Robust and scalable h-adaptive aggregated unfitted finite elements for interface elliptic problems.

This work introduces a novel, fully robust and highly-scalable, h -adaptive aggregated unfitted finite element method for large-scale interface elliptic problems. The new method is based on a recent distributed-memory implementation of the aggregated finite element method atop a highly-scalable Cartesian forest-of-trees mesh engine. It follows the classical approach of weakly coupling nonmatching discretisations at the interface to model internal discontinuities at the interface. We propose a natural extension of a single-domain parallel cell aggregation scheme to problems with a finite number of interfaces; it straightforwardly leads to aggregated finite element spaces that have the structure of a Cartesian product. We demonstrate, through standard numerical analysis and exhaustive numerical experimentation on several complex Poisson and linear elasticity benchmarks, that the new technique enjoys the following properties: well-posedness, robustness with respect to cut location and material contrast, optimal (h-adaptive) approximation properties, high scalability and easy implementation in large-scale finite element codes. As a result, the method offers great potential as a useful finite element solver for large-scale interface problems modelled by partial differential equations. • Novel approach enabling large-scale simulations for unfitted/embedded interfaces. • Methodology based on the approach of weakly coupling interface-overlapping meshes. • AgFEM is endowed with optimal (h-adaptive) approximation properties. • AgFEM shows remarkable robustness to cut location and material contrast. • Weak scaling tests up to 90M degrees of freedom and 2K CPU cores on complex domains. [ABSTRACT FROM AUTHOR]