ASYMPTOTICS VIA STEEPEST DESCENT FOR AN OPERATOR RIEMANN-HILBERT PROBLEM.pdfVIP

Max-Planck-Institutfur Mathematikin den NaturwissenschaftenLeipzig Asymptotics via steepest descent for anoperator Riemann-Hilbert problembySpyridon KamvissisPreprint-Nr.: 38 1998 ASYMPTOTICS VIA STEEPEST DESCENT FORAN OPERATOR RIEMANN-HILBERT PROBLEMSpyridon KamvissisABSTRACTIn this paper, we take the rst step towards an extension of the nonlinear steepestdescent method of Deift and Zhou to the case of operator Riemann-Hilbert prob-lems. In particular, we provide long range asymptotics for a Fredholm determinantarising in the computation of the probability of nding a string of n adjacent paral-lel spins up in the antiferromagnetic ground state of the spin 1/2 XXX HeisenbergChain. Such a determinant can be expressed in terms of the solution of an operatorRiemann-Hilbert factorization problem. Typeset by AMS-TEX1 2 SPYRIDON KAMVISSIS1. INTRODUCTION-MOTIVATIONIt has been established by Korepin, Izergin and Bogoliubov [KIB] that manyimportant problems arising in the theory of Quantum Inverse Scattering can bereduced to Classical Inverse Scattering Problems. In particular, the correlationfunctions for quantum solvable models can be described by solutions of classical in-tegrable equations, or, equivalently, are reducible to the solution of Riemann-Hilbertproblems. An important observation is that many such quantities are described byFredholm determinants of particular (`integrable) integral operators. As shownby dierent authors (see [DIZ] for a concise proof) the computation of these de-terminants reduces to the solution of a Riemann-Hilbert problem. Long range orlong time asymptotics for such a quantity can then be recovered via an asymptoticanalysis of an oscillatory Riemann-Hilbert problem.1 In general, such problems arenot local; in fact they can be considered as operator Riemann-Hilbert problems,which makes their analysis harder. In this paper we present a solution of one ofthese problems, by reducing it to two distinct (and essentially dierent) matrix(2x2) pr