Unveiling the dark Universe with HI and EMBER-2

Abstract

Next-generation radio telescopes will provide unprecedented data volumes of the neutral hydrogen (HI) distribution across cosmic time. The spatial and kinematic distribution of HI is a biased tracer of the underlying matter field, and as such contains information on the distribution of dark matter over a wide range of scales. Extracting dark matter properties from HI, however, is non-trivial because baryonic processes linked to galaxy formation significantly modify the HI distribution. Additionally, methods that use empirical relations, often calibrated via numerical simulations, do not use the full field-level information to model the complex relation between HI and dark matter. We use the recently introduced EMBER-2 model to directly predict dark matter distributions from HI tracers over a wide redshift range, z=0-6. After training on cosmological galaxy formation simulations run with FIRE-2, our method accurately recovers key statistics, including dark matter mass fractions, surface density profiles and cross-correlations, where the latter are reconstructed at an accuracy of 20% down to scales of k = 100 h/cMpc constituting a significant improvement over traditional approaches. The presented method may become a key ingredient in future inference pipelines as it can be readily integrated into downstream analysis tasks of radio surveys.

Figures (8)

• Figure 1: Overview figure showing an example region of the simulated HI and DM fields in the FB30 simulation and the corresponding reconstructed field from EMBER-2 at $z=1$. The figure shows both normalized channels, surface density $\Sigma$ and line-of-sight velocity $v$, in a composite manner, where increased brightness indicates higher $\Sigma$, color-coded by the corresponding velocity of each pixel (from red to blue). For better visibility we show the normalized field values, which have $\mathcal{O}(1)$. This visual comparison highlights that the HI density and velocity fields contain significant amounts of information across different phase-space regimes, which can be used to accurately reconstruct the underlying DM distribution and kinematics from large down to small scale structures. The two insets (teal and orange) showcase the model's reconstruction power on small scales for two example regions.
• Figure 2: Structure mass functions $\phi$ for simulated and reconstructed dark matter maps at different redshifts. Lines and shaded bands indicate median and 16th to 84th percentiles for the test dataset. Across most redshift and structure-mass regimes, the reconstructed $\phi$ is in excellent agreement with simulations.
• Figure 3: Dark matter to HI ratios as a function of peak $N_{\rm HI}$ and redshift (bottom left in each panel). Vertical shaded regions indicate different systems, showing the IGM regime, LLSs, sub-DLAs (sDLAs) and DLAs. Simulated (reconstructed) results are shown as solid (dashed) lines indicating the median and the 16th to 84th percentiles. The reconstructed median ratios and their scatter are in excellent agreement with the simulated counterparts.
• Figure 4: Simulated (solid) and reconstructed (dashed) dark matter surface density profiles as a function of radius at $z=0$. Colors correspond to the same $N_{\rm HI}$ thresholds as in figure \ref{['fig:f_dm']}. Lines indicate median relations whereas shaded regions show the 16th to 84th percentiles for the emulation. The vertical gray line indicates the pixel resolution of the maps, below which, values are interpolated. The dashed line represents the ratio of total dark matter to HI mass outside of structures.
• Figure 5: Median errors on the cross-correlation coefficients between the HI and dark matter surface densities ($\varphi = \Sigma$), as well as for the kinetic energy density surface density ($\varphi = \pi$) for selected redshifts. Solid and dashed lines represent the results on the FB15 and the FB30 test set, while for FB15 we also show the 16th to 84th percentiles as shaded regions. The horizontal dashed lines indicate the 20% error band.