Exploring the interplay between star formation efficiency and dust in regulating the UV luminosity of early systems in the JWST and ALMA era

TL;DR

This paper develops an analytic model that jointly links early galaxy star-formation efficiency and dust enrichment to their UV visibility during the JWST/ALMA era. By starting from the halo mass function and incorporating bursty star formation, SNII-driven dust production, and ISM dust processing (growth, reverse-shock destruction, ejection, and sputtering), the authors reproduce the observed UV luminosity function across z ~ 5–20 and predict how dust modulates UV observables. They find an evolving star-formation efficiency f_*(z) ≈ 10^{0.13 z − 3.5} and a dust-radius scaling α(z) ≈ ((1+z)/11.37)^{2.46}, with dust attenuation diminishing at higher redshift due to more extended dust distributions and stronger reverse-shock effects; at z ≳ 9, matching the UV LF suggests galaxies form stars at ≈10× the fiducial efficiency. The work further predicts the dust-to-stellar mass relation, the UV-to-total-SFR connection, and the dust mass function, concluding that ALMA-detected dusty galaxies at z ~ 5–7 are not representative of the average high-z population and offering a framework to interpret upcoming multiwavelength data.

Abstract

Recent observations by the James Webb Telescope (JWST) have unveiled numerous galaxy candidates between $z \sim 9 - 16.5$, hinting at an over-abundance of sources at the bright-end of the UV Luminosity Function (UV LF) at z $\gtrsim$ 11. Complementarily, the Atacama Large Millimetre Array (ALMA) has started yielding dust mass estimates at $z \sim 5 - 7$. In this work, we develop an analytic formalism baselined against the latest ALMA results, jointly exploring the impact of bursty star formation and its associated dust enrichment, on the visibility of early galaxies, while also modelling sources scattered off the main sequence of star formation. We incorporate dust production in type II Supernovae, dust destruction, ejection, growth and sputtering. Our key results are: (i) explaining the UV LF at $z \sim 5 - 13$ requires an average star formation efficiency that evolves as $f_*(z) = 10^{0.13z-3.5}$, with a number of observations exceeding this main sequence by a factor of 10. (ii) The dust enrichment of early systems is driven by dust production in SNII ejecta, while growth and sputtering impact the dust mass by 60\% and 40\% respectively at $z \sim 7$. (iii) galaxies at $z \gtrsim 9$ can retain significant dust, reaching average dust-to-stellar mass ratios of 0.19\% (0.14\%) at $z \sim 9$ ($z \sim 11$). Dust attenuation decreases with redshift as dust becomes more dispersed within halos. (iv) observations by ALMA at $z \sim 5$ and 7 are not representative of the average population that makes up the UV LF.

Figures (8)

• Figure 1: The evolving UV LF at $z \sim 5 - 16$. In each panel, lines show theoretical results for intrinsic and dust-attenuated UV LFs for the fiducial model as well as those allowing for a scatter of $0.5$ and 1 dex on $f_*$, as marked in panel a; the dark and light shaded areas show the corresponding $1 \sigma$ scatter for the $f_*\pm0.5$ and $f_*\pm 1$ dex cases, respectively. In every panel, the solid green and yellow lines show the theoretical upper limits i.e. the "maximal intrinsic UV LF" for a Salpeter and flat-ish IMF, respectively. Finally, points show observational data, as marked, from atek2015, bowler2017, Atek2018, ishigaki2018, bowler2020, bouwens2021, Bouwens_2022b, bouwens2022jwst, harikane2022a, donnan2022, Willot2023, Adams2023a, Casey2023a, Finkelstein2023a, Bouwens2023b, Harikane2024b, Harikane2023b, Harikane2024b, Leung2023, Robertson2023a, PerezGonzalez2023, Harikane2024, donnan2024jwst and Mcleod2024, Whitler2025, as marked in the panels.
• Figure 2: The redshift evolution of the UV luminosity density ($\rho_{\rm UV}$). As marked, the short dashed and solid lines show the intrinsic and dust-attenuated values of $\rho_{\rm UV}$ in the fiducial model; the dot-dashed and dotted lines show models allowing for 0.5 and 1 dex of scatter on $f_*$ respectively. The solid green and yellow lines show the theoretical upper limits in the "maximal" model for a Salpeter and flat-ish IMF, respectively. All of the theoretical models have been integrated down to galaxies with $\rm M_{UV} \hbox{$\; \buildrel < \over \sim \;$} -17$ to be able to compare to observations. As marked, points show observational results from mcleod2016, Bouwens2023b, donnan2022, Willot2023 and donnan2024jwst.
• Figure 3: The dust mass as a function of the stellar mass at $z \sim 5-11$, as marked, colour-coded by the observed specific SFR, $\psi_{\rm UV}/M_*$. The solid (black) line shows the results of the fiducial model. The other lines show this relation successively accounting for the processes of dust production, reverse shocks, dust destruction, dust ejection and grain growth in the ISM. Filled circles are obtained by scattering the star formation efficiency ($f_*$) by $\pm$ 1 dex. Filled squares show observational data from the ALPINE survey fudamoto2020 at $z \sim 5$ and from the REBELS survey bouwens2022 at $z \sim 7$, in panels a and b, respectively. The blue dashed lines and shaded areas represent the fiducial model in Mauerhofer2025 at the respective $z$, while orange dashed lines and shaded areas correspond to the same model, adopting an IMF evolving with $z$.
• Figure 4: The SFR inferred from the UV as a function of the total intrinsic SFR for $z \sim 5-11$, as marked, colour-coded by the host halo mass. In each panel, the solid (black) line shows results from the fiducial model; filled circles show results obtained by scattering the star formation efficiency ($f_*$) by $\pm$ 1 dex. Square points show observational results from the ALPINE survey fudamoto2020 at $z \sim 5$ and from the REBELS survey bouwens2022 at $z \sim 7$.
• Figure 5: The SFR inferred from the UV as a function of the total intrinsic SFR for $z \sim 11$, colour-coded by the host halo mass. In each panel, the solid (black) line shows results from the fiducial model; filled circles show results obtained by scattering the star formation efficiency ($f_*$) by $\pm$ 1 dex. The teal data points correspond to JADES-GS-z14-0 and JADES-GS-z14-1 Carniani2024, as shown in the legend.