Dust-UV offsets in high-redshift galaxies in the Cosmic Dawn III simulation

TL;DR

This study addresses the origin of spatial offsets between dust continuum and UV emission in high-redshift galaxies by leveraging the Cosmic Dawn III simulation, a large-volume, fully coupled radiation-hydrodynamics run that includes a dynamical dust model. After calibrating the dust masses with a factor $K_{\rm dust}=0.075$ to match the observed UV luminosity function and UV slopes, the authors generate attenuated UV maps and dust-density maps to study UV–dust offsets as a function of halo mass, UV magnitude, and stellar mass. They find that, in massive halos ($M_{\rm DM} \gtrsim 10^{11.5} M_\odot$), offsets up to ~2 pkpc arise primarily from severe dust extinction in central regions, not from a misalignment between dust and the stellar distribution; dust remains largely aligned with the bulk stellar/NIR component. The results reproduce key observational trends from ALPINE/REBELS at $z\sim5$, while predicting smaller offsets for fainter galaxies and highlighting the need for higher spatial resolution and improved dust physics in future simulations. Overall, the work underscores dust’s critical role in shaping the appearance of early galaxies and provides a framework to interpret UV–dust morphologies in the Epoch of Reionization.

Abstract

Recent observations have revealed puzzling spatial disparities between ALMA dust continuum and UV emission as seen by HST and JWST in galaxies at $z=5-7$ (e.g. ALPINE and REBELS surveys), compelling us to propose a physical interpretation of such offsets. We investigate these offsets using the Cosmic Dawn III (CoDa III) simulation, a state-of-the-art fully coupled radiation-hydrodynamics cosmological simulation, which incorporates a dynamical dust model. First of all, we find that our simulated dust masses, while calibrated to match observed ones, yield unrealistically large UV attenuations. In fact, the bright-end galaxy UV Luminosity function is best reproduced using only 7.5\% of the dust content of CoDa III galaxies. With this recalibration, we obtain populations of massive galaxies matching ALPINE and REBELS magnitudes and UV slopes, but with smaller dust masses than observed. In this framework, we also find significant dust-UV offsets in massive, UV-bright galaxies ($\mathrm{M}_\mathrm{DM}> 10^{11.5}$ M$_\odot$, M$_*>10^{10}$ M$_\odot$, M$_{\rm AB1600}<-21.5$), reaching up to $\sim 2$ pkpc for the most massive systems. Our analysis reveals that these offsets primarily result from severe dust extinction in galactic centers rather than a misalignment between dust and stellar mass distributions. At the spatial resolution of CoDa III (1.65 pkpc at z=6), the dust remains in majority well-aligned with the bulk stellar component, and we predict the dust continuum should therefore align well with the stellar rest-frame NIR component, less affected by dust attenuation. This study highlights the importance of dust in shaping the appearance of early galaxies at UV wavelengths, even as early as in the Epoch of Reionization.

Figures (9)

• Figure 1: Physical properties of a representative galaxy at z=7 with M$_{\rm DM} = 6.1 \times 10^{11}$ M$_{\odot}$. The red circle in each panel indicates R$_{\rm FoF}$. Top row: Projected quantities integrated along the line of sight. Bottom row: 1-cell thick slice at the position of maximum density. (a) H column density, (b) Dust column density, (c) Stellar mass surface density, (d) Intrinsic (unattenuated) absolute UV magnitude at 1600 Å $$, (e) H number density, (f) Temperature, (g) HI fraction, (h) Metallicity (absolute scale).
• Figure 2: Bright end UV luminosity function of CoDa III galaxies for redshifts z=5,6,7 and various dust contents. The shaded areas represent the average dust-attenuated LFs plus or minus the Poisson noise for K$_{\rm dust}$=0.075. The dotted lines represent the unattenuated LFs. The dashed lines represent the full dust content LFs (i.e. K$_{\rm dust}$=1.0). The horizontal dot-dashed line represents 1/V limit where V is the volume of the simulated domain in Mpc$^3$. The observed LFs are from bouwens21LF.
• Figure 3: M$_{\rm AB1500}$-$\beta$ distributions of CoDaIII galaxies (orange: dust-attenuated, green: unattenuated), along with the ALPINE and REBELS samples (respectively red and blue symbols with error bars). The gray-shaded colors indicate how many galaxies (dust-attenuated) reside in each cell, as a visual help. Note that nebular continuum emission is not accounted for, and can redden the spectra by up to $\Delta \beta=+0.3$wilkins2016.
• Figure 4: Dust masses in the Cosmic Dawn III simulation at z=5 as a function of stellar mass, along with the ALPINE and REBELS samples. Green and orange colors correspond to 100% and 7.5% of the simulated galaxies' dust masses, respectively. Our fiducial model for the post-processing of this study is the 7.5% case.
• Figure 5: From left to right: restframe-UV absolute M$_{\mathrm{AB1600}}$ magnitudes, attenuated and intrinsic (unattenuated), dust column density and projected stellar density maps of a sample of 4 Cosmic Dawn III galaxies with dark matter halo masses M$_{\mathrm{DM}}$=$5\times10^{11}-10^{12}$ M$_{\odot}$ at z=5-7. In the left panel, for each galaxy, symbols mark the positions of maximum attenuated UV flux (green 'x'), maximum NIR flux, using as proxy the maximum of the stellar projected density (green '+'), and maximum dust column density (green 'Y'). For the 2 leftmost plot columns, the color indicates the absolute magnitude (attenuated or intrinsic) of each cell. The dust column density and projected stellar density are given in log$_{10}$(M$_{\odot}$/pkpc$^2$).