The THESAN-ZOOM project: Mystery N/O more -- uncovering the origin of peculiar chemical abundances and a not-so-fundamental metallicity relation at $3<z<12$

TL;DR

THESAN-ZOOM investigates metal production, retention, and mixing in the early universe ($3<z<12$) using high-resolution zoom simulations that resolve ISM physics and track nine elements. It finds smoothly curved, slowly evolving MZRs for gas and stars, with metal retention and gas fractions jointly shaping evolution, and skewed MZR scatter toward higher redshift. A key result is the inversion of the gas-phase FMR in low-mass galaxies due to dilution from pristine inflows, while the stellar FMR remains intact; bursty star formation combined with differential winds naturally produces order-of-magnitude N/O excursions and nitrogen-rich episodes without exotic yields. Nitrogen-rich galaxies and nitrogen-rich GMCs emerge as robust predictions, reconciling high-redshift JWST observations with standard nucleosynthesis channels and highlighting the importance of burst-driven, spatially and temporally resolved enrichment for interpreting nebular abundances in the early universe.

Abstract

We present an analysis of metallicities and chemical abundances at $3<z<12$ in the THESAN-ZOOM simulations. We find that smoothly curved gas-phase and stellar mass-metallicity relations (MZR) are already in place at $z\approx12$ and evolve slowly ($\sim$0.2 dex increase for gas, $\sim$0.4 dex increase for stars at a fixed stellar mass) down to $z=3$, governed largely by the efficiency with which galaxies retain their metals, rather than gas fraction. The canonical fundamental metallicity relation (FMR) survives in stars but breaks down and inverts for gas in low-mass galaxies ($M_\ast\lesssim10^{9}\mathrm{M_\odot}$) due to regular dilution by low-metallicity gas inflow. We find broad agreement of gas-phase N/O, Fe/O, and C/O with high-redshift observations, including the presence of nitrogen-rich galaxies (NRGs; $\log(\mathrm{N/O})>-0.6$) without the need for exotic yields in our chemical network. Instead, bursty star formation naturally generates order-of-magnitude excursions in N/O on $\lesssim$100 Myr timescales due to temporally differential galactic winds; after a starburst, stellar feedback expels gas, leaving a large population of asymptotic-giant-branch stars to dominate the enrichment of the relatively low-mass interstellar medium. NRGs lie below the main sequence and typically exhibit $\mathrm{EW}[H$β$]\lesssim40$ Å, in apparent tension with observed high-EW NRGs. This tension is reconciled if observed NRGs are in the initial stages of a subsequent starburst, illuminating previously enriched gas, which is supported by the finding of high SFR surface density nitrogen-rich giant molecular clouds.

Figures (19)

• Figure 1: Maps of the central subhalo of m12.6 at $z=6$, with $M_\ast=10^{9.3}\,\mathrm{M_\odot}$. The large panel shows gas surface density and the smaller panels show, clockwise from the upper left, stellar mass surface density, gas-phase metallicity, iron-to-oxygen, carbon-to-oxygen, and nitrogen-to-oxygen. The metallicity and abundance ratios are weighted by gas mass. All maps show the same spatial scales. Considerable variations are seen spatially in the abundance ratios, and in particular the nitrogen-to-oxygen ratio, and such variations are not obviously correlated with the metallicity.
• Figure 2: The mass-metallicity relation across redshift, where we show both the gas-phase oxygen abundance (left panel) and the stellar iron abundance (right panel). The black histograms show log-scaled distribution of all thesan-zoom galaxies included in this work ($3<z<12$). Curved, coloured lines show our best fits to a redshift-dependent MZR (Eq. (\ref{['eq:mzr_eq']})) at $z=12$, $z=3$, and extrapolated down to $z=0$. We compare to observational results at $z=0$ for stellar metallicities of local dwarf galaxies Kirby:2013aa and at $z\approx3.5$ from the NIRVANDELS survey Cullen:2021aaStanton:2024aa. We also show estimates for the MW and GSE at $z=3$ as black and blue points, respectively Monty:2025aa. For the gas-phase, we compare with strong line-derived metallicities of SDSS galaxies Curti:2020aa and electron temperature method-derived metallicities Yates:2020aa. We also show higher-redshift results from MOSDEF at $z=3.3$Sanders:2021aa, JWST at $3<z<10$Curti:2024ab, and JWST at $4<z<10$Nakajima:2023aa. Our gas-phase MZR extrapolated to $z=0$ is within $\sim$0.05 dex of the Curti:2020aa MZR at $M_\ast\approx10^{11}~\mathrm{M_\odot}$ and within $\sim$$0.1-0.2$ dex of the Yates:2020aa MZR at $M_\ast\lesssim10^{10}~\mathrm{M_\odot}$. Although we are extrapolating and there is clear tension between different observational MZR fits at $z=0$, this relatively good agreement may suggest a consistent and shallow redshift evolution of the MZR from $z=0$ to $z=12$. In terms of the high-redshift observations, our MZR is in excellent agreement with Sanders:2021aa at $z\approx3.3$. The Curti:2024ab and Nakajima:2023aa MZR fits have extremely shallow slopes, cutting across our relations. This indicates either that there is a dramatic evolution in the MZR from $z=3$ which we do not capture in our simulation, or biases in the observations due to, for example, selection effects.
• Figure 3: The mass-metallicity relation across redshift, where we show both the gas-phase total metallicity (left panel) and the stellar total metallicity (right panel). The black histograms show log-scaled distribution of all thesan-zoom galaxies included in this work. Curved, coloured lines show our best fits to a redshift-dependent MZR at $z=12$, $z=6$, and $z=3$. The MZR from FIRE simulations Ma:2016aa are shown at $z=3$ and $z=6$ as dashed lines, and MZR from FIRE-2 simulations are shown as dotted lines at $z=6$ and $z=12$Marszewski:2024aa. The total metallicity-based gas-phase MZR has a steeper redshift dependence compared to the oxygen abundance-based fit, whereas the total metallicity-based gas-phase MZR has a shallower redshift dependence than the iron abundance-based fit. This is due to the characteristic time delay in enrichment for each element. Oxygen-enrichment occurs almost immediately via CC SNe, leading to a flatter redshift evolution, whereas iron is enriched on $\sim$Gyr timescales by SNe Ia, causing a steeper redshift evolution. Even for the total metallicity-based MZR fits, the gas-phase still has a somewhat steeper redshift evolution. Our gas-phase MZR shows a weaker redshift evolution compared to FIRE between $z=3$ and $z=6$, although the stellar MZR evolution is more comparable. The redshift evolution of the FIRE-2 gas-phase MZR is similarly weak to thesan-zoom. Both FIRE and FIRE-2 find a shallower low-mass slope than thesan-zoom, but they do not fit a curved mass-dependence, which makes it difficult to compare directly.
• Figure 4: The metal retention efficiency in gas, $y_\mathrm{gas} \equiv M_{Z,\text{gas}} / M_\ast$ (left), and stars, $y_\mathrm{\ast} \equiv M_{Z,\mathrm{\ast}} / M_\ast$ (right) as a function of redshift for several stellar mass bins. The gas-phase metal retention efficiency remains remarkably constant with stellar mass. $y_\mathrm{gas}$ tends to increase with cosmic time down to $z\approx5$, where it remains approximately constant down to $z=3$. The stellar metal retention efficiency, $y_\ast$, decreases with redshift and increases strongly with stellar mass. In the mass and redshift range we consider here, the stellar retention efficiency is small enough that it does not make up a significant fraction of the metals compared to the gas, however, we expect it to be a more important component in more massive and lower redshift galaxies.
• Figure 5: The gas mass to stellar mass ratio, $M_\mathrm{gas}/M_\ast$, as a function of redshift for several stellar mass bins. $M_\mathrm{gas}/M_\ast$ increases with cosmic time down to $z\approx5$, and decreases rapidly down to $z=3$. The behavior of $M_\mathrm{gas}/M_\ast$ at $z\gtrsim5$ is at odds with the generally expected trend of more gas-rich galaxies at higher redshifts. This is because thesan-zoom galaxies are more bursty at higher redshift and therefore there is a larger population of mini-quenched galaxies with very little gas, dragging down the median. $M_\mathrm{gas}/M_\ast$ shows a strong stellar mass dependence, where more massive galaxies have lower $M_\mathrm{gas}/M_\ast$ because they have converted more of their gas reservoir into stars.