How do adaptive mesh codes like sledgehamr handle singularities or extreme density gradients?

Sledgehamr handles scale separation—situations where some features are tiny and others are enormous—through adaptive mesh refinement (AMR), a technique that uses finer computational grids only where they're needed [AG-2024.04-069].

Think of it like a microscope with a variable zoom: you don't need high magnification everywhere, only where interesting dynamics happen. Sledgehamr builds on the AMReX framework to make this manageable for simulations of coupled scalar fields, such as axion string networks where filaments of exotic matter can be extremely thin but gravitationally significant [AG-2024.04-069].

Handling singularities and discontinuities

The real challenge comes when simulations contain sharp features—density jumps, shock waves, or even mathematical singularities like puncture black holes. Sledgehamr's documentation doesn't detail these mechanisms, but related codes show the state of the art:

High-order methods with limiters. Discontinuous Galerkin (DG) schemes are becoming the standard because they naturally accommodate adaptive meshes and can achieve high accuracy on smooth regions [AG-2024.05-180]. However, unlimited DG schemes produce spurious oscillations near discontinuities. The solution is a subcell finite volume limiter—a safety mechanism that locally "downshifts" to lower-order, oscillation-free calculations only where shocks or singularities appear [AG-2024.06-410]. This preserves high accuracy in smooth regions while staying stable in rough ones.

Constraint preservation. For gravitational systems, codes must also maintain physical admissibility—positive density, positive pressure, and subluminal velocities. Methods like constraint-preserving oscillation-eliminating DG schemes use geometric quasi-linearization to guarantee these properties even across discontinuities [AG-2024.10-129].

Minimal communication. Hybrid finite-volume/DG methods reduce inter-element communication from 26 neighbors to 6, avoiding deep ghost zones and making AMR implementation simpler [AG-2025.02-506].

The preprints don't directly describe sledgehamr's singularity-handling strategy beyond mentioning it works with coupled scalar fields; you'd need the code documentation or a companion methods paper for those specifics.