---
title: 'THESAN-ZOOM: High-Res Cosmic Simulation'
url: https://www.emergentmind.com/topics/thesan-zoom-simulations
type: topic
---

# THESAN-ZOOM: High-Res Cosmic Simulation

THESAN-ZOOM is a suite of high-resolution cosmological zoom-in radiation-hydrodynamic simulations developed as the galaxy-scale extension of the broader THESAN reionization program. Its purpose is to preserve the cosmological and radiation-field context of THESAN while resolving the multiphase interstellar medium, local stellar feedback, and circumgalactic structure of high-redshift galaxies at a level inaccessible to the uniform large-volume runs. Across the project, THESAN-ZOOM is used to study star-formation variability, galaxy sizes, star-formation efficiency, metallicity structure, ionizing-photon escape, the impact of external reionization, and the origin of metal absorbers during the epoch of reionization and cosmic dawn [2503.00106, 2503.04894, 2510.13977].

## 1. Project architecture and relation to THESAN

THESAN-ZOOM extends the original THESAN simulations into a regime where the multiphase ISM, local stellar feedback, and short-timescale star-formation variability are explicitly resolved. The parent THESAN simulations are large-volume radiation-hydrodynamic calculations of the epoch of reionization, whereas THESAN-ZOOM resimulates selected regions at much higher resolution while keeping those regions embedded in the parent radiation field rather than in a uniform ultraviolet background [2503.00106, 2510.13977].

The project contains 14 zoom regions selected from the parent dark-matter-only run thesan-dark-1. In one project description, 14 haloes were selected at \(z=3\), spanning environments that evolve into haloes of mass \(10^8\)–\(10^{13}\,M_\odot\); for each target, all dark matter particles within \(4R_{\rm vir}\) were traced back to \(z=127\) and the corresponding Lagrangian region was resampled [2510.13977]. A later project application states that the full THESAN-ZOOM project contains 14 zoom regions and 60 simulations across three resolution tiers, while another analysis of galaxy sizes uses all subhaloes found in the zoom regions over \(3<z<13\) [2606.14671, 2503.04894].

The suite is not a uniform cosmological volume. It is a set of targeted environments embedded in the radiation field of the parent THESAN simulation. This design is central to the project’s scientific logic: it combines halo-by-halo environmental specificity with a self-consistent large-scale reionization context, allowing the zooms to retain patchy external radiation histories and not merely isolated internal galaxy evolution [2510.13977, 2503.02927].

## 2. Numerical framework and baryonic physics

THESAN-ZOOM is run with AREPO-RT, the radiation-hydrodynamic extension of the moving-mesh code AREPO. Gas dynamics are solved on a quasi-Lagrangian Voronoi mesh, and radiative transfer is evolved on the fly with a moment-based scheme. One project description specifies a reduced speed of light \(\tilde c = 0.01\,c\) and 16 RT subcycles per hydro step, while another emphasizes the same on-the-fly RHD framework with non-equilibrium thermochemistry and local radiative feedback [2510.13977, 2503.04894].

A defining distinction from the large-volume THESAN box runs is the use of a resolved multiphase ISM model. Several THESAN-ZOOM papers describe the baryonic model as based on a tailored version of SMUGGLE rather than an effective equation of state. Star formation occurs only in dense, self-gravitating, Jeans-unstable gas, with a threshold \(n_{\rm H}>10\,{\rm cm^{-3}}\). In the fiducial GMC and SFE studies, the local law is written as
\[
\dot{\rho}_\star = \epsilon^{\rm cl}_{\rm ff}\frac{\rho_{\rm gas}}{t_{\rm ff}},
\qquad
t_{\rm ff}=\sqrt{\frac{3\pi}{32G\rho}},
\]
with \(\epsilon^{\rm cl}_{\rm ff}=1\) in the fiducial runs [2505.05554, 2503.01949].

The stellar feedback model is explicitly multi-channel. Across the project papers it includes photoionization, radiation pressure, photoelectric heating, stellar winds, supernovae, and an additional early stellar-feedback or early momentum-injection component during the first \(5\) Myr. Stellar spectra are taken from BPASS binary stellar population models. Cooling includes primordial non-equilibrium processes, molecular hydrogen, dust-gas interactions, and metal-line cooling, and the simulations track elemental abundances including C, N, O, Mg, Ne, Si, and Fe [2510.13977, 2505.05554, 2503.02927].

The suite uses three resolution tiers. Reported baryonic mass resolutions are \(9.09\times10^3\), \(1.14\times10^3\), and \(1.42\times10^2\,M_\odot\), corresponding to 4x, 8x, and 16x zooms in some papers, or L4, L8, and L16 in others. Different analyses use different tiers: the OI-absorber study uses the standard-resolution 4x zooms; the external-reionization study adopts the 8x runs as its main set; the GMC analysis uses 8x as fiducial and 4x/16x for robustness tests; and several morphology and SFMS papers use the highest-resolution version available for each target [2510.13977, 2503.02927, 2505.05554, 2503.04894].

## 3. External radiation, patchy reionization, and environmental embedding

A key feature of THESAN-ZOOM is that the local radiation field is not treated in isolation. Each zoom region receives radiation from stars inside the zoom plus an external contribution inherited from the parent THESAN run, so the simulations remain embedded in a self-consistent large-scale reionization background rather than a uniform UVB [2510.13977]. Other THESAN-ZOOM papers describe the same mechanism as time-varying radiation maps from the parent THESAN simulation imposed as boundary conditions, allowing galaxies to feel a realistic, patchy external radiation environment [2503.00106, 2503.04894].

This design becomes scientifically decisive in the dedicated study of external reionization. That work compares three treatments of the external ionizing field: the standard THESAN-1 boundary-condition setup, a uniform UV background switched on at \(z=10.6\), and a late uniform UV background switched on at \(z=5.5\). It finds that external UV radiation efficiently unbinds gas in haloes with masses below \(10^9\,M_\odot\) and suppresses subsequent star formation, but also that the patchy, self-consistent radiative-transfer treatment permits more realistic shielding and a cold but low-density gas phase down to \(z=3\) [2503.02927].

The same study argues that a patchy reionization history matters not only for instantaneous gas loss but for subsequent chemical evolution. In simulations with early reionization, minihaloes fail to form stars from pristine gas, which reduces the metal enrichment of gas later accreted by more massive haloes. Consequently, haloes with masses below \(10^{10}\,M_\odot\) at all simulated epochs \(z>3\) exhibit lower metallicities and altered metallicity distributions. The authors conclude that, at minimum, a semi-numerical model that incorporates spatially inhomogeneous reionization and a non-uniform metallicity floor is necessary to accurately emulate metal enrichment in minihaloes [2503.02927].

A plausible implication is that THESAN-ZOOM’s environmental realism is not merely a boundary-condition refinement. Within the project, it functions as a mechanism for preserving the causal link between local ISM evolution and large-scale ionization topology, especially in low-mass systems for which external radiation can regulate both baryon retention and enrichment history [2503.02927, 2510.13977].

## 4. Galaxy-scale predictions: star formation, sizes, efficiency, and metal mixing

A major set of THESAN-ZOOM results concerns the star-forming main sequence and the time variability of star formation. One analysis finds that the high-redshift SFMS normalization scales as \(\propto (1+z)^{2.64\pm0.03}\) for the 10 Myr-averaged intrinsic sequence, with a very weak stellar-mass dependence \(\propto M_\ast^{0.041\pm0.004}\). The same work argues that observationally inferred flatter evolution can be reproduced by filtering out weakly star-forming “lulling” galaxies, implying that current observational fits may be biased by missing low-SFR systems or overestimated star-formation rates. It further distinguishes two starburst modes: an externally driven mode powered by rapid large-scale inflows and an internally driven mode linked to cyclical ejection and re-accretion of the ISM in low-mass galaxies [2503.00106].

THESAN-ZOOM also predicts strongly burst-regulated structural evolution. The galaxy-size analysis reports that galaxies above the star-forming main sequence undergo rapid central compaction during starbursts, followed by inside-out quenching and spatially extended star formation that drives expansion. This produces oscillatory evolution around the size–mass relation and yields a positive intrinsic size–mass relation at high redshift, in contrast to the negative intrinsic trends found in several large-volume simulations using effective equation-of-state models. The same study finds that H\(\alpha\) emission is systematically extended relative to the UV continuum by a median factor of \(1.7\), while LyC emission is spatially disconnected from H\(\alpha\), and argues that a simple Strömgren-sphere model captures the median H\(\alpha\)-size behavior [2503.04894].

On halo and galactic scales, the project’s SFE papers present a consistent feedback-regulated picture. The halo-scale star-formation efficiency \(\epsilon^{\ast}_{\rm halo}\) follows a double power-law dependence on halo mass, with a slope roughly \(1/3\) at large halo masses and roughly \(2/3\) at lower masses. The paper interprets these limits as broadly consistent with momentum-driven and energy-driven outflow scenarios, respectively, and states that \(\epsilon^{\ast}_{\rm halo}\) is a factor of \(2\)–\(3\) larger than commonly assumed in empirical galaxy-formation models at \(M_{\rm halo}\lesssim 10^{11}\,M_\odot\). On kpc scales, the neutral-gas Kennicutt–Schmidt relation is approximately universal and follows \(\Sigma_{\rm SFR}\propto \Sigma_{\rm gas}^2\) [2503.01949].

The GMC analysis extends this hierarchy to cloud scales. It finds that the galaxy-scale average efficiency scales as \(M_{\rm halo}^{1/3}(1+z)^{1/2}\sim V_{\rm vir}\), while GMCs themselves show nearly universal properties across host mass and redshift. The cloud mass function follows \(\mathrm{d}N/\mathrm{d}M\propto M^\gamma\) with \(\gamma=-2.49^{+0.13}_{-0.07}\); the size–linewidth relation has \(b\simeq 0.6\); the median cloud surface density is \(\Sigma_{\rm GMC}\approx 70\,M_\odot\,{\rm pc}^{-2}\); and the median instantaneous cloud-scale efficiency is \(\epsilon_{\rm ff}^{\rm GMC}\sim 2\%\)–\(3\%\). The paper’s central conclusion is that the dominant driver of global variation is the GMC mass fraction in the ISM, \(f_{\rm GMC}\), not large changes in the intrinsic efficiency of individual clouds [2505.05554].

THESAN-ZOOM is also used externally as one of the high-redshift “bursty feedback” suites in a cross-simulation study of gas-phase metallicity gradients. In that comparison, Thesan Zoom, FIRE-2, and SPICE Bursty produce systematically flatter gradients than Thesan Box and SPICE Smooth, with the median gradient for Thesan Zoom reported as \(-0.021^{0.008}_{-0.072}\,{\rm dex/kpc}\), compared with \(-0.205_{-0.373}^{-0.124}\,{\rm dex/kpc}\) for Thesan Box. The study interprets THESAN-ZOOM’s flatter gradients as the consequence of bursty feedback driving turbulence, outflows, and radial metal mixing [2510.26877].

## 5. Circumgalactic gas, absorber populations, and ionizing escape

THESAN-ZOOM is designed not only for internal galaxy structure but also for CGM and near-IGM applications. The OI-absorber study uses the simulations to connect neutral oxygen absorbers at \(z=5\)–8 to galaxies and environments. It finds that the circumgalactic medium becomes progressively ionized with cosmic time, such that for haloes \(M_h\ge 10^9\,M_\odot\) the median neutral-oxygen covering fraction declines by about \(0.2\) from \(z=8\) to \(z=5\), while the total oxygen covering fraction remains roughly constant. The paper argues that this contrast shows that the changing OI signal is driven mainly by ionization-state evolution rather than by loss of oxygen from halo gas [2510.13977].

A second major result of that analysis is geometric rather than chemical. Observable OI absorbers with \(N_{\rm OI}\gtrsim 10^{13}\,{\rm cm^{-2}}\) are not confined to haloes. At \(z\ge 5\) the majority, \(\gtrsim 60\%\), arise beyond \(R_{\rm vir}\), and by \(z=8\) the paper emphasizes that the median position of roughly 90% of absorbers is outside the CGM. The absorber environments are overdensities rich in faint galaxies with \(M_\star\le 10^8\,M_\odot\), which explains the scarcity of directly detected counterparts while still allowing occasional nearby brighter star-forming galaxies similar to those suggested by ALMA and JWST observations [2510.13977].

A later THESAN-ZOOM application uses more than 35,000 galaxy realizations from the m12.6 zoom region over \(3<z<16\) to identify indirect diagnostics of LyC escape. In that work, random-forest regressors trained on simulation-derived indicators find that the strongest diagnostics of \(f_{\rm esc}\) are \(\mathrm{SFR}_{10}/\mathrm{SFR}_{100}\) and \(M_{\rm gas}/M_\ast\), whereas \(M_{\rm UV}\) dominates prediction of the escaping ionizing photon rate \(\dot N_{\rm ion,esc}\). The paper interprets this as evidence that ionizing-photon escape is strongly related to burst-driven gas clearing, while the ionizing emissivity that matters for reionization is more tightly tied to UV luminosity [2606.14671].

That same analysis uses the learned \(\dot N_{\rm ion,esc}\)–\(M_{\rm UV}\) relations together with observed UV luminosity functions to construct reionization histories consistent with observational constraints. It argues that the bulk of reionization occurred rapidly after \(z\approx 8\) and that galaxies with \(M_{\rm UV}<-17\) provide the dominant contribution, with the \(M_{\rm UV}\le -20\) population supplying more than 50% of the ionizing photon budget by \(z\sim 6\) in the THESAN-ZOOM-based model [2606.14671].

## 6. Comparative context, limitations, and methodological boundaries

Within the broader THESAN program, THESAN-ZOOM is consistently presented as complementary to the uniform large-volume simulations rather than as a replacement for them. The large THESAN box provides representative reionization topology, mass functions, and radiation-field context, while THESAN-ZOOM resolves the cold and clumpy gas, bursty star formation, and local feedback physics that the effective equation-of-state THESAN runs cannot capture. Several papers explicitly frame the zooms as a way to distinguish changes in metal mass distribution from changes in ionization state, or to recover galaxy properties that depend on multiphase ISM structure and short-timescale variability [2503.00106, 2510.13977].

The project’s limitations are likewise recurrent. The OI-absorber analysis does not evolve oxygen ionization states in non-equilibrium on the fly; instead it uses the approximation \(n_{\rm OI}/n_{\rm O}\approx n_{\rm HI}/n_{\rm H}\), which the authors note may overestimate OI in cold dense gas and neglect higher oxygen stages at very high temperatures [2510.13977]. The external-reionization study is confined to seven low-mass zoom targets and does not include baryon–dark matter streaming velocities, an X-ray background, or a dedicated Pop III stellar model [2503.02927]. The SFE study does not include black holes or AGN feedback, which may become important above \(M_{\rm halo}\gtrsim 10^{12}\,M_\odot\) [2503.01949]. The GMC paper notes that the snapshot cadence of \(\sim 10\) Myr is too coarse to measure integrated GMC efficiencies directly, making any inference about lifetime cloud efficiencies indirect [2505.05554].

A further methodological boundary is statistical. THESAN-ZOOM yields very large numbers of galaxy realizations when many snapshots and subhaloes are combined, but it is still a zoom suite rather than a representative cosmological volume. The OI study explicitly notes that incidence calculations use an external halo mass function because the zooms are not suited to measuring the halo mass function directly [2510.13977]. The 2026 LyC-escape study exploits one very large zoom region, m12.6, and therefore derives its machine-learning mappings from a single environment at the lowest resolution tier, even though the resulting catalog contains tens of thousands of galaxy realizations [2606.14671].

Taken together, these constraints define the project’s scientific niche. THESAN-ZOOM is most powerful when the question depends on resolved multiphase gas, bursty stellar feedback, local radiative transfer, or the detailed structure of the CGM and near-IGM, but still requires embedding in a realistic, patchy reionization environment. Its published results repeatedly show that these ingredients alter predictions for the SFMS, galaxy sizes, star-formation efficiency, metallicity gradients, LyC escape, external-reionization imprints, and metal-absorber origin in ways that differ qualitatively from both uniform-UVB treatments and large-volume effective-equation-of-state simulations [2503.00106, 2503.04894, 2503.01949, 2503.02927, 2510.13977].

Source: https://www.emergentmind.com/topics/thesan-zoom-simulations