Multiresolution Radiative Transfer for Line Emission

Abstract

Radiative transfer calculations are crucial for modeling interstellar clouds because they provide the link between physical conditions in a cloud and the radiation observed from it. Three-dimensional simulations of magnetohydrodynamic (MHD) turbulence are used to study the structure and dynamics of interstellar clouds and even to follow the initial stages of core collapse leading to the formation of new stars. The wide range of size scales in such models poses a serious challenge to radiative transfer calculations. In this paper we describe a new computer code that solves the radiative transfer problem on multiresolution grids. If the cloud model is from an MHD simulation on a regular Cartesian grid, criteria based, for example, on local density or velocity gradients are used to refine the grid by dividing selected cells into subcells. Division can be repeated hierarchically. Alternatively, if the cloud model is from MHD simulations with adaptive mesh refinement, the same multiresolution grid used for the MHD simulation is adopted in the radiative transfer calculations. High discretization is often needed only in a small fraction of the total volume. This makes it possible to simulate spectral line maps with good accuracy, also minimizing the total number of cells and the computational cost (time and memory). Multiresolution models are compared with models on regular grids. In the case of moderate optical depths (e.g., τ ~ few) an accuracy of 10% can be reached with multiresolution models where only 10% of the cells of the full grid are used. For optically thick species (τ ~ 100), the same accuracy is achieved using 15% of the cells. The relation between accuracy and number of cells is not found to be significantly different in the two MHD models we have studied. The new code is used to study differences between LTE and non-LTE spectra and between isothermal and nonisothermal cloud models. We find significant differences in line ratios and individual spectral line profiles of the isothermal and LTE models relative to the more realistic nonisothermal case. The slope of the power spectrum of integrated intensity is instead very similar in all models.