Auriga Superstars: Improving the resolution and fidelity of stellar dynamics in cosmological galaxy simulations

TL;DR

Cosmological galaxy simulations are limited by stellar-resolution constraints, hindering detailed stellar dynamics analyses of discs and halos. The authors introduce the Superstars method, which increases stellar mass resolution by forming $N$ lower-mass star particles per star-formation event with total mass $m_ ext{cell}$ and imparting birth kicks with width $\sigma_ ext{kick}$, yielding initial velocity $ extbf{v}_*= extbf{v}_ ext{cell}+ extbf{v}_ ext{kick}$ while conserving momentum and leaving winds unchanged. They demonstrate eightfold and sixty-fourfold improvements on a Milky Way–mass Auriga halo (Au-6) at modest to substantial computational cost, with global properties and satellite statistics preserved and substantially improved sampling of disc and halo substructure. This approach enables high-fidelity stellar dynamics in cosmological contexts without re-calibrating gas physics, facilitating detailed studies of bars, spirals, streams, and other dynamical features in realistic environments.

Abstract

Cosmological hydrodynamical simulations have become an indispensable tool to understand galaxies. However, computational constraints still severely limit their numerical resolution. This not only restricts the sampling of the stellar component and its direct comparison to detailed observations, but also the precision with which it is evolved. To overcome these problems we introduce the \emph{Superstars} method. This method increases the stellar mass resolution in cosmological galaxy simulations in a computationally inexpensive way for a fixed dark matter and gas resolution without altering any global properties of the simulated galaxies. We demonstrate the \emph{Superstars} method for a Milky Way-like galaxy of the Auriga project, improving the stellar mass resolution by factors of $8$ and $64$ at an additional cost of only $10\%$ and $500\%$, respectively. We show and quantify that this improves the sampling of the stellar population in the disc and halo without changing the properties of the central galaxy or its satellites, unlike simulations that change the resolution of all components (gas, dark matter, stars). Moreover, the better stellar mass resolution reduces numerical heating of the stellar disc in its outskirts and keeps substructures in the stellar disc and inner halo more coherent. It also makes lower mass and lower surface brightness structures in the stellar halo more visible. The \emph{Superstars} method is straightforward to incorporate in any cosmological galaxy simulation that does not resolve individual stars.

Figures (10)

• Figure 1: Quantification of the additional velocity kick that stars get at formation with the Superstars method. The panels show a simulation with $8$x better stellar mass resolution (Stars x8). The left panel shows the average absolute velocity kick imparted on star particles at formation as a function of their radial coordinate in the disc at $z=0$. The shaded area shows the $16$th and $84$th percentiles. We include all star particles with a vertical position $|z_\mathrm{disc}|<30~\mathrm{kpc}$ at $z=0$. The middle panel shows the average absolute velocity kick as a function of the input velocity dispersion of the birth cell. The right panel shows the average absolute velocity kick as a function of the density of the birth cell. All panels use mass-weighted averages for the Superstars particles. In the 2D histograms (middle and right panels) lighter colours correspond to more stars/higher total stellar mass per bin. Note the different range of values on the vertical axes. The typical additional velocity kick imparted on star particles at their birth is $5{-}10\,\mathrm{km/s}$.
• Figure 2: Face-on and edge-on stellar light projections at $z=0$ for three different resolution runs of the Auriga project (top rows) and for three runs increasing the stellar mass resolution only. The images show the $K$-band, $B$-band, and $U$-band as RGB channels. The left column shows the galaxy at the exact same resolution, but for a run with a different random number seed. Changing the resolution of all components (gas, dark matter, stars) clearly changes the global properties of the galaxy in a systematic way, but this is not the case when we only increase the stellar mass resolution with the Superstars method.
• Figure 3: Star formation history of simulations that only change the mass resolution of one component, stellar mass or dark matter, (left panels) and for simulations that change the resolution of all components, that is, gas, stars, and dark matter (right panel). The lines show the median and the shaded bands the $16$th-$84$th percentiles of all realisations run for the different configurations (see Table \ref{['tab:simulations']}; note that for the computationally expensive setups "Stars $\times64$" (orange), "All $\times8$ (L3)" (purple) and "All $\times64$ (L2)" (pink) only one realisation is available and therefore only a single line is shown). The star formation history remains the same if we change only the mass resolution of the dark matter or only the stars. In contrast, the star formation rate is significantly and systematically increased at times before $z=0.5$ if we change the gas resolution as well.
• Figure 4: Comparison of the properties of the galactic stellar discs at $z=0$ for realisations with different stellar mass and dark matter mass resolutions, but the same gas mass resolution. The columns show the face-on stellar mass surface density computed with a depth $|z|<30\,\mathrm{kpc}$ (left panel), the vertical velocity dispersion (middle panel), and the radial velocity dispersion of the stars in the disc. The lines show the median and the shaded bands the $16$th-$84$th percentiles of all realisations run for the different configurations (see Table \ref{['tab:simulations']}). The vertical dashed gray lines show the average optical radius of the reference simulations (blue). The bottom row shows the deviation from the median of the reference simulations. Neither increasing the stellar mass resolution, nor the dark matter mass resolution changes any properties of the stellar disc significantly for the runs with $8\times$ better stellar mass or dark matter mass resolution. At large radii $R_\mathrm{disc}>20\,\mathrm{kpc}$ the simulation with the $64$x better stellar mass resolution (orange) shows an increases in the surface density and a decrease in the velocity dispersions. However, it is unclear from this single, computationally expensive realisation how systematic this difference is.
• Figure 5: Satellite stellar mass function at $z=0$ for the reference run and simulations with improved stellar mass and dark matter mass resolution. The lower panel shows the satellite stellar mass function relative to the median of the reference runs. For a satellite stellar mass $M_{*,sat}>10^6\,\mathrm{M_\odot}$ the mass function is identical. Only for lower satellite stellar masses is there an indication that there are systematically slightly more satellites in the simulations with higher dark matter resolution (grey) and slightly fewer satellites in the simulations with higher stellar mass resolution (yellow, orange).