Spacetime Wavelet Method for Linear Boundary-Value Problems in Sylvester Matrix Equation Form

Abstract: We present a high-order spacetime numerical method for discretizing and solving linear initial-boundary value problems using wavelet-based techniques with user-prescribed error estimates. The spacetime wavelet discretization yields a system of algebraic equations resulting in a Sylvester matrix equation. We solve this system with a Global Generalized Minimal Residual (GMRES) method in conjunction with a wavelet-based recursive algorithm to improve convergence. We perform rigorous verification studies using linear partial differential equations (PDEs) with both convective and diffusive terms. The results of these simulations show the high-order convergence rates for the solution and derivative approximations predicted by wavelet theory. We demonstrate the utility of solving the Sylvester equation through comparisons to the commonly-used Kronecker product formulation. We show that our recursive wavelet-based algorithm that generates initial guesses for the iterative Global GMRES method improves the performance of the solver.