Observing gas and dust in simulations of star formation with Monte Carlo radiation transport on Voronoi meshes

TL;DR

This paper introduces a novel Monte Carlo radiative transfer method on Voronoi meshes for simulating gas and dust in star formation regions, improving accuracy over traditional Cartesian grid approaches.

Contribution

It presents the first implementation of Voronoi tessellation-based radiative transfer in MOCASSIN, enhancing simulation resolution in complex, non-uniform density fields from hydrodynamical models.

Findings

Voronoi-based method matches Cartesian results on uniform grids.

Improves resolution in dense regions with non-uniform initial conditions.

Provides synthetic spectra and emission maps of star-forming regions.

Abstract

Ionising feedback from massive stars dramatically affects the interstellar medium local to star forming regions. Numerical simulations are now starting to include enough complexity to produce morphologies and gas properties that are not too dissimilar from observations. The comparison between the density fields produced by hydrodynamical simulations and observations at given wavelengths relies however on photoionisation/chemistry and radiative transfer calculations. We present here an implementation of Monte Carlo radiation transport through a Voronoi tessellation in the photoionisation and dust radiative transfer code MOCASSIN. We show for the first time a synthetic spectrum and synthetic emission line maps of an hydrodynamical simulation of a molecular cloud affected by massive stellar feedback. We show that the approach on which previous work is based, which remapped hydrodynamical density fields onto Cartesian grids before performing radiative transfer/photoionisation calculations, results in significant errors in the temperature and ionisation structure of the region. Furthermore, we describe the mathematical process of tracing photon energy packets through a Voronoi tessellation, including optimisations, treating problematic cases and boundary conditions. We perform various benchmarks using both the original version of MOCASSIN and the modified version using the Voronoi tessellation. We show that for uniform grids, or equivalently a cubic lattice of cell generating points, the new Voronoi version gives the same results as the original Cartesian-grid version of MOCASSIN for all benchmarks. For non-uniform initial conditions, such as using snapshots from Smoothed Particle Hydrodynamics simulations, we show that the Voronoi version performs better than the Cartesian grid version, resulting in much better resolution in dense regions.