Local Taylor-based polynomial quasi-Trefftz spaces for scalar linear equations

Abstract

Trefftz-type of Galerkin methods for numerical PDEs use discrete spaces of problem-dependent functions. While Trefftz methods leverage discrete spaces of local exact solutions to the governing PDE, Taylor-based quasi-Trefftz methods leverage discrete spaces of local approximate solutions to the governing PDE. This notion of approximate solution, understood in the sense of a small Taylor remainder, is defined for differential operators with smooth variable coefficients. In both cases, it is possible to use discrete spaces much smaller than standard polynomial space to achieve the same orders of approximation properties. The present work is the first systematic study of local Taylor-based polynomial quasi-Trefftz spaces characterized as the kernel of the quasi-Trefftz operator, defined as the composition of Taylor truncation with the differential operator. The proposed linear algebra framework reveals the general structure of this linear operator and applies to any non-trivial linear scalar differential operator with smooth coefficients. It results in a fully explicit procedure to construct a local quasi-Trefftz basis valid in all dimension and for operators of any order, guaranteeing a minimal computational cost for the construction of these equation-dependent bases. The local quasi-Trefftz space is formally defined as the kernel of a linear operator between spaces of polynomials. The systematic approach relies on a detailed study of the structure of this operator, strongly leveraging the graded structure of polynomial spaces.