Sparse Grid Approximation Spaces for Space-Time Boundary Integral Formulations of the Heat Equation
This work provides improved numerical methods for solving heat equation boundary value problems, which is important for computational science and engineering applications.
The paper develops stable and accurate numerical schemes for boundary integral formulations of the heat equation, achieving higher convergence rates using anisotropic sparse grid discretisations compared to full tensor product methods.
The aim of this paper is to develop stable and accurate numerical schemes for boundary integral formulations of the heat equation with Dirichlet boundary conditions. The accuracy of Galerkin discretisations for the resulting boundary integral formulations depends mainly on the choice of discretisation space. We develop a-priori error analysis utilising a proof technique that involves norm equivalences in hierarchical wavelet subspace decompositions. We apply this to a full tensor product discretisation, showing improvements over existing results, particularly for discretisation spaces having low polynomial degrees. We then use the norm equivalences to show that an anisotropic sparse grid discretisation yields even higher convergence rates. Finally, a simple adaptive scheme is proposed to suggest an optimal shape for the sparse grid index sets.