Multiscale expansions of difference equations in the small lattice spacing regime, and a vicinity and integrability test. I

TL;DR

The paper introduces an algorithmic method to evaluate how closely discretized versions of integrable PDEs resemble their continuous counterparts and to assess their integrability properties, aiding in the development of better numerical schemes.

Contribution

It presents a systematic procedure combining multiscale expansion and integrability testing to analyze discretizations of integrable PDEs, including a new way to construct approximate symmetries and constants of motion.

Findings

The method effectively measures the 'distance' between discrete and continuous integrable systems.

It can test the asymptotic integrability of discretizations.

The approach helps improve numerical schemes for integrable PDEs.

Abstract

We propose an algorithmic procedure i) to study the ``distance'' between an integrable PDE and any discretization of it, in the small lattice spacing epsilon regime, and, at the same time, ii) to test the (asymptotic) integrability properties of such discretization. This method should provide, in particular, useful and concrete informations on how good is any numerical scheme used to integrate a given integrable PDE. The procedure, illustrated on a fairly general 10-parameter family of discretizations of the nonlinear Schroedinger equation, consists of the following three steps: i) the construction of the continuous multiscale expansion of a generic solution of the discrete system at all orders in epsilon, following the Degasperis - Manakov - Santini procedure; ii) the application, to such expansion, of the Degasperis - Procesi (DP) integrability test, to test the asymptotic integrability properties of the discrete system and its ``distance'' from its continuous limit; iii) the use of the main output of the DP test to construct infinitely many approximate symmetries and constants of motion of the discrete system, through novel and simple formulas.