Professional Documents
Culture Documents
Acta Applicandae Mathematicae Volume 36 Issue 1-2 1994 (Doi 10.1007 - bf01001546) A. O. Smirnov - Finite-Gap Elliptic Solutions of The KDV Equation
Acta Applicandae Mathematicae Volume 36 Issue 1-2 1994 (Doi 10.1007 - bf01001546) A. O. Smirnov - Finite-Gap Elliptic Solutions of The KDV Equation
125
9 1994 KluwerAcademic Publishers. Printed in the Netherlands.
A. O. S M I R N O V
Department of Mathematics, St. Petersburg Institute of Aircraft lnstrumentation, Gerzena str. 67,
190000 St. Petersburg, Russia
Abstract. A method is proposed for constructing finite-gap elliptic in x or/and in t solutions of the
Korteweg-de Vries equation. Dynamics of poles for two-gap elliptic solutions of the KdV equation are
considered. Numerous examples of new elliptic solutions of the KdV equation are given.
Key words. Korteweg--de Vries equation, soliton, theta function, integrable nonlinear equation,
Riemann surface, covering.
O. I n t r o d u c t i o n
In 1974, a new branch of the theory of nonlinear equations integrable by the inverse
scattering method appeared - t h e theory of finite-gap integration [1]. During the last
15 years, this theory came through a period of intensive development and found its
application in various branches of theoretical physics [2-7]. We are interested, in
particular, in the problem which can be formulated with respect to the Korteweg-
de Vries equation (KdV) in the following way. The general periodic potential of the
Schrtdinger operator of auxiliary linear problem [2-7]
r - ur = zr (0.1)
has infinitely large numbers of lacunae in its spectrum, i.e. is not a finite-gap
potential. To the contrary, the general finite-gap potential, which has a finite number
of lacunae in its spectrum and is constructed by using the Matveev-Its formula
[7-10]
u(x,t) = -202 In O(Ux + Vt + Z I B) + const, (0.2)
appears to be a quasi-periodic function. Naturally, the following problem has imme-
diately arisen: what is the specific structure of the spectrum of the periodic finite-gap
potential?
Up to recent times, only one series of such potentials was known, namely Lam6
potentials:
u(x, O) = g(g + 1)p(x), (0.3)
126 A.O. SMIRNOV
ml
B = B t, Im B > 0. (1.3)
W~CoW = Co,
with g x 9 submatrices S, P, Q, R
W= R '
W o B = (Q + R B ) ( S + p B - 1 ) , (1.4)
defines the action of the real simplectic group SP2g(~) in S a. The matrices W o B
and B E Sg are called equivalent.
128 ~ O. SMIRNOV
0[4; ~](p I B)
(1.5)
= Z e x p { l r i ( B ( m + a ) , m + 4 ) + 27ri(m+4,p+fl)},
mEgg
where p E C a , the summation runs over the integer g-dimensional lattice and ( , ) is
a standard scalar product in ~a. For the sake of brevity, let us write
0[4; fl](p + N + BM I B)
(1.6)
= exp { - ~ri(BM, M) - 27ri(p + 5, M) + 2;ri(4, N)}O[4; fl](p I B).
Xj E Z. (1.10)
./---1
If B E Sz, p E Cg, W E SP2g(~), then the theta functions (1.5) of the equivalent
matrices are related by
014'; Z'I(v' I w o B)
(i.11)
= xexP ( 2 ((S + PB)-IPp, p) } deI(S + pB)I/2ot4; fll(P [ B),
FINITE-GAP ELLIPTIC SOLUTIONS 129
where/~ is the vector of Riemann constants [10, 20], then the function O(Li(P) - Z I
B) will have precisely g first-order zeros, located at the points ~'j and no other zeros
will be present.
Let us define on F also (i) a second kind Abelian differentials d~,~,Q with a
single pole at the point Q E F of the order n + 1 and the principal part of the form
- n k - ' ~ - l d k , n >1 1; (ii) a third kind Abelian differential dwp,Q, which possesses a
pair of first-order poles at the points P, Q E F with residues +1 and -1, respectively.
The b-periods of the normalized Abelian differentials of the second and third kind
= O, ~a
J~ak
d~n,Q
k
dwp,Q --- O, k = 1 , . . . , g, (1.13)
are related to the normalized holomorphic differentials d/dj (1.1) by the relations
1 i dff2n,q = (n -1
2rci
dn
- 1)! ----g
dk Hm
k=0
, m = 1,...,g. (1.14)
and the integration path, connecting the points P and Q does not intersect any of
the cycles aj and bj, j = 1 , . . . , g .
0"71 = ( SQ.R
P) 70' o., d//0 = T t dUX
where I is a path on F and dff2 is a differential on F0, it follows that the vectors of
b-periods of these differentials satisfy the equation:
O(UlX + g I B1)
or
EXj - ~ e Z. (2.12)
j=l
One of the main properties of the INE is that they can be obtained as the compatibility
condition of a system of linear differential equations [2-7]. In our case, these are
given by
~ (P, x, u, t)
(3.7)
= exp ( ai, q(P)~ + a2,q ( P ) y / ~ - 4 a ~ , Q ( P ) t } n ( e , ~, y,,),
o(u(P) + u~ + v v + w t - z I B)e(Z I B)
H(P,x,V,t)= O(U(P)- Z I B)~-(U~TVy+ W-Y-~Z [ B)'
where
B is the matrix of periods of the Riemann surface F; 21riU, 27riV, 2~riW are the
vectors of the b-periods of the normalized Abelian differentials of the second kind
dff21,Q, or-1 dff22,Q and -4dff23,Q, respectively; g E C. The solution of the KP equa-
tion (3.2), associated with the given surface in these notations acquires the form [9,
10, 37, 38]:
But if the Riemann surface F is hyperelliptic and Q is a branching point, then the
Abelian differential dQz, Q will necessarily be a differential of some meromorphic
function d#Q. And since all the periods of a meromorphic function are equal to zero,
then the solution v(x, y, t) (3.8) will not depend on Y and, therefore, will become a
solution of the KdV equation (3.1) [9, 10, 37, 38]
In this case, the system (3.5)-(3.6) will go into (3.3)-(3.4), where t3 = #q(P), and
the solution will be given by the formula
Cq(P,~,t)
o(u(P) + u~ + w t - z ) o ( z ) (3.10)
= exp { ~'$1,Q(.P)~ -- 4923,q(P)t} 8(lt(P) --'~O-(U~%" ~V-t= "~)"
The hyperelliptic surface, associated with the finite-gap solution of the KdV equation
has a canonical form of a two-sheeted covering over the Riemann sphere
2g+l
w 2 = H ( E - Ej). (3.11)
j=l
Its branching points Ej are edges of the lacunae of the Schrtidinger operator spectrum
(3.3) with the potential (3.9); E = #Q(P), #Q(Q) = oo.
Let us now suppose that some Riemann surface F is n-sheeted covering the toms
F0 and satisfies the conditions [39]:
n
where fi(e 0 are elliptic functions with pole singularities at the point a = 0.
FINITE-GAP ELLIPTIC SOLUTIONS 133
(2) In the vicinity of a = 0, the function R(k, ~) can be represented in the form:
E = + ~ + O(c~2); (3.20)
and w, which has a single pole of order 29 + 1 located at the same point,
w= - 2 + 0(~3-29), (3.21)
and 2g + 1 simple zeros at other points, remaining static with respect to the involution
r (3.18).
COROLLARY 1. Under the conditions of Theorem 3
Further on, the n-sheeted covering (3.12)-(3.14), with the involution r (3.18), will
be denoted by G '~ and the hyperelliptic surfaces G n of genus g by F~.
L E M M A 1. The coefficients fi(c~) in equation (3.12) of the Riemann surface G ~
have the following structures
f2j(cO = _ n ! ( 2 j - 1) J
(2j)!(n 2j)! pS(~)+ ~ 4 ~ - k ( a ) , (3.23)
k=l
n!j J ) (3.24)
d~ "-kfot'l
= -ff(~ '(2j + 1)!(n - 2j - 1)!
k=2
denote these polynomials by 7~n(k, c0, if the following condition holds: ~,~(k, a) in
the neighbourhood of a = 0 can be represented in the form
n
7~n (~-1 + s + 0(~), a) = s'~ + n s,_ 1 + E hi(a) sn-j' (3.25)
Ot
j=2
In~Z]
~.(k,~) = ~~ ~ A f l ~ . _o2 j @ , ~). (3.27)
j=l
f2~(~) = -
n ! ( 2 j - 1) ( j(j - 1)(4j- 11)
(2j)!(n 2j)! f f ( ~ 30 gJ-%~) +
(3.23a)
+ 2j(j - 1)(j - 2 ) ( 1 2 j 2 - 6 9 j + 94) "1
105 93~-3(a) + ' "' f'
n~j
f2j+l(~) =
(2j + 1)!(n - 2j - 1)! ~~176
X {t.,j_l[ot
- , , A + (j - - 1)(j - 2)(4j + 3) 92M-3r~'s~t)+
30 (3.24a)
We shall omit the proof of this lemma. We note only that 7~~ a) are uniquely
determined, while 7~~ c~) are determined up to an additive constant, which is fixed
by the equality (3.28). We also point out the obvious analogy between this lemma
and the expansion theorem for a polynomial, which is an even (odd) function, i.e.
expands over the even (odd) powers of the argument. As an illustration to Lelmna 2,
we list the expressions for the coefficients of the 'basic' element R~ o0, which
allow one to construct all the polynomials R ~ cO, n <~ 13:
f6(~) = -858o
(1 4)
[03 + ~ g 2 ~ - ~ g2 , f 7 ( ~ ) = - 5 1 4 8 ( [ 0 ~ + g2)[0',
( 32 9310 - 53g2),2
f8(o~) = -9009 \[04 + 292102 + 'if
( 1 9 )
f9(~) = -2860 [02 + -5 ga[0 + 20g3 [0',
200 '~
flO(O0 = -2574 [05 + 692103 + 5693[02 + 5g210 _ " ~ 9203),
( 46 ._2+ 153)
f11(o0 = - 3 9 0 [04 + -~- 92~o 128g310+ --if-- g2 if,
~ o = 1, ~o = ~,
THEOREM 4. For any Riemann surface G n, the equation (3.12) has the form
[n/21
=
A j R ,0_ z j ( k , c0 = 0. (3.29)
j=l
and vice versa, for nearly all values of Aj E C (j = 1,..., [n/2l) the relations
(3.29), (3.13) determine the surface G".
The proof follows from Equations (3.14), (3.23)-(3.26) and from Lemma 2. The
coefficients Aj can not take a finite number of values, for which F (3.29), (3.13)
will have branching point 'above' a = 0. The branching may be most conveniently
controlled by means of the Newton diagram (Newton polygon) for the polynomial
(3.25).
COROLLARY 2. The manifold of branching points of the Riemann surface G ~ is
invariant with respect to r (3.18). In general, the genus of G" coincides with the
number of she~es, g = n. The Riemann surface G 2j (G 2j+l) has, in general, two
(four) static with respect to r points (3.18).
Remark 4. The hyperelliptic surface F~ are particular cases of the Riemann surface
G". Therefore, in contrast to the case of the genetic G", for each n there exist only
a finite number of sets Aj, determining some of the surface F~; the genus of F~
satisfies the inequality g ~< ( v ' ~ + 1 - 1)/2 and the number of static points is equal
to 2 g + 2 .
Thus the problem of constructing the n-elliptic solutions of the KdV equation
reduces to the calculation of [n/2] constants Aj.
and have the form (the generalization of equation (3.19) for the case of multiple
zeros, rj-multiplicity of the zero zj),
In particular, there were found in [21, 43, 44] the following new two-gap elliptic
potentials which differ from the two-gap Lam6 potential,
dO = U, J 1 "- u x x - 3 u 2, ]2 = u x z ~ z - l(]'uxxu - 5 u 2 + l O u 3,
w2 = Es_35aE3
2 63bE2+ ( ~
+-~- 567 )
c+--~--a 2 E-
(4.6)
13774b 1287d
4 2
gives the canonical form (3.11) of Riemann surface which is associated with the
given potential.
It also should be mentioned that Treibich and Verdier considered only 'primitive
potentials', i.e. such potentials which cannot be obtained by increasing the size of
the period lattice using the formula
(4.7)
= Z2C ,+--r+--r - -
m----0 s=0
-
y~, f 2s~ 2rn~' 2~, 2J),
from other elliptic potentials (where the primed symbol of the sum ~ ' means, as
usual, that the term with m 2 + s 2 = 0 is omitted). And we shall not restrict ourselves
to the 'primitive potentials' only because the Riemann surface, which is associated
with them does not explicitly differ, to our mind, from other surfaces of the type F '~
g-
I n particular, formula (4.7) shows, that for any number of sheets of the cover, there
exists a one-gap potential and an elliptic surface F~ which is associated with it.
Remark 7. The coverings F~, which correspond to Lam6 and Treibich-Verdier
potentials, satisfy the following condition: OkR(k,~) and O~TC(k,~) have common
zeros only 'above' the half-periods of the toms/70.
The simplest surface, which belongs to this class, has n = 2, g -- 1 and is described
by the polynomial
where nj is the number of points staticwith respect to 7- (3.18) lying 'above' the point
c~ = 0 (without taking Q into account) and 'above' the points ~ = wj (j = 1, 2, 3),
140 A.O. SMIRNOV
pl + 2pk - ej k dE
E = k ' w = 2 ( p ' + 2pk + ejk), or* da = w (4.10)
E1 -- - 2 e j , E z = e.i - H i , E3 = ej + H i , (4.11)
are edges of the gaps of the spectrum of Schr6dinger operator (3.3) with a two-elliptic
one-gap potential (see equation (4.7))
The next surface is also sufficiently simple and well known [14, 15, 20, 24]
The hyperelliptic structure and the edges of the gaps are defined by the formulae
[14]
3E dE
o-* da =
2w '
Ej = -3ej, j = 1, 2, 3, /34,5 = 4 - X ~ 2 . (4.16)
At n = 4, there exist only two surfaces of the considered class. One of them has
genus one:
For n = 5, only surface of the considered class exists [281 - - F52, #~ = (0, 3, 1, 1),
k 5 - (10p + 2 e j ) k 3 - 10p'k 2 -
(4.27)
- (15p 2 - 6 e j p - 9eZ)k - 2(p - ej) 9' = 0.
w 2 = ( E 2 - 3 e j E + 8 1 H 2 - 180e~)
(4.28)
x (E 3 + 3e./E 2 + (80H 2 - 189eY)E + 5 2 8 e j H 2 - 1215e3).
142 ~ O. SMIRNOV
It is not difficult to show that the considered surface is associated with Treibich-
Verdier potential
For example, this can be done by calculating the spectrum of the potential (4.31)
from the relations (4.5), (4.6) and comparing the result with equation (4.28).
r3: k - k - o, (5.1)
or, equivalently,
Its topological characteristic equals #3 = (0, 1, 1, 1), and the corresponding potential
is the three-elliptic one-gap potential
The transition to the canonical f o r m for this surface can be realized by means of the
functions ( p - O(X)):
3k(E + 3p)
w - 2 ' (5.5)
20p 2 -- 32pes + g2
(5.6)
lOpes + 6e} -- 4P 2 -- g2
2(39 -- e j )
3 dE
~r* dc~ = ......
2w
For n = 4, one can point to only one surface belonging to the considered class [28]
Depending on the sign +, this surface will be associated with one of the two one-gap
potentials which are derived by four-folding of the period
or
3
u(x,O):8~=oP(X+s(co,+~))-2ej-4g~ (5.8b)
The functions E, w and edges of the gaps Ej are defined by the relations
k2 - p + ej
w - "2k (3//7 + 9ej :q: 10Hi), (5.10)
E1 = --2ej -4- 4 H i ,
where p # 0 is one-fifth of the period of the toms (3.13). The second is the surface of
genus two F25 with the topological characteristic #52 = (2, 1, 1, 1). It is evident that the
second surface neither corresponds to the Lam6 potential (4.1), nor to the Treibich-
Verdier potential (4.3) and it can correspond even less to the one-gap potential (4.7).
Consequently, the 5-elliptic two-gap potential of the Schr6dinger operator which
is associated with the Riemann surface F~, has some other form. The following
theorem will help us to derive this form.
THEOREM 5 (see [26]). Let the g-gap double-periodic potential (4.2) of the Schri~-
dinger operator at z ---* 0 have the asymptotics
u(x, O) _ g(gz-------2~
+ 1) + 0(22). (5.13)
3
u(z,O)= g(g+l)~(z)+Ehj(hj+l){v(x+wj)-~}+
j=l (5.14)
+ E Ik(lk + 1){p(z + (k) + p(z - (k) - 2p((k)},
k
and if
z~ # 2rnw + 2 m l w', m, m I E Z,
then it is a "nonprimitive" potential.
FINITE-GAP ELLIPTIC SOLUTIONS 145
Remark 8. The invariants g2 and g3 of the covered elliptic surface F0 (3.13) are
connected with p(6) by the following relations (p = p(6), q = p(26)):
g 2 - 2 0 p 2 --4pq -- 4q 2, (5.17)
Equation (3.12), which defines the Riemann surface F25, has, in these notations,
the form
k 5 - (1Or(a) - 6 p - q ) k 3 - lOv'(a)k 2 -
- ( 2 p ( a ) + 6p + q) p ' ( . ) = O.
Remark 9 (see [40]). The first coefficient A1 of the polynomial (3.29) written for
the Riemann surface, which is associated with the potential
A1 = Z p ( x i - xj). (5.20)
i<j
Remark 10. In the case of the potential (4.2), i.e. in the case of multiple zeros xj
of the theta function O(Uz + Wt - Z [ B), it is necessary to use the regularization
1
A1 = -~ Z ri(ri + 1)rj(rj + 1)~(xi - xj), (5.20a)
~<j
THEOREM 7. Let F'~ be the Riemann surface given by equations (3.12), (3.13),
n >. 3, and having the topological characteristic #"2 = (2, 1, 1, 1). Then there exists
such a polynomial
j-1
o o
= $Zn_3(k, B, gn_s_z,(k , o0, (5.21)
s=l
that
E - 30kl~n(k, a) + c, (5.22)
n R,~-a(k, a)
3~ (5.24)
C~ - - T
Remark 11. For many of the Riemann surfaces F~ (3.22), (3.13) there exists some
fixed values of the invariants g2, g3 of the covered toms, for which the coincidence
of some edges of the gaps (Ej ~ Ek) for the spectrum of the Schr6dinger operator
with a corresponding potential takes place. Each of such cases needs a separate
consideration.
COROLLARY 5. Let the function E, given by equation (3.20), and defined on F~,
have the form (5.22). Also let
B1 = E p ( X i - Xj), (5.25)
3<i<j
where the summation spreads only simple zeros zi, zj of the theta function O(Uz +
Wt - Z [B).
Remark 12. Th e analogous statement can be proved also for the other five types
of surfaces F~ which are parametrized by thetopological characteristics #2
type 1 #2 = (2, 1, 1, 1), n/> 3;
FINITE-GAP ELLIPTIC SOLUTIONS 147
18(q 2 + Pq -- 2/92)
E j = 9ej + 12/)+ 1 2 q - j = 1,2,3,
q - ej
or* da - 5E - 6p - 6q dE.
2w
A full description of the hyperelliptic surfaces F~ (n -<. 10) will be given in a more
extended variant of this paper.
u = g pj- (6.1)
j=l i#j
It is not difficult to see that the simplest dynamics of the Calogero-Moser system
(6.1) corresponds to the initial data in the form of the one-gap n-elliptic potential. In
this case motion of the particles is uniform and rectilinear, their speed being v = 2c0,
where co is the sum of the gap edges Ej of the spectrum.
148 ~ O. SMIRNOV
The description of pole dynamics of two-gap n-elliptic solutions for the KdV
equation is more complicated, nevertheless it can be given in a relatively simple
form.
The evolution of the general finite-gap potentials of the Schrfdinger operator under
the action of KdV flow can be described in two ways (see equations (1.12), (3.9)):
where P j ( x , t ) are zeros of the C-function (3.10) on F. From equations (6.2) and we
obtain after straightforward transformations (taking the constant co = 0),
g /~r Eg-1
--
dE = 2x, (6.3a)
E J~,~(o,o) w
j=l
Eg-2
~-~/~'J(z'*) dE -8t, (6.3b)
j=x J~j(o,o) w
Differentiating equation (6.3) and resolving them with respect to the derivatives,
we can prove the following statement.
THEOREM 8 (see [7, 9, 10]). The projections #j(z,t) of the zeros "Pj(x,t) to the
complex plane C satisfy the system of differential equations
2w(/~j) j = 1 , . . . , g, (6.4a)
s#j
8~(~,~) ~ ~,.,
m#j , j = l,...,g. (6.4b)
o,.j(~,O= 1I(~ - . . )
s#j
THEOREM 9 (see [7, 9, 10]). The function
g
W -" H ( E - #j(x,t)) = E a + E TjE a-j, (6.5)
j=l j=l
FINITE-GAP ELLIPTIC SOLUTIONS 149
where u(x, t) is a g-gap potential (6.2) of the SchrOdinger operator whose solutions
are products of any two solutions of equation (3.3). The following equality is valid
for the functions 7j (x, t):
j-1 )
2 Jj-x "q-Z CkJj-k-1- ~ , (6.7)
")'j(X, t) -- - - ~ " k=l
2nE + 3n~
dE = ~r~da, (6.8)
4w
3dE
2W = tr~ d/3, (6.9)
where ~ E F0 (3.13),
The dependence of the invariants g2, g3 of the torus Fo on the invariants g2, g3 of
the surface (3.13) is different for different Riemann surfaces Ft.
COROLLARY 6. The hypereUiptic surface F'~ has the following form in terms of
the variables E,/3,
3 n~En_ 1 + En Dj En-J,
Pn (Z) = E n + "~ (6.12)
.i=2
W 2 "-- E 5 + AE 3 + BE 2 + CE + D
3 1 3 ( 12A2~E_lq_
r = E3 + 5 A E + ~ B + -~ C - 49 /
(6.14)
4
+ +
The last statement permits us to link the invariants of elliptic surfaces F0 and F0
given by equations (3.13) and (6.10), respectively.
Direct computation of the alternative representation (6.11) for the Riemann surface
F~ (n ~< 5) gives the following results.
The hyperelliptic surface F 3 (v~ = (0, 3, 1, 1), ~ = 0), which is associated with the
two-gap Lam6 potential (4.14), has the following form in the variables E,/3 [14, 15]
9 9
F~: E 3 - ~ g2E + 4 g3 - r = 0. (6.15)
W -- r - 23 g2}-l, (6.16)
I
W-- ff(/3)(E - 6 e j ) Z { E 4 - 12ejE 3 - (2Hf - 27eZ)E z +
(6.19)
2 2+
+ 12ej(9e ] + 2 H 2 ) E - 3(11] + 24ejH~ 108e~)}-1,
12ej( 4H2 + 3e2), ek = 36e~ek - 16Hf(2ej + ek),
(6.20)
el "- 3 6 e Z e t - 16H2(2ej + e,).
For the Treibich-Verdier potential (4.31), n = 5 and v25 = (2, 1, 1, 1), we have
1
w = - ff(/3)(E 2 - 3 e j E + 81H 2 - 180eZ)Z{E6 - 6 e j E 5 +
2
~, = - 3 e j E 2 + ( 2 5 H f - 5 4 e 2 ) E , - 1 3 5 e j H f + 333e 3, (6.23)
where E, (s = 1, 2, 3) are the roots of (see equation (4.28))
The second five-sheeted covering, which is associated with the potential (5.15),
(5.16), has, in these notations, the following form (5~ = - 4 p - 4q)
--~ 0,
1
w=- 2 ~t(fl)(E - 3p - 3q)3{E 5 9(p+q)E4+
The topological characteristic of F5z, as a cover over F0, equals v25 = (0, 3, 1, 1).
Let us consider an arbitrary two-gap n-elliptic solution (3.19) with the initial value
n
where aj (t) are determined from the asymptotic of the potential u(z, t) in the vicinity
of the pole xj(t):
2
U(X,t) -- (~ -- Xj)2 "4- aj(t) + O(z - xj), (6.29)
FINITE-GAPELLIPTICSOLUTIONS 153
n n
aj(t) = 2 E p(~(t)- xj(t)) - 2 E p(zk(0)), (6.30)
k#j,k=l k=4
n
aj(O) = 4p(xj(O)) + 2 E {p(zk(O)- zj(O)) - p(xk(O))}.
k#j,k=4
:i E dE
= 2xj(t), j = 1,...,n, (6.31)
1/3
: J dE
--=-St, j=l,...,n. (6.32)
W
Substituting equation (6.34) into (6.11), we obtain that the function Ej(t), j =
1,..., n, are the roots of the algebraic equation on the nth degree.
Resolving the equation and taking into the account the link between the functions w
and ~'(/3), we find the dependence
o~j(t) = ,,oQ
ft~(%) dp
X/4p 3 -- g2~o -- gs
154 A.O. SMIRNOV
which describe the pole dynamics of the elliptic solution for KdV equations (3.19),
(6.27).
Let us finish this section with explicit formulae which link the coordinates of the
points P = (E, w) E F~, n ~< 5 and ~ E F0.
E 3 + 2793 (6.37)
F32: P(a) - 9E 2 - 2792'
F4: p ( ~ ) = {E 4 + 10ejE 3 + 9 ( 1 3 e Z - 6 H Z ) E 2 +
+ 144e./(17H f - 2
36e2)E 2 + 27ej(235ej2 - l12HZ)E +
(6.41)
+ 54ej(3456H~ - 14360eZH 2 + 14 775e4)}
+ 18(645e i f - 304H])E 2 +
i f ( a ) = 2w H \ 5 E + 4 8 p + 3q
j=l (6.44)
where
A j = q + 4p - 8ej + 2 D j ,
The spectra of Schr6dinger operator with the potentials u(z, 0) and tie(z, 0) differ
only by the shift of c, and if these are finite-gap potentials, then the hyperelliptic
surfaces F = {(w, E)} and Fc = {(w,/~)} associated with them are birationally
equivalent:
2g+l
r: = I-[ ( E - Ej), (7.2)
j----1
156 a.o. SMIRNOV
2g+l
= w, E = E - c, Ej = Ej - c. (7.4)
It follows from the equation (7.1) that there is an essential difference between such
properties of the solutions of the KdV equation, as their periodicity with respect to
x and that with respect to L
Really, if a fmite-gap solution u(x, t) is a periodic one in x, then all other finite-gap
solutions ~c(x,t)of the form (7.1), associated with birationally equivalent hyper-
elliptic surface (7.3), appears to be also periodic in x with the same period. In
other words, the property of the z-periodicity is connected with the whole class of
birationally equivalent hyperelliptic surfaces.
On the contrary, if there is some periodic in t finite-gap solution for the KdV
equation, then in the general case, none of the other fmite-gap solutions fir(z, t),
c r 0 of the form (7.1) will possess this property of periodicity. The only exception
concerns the solutions periodic with respect to both x and t
for which a countable set of periodic in x and in t solutions of the form tic(x, t)
exists, where
nX
e=6mT, n, m E Z, mOO.
Therefore, the t-periodicity of the finite-gap s~lution does not correspond to the all
birationally equivalent hyperelliptic surfaces F,, but only to one of them, or to their
countable set.
LEMMA 6. The equality (7.1) follows from equations (7.2)-(7.4) in the class of
finite-gap potentials.
Proof Let us denote by r the mapping (7.4), r: F ~ F~, 7"(w,E) = (~,/~) and
consider the finite-gap solution
where B is a period matrix for the hyperelliptic surface Fr 27riU, 27riW are vectors
of b-periods of normalized Abelian differentials of the second kind d ~ l and -4d~3,
respectively, which have a unique pole in the point/~ = ~ in the vicinity of which
the following expansions by powers of local parameter ~ are valid
~, = g-2; (7.8)
el ks a constant.
It is well known (see, for example, [34]) that matrices of period of birationally
equivalent Riemann surfaces in the case of an equivalent choice of canonical bases
of cycles are equal: B = B. Therefore, the dependence of the solution (7.5) on the
parameter c can be realized only by the aid of the vectors U and W and the constant
61.
It follows from equations (7.4), (7.8) that the local parameters r and g are connected
by the relation
= V/.i-Z..~. (7.9)
r * d ~ 3 = d f f 2 3 - 23 c dff21, (7.11)
and therefore the vectors of b-periods of the differentials dff21, -4dg23 and d~l,
- 4 d ~ 3 are connected by the relatiolts
Now, after such an introduction, let us consider a two-gap and elliptic in z solution
of KdV equation (3.19) with the initial condition (see equation (5.20a))
Remark 13. The sum of edges of gaps of the spectrum of the Schr6dinger operator
with potential (7.14) equals
5
=
j=l
Therefore, considering the solutions (3.19), (7.14), we have to replace the relations
(6.19), (6.22) by the following ones
3 5
#l(xj(t),t) - s = -~ aj(t) - ~ , (7.15)
EdE
fQ~ - 2zj(t) + 20~t. (7.16)
w
COROLLARY 10. The dynamics of the poles zj(t) of the elliptic two-periodic by t
two-gap solution (3.19), (7.14)for K d V equation (3.1) is described by the equations
~J
nzj (t) = 2nE + 5n~ dE = aj (t), (7.17)
4w
~ 3dE
- 12t =
fr 2w
--/3(t), (7.18)
Pn(gj + ~) - (~(-12t)Q,~_3($j + ~) = O,
E = + O(o~ 2) for c~ ~ O.
Two-gap elliptic solutions for KdV equations (3.19), (7.14), quite naturally, are
not unique representatives of two-periodic in t solutions for KdV equations. The
other fmite-gap t-elliptic solutions of KdV equations one can obtain by considering
the Riemann surface F of genus g, which covers in an n-sheeted manner (n >/4) an
elliptic surface F0 (see equation (6.10)) and satisfying the following requirements:
n
where /3 E F0 (6.10), si(fl) are elliptic functions of/3 with the poles in the
point/3 = 0;
FINITE-GAP ELLIPTIC SOLUTIONS 159
(2) On the surface F 'above'/3 = 0, there is a branching point of the second order
Q in which
A_n-3 3
7 ~0/3 -1/3 4- O(/31/3), (7.20)
3
= - ~ + 0(1); (7.21)
where ~ is an elliptic Weierstrass function with periods of the covered surface (6.10),
tj (z) are zeros of the theta function O(Uz + Wt - Z I B) considered as a function
of the variable t. Edges of the gaps Ej are fixed by the asymptotics of the function
E(P), P E F, at P ---* Q
Proof It follows from Lemma 6 that to construct the hyperelliptic surface associ-
ated with the periodic in t solutions of the KdV equation, first of all it is necessary
to fix a local parameter e in the vicinity of the point Q, e(Q) = 0,
=
1 =/31/3 + /3+ (7.25)
160 A.O. SMIRNOV
The normalized Abelian differential of the second kind dff23,Q will have the fol-
lowing asymptotic in the vicinity of the point Q for the considered choice of local
parameter e
dff23,O = ( -~'X
3 +O(1) ) d r ( 1- ~ + - ~ - /~0
3 _4/3 + 0(/3_2/3)) d/3" (7.26)
It is easy to see now that the normalized Abelian differentials of the second kind
-4dff23, Q and
Therefore
where d ~ ~ (7.27) and d~23,Q^(7.26) are normalized Abelian differentials of the second
kind, which are defined on F0 given by equation (6.10) and on F 3 given by equation
(6.15), respectively.
Concluding this section, we shall give a number of examples of covers which are
associated with elliptic in t solutions of KdV equations (n i> 4, fl E F0 is given be
equation (6.10)). To avoid misunderstanding, we shall use the normalization
2#+1
~-~ E j = 0,
j=l
for describing the two-gap solutions. The first covering, which we shall consider
here, will be the four-sheeted covering corresponding to two-gap Treibich-Verdier
potential (4.21)-(4.26).
~/)
A= E-6ej' ~ =-ej" (7.33)
Expressions for the functions ~3(fl) and ~3'(fl) and for the differential dfl are given by
the relations (6.18)-(6.20), and (6.9).
The next Treibich-Verdier potential (n = 5, g = 2) corresponds to the five-sheeted
cover
where
6 3
D 1 = - ~(384ej - 175ej Hf),
3 3
D2 = - ~ ( 5 3 6 1 e j - 2450ejH2),
25
D3 = - 4 ( 1 1 9 2 3 2 e 6 - 166 320e~Hf + 77 589e~H 4 -
1
- 12100H 6) - ~ 92,
3 3
(7.35)
D4 = - ~ ( 1 5 5 1 e j - 700ejHf),
2E - 3ej
E 2 - 3 e j E + 81H 2 - 180e 2"
162 A.O. SMIRNOV
where
3
D1 = - ~ ( 4 0 6 p 3 + 1303p2q + 88pq 2 - 309q3),
1
D2 = - ~ ( 2 3 7 4 p 3 . + 237p2q - 3348pq 2 - 1711q3),
2w
A= (7.37)
E - 3p - 3q"
The topological characteristics for these three coverings were pointed out earlier.
Further analysis shows that there exists one more five-sheeted covering (7.19),
(6.10) of genus three with the topological characteristic v~ = (2, 3, 1, 1):
where
~j = - 1 1 250p 3 + 750ph 2,
(7.39)
ei,k = 5625p 3 - 375p hz + 50v~h3,
The link between the variables A,/3 and E, w is given by the relations (~0 = - 4 p )
2w
= _ (7.41)
E 2 - 38pE + 16h 2 + 37p 2'
•
E - 7p
(E 2 - 38pE + 16h 2 + 37p2) 2'
u(x, O) = g(gx-------T---
+ 1) + O(1).
7j(x, 0) =
( 1 )J(g+j),(2j-1)!,
- ~ x2 (g_j)!j!
(1),
+ O x2-27:2_2 j = 1,...,go (A.1)
(1)
#j(x,O)= --~ +o -~ , j = l,...,g, (A.2)
ma + Za ( - - ~)j(g+j),(2j_l),!mg_J=O"
(g_j)!j!
j=l
Therefore
where
# #
M=~Mj,
j=l j=l
FINITE-GAP ELLIPTIC SOLUTIONS 165
From equations (A.6), (6.2) and (1.6), it follows that for any z and for any t
- x,, ,) = t).
Acknowledgements
I am grateful to V. B. Matveev, I. M. Krichever, and V. Z. Enol'skii for their attention
to the present work and for valuable discussion. I also wish to thank A. Treibich for
sending me his reprint and a copy of his doctoral thesis.
References
1. Novikov, S. P.: Funct. Analiz i Prilozhen 8(35) (1974), 54 (in Russian).
2.. Ablowitz, M. J. and Segur, H.: Solitons and Inverse Scattering Transform, SIAM, Philadelphia,
1981.
3. Faddeev, L. D. and Taktadjan, L. A.: Hamiltonian Method in the Theory of Solitons, Springer,
Berlin, Heidelberg, New York, 1987.
4. Calogero, E and Degasperia, A.: Solitons and the Spectral Transforr,~ 1, North-Holland, Amster-
dam, 1982.
5. Dodd, R. K., Eilbeck, J. C., Gibbon, J. D. and Morris, H. C.: Solitons and Nonlinear Wave
Equations, Academic Press, London, 1982.
6. Newel, A. C.: Solitons in Mathematics and Physics, SIAM, Philadelphia, 1985.
7. Zakharov, V. E., Manakov, S. V., Novikov, S. P. and Pitaevsldi, L. P.: in S. P. Novikov (ed.),
Soliton Theory: Inverse Scattering Method, Nauka, Moscow, 1980.
8. Its, A. R. and Matveev, V. B.: Teoret. i Mat. Fiz. 23(1) (1975), 51 (in Russian).
9. Matveev, V. B.: Abelian Functions and Solitons, Preprint No. 373, Univ. of Wroclaw, 1976.
10. Dubrovin, B. A., Matveev, V. B. and Novikov, S. E: Russian Math. Surveys 31(1) (1976), 55.
11. Dubrovin, B. A. and Novikov, S. E: SovietJETF 67(12) (1974), 2131.
12. Babieh, M. B., Bobenko, A. I. and Matveev, V. B; Dokl. A N SSSR 272(1) (1983), 13 (in Russian).
13. Babieh, M. B., Bobenko, A. I. and Matveev, V. B.: Izv. An SSSR 49(13) (1985), 511 (in Russian).
14. Belokolos, E. D. and Enol'skii, V. Z.: Teoret. i Mat. Fiz. 53(2) (1982), 271 (in Russian).
15. Belokolos, E.D.,Bobenko, A. I.,Enol'skii, V. Z. andMatveev, V.B.: UspekhiMat. Nauk41(2)
(1986), 3 (in Russian).
16. Matveev, V. B. and Smirnov, A~ O.: Dokl. A N S S S R 203(1), (1987), 73 (in Russian);
Matveev, V. B. and Smirnov, A. O.: Lett. Math. Phys. 14 (1987), 25.
17. Smirnov, A. O.: Mat. sb. 133(7) (1987), 382 (in Russian).
18. Smirnov, A. O.: Teoret. i Mat. Fiz. 66(1) (1986), 30 (in Russian).
19. Smirnov, A. O.: Teoret. i Mat. Fiz. 78(1) (1989), 11 (in Russian).
20. Belokolos, E. D., Bobenko, A. I., Enol'skii, V. Z., Its, A. R. and Matveev, V. B.: Algebro-
geometrical Approach to Nonlinear Evolution Equations, Springer Ser. Nonlinear Dynamics,
Springer, Berlin, Heidelberg, New York, 1994.
21. Verdier, J.-L.: New elliptic solitons,AlgebraicAnalysis, vol. II (Micio Sato 60th birthday volume),
Academic Press, New York, 1988, pp. 901-910.
22. Treibich, A. and Verdier, J.-L.: Solitons elliptiques, in the special volume for 60th anniv, of
Prof. A. Grothendieck, Birkh~iuser, Boston, 1990.
23. Treibich, A. and Verdier, J.-L.: Superpositions de solitons et sous-jacobiennes, Preprint, Paris,
1988.
24. Its, A. R. and Enol'skii, V. Z.: Funktsional. Anal. i Prilozhen. 20(1) (1986), 73 (in Russian).
166 A.O. SMIRNOV
25. Belokolos, E. D. and Enol'skii, V. Z.: Algebraicallylntegrable Nonlinear Equations and Humbert
Surfaces, Proc. Workshop on Plazma Theory and Nonlinear Waves, ~vbl. 1, Kiev 1987, World
Scientific, Singapore, 1988, p. 20.
26. Belokolos, E. D. and Enol'skii, V. Z.: Usp. Mat. Nauk44(5) (1989), 155 (in Russian).
27. Belokolos, E. D. and Enol'skii, V. Z.: Funktsional. Anal. i Prilozhen. 23(2) (1989), 70 (in
Russian).
28. Smirnov, A. O.: Matem. Zametki 45(6) (1989), 66 (in Russian).
29. Smirnov, A. O.: Matent Zametki 46(5) (1989), 100 (in Russian).
30. Smirnov, A. O.: Mat. Sb. 181(6) (1990), 804 (in Russian).
31. Farkas, H. M. and Kra, I.: Riemann Surfaces, Springer, Berlin, Heidelberg, New York, 1980.
32. Fay, J. D.: Them Functions on Riemann Surfaces, Lecture Notes in Math. 352, Springer, Berlin,
Heidelberg, New York, 1973.
33. Igusa, J.: Them-Functions, Grund. Math. Wiss., Vol. 194, Springer, Berlin, Heidelberg, New York,
1972.
34. Krazer, A.: Lehrbuch der Thetafunktionen, Teubner, Leipzig, 1903.
35. Mumford, D.: Tam Lectures on Theta, T~bl. I, Birkh~iuser, Basel, 1983; Vol. II, Birkh~iuser, Basel,
1984.
36. Dubrovin, D. A.: Russian Math. Surveys, 36(2) (1981), 11.
37. Krichever, I. M.: Dokl. A N S S S R 227(2) (1976), 291 (in Russian).
38. Krichever, I. M.: FunktsionaL Anal. i Prilozhen. 11(1), (1977), 15 (in Russian).
39. Krichever, I. M.: FunktsionaL Anal. i Prilozhen. 14(4) (1980), 45 (in Russian).
40. Enol'skii, V. Z. and Kostov, N. A.: DifferentialEquations in a SpectralParameter and Coverings
of a Toms, Preprint of Danish Technical University, 1989.
41. Airault, H., McKean, H. P. and Moser, J.: Comn~ PureAppl. Math. 30 (1977), 94.
42. Treibich, A.: C. R. Acad. Sci. Paris, S~rie I, 307 (1988), 531.
43. Treibich, A.: Duke Math. J. 59(3) (1989), 611.
44. Treibich, A. and Verdier, J.-L.: C. R. Acad. Sci. Paris, Skrie I, 311 (1990), 51.