The THESAN-ZOOM project: central starbursts and inside-out quenching govern galaxy sizes in the early Universe

TL;DR

The THESAN-ZOOM study tackles how galaxy sizes evolve at $3<z<13$ by leveraging high-resolution radiation-hydrodynamics simulations that resolve a multi-phase ISM and bursty star formation. The authors show that size growth is tied to cycles of central compaction during starbursts and inside-out quenching that expands the system, producing size oscillations around a positive intrinsic size–mass relation. They demonstrate that, when observational biases are accounted for, their simulated size–mass relations align with JWST results, addressing tensions with larger-volume simulations. They further reveal extended nebular emission driven by extreme LyC production, with a simple Strömgren-sphere framework explaining the observed trends, underscoring the necessity of realistic ISM physics and burstiness for modeling early galaxies.

Abstract

We explore the evolution of galaxy sizes at high redshift ($3 < z < 13$) using the high-resolution THESAN-ZOOM radiation-hydrodynamics simulations, focusing on the mass range of $10^6\,\mathrm{M}_{\odot} < \mathrm{M}_{\ast} < 10^{10}\,\mathrm{M}_{\odot}$. Our analysis reveals that galaxy size growth is tightly coupled to bursty star formation. Galaxies above the star-forming main sequence experience rapid central compaction during starbursts, followed by inside-out quenching and spatially extended star formation that leads to expansion, causing oscillatory behavior around the size-mass relation. Notably, we find a positive intrinsic size-mass relation at high redshift, consistent with observations but in tension with large-volume simulations. We attribute this discrepancy to the bursty star formation captured by our multi-phase interstellar medium framework, but missing from simulations using the effective equation-of-state approach with hydrodynamically decoupled feedback. We also find that the normalization of the size-mass relation follows a double power law as a function of redshift, with a break at $z\approx6$, because the majority of galaxies at $z > 6$ show rising star-formation histories, and therefore are in a compaction phase. We demonstrate that H$α$ emission is systematically extended relative to the UV continuum by a median factor of 1.7, consistent with recent JWST studies. However, in contrast to previous interpretations that link extended H$α$ sizes to inside-out growth, we find that Lyman-continuum (LyC) emission is spatially disconnected from H$α$. Instead, a simple Strömgren sphere argument reproduces observed trends, suggesting that extreme LyC production during central starbursts is the primary driver of extended nebular emission.

Figures (10)

• Figure 1: RGB images of 3 thesan-zoom galaxies at three redshifts in intrinsic H$\alpha$ (red), Lyman continuum (LyC, green), and UV (blue). We show m12.6 (left, subhalo 3), m12.6 (middle, subhalo 4), and 11.1 (right, subhalo 0). The white cross shows the UV center of light. The different tracers generally show remarkably distinct morphologies. H$\alpha$ traces ionised gas and therefore appears near LyC clumps, however a strong H$\alpha$ haze can also be identified around several galaxies. LyC emission tends to arise in compact star-forming regions scattered around the galaxy. The UV morphology is smoother than LyC, owing to the fact that stars from previous bursts can still contribute strongly to UV emission.
• Figure 2: The half-light and stellar half-mass radii of thesan-zoom galaxies as a function of stellar mass. We show stellar mass ($r^\ast_{1/2}$, upper left), H$\alpha$ ($r^{\mathrm{H}\alpha}_{1/2}$, upper right), UV ($r^\mathrm{UV}_{1/2}$, lower left), and optical ($r^\mathrm{opt}_{1/2}$, lower right) sizes. The log-scaled contour plots show the distribution of galaxies across $3<z<13$. Colored lines represent the median relation in redshift bins, where at least 10 galaxies are required for each point to be plotted. We see a positive size--mass relation for the stellar half-mass across the entire stellar mass and redshift range. This naturally leads to positive size--mass relations for $r^{\mathrm{H}\alpha}$, $r^\mathrm{UV}_{1/2}$, and $r^\mathrm{opt}_{1/2}$. This is in contrast to the negative size--mass relations from thesan-1Shen:2024aa, Illustris-TNG Genel:2018aa, FLARES Roper:2022aa, and Bluetides Marshall:2022aa, where concentrated dust must be invoked to recover a positive observed size--mass relation. We show the median $r^\ast_{1/2}$ for these simulations at $5\leq z\leq7$ in the upper left panel.
• Figure 3: The intrinsic 2D half-light radii of thesan galaxies in H$\alpha$ (left, $R^{\mathrm{H\alpha}}_{\mathrm{eff}}$), UV (center, $R^{\mathrm{UV}}_{\mathrm{eff}}$), and optical (right, $R^{\mathrm{opt}}_{\mathrm{eff}}$) emission as a function of stellar mass. Log-scaled contour plots show the distribution of galaxies at $5<z<7$. We also show our best fit line in the redshift bin $5<z<7$ (solid black) and our best fit after biasing for galaxies observable in typical JWST surveys (dash black, see text for details). Projection effects cause 2D half light sizes to be smaller and observations are unable to detect dim, extended galaxies. Accounting for these brings our biased best fit line within 0.1--0.2 dex of JWST observational results for galaxy H$\alpha$ (Danhaive et al. in prep.), optical Miller:2024aaAllen:2024aa, and UV sizes Morishita:2024aaAllen:2024aa.
• Figure 4: The stellar half-mass radii, $r^\ast_{1/2}$, of thesan-zoom galaxies as function of stellar mass. The green log-scaled contours show the distribution of galaxies. The line shows the size--mass evolution of an individual galaxy (m11.9, subhalo 0) and is coloured by its offset from the SFMS, $\Delta_{\mathrm{MS, 10}}$, where blue is above the SFMS and red is below. The dashed black lines show the mass of the galaxy at different redshifts. When the galaxy is heavily star-forming and lies above the SFMS, it goes through a phase of compaction where $r^\ast_{1/2}$ rapidly decreases due to strongly centrally concentrated star formation. Feedback and gas consumption cause the galaxy to quench inside out, leading to extended star formation and an increase in size in an expansion phase as the galaxy falls below the SFMS. At $\log_{10}(M_{\ast}\,[\mathrm{M}_\odot])\approx8.4$, the galaxy increases in size dramatically as it emerges from a quiescent period because the UV center-of-light is shifted to star formation which is proceeding away from the galactic center.
• Figure 5: The half-mass radius of stars formed in the last 10 Myr, $r^\mathrm{SFR}_{1/2}$, relative to the stellar half-mass radius, $r^\ast_{1/2}$, as a function of the offset from the SFMS, $\Delta_{\mathrm{MS, 10}}$. The log-scaled contours show the distribution of galaxies. The black points and errorbars show the medians and $16^\text{th}$--$84^\text{th}$ percentile range, respectively. The black dashed line shows $r^{\mathrm{SFR}}_{\mathrm{1/2}}/r^{\ast}_{\mathrm{1/2}}=1$, on which galaxies fall if their star formation occurs on the same scale as the existing stellar population. Galaxies above the SFMS tend to form stars in a central starburst, causing compaction of the galaxy size. These starbursts tend to quench inside-out, meaning that most galaxies below the SFMS are expanding their size. As a galaxy oscillates about the SFMS, it also oscillates around the size--mass relation.