2023MNRAS.525.5932B

The feedback loop between the galaxies producing the background radiation field for reionization and their growth is crucial, particularly for low-mass haloes. Despite this, the vast majority of galaxy formation studies employ a spatially uniform, time-varying reionizing background, with the majority of reionization studies employing galaxy formation models only required to work at high redshift. This paper uses the well-studied TNG galaxy formation model, calibrated at low redshift, coupled to the AREPO-RT code, to self-consistently solve the coupled problems of galaxy evolution and reionization, evaluating the impact of patchy (and slow) reionization on early galaxies. THESAN-HR is an extension of the THESAN project to higher resolution (a factor of 50 increase, with a baryonic mass of m_b ≈ 10⁴ M_⊙), to additionally enable the study of 'mini-haloes' with virial temperatures T_vir < 10⁴ K. Comparing the self-consistent model to a uniform UV background, we show that galaxies in THESAN-HR are predicted to be larger in physical extent (by a factor ∼2), less metal enriched (by ∼0.2 dex), and less abundant (by a factor ∼10 at M₁₅₀₀ = - 10) by z = 5. We show that differences in star formation and enrichment patterns lead to significantly different predictions for star formation in low mass haloes, low-metallicity star formation, and even the occupation fraction of haloes. We posit that cosmological galaxy formation simulations aiming to study early galaxy formation (z ≳ 3) must employ a spatially inhomogeneous UV background to accurately reproduce galaxy properties.