A Non-Archimedean Approach to Prolongation Theory
H. Eck
Abstract:Some evolution equations possess infinite-dimensional prolongation Lie algebras which can be made finite-dimensional by using a bigger (non-Archimedean) field. The advantage of this is that convergence problems hardly exist in such a field. Besides that, the accompanying Lie groups can be easily constructed. The prolongation theory of Wahlquist and Estabrook [ 1 ] can be described as follows: Find a 1-form o9 = A dx + B dt, defined on the (x, t)-space with values in a Lie algebra g such that d e ) + ~ o ^ o9= + + [A,B d x A d t ~t Ox is a linear combination of the 2-forms which describe the evolution equation in question. If u is an analytic solution of this equation, then A and B are analytic functions of u and its x-derivatives, g is called a prolongation algebra and is found, as a presentation, as a part of the prolongation problem. In this formulation the prolongation variables seem to have disappeared but they turn out to be coordinates of a Lie group G which has (a completion of) g as a Lie algebra. One can find conservation laws by using the following theorem. THEOREM (th. 5 of 4.6 of [3]). Let G be a Lie group (local or global), M an analytic manifold, o9 a 1-form on M with values in the Lie algebra g o f G such that do) + 89 ^ o9 = O, then for every x ~ M and g ~ G there is an analytic map f , defined on an open neighbourhood of x, with values in G, such that f ( x ) = g and that f 1 d f = oJ. Two such r s coincide in the neighbourhood o f x. REMARK. G and M are analytic manifolds over the same field K where K can be ~, C, or a complete non-Archimedean field. An example of the latter will be given below. We shall now construct f explicitly for the Korteweg-de Vries equation Ut -~12UUx + Uxxx ----O, the nonlinear SchrOdinger equation ~0q, = o , i ~ , + r 2 2 3 2 H . N . VAN ECK and the Burgers equation qt + 2qqx + qxx = O. Here the following fact is used: if b~ . . . . . bn is a basis of the Lie algebra g, then e~ 'b ' . . , e ~"b" sweeps out from a neighbourhood of 1 in G when the ~; progress through a neighbourhood of 0 in K [3]. Moreover, the formula for U and V~ g 1 e U V e -U = e ad u ( v ) = V + [U, V] + ~ [U, [U, V]] + . . . is requisite [2]. K d V : The solution of the prolongation problem is A = 2 X 1 2 u X 2 3u2X3 and B = 2(Uxx + 6ua)X2 + 3 ( 8 u 3 U 2 + 2uuxx)X3 8 X 4 8 u S 5 4u2X6 4UxX 7 (see [1]) and the prolongation algebra g = H x ( A I | H has a basis ( r _ 3 , r _ l , ro, r l , r3) with multiplication table [r 1, r _ l ] = [ r _ 3 , r3] = r o ; the other commuta tors being zero. A 1 has a basis (h, y, z) with commuta torable [h, y] = 2y, [h, z] = 2z, and [ y, z] = h. C [ T ] is the ring of polynomials in the indeterminate T with complex coefficients. Every element of H commutes with any o f A 1 | C [ T ] . Details are found in [4] and [5]. We can decompose X1 . . . . . XT: 1 = 1 2 X 1 = r, l y + ~ T z , X 2 = r_ l + z , X 3 r _ 3 , X 4 = r 3 89 + ~T z , X5 = I 89 T z , X 6 = Z , ~y + X7 = ro + 89 We can split A and B accordingly, so we can look upon the KdV as having two prolongation algebras, H and A 1 | C [ T ] .