📝 SummarXiv

Abstract

The halo model, combined with halo occupation distribution (HOD) prescriptions, is widely used to interpret cosmic infrared background (CIB) anisotropies and connect star-forming galaxies to large-scale structure. Recent implementations adopt more physical parameterizations, but it remains unclear whether these models reliably recover astrophysical quantities. We test whether current CIB halo models can constrain the star formation efficiency, $\eta(M_h,z)$, and the halo mass where it peaks, $M_{\rm max}$, when fitted to mock data. We examine whether discrepancies arise from emission assumptions (the HOD ingredients) or from more fundamental components, such as bias and matter clustering. Using the M21 CIB HOD model within the halo framework, we fit mock CIB power spectra and star formation rate density (SFRD) data from the SIDES-Uchuu simulation, then repeat the analysis with a simplified simulation (SSU) matched to the HOD assumptions. Comparing best-fit parameters to known inputs, we find that although the M21 model fits the mock data well, it fails to recover intrinsic parameters, especially $M_{\rm max}$, even when applied to data generated with consistent assumptions. Emission-related quantities (SFRD, emissivity) agree within 5%, but the two-halo term shows a redshift- and scale-dependent offset exceeding 20%, likely due to the linear treatment of halo bias and matter clustering. Scatter in the SFR-halo mass relation and spectral energy distributions significantly impacts shot noise ($\sim 50\%$) but only modestly ($<10\%$) the clustered signal. We conclude that robust recovery of physical parameters from CIB clustering requires improved cosmological ingredients in halo models, including scale-dependent halo bias and nonlinear matter power spectra, alongside refined emission modeling.