You are on page 1of 6

LECTURE II-8, PART II: ABELIAN DUALITY IN FOUR

EDWARD WITTEN

1. Duality and Sl(2; Z)

In this second part of lecture II-8, we discuss abelian duality in four dimensions, and
give an application to an Sl(2; Z) symmetry of the free U (1) theory in four dimensions.
We postpone discussion of Sl(2; Z) symmetries of non-free theories to a later lecture,
since all known examples of that involve supersymmetry.
We work with a U (1) bundle L on a 4-manifold M , and a connection A on L, whose
curvature is F =FA. The gauge theory Lagrangian (in Euclidean signature) including the
topological term is
L(A) = 4e2 d x gFmnF + 162 d4x pgmnpq F mnF pq
p i
Z Z
1 4 mn

1 Z
i Z (1.1)
= 2e2 FA ^  FA + 42 FA ^ FA:
We have used the standard normalization on the kinetic term, and have normalized the
topological term so that replacing  by  +2 does not changeR the physics. (This property
of the topological term derives from the fact that c1(L)2 = (FA=2)^(FA=2) is always
an integer. Notice that on a spin manifold, c1(L)2 is always an even integer, and we
gain an additional equivalence under replacement of  by  + .)
Let  =  + 2ei2 2 h. As we have just observed,  7!  + 2 is a symmetry of this
theory, and  7!  + 1 is a symmetry when working on a spin manifold. To extend this
to an Sl(2; Z) action (in the spin manifold case) we also need a symmetry which maps
to 1= ; this will be given by a duality transformation FA \$  FC (with C being a new
\dual" connection).
The computations for this duality transformation are similar to those in lecture II-7.
We begin by de ning F = 12 (FA   FA), and rewriting our Lagrangian (1.1) as
i Z
i Z
L(A) = 4 F+ ^ F+ + 4 F ^ F
(1.2)
= 4i kF+k2 4i kF k2:
Z Z

Date : 6 March 1997.

 Notes by David R. Morrison.
1
2

Letting G denote the gauge group associated to A, the partition function for this theory
can be written as
XZ
Z ( ) = vol(1 G )
R R
DA e 4i kF+ k2+ 4i kF k2 : (1.3)
L
Our earlier examples of duality began with a theory of a scalar eld  which entered
into the Lagrangian only through its derivative d so that the theory had a symmetry
under  7!  + c (with c constant); the rst step in the duality transformation was to
gauge this symmetry, introducing also an appropriate Lagrange multiplier eld.
The present theory is already a gauge theory, being a theory of a connection A which
enters into the Lagrangian only through its curvature FA, so that there is a symmetry
under A 7! A + B (with B a at connection). We want to do the analogue of gauging
this symmetry, by allowing B to be an arbitary connection on an arbitrary bundle,
introducing a kind of \exotic gauge eld" G which is a 2-form eld, and extending the
symmetry to
A!A+B
G ! G + FB : (1.4)
Then F := FA G plays the role of the \gauge-invariant" quantity, analogous to the
covariant derivative of a scalar eld. It is to be stressed that two G elds will be
considered gauge-equivalent if they di er by G ! G + FB for FB the curvature of any
connection on any line bundle. In our analysis, we will assume for simplicity that there
is no torsion in H 2(M ).
We need a \gauge-invariant" extension of our Lagrangian. We might try
L(A; G) = 4i kF+ k2 4i kF k2;
Z Z
(1.5)
but this is too simple (because, for example, we could gauge F to zero). To improve this,
we introduce a new connection C on a line bundle N, with curvature FC , and consider
the Lagrangian
L(A; G; C ) = 4i kF+k2 4i kF k2 2i FC ^ G:
Z Z Z
(1.6)
The partition function for this new theory can be represented as a path integral, which
includes sectors associated to all choices of bundles L and N:
1 1 1 X Z DA DG DC e 4i R kF+k2+ 4i R kF k2+ 2i R FC ^G ;
vol(Ge) vol(G ) vol(GC ) L;N (1.7)
where G and GC denote the gauge groups associated to A and C , respectively, and Ge
denotes the \exotic" gauge group.
To see that this new theory is equivalent to the original one, we rst do the C -integral
in (1.7): write C = C0 + C 0, for C0 a xed connection on the line bundle N. Then the
3

C 0 integral is
1 Z DC 0 e 2i R C0 ^dG = (dG): (1.8)
vol GC
Thus, when we sum over N we nd
1 X Z DCe 2i R FC ^G = X ei(x;G)(dG) = ( G  2 Z)(dG):
vol GC N x2H 2 (M ) 2 (1.9)
The conditions that dG = 0 and that [G=2] is an integral class precisely mean that G is
of the form FB for some connection on some line bundle and hence that G can be gauged
to zero. After doing this, it follows that the partition function (1.7) coincides with Z ( ),
and we recover the original theory.
Alternatively, we can evaluate the partition function (1.7) by gauging A to 0, using
the \exotic" gauge invariance (which has an ordinary gauge invariance as an ambiguity).
This leaves the path integral
1 X Z DG Z DC e 4i R kG+k2+ 4i R kG k2+ 2i R FC ^G : (1.10)
vol GC N
To evaluate the G integral, we complete the square, bearing in mind that
Z Z
FC ^ G = (FC + ^  G+ FC ^  G ): (1.11)
In fact, if we de ne G0 = G 1F
C+ + 1 FC , then we can write the exponent from
eq. (1.10) as
i Z kG0 k2 + i Z kG0 k2 + i Z kF k2 i Z kF k2:
C+ C
4 + 4 4 4 (1.12)
When we carry out the G0 integral, the rst two terms give a Gaussian integral which
contributes to the overall normalization; integrating out G0 leaves the path integral
1 X Z DC e 4i R kFC + k2 4i R kFC k2 : (1.13)
vol GC N
This is the same as the original path integral, but with  replaced by 1= , precisely
what we wanted to show.
As we did in the case of two dimensions, it is possible to analyze the  -dependence of
the normalization of the path-integral, and obtain further interesting results. Some hint
of the avor of the results to be obtained this way is seen if we evaluate the Gaussian
integral indicated above, which yields
!n2+ !n2
p2  p 2  ; (1.14)
i i
where n2 denote the numbers of self-dual and anti-self-dual 2-forms. Of course, these
numbers are in nite, so there must be some cancellation against other normalization
4

factors. When this is worked out in detail,1 the result is found to be

Z ( ) =  +4    4  Z ( 1= ); (1.15)
where  and  are the Euler number and signature of M , respectively. Thus, the partition
function Z ( ) is actually a modular form for Sl(2; Z) (or for a subgroup, when the
manifold is not spin) of weight ( +4  ;  4  ).
We can also follow certain operator insertions through the duality transformation, as
we did in lower dimensions. An insertion of F in the original theory can be realized by
inserting F = F G in the extended theory, which can be written
F = F G0 1 F ; or F = F G0 + 1 F ;
+ + +  C+
C (1.16)
respectively, after making the change of variables to G0. Thus, when we gauge A to zero,
and integrate out G0, we are left with operator insertions proportional to FC  , namely:
F+ 7! ( 1= )FC +; and F 7! (1= )FC : (1.17)
Notice that as a consequence of the  -dependence of these mappings, a correlation func-
tion involving insertions of F+ and F will have a di erent modular weight than that of
the partition function.
2. The Hamiltonian formalism
Returning to the case that the gauge group is U (1), let us brie y discuss abelian four-
dimensional duality in a Hamiltonian framework. Take a 4-manifold of the form M3 R,
where R is a timelike direction. Note that this is a spin manifold, so we expect full
Sl(2; Z) symmetry. For simplicity we suppose that there is no torsion in H1(M3). Each
class x 2 H 2(M3) determines a complex line bundle Lx on the 3-manifold M3 (satisfying
c1(Lx) = x). The Hilbert space for our theory on the 3-manifold M3 can be written in
the form
M
H ( M3 ) = Hx ; (2.1)
x2H 2(M3 ;Z)
where Hx is the Hilbert space which comes from quantizing connections on Lx. (On
the left, we have explicitly indicated the dependence on the coupling constant  .) To
construct Hx, write an arbitrary connection in the form A = A0 + , where A0 is a
harmonic connection (a connection whose curvature is a harmonic two-form) and is
a 1-form which is co-closed. Let Tx be the space of harmonic connections on the line
bundle Lx. Then the quantization yields
Hx = H
L2(Tx): (2.2)
Here H is a Hilbert space obtained by quantizing the space of 's, and L2(Tx) is just
the space of L2 functions on Tx.
1 E. Witten, On S -duality in abelian gauge theory, Selecta Math (N.S.) 1 (1995), 383{410.
5

Note that the factor H is independent of x, since the space of co-closed one-forms
is de ned with no reference to x. Duality maps H to itself while acting separately on
H0 = xL2(Tx). The duality action on H follows from the operator mapping in (1.17).
The action of duality on H0 can be described as follows. Note that the Tx's are all
principal homogeneous spaces acted on by the torus H 1(M3; R=Z), which parametrizes
at line bundles on M3; the action is de ned by tensoring any given line bundle with
connection by a at line bundle determined by a point in H 1(M3; R=Z). Let y denote a
character of the abelian group H 1(M3; R=Z). There is a decomposition L2(Tx) = y Tx;y ,
where Tx;y is the subspace of L2(Tx) transforming in the character y. Each Tx;y is one-
dimensional. Hence
H0 = x;y Tx;y (2.3)
Note that, by Poincare and Pontryagin duality, the character group of H 1(M3; R=Z) is
H 2(M3; Z). Thus, x and y take values in the same space. It is hence possible to exchange
them, and this is what the  ! 1= transformation does (more precisely, it acts by
(x; y) ! ( y; x)). Thus duality exchanges a classical notion { the decomposition with
respect to x { with a quantum notion { the decomposition with respect to y. The claim
about how the duality acts will be justi ed below where we introduce the operators QE
and QM .
Upon quantization|and suppressing  for a moment|one writes the four-dimensional
curvature as FA0 + e2Adt, where FA0 is a two-form on M3 and A|a one-form on M3|is
the momentum conjugate to the connection A. The Hamiltonian becomes
Z
1
H = 2e2 FA20 + e2 rA0 + H ( ):
2
(2.4)
Here H ( ) is the part of the Hamiltonian that acts on H . The other terms act on H0.
The rst term is the magnetic energy of the harmonic connection R
A0; it comes from the
part of the Lagrangian quadratic in FA and is a multiple of M3 x^  x. The second term,
0
which comes from the part of the Lagrangian
R
quadratic in A, is the electric energy, the
Laplacian on Tx; it is a multiple of M3 y^  y.
Including the  term shifts the quantization. In fact, the canonical momentum FA_ =
 A as determined from the original Lagrangian (1.1) is
FA_ = 2i FS = 2i  F  F (2.5)
A e2 A  A
At non-zero , one has not H0 = xL2(Tx) but H0 = x L2 (Tx; S ), where S is a certain
at line bundle over Tx and L2 is the space of L2 sections. I leave it as an exercise to
the reader to identify S . Of course, S is trivial at  = 0 and only depends on  modulo
2 .
Rewriting the formula for FA_ in terms of  , we can determine how this operator
transforms under  ! 1= . Indeed, under the operator mapping (1.17) one gets
FA_ = F+ + F 7! FC + + FC = FC : (2.6)
6

We will also need the \dual" version of this computation:

FA = F+ + F 7! ( 1= )FC + ( 1= )FC = FC_: (2.7)
In the Hamiltonian formalism, for any 2-cycle   M3 we can de ne associated \elec-
tric" and \magnetic" operators on the Hilbert space H (M3) for any 2-cycle   M3,
by
Z
F _ Z FC
QE () =  2 =  2
A
Z
F Z
F _ (2.8)
A
QM () =  2 =  2 : C

(These operators only depend on the class of  in H2(M3; Z).) Clearly, under  ! 1= ,
that is under A ! C , one has QE ! QM , QM ! QE . Since x and y are the eigenvalues
of QM and QE , this means that  ! 1= acts by (x; y) ! ( y; x). Moreover, from the
explicit formula (2.5) for FA_, one sees that when  is increased by 2 (an operation which
leaves the Hilbert space unchanged), the operator QM is unaltered, but the operator QE
maps to QE + QM .
The statements made in the last paragraph can be combined to the following: Sl(2; Z)
acts on H0 via the natural action of Sl(2; Z) on the pair (x; y).