hep-th/9509162

KUCP-0083
September 1995

Whitham-Toda Hierarchy
arXiv:hep-th/9509162v1 28 Sep 1995

And
N = 2 Supersymmetric Yang-Mills Theory

Toshio NAKATSU 1 and Kanehisa TAKASAKI 2

1 Departmentof Mathematics and Physics, Ritsumeikan University

Kusatsu,Shiga 525-77,Japan
2 Department of Fundamental Sciences, Faculty of Integrated Human Studies, Kyoto University

Yoshida-Nihonmatsu-cho,Sakyo-ku,Kyoto 606,Japan

ABSTRACT

The exact solution of N = 2 supersymmetric SU (N ) Yang-Mills theory is studied in the

framework of the Whitham hierarchies. The solution is identified with a homogeneous solution
of a Whitham hierarchy. This integrable hierarchy (Whitham-Toda hierarchy) describes mod-
ulation of a quasi-periodic solution of the (generalized) Toda lattice hierarchy associated with
the hyperelliptic curves over the quantum moduli space. The relation between the holomorphic
pre-potential of the low energy effective action and the function of the (generalized) Toda lattice
hierarchy is also clarified.

E-mail: [email protected], [email protected]

1
 Recently Seiberg and Witten [1] obtained exact expressions for the metric on the moduli
space and the dyon spectrum of N = 2 supersymmetric SU(2) Yang-Mills theory by using
a version of the Olive-Montonen duality [2] and holomorphy [3] of 4d supersymmetric the-
ories. Their approach has been generalized to the case of other Lie group [4],[5]. Especially
surprising in these results is unexpected emergence of elliptic (or hyperelliptic) curves and
their periods. Although these objects appear in the course of determining the holomorphic
pre-potential F of the exact low energy effective actions, physical significance of the curves
themselves is unclear yet. It will be important to clarify their physical role. An interesting
step in this direction has been taken [6] from the view of integrable systems, in which
the correspondence between the Seiberg-Witten solution [1] and the Gurevich-Pitaevsky
solution [7] to the elliptic Whitham-KdV equation [8] is pointed out.
In this article we shall consider the exact solution of N = 2 supersymmetric SU(N)
Yang-Mills theory in the framework of the Whitham hierarchies [9] and show that it can be
identified with a homogeneous solution of a Whitham hierarchy. This Whitham hierarchy
(Whitham-Toda hierarchy) turns out to be modulation equations [10] of a quasi-periodic
solution of the (generalized) Toda lattice hierarchy associated with the hyperelliptic curves
over the moduli space of Yang-Mills theory. In particular the relation between the pre-
potential of this supersymmetric theory and the function of the (generalized) Toda lattice
hierarchy will be clarified.

1
One of the main ingredients in the analysis [1], [4] of the exact solution of N = 2 super-
symmetric SU(N) Yang-Mills theory is the meromorphic differential dS

x dPdx(x)
dS = dx (1)
y
on the family of the hyperelliptic curves
N 2
C : y 2 = P (x)2 2N , P (x) = xN + uN k xk .
X
(2)
k=0

Each uk is an order parameter of N = 2 supersymmetric SU(N) Yang-Mills theory. u =

(u2 , , uN ) become the parameters of the flat moduli. is the lambda-parameter of this

2
gauge theory and we henceforth fix its value. The spectrum of excitations in the theory
will be measured by the units
I I
ai = dS , aD,i = dS (1 i N 1), (3)
i i

where i and i (1 i N 1) are the standard symplectic basis of homology cycles of

the hyperelliptic curve C. In particular each i is a cycle which encircles counterclockwise
the cut between two neighboring branch points in xplane. Let p and p be the two
points at infinity of the Riemann sheets of C : x(p ) = x(p ) = . Since the pole divisor
of dS (1) is 2p + 2p and it has no residues at these points, one can decompose it into
the sum :
N
X 1
dS = dX,1 + ai dzi , (4)
i=1

where dzi (1 i N 1) is the holomorphic differential normalized by the condition

H
i dzj = i,j . dX,1 is the meromorphic differential of second kind (its second order poles
H
are at p and p ) with vanishing periods along the -cycles, that is, i dX,1 = 0 for

i.
Let us investigate the role of dS from the view of integrable system. For this purpose
we introduce the following meromorphic functions h and h on the curve C,

h = y + P (x) , h = y + P (x) , (5)

and then consider the effect of infinitesimal deformation of the moduli parameters u with
h (or h) being fixed. Under this condition one can obtain

 xk
dS 
= dx . (6)
uN k y


f ix h

After changing the moduli parameters from u to a = (a1 , , aN 1 ) the above equation
reads


dS = dzi , (7)

ai


f ix h

which implis that dS satisfies the following system of differential equations:

dzj = dzi (1 i, j N 1), (8)
ai aj

3
where the derivation by the moduli parameters a is understood to be partial derivation fix-
ing h and the other as. Differential equations of this form (8) were first derived by Flashka,
Forest and MacLaughlin [11] as modulation equations of quasi-periodic solutions in soliton
theory. The concept of modulated quasi-periodic solutions originates in Whithams work
on the KdV equation [8], and because of this, this type of systems are called Whitham
equations. Remarkably, Whitham equations themselves are integrable systems. In par-
ticular, as soliton equations admit infinitely many commuting flows ( which constitute an
integrable hierarchy ), the associated Whitham equations also have infinitely many extra
commuting flows [9]. These commuting flows are generated by a set of meromorphic dif-
ferentials {dA }AI , where flows are labeled by indices A, B, , and the hierarchy can be
written as integrability conditions of those differentials. In the above case, these equations
are given by

dzj = dzi , dzi = dA , dB = dA , (9)
ai aj TA ai TA TB
where, as in (8), the derivation by the time variables mean derivation fixing h.

2
Now we specify the meromorphic differentials dA . They consist of two types of meromor-
phic differentials d,n and d,n of second-kind (n 1): d,n (d,n ) has a pole of
order n + 1 at p (p ) and is holomorphic elsewhere. We also introduce a meromorphic
differential d,0( d,0 ) of third-kind : d,0 has simple poles at p and p with
resp d,0 = resp d,0 = 1 and is holomorphic elsewhere. All these differentials are
H H
normalized such that they have no periods along any i -cycle ; i d,n = i d,n = 0
for i, and determined by h and h through the following prescription. Let us define local
coordinates z and z in neighborhoods of p and p as
N N
z = h1 , z = h1 . (10)
1
Because the divisors of h and h are respectively N p Np and Np N p it follows
that z (p ) = 0 and z (p ) = 0. In a neighborhood of p , they can be written
n1 m1
X
d,n = {nz qm,n z }dz ( n 1 ),
m1
1
This follows from the relation: hh = 2N .

4
 1 m1
X
d,n = {n,0 z rm,n z }dz ( n 0 ), (11)
m1

while, in a neighborhood of p ,

1 m1
X
d,n = {n,0 z rn,m z }dz ( n 0 ),
m1
n1 m1
X
d,n = {nz qm,n z }dz ( n 1 ). (12)
m1

Integrability conditions (9) clearly ensure the existence of a differential dS which satis-
fies

dS = dzi , dS = d,n , dS = d,n , dS = d,0 , (13)
ai Tn Tn T0
where 1 i N 1 and n 1. We denote time variables of the flows generated by d,n ,
d,0 and d,n as Tn , T0 and Tn respectively. Now let us construct a function F (a, T, T )
from this differential dS by the following equations
F 1 aD,i
I
= dS ( ),
ai 2 1 i 2 1
F n F n
= resp z dS , = resp z dS ,
Tn T n
F
= resp ln z dS + resp ln z dS , (14)
T0
where 1 i N 1 and n 1. Notice that consistency of this definition of the F -function
can be checked by using integrability conditions (9) and the Riemann bilinear relation for
meromorphic differentials [12]. As an example let us prove the equality:
!
1 
I 
n
dS = resp z dS (n 1). (15)
Tn 2 1 i ai
Rp
In fact, by setting ,n (p) = d,n , the L.H.S of eq.(15) can be transformed as
!
1 1
I I
dS = d,n
Tn 2 1 i 2 1 i

1 NX1
(I I I I )
= d,n dzi dzi d,n
2 1 j=1 j j j j

= resp ,n dzi
n
= resp z dzi ,

5
which is equal to the R.H.S of eq.(15). We also notice that, owing to definition (14), the
local behaviors of dS can be described by the F -function as

X
n1 1
X F n1
dS = nTn z + T0 z z dz around p ,

n1 Tn

n1

X
n1 1 F n1 X
dS = nTn z T0 z z dz around p . (16)

n1 Tn

n1

An interesting class of solutions of the Whitham hierarchy are solutions that enjoy the
homogeneity condition
N 1
X F X F X F
ai + Tn + Tn = 2F. (17)
i=1 ai n0 Tn n1 Tn

In this case, dS, which is introduced by eq.(13), has a form analogous to (4)
N
X 1 X X
dS = ai dzi + Tn d,n + Tn d,n . (18)
i=1 n0 n1

Therefore one can reproduce dS of (1) by simply setting T1 = T1 = 1 and other TA -

variables zero. The pre-potential F of N = 2 supersymmetric SU(N) Yang-Mills theory

is now given by F = 2 1F . Let us prove that (18) is consistent with (16). We first
notice that eqs. (13) and (14) make it possible to express the coefficents in (11) and (12)
in terms of F . As for the holomorphic differentials dzi let us write their local expansion as
m1
X
dzi = i,m z dz around p ,
m1
m1
X
dzi = i,m z dz around p , (19)
m1

where 1 i N 1. Then, by using eqs.(13) and (14), we can also express the coefficients
in (19) in terms of F . Inserting these local expansions into the R.H.S of (18), and recalling
the homogeneity of F (17), one can reproduce (16).
For this homogeneous solution, we can write the F -function in terms of the -periods of
the differentials dzi , d,n and d,n as follows. First notice that by homogeneity relation
(17) and definition of F -function (14), F can be written as a sum of periods and residues
of dS

1 NX
1
F X F X F
F = ai + Tn + Tn
2 i=1 ai n0 Tn n1 Tn

6

1 NX
1
a
I
i
X
n
= dS Tn resp z dS
2 i=1 2 1
i n1


n
X
Tn resp z dS T0 (resp ln z dS resp ln z dS) . (20)
n1

Since dS has the form given in (18), this expression of F can be evaluated further to become
the aforementioned form. Eventually, we obtain

N 1 N 1
1 X X X  
F = i,j ai aj + ai i,k Tk + i,k Tk + i,0 T0
4 1 i,j=1 i=1

k1

1 X 1 X 1
r0,0 T02
X
+ qk,l Tk Tl + rk,l Tk Tl + qk,l Tk Tl +
2 k,l1 k,l1 2 k,l1 2
X 
+T0 rk,0Tk + r0,k Tk , (21)
k1

where the quantities i,j , i,0 and r0,0 are introduced as

I
i,j = dzj (1 i, j N 1),
i
1
I
i,0 = d,0 (1 i N 1),
2 1 i
r0,0 = resp ln z d,0 + resp ln z d,0 , (22)

and other quantities in (21) are those appearing in expansions (11),(12) and (19).

3
Now we will discuss the relation between this homogeneous solution of the Whitham hi-
erarchy (9) and the Toda lattice hierarchy. Let us denote Toda lattice time variables by
t = (t1 , t2 , ) , t = (t1 , t2 , ) and n (n Z). It is well-known [13] that the Toda lattice
hierarchy has a quasi-periodic solution associated with a hyperelliptic curve C as in (2). In
fact, this is a dimensionally reduced solution in the lowest sector of the hierarchy, the
solution gives an N-periodic solution of the Toda chain rather than the two-dimensional
Toda field equations 2 . Since the flows of the Toda lattice hierarchy do not change this
curve itself, the moduli parameters u are invariants (integrals of motion) of these solutions
of the Toda lattice hierarchy.
2
This dimensional reduction is irrelevant in our discussion.

7
 We introduce an additional new time variables = (1 , , N 1 ) into this solution.
This is achieved by modifying the associated Baker-Akhiezer function [13] as

(p; t, t, n, )

Z p X Z p X Z p N
X 1 Z p
= exp n d,0 + tn d,n + tn d,n + 1 i dzi

n1 n1 i=1
 
1 PN 1
z(p) z(D) + n(z(p ) z(p )) + n1 tn n + n1 tn n + i=1 i i
P P
2
 1 N 1

z(p ) z(D) + n(z(p ) z(p )) + n1 tn n + n1 tn n + 2 i=1 i i
P P P

(z(p ) z(D) + )
. (23)
(z(p) z(D) + )
Rp
z is the Abel mapping, that is, z(p) =T (z1 (p), , zN 1 (p)) where zi (p) = dzi . D is
a positive divisor of degree N 1 and is the Riemann constant. l , l (l 1) and i
(1 i N 1) are N 1 dimensional complex vectors whose components are given by
l =T (l,1 , , l,N 1 ), l =T (l,1 , , l,N 1 ), and i =T (i,1 , , i,N 1 ). Notice that
Baker-Akhiezer function (23) reduces to that of the ordinary Toda lattice as 0. For
non-zero values of , is quasi-periodic along the -cycles,

i 1i
(p; t, t, n, ) 7 e (p; t, t, n, ), (24)

and periodic along the -cycles. Hence the Baker-Akhiezer function can also be regarded
as a section of a flat line bundle over C now labeled by -variables. We also notice that
this solution of the (generalized) Toda lattice is N-periodic (with respect to n). One can
check the following periodicity at the level of Baker-Akhiezer function

(p; t, t, n + N, ) = const. h(p) (p; t, t, n, ). (25)

One may introduce the -function of this solution following the prescription of Toda
lattice hierarchy [14]. In particular the local expansions of around p and p will be
described by the -function : In a neighborhood of p

n
P
t z l (t [z ], t, n, )
(p; t, t, n, ) = z e l1 l , (26)
(t, t, n, )
and in a neighborhood of p

n
P
t z l (t, t [z ], n + 1, )
(p; t, t, n, ) = z e l1 l , (27)
(t, t, n, )

8
 2 3
where [z ] = ( z1 , z2 , z3 , ).
By matching this expression of with (23), we can write down the -function in terms
of the theta function and the coefficients of local expansion of dzi , etc. We thus obtain the
following expression of the function :

(t, t, n, )
= eF (t,
t,n.)

1 NX
1 X X
(n 1)z(p ) + nz(p ) z(D) + + tn n + tn n + i i ,
n1 n1 2 i=1
(28)
3
where F is a polynomial of second degree in t, t, n and given by
N 1
1 X X 1 X 1 X
F (t, t, n, ) = qk,l tk tl + rk,l tk tl + qk,l tk tl i,j i j
2 k,l1 k,l1 2 k,l1 4 1 i,j=1

n(n 1) N
X 1 X
+ r0,0 + 1 i (i,k tk + i,k tk ) ni,0
2 i=1

k1

dk tk + d0 n.
X X X
n rk,0 tk (n 1)r0,k tk + dk tk +
k1 k1 k1
(29)

Note that the quadratic part of F (29) has almost the same form as F in (21). This is not
accidental. In fact it turns out that the Whitham hierarchy in the previous section gives
modulation equations of the aforementioned quasi-periodic solution of the (generalized)
Toda lattice hierarchy. Let us explain this relation in more detail following Bloch and
Kodama [15] 4 . For this purpose we introduce slow time variables 5
Tl , Tl (l 1), T0 and
ai (1 i N 1) by

Tl = tl , Tl = tl , T0 = n , ai = 1i . (30)

and consider the asymptotics of Baker-Akhiezer function (23) and -function (28) as 0.
In the slow time variables, they turn out to have the following expression
 
d ln p; T /, T /, T0 /, a/ 1 = 1 n dS (n) (p; T, T , T0 , a),
X

n0
3
In (29) dk (k 1) and dk (k 0) are the constants irrelevant to our discussion and we do not describe
their explicit forms.
4
Their Whitham hierarchy for the Toda lattice is different from ours, but the idea of deriving modulation
equations is the same.
5
is arbitary small. So these variables are small relative to the (generalized) Toda time variables.

9
  
ln T /, T /, T0 /, a/ 1 = 2 n F (n) (T, T , T0 , a).
X
(31)
n0

The leading-order terms dS (0) and F (0) are given by the same expressions as dS in (18)
and F in (21) respectively. We now allow the moduli parameters u to depend on the slow
variables,

uk = uk (T, T , T0 , a). (32)

In other words, the hyperelliptic curve C can now slowly vary as C = C(T, T , T0 , a).
We further require that this modulated
quasi-periodic wave be still a solution of the
(generalized) Toda lattice hierarchy. This induces a system of differential equations (mod-
ulation equations) to the moduli parameters. By the theory of Whitham hierarchies [9],
[10], [11], these modulation equations turn out to be nothing but our Whitham hierarchy
(9).

4
So far, we have considered the integrable structure of N = 2 supersymmetric Yang-Mills
theory by embedding the system into the Whitham hierarchy (9). The hierarchy is con-
structed by adding the additional T - and T -flows to the flows generated by the holomorphic
differentials. Meanwhile, the time variables a of the latter flows, along with their dual vari-
ables aD , constitute the units of spectrum in the theory. Though all these flows can be
identified with those of the (generalized) Toda lattice, physical roles of the additional ones
are not clear. In this final section we discuss their physical implication. In particular we
show that the solution of the Whitham hierarchy that we have considered satisfies the same
Virasoro constraints as appear in topological string theory.
For this purpose we define the following function Q by using the differential dS intro-
duced in (13) ,
dS
Q= . (33)
dh
Rp
Since S(p) = dS is multi-valued, Q can also be multi-valued in general. Actually,
it is known [9] that a class of solutions of the Whitham hierarchy are characterized by
geometric conditions concerning single-valuedness and regularity of Q. For an example

10
one can determine a solution of the Whitham hierarchy by the condition that Q be single-
valued and have no singular points other than p and p . This is indeed the solution
that we have considered in the preceding sections. Let us now impose a stronger condition
that this single-valued Q be regular at p , too. This regularity condition forces Tk = 0.
Furthermore, by the single-valuedness, it follows that

resp hn+1 QdS = 0 (34)

for n 1. This is a consequence of the Riemann bilinear relation. Since we already know
how the differentials hn+1 QdS behave around p , eqs.(34) can be rephrased as constraints
on the F -function. They are given by
X F 1 X F F 1 X
kTk + + klTk Tl n+1,0 = 0 (35)
k1 Tk+nN 2 k+l=nN Tk Tl 2 k+l=N
for n 1. Notice that we can make F independent of TkN (k 1) in (35) without any
loss of generality. This is because the corresponding generators become exact forms,

d,kN = dhk . (36)

Hence constraints (35) now become precisely same as those on the free energy of the
AN 1 topological string (in spherical limit) [16],[17]. In this topological string theory, time
variables such as Tn play the role of coupling constants of the chiral primary fields or their
gravitational descendants.
From this observation it would be a very fascinating idea to interpret the meromorphic
differentials d,n (n 1) as the chiral primary fields or their gravitational descendants
of the AN 1 topological string. It should be noticed that one can also give the same
interpretation on the differentials d,k (k 1). So it might be possible to interpret N=2
supersymmetric Yang-Mills theory as a coupled system of two topological string models.
As we have mentioned above, the corresponding solution of the Whitham hierarchy is
characterized by the condition that Q be single-valued and have singularities at most at
p and p . Again by the Riemann bilinear relations, this condition can be restated as

resp hn+1 QdS + resp hn+1 QdS = 0 (37)

for n. We can rewrite (37) as constraints on the F -function. They turn out to be

F 1 X F F X F 1 X
= 2nN
X
kTk + (k + nN)Tk+nN + klTk Tl ,
k1 Tk+nN 2 k+l=nN Tk Tl
k1 Tk 2 k+l=nN

11

F 1 F F X F 1
= 2nN
X X X
k Tk + (k + nN)Tk+nN + klTk Tl ,
k1 Tk+nN 2 k+l=nN Tk Tl
k1 Tk 2 k+l=nN

(38)

where n 0.

4.1 Acknowledgements
The second author (K.T.) would like to thank A. Gorsky, S. Kharchev, A. Marshakov, A.
Mironov, and A. Morozov for fruitful discussions. The idea to connect the hyperelliptic
cirve C with the Toda lattice was inspired by this Moscow group.

References
[1] N.Seiberg and E.Witten, Nucl. Phys. B426 (1994) 19.

[2] C.Montonen and D.Olive, Phys. Lett. 72B (1977) 117.

P.Goddard, J.Nuyts and D.Olive, Nucl. Phys. B125 (1977) 1.

[3] N.Seiberg, Nucl. Phys. B435 (1995) 129, and references therein.

[4] A.Klemm,W.Lerche,S.Theisen and S.Yankielowicz, Phys. Lett. 344B (1995) 169.

P.Argyres and A.Faraggi, Phys. Rev. Lett. 73 (1995) 3931.
M.Douglas and S.Shenker, Nucl. Phys. B447 (1995) 271.

[5] U.H.Danielsson and B.Sundborg, Preprint USITP-95-06,UUITP-4/95.

A.Brandhuber and K.Landsteiner, Preprint CERN-TH/95-180.

[6] A.Gorsky, I.Krichever, A.Marshakov, A.Mironov and A.Morozov, Phys. Lett. 355B
(1995) 466.

[7] A.Gurevich and L.Pitaevsky, JETP 65 (1973) 65.

[8] G.B.Whitham, Linear and Nonlinear Waves, John Wiley, New York, 1974.

[9] I.M.Krichever, talk at the 1st Sakharov Conference, Moscow 1991 ; Comm. Pure Appl.
Math. 47-4 (1994) 437 .
B.A.Dubrovin, Commun. Math. Phys. 145 (1992) 195 ; Nucl. Phys. B379 (1992) 627..

12
[10] B.A.Dubrovin and S.P.Novikov, Russian Math. Surveys 44:6 (1989) 35.
P.D.Lax, C.D.Levermore and S.Verakides, in Important Developments in Soliton
Theory, A.S.Fokas and V.E.Zakharov (eds.), Springer Series in Nonlinear Dynamics
(Springer-Verlag,1994)

[11] H.Flaschka, M.G.Forest and D.W.McLaughlin, Comm. Pure Appl. Math. 33 (1980)
739.

[12] H.Farkas and I.Kra, Riemann surfaces, (Springer-Verlag, 1980).

[13] I.M.Krichever, Russian Math. Surveys 33:4 (1978) 255.

P.van Moerbeke and D.Mumford, Acta Math. 143 (1979) 93.

[14] K.Ueno and K.Takasaki, Adv. Stud. Pure Math. 4 (1984) 1.

[15] A.M.Bloch and Y.Kodama, SIAM J. Appl. Math. 52-4 (1992) 909.

[16] I.M.Krichever, Commun. Math. Phys. 143 (1991) 415.

K.Takasaki and T.Takebe, Intern. J. Mod. Phys. A7, suppl.1B (1992) 889.

[17] T.Nakatsu, A.Kato, M.Noumi and T.Takebe, Phys. Lett. 322B (1994) 192.

13