The DREAMS Project: Disentangling the Impact of Halo-to-Halo Variance and Baryonic Feedback on Milky Way Satellite Galaxies

TL;DR

The DREAMS study investigates satellite galaxies around Milky Way–mass hosts to quantify the relative importance of halo-to-halo variance versus baryonic feedback uncertainties. By combining a large suite of 1,024 hydrodynamic DREAMS realizations with an emulator (NeHOD) that uses a normalizing flow and a variational diffusion model, the authors robustly assess how changes in SN wind energy, wind speed, AGN feedback, and cosmology affect satellite mass functions, radial distributions, inner DM slopes, and half-light radii. They find halo-to-halo variance generally dominates the scatter in most satellite properties, while baryonic physics, particularly SN feedback, can modulate stellar masses and sizes, though the inner DM density remains cuspy due to adiabatic contraction. A notable discrepancy remains in the sizes of high-mass satellites compared with SAGA, likely pointing to missing ISM physics or alternative feedback implementations, which motivates future studies across different galaxy-formation models and environment priors. These results establish a framework to quantify theoretical uncertainties in MW-mass satellite populations and enable efficient exploration of large parameter spaces with emulators.

Abstract

We analyze the properties of satellite galaxies around 1,024 Milky Way-mass hosts from the DREAMS Project, simulated within a $Λ$CDM cosmology. Utilizing the TNG galaxy-formation model, the DREAMS simulations incorporate both baryonic physics and cosmological uncertainties for a large sample of galaxies with diverse environments and formation histories. We investigate the relative impact of the physical uncertainty from the galaxy-formation model on predicted satellite properties using four metrics: the satellite stellar mass function, radial distribution, inner slope of dark matter density profile, and stellar half-light radius. We compare these predictions to observations from the SAGA Survey and the DREAMS N-body simulations and find that uncertainties from baryonic physics modeling are subdominant to the scatter arising from halo-to-halo variance. Where baryonic modeling does affect satellites, the supernova wind energy has the largest effect on the satellite properties that we investigate. Specifically, increased supernova wind energy suppresses the stellar mass of satellites and results in more extended stellar half-light radii. The adopted wind speed has only a minor impact, and other astrophysical and cosmological parameters show no measurable effect. Our findings highlight the robustness of satellite properties against uncertainties in baryonic physics modeling.

Figures (8)

• Figure 1: A gallery of the first 10 (out of 1,024) simulated MW-mass systems from the DREAMS CDM suite. The top row displays DM density projections for the host and its surrounding satellites out to 300 kpc, ordered from left to right by increasing SN wind energy, $\widebar{e}_w$. Each column below the top row shows the corresponding population of resolved satellite galaxies ($M_*>10^7~\mathrm{M}_\odot$) for that host. The satellite images are shown in order from most to least massive and have a $20\times20$ kpc field of view. The red, green, and blue color channels for the satellite images are created from the $i$, $r$, and $g$ Sloan filter luminosities. The figure highlights the significant variation in satellite numbers, sizes, and overall appearance.
• Figure 2: Left: SMF for MW-mass galaxies in the emulated DREAMS Fiducial (black) and Varied (red) datasets, compared to the observations from SAGA SAGAIII. SAGA hosts are down-selected to fall within the mass range of the DREAMS host galaxies, reducing the observational sample from 101 to 80 MW-mass systems. The satellites are taken from their Gold sample, which has $M_* \geq 10^{7.5}~\mathrm{M}_\odot$ and is 94% complete. Right: Normalized 2D projected radial distribution, $R_{\rm sat}$, for the same satellites shown in the left panel. For both panels, lines indicate the average value, while the shaded bands show the 1$\sigma$ spread. The emulated DREAMS data agrees well with the SAGA results for both distributions. The primary exception is the radial distribution below 2D $R_{\rm sat} \lesssim 25$ kpc. Notably, the DREAMS Fiducial and Varied results are comparable to each other, suggesting that uncertainties in the cosmological and astrophysical parameter variations are subdominant to intrinsic halo-to-halo variance and uncertainties in the host halo mass.
• Figure 3: Left: DM density profiles for three randomly selected satellites in the DREAMS CDM suite, covering a wide range of satellite mass. Black lines show the profiles from the hydrodynamic suite, and the dot-dashed purple lines show the same satellites from the N-body suite. The vertical gray band shows the region over which the slope of the density profile is calculated. Signs of adiabatic contraction are evident within 2 kpc as the hydrodynamic density profile becomes cuspier than the corresponding N-body profile. Right: The slope of the DM density profile at 1.5 kpc, $\alpha_{{\text{1.5~kpc}}}$, versus total satellite mass for the emulated DREAMS Fiducial (black) and Varied (red) datasets. For comparison, the results from the DREAMS N-body simulations are also provided in purple. Lines indicate the mean value with the band covering the $1\sigma$ spread. The density slope is evaluated between 1.3--1.7 kpc of the satellite center, where effects from gravitational softening are reduced. No stellar mass cut is applied for the satellites in this figure to allow for a fair comparison with the N-body suite. In general, the inclusion of baryons steepens the slope across all halo masses, but the range in $\alpha_{{\text{1.5~kpc}}}$ appears to be largely driven by halo-to-halo variance and uncertainty in the host halo mass.
• Figure 4: Left: Satellite size-mass relation for the emulated DREAMS Fiducial (black) and Varied (red) datasets. The observed half-light radii from SAGA ($1\sigma$ spread) are shown by the blue band SAGAVI. Size is measured using the r-band 2D circularized stellar half-light radius for all datasets, see Section \ref{['sec:calcs']} for details. Satellites falling below the pink-dashed line or to the left of the green dashed line are subject to numerical resolution or heating effects. The DREAMS simulations do not produce the SAGA results in the top-right of the panel, where numerical effects should be under control. Right: The size-mass relation is shown for different values of the wind energy $\widebar{e}_w$, the dominant source of physical uncertainty in the model. The varied astrophysical and cosmological parameters in the DREAMS model are not sufficient to bring the results into correspondence with SAGA. For both the left and right panels, the mean size-mass relation is indicated by the colored lines, while the bands indicate the $1\sigma$ spread.
• Figure A1: A validation corner plot comparing the distributions of five key satellite properties from the original DREAMS simulations (blue) and those generated with the emulator (red; shown as 'DREAMS Varied' in the main text). The diagonal panels show the 1D probability distributions and the off-diagonal panels show the 2D joint distributions. The agreement between the simulations and the emulator demonstrates that NeHOD reproduces not only the distribution of individual properties, but also the complex correlations between them.