2025A&A...703A..98O

Recent observations reveal puzzling spatial offsets between the ALMA dust continuum from bright galaxies at z = 5-7 and their UV emission seen by HST and JWST (e.g. ALPINE and REBELS surveys). We investigated this using Cosmic Dawn ('CoDa') III, a state-of-the-art radiation-hydrodynamics simulation of fully coupled galaxy formation and re-ionisation, which includes a dynamical dust model. Dust-UV offsets are predicted for massive, UV-bright galaxies, up to ∼2 pkpc for the most massive haloes (M_DM > 10^11.5 M_⊙, M_* > 10¹⁰ M_⊙, and M_AB1600 < -21.5). Observed offsets are larger for ALPINE (4.4 < z < 5.9) than REBELS (z > 6.5), consistent with those in CoDa III, where higher-mass haloes appear at lower redshift. These offsets result primarily from severe dust extinction in galactic centres rather than from a misalignment in the dust- and stellar-mass distributions. At the CoDa III spatial resolution (1.65 pkpc at z = 6), dust and stellar components typically remain aligned. We predict that dust emission should therefore align well with stellar rest-frame Near Infra-Red, which is less affected by dust attenuation than the UV. This highlights the importance of dust in shaping galactic appearance. Our simulated dust masses for such bright galaxies agree with observational estimates, but to match the observed UV luminosity function, we reduced our dust attenuation. For SMC-like dust, the bright-end of the UV luminosity function at these redshifts is best fit, for example, with only 7.5% of the dust in CoDa III galaxies. Our massive galaxies thus reproduce the UV spectral slopes of ALPINE and REBELS galaxies, while preserving their dust-UV offsets, but exhibit dust masses lower than those suggested by observations.