Speaker
Description
Estimating the amplitude of Newtonian Noise due to elastic waves in the ground requires accurate modelling of the medium and the surrounding geometry. Time-domain methods based on spectral elements [1] are popular but rely on hexahedral elements which may struggle to represent accurately complex shapes. Furthermore, small elements near caves and tunnels may overconstrain the time step (due to the CFL condition). On the other hand, frequency-domain approaches remove the time-stepping constraint and can easily handle damping.
We present our frequency-domain solver for Navier equations based on non-overlapping Domain Decomposition Methods (DDM), which has been shown to be very memory-efficient [2]. It can simulate elastic waves with high-order elements on an arbitrary tetrahedral mesh that can easily represent the Einstein Telescope's underground infrastructure and is naturally suited to multiple right-hand sides to evaluate various sources of noise. In this context, the post-processing of the solution to estimate the gravitational field perturbation is natural and embedded in the workflow.
We present scalability results on large CPU clusters and discuss perspectives on GPU acceleration.
[1] D. Komatitsch and J.-P. Vilotte, "The spectral-element method: an efficient tool to simulate the seismic response of 2D and 3D geological structures", Bull. Seismol. Soc. Am., 88(2), 368–392, 1998.
[2] B. Martin, P. Jolivet, and C. Geuzaine, "Comparison of substructured non-overlapping domain decomposition and overlapping additive Schwarz methods for large-scale Helmholtz problems with multiple sources", J. Comput. Phys., 548, 114557, 2026, doi:10.1016/j.jcp.2025.114557.