Modeling the orbital histories of satellites of Milky Way-mass galaxies: testing static host potentials against cosmological simulations

TL;DR

This study evaluates the accuracy of static, axisymmetric host potential models in reconstructing satellite galaxy orbits within cosmological simulations, revealing significant uncertainties especially for individual orbital histories.

Contribution

It provides a rigorous benchmark of static potential models against cosmological simulations, highlighting the limitations and uncertainties in orbital reconstructions of satellite galaxies.

Findings

Orbital energy and angular momentum vary by 25% over time.

Most orbital properties are biased by less than 10%.

Uncertainties in recent orbital parameters are around 20%, earlier events up to 80%.

Abstract

Understanding the evolution of satellite galaxies of the Milky Way (MW) and M31 requires modeling their orbital histories across cosmic time. Many works that model satellite orbits incorrectly assume or approximate that the host halo gravitational potential is fixed in time and is spherically symmetric or axisymmetric. We rigorously benchmark the accuracy of such models against the FIRE-2 cosmological baryonic simulations of MW/M31-mass halos. When a typical surviving satellite fell in ($3.4-9.7$ Gyr ago), the host halo mass and radius were typically $26-86$ per cent of their values today, respectively. Most of this mass growth of the host occurred at small distances, $r\lesssim50$ kpc, opposite to dark-matter-only simulations, which experience almost no growth at small radii. We fit a near-exact axisymmetric gravitational potential to each host at $z=0$ and backward integrate the orbits of satellites in this static potential, comparing against the true orbit histories in the simulations. Orbital energy and angular momentum are not well conserved throughout an orbital history, varying by 25 per cent from their current values already $1.6-4.7$ Gyr ago. Most orbital properties are minimally biased, $\lesssim10$ per cent, when averaged across the satellite population as a whole. However, for a single satellite, the uncertainties are large: recent orbital properties, like the most recent pericentre distance, typically are $\approx20$ per cent uncertain, while earlier events, like the minimum pericentre or the infall time, are $\approx40-80$ per cent uncertain. Furthermore, these biases and uncertainties are lower limits, given that we use near-exact host mass profiles at $z=0$.