Emulating InterStellar Medium Chemistry with Physics Informed Neural Networks

Explore the application of Physics Informed Neural Networks (PINNs) in emulating InterStellar Medium (ISM) chemistry, focusing on the formation of Giant Molecular Clouds (GMCs) and stars. Delve into the challenges of incorporating non-equilibrium chemistry in astrophysical simulations, including the high number of reactions, brief evolutionary timescales, and complex ODEs. Examine the potential of emulators as fast alternatives to traditional chemical network solvers. Learn about a PINN model designed to solve thermo-chemical ODEs within specific ranges of chemical species and temperature initial conditions. Evaluate the model's performance in terms of accuracy and efficiency, comparing it to conventional solvers. Discover recent improvements in accounting for stellar radiation in GMC simulations using modified PINN models. Gain insights into cutting-edge research at the intersection of astrophysics and machine learning, presented by Lorenzo Branca from the Alan Turing Institute in this 46-minute talk.