Monte Carlo radiative transfer in optically thick regions
A.Krieger, S. Wolf
Radiative transfer simulations based on the Monte Carlo method play a leading role in the calculation of the radiative field and corresponding observable properties of a wide variety of astrophysical objects. The core principle of this approach is the simulation of photon packages that follow individual probabilistically determined paths through a model space in which they interact with matter, like dust or gas. Eventually, Monte Carlo radiative transfer simulations can be used to calculate temperature and flux maps of the simulated object. However, in the regime of high optical depths, this approach encounters difficulties since a proper representation of the various physical processes can only be achieved by considering high numbers of simulated photon packages. As a consequence, the demand for computation time rises accordingly and thus practically puts a limit on the optical depth of models that can be simulated. Despite its enormous demand for computation time, temperature maps often still show high levels of noise and calculated flux values are severely underestimated. We present different methods that aim to solve the problem of high optical depths in dusty media. To improve the temperature calculation, we identified and precalculated repeatedly occuring simulated processes, stored their outcome in a multidimensional cumulative distribution function, and immediately replaced the basic Monte Carlo transfer during a simulation by that outcome. Additionally, we identify two problems that arise for Monte Carlo radiative transfer simulations that hinder a proper determination of flux: (i) a mismatch between the probability and weight of a photon package's path and (ii) the necessity for simulating a wide range of high scattering orders. We argue that the basic peel-off method already partly solves these problems and, futhermore, propose an advanced version of it, the extended peel-off method. Carried out performance tests of our proposed methods confirm their validity and show a significant boost in computation speed which, eventually, enables us to perform Monte Carlo radiative transfer simulations of increasingly complex systems.