Efficient Monte Carlo methods for continuum radiative transfer

Abstract

We discuss the efficiency of Monte Carlo methods in solving continuum radiative transfer problems. The sampling of the radiation field and convergence of dust temperature calculations in the case of optically thick clouds are both studied. For spherically symmetric clouds we find that the computational cost of Monte Carlo simulations can be reduced, in some cases by orders of magnitude, with simple importance weighting schemes. This is particularly true for models consisting of cells of different sizes for which the run times would otherwise be determined by the size of the smallest cell. We present a new idea of extending importance weighting to scattered photons. This is found to be useful in calculations of scattered flux and could be important for three-dimensional models when observed intensity is needed only for one general direction of observations. Convergence of dust temperature calculations is studied for models with optical depths 10-10000. We examine acceleration methods where radiative interactions inside a cell or between neighbouring cells are treated explicitly. In optically thick clouds with strong self-coupling between dust temperatures the run times can be reduced by more than one order of magnitude. The use of a reference field was also examined. This eliminates the need for repeating simulation of constant sources (e.g., background radiation) after the first iteration and significantly reduces sampling errors. The applicability of the methods for three-dimensional models is discussed.