The ionised interstellar medium of DSFGs revealed by JWST/NIRSpec and ALMA: Super-solar metallicity, low ionisation parameters and, typical electron densities

Abstract

We present a detailed study of near-infrared (2-4$\rm μ$m) JWST/NIRSpec spectra of 48 high-redshift ($z=2.53^{+1.32}_{-0.70}$) galaxies detected with ALMA at $>3σ$. From a multi-wavelength SED analysis we establish the sample has a a median stellar mass of $\rm\log_{10}(M_\ast/M_\odot)=10.8\pm0.1$ and dust mass of $\rm\log_{10}(M_{\rm d}/M_\odot)=8.7\pm0.1$, covering a broad range of far-infrared luminosity $\rm (\log_{10}(L_{FIR}/L_\odot)=10.9-12.7)$. The majority of sources show no signs of AGN activity, with 40% having either X-ray counterparts $(\rm L_{Xc}>10^{42}erg/s)$, elevated optical line ratios, or broad (FWHM>800 km/s) H$α$ profiles, although we note this is a lower limit due to the stochastic placement of NIRSpec slits. We establish the sample has a median gas-phase metallicity of $12+\log({\rm O/H})=8.71\pm0.02$, as derived from the [NII]/H$α$ ratio, with the most FIR-luminous galaxies ($\rm\log_{10}(L_{\rm FIR}/L_\odot)>12$) falling $0.15\pm0.03$dex above the fundamental metallicity relation. From the [SII] emission-line doublet ratio, we measure a median electron density of $\log_{10}(n_{\rm e}/{\rm cm}^{-3})=2.53\pm0.07$ consistent with less-massive, star-forming, galaxies at the same epoch. For nine galaxies with [OII] and H$β$ detections (median $\rm\log_{10}(L_{\rm FIR}/L_\odot)=11.81\pm0.15$), we derive a median observed (dust-uncorrected) ionisation parameter of $\rm\log_{10}(U)=-2.84\pm0.06$. Our results indicate that luminous far-infrared galaxies are massive, chemically evolved systems that appear to deviate from the standard dust and metal production equilibrium observed in less obscured galaxies. This study demonstrates the synergy of JWST and ALMA in unveiling the nature of DSFGs, and highlights the need for a NIRSpec survey of uniformly selected, massive, dust-obscured, galaxies to fully characterise their interstellar medium.

Figures (16)

• Figure 1: JWST/NIRCam false colour images for the final sample of 48 ALMA -- NIRSpec galaxies ranked in descending far-infrared luminosity (L$_{\rm FIR}$). For each source we show the 40$\,\times\,$40 cutout as labelled with the DJA-ID, spectroscopic redshift, far-infrared luminosity, ALMA peak flux centroid (purple star) and NIRSpec MSA slits (white boxes) for one dither position. For each source we select the NIRCam filter that corresponds to rest-frame 2µm (R, near-infrared), 0.7µm (G, optical) and 0.3µm (B, ultra-violet). Sources identified as AGN (see Sec. \ref{['Sec:AGN']}) are highlighted, where $\dagger$ indicates broad emission lines, $\ddagger$ indicates an X-ray counterpart and $*$ highlights elevated optical-line ratios.
• Figure 2: Three examples of the mercurius far-infrared modified blackbody modelling. For each source show the photometry (black points; upper limits shown as arrows) and best-fit blackbody curve (purple-dashed line) and 1$\sigma$ uncertainty (purple shaded region). We report the best-fit parameter in each panel, as well as NIRSpec (and ALMA) source ID, redshift and CMB temperature. Sources identified as AGN (see Sec. \ref{['Sec:AGN']}) are highlighted, where $\dagger$ indicates broad emission lines, $\ddagger$ indicates an X-ray counterpart and $*$ highlights elevated optical-line ratios. The SEDs, and MCMC corner plots, of the full sample are given in Section \ref{['App:FIR']}.
• Figure 3: JWST/NIRSpec spectra for four galaxies in the sample across a range of far-infrared luminosity. For each source we show the NIRSpec spectra (and 1$\sigma$ uncertainty) from rest-frame 0.36 -- 0.54µm highlighting the [Oiii]/H$\beta$ complex and 0.65 -- 0.74µm highlighting the [Nii]/H$\alpha$ complex, with detected ($>$3$\sigma$) emission lines labelled. We further indicate the NIRSpec-ID, bagpipes derived stellar mass, observed [Nii]/H$\alpha$ ratio, spectroscopic redshift as well as ALMA-ID, ALMA flux density and far-infrared luminosity of each source. Sources identified as AGN (see Sec. \ref{['Sec:AGN']}) are highlighted, where $\dagger$ indicates broad emission lines, $\ddagger$ indicates an X-ray counterpart and $*$ highlights elevated optical-line ratios.
• Figure 4: Left: The offset to the star-forming main sequence relation from Popesso2023 as function of stellar mass. The ALMA-NIRSpec galaxies are coloured by their far-infrared luminosity (L$_{\rm FIR}$) and we indicate the intrinsic 0.3 dex scatter in the main-sequence (grey shaded region). We include comparison samples of high-redshift star-forming galaxies from the CEERS and AURORA surveys (grey histogram), as well as H$\alpha$-detected SMGs from the KAOSS survey Birkin2022Birkin2024 (grey plusses) and AS2UDS survey Dud2020 (grey contours), in addition to far-infrared bright ($S_{\rm 870\mu m}$$>$ 1mJy) galaxies at $z$ = 1 -- 5 (blue density region) from the L-Galaxies SAM Araya2025. The ALMA-NIRSpec sample shows a strong correlation between main sequence offset and far-infrared luminosity, with far-infrared bright log$_{10}$(L$_{\rm IR}$[L$_{\odot}$]) $>$ 12.0 galaxies falling ($\gtrsim$0.5 dex) above the main-sequence relation, consistent with other bright far-infrared detected, and simulated, galaxies at the same epoch. Right: The dust to stellar mass ratio as a function of redshift with the ALMA -- NIRSpec galaxies coloured by far-infrared luminosity. We again include comparison samples of high-redshift SMGs from KAOSS, AS2UDS as well as CO-selected SMGs from Birkin2021. Our sample has similar dust to stellar mass ratios when compared to traditional SMG samples at the same redshift. We also show the observational scaling relation from Magnelli2020 (pink lines) and the Santa-Cruz SAM (blue lines) Popping2017 for a range of stellar masses, in addition to the observationally-derived scaling from Kokorev2021 at fixed stellar mass ($\log_{10}$(M$_{\rm \ast}$[M$_{\odot}$]) = 10.8) as a function of main sequence offset (green lines).
• Figure 5: Observed (dust-uncorrected) [Oiii]/H$\beta$ as function of [Nii]/H$\alpha$ (left) and [Sii]/H$\alpha$ (right). Sources with S/N$<$3 in the respective emission-line are shown as limits. The distribution of [Nii]/H$\alpha$ and [Sii]/H$\alpha$ for the full sample is shown by the grey histograms. We show comparison samples of typical star-forming galaxies at $z$$\sim$ 0 (red density) from the Sloan Digital Sky Survey Bricnhmann2004, as well as a redshift matched sample of star-forming galaxies from the CEERS Shapley2023 and AURORA Surveys Shapley2025. The black solid curve shows the division between star-formation acticity and AGN Kauffmann2003 at $z$ = 0 whilst the theoretical demarcation at $z$ = 2 and $z$ = 3 from Kewley2013 is given by the blue- and red-dashed lines respectively. The black arrow indicates the impact of A$_{V}$ = 5 dust correction using a Calzetti2000 dust law. The majority of ALMA -- NIRSpec galaxies are consistent with typical SFGs at the same epoch, pushing to lower [Oiii]/H$\beta$ line ratios, with only three galaxies showing AGN-like optical line diagnostics.