The Gas and Stellar Content of a Metal-poor Galaxy at z = 8.496 as Revealed by JWST and ALMA

We include the JWST NIRCam and NIRISS photometric data of S04590 from the catalog presented by G. Brammer et al. (2023, in preparation), which provides a compilation of the JWST ERO photometric data released to date. Briefly, the raw data entered into this catalog has been reduced using the public software package grizli (Brammer 2019), which masks imaging artifacts, provides astrometric calibrations based on the Gaia Data Release 3 catalog, and shifts images to a common pixel scale of 004 pixel⁻¹ using astrodrizzle. Cutouts of the reduced images from NIRCam and NIRISS used in this analysis are provided on a dedicated repository. ²⁷ S04590 is detected in the five deep NIRCam imaging filters F150W, F200W, F277W, F356W, and F444W, with an F150W (≈H-band) magnitude = 27.3. The nondetection in the bluest filter F090W makes this galaxy a clear photometric dropout. We also include two shallower images obtained with NIRISS with the F115W and F200W filters. This set of photometric measurements is derived using the updated zero-points, and corrected for Milky Way extinction. Since the source is extended in the images, we use aperture photometry derived using a diameter of 05 from the catalog by G. Brammer et al. (2023, in preparation).

The centroid of the galaxy is localized to R.A., decl. =7^h23^m2624, −73°26'570. A zoomed-in image showing the false RGB color (F150W+F277W+F444W) of S04590 is shown in Figure 1, together with the spectral energy distribution (SED) inferred from modeling of the photometric data points (see also Section 3.2).

Zoom In Zoom Out Reset image size

The JWST ERO spectroscopic data of S04590 have already been presented and examined by a number of independent studies (Arellano-Córdova et al. 2022; Brinchmann 2022; Carnall et al. 2023; Curti et al. 2023; Katz et al. 2023; Rhoads et al. 2023; Schaerer et al. 2022; Tacchella et al. 2022; Taylor et al. 2022; Trump et al. 2022). The NIRSpec spectroscopy is obtained with the medium-resolution gratings G235M/F170LP (1.75–3.1 μm) and G395M/F290LP (2.9–5.2 μm), with ${ \mathcal R }=\lambda /{\rm{\Delta }}\lambda =1000$ (Jakobsen et al. 2022), for a total exposure time of 8754 s. We use the JWST pipeline to perform the standard wavelength, flat-field, and photometric calibrations to the individual NIRSpec exposure files. ²⁸ We generate the full combined 2D spectra and optimal (Horne 1986) 1D extractions with scripts that extend the standard pipeline functionality (Brammer 2022). We further scale the overall flux density of the spectrum to the derived photometry to improve the absolute flux calibration by matching the integrated flux within the F444W passband to the derived photometry of the same filter. The reduced spectrum is made available on the dedicated repository. ²⁷ Overall, our extracted version of the spectrum appears consistent with the publicly available 1D spectrum from Curti et al. (2023), though we detect a slightly weaker [O ii] λλ3726, 3729 doublet and [O iii] λ4363 line transition (see Section 3.3 for further discussion). We confirm a similar trend in another up-to-date NIRSpec study in Nakajima et al. (2023). Both of these spectra were derived from an independent source extraction and calibration, which alleviated the issues regarding flux calibration and the inferred line ratios from the Phase 3 MAST products due to an error in the first s007 exposure. The spectrum is presented in Figure 1, where the most prominent nebular emission lines used to determine the redshift and infer the physical properties of the galaxy are marked as well.

ALMA Band 5 follow-up observations of [C ii] spectroscopy were performed as a part of the Cycle 8 DDT program (#2022.A.00022.S, PI: S. Fujimoto). The details of the observations, data reduction, calibrations, and imaging procedures are presented in Fujimoto et al. (2022b). Briefly, we reduced and calibrated the ALMA data with the Common Astronomy Software Applications package version 6.4.1.12 (CASA; McMullin et al. 2007) by using the pipeline script in the standard manner. We adopted the natural weighting, resulting in an FWHM size of the synthesized beam of 135 × 125 with 1σ sensitivities for the continuum map and the line cube with a 40 km s⁻¹ channel width of 11.6 μJy beam⁻¹ and 165 μJy beam⁻¹, respectively.

The [C ii] line emission originating from S04590 is detected at 4.5σ, but is blueshifted by 90 km s⁻¹ from the rest-frame optical spectrum and with a spatial offset of 05 from the rest-frame UV component (Fujimoto et al. 2022b; see also Figure 1). The robustness of the detection is substantiated by the simultaneous detection of the FIR [O iii] 88 μm emission line, which, however, is spatially and spectroscopically consistent with the UV, star-forming component.

The [C ii] line is detected with a velocity-integrated flux of S_[CII] = 0.094 ± 0.027 Jy km s⁻¹. This corresponds to an intrinsic luminosity (following, e.g., Solomon & Vanden Bout 2005) of L_[CII] = (1.67 ± 0.37) × 10⁷ L_⊙, corrected for the magnification factor due to gravitational lensing, which we here assume to be μ = 8.69 from the updated Glafic lens model (Harikane et al. 2022). We note the 10%–30% differences among the most recent mass models for this cluster (Caminha et al. 2022; Mahler et al. 2022). To represent this uncertainty and for internal consistency with Fujimoto et al. (2022b) we adopt μ = 8.69 ± 2.61 here. Modeling the [C ii] emission line in the 1D spectrum extracted from the higher-resolution data cube, with a resolution of 30 km s⁻¹, yields an FWHM_[CII],obs =118 ± 36 km s⁻¹ (Fujimoto et al. 2022b), or a line width of σ_[CII],obs = 50 ± 15 km s⁻¹. Correcting for the instrumental resolution this corresponds to an intrinsic line width of ${\mathrm{FWHM}}_{[\mathrm{CII}],\mathrm{intr}}=\sqrt{{\mathrm{FWHM}}_{[\mathrm{CII}],\mathrm{obs}}^{2}-{\mathrm{FWHM}}_{\mathrm{inst}}^{2}}=114\pm 35\,$ km s⁻¹, as reported in Table 1 and used throughout.

Table 1. Measurements and Physical Properties of S04590

R.A. (J2000)	07:23:26.24 (11085933)
Decl. (J2000)	−73:26:57.0 (−7344917)
zspec	8.4959 ± 0.0003
a $\mathrm{log}({M}_{\star }/{M}_{\odot })$	7.15 ± 0.15
a Mean stellar age (Myr)	${1.3}_{-0.3}^{+0.4}$
a AV (mag)	${0.20}_{-0.03}^{+0.04}$
a ${\mathrm{log}}_{10}(U)$	$-{1.12}_{-0.18}^{+0.08}$
a βUV	–1.81 ± 0.08
b S[CII] (Jy km s−1)	(1.1 ± 0.3) × 10−2
b L[CII] (L⊙)	(1.67 ± 0.37) × 107
σ[CII] (km s−1)	50 ± 15
FWHM[CII] (km s−1)	114 ± 35
SFRHβ (M⊙ yr−1)	2.9 ± 0.9
Te (K)	(2.4 ± 0.4) × 104
$12+\mathrm{log}({\rm{O}}/{\rm{H}})$	${7.16}_{-0.12}^{+0.10}$
$\mathrm{log}Z/{Z}_{\odot }$	$-{1.53}_{-0.12}^{+0.10}$
Mdyn,tot (M⊙)	(9.0 ± 4.5) × 108
Mgas (M⊙)	(8.0 ± 4.0) × 108
MZ,ISM (M⊙)	(3.2 ± 1.6) × 105
b M d (M⊙)	<1.2 × 105

Notes. All values reported here that are affected by magnification due to gravitational lensing have been corrected by the magnification factor μ = 8.69.

^a Inferred physical properties from the SED model are based on the integrated photometry (see text for caveats). ^b Measurements from Fujimoto et al. (2022b).

Download table as:  ASCIITypeset image

We first model the morphology of S04590 in all the NIRCam images in which it is detected (excluding F090W) simultaneously using Galfit (Peng et al. 2002). We find that a simple Sersíc profile is able to reproduce the observed light profile, with an effective radius around 1.5–2 pixels depending on the filter used. In the F150W filter (rest-frame UV ≈1600 Å) we derive an effective radius of r_e,UV = 007, corresponding to r_e,UV = 0.33 kpc in the image plane at z = 8.496 (consistent with the results of Rhoads et al. 2023). Correcting for the magnification factor, we estimate a physical size of ${r}_{{\rm{e}},\mathrm{UV}}/\sqrt{\mu }=0.11\,\mathrm{kpc}$ in the source plane.

To infer the physical properties as well as constrain the SFH of S04590, we model the integrated SED using the fitting code Bagpipes (Carnall et al. 2018). We fix the redshift to the spectroscopic NIRSpec redshift z_spec = 8.496 (see Section 3.3). We include nebular emission via Cloudy (Ferland et al. 2017), allowing the ionization parameter, U, to vary within $-3\lt {\mathrm{log}}_{10}(U)\lt -1$ to account for the typically higher ionization parameter at z ≳ 6 (Sugahara et al. 2022), and the derived electron temperature (see, e.g., Arellano-Córdova et al. 2022; Curti et al. 2023; Katz et al. 2023; Trump et al. 2022; and the section below). We assume a Salim et al. (2018) attenuation curve, which is notably steeper than the typically adopted Calzetti extinction curve (Calzetti et al. 2000). The steeper Salim et al. (2018) curve better represents low-mass, low-metallicity galaxies (Reddy et al. 2018) and effectively outputs a lower A_V . We use a constant SFH model, which mimics a bursty SFH on short timescales, expected for young galaxies at this redshift. A more detailed description of the priors used in the modeling, together with the resulting posterior distributions, is provided in Appendix A.

The best-fit SED model based on the integrated photometry shown in Figure 1 yields a stellar mass ${\mathrm{log}}_{10}({M}_{\star }/{M}_{\odot })\,=7.15\pm 0.15$, a young stellar population with a mean stellar age of ${1.3}_{-0.3}^{+0.4}$ Myr, a relatively blue UV continuum slope β = −1.81 ± 0.08, and a low dust attenuation ${A}_{V}={0.20}_{-0.03}^{+0.04}$ mag for S04590, as summarized in Table 1. This is overall consistent with the results from Carnall et al. (2023). We note that Giménez-Arteaga et al. (2022) found a substantially higher stellar mass of M_⋆ ≈ 10⁸ M_⊙ based on a resolved SED study of a set of galaxies at z ≳ 5 (including S04590), which has also been recovered assuming a burstier SFH (Tacchella et al. 2022). To take into account any potential systematic uncertainties on the stellar mass estimates from varying SFH models (Whitler et al. 2023) or in general across SED fitting codes (Pacifici et al. 2022), we assume a more conservative stellar mass of M_⋆ = 10^7.2–10^8.0 M_⊙ in the following analysis.

Based on the extracted 1D NIRSpec spectrum, we derive the line fluxes of the following nebular and auroral emission lines: [O iii] λλλ4363, 4959, 5007; the doublet [O ii] λλ3726, 3729; and the Balmer lines Hδ, Hγ, and Hβ (as also highlighted in Figure 1). We note that He i and [Ne iii] have also been detected in the spectrum of S04590 (e.g., Carnall et al. 2023; Schaerer et al. 2022), but we focus on the above transitions, which are used in the primary analysis to infer the physical properties of S04590. For each line we derive the fluxes by modeling them with Gaussian profiles, keeping each flux as a free parameter but tying the redshift and the FWHM broadening together across transitions. This effectively assumes that all the nebular lines originate from the same H ii region within the galaxy. In the following, however, we only consider the line ratios of transitions close together in wavelength space to minimize the propagation of any potential offsets in flux calibration in the derived results. The regions of the spectrum zoomed in on for these particular transitions with the best-fit models included are provided in Appendix B.

From the joint fit, we measure a redshift of z = 8.4959 ± 0.0003. The derived Balmer line ratio, Hγ/Hβ = 0.54 ± 0.04, which is consistent with the theoretically predicted ratio for a Case B recombination scenario (Osterbrock & Ferland 2006), further implies a low dust attenuation in S04590 as revealed by the best-fit SED model. We infer the SFR based on the Hβ flux measurements as

$\begin{eqnarray}&&{\mathrm{SFR}}_{{\rm{H}}\beta }({M}_{\odot }\,{\mathrm{yr}}^{-1})=5.5\times {10}^{-42}{L}_{{\rm{H}}\beta }(\mathrm{erg}\,{{\rm{s}}}^{-1})\times {f}_{{\rm{H}}\alpha /{\rm{H}}\beta },\end{eqnarray} \tag{ 1 }$

assuming a Kroupa (2001) IMF and the predicted f_Hα/Hβ = 2.76 from the Case B recombination scenario  (at T_e = 20,000 K). We derive an intrinsic SFR of SFR_Hβ = 2.9 ± 0.9 M_⊙ yr⁻¹ based on the photometrically calibrated spectrum and taking into account the magnification factor. The derived SFR and stellar mass are consistent with the star-forming main sequence implied by the parameterization of Speagle et al. (2014) when extrapolated to z = 8.5, in addition to recent results presenting the SFR–M_⋆ main sequence of galaxies at z ∼ 7–10 (Heintz et al. 2022b). Additionally, the specific SFR (sSFR = SFR/M_⋆) of log sSFR = −6.7 ± 0.4 yr⁻¹ is consistent with recent estimates of galaxies at z ≳ 7 (Stefanon et al. 2022; Topping et al. 2022) and the observed redshift evolution of sSFR(z) out to z ∼ 8.5.

To determine the electron temperature T_e and the gas-phase metallicity through the oxygen abundance $12+\mathrm{log}$(O/H) we follow the prescriptions by Izotov et al. (2006). We derive T_e = (2.4 ± 0.4) × 10⁴ K and $12+\mathrm{log}({\rm{O}}/{\rm{H}})={7.16}_{-0.12}^{+0.10}$ (i.e., $\mathrm{log}Z/{Z}_{\odot }\,=\,$−1.53, assuming $12+\mathrm{log}{({\rm{O}}/{\rm{H}})}_{\odot }=8.69;$ Asplund et al. 2009) for all n_e ≲ 10⁴ cm⁻³, which is well beyond the inferred electron densities of H ii regions in local galaxies (Hunt & Hirashita 2009) and of galaxies at higher redshifts (n_e ≈ 10² cm⁻³; Gillman et al. 2022). Fujimoto et al. (2022b) also found ${n}_{e}={220}_{-100}^{+170}\,$ cm⁻³ for S04590 directly from the [O iii] 88 μm/λ5007 line ratio. Our results are generally consistent with previous estimates (Arellano-Córdova et al. 2022; Brinchmann 2022; Curti et al. 2023; Rhoads et al. 2023; Schaerer et al. 2022; Trump et al. 2022), though we note the slightly lower T_e and higher $12+\mathrm{log}$(O/H), which represent the weaker [O ii] λλ3726, 3729 and [O iii] λ4363 line transitions detected in our extracted version of the spectrum. The oxygen abundance is found to be dominated by O²⁺, with only a small contribution from O⁺. We further note that the ratio [O iii] λ5007/Hβ = 3.65 ± 0.11 places S04590 in the upper region of the Baldwin–Phillips–Terlevich diagram (Baldwin et al. 1981), suggesting high ionization parameters from harder ionization fields or potential active galactic nucleus contributions. The narrow emission lines seem to exclude the latter, however. Our results thus mainly point to S04590 being a dense, highly ionized star-forming galaxy.

We find that the combined mass and metallicity of S04590 is consistent with those predicted from the Astraeus (Ucci et al. 2023) and Feedback in Realistic Environments (FIRE) (Ma et al. 2016) simulations, extrapolated to z ≈ 8 (see Figure 2), but slightly lower than the recently observed mass–metallicity (MZ) relation of galaxies at z ≈ 7–10 (Heintz et al. 2022b). We note that none of the empirical calibrations of the fundamental SFR–Z–M_⋆ plane relation derived for galaxies at z ≈ 0–3 (e.g., Curti et al. 2020; Sanders et al. 2021; Li et al. 2022) are able to recover the derived low metallicity of S04590 (see, e.g., Curti et al. 2023). Our SFR and M_⋆ measurements for instance predict $12+\mathrm{log}$(O/H) = 7.76, about 0.6 dex higher than that derived from the strong-line diagnostics. This offset appears to be ubiquitous in high-z galaxies at z ≳ 5 (Heintz et al. 2022b), likely indicating substantial infall of neutral, pristine gas diluting the metallicity at z ≳ 5. Intriguingly, S04590 shows a pair of mass and metallicity measurements similar to those of the local, extremely metal-poor galaxy I Zwicky 18 (I Zw 18), which is characterized by an oxygen abundance of $12+\mathrm{log}({\rm{O}}/{\rm{H}})=7.14$ and M_⋆ = 2 × 10⁷ M_⊙ (Madden et al.2013).

Zoom In Zoom Out Reset image size

Determining the gas-phase metallicity from the description by Jones et al. (2020, their Equation (2)) relying on the FIR [O iii] 88 μm transition yields $12+\mathrm{log}({\rm{O}}/{\rm{H}})=7.82$. This method thus overpredicts the metallicity by 0.65 dex as compared to the results based on the direct T_e method derived from the redshifted optical nebular and auroral emission lines here. Comparing this metallicity to other strong-line diagnostics defined for local "analogs" of high-redshift galaxies (Bian et al. 2018) yields offsets of −0.2, 0.4, and −0.3 dex for the R23, O32, and O3Hβ calibrations, respectively. The most accurate calibration for this particular galaxy seems to be the O32 calibration from the Sloan Digital Sky Survey reference sample in Bian et al. (2018), similar to the O32 diagnostics derived by Curti et al. (2020), which seems to be consistent with the direct metallicity within 0.1 dex (as is also evident from the galaxy at z = 9.5 presented by Heintz et al. 2022b).

Assuming that [C ii] traces the ISM of S04590 we can further derive the total dynamical mass M_dyn of this galaxy. We caution, however, that the spectroscopic blueshift and the ≈2σ spatial offset of [C ii] relative to the UV component and [O iii] emission (Fujimoto et al. 2022b) indicate that the origin of the [C ii] emission is not fully co-spatial with the most intense, star-forming region of the galaxy. Such extended, blueshifted [C ii] emission has been observed in other high-z galaxies, and may hint at an offset an abundant gas reservoir, a shock-heated outflowing gas, or minor mergers (e.g., Maiolino et al. 2015; Kohandel et al. 2019; Arata et al. 2020; Akins et al. 2022; Katz et al. 2022). The latter scenario would typically produce much broader (>1000 km s⁻¹) [C ii] line profiles (Appleton et al. 2013), whereas the observed narrow-line feature is consistent with the low stellar mass of S04590. We also caution that we are not able to firmly exclude the presence of a merging system due to the limited spatial resolution of our ALMA observations (Rizzo et al. 2022). However, given that the [C ii] line width and luminosity are consistent with predictions based on the stellar mass and SFR of S04590, we assume in the following that the [C ii] emission is physically associated with the ISM of the galaxy.

We infer M_dyn based on the measured FWHM of the [C ii] emission line corrected by the instrumental resolution, FWHM_[CII] = 114 ± 35 km s⁻¹, as

$\begin{eqnarray}&&{M}_{\mathrm{dyn}}=1.16\times {10}^{5}{v}_{\mathrm{circ}}^{2}{D}_{[\mathrm{CII}]}\end{eqnarray} \tag{ 2 }$

following, e.g., Wang et al. (2013) and Willott et al. (2015). Here we approximate the circular velocity to v_circ ≈ 0.52 ×FWHM_[CII], appropriate for velocities dominated by nonordered rotation (Decarli et al. 2018; Neeleman et al. 2021). Fujimoto et al. (2022b) measured a circularized [C ii] effective radius of 069 ± 042 in the image plane, which corresponds to 023 ± 014 in the source plane or to a physical size of r_e,[CII] = 1.1 ± 0.7 kpc at z = 8.496. Such extended [C ii] halos reaching beyond the UV components of the galaxies are typically observed at z ≳ 6, with r_e,[CII] ≈ 2−3 × r_e,UV (Smit et al. 2018; Fujimoto et al. 2019; Matthee et al. 2019; Carniani et al. 2020; Akins et al. 2022; Fudamoto et al. 2022), though S04590 presents one of the most notable cases with r_e,[CII] ≈ 10 × r_e,UV. Assuming the "disk" diameter is D_[CII] = 2 ×r_e,[CII], we infer D_[CII] = 2.2 ± 1.4 kpc. This yields a dynamical mass of M_dyn = (9.0 ± 4.5) × 10⁸ M_⊙. We note that if we assume ordered, rotation-dominated motion and an average inclination angle of 45°, the dynamical mass increases by ≈50%. However, since there is no strong evidence of ordered motion in S04590, we adopt the above-derivedvalue for the dynamical mass.

Based on the assumption that the dynamical mass is dominated by the baryonic matter content within the [C ii]-emitting region (which is confirmed to be the case for dusty, massive galaxies at z ∼ 4–5; Rizzo et al. 2020, 2021), the total dynamical mass of S04590 subtracted by its stellar mass should represent the total gas content of this galaxy (in molecular and/or neutral, atomic form); M_gas = M_dyn − M_⋆. For S04590, the stellar mass derived from the SED fitting only constitutes ≈1%–2% of the total dynamical mass, such that M_gas ≈ M_dyn. However, using more conservative estimates of M_⋆ ≈ 10⁸ M_⊙ derived from more complex SFHs or spatially resolved analyses (Giménez-Arteaga et al. 2022; Tacchella et al. 2022), we estimate a total gas mass of M_gas = (8.0 ±4.0) × 10⁸ M_⊙. This yields a gas mass excess and gas fraction of $\mathrm{log}({M}_{\mathrm{gas}}/{M}_{\star })\,=\,$ 0.9–1.7 and f_gas = M_gas/(M_gas +M_⋆) ≳ 90%. This gas mass excess is substantial but not too surprising given the overall increase of M_gas/M_⋆ with increasing redshift (Geach et al. 2011; Carilli & Walter 2013; Tacconi et al. 2013; Scoville et al. 2017; Tacconi et al. 2018; Heintz et al. 2021, 2022a) and decreasing metallicity (Catinella et al. 2018; Heintz et al. 2021; Stark et al. 2021).

In Figure 3 we show the total gas mass excess and gas depletion of S04590 together with a set of literature data as a function of redshift. We include the local sample of galaxies at z ∼ 0 by Leroy et al. (2008), for which we estimate ${M}_{\mathrm{gas}}={M}_{\mathrm{HI}}+{M}_{{{\rm{H}}}_{2}}$. At high z (z ≳ 4), we compile galaxy samples for which the dynamical mass and the [C ii]-emitting size can be inferred, which include the ALMA-ALPINE survey (z ∼ 4–6; Béthermin et al. 2020; Faisst et al. 2020; Le Fevre et al. 2020), the sample from Capak et al. (2015; z ∼ 5–6), and ALMA-REBELS (z ∼ 6–8; Bouwens et al. 2022). These all target "regular" star-forming galaxies at the respective survey redshifts. To estimate M_gas for the high-z sample compilation, we again assume M_gas = M_dyn − M_⋆ with M_dyn estimated using Equation (2) and M_⋆ as provided in the catalogs. If they are not reported, we assume inclination angles of 45° for the [C ii]-emitting galaxy compilation. We observe a strong increase in the gas mass excess as a function of redshift, with averages of $\mathrm{log}({M}_{\mathrm{gas}}/{M}_{\star })=\,$−0.3 at z ≈ 0 and $\mathrm{log}({M}_{\mathrm{gas}}/{M}_{\star })=0.7$ at z ≈ 6–8. Similarly, we compute the gas depletion times for the compiled galaxy samples and find averages of t_depl = 7 Gyr at z ≈ 0, declining to t_depl = 0.2 Gyr at z ≳ 4. In both cases, S04590 is a clear outlier owing to its large inferred gas mass. However, this can naturally be explained by its low metallicity and stellar mass, for which a large gas fraction is expected, as shown in Figure 4, which are consistent with the properties of I Zw 18.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We now attempt to further disentangle the baryonic mass components of S04590 by estimating the contributions from the cold molecular and neutral atomic gas phases to the total gas mass. Recently, Zanella et al. (2018) calibrated the [C ii] luminosity to the total molecular gas mass M_mol for a set of galaxies at lower redshifts (z ∼ 0–2). They found a conversion factor of α_[CII] = [C ii]-to-M_mol = 30 M_⊙/L_⊙ with a scatter of 0.3 dex (see also Madden et al. 2020), independent of the metallicity. Simulations of galaxies at z ∼ 6 suggest a slightly lower but consistent average conversion of α_[CII] = 18 M_⊙/L_⊙ (Vizgan et al. 2022b). Applying these calibrations to L_[CII] yields M_mol = (3.0–5.0) × 10⁸ M_⊙ (depending on the conversion used), with a combined uncertainty of ≈40% taking into account the statistical uncertainty and scatter in the calibration. The molecular gas content thus comprises ≈30%–50% of M_gas by mass.

However, [C ii] can also be used as a proxy for the neutral atomic H i gas mass (Heintz et al. 2021, 2022a; Vizgan et al. 2022a) since it has been shown observationally to predominantly originate from the neutral gas phase in the Milky Way and nearby galaxies (Madden et al. 1993, 1997; Pineda et al. 2014; Croxall et al. 2017), with supporting evidence from high-redshift galaxy observations (e.g., Novak et al. 2019; Meyer et al. 2022) and simulations (Katz et al. 2017; Ramos Padilla et al. 2022) as well. We adopt the [C ii]-to-H i calibration from Heintz et al. (2021), prescribed as

$\begin{eqnarray}\begin{array}{rcl}\mathrm{log}{\beta }_{[\mathrm{CII}]} & = & \mathrm{log}{M}_{\mathrm{HI}}/{L}_{[\mathrm{CII}]}=(-0.87\pm 0.09)\\ & & \times \,\mathrm{log}(Z/{Z}_{\odot })+(1.48\pm 0.12)\end{array}\end{eqnarray} \tag{ 3 }$

(see also Heintz et al. 2022a). In this previous work, the M_HI of [C ii]-emitting galaxies at z ∼ 6–8 from REBELS (e.g., Bouwens et al. 2022) could only be inferred based on uncertain metallicity estimates assuming an SFR–Z–M_⋆ fundamental plane relation (e.g., Curti et al. 2020). With the accurate gas-phase metallicity from the NIRSpec analysis performed here, we obtain the first robust high-z M_HI estimate of M_HI ≈ 10¹⁰ M_⊙. This is ≈10× more massive than the inferred dynamical mass. This can be explained by (i) the [C ii]-to-H i relation being less robust at these high redshifts or low metallicities; (ii) the existence of a much larger H i gas component situated at larger radii than assumed for the [C ii]-emitting disk, which then causes the total dynamical mass of the system to be underestimated; or (iii) an increased excitation of L_[CII]. However, the derived ratio L_[CII]/SFR ≈ 10⁷ L_⊙/(M_⊙ yr⁻¹) (see also Fujimoto et al. 2022b) appears consistent with the ubiquitously observed SFR−L_[CII] relation (e.g., De Looze et al. 2014; Schaerer et al. 2020). We also note that I Zw 18 again has remarkably comparable properties in terms of the metallicity ($12+\mathrm{log}$(O/H) = 7.14), and with an estimated H i gas mass of M_HI = 10⁸ M_⊙ and [C ii] luminosity L_[CII] = 1.1 × 10⁵ L_⊙ (e.g., Rémy-Ruyer et al. 2014; Cormier et al. 2015), yielding an equally high [C ii]-to-H i ratio of $\mathrm{log}{\beta }_{[\mathrm{CII}]}=\mathrm{log}({M}_{\mathrm{HI}}/{L}_{[\mathrm{CII}]})\,=\,2.95\,{M}_{\odot }/{L}_{\odot }$. In any case, since M_⋆ and M_mol only constitute 2%–10% and 30%–50% of the total dynamical mass of S04590, respectively, we surmise that the largest baryonic mass component is likely in the form of neutral atomic H i gas.

With the first high-z combination of the gas-phase metallicity and the total gas mass, we can now infer the ISM metal mass, M_Z,ISM, of S04590 at an unprecedented redshift of z = 8.5 (previously only possible up to z ≈ 3; Sanders et al. 2023). We determine this as

$\begin{eqnarray}&&{M}_{Z,\mathrm{ISM}}={M}_{\mathrm{gas}}\times {10}^{12+\mathrm{log}({\rm{O}}/{\rm{H}})-8.69}\times {Z}_{\odot },\end{eqnarray} \tag{ 4 }$

where Z_⊙ = 0.0134 is the solar metallicity by mass. This yields M_Z,ISM = (3.2 ± 1.5) × 10⁵ M_⊙. The expected total Type II supernova metal production estimated by Peeples et al. (2014), assuming the prescribed SFHs by Leitner (2012) and a nucleosynthetic yield of y = 0.033, is ${M}_{Z,\mathrm{SN}}=(7.6\pm 1.9)\,\times {10}^{5}\,{M}_{\odot }$ marginally consistent with our estimate. Assuming a briefer SFH following Sanders et al. (2023), which is likely more appropriate for S04590 given its high redshift, we obtain consistent results. This indicates that the majority of the produced metals from core-collapse supernovae in S04590 are retained in the ISM. This is in contrast with star-forming galaxies at z ∼ 2–3, which only retain ≈20% of the produced metals in the ISM (Sanders et al. 2023), and even more so with local galaxies, which have most of their metals captured in stars. Further, this puts strong constraints on the possibility of significant outflows in S04590, as most of the produced metals appear to still be confined to the ISM.

In combination with the constraints placed on the dust mass of M _d < 1.2 × 10⁵ M_⊙ utilizing the FIR continuum coverage provided by ALMA from Fujimoto et al. (2022b), we further determine upper bounds on the mass ratios of the dust-to-gas (DTG, M _d /M_gas) and dust-to-metal (DTM, M _d /M _Z ) content of S04590. We find logDTG < −3.8 and $\mathrm{log}$DTM < −0.5, as also shown in Figure 5. These DTG and DTM mass ratios are substantially lower than the Milky Way average of DTG ≈ 10⁻² and DTM ≈ 0.5. However, this is likely due to the metallicity dependence of the DTG and DTM ratios predicted from simulations (Popping et al. 2017; Hou et al. 2019; Li et al. 2019; Vijayan et al. 2019; Graziani et al. 2020; Triani et al. 2020) and those observed in local galaxies (Rémy-Ruyer et al. 2014; De Vis et al. 2019) and high-z sightlines in absorption (De Cia et al. 2016; Wiseman et al. 2017; Péroux & Howk 2020; Popping & Péroux 2022). Comparing our measurements to those on the DustPedia ²⁹ sample of local galaxies at z ≈ 0 (De Vis et al. 2019) and to absorption-derived ratios from quasar absorption line systems (DLAs) at z ≈ 1–4 (Péroux & Howk 2020; Popping & Péroux 2022) in Figure 5, we find that they are indeed consistent with the expected evolution with metallicity that appears universal across redshifts. Moreover, these results are again in good agreement with the mass ratios derived for I Zw 18, with an estimated dust mass of M_dust ≈ 2 × 10³ M_⊙ (Cannon et al. 2002), yielding a DTG ratio of 2 × 10⁻⁵ and a DTM ratio ≈ 5%. These results highlight the strong deviation of the DTG and DTM mass ratios from the Milky Way average at high redshifts and low metallicities.

Zoom In Zoom Out Reset image size