Transferring Data from a Voronoi Mesh to an Adaptive Cartesian Grid in Pursuit of Self-consistent Top-down Star Formation

TL;DR

This paper introduces VorAMR, a novel tool for transferring data from unstructured Voronoi meshes to adaptive Cartesian grids, enabling multi-scale star formation simulations with minimal errors and consistent evolution.

Contribution

VorAMR is the first tool to transfer data from Voronoi meshes to AMR Cartesian grids, integrated into Torch for improved multi-scale star formation modeling.

Findings

VorAMR achieves a few percent error in mass and energy conservation.

Torch simulations produce similar star formation patterns as original Voronoi-based simulations.

Preliminary results show more compact clusters and increased stellar material in Torch simulations.

Abstract

Unstructured Voronoi mesh simulations offer many advantages for simulating self-gravitating gas dynamics on galactic scales. Adaptive mesh refinement (AMR) can be a powerful tool for simulating the details of star cluster formation and gas dispersal by stellar feedback. Zooming in from galactic to local scales using the star cluster formation simulation package Torch requires transferring simulation data from one scale to the other. Therefore, we introduce VorAMR, a novel computational tool that interpolates data from an unstructured Voronoi mesh to an AMR Cartesian grid. VorAMR is integrated into the Torch package, which integrates the FLASH AMR magnetohydrodynamics code into the Astrophysical Multipurpose Software Environment. VorAMR interpolates data from an AREPO simulation to a FLASH AMR grid using a nearest-neighbor particle scheme, which can then be evolved within the Torch package, representing the first ever transfer of data from a Voronoi mesh to an AMR Cartesian grid. Interpolation from one numerical representation to another results in an error of a few percent in global mass and energy conservation, which could be reduced with higher-order interpolation of the Voronoi cells. We show that the postinterpolation Torch simulation evolves without numerical abnormalities. A preliminary Torch simulation is evolved for 3.22 Myr and compared to the original AREPO simulation over the same time period. We observe similarly distributed star cluster formation between the two simulations. More compact clusters are produced in the Torch simulation as well as 2.3 times as much stellar material as in AREPO, likely due to the differences in resolution.