The Dust Grain Ice Formation Inverse Problem

Inverse Problem Figure

The formation of interstellar dust grain ices seems to trigger the formation of complex molecules, such as pre-biotic molecules, observed in the gas phase during star formation. Therefore, understanding ice formation is crucial to our knowledge of the chemistry of star forming regions. The dust grain ice formation inverse problem is to infer the physical parameters of starless and star forming cores that govern the formation of interstellar dust grain ices from ice observations. To deal with this problem, we use observations [2] from the existing literature for water ice, methanol ice and other ice species in dark molecular clouds and UCL_CHEM, a time dependent, gas grain chemical model [1]. The model simulates the gas-grain chemistry covering broad astronomical physical conditions, typical of starless and star forming cores. The observations provide the guideline for constraining the physical parameters of the model and get to understand under what conditions icy mantles form.

We work on the inverse problem by employing machine learning techniques in conjunction with bayesian statistics. Bayesian inference is basically employed to calculate the posterior probability of a parameter set to produce the observed abundances. To explore the parameter space, either a Markov Chain Monte Carlo sampler is used or a pre-determined grid of physical parameters. Likelihood, prior probabilities calculation as well as various data mining tasks are implemented using machine learning techniques such as nearest neighbour analysis, kd-trees, hierarchical clustering etc.


Viti et al. 2004

Whittet et al. 2011

Page last modified on 11 apr 17 16:18 by Roger Wesson