Renaissance Simulations and Early Galaxy Evolution
- Renaissance Simulations are state-of-the-art adaptive-mesh refinement models that simulate early universe physics and first galaxy formation using high-res hydrodynamics and radiation transport.
- They integrate detailed treatments of Population III star formation, metal enrichment, and feedback to analyze reionization processes and UV luminosity functions.
- The simulations combine extreme resolution with varied cosmic environments, offering actionable insights for calibrating semi-analytic models and interpreting JWST observations.
The Renaissance Simulations are a suite of extremely high-resolution and physics-rich radiation transport hydrodynamics cosmological adaptive-mesh refinement simulations of high redshift galaxy formation, designed to follow the first generations of stars and galaxies during the epoch of reionization. Performed with the Enzo AMR code on the Blue Waters supercomputer, the suite has been used to study ultraviolet luminosity functions, Population III star formation, ionizing escape fractions, synthetic spectra, black-hole-remnant growth, and the interpretation of JWST observations at (O'Shea et al., 2015, Xu et al., 2016, Xu et al., 2016, Barrow et al., 2017, McCaffrey et al., 2023).
1. Numerical architecture and simulated environments
The suite consists of three zoom-in regions selected from a parent cosmological box of (comoving), sampling an overdense, a near mean-density, and an underdense environment. These are commonly denoted Rarepeak, Normal, and Void, and were evolved to different final redshifts in order to follow structure formation across different environments (Xu et al., 2016, Hazlett et al., 2024).
| Region | Comoving volume | Final redshift |
|---|---|---|
| Rarepeak | $15$ | |
| Normal | $12.5$ | |
| Void | $8$ |
Initial conditions are set at with WMAP7 parameters , 0, 1, 2, 3, and 4. In the refined regions the simulations reach an effective 5 equivalent resolution, a dark matter particle mass of 6, and maximum AMR resolution of about 7 comoving pc, with up to 8 levels of refinement (Xu et al., 2016, Xu et al., 2016).
At their final redshifts, each region yields a sample of more than 9 halos in the mass range 0. Across the full suite, the simulations follow the feedback from 1 to 2 metal-free Population III stars in each simulation, which makes the suite simultaneously a statistical sample of early galaxies and a resolved laboratory for first-star feedback (Xu et al., 2016).
2. Star formation, chemistry, and radiative feedback
The Renaissance Simulations self-consistently model 3D gravitational dynamics, adaptive mesh refinement, hydrodynamics, primordial and metal-line cooling, nonequilibrium chemistry for 3 species, radiative transport, and local stellar feedback. The chemistry and cooling treatment includes metal-dependent cooling, while star formation prescriptions distinguish between metal-free Population III stars and metal-enriched star formation using density and metallicity criteria (Xu et al., 2016).
Population III stars form in gas with metallicity 4 and high density. Their IMF is described as Kroupa-like but shifted, with characteristic mass 5; a related formulation used in the synthetic-spectra work is
6
Metal-enriched stars and Population III stars inject energy, metals, and radiation, and the simulations follow stars until end-of-life or remnant formation (Xu et al., 2016, Barrow et al., 2017).
Radiative transfer is treated with adaptive ray tracing, including ionizing UV and Lyman-Werner radiation. The full-suite reionization analysis uses the Moray adaptive ray-tracing module and includes local and background LW radiation, ionization, and photoheating of the IGM, while the late-Population-III analysis in the Void region explicitly used only local LW radiation rather than a uniform metagalactic LW background within the refined region. Large-scale LW attenuation is modeled with the “picket-fence” approximation (Xu et al., 2016, Xu et al., 2016).
3. Galaxy populations, luminosity functions, and reionization
The first major results from the suite showed that the ultraviolet luminosity function of simulated galaxies is consistent with observations at the bright end, 7, but becomes essentially flat at lower luminosities rather than rising steeply as a Schechter extrapolation would suggest. The simulations predict a flattening at 8 and a lower limit around 9 (O'Shea et al., 2015).
This behavior was traced to two coupled relations: a declining stellar fraction with decreasing halo mass, and a halo occupation fraction that falls rapidly below the atomic-cooling scale. In the Renaissance analysis these were written as
$15$0
and
$15$1
with $15$2 and $15$3. Halos with virial masses below $15$4 do not universally contain stars, and the fraction of halos containing stars drops to zero at $15$5 (O'Shea et al., 2015).
A complementary statistical analysis across the full suite found that galaxy formation and ionizing escape are primarily halo-mass driven. The occupation fraction of halos with recent star formation drops from unity above $15$6 to $15$7 at $15$8 and $15$9 at 0, quite independent of redshift and region. Key galaxy properties, including stellar mass, star formation rate, gas fraction, and UV magnitude, are largely determined by halo mass rather than environment or redshift up to 1; the most massive galaxies reach 2 by 3 (Xu et al., 2016).
For reionization, the central quantity is the escape fraction of ionizing photons,
4
The suite finds median 5 in halos near 6, an average 7 in the range 8, and 9 in halos with virial masses $12.5$0. Small halos therefore contribute strongly to the earliest escaped ionizing photon budget, while larger galaxies dominate later as feedback suppresses star formation in the smallest systems (Xu et al., 2016).
4. Late Population III star formation and the metal–LW interplay
One of the most distinctive Renaissance results is the demonstration that Population III star formation continues to $12.5$1 in the Void region. In a survey volume of $12.5$2 comoving $12.5$3, $12.5$4 Pop III galaxies with recent star formation were identified at $12.5$5, compared to roughly $12.5$6 metal-enriched systems in the same volume. Their halo virial masses span $12.5$7, with some systems hosting more than $12.5$8 active Pop III stars and active Pop III stellar mass exceeding $12.5$9 in a single galaxy (Xu et al., 2016).
The inferred number density in that survey is
0
The physical explanation is a two-factor combination: metal enrichment is local and slow, while Lyman-Werner radiation is strong and widespread. At 1, only 2 of the volume and 3 of the gas mass are enriched to 4, but LW intensity is 5 everywhere and exceeds 6 over more than 7 of the volume. Strong LW fields suppress 8 cooling and raise the Pop III formation threshold to halos of mass 9, while incomplete external enrichment leaves pristine gas available even at late times (Xu et al., 2016).
The same study examined analogues of the luminous Ly$8$0 emitter CR7, discovered by Sobral et al. 2015. A partial analogue was found in Halo 528, with $8$1, forming a single Pop III star and lying $8$2 kpc from a metal-enriched galaxy with $8$3. However, the Pop III cluster in the simulation is far less massive and luminous than CR7 is inferred to be, and rapid local enrichment terminates further pristine star formation. The simulated result therefore constrains, rather than broadly supports, a massive Pop III cluster interpretation for CR7 (Xu et al., 2016).
5. Synthetic observations and relation to JWST data
Renaissance outputs have been repeatedly post-processed into synthetic observables. A detailed spectral study used a pipeline combining FSPS, Hyperion, Cloudy, and custom tools to generate spectra and IR images for $8$4 high-redshift galaxies. The emergent spectra exhibit strong halo-to-halo variability, especially at low stellar mass, with line ratios and equivalent widths depending on stellar mass, metallicity, gas mass fraction, and formation history. Halos with higher masses show more consistent spectra, and their [OIII] equivalent widths lie between $8$5 and $8$6. The same analysis found that viewing angle alone can produce a three-fold difference in flux because ionized gas channels create anisotropic escape paths, and the simulated galaxies occupy the star-forming locus in BPT space, well below the typical AGN region (Barrow et al., 2017).
A later photometric study used Powderday with FSPS, Cloudy, Hyperion, JWST NIRCam filters, WebbPSF, and $8$7 viewing angles to generate mock observations and size measurements for $8$8 galaxies. In that catalog, stellar masses range from $8$9 to 0, while star formation rates span 1 to 2. The majority of half-light radii lie between 3 and 4 pc, the majority of half stellar mass radii are around 5 kpc, and Sérsic indices vary between 6 and 7. The simulated galaxies are generally bluer than the JWST sample, and the photometric analysis notes that the synthetic photometry does not include IGM attenuation blueward of Ly8 at observed redshifts (Hardin et al., 28 May 2026).
These mock-observation programs have been used directly in the interpretation of JADES and CEERS. A 2023 comparison argued that the most massive Renaissance galaxies have stellar masses and star formation rates entirely consistent with spectroscopically confirmed JWST galaxies at 9, after extrapolating simulated growth histories with measured specific star formation rates. In that analysis, the suite showed excellent agreement with JADES and CEERS and found no tension between the 0CDM model and current JWST measurements (McCaffrey et al., 2023).
6. Black-hole remnants and high-energy backgrounds
Renaissance has also been used to study compact remnants from Population III stars. A post-processing analysis followed the growth of more than 1 black holes seeded by Pop III remnants in halos from 2 to 3. The result was overwhelmingly inefficient accretion: on average the black holes increase their initial mass by a factor 4, the most active objects increase their mass by approximately 5, and only a single black hole experiences any super-Eddington episode, which is brief and not repeated. No correlation was found between black-hole accretion and host halo mass in the sampled range, and the analysis attributes the low growth rates to the fact that star formation consumes dense gas and stellar feedback suppresses accretion (Smith et al., 2018).
A separate high-energy application used the Pop III star formation rates and spatial distributions in Renaissance to calculate the X-ray luminosity density and background from Pop III binaries over 6. In the preferred method, the X-ray luminosity density rises from 7 at 8 to 9 at 00. The integrated output corresponds to approximately 01 eV of X-ray energy per hydrogen atom, sufficient to heat the IGM to 02 K and ionize a few percent of the neutral hydrogen. The resulting keV background was found to accumulate into a spectrum extending to roughly 03 eV, with implications for the 21-cm signal, the Ly04 forest, and the CMB Thomson optical depth (Xu et al., 2016).
7. Semi-analytic calibration and later methodological uses
Because the suite couples high physical fidelity to moderate cosmological volume, it has become a calibration set for semi-analytic models of the first stars and galaxies. A 2024 framework calibrated a semi-analytic model to Renaissance star formation histories from the Normal region and constructed a four-parameter prescription for metal-enriched star formation with an initial bursty stage followed by a steady stage. Between 05, the calibrated model reproduces metal-enriched star formation rate densities within a factor of 06 for the average-density region, and the total metal-enriched stellar mass differs from Renaissance by about 07 at 08. The same study reports larger deviations at later times in overdense or rarefied environments, which it attributes to environmental dependencies not included in the model (Hazlett et al., 2024).
An extended 2025 semi-analytic model calibrated jointly to Aeos and Renaissance used Renaissance primarily to calibrate metal-enriched Population II star formation and feedback in the resolved-halo regime. That work found that agreement with the hydrodynamical calibrators requires increasing the normalization of, flattening the redshift dependence of, and adding scatter to the commonly used critical mass threshold 09, and it emphasized the importance of modeling the delay between Pop III stellar death and later Pop II star formation (Hazlett et al., 13 Oct 2025).
Renaissance has also been used as training data for machine-learning classification of rare black-hole-seed environments. In an SVM-based search for direct-collapse black-hole candidates, the best-performing model relied on quantities tied to star formation and chemical state, especially metallicity, incident flux of Lyman-Werner radiation, and halo stellar mass. That use of the suite is methodologically distinct from the original hydrodynamics calculations, but it remains anchored in the same resolved chemical, radiative, and halo-scale information that motivated the simulations in the first place (Pries et al., 11 Nov 2025).