JADES: Insights on the low-mass end of the mass--metallicity--star-formation rate relation at $3 < z < 10$ from deep JWST/NIRSpec spectroscopy

TL;DR

This study leverages deep JWST/NIRSpec spectroscopy from the JADES program, combined with higher-mass literature samples, to map the mass–metallicity relation (MZR) and probe the fundamental metallicity relation (FMR) at 3 < z < 10 for 146 galaxies spanning log(M*/M⊙) ≈ 6.5–9.5. Using Te-based calibrations and multiple strong-line diagnostics, the authors find a flattening of the MZR at the low-mass end with a slope of β ≈ 0.17, and a mild evolution in normalization up to z ~ 6, with potential further flattening at z > 6. The data show a persistent deviation from the local FMR, especially at high redshift, suggesting enhanced gas accretion dilution and/or stronger metal outflows in the early Universe, challenging the universality of the FMR. Comparisons with simulations indicate momentum-driven winds best reproduce the observed low-mass MZR slope, while analytic models highlight the role of evolving gas fractions; overall, the results imply evolving feedback and gas-flow physics in dwarf galaxies during early cosmic times.

Abstract

We analyse the gas-phase metallicity properties of a sample of low stellar mass (log M*/M_sun <= 9) galaxies at 3 < z < 10, observed with JWST/NIRSpec as part of the JADES programme in its deep GOODS-S tier. By combining this sample with more massive galaxies at similar redshifts from other programmes, we study the scaling relations between stellar mass, oxygen abundance (O/H), and star-formation rate (SFR) for 146 galaxies, spanning across three orders of magnitude in stellar mass and out to the epoch of early galaxy assembly. We find evidence for a shallower slope at the low-mass-end of the mass-metallicity relation (MZR), with 12 + log(O/H) = (7.72+-0.02) + (0.17+-0.03) log(M* / 10^8 M_sun), in good agreement with the MZR probed by local analogues of high-redshift systems like 'Green Pea' and 'Blueberry' galaxies. The inferred slope is well matched by models including 'momentum-driven' SNe winds, suggesting that feedback mechanisms in dwarf galaxies (and at high-z) might be different from those in place at higher masses. The evolution in the normalisation is observed to be relatively mild compared to previous determinations of the MZR at z~3 (~ 0.1 - 0.2 dex across the explored mass regime). We observe a deviation from the local fundamental metallicity relation (FMR) for our sample at high redshift, especially at z > 6, with galaxies significantly less enriched (with a median offset in log(O/H) of ~ 0.5 dex, significant at ~ 5 sigma) than predicted given their M* and SFR. These observations are consistent with an enhanced stochasticity in the star-formation history, and/or with an increased efficiency in metal removals by outflows, prompting us to reconsider the nature of the relationship between M*, O/H, and SFR in the early Universe.

Figures (14)

• Figure 1: Top panel: Example fit to the PRISM (R100) spectrum (in black) of the $z=4.805$ galaxy JADES-GS+53.16268-27.80237, zoomed-in on the region of the main rest-frame optical emission lines. The best-fit from ppxf is shown in red. Bottom panels: Medium resolution grating (R1000) spectrum (and its best-fit) for the same galaxy. From left to right, the panels show a zoom-in on the regions of the [O ii] and [Ne iii], H$\beta$ and [O iii], H$\alpha$ [N ii] and [S ii] emission lines, respectively. In all panels, the error spectrum is marked by the cyan shaded region.
• Figure 2: Comparison between the metallicity derived from medium resolution gratings and PRISM spectra, for objects in which both configurations satisfy the requirements described in Section \ref{['sec:metallicity']}. The two distributions scatter across the equality line (in red), with a median offset of $0.01$ dex and a standard deviation of $0.15$ dex. In $75\%$ of the cases the two measurements are consistent within their $1\sigma$ uncertainties.
• Figure 3: The redshift distribution of the selected sample of 62 JADES galaxies with metallicity measurements analysed in this paper, compared to the distribution of the total 'JADES-Deep' sample presented in Section \ref{['sub:observations']} and bunker_hst_deep_DR_2023 with spectroscopic redshift determination.
• Figure 4: The distribution in the stellar mass versus star-formation rate plane for our combined JWST sample, including JADES galaxies presented in this work and the literature sample compiled from nakajima_mzr_ceers_2023 (CEERS), curti_smacs_2023 (EROs), and bunker_gnz11_2023 (GN-z11), respectively. The top and right-hand inset panels show the histograms of the distribution in both parameters. The SFR for JADES galaxies is derived from beagle fitting to both PRISM spectra and NIRCAM photometry (where available, and to PRISM spectra only otherwise). For CEERS galaxies, the SFRs are compiled from nakajima_mzr_ceers_2023 as inferred from the H$\beta$ luminosity and the kennicutt_star_2012 calibration, but have been here scaled down by $0.23$ dex to account for the mean offset between the H$\alpha$(H$\beta$)-based and beagle-based SFRs measured for the JADES sample (we refer to Section \ref{['sec:mass_sfr']} for more details). For both ERO galaxies and GN-z11 the M$_{\star}$ and SFR have been derived via beagle fitting, consistently with what done for JADES sources. The parametrisation (and its extrapolation at low-mass)
• Figure 5: The mass-metallicity relation (MZR) for our full JWST sample. Filled circles represent individual galaxies presented in this work from JADES, whereas filled crosses are galaxies observed in the framework of the CEERS programme and compiled from nakajima_mzr_ceers_2023, with metallicity derived as detailed in Section \ref{['sec:metallicity']}. The star symbol reports the JADES/NIRSpec observations of GN-z11 (bunker_gnz11_2023), whereas 'X' symbols mark galaxies from the EROs as compiled from curti_smacs_2023 and laseter_auroral_jades_2023. Large squared and diamond symbols mark the median values computed in bins of M$_{\star}$ (full sample, in purple), and (M$_{\star}$ z), as described in Table \ref{['tab:binned_values']}. An orthogonal linear regression fit to the median values in bins of M$_{\star}$ for the different redshift sub-samples is shown by the purple (full sample), yellow ($z=3-6$) and blue ($z=6-10$) lines, respectively. We include a comparison with previous determinations of the MZR at lower redshifts from curti_massmetallicity_2020 (SDSS at $z\sim 0.07$), and sanders_mosdef_mzr_2021 (MOSDEF at $z\sim 2-3$), as well as the best-fit of the low-mass end of the MZR at $z\sim 3$ provided by Li_mzr_dwarfs_z3_2022 and based on JWST/NIRISS slitless spectroscopy. The MZR curves at $z\sim2-3$ have been scaled down by $\sim0.1$ dex to account for the systematics differences between the metallicity calibrations used in this work and the bian_ldquodirectrdquo_2018 calibrations adopted in the original papers.