The solution to the q-KdV equation

M. Adler

E. Horozov

P. van Moerbeke

October 15, 1997

Abstract: Let KdV stand for the Nth Gelfand-Dickey reduction of the KP
hierarchy. The purpose of this paper is to show that any KdV solution leads
eectively to a solution of the q-approximation of KdV. Two dierent q-KdV
approximations were proposed, by E. Frenkel [7] and Khesin, Lyubashenko
and Roger [12]. We show there is a dictionary between the solutions of q-
KP and the 1-Toda lattice equations, obeying some special requirement; this
is based on an algebra isomorphism between dierence operators and D-
operators, where Df(x) = f(qx). Therefore every notion about the 1-Toda
lattice can be transcribed into q-language. So, q-KdV is yet another Toda
discretization of KdV.
Consider the q-dierence operators D and D
q
, dened by
Df(y) = f(qy) and D
q
f(y) :=
f(qy) f(y)
(q 1)y
,

Appeared in: Physics letters A, 242, 139-151, (1998)

Department of Mathematics, Brandeis University, Waltham, Mass 02254, USA. E-

mail: adler@math.brandeis.edu. The support of a National Science Foundation grant #
DMS-9503246 is gratefully acknowledged.

Department of Mathematics and Informatics, Soa University, 5 J. Bourchier Blvd.,

Soa 1126, Bulgaria.

Universite de Louvain, 1348 Louvain-la-Neuve, Belgium and Department of

Mathematics, Brandeis University, Waltham, Mass 02254, USA. E-mail: vanmoer-
beke@geom.ucl.ac.be or @math.brandeis.edu. The support of a NSF grant # DMS-
9503246, a Nato, a FNRS and a Francqui Foundation grant is gratefully acknowledged.
Part of this work was carried out at the Centre Emile Borel, Paris during fall 96.
1
and the q-pseudo-dierential operators
Q = D+u
0
(x)D
0
+u
1
D
1
+... and Q
q
= D
q
+v
0
(x)D
0
q
+v
1
(x)D
1
q
+...
The following q-versions of KP were proposed by E. Frenkel [7] and by Khesin,
Lyubashenko and Roger [12], for n = 1, 2, ...:
Q
t
n
= [(Q
n
)
+
, Q] (Frenkel system) (0.1)
Q
q
t
n
= [
_
Q
n
q
_
+
, Q
q
], (KLR system) (0.2)
where ( )
+
and ( )

refer to the q-dierential and strictly q-pseudo-dierential

part of ( ). The two systems are identical, after a (constant) upper-triangular
linear transformation from the u
i
s to the v
i
s, as will become clear from the
isomorphism between q-operators and dierence operators, explained below.
The purpose of this paper is to give a large class of solutions to both systems.
The -function (z) :=

iZ
z
i
; enjoys the property f(, )(/) =
f(, )(/). Consider an appropriate space of functions f(y) representable
by Fourier series in the basis
n
(y) := (q
n
x
1
y) for xed q = 1,
f(y) =

f
n

n
(y);
the operators D, dened by Df(y) = f(qy), and multiplication by a function
a(y) act on the basis elements, as follows:
D
n
(y) =
n1
(y) and a(y)
n
(y) = a(xq
n
)
n
(y).
Therefore, the Fourier transform,
f

f = (..., f
n
, ...)
nZ
,
induces an algebra isomorphism, mapping D-operators onto a special class
of -operators in the shift :=
_

i,j1
_
i,jZ
, as follows:

i
a
i
(y)D
i

i
diag (..., a
i
(xq
n
), ...)
nZ

i
; (0.3)
2
conversely, any dierence operator, depending on x, of the type (0.3) i.e.,
annihilated by D Ad

, where (Ad

)a = a
1
, maps into a D-operator.
This is the crucial basic isomorphism used throughout this paper.
To the shift and to a xed diagonal matrix = diag(
n+1
)
nZ
, we
associate new operators

= and

=

+ = ( 1).
Observe that, under the isomorphism (0.3),
D ,
1
(q 1)x
D

and D
q

,
upon setting
1
n
= (1 q)xq
n1
.
Dening the simple vertex operators
X(t, z) := e

1
t
i
z
i
e

1
z
i
i

t
i
, (0.4)
we now make a statement concerning the so-called one-Toda lattice; the latter
describes deformations of of a bi-innite matrix L, which is lower-triangular,
except for 1s just above the main diagonal. The rst formula (0.6) below
gives a solution to the Frenkel system (Theorem 0.1), upon replacing

by ,
which amounts to conjugating L by a constant diagonal matrix ; see (2.2).
The second formula (0.6) gives, via the isomorphism (0.3), a solution to the
KLR system (Theorem 0.2). Thus in the L-representation the two systems
are related by a trivial diagonal conjugation. Note, Theorem 0.1 is given for
arbitrary = (...,
1
,
0
,
1
, ...).
We shall need the well-known Hirota symbol for a polynomial p,
p(

)f g := p
_

y
1
,
1
2

y
2
, ...
_
f(t + y)g(t y)

y=0
.
Note A
+
refers to the upper-triangular part of a matrix A, including the
diagonal, and for C, set [] := (,
1
2

2
,
1
3

3
, ...) C

.
Theorem 0.1. Given an integer N 2, consider an arbitrary -function for
the KP equation such that /t
iN
= 0 for i = 1, 2, 3, ... (N-KdV hierarchy).
3
For a xed , , c C

, the innite sequence of -functions

1

n
:= X(t,
n
)...X(t,
1
)(c + t),
0
= (c + t), for n 0;
satises the 1-Toda bilinear identity for all t, t

and all n > m:

_
z=

n
(t [z
1
])
m+1
(t

+ [z
1
])e

1
(t
i
t

i
)z
i
z
nm1
dz = 0.
The bi-innite matrix (a full matrix below the main diagonal), where p

are
the elementary Schur polynomials,
L =

=0
diag
_
p

)
n+2

n

n+2

n
_
nZ

1
(0.5)
has the following properties:
(i) L
N
satises the 1-Toda lattice
L
N
t
n
= [(L
n
)
+
, L
N
], n = 1, 2, ..,
(ii) L
N
is upper triangular and admits the following expression in terms
2
of

and

:
L
N
=

N
+
N

1
(
j
+ b
j
)

p1
+
_
N1

0
a
j
+

1ijN1
(
i
+ b
i
)(
j
+ b
j
)
_

N2
+... +
N
1

0
=

N
+
_
N

1
b
j
_

N1
+
_
_
N1

0
a
j

N1

1
(b
N
b
i
)
i
+

1ijN1
b
i
b
j
_
_

N2
+ ...(0.6)
1

n
for n < 0 is dened later in (3.3).
2
in the expressions below, the coecients of the

s are diagonal matrices, whose

0th component is given by the expression appearing below; e.g.,

N
1
b
j
stands for
diag(

N
1
b
j+n
)
nZ
and
N
1
stands for diag(
N
1+n
)
nZ
.
4
with
b
k
=

t
1
log
(c + t
k

1
[
1
i
])
(c + t
k1

1
[
1
i
])
, a
k
=
_

t
1
_
2
log
_
c + t
k

1
[
1
i
]
_
,
(0.7)
for k 1. These expressions for k 0 will be given in (3.4) and (3.5).
In view of (0.7), the shift
: b
k
b
k
= b
k+1
and a
k
a
k
= a
k+1
corresponds to the following transformation,
: c c [
1
1
] and
i

i+1
. (0.8)
Therefore, in order that L
N
satises the form of the right hand side of (0.3),
we must make c and
i
depend on x and q, such that the map on a, b,
corresponds to D, in addition to the fact that all
i
must tend to simul-
taneously and c to (x, 0, 0, ...), when q goes to 1. So, c(x) and (x) must
satisfy:
_

_
Dc(x) = c(x) [
1
1
(x)]
D
n
(x) =
n+1
(x)
lim
q1

i
=
lim
q1
c(x) = x := (x, 0, 0, ...);
(0.9)
its only solution is given by:

1
n
= (1 q)xq
n1
and c(x) =
_
(1 q)x
1 q
,
(1 q)
2
x
2
2(1 q
2
)
,
(1 q)
3
x
3
3(1 q
3
)
, ...
_
,
(0.10)
and thus D
n
c(x) = c(x)

n
1
[
1
i
]. With this choice of
n
,
1
(q 1)x
D

and D
q
:=
D 1
(q 1)x

. (0.11)
In analogy with (0.4), we dene the simple q-vertex operators:
X
q
(x, t, z) := e
xz
q
X(t, z) and

X
q
(x, t, z) :=
_
e
xz
q
_
1
X(t, z). (0.12)
5
in terms of (0.4) and the q-exponential e
x
q
:= e

1
(1q)
k
x
k
k(1q
k
)
. Therefore under
the isomorphism (0.3), Theorem 0.1 can be translated into q-language, to
read:
Theorem 0.2. Any KdV -function leads to a q-KdV -function (c(x)+t);
the latter satises the bilinear relations below, for all x R, t, t

, and
all n > m, which tends to the standard KP-bilinear identity, when q goes to
1:
_
z=
D
n
(X
q
(x, t, z)(c(x) + t)) D
m+1
_

X
q
(x, t

, z)(c(x) + t

)
_
dz = 0

_
z=
X(t, z)( x + t) X(t

, z)( x + t

) dz = 0
(0.13)
Moreover, the q-dierential operator Q
N
q
has the form below and tends to
the dierential operator L
N
of the KdV hierarchy, when q goes to 1:
Q
N
q
= D
N
q
+

t
1
log
(D
N
c + t)
(c + t)
D
N1
q
+
_
N1

i=0

2
t
2
1
log (D
i
c + t)

N2

i=0

i+1
_

t
1
log
(D
N
c + t)
(D
N1
c + t)


t
1
log
(D
i+1
c + t)
(D
i
c + t)
_
+

0ijN2

t
1
log
(D
i+1
c + t)
(D
i
c + t)

t
1
log
(D
j+1
c + t)
(D
j
c + t)
_
D
N2
q
+ ...

_

x
_
N
+ N

2
t
2
1
log ( x + t)
_

x
_
N2
+ ... . (0.14)
M.A. and PvM thank Edward Frenkel for kindly discussing this problem
during spring 1996. For a systematic study of discrete systems, see Kupersh-
midt [13] and Gieseker [8]. It is an old observation (see [11]) that the Toda
lattice discretizes the KdV equation, and this in many dierent ways. There-
fore it is not surprising that q-KdV is yet another Toda discretization of KdV.
In an elegant recent preprint, Iliev [10] has obtained q-bilinear identities and
q-tau functions, as well, purely within the KP theory.
6
1 The KP -functions and Grassmannians
KP -functions satisfy the dierential Fay identity for all y, z C, in terms
of the Wronskian {f, g} := f

g fg

, as shown in [1, 16]:

{(t [y
1
]), (t [z
1
])}
+(yz)((t[y
1
])(t[z
1
])(t)(t[y
1
] [z
1
]) = 0.
(1.1)
In fact this identity characterizes the -function, as shown in [15]. We shall
need the following, shown in [1]:
Proposition 1.1. Consider -functions
1
and
2
, the corresponding wave
functions

i
= e

i1
t
i
z
i
i
(t [z
1
])

i
(t)
= e

i1
t
i
z
i
_
1 + O(z
1
)
_
(1.2)
and the associated innite-dimensional planes, as points in the Grassmannian
Gr,

W
i
= span
_
_
_
_

t
1
_
k

i
(t, z), for k = 0, 1, 2, ...
_
_
_
;
then the following statements are equivalent
(i) z

W
2


W
1
;
(ii) z
2
(t, z) =

t
1

1
(t, z)
1
(t, z), for some function = (t);
(iii)
{
1
(t [z
1
]),
2
(t)} + z(
1
(t [z
1
])
2
(t)
2
(t [z
1
])
1
(t)) = 0 (1.3)
When (i), (ii) or (iii) holds, (t) is given by
(t) =

t
1
log

2

1
. (1.4)
Proof: To prove that (i) (ii), the inclusion z

W
2


W
1
implies z

W
t
2

W
t
1
, where

W
t
=

We

1
t
i
z
i
; it follows that
z
2
(t, z) = z(1 + O(z
1
)) W
t
1
7
must be a linear combination, involving the wave functions
i
= e

1
t
i
z
i

i
:
z
2
=
_

x
+ z
_

1
(t)
1
, and thus z
2
=

t
1

1
(t)
1
. (1.5)
The expression (1.4) for (t) follows from equating the z
0
-coecient in (1.5),
upon using the -function representation (1.2). To show that (ii) (i), note
that
z
2
=

t
1

1
W
0
1
and taking z-derivatives, we have
z
_

t
1
_
j

2
=
_

t
1
_
j+1

1
+
1
_

t
1
_
j

1
+ +
j+1

1
,
for some
1
, ,
j+1
depending on t only; this implies the inclusion (i). The
equivalence (ii) (iii) follows from a straight forward computation using
the -function representation (1.2) of (ii) and the expression for (t).
2 The one-Toda lattice
For details on this sketchy exposition, see [3]. The one-Toda lattice equations
L
t
n
= [(L
n
)
+
, L], (2.1)
are deformations of an innite matrix
L =

<i0
a
i

i
+

, with

:= =
1
, (2.2)
for diagonal matrices and , with non-zero entries, and diagonal matrices a
i
,
depending on t = (t
1
, t
2
, . . .). Note the conjugation by the constant diagonal
matrix is harmless, but it is necessary to capture the KLR-system. One
introduces wave and adjoint wave vectors (t, z) and

(t, z), satisfying

L = z and L

= z

8
and

t
n
= (L
n
)
+

t
n
= ((L
n
)
+
)

. (2.3)
The wave vectors and

can be expressed in terms of one sequence of

-functions (n, t) :=
n
(t
1
, t
2
, . . .), n Z, to wit:
(t, z) =
_
e

1
t
i
z
i
(t, z)
_
nZ
=
_

n
(t [z
1
])

n
(t)
e

1
t
i
z
i

n
z
n
_
nZ
,

(t, z) =
_
e

1
t
i
z
i

(t, z)
_
nZ
=
_

n+1
(t + [z
1
])

n+1
(t)
e

1
t
i
z
i

1
n
z
n
_
nZ
(2.4)
It follows that, in terms of (z) := (z
n
)
nZ
and the notation a

:= diag(a
k+1
)
kZ
:
= e

1
t
i
z
i
S(z), with S =

0
p
n
(

)(t)
(t)

n
,

= e

1
t
i
z
i
(S

)
1

1
(z
1
), with S
1
=

n
_
p
n
(

)(t)
(t)
_

.
Moreover, as will follow from Proposition 2.1 below, and

satisfy the
bilinear identities:
_
z=

n
(t, z)

m
(t

, z)
dz
2iz
= 0, for all n > m.
From the representation of S and S
1
above, it follows that
L
k
= S

k
S
1
=

=0
diag
_
p

)
n+k+1

n

n+k+1

n
_
nZ

k
=

k
+ diag
_

t
1
log

n+k

n
_
nZ

k1
+ ...
+diag
_

t
k
log

n+1

n
_
nZ

0
+ diag
_

2
t
1
t
k
log
n
_
nZ

1
+ ...
(2.5)
9
For instance, the
0
-term in the last expression follows from setting m =
n 2, t t + [], t

t [] in the bilinear identity above, yielding

0 =

n
(t + [])
n1
(t [])

n
(t)
n1
(t)
_
z=

n
(t + [], z)

n2
(t [], z)
dz
2iz
=
1

n1

j0

j
_

t
j+2
p
j+2
(

)
_

n

n1
,
and thus,

t
k
log

n+1

n
=
p
k
(

)
n+1

n

n+1

n
=
_
L
k
_
nn
.
With each component of the wave vector , we associate a sequence of
innite-dimensional planes in the Grassmannian Gr
W
n
= span
C
_
_
_
_

t
1
_
k

n
(t, z), k = 0, 1, 2, ...
_
_
_
= e

1
t
i
z
i
span
C
_
_
_
_

t
1
+ z
_
k

n
(t, z), k = 0, 1, 2, ...
_
_
_
(2.6)
and planes
W

n
= span
C
_
_
_
_

t
1
_
k

n1
(t, z), k = 0, 1, 2, ...
_
_
_
,
which are orthogonal to W
n
by the residue pairing
_
z=
f(z)g(z)
dz
2iz
. (2.7)
Note that the plane z
n
W
n
has so-called virtual genus zero, in the terminol-
ogy of [14]; in particular, this plane contains an element of order 1 +O(z
1
).
The following statement is contained in [3]:
Proposition 2.1. The following ve statements are equivalent
(i) The 1-Toda lattice equations (2.1)
10
(ii) and

, with the proper asymptotic behaviour, given by (2.4), satisfy

the bilinear identities for all t, t

_
z=

n
(t, z)

m
(t

, z)
dz
2iz
= 0, for all n > m; (2.8)
(iii) the -vector satises the following bilinear identities for all n > m and
t, t

:
_
z=

n
(t [z
1
])
m+1
(t

+ [z
1
])e

1
(t
i
t

i
)z
i
z
nm1
dz = 0; (2.9)
(iv) The components
n
of a -vector correspond to a ag of planes in Gr,
W
n1
W
n
W
n+1
.... (2.10)
(v) A sequence of KP--functions
n
satisfying the equations
{
n
(t [z
1
]),
n+1
(t)} + z(
n
(t [z
1
])
n+1
(t)
n+1
(t [z
1
])
n
(t)) = 0
(2.11)
Proof: The proof that (i) is equivalent to (ii) follows from the methods in
[4, 16]. That (ii) is equivalent to (iii) follows from the representation (2.4) of
wave functions in terms of -functions. Finally, we sketch the proof that (ii)
is equivalent to (iv). The inclusion in (iv) implies that W
n
, given by (2.6), is
also given by
W
n
= span
C
{
n
(t, z),
n+1
(t, z), ...};
Since each
n
is a -function, we have that
_
z=

n
(t, z)

n1
(t

, z)
dz
2iz
= 0,
implying that, for each n Z,

n1
(t, z) W

n
. Moreover the inclusions
... W
n
W
n+1
... imply, by orthogonality, the inclusions ... W

n

W

n+1
..., and thus
W

n
= {

n1
(t, z),

n2
(t, z), ...}.
11
Since
W
n
W
m
= (W

m
)

, all n m,
we have the orthogonality W
n
W

m
by the residue pairing (2.7) for all n m,
i.e.,
_
z=

n
(t, z)

m1
(t

, z)
dz
2iz
= 0, all n m.
Note (ii) implies W

m
W

n
, n > m, hence W
n
W
m
, n > m, yielding
(iv). That (iv)(v) follows from proposition 1.1, by setting
1
:=
n
and

2
=
n+1
. Then (v) is equivalent to the inclusion property
z(z
n1
W
n+1
) (z
n
W
n
), i.e. W
n+1
W
n
,
thus ending the proof of proposition 2.1.
3 Proof of Theorems 0.1 and 0.2
At rst, we exhibit particular solutions to equation (2.11), explained in [1].
Lemma 3.1. Particular solutions to equation
{
1
(t [z
1
]),
2
(t)} + z(
1
(t [z
1
])
2
(t)
2
(t [z
1
])
1
(t)) = 0
are given, for arbitrary C

, by pairs (
1
,
2
), dened by:

2
(t) = X(t, )
1
(t) = e

t
i

1
(t [
1
]), (3.1)
or

1
(t) = X(t, )
2
(t) = e

t
i

2
(t + [
1
]). (3.2)
Proof: Using
e

1
1
i
(

z
)
i
= 1

z
,
it suces to check that
2
(t) satises the above equation (2.11)
e

t
i

i
_
{
1
(t [z
1
]),
2
(t)} + z(
1
(t [z
1
])
2
(t)
2
(t [z
1
])
1
(t))
_
= e

t
i

i
{
1
(t [z
1
]), e

t
i

1
(t [
1
])}
12
+ z(
1
(t [z
1
])
1
(t [
1
]) (1

z
)
1
(t)
1
(t [z
1
]
1
]))
= {
1
(t [z
1
]),
1
(t [
1
])}
+ (z )(
1
(t [z
1
])
1
(t [
1
])
1
(t)
1
(t [z
1
] [
1
]))
= 0,
using the dierential Fay identity (1.1) for the -function
1
; a similar proof
works for the second solution, given by (3.2).
Proof of Theorems 0.1 and 0.2: From an arbitrary N-KdV -function,
construct, for , c, C

, the following sequence of -functions, for n 0,

as announced in Theorem 0.1:

0
(t) = (c + t)

n
= X(t,
n
)...X(t,
1
)(c + t)
=
(
1
, ...,
n
)

n
1

i1
i
n

k=1
e

i=1
t
i

i
k
(c + t
n

1
[
1
i
]),

n
= X(t,
n+1
)...X(t,
0
)(c + t)
=
(
0
, ...,
n+1
)

n
1

i1
i+1
n

k=1
e

i=1
t
i

i
k+1
(c + t +
n

1
[
1
i+1
])
(3.3)
and so, each
n
is dened inductively by

n+1
= X(t,
n+1
)
n
;
thus by Lemma 3.1, the functions
n+1
and
n
are a solution of equation (v)
of proposition 2.1. Therefore, by the same proposition 2.1, the
n
s form a
-vector of the 1-Toda lattice. By removing the harmless exponential factor

n
k=1
exp(

1
t
iN

iN
k
), each
n
has the property that
n
/t
iN
= 0 for i =
1, 2, ...; therefore
z
N
W
n
W
n
.
In particular, the representation
W
n
= span{
n
(t, z),
n+1
(t, z), ...},
13
which follows from the inclusion ... W
n
W
n+1
..., implies that, since
L = z,
z
N

k
=

jk
a
j

j
= (L
N
)
k
,
and thus L
N
is upper-triangular.
Therefore, we conclude that the matrix L, dened by (2.5), from the
sequence of -functions (3.3),
L =

+
_

t
1
log

n+1

n
_
nZ
+
_
_
_

t
1
_
2
log
n
_
_
nZ

1
+ ...
=

+ (
n+1
+ b
n+1
)
nZ

0
+ (a
n
)
nZ

1
+ ...,
satises the 1-Toda lattice equations, where
b
n+1
=

t
1
log
(c + t

n+1
1
[
1
i
])
(c + t

n
1
[
1
i
])
for n 1
=

t
1
log
(c + t [
1
1
])
(c + t)
, for n = 0,
=

t
1
log
(c + t +

n+2
0
[
1
i
](1
1,n
))
(c + t +

n+1
0
[
1
i
])
, for n 1, (3.4)
and
a
n
=

2
t
2
1
log (c + t
n

1
[
1
i
]) for n 1
=

2
t
2
1
log (c + t) for n = 0
=

2
t
2
1
log (c + t +
n+1

0
[
1
i
]) for n 1, (3.5)
conrming (0.7). Using the fact that, in view of (2.5), the diagonal terms of
L
N
are given by

t
N
log

n+1

n
=
N
n+1
,
and the fact that, in the notation of footnote 2,

n
= (

+ )
n
=

n
+
_
n

i
_

n1
+
_
_

1ijn1

j
_
_

n2
+ ...,
14
one nds that the upper-triangular matrix L
N
has the following expression:
L
N
=

N
+
N

1
(
j
+ b
j
)

p1
+
_
N1

0
a
j
+

1ijN1
(
i
+ b
i
)(
j
+ b
j
)
_

N2
+... +
N
1

0
=

N
+
_
N

1
b
j
_

N1
+
_
_
N1

0
a
j

N1

1
(b
N
b
i
)
i
+

1ijN1
b
i
b
j
_
_

N2
+ ...
(3.6)
in terms of b
k
and a
k
dened in (0.7), thus proving Theorem 0.1.
To prove Theorem 0.2, note at rst:
z
nm1

n
k=m+2
(
k
)
n

k=m+2
e

i=1
1
i
_

k
z
_
i
=
z
nm1

n
k=m+2
(
k
)
n

k=m+2
_
1

k
z
_
=
n

k=m+2
_
1
z

k
_
=
n

k=m+2
e

i=1
1
i
_
z

k
_
i
=
e
xzq
n
q
e
xzq
m+1
q
= D
n
e
xz
q
D
m+1
_
e
xz
q
_
1
.
The function
n
, dened in Theorem 0.1, satises the bilinear identity of
Theorem 0.1; therefore, using (3.3) and the above in the computation of
n
(t
[z
1
]), the following relations hold, up to a multiplicative factor depending
on and :
(, )
_
z=

n
(t [z
1
])
m+1
(t

+ [z
1
])e

1
(t
i
t

i
)z
i
z
nm
dz
z
=
_
z=
(c(x) + t [z
1
]
n

1
[
1
i
])(c(x) + t

+ [z
1
] +
m+1

1
[
1
i
])
n

k=m+2
_
1
z

k
_
e

1
(t
i
t

i
)z
i
dz
=
_
z=
D
n
(X
q
(x, t, z)(c(x) + t)) D
m+1
_

X
q
(x, t

, z)(c(x) + t

)
_
dz = 0.
15
When q 1, the second expression above tends to the standard KP-bilinear
equation, upon using (0.10). Moreover, one checks by induction, using the
expression (2.5) for L and (3.3), that (L
N
)
+
for N = 1, 2, 3, ... has the q-form
(0.3). Also, note that a
k
and b
k
can be expressed in terms of the D-operator,
using (0.7); to wit:
b
k
=

t
1
log
(D
k
c + t)
(D
k1
c + t)
, a
k
=
_

t
1
_
2
log
_
D
k
c + t
_
.
So, the expression for Q
N
q
in Theorem 0.2 follows at once from (3.6). The
fact that

t
1
log
(D
j+1
c + t)
(D
j
c + t)


2
x
2
log ( x + t)
implies that all terms in (0.14) vanish in the limit q 1, except for the
term

N1
i=0

2
t
2
1
log (D
i
c + t); so we have that
lim
q1
Q
N
p
=
_

x
_
N
+ N

2
x
2
log ( x + t)
_

x
_
N2
+ ...,
thus ending the proof of theorem 0.2.
4 Examples and vertex operators
The isomorphism (0.3) enables one to translate every 1-Toda statement, hav-
ing the form (0.3) into a D or D
q
statement. Also every -function of the
KdV hierarchy leads automatically to a solution of q-KdV. For instance, by
replacing t c(x) +t in the Schur polynomials, one nds q-Schur polynomi-
als. The latter were obtained by Haine and Iliev [9] by using the q-Darboux
transforms; the latter had been studied by Horozov and coworkers in [5, 6].
The n-soliton solution to the KdV (for N = 2) (for this formulation, see
[4]),
(t) = det
_

i,j

a
j
y
i
+ y
j
e

k:odd
t
k
(y
k
i
+y
k
j
)
_
1i,jn
,
leads to a q-soliton by the shift t c(x) + t, with c(x) as in (0.8), namely
(x, t) = det
_
_
_
ij
a
j
_
e
xy
i
q
e
xy
i
q
_
1
y
i
+ y
j
e

k=1
t
k
(y
k
i
+y
k
j
)
_
_
_
1i,jn
.
16
Moreover the vertex operator for the 1-Toda lattice is a reduction of the
2-Toda lattice vertex operator (see [2]), given by
X(t, y, z) =

(z)X(t, z)X(t, y)(y)

=
z
y z
e

1
t
i
(y
i
z
i
)
e

1
(y
i
z
i
)
1
i

t
i
_
y
n
z
n
_
nZ
;
in particular, if is a 1-Toda vector, then a +bX(t, y, z) is a 1-Toda vector
as well. Using the dictionary, this leads to q-vertex operators
X
q
(x, t; y, z) = e
xy
q
(e
xz
q
)
1
e

t
i
(y
i
z
i
)
e

(y
i
z
i
)
1
i

t
i
for q-KP,
and, for any Nth root of 1,
X
q
(x, t; z) = e
xz
q
(e
xz
q
)
1
e

t
i
z
i
(
i
1)
e

z
i
(
i
1)
1
i

t
i
for q-KdV,
having the typical vertex operator properties.
References
[1] M. Adler and P. van Moerbeke: Birkho strata, Backlund transforma-
tions and limits of isospectral operators , Adv. in Math., 108 140204
(1994).
[2] M. Adler and P. van Moerbeke: The spectrum of coupled random ma-
trices, preprint (1997), Ann. of Math (1999).
[3] M. Adler and P. van Moerbeke: Vertex operator solutions of the discrete
KP equation , preprint (1997), Comm. Math. Phys. (1999).
[4] M. Adler, T. Shiota and P. van Moerbeke : A Lax representation for
the vertex operator and the central extension, Comm. Math. Phys. 171,
547588, (1997).
[5] B. Bakalov, E. Horozov and M. Yakimov: Bispectral algebras of com-
muting ordinary dierential operators , Comm. Math. Phys., to appear.
[6] B. Bakalov, E. Horozov and M. Yakimov: General methods for con-
structing bispectral operators , Phys. Letters A, 222 5966 (1996).
17
[7] E. Frenkel: Deformations of the KdV hierarchy and related soliton equa-
tions , Int. Math. Res. Notices, 2 5576 (1996).
[8] D. Gieseker: The Toda hierarchy and the KdV hierarchy , preprint,
alg-geom/9509006.
[9] L. Haine and P. Iliev: The bispectral property of a q-deformation of the
Schur polynomials and the q-KdV hierarchy , J. of Phys. A: Math. Gen,
30, 7217-7227 (1997).
[10] P. Iliev: Tau-function solutions to a q-deformation of the KP-hierarchy,
preprint (1997).
[11] M. Kac, P. van Moerbeke: The solution of the periodic Toda lattices,
Proc. Nat. Acad. Sci., USA, 72, (4), 1627-1629 (1975) and The solution
of the periodic Toda lattices, Proc. Nat. Acad. Sci., USA, 72, (8), 2879-
2880 (1975) .
[12] B. Khesin, V. Lyubashenko and C. Roger: Extensions and contractions
of the Lie algebra of q-Peudodierential symbols on the circle , J. of
functional analysis, 143 5597 (1997).
[13] B. A. Kupershmidt: Discrete Lax equations and dierential-dierence
calculus , Asterisque, 123 (1985).
[14] G. Segal, G. Wilson: Loop groups and equations of KdV type, Publ.
Math. IHES 61, 565 (1985).
[15] K. Takasaki, T. Takebe: Integrable hierarchies and dispersionless limit,
Reviews in Math. Phys. 7,743808 (1995) .
[16] P. van Moerbeke: Integrable foundations of string theory, in Lecures
on Integrable systems, Proceedings of the CIMPA-school, 1991, Ed.:
O. Babelon, P. Cartier, Y. Kosmann-Schwarzbach, World scientic, pp
163267 (1994).
18