Dwarf galaxy archaeology from chemical abundances and star formation histories

TL;DR

This paper models the chemical abundances and star formation histories of disrupted dwarf galaxies in the Milky Way halo, deriving their evolutionary parameters and demonstrating a robust, binning-free likelihood fitting method applicable to various astrophysical models.

Contribution

It introduces a likelihood-based fitting method for chemical evolution models that avoids data binning and is validated against mock data, applicable to diverse astrophysical scenarios.

Findings

GSE formed stars for about 5.4 Gyr, with star formation lasting 1.5-2 Gyr after infall.

Wukong/LMS-1's star formation lasted approximately 3.36 Gyr, with no direct age measurements in the sample.

Inferred outflow mass-loading factors align with predictions from galactic wind models.

Abstract

We model the stellar abundances and ages of two disrupted dwarf galaxies in the Milky Way stellar halo: Gaia-Sausage Enceladus (GSE) and Wukong/LMS-1. Using a statistically robust likelihood function, we fit one-zone models of galactic chemical evolution with exponential infall histories to both systems, deriving e-folding timescales of $\tau_\text{in} = 1.01 \pm 0.13$ Gyr for GSE and $\tau_\text{in} = 3.08^{+3.19}_{-1.16}$ Gyr for Wukong/LMS-1. GSE formed stars for $\tau_\text{tot} = 5.40^{+0.32}_{-0.31}$ Gyr, sustaining star formation for $\sim$$1.5 - 2$ Gyr after its first infall into the Milky Way $\sim$10 Gyr ago. Our fit suggests that star formation lasted for $\tau_\text{tot} = 3.36^{+0.55}_{-0.47}$ Gyr in Wukong/LMS-1, though our sample does not contain any age measurements. The differences in evolutionary parameters between the two are qualitatively consistent with trends with stellar mass $M_\star$ predicted by simulations and semi-analytic models of galaxy formation. Our fitting method is based only on poisson sampling from an evolutionary track and requires no binning of the data. We demonstrate its accuracy by testing against mock data, showing that it accurately recovers the input model across a broad range of sample sizes ($20 \leq N \leq 2000$) and measurement uncertainties ($0.01 \leq \sigma_\text{[$\alpha$/Fe]}, \sigma_\text{[Fe/H]} \leq 0.5$; $0.02 \leq \sigma_{\log_{10}(\text{age})} \leq 1$). Our inferred values of the outflow mass-loading factor reasonably match $\eta \propto M_\star^{-1/3}$ as predicted by galactic wind models. Due to the generic nature of our derivation, this likelihood function should be applicable to one-zone models of any parametrization and easily extensible to other astrophysical models which predict tracks in some observed space.