David Robinson

David Robinson

Physics Ph.D. Candidate

Talks and presentations

Exploring the Dependence of Gas Cooling and Heating Functions on the Incident Radiation Field with Machine Learning

May 22, 2024

Lightning talk, First Stars VII, CCA, New York, NY

Abstract: Gas cooling and heating functions control gas thermodynamics and play a crucial role in galaxy formation. These functions depend on both gas properties (temperature, density, metallicity) and the incident radiation field. While they can be computed exactly with photoionization codes, this is computationally expensive and impractical to do on-the-fly in hydrodynamic simulations. As an alternative to interpolation tables of pre-computed values, we explore the capacity of machine learning to approximate cooling and heating functions with a generalized radiation field. Specifically, we use the machine learning algorithm XGBoost to predict cooling and heating functions calculated with the photoionization code Cloudy at fixed metallicity, using different combinations of photoionization rates as features. Our XGBoost models outperform a traditional interpolation approach at each fixed metallicity, regardless of feature selection. At arbitrary metallicity, we are able to reduce the frequency of the largest cooling and heating function errors compared to an interpolation table. We find that the primary bottleneck to increasing accuracy lies in accurately capturing the metallicity dependence.

The dependence of cooling and heating functions on local radiation fields

June 27, 2023

Contributed talk, New Perspectives 2023, Fermilab, Batavia, IL

Abstract: Cooling and heating functions of gas determine its energy budget and the thermal pressure support it can provide. These functions are thus a key ingredient in the physics that control how stars and galaxies form. The radiative transfer physics shaping cooling and heating functions is known, but is too computationally expensive to include in hydrodynamic simulations for realistic local radiation fields within galactic halos. Instead, a fast approximation scheme is needed. We use machine learning to investigate which wavelength bands of the radiation field most strongly affect cooling and heating functions. We use these results to develop more accurate approximation schemes to cooling and heating functions in the presence of a local radiation field.

The dependence of cooling and heating functions on local radiation fields

October 22, 2022

Contributed talk, Michigan Space Grant Consortium (MSGC) Fall Conference, University of Michigan, Ann Arbor, MI

Abstract: Cooling and heating functions determine the thermal pressure support of gas clouds and the energy budget of gas. These functions are a key component in how gas clouds collapse to form stars and subsequently galaxies. The radiative transfer physics underlying cooling and heating functions is known but is too computationally expensive to include in hydrodynamic simulations for realistic local radiation fields within galactic halos. Hence, a fast approximation to the dependence on the incident radiation field is needed to include local effects. We first discuss results from an existing approximation to heating and cooling functions in a simulation from the Cosmic Reionization on Computers project. We find that the simulated gas thermodynamics cannot be adequately described by functions computed with a spatially constant radiation field. We also discuss ongoing work using machine learning to investigate what wavelength bands of the radiation field most strongly affect cooling and heating functions.