MEGATRON: Disentangling Physical Processes and Observational Bias in the Multi-Phase ISM of High-Redshift Galaxies

TL;DR

The paper addresses how to interpret JWST-like rest-frame UV–optical spectra of high-redshift galaxies when the ISM is dense, multi-phase, and influenced by non-equilibrium thermochemistry. It introduces MEGATRON, a suite of high-resolution cosmological radiation-hydrodynamics simulations with on-the-fly chemistry, plus self-consistent synthetic observations that mimic observational pipelines. The study finds that high-z ISM is denser, less metal-enriched, and has higher ionization parameters than local analogs; line diagnostics are strongly affected by the density-temperature structure and by the chosen sub-grid physics, with O32 acting as a density tracer at high densities and C43/N43 providing more robust ionization-parameter diagnostics in some regimes. These results show that emission-line ratios can constrain the underlying star-formation and feedback physics in the early universe and guide the interpretation of JWST data across large samples.

Abstract

Now detected out to redshifts of $z\sim 14.5$, the rest-frame ultraviolet and optical spectra of galaxies encode numerous physical properties of the interstellar medium (ISM). Accurately extracting these properties from spectra remains a key challenge that numerical simulations are uniquely suited to address. We present a study of the observed ISM of galaxies in MEGATRON: a suite of cosmological radiation hydrodynamics simulations coupled to on-the-fly non-equilibrium thermochemistry, with multiple prescriptions for star formation/feedback and parsec-scale resolution; capable of directly predicting spectroscopic properties of early galaxies. We find that irrespective of feedback physics used, the ISM of high-redshift galaxies is denser, less metal enriched, and subject to higher ionization parameters and radiation fields compared to similar mass galaxies in the local Universe -- in agreement with interpretations of JWST observations. Using common observational techniques to infer bulk galaxy properties, we find that ISM gas density controls the slope of the mass-metallicity relation. Similarly, at the densities reached in some high-redshift galaxies, O32 becomes a density tracer rather than one of ionization parameter. This motivates the use of other line ratios like C43 and N43 to infer the ionization state of the gas. Finally, various feedback models populate different regions of strong-line diagnostic diagrams as the line ratios are sensitive to the feedback-modulated density-temperature structure of the ISM. Therefore, observed strong-line diagnostics can provide a strong constraint on the underlying physics of star formation and feedback in the high-redshift Universe.

Figures (14)

• Figure 1: Projected UV-optical emission line maps normalized by the total $\rm H\beta$ flux for the galaxy with the largest specific star-formation rate in the bursty star formation simulation. Low surface-brightness gas has been brightened. On the bottom, we show the total intrinsic spectrum (in arbitrary $\log\ f_{\nu}$ units), highlighting the relative strengths of each line. This galaxy has a stellar mass of $10^{7.9}M_{\odot}$ at $z=8.93$.
• Figure 2: Temperature-density phase diagrams for an example $\sim10^{8.5}M_{\odot}$ stellar mass galaxy at $z=8.6$ in the efficient star formation run, with all cells colored by their contribution to the total luminosity of a given emission line. We include UV, optical, and FIR lines in blue, green, and red respectively. No two emission lines are produced by the same gas, introducing potential biases when emission lines are combined to make measurements of ISM properties.
• Figure 3: Distributions of common UV, optical, and FIR emission line ratios for each MEGATRON simulation, highlighting the impacts that different subgrid prescriptions for star-formation/feedback have on the ISM, and therefore on its spectroscopic properties.
• Figure 4: Smoothed BPASS spectral energy distributions for a $1~{\rm Myr}$, $10^6\ {\rm M_{\odot}}$ stellar cluster for a variety of metallicities, normalized at 1500Å Eldridge:2017Stanway:2018. This illustrates the fact that lower-metallicity stellar populations show characteristically harder and more dominant ionizing spectra. These differences are important for higher-ionization species (e.g. $\rm N~\small IV$, $\rm C~\small IV$, $\rm O~\small III$, $\rm N~\small III$, $\rm C~\small III$, etc.) which are often observed in the high-redshift Universe. In the top panel, we show the ionization energies for species of elements considered in this work.
• Figure 5: Mass-metallicity relations for each of the high-redshift MEGATRON simulations. Unless specified as intrinsic, coloured points and lines represent masses and metallicities inferred using observational techniques (i.e. the direct method for metallicities and BAGPIPES SED fitting for stellar masses). Left: Observed gas-phase metallicity as a function of observed stellar mass, coloured by the $\rm H\alpha$-derived star formation rate for all galaxies in MEGATRON with M$_{\rm UV}\leq-15$. We show comparisons with observational data of galaxies with direct $\rm [O~{\small III}]~\lambda 4363$ measurements above redshift $z = 5$Curti:2020Curti:2023Arellano-Cordova:2022Arellano-Cordova:2025Cameron:2023cMorishita:2024Pollock:2025, running means (colored lines), and contours showing the distribution of intrinsic values (i.e. mass-weighed metallicity and true stellar mass) for each simulation (pale blue). Observational measurements of these quantities tend to imprint a slight bias on the MZR, by flattening the slope of the relation. Right: We show comparisons between the running mean of each simulation with the same observational data. We also indicate the impact of key physical processes, highlighting intrinsic biases (e.g. different feedback models) in gray and observational biases (e.g. temperature/density fluctuations) in blue.