Monte Carlo radiative transfer in embedded prestellar cores

We implement a Monte Carlo radiative transfer method, that uses a large number of monochromatic luminosity packets to represent the radiation transported through a system. These packets are injected into the system and interact stochastically with it. We test our code against various benchmark calculations and determine the number of packets required to obtain accurate results under different circumstances. We then use this method to study cores that are directly exposed to the interstellar radiation field (non-embedded cores). Our code predicts temperature and intensity profiles inside these cores which are in good agreement with previous studies using different radiative transfer methods. We also explore a large number of models of cores that are embedded in the centre of a molecular cloud. We study cores with different density profiles embedded in molecular clouds with various optical extinctions and we calculate temperature profiles, SEDs and intensity profiles. Our study indicates that the temperature profiles in embedded cores are less steep than those in non-embedded cores. Deeply embedded cores (ambient cloud with visual extinction larger than 15-25) are almost isothermal at around 7-8 K. The temperature inside cores surrounded by an ambient cloud of even moderate thickness ( $A_{\rm V}\sim 5$) is less than 12 K, which is lower than previous studies have assumed. Thus, previous mass calculations of embedded cores (for example in the $\rho$ Ophiuchi protocluster), based on mm continuum observations, may underestimate core masses by up to a factor of 2. Our study shows that the best wavelength region to observe embedded cores is between 400 and 500 $\mu {\rm m}$, where the core is quite distinct from the background. We also predict that very sensitive observations (~ 1-3 MJy ${\rm\, sr^{-1}}$) at 170-200 $\mu {\rm m}$ can be used to estimate how deeply a core is embedded in its parent molecular cloud. The upcoming Herschel mission (ESA, 2007) will, in principle, be able to detect these features and test our models.