The impact of cosmic ray feedback during the epoch of reionisation

TL;DR

This study investigates whether cosmic ray (CR) feedback shapes galaxy evolution and the epoch of reionisation (EoR) by comparing two non-zoomed, 10 Mpc^3 sphinx RMHD simulations: one with strong supernova (SN) feedback and another that reduces SN feedback while including CR transport and injection. Both models reproduce observed high-z UV luminosity functions and regulate star formation similarly, yet CR feedback lowers the escape fraction of Lyman continuum photons, delaying and partially inhibiting reionisation. The analysis shows that CRs make the interstellar and circumgalactic medium more gas-rich near star-forming regions, absorbing LyC photons within ~50 pc and reducing their contribution to the IGM, particularly in massive halos. The results highlight a tension between incorporating CR physics and achieving the observed reionisation timeline, underscoring the need for refined CR transport modeling and complementary feedback processes to accurately capture galaxy evolution and reionisation in the early universe.

Abstract

Galaxies form and evolve via a multitude of complex physics. In this work, we investigate the role of cosmic ray (CR) feedback in galaxy evolution and reionisation, by examining its impact on the escape of ionising radiation from galaxies. For this purpose, we present two Sphinx cosmological radiation-magneto-hydrodynamics simulations, allowing for the first time a study of the impact of CR feedback on thousands of resolved galaxies during the Epoch of Reionisation (EoR). The simulations differ in their feedback prescriptions: one adopts a calibrated strong supernova (SN) feedback, while the other simulation reduces the strength of SN feedback and includes CR feedback instead. We show that both comparably regulate star formation, reasonably match observations of high-redshift UV luminosity functions, and produce a similar amount of hydrogen ionising photons. In contrast to the model with strong SN feedback, the model with CRs lead to incomplete reionisation, which is in strong disagreement with observational estimates of the reionisation history. This is due to CR feedback shaping the ISM differently, filling with gas the low density cavities otherwise carved by SN explosions. As a result, this reduces the escape of ionising photons, at any halo mass, and primarily in the close vicinity of the stars. Our study indicates that CR feedback regulates galaxy growth during the EoR, but negatively affects reionisation, a tension which paves the way for further exploration and refinement of existing galaxy formation and feedback models. Such improvements are crucial in order to capture and understand the process of reionisation and the underlying evolution of galaxies through cosmic time.

Figures (13)

• Figure 1: Histogram of the number of $z=5$ dark matter halos per virial mass for sphinx-sn (in red) and sphinx-cr (in dark blue). The number of halos in each bin is written with the same colour code, and roughly corresponds to a total of 3600 resolved halos (i.e. with a DM mass higher than $7.5\times10^7\rm\,M_\odot$) for both simulations.
• Figure 2: Hydrogen gas column density (first rows) and mass-weighted fraction of ionised hydrogen (second rows) maps centred on the most massive halo at $z=5$ from sphinx-sn (top) and sphinx-cr (bottom). From left to right, maps are 500, 100 and $10\,\rm kpc$ wide, and a 50, 10 and $1\,\rm kpc$ width scale bar is respectively plotted in the lower left corner of each panel.
• Figure 3: From top to bottom: Time evolution of the SFRD, total stellar mass, and sSFR in sphinx-sn (red) and sphinx-cr (dark blue) runs. The SFR is calculated from all stellar particles formed in the 10 cMpc boxes, averaged over 10 Myr.
• Figure 4: SMHM relation (upper panel) and 10-Myr averaged SFR (lower panel) versus halo mass in sphinx-sn (red) and sphinx-cr (dark blue) at $z=5$. The curves respectively show the averaged stellar mass and SFR per bin of halo virial mass, and the coloured shaded regions represent the standard deviation. We also show observational constraints from Read2017 at $0.2\leq z\leq0.4$ with black crosses, from Behroozi2019 at $z=5$ with a dark grey shaded area and from Stefanon2021 at $z=5.6$ with a light grey shaded region. At any halo mass, the stellar mass is roughly the same in the two simulations, but tends to be higher than the observational constraints. At the massive end, galaxies are more massive in sphinx-cr and have higher SFRs.
• Figure 5: Dust-attenuated UV luminosity function in sphinx-sn (red) and sphinx-cr (dark blue), with Poissonian error-bars. From the top left to the bottom right panel, we show increasing redshift from 5 to 10. The references of the observations shown in each panel are written in the legend. At any time, the UV luminosity functions of the two simulations are very similar and consistent with the observational constraints.