Starlight from JWST: Implications for star formation and dark matter models

TL;DR

This work links the high-redshift UV luminosity function measured by JWST and HST to the underlying dark matter physics through an improved excursion-set halo framework with ellipsoidal collapse and a new window function. It finds strong evidence for a sharp transition from Pop-II/I to Pop-III stars around $z\sim10$ and a progressive reduction of feedback up to $z\sim25$, enabling substantial star formation in the early universe within CDM. When exploring non-CDM scenarios, the authors derive lower bounds of $m_{\rm FDM} > 5.6\times10^{-22}$ eV and $m_{\rm WDM} > 1.5$ keV (95% CL), and constrain white-noise power to $k_c > 25\,\mathrm{Mpc}^{-1}$, disfavoring axion miniclusters with $m_a < 6.6\times10^{-17}$ eV and PBHs with large $f_{\rm PBH}$; nonetheless, DM properties beyond CDM leave only modest additional imprint given current uncertainties. The results indicate that the observed UVLF is largely governed by star-formation physics—populations, feedback, and dust—rather than by modest departures from CDM within the probed parameter space. As JWST data improve, these bounds on DM and the high-$z$ star-formation history will tighten further, enabling sharper tests of early-universe growth and SMBH formation in non-CDM cosmologies.

Abstract

We confront the star formation rate in different dark matter (DM) models with UV luminosity data from JWST up to $z\simeq25$ and legacy data from HST. We find that a transition from a Salpeter population to top-heavy Pop-III stars is likely at $z\simeq10$ and that beyond $z=10-15$ the feedback from supernovae and active galactic nuclei is progressively reduced, so that at $z\simeq25$ the production of stars is almost free from any feedback. We compare fuzzy and warm DM models that suppress small-scale structures with the CDM paradigm, finding that the fuzzy DM mass $> 5.6 \times 10^{-22}{\rm eV}$ and the warm DM mass $> 1.5\, {\rm keV}$ at the 95% CL. The fits of the star formation rate parametrisation do not depend strongly on the DM properties within the allowed range. We find no preference over CDM for enhanced matter perturbations associated with axion miniclusters or primordial black holes. The scale of the enhancement of the power spectrum should be $> 25\,{\rm Mpc}^{-1}$ at the 95% CL, excluding axion miniclusters produced for $m_a < 6.6 \times 10^{-17}\,{\rm eV}$ or heavy primordial black holes that constitute a fraction $f_{\rm PBH} > \max[105 M_\odot/m_{\rm PBH}, 10^{-4} (m_{\rm PBH}/10^4 M_\odot)^{-0.09}]$ of DM.

Figures (10)

• Figure 1: Left panel: The unconditional first-crossing distribution for walks starting from $S = 0$, $\delta = 0$. The histogram shows the results of random walks, the solid black and dashed orange curves show the ansatz \ref{['eq:ST']} with $q=0.80$ and $q=0.707$, respectively, and the dotted grey curve shows the spherical collapse result. Other panels: The conditional first-crossing distribution at $z>z_0$ for walks starting from $S_0>0$ and $\delta_{\rm ell}(z_0,S_0) > 0$ as indicated in the plot. The histogram again shows the results of random walks, the solid black curves show the ansatz \ref{['eq:ST2']}, the dashed orange curves show the ansatz \ref{['eq:ST']} with $\delta_{\rm sp}(z) \to \delta_{\rm sp}(z)-\delta_{\rm sp}(z_0)$ and $S \to S-S_0$, and the dotted gray curve shows the spherical collapse result.
• Figure 2: The variance $S=\sigma^2$ of matter perturbations (upper panel) and the halo growth rate (lower panel) in the models considered in this work: CDM (dashed black line), FDM (green line), WDM (orange line) and an enhanced matter power spectrum (red line).
• Figure 3: The HMF at various redshifts $z$ in the FDM (left) and WDM (middle) models, and in models where white noise dominates the matter perturbations at $k>k_c$ (right) . The dashed contours show for comparison the HMF in the CDM model.
• Figure 4: The distributions of lensing magnifications at different source redshifts.
• Figure 5: The UV luminosity function at $z=4$ (black) and at $z=17$ (blue) with and without corrections due to lensing magnification and dust attenuation. The cyan and pink points show, respectively, the HST $z=4$ data from 2021AJ....162...47B and 2022ApJS..259...20H, and the brown and gray points the JWST $z=17$ data from Perez-Gonzalez:2025bqr and Castellano:2025vkt.