skip to content

Estimating the effective column density of three-dimensional structures using machine-learning

Relation of the local number density (nH) with the most probable visual extinction (Av) as derived from 3D hydrodynamical models (Bisbas et al. 2019, MNRAS, 485, 3097). The visual extinction is connected with column density via a constant factor.

Supervisors: Dr. Thomas Bisbas & Prof. Dr. Stefanie Walch-Gassner

To account for chemical processes in hydrodynamical simulations, the knowledge of the local, "effective", column density is required. This is a common problem in the astronomical community focusing on numerical models of the interstellar medium due to its nature of being computationally very demanding. The existence of such a relation is evident from various studies that attempted to extract this information, both computationally and theoretically (see figure).

In the SILCC project, a fast and robust routine allows for an accurate calculation of the effective column density for each computational cell. Throughout the simulations that have been performed the last few years, a wealth of outputs has been produced which includes accurate correlations of the above quantities for all different dynamical times and for different cases.

The task of this project is to focus on the development of a machine learning algorithm that will use the SILCC outputs to "learn" how to populate each computational cell of a given density distribution, with the most probable effective column density. This will provide significant help in the astronomical community modelling the ISM and will also lead to the design of next-generation radiative transfer algorithms. The project requires basic background knowledge of ISM physics and excellent computational skills in Python language.