Sparse grid quadrature on products of spheres (Q891778)

Denoting the unit sphere in \(\mathbb R^3\) by \(\mathbb S^2\), the paper is devoted to the development and investigation of algorithms for the approximate computation of integrals over \((\mathbb S^2)^k\) with (typically large) \(k \in \mathbb N\). The integrand functions are assumed to be contained in a weighted tensor product of reproducible kernel Hilbert spaces over this domain. In contrast to the frequently considered quasi-Monte Carlo methods, the authors construct formulas with arbitrary weights and use the additional flexibility obtained in this way to optimize the convergence behaviour. Specifically, a dimension adaptive method is developed and shown to be optimal. Numerical examples illustrate the theoretical results.