GalactiKit: reconstructing mergers from $z=0$ debris using simulation-based inference in Auriga

TL;DR

GalactiKit is a simulation-based inference method that reconstructs the properties of galaxy mergers from debris data, improving understanding of Milky Way formation by combining chemical and dynamical information.

Contribution

This work introduces GalactiKit, a novel data-driven approach that uses simulation-based inference to accurately reconstruct merger histories from debris properties.

Findings

Kinematic data traces the merger infall time.

Chemical data is essential for mass estimates.

Combining chemo-dynamical data yields the best reconstruction accuracy.

Abstract

We present GalactiKit, a data-driven methodology for estimating the lookback infall time, stellar mass, halo mass and mass ratio of the disrupted progenitors of Milky Way-like galaxies at the time of infall. GalactiKit uses simulation-based inference to extract the information on galaxy formation processes encoded in the Auriga cosmological MHD simulations of Milky Way-mass halos to create a model that relates the properties of mergers to those of the corresponding merger debris at $z=0$. We investigate how well GalactiKit can reconstruct the merger properties given the dynamical, chemical, and the combined chemo-dynamical information of debris. For this purpose, three models were implemented considering the following properties of merger debris: (a) total energy and angular momentum, (b) iron-to-hydrogen and alpha-to-iron abundance ratios, and (c) a combination of all of these. We find that the kinematics of the debris can be used to trace the lookback time at which the progenitor was first accreted into the main halo. However, chemical information is necessary for inferring the stellar and halo masses of the progenitors. In both models (b) and (c), the stellar masses are predicted more accurately than the halo masses, which could be related to the scatter in the stellar mass-halo mass relation. Model (c) provides the most accurate predictions for the merger parameters, which suggests that combining chemical and dynamical data of debris can significantly improve the reconstruction of the Milky Way's assembly history.