Poster Presentation

Probing Environmental Impact on Galaxy Star Formation with Graph Neural Networks

Author(s): Keita Fukushima (Seoul National University)

Presenter: Keita Fukushima (Seoul National University)

In the local universe, galaxies in high-density environments are observed to exhibit suppressed star formation activity. Previous studies have investigated this environmental effect using indicators such as galaxy overdensity or the distance to the 5th-nearest neighbor. However, these approaches have limitations in capturing the direct spatial relationships among galaxies.

In this study, we perform a cosmological hydrodynamic simulation of a (200 Mpc)^3 volume using the GADGET4-Osaka code to model galaxy evolution in both high- and low-density environments. We construct a predictive model for galaxy-specific star formation rates using a Graph Neural Network (GNN), which allows us to directly incorporate galaxy positions and mutual distances into the learning process. We evaluate the importance of each input feature in the GNN model and investigate how environmental factors contribute to the variation in sSFR.