Scalable angular adaptivity for Boltzmann transport

This paper describes an angular adaptivity algorithm for Boltzmann transport applications which for the first time shows evidence of O(n) scaling in both runtime and memory usage, where n is the number of adapted angles. This adaptivity uses Haar wavelets, which perform structured h-adaptivity built on top of a hierarchical P0 FEM discretisation of a 2D angular domain, allowing different anisotropic angular resolution to be applied across space/energy. These wavelets can be mapped back to their underlying P0 space scalably, allowing traditional DG-sweep algorithms if desired. Instead we build a spatial discretisation on unstructured grids designed to use less memory than competing alternatives in general applications and construct a compatible matrix-free multigrid method which can handle our adapted angular discretisation. Fixed angular refinement, along with regular and goal-based error metrics are shown in three example problems taken from neutronics/radiative transfer applications. •Shows evidence of scalable adaptivity in runtime/memory up to 15 levels of refinement.•Produces angular discretisation with solid angle of 10−9.•Two problems shown would take 1013-1014 DOFs to resolve with our Haar discretisation if adaptivity were not used.