (4.43) consist in (the D are GP covariant derivatives)
conventional vierbeins instead of the inverse 14 x 14 matrix
 67 68

4pab - Deab - pCedR ” ab . pci“R

co ca Conclusions

6p® - Dea The following facts have emerged from the foregoiag
discussion.

6p ■ De - pCe 1) It is possible and indeed convenient to develop the

formalism of gravitational theories as gauge theories on a
which are the conventional transformations of -jpergravity group manifold. The groups are P or GP for Gravity and Super -
(see (4.42a-c) for the variations contained in Dea and De). gravity, but further enlargements should follow the same

Supergravity however has meaning on a 4-dimensional mani- method.

When this is done, the connection for'is and the vierbeins
foldt/tff* only which we may take to be « 0. (4.43) or (7.10)
appear as a single c-bein on G (c--the dimension of G) which
are then relating the restriction of the forms pA on«^*(<; “ « 0)
also play the role of connections on G in computing covariant
before AGCT to the corresponding restriction o n t ^ 1 (£° ■ 0)
derivatives and curvatures. In particular, ordinary curva­
after AGCT clearly any covariant theory admits formulae (4.43)
tures and torsions are naturally unified as expected in
A
but so far only in super»ravity the R^* components are all
Einstein-Cartan theories.
functionals of the restriction P^LJ and (7.10) is a true trans- 2) An action is formed under the assumption that it reduces
Off
formations. In this sense supergravity is the only known to the Einstein action in the appropriate cases, and be com­

svpersymaetric theory with structural group GP. patible with the vanishing of curvature ("flat" empty space).
The notion of pseudocurvature is introduced and related to
Chevallev cohomology and to the Segal-Wigner-Inonu theory of
group contraction.
Pseudocurvatures provide an interesting way of achieving
the same ends as the MacDowell-Mansouri23^ procedure, but with
methods which are purely intrinsic to G. The method repro­
duces conventional Supergravity.
3) In all these theories, the action is strictly gauge
invariant under SO(4) gauges only. Moreover, the Lagrangian
 69
70

density is always a 4-form and is integrated over an arbi­ Appendix

trary variety in G. However, physics is seen to be com­

pletely determined by what happens on a single . The The (Kuclidean) Dirac Algebra
transfer of information from any to any other C G Metric: fl, 1,1,1)
implies either ordinary SO(4) gauged symmetry or supersym- a b . b a ...ab (A. 1)
Y Y + Y Y * to
metry as an AGCT gauge. In a way, restricting a GP S0(4)- ab I, a b b a. (A. 2)
0 = yCY Y - Y Y )
factorized theory to A A/4 includes as partial rt,4 slices O .4 , 0,-1
Y * -1Y ■ -(Y } CA.3)
all possible supersymmetry related conventional supergravity ys * yV yV ■ -rY5r l (A.4)
theories. "determinism" has been realized else­ ly * -y C (A. S)
where."**^ Ay M - V A (M » 0,1,2,3) (A.6)
4) Factorization is seen to be a key simplifying feature With the choice
in these theories. We conjecture that if a Lagrangian is A = -C = y 4 (A.7)
gauge-invariant under any H C G, then it is H-factorizable we have
as a consequence of the equations of motion. In this sense, , a<+ , a,-l a (A.8)
(Y ) » (Y ) = Y
factorization should not be an independent postulate. CT = -C (A.9)
It is possible to give a heuristic proof of the fac­ (CYa)T « Cy 3 (A.10)
torization hypothesis for solutions infinitesimally close to (Caab)T = Coab (A.11)
a factorized one. All these solutions can be reduced to (Cy 5)t = -Cy 5 (A.12)
factorized ones by an infinitesimal coordinate transformation ( C Y j r V = -Cr5Ya (A.13)
on G. In the large, however, it may be that there are discrete C2 = 1 (A.14)
families of factorized solutions with the same boundary condi­ and the adjoint (= charge conjugate) Majorana spinor is defined:
tions and still topologically distinct. This interesting pos­ if = (C*)T (A.IS)
sibility remains an open question.
5) The procedure can be in principle extended to other Commutators
groups. It would also be desirable to extend the theory so as [oab,ocd] = Sadobc - «acobd + 6bcoad - «bdoac (A.16)
to include "matter" distributions whose energy-momentum, super- (ra,<7Cdj = aaCYd - 6aV (A.17)
synunetry and spin density tensors are non-trivial functions on G.
 71

References
The generators:
1) See for example, B. Zumino, Free. 1'th Intern. Conf. on
S r?) - va 3 „b 3 (A.18)
x & ' x 3? High-Energy Physics, 19~4 (fcd. J. R. Smith), Rutherford
Lab., Chilton, pidcot, u. K., p. 1-254.
Sab^) - ^ ° ab’V 8 (A.19)
2) A. Salam and J. Strathdee, Phys. Rev. D11, 521 (1975).
3) L. Corwin, Y. Ne'eman and S. Sternberg, Rev. Mod. Phys.
ISa b ^ - Sab^)l = 0 (A.20)
47, 573 (1975).
4) S. Ferrara. Rivista Nuovo Ci.nento 6, 105 (1976).
obey the same algebra as (A.16)
P. Fayet and S. Ferrara, Physics Reports 32C, 249 (19” ).
5) For a previous exposition of the theory for physicists, see
(A.21)
•Sab,Scd* = 4adSbc ' 6acSbd * 6bcSad ' 6bdSac B. S. DeWirt, In Relativity Groups and Topology (Proc.
Les. Houches 1963 Seminar), ed. C. and B. DeKitt, Gordon
Fierz-transformations utilized in Chapter 6
6 Breach, N. Y (1964).
Curex)eabfe * i(xoabY sirfw) - iCuoabY sY fx) (A.22)
6) F. A. Berezin and D. A. Leites, Dokl. AkaJ. Xauk SSSR,
for the case of
224 (.1975); Engl, translation in Soviet Math. Dokl. lt>,
(x.u) * 0
1218 (1975).
and
") B. Zumino, Proc. Conf. on Gauge Theories and Modern Field
(wYe«>)eab^e ■ -^iCuo^YjY^u) (A.23)
Theory, Northeastern University, Boston 19*5 (eds. R.
for [o>,u] ■ 0.
Arnowitt and P. Nath), M.I.T. Press, Cambridge, Mass.,
1976, p. 255.
8) F. A. Berezin and G. r. Katz, Mat. Sb. (SSSR) S2, 343 119"0).
Fngl. translation 1_1, 311 (197C).
9) T.h’.B. KibbJe, J. Math. Phys. 2, 212 (1961).
D. K. Sciama, in Recent Developments in General Relativity
(Pergamon & PWN, Oxford, 1962), p. 415.
10) P. Von der Heyde, Physics Letters 58A, 141 (1976).
11) Yu. A. Golfand and F.. P. Likiitman, JETP Letters 13, 452 (1971 1.
 73 74

12) J. Wess and B. Zumino, Nucl. Phys. B70, 39 (1974). 27) D. Z. Freedman and P. von Nieuwenhuiien, Phys. Rev. D14,
13) A. Salam and J. Strathdee, Nucl. Phys. B76, 477 (1974). 912 (1976).
14) p. Nath and R. Arnowitt, Phys. Letters 56B, 177 (1975). 28) P. Nath and R. Arnowitt, Physics Letters 65B, 73 (1976).
15) V. P. Akulov, D. V. Volkov and V. A. Soroka, JETP Letters 29) R. Arnowitt, P. Nath anJ B. Zuraino, Physics Letters 50E,
22, 396 (1975) (Engl, translation p. 187). 81 (1975).
16) J. Kess and B. Zumino, Phys. Letters 66B, 361 (1977). 30) C. Teitelboim, Phys. Rev. Letters 38, 1106 (1977).
17) C. Woo, Lett. Nuovo Cimento 13, 546 (1975).
31) L. Brink, M. Gell-Mann, Y. Ne'eman, P. Ramond and
18) P. P. Srivastava, Lett. Nuovo Cimento ^3, 161 and 1^, 657
J. Schwarz, "Superspace Lecture Notes," Aspen Center
(1975).
for Physics, 1977 Supergravity Seminar (unpub.).
19) E. farta-i, C. R. Acad. Sci. (Paris), 174, 593 (1922).
Also, Ann. Ec. Norm. Sup. 40, 325 (1923); « , 1 (1924);
£2, 17 (1925).
20) F. W. Hehl, Thesis, Techn. Univ. Clausthol, 1970 and Phys.
Lett. A36, 225 (1971).
21) A. Trautman, Bull. Acad. Pol. Sci., Ser. Sci. Math.
Astron. Phys. 20, 185, 503, 895 (1972), and 21, 345 (1973);
also Symp. Math. ^2, Bologna, 1973, p. 30.
22) F. W. Hehl, P. Von deT Heyde, G. D. Kerlick, J. M. Nester,
Rev. Mod. Phys., 48, 393 (1976).
23) S. K. MacDowe 11 and Mansouri, Phys. Rev. Letters 38,
739 (1977).
24) C. Chevalley and S. Eilenberg, Trans. Am. Math. Soc. 63,
85 (1948).
25) P. Z. Freedman, P. von Nieuwenhui:en and S. Ferrara,
Phys. Rev. D13, 3214 (1976).
26) S. Deser and B. Zumino, Phys. Letters 62B, 335 (1976).
 75

Acknowledgements

The authors would like to thank Professor J. Prentki and

the Theory Division of CERN for their hospitality during part

of this work. They would also like to thank Professor J. A.

Wheeler and the Center for Theoretical Physics at the Univer­

sity of Texas, Austin, for his hospitality to one of us (T. R.)

during part of the continuation of this work.