The DREAMS Project: Disentangling the Impact of Halo-to-Halo Variance and Baryonic Feedback on Milky Way Dark Matter Density Profiles

Abstract

Astrophysical searches for dark matter in the Milky Way require a reliable model for its density distribution, which in turn depends on the influence of baryonic feedback on the Galaxy. In this work, we utilize a new suite of Milky Way-mass halos from the DREAMS Project, simulated with Cold Dark Matter (CDM),to quantify the influence of baryon feedback and intrinsic halo-to-halo variance on dark matter density profiles. Our suite of 1024 halos varies over supernova and black hole feedback parameters from the IllustrisTNG model, as well as variations in two cosmological parameters. We find that Milky Way-mass dark matter density profiles in the IllustrisTNG model are largely insensitive to astrophysics and cosmology variations, with the dominant source of scatter instead arising from halo-to-halo variance. However, most of the (comparatively minor) feedback-driven variations come from the changes to supernova prescriptions. By comparing to dark matter-only simulations, we find that the strongest supernova wind energies are so effective at preventing galaxy formation that the halos are nearly entirely collisionless dark matter. Finally, regardless of physics variation, all the DREAMS halos are roughly consistent with a halo contracting adiabatically from the presence of baryons, unlike models that have bursty stellar feedback. This work represents a step toward assessing the robustness of Milky Way dark matter profiles, with direct implications for dark matter searches where systematic uncertainty in the density profile remains a major challenge.

Figures (14)

• Figure 1: Milky Way Mass Dark Matter Halos from the SB5 DREAMS CDM Suite. Projections of the dark matter density of a small fraction ($3\%$) of DREAMS halos. Each row contains variations of the five DREAMS parameters ($\bar{e}_w$, $\kappa_w$, $\epsilon_{f,\,\mathrm{high}}$, $\Omega_{\rm M}$, and $\sigma_8$, top-to-bottom, respectively; see Table \ref{['tab:dreams_parameters']} and Section \ref{['subsec:astro_variations']}). Each column represents the minimum, $20^{\rm th}$ percentile, $40^{\rm th}$ percentile, $60^{\rm th}$ percentile, $80^{\rm th}$ percentile, and maximum variation for each parameter, noting that in each row the other DREAMS parameters are also co-varied. The halos simulated here are all generated using different initial conditions (see Rose_2025Rose_2025). Thus, these simulations capture: (i) the impact of uncertainty in cosmology and astrophysics, (ii) the uncertainty in estimation of Milky Way halo mass, and (iii) intrinsic halo-to-halo variation.
• Figure 2: Dark Matter Density Profiles in the DREAMS CDM Milky Way-Mass Suite. The large left-hand panel shows the dark matter density as a function of radius for all halos in the DREAMS CDM Milky Way-mass suite from $\sim1.2$ kpc to $600$ kpc (normalized by $R_{200}$). The red line and shaded region corresponds to the weighted mean and plus/minus standard deviation of the individual profiles (see Section \ref{['subsec:weighting']} and CDM_Centrals_2025 for details on the weighting procedure). The values in the bottom left of the large panel correspond to the weighted standard deviation of densities at $0.01R_{200}$ ($\sim2~{\rm kpc}$), $0.1R_{200}$ ($\sim20$ kpc), and $R_{200}$ ($\sim200$ kpc). The panels on the right show how the density depends on the upper (dotted orange) and lower (dashed blue) $10\%$ range for each of the parameters (compared to the average $10\%$ of the parameter ranges, $\rho_{\rm avg}$): $\bar{e}_w$ (top left), $\kappa_w$ (top middle), $\epsilon_{f,\,\mathrm{high}}$ (top right), $\Omega_{\rm M}$ (bottom left), and $\sigma_8$ (bottom middle). The colored line in each panel is the median density profile while the colored shaded region represents the $25^{\rm th}$ and $75^{\rm th}$ percentile. Overall, the profiles are self similar, with a modest amount of scatter. Moreover, the baryonic parameters have the strongest effect on the central densities.
• Figure 3: Dependence of generalized NFW scale density ($\rho_s$) on Astrophysics, Cosmology, and Halo Mass. Predictions from an ensemble of emulators (see Section \ref{['subsec:emulator']}) for the scale density, $\rho_s$, of the DREAMS CDM Milky Way-mass density profiles. We show the predictions made by the emulator as a function of $\bar{e}_w$ (top left), $\kappa_w$ (top middle), $\epsilon_{f,\,\mathrm{high}}$ (top right), $\Omega_{\rm M}$ (bottom left), $\sigma_8$ (bottom middle), and $M_{\rm halo}$ (bottom right). The shaded regions represent the one standard deviation uncertainty of the predictions based on the individual model prediction uncertainty as well as variance across the emulators (via Equation \ref{['eqn:emulator_uncertainty']}). The vertical red solid line at the top of each panel corresponds to the fiducial TNG value (see Table \ref{['tab:dreams_parameters']}). In every panel, the non-varied parameters are fixed to their fiducial values and $M_{\rm Halo} = 10^{12}~{\rm M}_\odot$ unless explicitly varied. While not shown, the emulator also predicts the scale radius. The trends for the scale radius are identical to those shown here, just in the opposite direction (i.e., increasing $\kappa_w$ increases $r_s$). We find that only $\kappa_w$ drives variation in $\rho_s$ more significant than the halo-to-halo variation.
• Figure 4: Dependence of generalized NFW shape parameters on $\mathbf{\bar{e}_w}$. Predictions from our neural network emulator for the dependence of the inner slope ($\gamma$; dot-dashed line), transition rate ($\alpha$; dashed line), and outer slope ($\beta$) of the best-fit gNFW profiles on the supernova wind energy ($\bar{e}_w$). As a point of reference, the dotted gray lines show the canonical NFW profile values ($\alpha=\gamma=1$ and $\beta=3$). The short vertical solid line at the top corresponds to the fiducial TNG value of $\bar{e}_w=3.6$ (see Table \ref{['tab:dreams_parameters']}; Section \ref{['subsec:astro_variations']}). The average prediction lines are the mean of our ten emulators while the shaded regions approximate the halo-to-halo variation (although there is also a small contribution from the combination of emulators, see Section \ref{['subsec:emulator']} for details). We note that, within the uncertainty due to halo-to-halo variation, the shape parameters have no strong dependence on any of the other DREAMS simulation parameters or halo mass.
• Figure 5: Central Mass Growth ($\mathbf\Gamma$) of Halos at $\mathbf{0.01R_{200}}$. Predictions from our emulators for the central mass growth, defined as the ratio of the dark matter mass in the hydro simulations to that of the DMO simulations (see Equation \ref{['eqn:mass_growth']}) at a radius of $0.01 R_{200}$, as a function of the three DREAMS baryon feedback parameters. The solid line represents the average prediction of the parameter from the ensemble of emulators for $M_{\rm Halo} = 10^{12}~{\rm M}_\odot$, while the shaded region is the uncertainty on the predictions as a proxy for halo-to-halo variance (via Equation \ref{['eqn:emulator_uncertainty']}). The dashed gray line represents a mass ratio of unity, where the halo mass is unchanged from the hydro to the DMO simulation. The short solid lines at the top of each panel correspond to the fiducial TNG value (see Table \ref{['tab:dreams_parameters']}). We find that the supernova feedback parameters decrease the central mass growth, whereas the AGN feedback parameter increases the central mass growth. The dashed and dotted lines show the average for halo masses of $10^{11.8}$ and $10^{12.2}~{\rm M}_\odot$, respectively (corresponding bands not shown).