Characterizing Starbursts: Mapping the Physical Conditions of NGC 253 with Neural Networks

Starburst galaxies exhibit powerful, complex networks of star-formation feedback processes that heat and interact with interstellar gas and dust. Recognizing how feedback from one generation of stars affects the material that will form the next is crucial for understanding both the life cycle of star formation in starburst galaxies as well as how starbursts fit into galaxy evolution overall. By comparing molecular gas measurements to chemical models, we can constrain the physical conditions of star-forming regions and investigate the dominant feedback mechanisms influencing gas properties. To model more star-forming regions at higher resolution, we implement neural network-based models to replace the role of time-consuming chemical models in parameter inferencing algorithms. I will demonstrate that we can use neural networks to reproduce the output of chemical models in a fraction of the time, allowing us to analyze molecular gas conditions in nearby galaxies more completely than ever before. We constrain our models with ALCHEMI measurements of HCN and HNC toward NGC 253 and infer key molecular gas properties, (e.g. volume density, cosmic-ray ionization rate) across NGC 253 on 50 pc scales. We find that cosmic-ray ionization plays a dominant role in the feedback budget in NGC 253's nucleus, while UV heating may prove more important farther from the nucleus. I will discuss potential causes of this distribution in feedback mechanisms and its impacts on future star formation.