IL NUOVO CIMENTO VOL. 108 B, N.

11 Novembre 1993

Non-Abelian Gauge Antisymmetric Tensor Fields.

S. N. SOLODUKHIN
Department of Theoretical Physics, Physics Faculty of Moscow University
Moscow 117234, Russia

(ricevuto il 14 Maggio 1993; approvato il 29 Maggio 1993)

Summary. - - We construct the theory of non-Abelian gauge antisymmetric tensor

fields, which generalize the standard Yang-Mills fields and Abelian gauge p-forms.
The corresponding gauge group acts on the space of inhomogeneous differential
forms and is shown to be a supergroup. The wide class of generalized Chern-Simons
actions is constructed.
PACS 04.50 - Unified fields theories and other theories of gravitation.
PACS 02.40 - Geometry, differential geometry, and topology.
PACS 11.15 - Gauge field theories.
PACS 12.10 - Unified fields theories and models.

1. - Introduction.

Recently, much attention has been paid to the study of theoretical models with
Abelian gauge antisymmetric tensor fields (AGATF). It turned out that AGATF play
an important role in string theory, supergravity and gauge theory of gravity [1]. An
important circumstance is that the AGATF can be described using differential
geometric methods as differential forms (in which case the parameters of gauge
transformations are also differential forms). On the other hand, the fermionic-matter
fields can be described with the help of a set of antisymmetric tensor fields of
a r b i t r a r y rank (i. e. inhomogeneous differential forms)[2, 3].
During the last years the surprising relation of gauge fields to the space-time
topology was discovered. In the works of Donaldson [4] and Witten [5] such a relation
was demonstrated for the standard gauge fields. On the other hand, it was shown by
Horowitz [6], and earlier by Schwarz [7], that AGATF give us an example of the
so-called ((topological field t h e o r y - and lead to the appearance of topological
invariants (such as the Ray-Singer torsion and linking numbers) under quantization.
In[6] the minimal non-Abelian generalization of AGATF was also considered.
However, so far there does not exist a satisfactory generalization of AGATF for the
non-Abelian case. Attempts of such a generalization[8] result in an essentially
non-local and non-linear theory. Hence, an actual problem is to construct a theory of
non-Abelian gauge antisymmetric tensor fields (NGATF). These fields could find

1275
1276 s.N. SOLODUKHIN

numerous applications in m a n y theoretical models where so far one has used the
AGATF. T h e r e is also a hope t h a t N G A T F will give us new tools for the study of
space-time topology. I t is a goal of this p a p e r to a t t e m p t the construction of N G A T F
theory.
At p r e s e n t we know two kinds of gauge fields. The first is the standard
(Yang-Mills) non-Abelian gauge field which is described as the Lie-algebra-valued
one-form A(1) with the infinitesimal gauge transformation law:

(1.1) ~A(1) = - d~(0) - A(1)~(o) + ~(0)A(1) ,

where A(~) = A~;~ ~ dx ,~, the matrices ;~ are the g e n e r a t o r s of a gauge group G, ~(0) =
= ~ ~)~a is an infinetisimal transformation p a r a m e t e r - - t h e zero-form with values in the
Lie algebra of the group G.
The second is the theory of A G A T F which are n-forms A(n ) with the gauge
transformation law:

(1.2) ~A(~) = - d~(~ _ 1) ,

where ( n - 1)-form ~(~_ 1) is an infinitesimal transformation p a r a m e t e r .

Our aim is to construct a theory of N G A T F which would be n-forms A(n) with
values in the Lie algebra of a group G:

1
(1.3) A ( , ) = ~.. nal~ 1 ....~-~,t ) a d x ~ A ... A d x '~" 9

Such a theory should generalize both the Yang-Mills model and A G A T F theory in
t h a t the transformation laws (1.1), (1.2) are recovered as the particular cases of a
gauge transformation law for fields (1.3).
One could naively expect the following transformation law for A(n):

(1.4) ~A(~)- - d~(~ 2) - A A ~ ( n - 1 ) + Zt(n ~ ) A A ,

where

(1.5) ~(~ - 1) - - - ~,,1 , , , ~ ) ~ d x " ~ A . . . A d x '~ '~

(n - 1)!

is an infinitesimal p a r a m e t e r (n - 1)-form with values in the algebra of group G.

However, the formula (1.4) is evidently senseless, since the last t e r m s in (1.4) are
differential forms of r a n k ( 2 n - 1), while the left-hand side of (1.4) is the n-form.
Only for the case n = 1 this formula has sense. In order to m a k e sense of (1.4), let us
consider a set of gauge forms A(~), A(i) such that (1.4) is replaced by

(1.6) -)~A(,,) - - d~(~, _ 1) - A(1) A ~(n - 2) + ~(n- 1) AA(~).

A natural generalization a p p e a r s to take a complete set of forms A(n), A(n 2), - " ,
A(p), ..., A(i) with a transformation law

(1.7) ~A(n) - -d~(~, 2) - ~ A(k) A ~(~-k) + ~(n k) AA(k),

k
N O N - A B E L I A N G AU GE A N T I S Y M M E T R I C T E N S O R F I E L D S 1277

where ~(0), a(~) . . . . , ~(p), ..., a(n ~) are infinitesimal gauge parameters--the set of
k-forms with values in algebra of group G.
In the standard Yang-Mills theory the gauge transformations act on a multiplet of
matter fields:

(1.8) 8~r = ~ijr j

Both r and the parameters are zero-forms.

A natural generalization is to describe the matter fields by n-forms, and the
parameters by n-forms. However,
s " i
(1.9) (n) ----) r ~n) "~ O~(k)j A r ~n)

transfers n-forms into (n + k)-forms. Hence, one should consider the set of forms of
all possible ranks on a given manifold, i.e. the multiplet of inhomogeneous differential
forms:

(1.10) Vj. i = ~k ~.
1 r ' ~ d x ~ A ... A dx ~r 9

Then the transformation generated by inhomogeneous forms

(1.11) ~ 9= ~ 1 ~ ijT,...~k dx '~IA "" A d x ~k

~.

acts on the field (1.10) as follows:

(1.12) 8 ~ , i = ~ijA ~v.J.

Notice that such a generalization was initiated by the study of fermionic-matter

description in terms of inhomogeneous differential forms[2,3]. In particular in
ref. [3] the example of supersymmetry transformation mixing the forms of different
rank was considered.

2. - I n f i n i t e s i m a l gauge transformations.

Let us consider an action for matter fields described by the differential form F"
taking values in a linear space _~N (1.10) which is invariant under the global
transformation (1.12):

(2.1) ~'---) ~' + ~ A ~',

where the infinitesimal-transformation parameter a (which is the matrix-valued

inhomogeneous form (1.11)) has constant coefficients. The kinetic term of an action
depends on d~'.
Under global transformations (2.1) one has

(2.2) d ~ ~ ---) d~F" + ~ A d ~ ~",

85 - ll Nuovo Uimento B
1278 s.N. SOLODUKtIIN

where the operator r, acts on the forms as follows:

(2.3) V~ = ~k ( - k!1)k ~.~l...,~kdx"~l A ... A dx ~k 9

This being the involution, one easily sees t h a t 7 2 = id. Then under local transfor-
mations (2.1) (when coefficients ~.,,1 .,~k depend on x) one obtains

(2.4) d7 ~--) du 'r' + ~a A d7 ~ + d~ A ~ .

In order to compensate the last term, one should introduce the gauge f i e l d - - a
matrix-valued inhomogeneous form A (1.3):

(2.5) A = ~ A(k),
k
and consider the covariant derivative

(2.6) VT" = du + A A 7 ~,

which t r a n s f o r m s under local transformations (2.1) as follows:

(2.7) V~ ~--) V~ '~+ r~ A V7 ~9

Then, the gauge-invariant action is obtained b y substituting the external derivative

d7 ~ by the covariant derivative V7 ~.
F r o m (2.7) one obtains the transformation law for the gauge field A:

(2.8) A - . A - d~ + r~ A A - A A a .

In t e r m s of the inhomogeneous components (2.5) this formula reads

(2.9) A(~)--)A(n) - d~(~ 1) + ~ [ ( - 1 ) k a(k)AA(n-k) - A ( ~ - k ) A ~(k)].

k=0

This formula m a k e s our preliminary discussion in the introduction m o r e precise (see

(1.7)).
I t is easy to see that, indeed, the transformation laws (2.8), (2.9) contains the laws
(1.1), (1.2) as particular cases.
The curvature for the AGATF,

(2.10) F(~ + 1) = dA(~),

is invariant under gauge t r a n s f o r m a t i o n (1.2), while the curvature for the standard
Yang-Mills gauge field

(2.11) F(2) = dA(~) + A(~)AA(1 )

under a gauge transformation (1.1) changes as

(2.12) ~ F ( 2 ) = ~(0)F(2) - - F ( 2 ) ~ ( 0 ) 9

L e t us find the generalized expression for the NGATF curvature with the
transformation law

(2.13) F-+F + ~AF- F A~.

NON-ABELIAN GAUGE ANTISYMMETRIC TENSOR FIELDS 1279

It is easy to see that the proper generalization reads

(2.14) F = dA - (~A) A A .

This satisfies all necessary conditions.

For homogeneous components F ( k ) : F = ~ F(k) one has
k

(2.15) F(n + 1) = dA(n) - ~ ( - 1)kA(k) A A(n - k + 1) 9

k=l
It is also easy to verify the Jacobi identity for the curvature F,
(2.16) dF+AAF- (vF) A A = 0.

In the standard Yang-Mills theory, one usually considers the gauge one-form A(1) to
be anti-Hermitian. However, there are no reasons for such a restriction in our
general construction.

3. - T h e gauge group.

Let us discuss now the group structure of transformations (1.12) acting on the
space of inhomogeneous forms ~" with values in an N-dimesional linear space (see
(1.10)).
Let G be an inhomogeneous form with values in a space of matrices G L ( N , C):

(3.1) G = E G(k) , G(k) = ~.1 Gi.j,~l ...[.tk dx ~1 A ... A dx ~k "

k

We determine the group multiplication of such forms as follows:

(3.2) G' o G" = G '~j A G"Jk.

In the general case, the exterior algebra is not a group, since not for every element
one can determine the inverse element with respect to the wedge product A. The
condition of existence for such an inverse element is the non-degenerate scalar part
G(0) ;~ 0 in the Abelian case (when {G} = A ' M ) , and

(3.3) det [G(~0)j] ~ 0

in the non-Abelian case. Consequently, one must consider the space of all forms (3.1)
satisfying the condition (3.3).
In order to construct an action for the matter fields (1.10), it is necessary to have
an invariant scalar product on the space of the inhomogeneous differential forms
(1.10). The standard bilinear form

(3.4) I * ~ A F"

is invariant under transformations

(3.5) ys'__~G A
if G has only a zero-rank component. Notice that (3.4) demands a metric on the
manifold M which enters via the Hodge operator *.
1280 s.N. SOLODUKHIN

Instead, let us consider the following scalar product for forms on a D-dimensional
manifold M D :
D
(3.6) (r ,r =
l
M D
~r k=o .
I
M D
"

where the operator ~ acts on differential forms as follows:

D
(3.7) ~r = ~ 1 _ 1)k(k - 1)/2 dx "~ A ..dx '~
"

Notice t h a t ~ is the anti-involution on A * M D,

(3.s) ~(r A r = (~r A (~r

F o r invariance of (3.6) under transformations (3.5) it is sufficient to set the

condition

(3.9) ~G t A G = 1.

Clearly, all such G form a group.

L e t us consider

(3.10) G = 1 + a,

where ~ is an infinitesimal differential form.

Then from (3.9) one finds

(3.11) ~t = _ a.

A general solution to this equation reads

D (i)k(k 1)/2
(3.12) ~ = E (~ (k)
a ~a ~_ ~(k) i I ) ,
k=o k!

where ,z~k) (,c(k), e(k)) are k-forms with real coefficients, and N • N matrices h a =
h a

= ( i I , ) a ) are anti-Hermitian g e n e r a t o r s of the unitary group U ( N ) with the structure

constants f~aa. I t should be noted t h a t there is no additional condition (like d e t g = 1
for usual unitary group) for group elements G(3.9) in order to take away the Abelian
subgroup g e n e r a t e d by ~(k)iI (3.12).
Thus the g e n e r a t o r s of the group (3.10) have the form

1
(3.13) X(~) = )~~I(8 ) , 1(8) = - - dx '~1A ... A dx "~ , k = 0, ..., D .
k!

These g e n e r a t o r s satisfy

(3.14) b ~ ) = f a .b.c.x.( s + s , )
X~s)X~s,b ) - ( - 1)181 18'] X~s,)X~s

and so they define a superalgebra. Hence the group (3.10) is a supergroup. Roughly,
it is the unitary group U ( N ) with inhomogeneous differential forms as p a r a m e t e r s .
Notice that the supergroups a p p e a r in ref. [9] in a similar manner.
NON-ABELIAN GAUGE ANTISYMMETRIC TENSOR FIELDS 1281

In the appendix we find the general form for a group G element satisfying
eq. (3.9):

(3.15) G = go(1 + H + ~ )

where the zero-form go is an element of the unitary group U(N):

(3.16) gCogo = 1
and H and ~ are inhomogeneous forms of rank /> 1 which satisfy the conditions

(3.17) ~H t = - H , ~:7-t = ~J.

The form ,~)~is constructed as a polynomial of H (see the appendix): ~ = ~ ak H 2k

with respect to wedge product.
We can also define the subgroup of group G (3.15) as follows:

(3.18) GD=- { G = g o ( I + H + J - ) I H ( 2 k _ i ) = O , k = I, . . . , K } .

This subgroup will be useful in the next section to write the Chern-Simons terms
NGATF.
At the end of this section let us write the explicit matter field actions invariant
under global transformations (3.5), (3.9):

(3.19) S= f ~Ad~, S = I d~ A de,

M D IF1D

where

(3.20) ~ = ~t.
The gauge-invariant version of these action is

(3.21) s= f
M D M D

vh e r e V is the covariant derivative (2.6).

Notice that we do not use metric on the space-time manifold M D to write the
actions (3.19), (3.21). Hence, one can expect to obtain the description of some
topological invariants as quantum observables in the models (3.19), (3.21).

4. - Generalized Chern-Simons for NGATF.

In order to treat NGATF as a dynamical field one must formulate the action for its
description. It should be noted that the usual action

(4.1) I T r ( * F A F)

which is invariant in the Abelian case cannot be directly generalized to the

non-Abelian case. This was the main problem in earlier works on NGATF (see [8]). In
our formalism there is a natural possibility to overcome the difficulty with the help of
1282 S. N. SOLODUKHIN

the above-introduced invariant scalar product for inhomogeneous differential forms

(3.6). One can suggest the two expression for the NGATF action

(4.2) (F,F)-- ~ Tr(FAF)

J
M~

and
f
(4.3) (F, F ) = / T r ( F A F ) .
J
MD

It is easy to see that both these expressions are invariant with respect to the gauge
transformations (2.13) if the space-time dimension D is odd. However, if D is even,
we have for the variation of (4.2), (4.3):

(4.4) J~(F, F ) = I Tr[(a- r,a)AFAF],

(4.5) A~(/~, F ) = I Tr[(~- ~)A/0AF].

Consequently, in order for (4.2), (4.3) to be gauge invariant, the infinitesimal form
shoul be even:
(4.6) r,a = a .
More precisely, since the curvature form F has rank 1> 2, so the homogeneous
components of ~ should satisfy the condition

D
(4.7) ~(2k+1) = 0, k = 0, ..., - - - 3.
2
In other words for even space-time dimension D the gauge group for (4.2)-(4.3) is
D - 3 (see (3.18)).
G(D/2)
Here we can restrict ourselves to the particular case in even D (see
also [10]):
(4.8) vA = - A , vF = F .
Then, from action (4.2), one obtains the equations of motion

(4.9)
I dF +A AF-
[dfi + A A P - F A A = 0
(r,F) A A = 0 (for odd D),
(for even D).

It is easy to see that if the curvature F satisfies one of the conditions

(4.10) /~ = + F , F = - F,
then eqs. (4.9) hold identically as a consequence of the Jacobi identity (2.16). These
conditions (4.10) are similar to the self- and antiself-duality conditions for instantons
in the standard Yang-Mills theory. As concerns the functional (4.3), one can show
that under the same conditions on the gauge group (4.7) and on the gauge field (4.8) in
NON-ABELIAN GAUGE ANTISYMMETRIC TENSOR F I E L D S 1283

the even-D case, (4.3) is the total derivative since we have

(4.11) doJ(2) = 0, ~o(2) = T r ( F A F ) .

Hence, integral over the form o~(2) (4.3) is not an action but is an analog of the
Pontryagin class. It is well known that, given the closed invariant differential form

(4.12) d(o = 0, / I co = 0,

one can determine the Chern-Simons term

(4.13) ~o = d F c s 9

For the form %(2) (4.11) the Chern-Simons term Fcs has the form

2
(4.14) I"
cs = A dA - -
3 A(-~A)A

independently of the value of D. Notice that both ~ (2) and Fcs have sense in arbitrary
dimension.
Now one can write the gauge-invariant Chern-Simons action for every closed
(D - 1)-dimensional manifold M D - 1 .

(4.15) Scs = ~ Fcs.

MD-1

This action exists in arbitrary space-time dimension starting with dimension 3. Such
a generalization of the Chern-Simons term was suggested in[10]. It is also a
particular result in our general construction.
To remind, for the standard Yang-Mills gauge fields (one-forms) the Chern-
Simons terms exist only in odd dimension D = ( 2 k - 1) and are determined as follows:

(4.16) d F c s = ~(2k),

where ~(2k) is a closed invariant 2k-form:

(4.17) ~(2k) = Tr [F k ] ;

here F is the curvature 2-form.

The form oJ(2) (4.11) is similar to ~(4). However, 0)(2) comprise components of any
rank N i> 4, so the generalized Chern-Simons term Fcs (4.14) exists in any dimension
D I> 3 and, in particular, in even dimension D, contrary to the standard gauge fields
case (the four-dimensional Chern-Simons example will be considered in the next
section).
However, the above discussion does not exhaust the rich structure of NGATF
Chern-Simons terms. Indeed, in dimension D i> 6 one can consider the D-form

(4.18) ~o(8) = Tr (F A F / ~ F ) ,

which is closed and invariant under the same conditions (4.7). Hence, in higher-
1284 s.N. SOLODUKHIN

dimension D along with (4.14) there exists also the following Chern-Simons term

(4.19) ~p~3)
uICS ---- ~o(3 ) 9

Further, in higher-dimension D any invariant closed D-form

(4.20) o~(k) = T r F k

determines the corresponding Chern-Simons r CS"

(k)"

(4.21) ~l
,~1 ~(k)
CS ---- O ) ( k ) "

Any such term F(k)

CS defines the gauge-invariant NGATF action

(4.22) S(k)= ~ F(k)

~CS
J
M D- i

for a closed (D - 1)-dimensional manifold M D - 1

As a direct generalization, one can also consider the NGATF taking values in a
Grassman algebra A and write the corresponding generalized Chern-Simons terms
F(k)
CS" In three dimensions, we then obtain the field theory whose partition function is
the Casson invariant [11, 12]. Hence, one can also expect the appearence of relevant
invariants under quantization of these super Chern-Simons actions in arbitrary
dimension.
Summarizing, we obtained a very rich Chern-Simons-term structure for NGATF.
One can use these along with (4.2) as actions in order to determine the dynamical
description of NGATF.

5. - The h i g h e r - d i m e n s i o n a l Chern-Simons.

In three dimensions (4.14) gives us the usual Chern-Simons term for the
Yang-Mills field [5]. As a next example, let us consider the generalized Chern-Simons
term in four dimensions. F o r standard Yang-Mills gauge fields such a term does not
exist in D = 4.
F r o m our general expression (4.14) we obtain

(5.1) Fcs = Tr (A(1) A dA(~) + A(2) A dA(1) + 2A(1) AA(1) A A(2)).

It is easy to see that the integral of (5.1) over a four-dimensional closed manifold
M 4,

r
(5.2) Scs = / FCS
J
M4

coincides with

(5.3) S c s = ] 2 Tr [F(2) A A(2) ]

M4
NON-ABELIAN GAUGE ANTISYMMETRIC TENSOR FIELDS 1285

and is thus invariant under the gauge transformations (2.9):

3~A(1) = - d~(0) + ~(0)A(1) - A(1)~(0) ,

(5.4)
~A(2) = - d~(1) - ~(1) AA(1) - A(1) A ~(1) ~- ~ ( 0 ) A ( 2 ) - A(2)~(o) 9

The Chern-Simons action (5.3) with respect to (5.4) is transformed by the total
differential.
It is worth noting that the action (5.3) is already well known in the literature as
the BF-system. This model was proposed in [6] (see also [11, 12]) as an example of an
exactly soluble diffeomorphism-invariant theory. (A similar action was also
considered by Freedman and Townsend in [8].) As shown in[6], this model, when
quantized, describes the topological invariants such as linking numbers. The
transformations (5.4) contrary to [6] are not accidental but appear as a part of a more
general and rich structure.
As another example, let us consider the gauge gravity. The corresponding
generalization of the Lorentz group will be described by (see the appendix for more
details)

(5.5) ~Gki A ~kzG~- = ~ j ,

D
where G~-9 = • i
G(p)y is the matrix-valued inhomogeneous differential form; and Vij =
p=0
=diag(+l,-1, ...,-1) is the Minkowski D-dimensional metric tensor, i , j =
= 1, 2, ..., D.
Localizing the group (5.5) along the lines discussed in sect. 2-3, one introduces the
generalized k-forms c o n n e c t i o n s O)ik)j. However, not all of these are dynamically
present in the Chern-Simons action. In four dimensions, for example, one has an
expression similar to (5.3):

(5.6) Scs = ~ I" c s , Fcs = Tr [d~o(1) A 09(2) ~- 09(1) A 09(1) A 0)(2)].

,J
M4

This action was also considered in detail by Horowitz [6].

It is worth noting that the group (5.5) is a super-generalization of the Lorentz
group. Hence, the corresponding gauge gravity theory may be treated as the new
version of a supergravity theory.
Now let us write explicitly the generalized Chern-Simons term (4.14) for the
closed manifold M D , D >! 5:
D=5

(5.7) S c s = f Tr [A(3) A F(2) ] ;

D=6

= J Tr A - /\ - A(2) A A(2) A A(2)

(5.8)
IF(3) dA(2) + A(1) AA(2) - A(2 ) A A(1) ;
1286 s.N. SOLODUKHIN

D=7
(
f S c s = J Tr [A(5) A F(2) - A(3) A F(4) ],
(5.9)
F(4) dA(a) +A(3)AA(1) +A(1)AA(3) ;

D=8
r
Scs = J Tr [A(~) A F(2) + A(5) A F(3) + A(4) A F(a)A(2) AA(3) AA(a)],
(5.10)
F(a) = dA(2) + A(1) AA(2) - A(2) AA(1),
F(4) dA(3) + A(1) A A(3) + A(3) A A(1) - A(2) A A(2) 9

One can see that the generalized Chern-Simons actions in D = 6, 8 do not have the
form of the BF-system in these dimensions. It should be noted that although our
Chern-Simons action for a closed manifold M D, D = 4, 5, 7, ... has the form of a
BF-system, it differs from a BF-system for a manifold with a boundary which is
appropriate for the canonical quantization of the model. For an open manifold this
action is similar to the standard Chern-Simons term for the Yang-Mills field in three
dimensions [5].

6. - F u r t h e r g e n e r a l i z a t i o n s and development.

There is a wide field for further generalizations and applications of the suggested
theory. Here we would like to mention some, in our opinion, most interesting.

6"1. Generalization of gauge group. - The gauge group in sect. 3 (see (3.9)) is the
super-generalization of the unitary group U(N). In general case, one can take any
group g (orthogonal, simplictic, etc.), which satisfies the condition

(6.1) g* ~g = ~,

where * and r~ are appropriate conjugation and invariant metric tensor. Then one can
determine the group on differential forms as fol]ows:

(6.2) ~G*/~ vG = r~,

which will give us the superization of group g.

Moreover, there is a further generalization of (6.2)[13]. Indeed, let P be an
inhomogeneous differential form with values in a matrix space. Let us consider the
differential form G, which satisfy the condition

(6.3) ~G* A P A G = P .
The differential form P plays the role of fundamental form on a manifold M which
defines the scalar product for m a t t e r fields. In a general case there are two natural
fundamental forms: zero-form and D-form of volume. If P is a zero-form Pij, one
obtains the definition (6.2). On the other hand, if P is a volume form, then G in (6.3)
has only zero-rank component G(o). Hence (6.3) is equivalent to eq. (6.1) in this case.
NON-ABELIAN GAUGE ANTISYMMETRIC TENSOR FIELDS 1287

Notice that in these two described cases (a zero-form P and a volume-form P) the
form P belongs to the cohomology space H(~ and H(D)(M), respectively. In a
general case, one may assume that P is closed form, d P = 0, and represents the
relevant cohomology classes
D

(6.4) pij= k~_o=p~)k), p~k) ~H(k)(M).

For example, if M is a compact K/ihler manifold, then the even cohomology groups
H(2k)(M) are not trivial[14]. So the construction, (6.3), (6.4), has a non-trivial
realization.
This superization of a group g (6.1) seems to be interesting since it inherits the
properties of initial classical group g (6.1) and feels the geometry of the underlying
manifold M D.

6"2. Quantization and new topological invariants. - It is well known that the
quantization of diffeomorphism-invariant metric-independent actions reproduces
topological invariants in terms of quantum expectation values [6, 7]. One obtains the
Ray-Singer torsion when quantizing AGATF[7] and linking numbers for the
BF-system described in sect. 5 [6]. All these theories are particular cases of the
above-considered NGATF theory. There exist two kinds of actions for NGATF:

(6.5) (F, F ) = ~ Tr (_P A F )

J
M D

and a wide class of the Chern-Simons actions

(6.6) ~(k)
~CS ----
~ ~r(k)
CS "

M D

Both these actions are metric independent, so one can expect that the topological
invariants arise as path integrals with any of these actions:

(6.7) Z((91, ..., (gj) = I ~)A (91... •j exp [~ S ~ A ~~ ~

where (91, ..., d)j are gauge-invariant operators. In the Abelian case, one usually takes
as Oi integrals over the k-dimensional cycles r(k)[6]:

(6.8) (9 = A(k).
*2
)'(k)

For the one-form A(1 ) o n e has the gauge-invariant operator[5]

Lr(~)

which is the holonomy of the connection A(1) around the loop Y(1). In our paper [15] we
find the corresponding generalization of these formulae for NGATF.
1288 s. N. SOLODUKHIN

Notice that actions (3.21) for matter fields are also metric independent. Hence
they are also subject for quantization leading to topological invariants.

6"3. Physical models with NGATF. - The Abelian gauge antisymmetric tensor
fields play an important role in the supergravity, in the gauge theory of gravity and
in the string theory [1]. The NGATF theory is the generalization of both AGATF and
the standard gauge fields. Hence NGATF may find a wide application.
As it was shown in sect. 3, the supergroup appears naturally in our approach. In
fact, it is not necessary to introduce additional Grassman variables in the theory. The
superspace has a natural realization as a space of variables x :~, dx ~, i.e. the space
A * (M). Hence, this approach is appropriate for the description of supergauge theory.
For instance, one may attempt the construction of the super-Poincar~ gravity in the
framework of the gauge approach with NGATF. In particular, it allows to write the
gauge gravity action without the help of metric using only the gauge fields. Such a
theory would describe a phase in which general covariance is unbroken. The metric
will appear as a Goldstone boson of a spontaneously broken (local) general covariance,
as discussed in[16].
It should be noted here that the possibility to describe gravity with the help of
SL(3)-algebra-valued two-form instead of metric was also considered in[17].
Another promising application of NGATF is the string field theory. There an
approach was developed which gives a non-perturbative description of strings with
the help of an infinite-dimensional analog of the Chern-Simons action [18]. One may
consider the corresponding gauge field as the inhomogeneous differential form on the
space of string configurations. The appropriate gauge group would be the
diffeomorphism group Diff (S 1) probably together with the Kac-Moody extension of
the Poincar~ group [19].
One may also expect the NGATF to play a role in condensed-matter systems
where the standard Chern-Simons approach has proved to be very useful [20].

I am grateful to Yu. N. Obukhov for valuable discussions and careful reading of

the manuscript. I also thank D. Leites, V. Serganova, D. Fursaev and O. Zakharov
for conversations.

APPENDIX

Let us obtain the general solution for the equation

(A.1) ~G* A G = 1
which determines the gauge group element G:

In a general case, G can be represented in the form

(A.3) G = go h,
N O N - A B E L I A N GAUGE A N T I S Y M M E T R I C T E N S O R F I E L D S 1289

where go is a matrix-valued zero-form, and h is a matrix-valued inhomogeneous

differential form with a zero-rank component equal to h(o) = 1.
Inserting (/L3) into (A.1) we assume that the matrix go is unitary:

(A.4) grog o = 1 .

Then, we obtain the equation for h:

(A.5) ~h* A h = 1, h(o) = 1.

Any form h with h(0) = 1 is decomposed into a sum

(A.6) h -- 1 + H + ,~',
where the forms H and ~ - are both of rank /> 1 and satisfy the conditions

(A.7) ~H* = - H , i~J-* = .~7-.

F r o m (A.6) and (A.7) we obtain that

(A.8) h - 1 = ~h r = 1 - H + v~-.

Inserting this expression into (A.5) one finds

(A.9) ~ - A ~:~+ 2 , ~ - - H A H = HA Z- JZAH.

It is a consequence of eqs. (A~5)-(A.8) that

(A.10) H = _1 (g _ g - 1 ) , /~'=: ~1 (g + g -1)

-- 1.
2
One concludes from this that H and 5~-comraute:
(A.11) H A ~ = ~-A H.
The last equation and conditions (A.7) suggest representing ,~-as an even polynomial
of H:

(A.12) ~J-= ~ a k H 2k .
k
Substituting (A.12) into (A.9) and taking into account (A.11), one gets

(A.13) ~ , 2 a k H 2k + ~ a p a , ~ H 2(p+'~) - H ~ = O.
k p, m

Then, one obtains the following r e c u r r e n t definition for ak:

1 1 1 xk- ~ 1
(A.14) al = - - , a2 - , ak = -- -- 2~ ap ak - p, k >I 2 .
2 8 2 p=l

Then the p a r a m e t e r space of the group G (A.1) consists of the parameters of the
unitary matrix go (A.4) and independently of differential form

(A.15) H = ~ (i)k(k - 1)/2

->1 k! H~l...~k~. a d x ~1A ... A d x ~k ,

where { ~ } is a set of anti-Hermitian matrices.

1290 s. N. SOLODUKHIN

I t is easy to obtain the group multiplication law for H . We should note here that
this is non-Abelian even in the Abelian case (when go e U(1)).
L e t us consider in brief the case of the Lorentz group. The gauge group is
determined in this case b y the condition

(A.16) ,zGT A vG = V,

where r~ij= d i a g ( + l , . . . , - 1 ) is the metric tensor, and G = ~ G~k)j is the

matrix-valued differential form. The element G (A.16) can also be r e p r e s e n t e d in the
form

(A.17) G = g o ( 1 + H + ~7-),

where go is the Lorentz group element (gWr~go = V), and H and ~ satisfy

(A.18) {Hij = - Hji , ~ = ~ji ,

where we denote Hij = H~.rjik.

As earlier the form ~ can be r e p r e s e n t e d as a polynomial of H (A.12). F r o m
(A.18) one obtains for the k-rank component H(k):

(A.19) H~k) = _ ( - 1)k(k - 1)/2H~k) "

In the four-dimensional case we have

H(,jk): - ) , k = 0, 1, 4,
(A.20)
HJ:), k : 2, a .

Hence H is not decomposed with respect to the basis of the Lorentz group generators,
as we had in the unitary group case (see eq. (A.15)). One should consider additionally
the s y m m e t r i c matrices basis for H (k), k = 2, 3. This is the main difference in the
definition of the group G for complex and real cases.

REFERENCES

[1] E. NAMBU: Phys. Rep. C, 23, 250 (1976); E. SEZGIN and P. VAN NIEUWENHUIZEN:Phys.
Rev. D, 22, 301 (1980); W. SIEGEL: Phys. Lett. B, 93, 170 (1980); P. K. TOWNSEND: Phys.
Lett. B, 90, 275 (1980); Yu. N. OBUKHOV:Phys. Lett. B, 109, 195 (1982); R. ROHM and
E. WITTEN: Ann. Phys. (N.Y.), 170, 454 (1986); D. OLIVIER:Phys. Rev. D, 33, 2462 (1986);
C. TEITELBOIM: Phys. Lett. B, 167, 63 (1986); J. M. RABIN: Phys. Lett. B, 172, 333
(1986).
[2] D. IVANENKO and L. LANDAU: Z. Phys., 48, 340 (1928); E. KAHLER: Rend. Math., 21,
4255 (1962); W. GRAF: Ann. Inst. H. Poincard, 29, 85 (1978); P. BECHER and H. Joos:
Z. Phys., C15, 343 (1983); I. M. BENN and R. W. TUCKER: Commun. Math. Phys. 89, 341
(1983); D. D. IVANENKOand Yu. N. OBUKHOV:Ann. Phys. (Leipzig), 42, 59 (1985); D. D.
IVANENKO, YU. N. OBUKHOV and S. N. SOLODUKHIN: Preprint IC/85/2. (ICTP, 1985).
J. A. BULLINARIA: Ann. Phys. (N.Y.), 168, 301 (1986); Yu. N. OBUKHOV and S. N.
SOLODUKHIN: Preprint IC/91/147. (ICTP, 1991).
[3] S. N. SOLODUKHIN:Ann. Phys. (Leipzig), 46, 439 (1989); Int. J. Theor. Phys., 31, 47
(1992).
NON-ABELIAN GAUGEANTISYMMETRICTENSOR FIELDS 1291

[4] S. DONALDSON: J. Diff. Geom., 18, 269 (1983); 26, 397 (1987); Topology, 29, 257
(1990).
[5] E. WITTEN: Commun. Math. Phys., 121, 351 (1989).
[6] G. T. HOROWITZ: Commun. Math. Phys., 125, 417 (1989); G. T. HOROWITZ and M.
SREDNICKI: Commun. Math. Phys., 130, 83 (1990).
[7] A. S. SCHWARZ:Lett. Math. Phys., 2, 247 (1978)
[8] P. K. TOWNSEND:in Supergravity, edited by P. VAN NIEUWENHUIZENand D. Z. FREEDMAN
(North-Holland Publishing Company, 1979); D. 21. FREEDMAN and P. K. TOWNSEND:Nucl.
Phys. B, 177, 282 (1981); R. I. NEPOMNECHIE: Nucl. Phys. B, 212, 301 (1983).
[9] P. F. PICZEN: J. Phys. A, 16, 3457 (1983).
[10] R. C. MYERS and V. PERIWAL: Phys. Lett. B, 225, 352 (1989).
[11] E. WITTEN: Nucl. Phys. B, 323, 113 (1989).
[12] D. BIRMINGHAM,M. BLAU and G. TOMPSON:Int. J. Math. Phys. A, 5, 4721 (1990); M. F.
ATIYAH and L. JEFFREY: J. Geom. Phys., 7, 119 (1990); C. H. TAUBES:J. Diff. Geom., 31,
301 (1990).
[13] I thank Yu. N. OBUKHOVfor this remark.
[14] S. KOBAYASHIand K. NAMIZU: Foundations of Differential Geometry, Vol. 2 (Interscience
Publishers, New York, N.Y., 1969).
[15] Yu. N. OBUKHOVand S. N. SOLODUKHIN:Non-Abelian gauge differential forms and their
holonomies, preprint MGU, Moscow, 1993, to be published.
[16] E. WITTEN: Commun. Math. Phys., 117, 353 (1988).
[17] G.'T HOOFT: Nucl. Phys. B, 357, 211 (1991).
[18] E. WITTEN: Nucl. Phys. B, 268, 253 (1986); J. L. GERVAIS: Nucl. Phys. B, 287, 815
(1987).
[19] M. J. BOWlCK and S. G. RAJEEV: Nucl. Phys. B, 293, 348 (1987); Y. BARS and S.
YANKIELOWICH: Phys. Rev. D, 35, 3878 (1987); L. CASTELLANI:Phys. Lett. B, 206, 47
(1988); Nucl. Phys. B, 317, 46 (1989).
[20] Y. N. CHEN, B. I. HALPERIN, F. WILCZEK and E. WITTEN: Int. J. Mod. Phys. B, 3, 1001
(1989); J. D. LYKKEN, J. SONNENSCHE~N and N. WEISS: The theory of anyonic
superconductivity: a review, FERMILAB-PUB-91/41-T, January 1991.