Oral Presentation

Interpretable Modeling of the Initial Density Field

Author(s): Shunya Uchida (Nagoya University)

Presenter: Shunya Uchida (Nagoya University)

"Galaxy formation is driven by a complex interplay of factors, including dark matter halo properties and the surrounding environment. However, the initial density field may also play a pivotal role, offering a more complete picture of the mechanisms governing galaxy evolution.

Previous studies (e.g., Dalal et al. 2008; Tinker et al. 2017) have demonstrated that assembly bias in massive haloes is imprinted in the fluctuations of the primordial density field. The formation history of haloes—including their formation times and growth trajectories—is significantly influenced by these initial fluctuations. This history, in turn, has been linked to quenching in massive galaxies, with older haloes more likely to host quenched galaxies. In contrast, such correlations appear weaker in lower-mass galaxies, where tidal interactions and other large-scale environmental effects dominate.

In this study, we extend these findings by developing a neural network framework to predict galaxy properties—specifically stellar mass and star formation rate—at z = 0 using the IllustrisTNG300 simulation. We incorporate the initial density field at z = 20 as a representation of the large-scale environment, along with dark matter subhalo properties of both host and neighboring subhaloes at z = 0. Using SHAP (SHapley Additive exPlanations), we quantify the relative contributions of the initial density field and local environmental features, and explore their influence on galaxy properties across different populations.

Our results shed light on the environmental imprint of the initial density field on galaxy evolution at z = 0, providing new insights into the galaxy–halo connection and guiding future efforts in reconstructing cosmological initial conditions."