Draft version February 23, 2024

Typeset using LATEX twocolumn style in AASTeX631

Diverse Oxygen Abundance in Early Galaxies Unveiled by Auroral Line Analysis with JWST
Takahiro Morishita ,1 Massimo Stiavelli ,2 Claudio Grillo ,3, 4 Piero Rosati ,5, 6 Stefan Schuldt ,3, 4
Michele Trenti ,7, 8 Pietro Bergamini ,3, 5 Kit Boyett ,7, 8 Ranga-Ram Chary ,1 Nicha Leethochawalit ,9
Guido Roberts-Borsani ,10 Tommaso Treu ,11 and Eros Vanzella 12
1 IPAC, California Institute of Technology, MC 314-6, 1200 E. California Boulevard, Pasadena, CA 91125, USA
arXiv:2402.14084v1 [astro-ph.GA] 21 Feb 2024

2 Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA
3 Dipartimento di Fisica, Università degli Studi di Milano, Via Celoria 16, I-20133 Milano, Italy
4 INAF - IASF Milano, via A. Corti 12, I-20133 Milano, Italy
5 INAF - OAS, Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, via Gobetti 93/3, I-40129 Bologna, Italy
6 Dipartimento di Fisica e Scienze della Terra, Università degli Studi di Ferrara, Via Saragat 1, I-44122 Ferrara, Italy
7 School of Physics, University of Melbourne, Parkville 3010, VIC, Australia
8 ARC Centre of Excellence for All Sky Astrophysics in 3 Dimensions (ASTRO 3D), Australia
9 National Astronomical Research Institute of Thailand (NARIT), Mae Rim, Chiang Mai, 50180, Thailand
10 Department of Astronomy, University of Geneva, Chemin Pegasi 51, 1290 Versoix, Switzerland
11 Department of Physics and Astronomy, University of California, Los Angeles, 430 Portola Plaza, Los Angeles, CA 90095, USA
12 INAF – OAS, Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, via Gobetti 93/3, I-40129 Bologna, Italy

Submitted to ApJ

ABSTRACT
We present deep JWST NIRSpec observations in the sightline of MACS J1149.5+2223, a massive
cluster of galaxies at z = 0.54 . We report the spectroscopic redshift of 28 sources at 3 < z <
9.1, including 9 sources with the detection of the [O iii]λ4363 auroral line. Combining these with 16
[O iii]λ4363 -detected sources from publicly available JWST data, our sample consists of 25 galaxies with
robust gas-phase metallicity measurements via the direct method. We observe a positive correlation
between stellar mass and metallicity, with a ∼ 0.5 dex offset down below the local relation. Interestingly,
we find a larger than expected scatter of ∼ 0.3 dex around the relation, which cannot be explained
by redshift evolution among our sample or other third parameter. The scatter increases at higher
redshift, and we attribute this to the enrichment process having higher stochasticity due to shallower
potential wells, more intense feedback processes, and a higher galaxy merger rate. Despite reaching to
a considerably low-mass regime (log M∗ /M⊙ ∼ 7.3), our samples have metallicity of log (O/H) +12 > ∼ 7,
i.e. comparable to the most metal poor galaxies in the local Universe. The search of primordial galaxies
may be accomplished by extending toward a lower mass and/or by investigating inhomogeneities
at smaller spatial scales. Lastly, we investigate potential systematics caused by the limitation of
JWST’s MSA observations. Caution is warranted when the target exceeds the slit size, as this situation
could allow an overestimation of “global” metallicity, especially under the presence of strong negative
metallicity gradient.

1. INTRODUCTION further enhancement in chemical abundances. The ac-

Gas cycle within a galaxy system carries an important cumulation of stars formed in such a cyclical way repre-
role. The ejecta from previously formed stars enrich sur- sents a major component of galaxies in a later epoch,
rounding gas, some fraction of which is subject to be culminating an established relation of metal contents
used in the following star formation cycles, leading to and other fundamental galaxy properties, such as stel-
lar mass, seen in the local universe (e.g., Tremonti et al.
2004; Gallazzi et al. 2005; Andrews & Martini 2013;
Corresponding author: Takahiro Morishita Kirby et al. 2013; Curti et al. 2020). Observations have
[email protected] reported the presence of the relationship out to z ∼ 3,
2 Morishita, Stiavelli, et al.

but with an offset to the local relation (e.g., Maiolino In this study, we present our Cycle 1 GTO observa-
et al. 2008; Zahid et al. 2011; Henry et al. 2013; Steidel tions (PID 1199; Stiavelli et al. 2023), executed along
et al. 2016; Sanders et al. 2021). the sightline of MACS J1149.5+2223, a massive clus-
However, galaxies are by no means closed systems, ter of galaxies at z = 0.54 . We report the detection of
and thus chemical enrichment on a galaxy lifetime scale the [O iii]λ4363 line in 9 galaxies, out of 28 sources at
is multifaceted. Star formation, ejection of enriched gas 3 < z < 9.1 spectroscopically confirmed in the program.
by stellar feedback and accretion of pristine gas from We combine this new sample with spectroscopic data
cosmic web have a complex interplay that ultimately for an additional 16 galaxies with [O iii]λ4363 available
defines a unique chemical enrichment history for a given in the literature, aiming to obtain robust oxygen abun-
object. The situation is expected to be even more com- dance measurements of these galaxies and to investigate
plicated at earlier times, where the gas infall rate is their relation with global properties. In Sec. 2 we present
higher, star formation is more intense, and the gravita- our data reduction, followed by our spectroscopic and
tional potential well is not deep owing to the lower mass photometric analyses in Sec. 3. We then characterize
of (proto)galaxies. As such, extending the observational their physical properties and investigate the relationship
frontier toward higher redshift is critical to our under- between stellar mass and oxygen abundance in Sec. 4.
standing of the origin of such a tight correlation between We discuss the inferred properties of early galaxies in
metallicity, which is often represented by oxygen abun- the context of galaxy evolution in Sec. 5. Where rele-
dance measurements, and other galaxy properties. vant, we adopt the AB magnitude system (Oke & Gunn
With the advent of JWST, galaxy interstellar medium 1983; Fukugita et al. 1996), cosmological parameters of
studies are now possible at z > ∼ 4 (Curti et al. 2023a; Ωm = 0.3, ΩΛ = 0.7, H0 = 70 km s−1 Mpc−1 , and the
Heintz et al. 2023; Laseter et al. 2023a; Nakajima et al. Chabrier (2003) initial mass function. Throughout the
2023; Sanders et al. 2023; Stiavelli et al. 2023). The manuscript, we use the term metallicity to refer to oxy-
unparalleled JWST sensitivity can capture spectral fea- gen abundance.
tures of high redshift galaxies at rest-frame optical
wavelengths, such as those from oxygen and hydro- 2. DATA
gen, enabling a fair comparison to the variety of well- 2.1. NIRSpec/MSA Spectroscopic Observations
established lower redshift studies in the literature. Most
The NIRSpec/MSA observations were configured with
importantly, the O/H abundance can be inferred from
two medium-resolution gratings (G235M and G395M),
the electron temperature (Te ), which in turn is mea-
aiming to capture rest-frame optical emission lines of
sured from the ratio of the [O iii]λ4363 auroral emission
galaxies at z ∼ 3–9. Our NIRSpec observations
line to [O iii]λ5007 . This measurement, often referred to
were executed over four different visits. The first two
as direct method, enables robust determination of oxy-
visits, #20 and #22, were taken at position angle
gen abundances. In contrast, most of the pre-JWST
PA=257.766 degree with G235M and G395M, respec-
high-z studies adopt indirect metallicity inference, e.g.,
tively. In the second group of visits, the same grat-
R23 (Pettini & Pagel 2004; Maiolino et al. 2008; Marino
ing pair was used, but with a slightly different position
et al. 2013; Curti et al. 2020). Such indirect methods
angle, namely PA=259.660 degree for visit #21/G235M
are in many cases constructed and calibrated from local
and PA=256.766 degree for visit #23/G395M. The total
measurements, and thus have limitation, e.g., are appli-
science time, excluding overhead, is 16632 sec for each
cable only within a moderate metallicity range ≳ 8 (e.g.,
grating, though not all sources have the full exposure
Curti et al. 2020). As such, their application to high-z
coverage due to the mask design.
galaxies is potentially subject to systematic biases.
We designed a slit mask for each of the four MSA vis-
Auroral lines are, however, faint. For example,
its. The parent source catalog was constructed by using
[O iii]λ4363 is of the order of ∼ 1/100 to [O iii]λ5007 ,
the existing HST imaging data, including CLASH, HFF,
hence the direct oxygen measurement was previously
GLASS, and BUFFALO (Postman et al. 2012; Treu
limited to a smaller fraction of z > 0 galaxies, and to
et al. 2015; Lotz et al. 2017; Kelly et al. 2018; Steinhardt
only a handful at z ∼ 2–3 (Sanders et al. 2016, 2020;
et al. 2020). Spectroscopically redshift measurements
Gburek et al. 2019). On the other hand, [O iii]λ4363 in-
available in the literature were also incorporated (Grillo
tensity increases as the electron temperature increases.
et al. 2016; Treu et al. 2016; Shipley et al. 2018, Schuldt
With galaxies being less enriched and hosting more
et al., in prep). In our mask design, the highest priority
intense star formation, the application of the direct
was given to JD1, a previously known spec-z source at
method becomes relatively more favorable within the
z = 9.1 (Hashimoto et al. 2018; Hoag et al. 2019), fol-
higher redshift range enabled by JWST.
lowed by z > 6 galaxies presented in Zheng et al. (2017)
 Metallicity of Galaxies at 3 < z < 9.5 3

and then z > 1 galaxies. The sources in our catalog the total quantities for subsequent analysis (i.e. deriva-
were firstly identified based on the existing WFC3-IR- tion of stellar mass and star formation rate).
stack detection. However, since the NIRSpec coverage
is larger than the WFC3-IR footprint, we expanded our 3. ANALYSIS
source coverage to the ACS F606W+F814W footprint, 3.1. Emission Line Fitting
which offers z <∼ 4 dropout sources as fillers. For each As shown in Fig. 1, the slit positions for some sources
source, the default 3 shutter slitlets were assigned, en-
do not necessarily match in the two grating observations.
abling exposure at three positions separated for 0.′′ 46
To accommodate potential systematic errors due to this,
from each. We selected the default (Entire Open Shut-
we correct for aperture losses. For each grating spectrum
ter Area) for source allocations in APT, which secures
of each source, we apply a global scale factor that is
to place the source center position within 35 milliarcsec
derived by comparing the continuum spectrum (with the
from the shutter boundary. Including fillers and open
wavelengths of strong emission lines being masked) to
shutters in empty sky regions, our MSA masks consisted
the best-fit spectral template from our spectral energy
of in total 341 unique sources.
distribution (SED) analysis (Sec. 3.3).
For the MSA data reduction, we use msaexp1 follow-
We then model the line profile of each emission line
ing Morishita et al. (2023). The one-dimensional spec-
of interest in the extracted 1d spectrum with a Gaus-
trum is extracted via optimal extraction. The source
sian function. Our basic strategy is to define a wave-
light profile along the cross-dispersion direction is di-
length window for each line, model the underlying con-
rectly measured using the 2-d spectrum. For sources
tinuum spectrum by a polynomial fit, and subtract it
with faint continuum or with any significant contamina-
from the 1-d spectrum before the Gaussian fit is carried
tion (i.e. from a failed open shutter), we visually inspect
out. The model includes amplitude, line width, and
the 2-d spectrum and manually define the extraction box
redshift parameters. However, when multiple lines are
along the trace where any emission lines are identified.
located close to each other in wavelength, we fit those
simultaneously using a common redshift parameter, so
blending can effectively be resolved. In addition, dou-
2.2. Imaging Data Reduction and Photometry blet lines (i.e. [O iii]λλ4959+5007 and [N ii]λλ6548+6583 )
NIRCam imaging observations were executed on June are modelled with a fixed line ratio (1:3 and 1:1.7, re-
6 – 8 2023 as part of the same program (Visit #5), with spectively), with single line width and redshift param-
six filters configured (F115W, F150W, F200W, F277W, eters. The total flux of the doublets estimated in this
F356W, and F444W) with the MEDIUM8 readout mode, way remains robust, while we note that the ratio could
totaling in ∼ 1.1 hr on-source exposure in each filter. have a systematic uncertainty ≲ 10 % (e.g., Storey &
We follow the same steps for JWST image reduction Zeippen 2000).
presented in Morishita & Stiavelli (2023). Briefly, we The integrated flux of each line is estimated by sum-
retrieve the raw-level images from the MAST archive ming the corresponding Gaussian component, and the
and reduce those with the official JWST pipeline, with flux error is estimated by summing the error weighted
several customized steps including 1/f -noise subtraction by the amplitude of the Gaussian model in quadra-
and snowball masking. The final drizzled images are ture. In the following analysis, we adopt flux measure-
aligned with the same WCS coordinate as for the HST ments when the line is detected (signal-to-noise ratio
images, with the pixel scale set to 0.′′ 03. In Fig. 1, we S/N ≥ 3); for those not detected, we adopt the 3-σ
show the NIRCam and NIRSpec coverage, on top of the upper limit.
footprint of the existing HST images.
3.2. Direct Metallicity Measurements
We identify sources in the detection image using SEx-
tractor (Bertin & Arnouts 1996), and measure their In general, the line intensity we measure needs to be
fluxes on the psf-matched mosaic images with a fixed corrected for dust absorption. We determine the correc-
aperture of radius r = 0.′′ 16 . The measured aperture- tion on the basis of the ratio of the strongest Balmer
based fluxes of each source is then scaled by a single fac- lines we measure, thus, Hα/Hβ is considered before
tor, defined as fluxauto,F444W /fluxaper,F444W , to the total Hβ/Hγ. We require the Balmer lines to be detected
flux. With this approach, colors remain as those mea- at 5-σ in order to apply such correction. In the absence
sured in the aperture, whereas the fluxes are scaled to of a high S/N detection of Balmer lines, we apply no cor-
rection. Generally, we find the AV values derived from
the Balmer line ratios to be in broad agreement with the
1 https://github.com/gbrammer/msaexp SED-derived AV values, after using the Calzetti (1997)
4 Morishita, Stiavelli, et al.

Figure 1. (Left): Pseudo RGB images of NIRCam in the cluster pointing and the parallel field are shown
(F115W+F150W/F200W+F277W/F356W+F444W for blue/green/red). The confirmed z > 3 sources in our program are
marked (cyan). The HST/ACS images are shown in the background. The Hubble Frontier Field’s parallel pointing is located to
the South of the image (not shown here). NIRSpec footprints are marked (red solid rectangles). (Right): Zoomed view around
the NIRSpec FoV. (Bottom): Cutout RGB stamps of z > 3 sources are shown (2′′ × 2′′ ). The positions of MSA shutters are
overlaid.
rescaling factor that accounts for continuum and emis- not measured at S/N≥ 3 in those cases we estimate Hβ
sion lines dust effects. from another Balmer line using Case B (Osterbrock &
The auroral line [O iii]λ4363 provides a reliable method Ferland 2006) and 10,000 K, assuming no attenuation.
to derive the electron temperature and, from that, the For 20018, 60001, 80001, and m39 we estimate Hβ from
oxygen metallicity by using the Izotov et al. (2006) it- Hγ, while for 80002 we estimate it from Hδ. For objects
erative method. The method requires knowledge of the with [O ii] detection we add the contribution of singly
electron density, ne , which we cannot derive directly for ionized Oxygen to the total Oxygen abundance by esti-
(most of) the objects in our sample. In our derivation we mating the temperature following Campbell et al. (1986)
assume ne = 300 cm−3 . We have estimated the effect of and otherwise following Izotov et al. (2006). These cor-
the density uncertainty by assuming also ne = 100, 1000 rection are always less than 0.17 and usually less than
to assess the range of variability. Usually this contribu- 0.1 in log (O/H) +12.
tion to the overall uncertainty is small. The calculations We derive Oxygen metallicity for objects with a 3-σ
require a well measured Hβ. For some objects the line is or higher S/N for [O iii]λ4363 . For the remaining objects
 Metallicity of Galaxies at 3 < z < 9.5 5

Figure 2. NIRSpec MSA spectra of our primary sample galaxies, zoomed in the wavelength region of the Hγ and [O iii]λ4363
lines. The modelled gaussian is shown when the line is detected at > 3-σ (red solid lines). The detection significance for each
line is shown in the label. Note that sources from Nakajima et al. (2023) are not shown, as we adopt their published flux
measurements.
we adopt a value of [O iii]λ4363 identical to three times
the error as an upper limit, which gives us a lower limit
to the metallicity.

3.3. Spectral Energy Distribution Analysis

We infer the spectral energy distribution (SED) of
our sample galaxies through SED fitting using the
JWST+HST photometric data. We use the SED fit-
ting code gsf (ver1.85; Morishita et al. 2019), which
allows flexible determinations of the SED by adopting
binned star formation histories (often referred to as non-
parametric). We follow the same procedure as described
in Stiavelli et al. (2023), but this time without fitting
spectroscopic data simultaneously as continuum fluxes
of a large fraction of the sources are too faint to be use-
ful.
The star-formation rate of individual galaxies is cal-
culated with the rest-frame UV luminosity (∼ 1600 Å)
using the determined SED model. The UV luminosity
is corrected for dust attenuation using the βUV slope, Figure 3. Our samples in the stellar mass–star formation
rate plane. Symbols are color-coded by redshift. Different
which is measured by using the posterior SED, as in
symbols represent references (circles for GTO1199, squares
Smit et al. (2016): JADES, diamonds Nakajima et al. 2023). The linear relation
derived for UV-selected galaxies at 5 < z < 10 (Morishita
A1600 = 4.43 + 1.99 βUV (1)
et al. 2023, solid line) is shown for comparison. The broad
line object, 20024, is marked separately (red star). Those ex-
The attenuation corrected UV luminosity is then con-
cluded from the mass-metallicity relation analysis (Sec. 4.3)
verted to SFR via the relation in Kennicutt (1998): are shown as open circles.
SFR [M⊙ yr−1 ] = 1.4 × 10−28 LUV [erg s−1 Hz−1 ]. (2)

4. RESULTS
6 Morishita, Stiavelli, et al.

8.5 This study (3<z<9.5)

8.0

12+log(O/H)
7.5
AM13 (z=0)
Berg+22 (z=0)
Curti+20 (z=0)
Li+23 (z=2&3)
7.0 Sanders+21 (z=2&3)
Nakajima+23 (z=4-10)
Heintz+23 (z=7-10)
7 8 9 10
Stellar Mass [log M ¯ ]
Figure 4. (Left): Our sample galaxies in in the stellar mass–metallicity plane. Symbols and colors are the same as in Fig. 3.
Those with an upper limit in [O iii]λ4363 are shown as a lower limit in log (O/H). The best-fit slope to our sample is shown
(orange solid line, with hatched regions for the 3-σ range). (Right): Same as in the left panel, but with the mass-metallicity
relation slopes determined at z ∼ 0 (Andrews & Martini 2013; Curti et al. 2020; Berg et al. 2022), z ∼ 2–3 (Sanders et al. 2021;
Li et al. 2023), and high-z (Heintz et al. 2023; Nakajima et al. 2023) in the literature.

4.1. Spectroscopic Sample from GTO 1199 has broad features in Hydrogen recombination lines.
In Fig. 1, we present the cutout images of 28 sources This source, however, shows no significant detection of
confirmed to z > 3 in the GTO1199 observations. The [O iii]λ4363 , and thus it is not included in our primary
fluxes are reported in Table 1. We note that redshift of metallicity analysis below. The derived physical prop-
all sources is measured by multiple lines, in most cases erties of our samples are reported in Table 2.
including the [O iii]-doublet.
Among the confirmed sources, 9 have > ∼ 3 σ detection
in [O iii]λ4363 , with ID 3 (JD1) holding the highest red- 4.2. Additional [O iii]λ4363 Samples
shift, z = 9.11 (Stiavelli et al. 2023). The [O iii]λ4363 In addition, we explore public datasets to expand the
line of each source is shown in Fig. 2. Two sources do sample size of [O iii]λ4363 sources. We include 6 galaxies
not have wavelength coverage for [O iii]λ4363 , due to the at 3 < z < 9.5 from the JADES program (Hainline
wavelength range falling in a detector gap or not being et al. 2023; Bunker et al. 2023; Eisenstein et al. 2023)
covered by the corresponding grating. observed in their medium resolution spectroscopy and
We note the observed high [O iii]λ4363 /[O iii]λλ4959+5007 10 at 4 < z < 8.6 from Nakajima et al. (2023).
ratio for our samples, with the median value of ∼ 0.027 For the JADES sources, we start with the public
(also Sec. 5.4). This is considerably higher compared 1-dimensional spectra and repeat the same line fit-
to those reported in lower-z studies — for comparison, ting analysis as for the GTO1199 sources. The mea-
we find ∼ 0.01 in the Sloan sample for the mass range sured fluxes are reported in Table 1. Four of the
of 108 < M∗ /M⊙ < 109 in the MPA-JHU catalog 2 ; JADES sources were previously reported in Laseter et al.
Sanders et al. (2016) reported 0.013 ± 0.003 for a galaxy (2023b). One of the Laseter et al. (2023b) sources has a
at z = 3.08. solid detection of [O iii]λ4363 , but has no spectroscopic
Overall, our sample galaxies are characterized by nar- coverage for the [O iii]λλ4959+5007 -doublet (NIRSpec ID
row lines. Their location on the stellar mass-SFR plane 00008083); as such, this source was excluded from our
is shown in Fig. 3. The exception is ID 20024, which metallicity analysis.
For the sources reported in Nakajima et al. (2023, from
2 https://www.sdss4.org/dr17/spectro/galaxy mpajhu/. public JWST fields, ERO (Pontoppidan et al. 2022),
GLASS (Treu et al. 2023), and CEERS (Finkelstein
 Metallicity of Galaxies at 3 < z < 9.5 7

1.0 Low-z sample mean:-0.06 High-z sample mean:-0.04

B(z) = βz + αz log(1 + z), (4)
std: 0.19 std: 0.31 where N (0, σlogMZR ) represents a normal distribution
with σlogMZR being the intrinsic scatter around the rela-

0.5
tionship. For the calculation of posterior distributions,
we employ the Markov chain Monte Carlo method by
using emcee (Foreman-Mackey et al. 2014). The likeli-
∆ log(O/H)

hood function is defined as:

0.0
!
X
2 2


lnL = −0.5 (∆i + ln 2πσi ) − 2χlim , (5)
i

where i runs through detected sources (i.e. S/N

0.5 [O iii]λ4363 > 3). ∆i is the normalized residual:

∆i = (log (O/H)model − log (O/H)i )/σi . (6)

9 8 7
sSFR [yr −1 ]
1.0
The term of χlim adds the likelihood of lower-limit
0.6 0.8 1.0 sources (also, Sawicki 2012):
log(1 + z) X
r
π log (O/H)model − σj

!
χlim = ln σj 1 + erf( √ )
j
2 2σj
Figure 5. Redshift distribution of ∆ log (O/H), metallicity
offset from the determined mass-metallicity relation in Fig. 4. (7)
The mean and standard deviation are shown for each of low- where j runs through lower-limit sources, σj lower limit
and high-redshift sub-samples, split at the median redshift on log (O/H), and erf the error function.
(vertical dotted line). Symbols are color-coded by specific We adopt the median mass of the sample as the pivot
star formation rate. Note that the lower limit samples are mass, M0 . From the regression analysis, we find the
not shown. slope α = 0.27 ± 0.10, similar to Nakajima et al. (2023).
We find a slightly smaller offset (∼ −0.15 dex) from
et al. 2023)), we adopt their published measurements the Nakajima et al. (2023)’s relation, but the scatter
for line fluxes and magnification. around our determined relationship is considerably large
To guarantee a uniform analysis, we derive oxy- (σlog MZR = 0.26 ± 0.05).
gen abundances using the same method presented in The derived intercept is ∼ 0.4–0.7 dex lower than that
Sec. 4.1, and stellar mass and star formation rate as in the local relation at the pivot mass, log M0 = 8.8 (e.g.,
in Sec. 3.3. The derived physical properties are sum- Andrews & Martini 2013; Curti et al. 2020; Berg et al.
marized in Table 2. We find overall good agreement of 2022, among those adopt the direct [O iii]λ4363 method).
the derived abundances with the previous studies. This This indicates evolution of the relation, from z ∼ 3–9 to
provides an independent verification of our method. It z ∼ 0, supporting the previous conclusion reported in
also enables a fair internally-consistent comparison with various studies at similar redshifts (e.g., Nakajima et al.
the objects from those studies. 2023; Heintz et al. 2023).
On the other hand, redshift evolution of metallicity
4.3. Stellar Mass – Metallicity Distribution of among our sample galaxies is not observed. The rele-
Galaxies at 3 < z < 9.5 vant parameter, αz , is consistent with no evolution (i.e.
In the left panel of Fig. 4, we show the distribution of αz = 0.0±0.5) in our regression analysis. Indeed, we find
our sample galaxies in the stellar mass-metallicity plane. that our samples are distributed in a fairly wide range of
Despite a wide redshift range, our samples overall dis- metallicity regardless of redshift. For example, the two
tribute around the linear relation reported in the liter- highest redshift sources (JD1 and 10058975) have dis-
ature, derived for z > 3 galaxies (Heintz et al. 2023; tinct values (log (O/H) +12 = 7.81 and 7.08). ERO 6355
Nakajima et al. 2023). at z = 7.655 is one of the most enriched galaxies in the
We investigate the distribution of our samples by using combined sample, with log (O/H) +12 = 8.29. We in-
the following formulae: vestigate the observed scatter in Sec. 5.2.
It has been known that at lower redshifts mass-
12 + log (O/H) = α log(M∗ /M0 ) metallicity relations have dependency on a third param-
(3) eter, such as SFR (e.g., Andrews & Martini 2013) and
+ B(z) + N (0, σlogMZR ),
8 Morishita, Stiavelli, et al.

gravitational potential (e.g., Sánchez-Menguiano et al. relation is also reported in other numerical work (e.g.,
2023). To understand the observed scatter, we repeat Ma et al. 2016; Langan et al. 2020).
the regression analysis by including additional physical
quantities to the x-axis, namely SFR, sSFR, rest-frame 5.2. On the scatter around the Mass-Metallicity
optical size, and gravitational potential. However, none relation
of those quantities offers significant reduction in the
We have found a considerably large scatter around
observed scatter. For example, with M∗ /SFR0.66 (i.e.
the determined relation. Our regression analysis above
the fundamental-metallicity; Andrews & Martini 2013;
concluded that the scatter cannot be attributed to a
Curti et al. 2020; Li et al. 2023), we only find a compa-
third quantity, including the redshift dependence of the
rable intrinsic scatter, σlog MZR = 0.28.
relation. For comparison, at z ∼ 0 the scatter is found
to be ∼ 0.1 around the fundamental mass-metallicity
5. DISCUSSION relation at the same mass range (Curti et al. 2020).
This suggests that metal enrichment process occurs
5.1. On the presence of the Mass-Metallicity relation rapidly, but also susceptible to mechanisms that di-
We have observed correlation between stellar mass and lute the enrichment over a similar time scale. Such
metallicity for our sample galaxies. Despite the large inverse processes are expected to be less significant at
scatter, the presence of a positive slope indicates that lower redshift, where a stronger gravitational potential
stellar mass formation plays a critical role in enrichment aids retaining the enriched gas within the galaxy system
process already at these redshifts. against feedback processes. We note that in the stellar
The observed slope of our galaxies is offset from the mass range of interest here, the halo mass of the system
local relation for ∼ 0.4–0.7 dex, depending on the local increases as redshift decreases until z ∼ 1 (and then it
reference. A similar offset is reported in other studies at decreases toward z ∼ 0; e.g., Behroozi et al. 2019).
similar redshifts (e.g., Heintz et al. 2023), and at z ∼ 3 Our data supports the interpretation. In Fig. 5, we
(e.g., Sanders et al. 2020; Curti et al. 2023b), the lower show the deviation of metallicity from the determined
bound of our redshift range. In contrast, we find no relation. Clearly, the amplitude of the scatter increases
clear evidence of redshift evolution among our sample. at increasing redshift. The standard deviation for the
We note that the time difference is much smaller at 3 < lower-half and higher-half redshift samples are 0.19 and
z < 9 (∼ 1.5 Gyr; whereas ∼ 10 Gyr from z ∼ 3 to 0.31, respectively. Notably, high-z galaxies are intense
0), while some studies reported the presence of weak in star formation. In Morishita et al. (2023), we re-
evolution using a sample of calibrated metallicity (e.g., ported that star formation rate surface density, ΣSFR ,
Curti et al. 2023b). of z > 5 galaxies is ∼ 10–100 times higher than z ∼ 2
One possible interpretation is that what regulates samples, driven both by compact morphology and high
metallicity changes around z ∼ 2–3, from bursty to more star formation rate. The decrease of scatter amplitude
continuous star formation (e.g., Asada et al. 2024). At can be explained such that interplay between the poten-
higher z, the relevant “clock” is driven by mass, rather tial well and feedback process becomes more constrained
than the specific redshift, and galaxies will enrich un- at decreasing redshift. We remind the absence of clear
til perhaps an outflow quenches star formation or ejects correlation between ∆ log (O/H) and sSFR in our sam-
metals. At decreasing redshift, the star formation his- ple. However, this could simply be due to the fact that
tory becomes more important, as galaxies begin retain- chemical enrichment may not necessarily proceed in the
ing metals or perhaps accreting metal rich gas so their same time scale for star formation at these redshifts
metallicity can be higher for a given mass. It remains (<∼ 200 Myr; Faisst & Morishita, in prep).
surprising that slopes remain similar across the cosmic Alternatively, an increased fraction of merger-driven
time. evolution would also explain the observed scatter trend.
The absence of redshift evolution at z ≳ 5 is reported As redshift increases, so does the merger rate (Fakhouri
in numerical simulations. Ucci et al. (2023) reported a & Ma 2008), and thus galaxy assembly is progressively
decrease of metallicity for log M∗ /M⊙ ∼ 7 galaxy sys- more driven by mergers than smooth accretion. As a
tems at decreasing redshift. They attribute this to var- result, the system’s chemical enrichment is more affected
ious factors, including the differential impact of feed- by independent building blocks rather than by smooth
back in different halo mass systems at low and high red- accretion of material and self-recycling.
shifts. The impact by feedback becomes smaller at a In our data, we did not confirm the presence of the
larger mass system, making the “observed” evolution in- third parameter that reduces the scatter. Simulations
significant. Similarly weak/no redshift evolution of the find a different conclusion. Ucci et al. (2023) reported a
 Metallicity of Galaxies at 3 < z < 9.5 9

1.0 1.0
dZ/dkpc
=-0.2
=0.2
dZ/dkpc
0.5 0.5
∆ log(O/H)

∆ log(O/H)
0.0 0.0

0.5 0.5

0.6 0.6
log(1 + z) log(1 + z)
1.00.0 0.2 0.4 0.6 1.0 0.0 0.2 0.4 0.6 0.8
Distance to MSA position [kpc] Fractional Flux Loss by MSA
Figure 6. Same as Fig. 5, but with different quantities for the x-axis. (Left): Distance from the MSA shutter position to
the light-weighted center of the target galaxy. The distribution of our samples does not reveal a clear trend with the offset
amplitude. The deviation from the mean relation cannot be explained by the offset, even under the presence of steep metallicity
gradient (0.2 dex/kpc and −0.2 dex/kpc; arrows). (Right): Fractional flux loss of MSA (i.e. flux that falls out of the MSA slit
over the total flux). The fraction flux loss estimated in F444W is shown here.

dependence of the relationship on specific star formation no impact to mass-metallicity relation if the metallicity
rate at z > 5, in a way that a lower metallicity galaxy measured at the position represents the entire system.
has a higher sSFR. Their observed trend is interpreted However, this is not necessarily the case. Studies have
as the strong star formation could blow the enriched gas found a pronounced gradient in gas-phase metallicity in
away from the system, while pristine gas may still feed active galaxies, up to z ∼ 3 (e.g., Belfiore et al. 2017;
the galaxy system. We do not observe this trend in our Bresolin 2019; Wang et al. 2020).
sample. A larger sample may be needed to confirm the We investigate this by examining the deviation in
predictions. metallicity from the mean relationship in relation to
the offset distance (Fig. 6, left). Our analysis reveals
5.3. Potential Systematic Uncertainties no clear indication that the offset amplitude is the pri-
mary cause for the observed scatter. In fact, the ob-
Slit-spectroscopy observations have two major limita-
served scatter amplitude surpasses what can be ac-
tions, namely the position of slit regarding the galaxy
counted for by an extreme gradient reported in the lit-
center and aperture loss due to the limited slit size.
erature (±0.2 dex/kpc).
Since the mass-metallicity relation concerns the global
We then investigate the impact by aperture loss. We
metallicity of the system, those systematic uncertain-
here measure fractional aperture loss (∆fMSA ), i.e. flux
ties need to be addressed. Those effects are particularly
that falls out of the slit over the total flux of the tar-
the case for JWST for its design of MSA masks, its re-
get, in the F444W image. The right panel of Fig-
solving power, and its relatively small size of the slit
ure 6 shows the distribution. This quantity ranges from
(0.′′ 2 × 0.′′ 46). Despite its critical importance, the litera-
∼ 0 (i.e. most flux is within the slit) to 0.8 (20 %
ture has not yet engaged in a comprehensive discussion
is within the slit) for our sample, with the median
on these systematic issues.
value 0.4. Interestingly, three sources that are least
In our observations, we find that several targets show
affected by the loss (∆fMSA < 0.25) have metallicity
a small offset in the MSA slit position from its light
below the mean relation. On the other hand, at the
center (Fig. 1). This is partially due to our selection
high-loss end (∆fMSA > ∼ 0.5), four (3, 30055, ERO 6355,
of “Entire Open Shutter Area” configuration in APT,
GLASS 10021) out of six are located above the mean
chosen to maximize multiplexity, leading to an offset of
metallicity relation.
∼ 0.25 arcsec at maximum. Such an offset would have
10 Morishita, Stiavelli, et al.

Through the robust determination of oxygen abun-

dance, our data reveal the distribution of early galax-
ies in the mass-metallicity plane. Despite probing
a wide range of both redshift and stellar mass, our
galaxies turn to be moderately enriched in metallic-
ity, log (O/H) +12 > ∼ 7. For comparison, the protoype
of low-metallicity galaxy in the local Universe, I Zw 18,
has log (O/H) +12 = 7.17 (Izotov et al. 1999). The
lowest mass galaxy among our mass-metallicity relation
sample, with log M∗ /M⊙ = 7.3 (ID 18322), has oxy-
gen abundance of log (O/H) +12 ∼ 7.3. This points us
to the need for further exploration toward lower mass,
log M∗ /M⊙ ≪ 7, even at such an early time of the uni-
verse, to find a less enriched galaxy system.
In Fig. 7, we show the stacked spectrum of our sam-
ples. Given that the temperature becomes high in low-
metal ISM, the line ratio [O iii]λ4363 /[O iii]λλ4959+5007
also becomes larger, leaving us a chance to apply the
direct-Te method down to a considerably low-metallicity
Figure 7. (Top panels): Stacked spectrum of the sample regime. Shown along with our stacked spectrum are a
galaxies that have [O iii]λ4363 detected (red lines). Individual few predictions for the lines. Depending on the elec-
spectra are shown in the background (gray lines). All spectra tron temperature, the [O iii]λ4363 line can be detected
were subtracted continuum and normalized by the integrated down to log (O/H)+12 ≈ 6 for the sensitivity enabled
Hβ line flux before being stacked. (Bottom panels): Same by JWST.
stacked spectra are shown (gray), along with three metal- Recently, Vanzella et al. (2023) reported an upper
poor models of for different temperatures (1.5, 3, and 5 ×
limit of log (O/H)+12 < 6.3 on a low-mass star com-
104 K, with Z = 10−3 Z⊙ ). 4
plex ( <
∼ 10 M⊙ ), through calibrated strong line meth-
ods. The system is a part of lensed arcs strongly ampli-
Despite the observed signal being statistically weak, fied by the foreground galaxy cluster. Sub-scale observa-
this reveals an important implication. Since the MSA tions, with an assist of strong gravitational lensing (also,
slit position is, in default, configured to be near the light Adamo et al. 2024), is a path toward the identification
center of the target, the inferred metallicity would be bi- of pristine systems (also, e.g., Bovill et al. 2024).
ased to the central part, where metallicity is likely most Lastly, we note that the MSA slits of our NIRSpec
enriched. The effect would become significant when observations were placed near the galaxy light-center,
there exists a negative gradient and the aperture loss as the default option. This is presumably the case for
is large (i.e. large-sized target), leading to an overesti- many other MSA programs. Given that the galaxy cen-
mated metallicity. In other words, the four sources of ter is likely the place where metal enrichment is more
high ∆fMSA may be observed to be metal rich because advanced, spectroscopy of off-centered regions may also
their low-metallicity part might have fallen out of the help such endeavor.
slit. Qualitatively, the impact should remain < ∼ 0.2 dex
reduction in metallicity at most for the typical apparent
size of our galaxies assuming a gradient of −0.2 dex/kpc. 6. SUMMARY
A more realistic estimate could be achieved upon the In this study, we have reported 28 galaxies at 3 <
consideration of, e.g., the actual light profile and metal- z < 9.1 in the sightline of the cluster of galaxies
licity gradient, which is beyond the scope of this study. MACS J1149. We detected the auroral [O iii]λ4363 line
We also repeat the regression analysis but this time in 9 sources, which allows robust oxygen abundance de-
with 24 galaxies with those fractional flux loss < 0.5. termination via the direct-Te method. By combining
We do not see decrease in the scatter (σlog MZR = 0.27; with 16 sources from publicly available data, we found
Table 3). We find positive redshift evolution of the in- that the mass-metallicity relation exists at the redshift
tercept, but the significance remains too tentative to range of interest, with ∼ 0.4–0.7 dex offset down be-
confirm the trend (αz = −0.3 ± 0.7). low the local relation. We found that the scatter of
our sources around the relation is not small. Inter-
5.4. Toward the identification of Pop-III galaxies
estingly, the deviation becomes smaller as redshift de-
 Metallicity of Galaxies at 3 < z < 9.5 11

creases, which is indirect evidence that the interplay be-

tween gravitational potential well and feedback intensity
converge. We find no clear evidence of redshift evolution
in our sample. While seen in previous lower-z observa-
tions and predicted in simulations, we did not find the
third parameter that can effectively reduce the observed ACKNOWLEDGEMENTS
scatter. Despite reaching to a considerably low-mass Some/all of the data presented in this paper were
with an assist of strong lensing, our samples exhibit a obtained from the Mikulski Archive for Space Tele-
relatively enriched ISM, indicating that the exploration scopes (MAST) at the Space Telescope Science Insti-
of lower-mass galaxies or sub-galactic scale is needed for tute. The specific observations analyzed can be accessed
the search of less enriched populations, even at the red- via 10.17909/q8cd-2q22. We acknowledge support for
shift range explored in this study. this work under NASA grant 80NSSC22K1294. C.G.,
We investigated potential systematic impacts arising P.R., and S.S. acknowledge support through grants
from the constraints of JWST’s MSA observations. It MIUR2017 WSCC32 and MIUR2020 SKSTHZ. KB and
is important to bear in mind that the metallicity mea- MT are supported by the Australian Research Coun-
sured using MSA slits may not necessarily represent the cil Centre of Excellence for All Sky Astrophysics in
“global” metallicity, which is the quantity of primary 3 Dimensions (ASTRO 3D), through project number
concern in mass-metallicity relation analyses. This ef- CE170100013.
fect is likely to be more pronounced at lower redshifts Software: Astropy (Astropy Collaboration et al. 2013,
or in lensing fields, where the anticipated flux loss is 2018, 2022), bbpn (Morishita 2023), EAzY (Brammer
greater. et al. 2008), EMCEE (Foreman-Mackey et al. 2013),
gsf (Morishita et al. 2019), numpy (Harris et al. 2020),
python-fsps (Foreman-Mackey et al. 2014).

REFERENCES
Adamo, A., Bradley, L. D., Vanzella, E., et al. 2024, arXiv Brammer, G. B., van Dokkum, P. G., & Coppi, P. 2008,
e-prints, arXiv:2401.03224, ApJ, 686, 1503, doi: 10.1086/591786
doi: 10.48550/arXiv.2401.03224 Bresolin, F. 2019, MNRAS, 488, 3826,
Andrews, B. H., & Martini, P. 2013, ApJ, 765, 140, doi: 10.1093/mnras/stz1947
doi: 10.1088/0004-637X/765/2/140 Bunker, A. J., Cameron, A. J., Curtis-Lake, E., et al. 2023,
Asada, Y., Sawicki, M., Abraham, R., et al. 2024, MNRAS, arXiv e-prints, arXiv:2306.02467,
527, 11372, doi: 10.1093/mnras/stad3902 doi: 10.48550/arXiv.2306.02467
Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., Calzetti, D. 1997, AJ, 113, 162, doi: 10.1086/118242
et al. 2013, A&A, 558, A33, Campbell, A., Terlevich, R., & Melnick, J. 1986, MNRAS,
223, 811, doi: 10.1093/mnras/223.4.811
doi: 10.1051/0004-6361/201322068
Chabrier, G. 2003, PASP, 115, 763, doi: 10.1086/376392
Astropy Collaboration, Price-Whelan, A. M., Sipőcz, B. M.,
Curti, M., Mannucci, F., Cresci, G., & Maiolino, R. 2020,
et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4f
MNRAS, 491, 944, doi: 10.1093/mnras/stz2910
Astropy Collaboration, Price-Whelan, A. M., Lim, P. L.,
Curti, M., D’Eugenio, F., Carniani, S., et al. 2023a,
et al. 2022, ApJ, 935, 167, doi: 10.3847/1538-4357/ac7c74
MNRAS, 518, 425, doi: 10.1093/mnras/stac2737
Behroozi, P., Wechsler, R. H., Hearin, A. P., & Conroy, C.
Curti, M., Maiolino, R., Curtis-Lake, E., et al. 2023b, arXiv
2019, MNRAS, 488, 3143, doi: 10.1093/mnras/stz1182
e-prints, arXiv:2304.08516,
Belfiore, F., Maiolino, R., Tremonti, C., et al. 2017, doi: 10.48550/arXiv.2304.08516
MNRAS, 469, 151, doi: 10.1093/mnras/stx789 Eisenstein, D. J., Johnson, B. D., Robertson, B., et al.
Berg, D. A., James, B. L., King, T., et al. 2022, ApJS, 261, 2023, arXiv e-prints, arXiv:2310.12340,
31, doi: 10.3847/1538-4365/ac6c03 doi: 10.48550/arXiv.2310.12340
Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 Fakhouri, O., & Ma, C.-P. 2008, MNRAS, 386, 577,
Bovill, M. S., Stiavelli, M., Wiggins, A. I., Ricotti, M., & doi: 10.1111/j.1365-2966.2008.13075.x
Trenti, M. 2024, ApJ, 962, 49, Finkelstein, S. L., Bagley, M. B., Ferguson, H. C., et al.
doi: 10.3847/1538-4357/ad148a 2023, ApJL, 946, L13, doi: 10.3847/2041-8213/acade4
 12 Morishita, Stiavelli, et al.

Table 1. Rest-frame optical line flux measurements of 3 < z < 9.5 galaxies from the GTO 1199 program.

ID [OII]3727 [NeIII]3869 [NeIII]3968 Hδ4102 Hγ4340 [OIII]4363 Hβ4861 [OIII]4959+5007 [NII]6548+6583 Hα6563

GTO 1199 Sample

30010 – – – – 34.5 ± 15.0 < 11.7 63.2 ± 12.5 330.3 ± 54.7 < 71.6 147.6 ± 59.1
30055 – – – 47.5 ± 28.2 79.3 ± 17.9 18.8 ± 16.9 118.4 ± 16.1 1358.7 ± 102.1 < 19.3 438.3 ± 20.2
20028 – – 63.0 ± 19.2 59.6 ± 19.0 123.6 ± 13.6 26.4 ± 9.8 267.1 ± 20.5 2061.2 ± 90.8 85.1 ± 32.9 954.4 ± 26.2
20036 – < 6.5 < 4.6 3.7 ± 3.5 < 2.8 < 2.6 1.5 ± 1.1 12.3 ± 5.2 < 1.9 4.7 ± 1.1
150742 92.1 ± 31.2 < 19.7 < 13.2 < 11.4 < 9.3 < 14.4 31.8 ± 11.4 155.5 ± 49.7 79.8 ± 26.1 219.4 ± 12.5
20018 < 1.4 < 0.8 < 0.9 < 0.8 < 0.8 < 0.9 – 12.5 ± 2.6 < 3.4 126.2 ± 7.7
20014 – – – – – – – – < 1.4 78.5 ± 4.1
20024 104.3 ± 19.1 23.8 ± 19.4 < 27.2 47.6 ± 24.7 41.3 ± 30.2 < 20.1 183.3 ± 30.1 345.3 ± 83.4 436.8 ± 75.7 752.6 ± 37.3
20044 < 48.3 67.9 ± 38.8 < 85.7 60.7 ± 24.8 127.3 ± 38.7 < 24.2 230.5 ± 21.8 1067.2 ± 84.2 – –
30093 100.0 ± 35.1 17.3 ± 17.1 < 29.7 < 13.4 14.5 ± 11.6 < 12.7 47.6 ± 15.5 332.7 ± 67.7 < 13.5 160.7 ± 28.1
150880 88.1 ± 12.3 52.0 ± 8.5 26.0 ± 9.4 24.6 ± 7.7 34.4 ± 6.8 11.9 ± 6.2 87.6 ± 8.2 898.7 ± 47.5 23.3 ± 11.0 286.9 ± 9.0
320108 495.6 ± 19.6 137.4 ± 12.0 74.4 ± 10.3 – 128.9 ± 5.6 13.3 ± 4.5 321.8 ± 11.9 2555.7 ± 61.7 – –
80002 < 0.9 1.7 ± 1.1 < 1.5 1.9 ± 1.0 1.3 ± 0.5 < 0.5 – 12.4 ± 3.0 – –
320002 8.8 ± 2.0 4.1 ± 1.6 3.5 ± 1.9 2.5 ± 1.9 4.0 ± 1.4 2.4 ± 1.9 9.5 ± 3.1 78.0 ± 15.3 < 0.7 37.6 ± 2.7
320005 82.6 ± 8.3 18.7 ± 5.2 8.1 ± 5.0 6.0 ± 3.9 16.7 ± 3.5 < 3.7 42.5 ± 5.6 363.0 ± 28.2 – –
150903 21.2 ± 4.8 12.6 ± 3.9 5.6 ± 4.1 4.5 ± 2.2 12.0 ± 2.5 3.5 ± 2.1 17.7 ± 4.2 192.1 ± 18.5 – –
60003 105.2 ± 26.7 30.6 ± 15.2 < 12.5 27.2 ± 19.3 19.0 ± 12.8 < 12.6 60.4 ± 30.3 254.9 ± 83.9 < 7.6 123.1 ± 32.6
10007 < 0.9 < 0.8 < 0.8 < 1.7 < 1.2 < 27.3 18.2 ± 12.6 53.8 ± 30.6 < 9.2 48.4 ± 9.2
60000 43.2 ± 6.4 12.8 ± 5.0 < 4.9 < 2.7 < 18.3 < 33.0 24.5 ± 6.9 190.0 ± 30.2 < 10.6 83.4 ± 7.7
60002 10.3 ± 6.6 < 5.6 < 1.9 – < 6.7 < 2.5 < 0.5 < 0.5 < 0.7 < 0.6
60001 133.4 ± 12.1 28.1 ± 6.9 < 5.7 11.9 ± 5.4 27.5 ± 7.6 < 22.1 72.3 ± 8.1 569.5 ± 38.6 43.3 ± 19.3 287.9 ± 13.1
80001 < 0.2 < 0.3 < 0.6 < 0.5 < 0.5 < 0.4 < 0.3 < 0.3 < 0.5 0.8 ± 0.4
10001 3.5 ± 3.0 3.8 ± 2.3 < 1.7 4.8 ± 2.9 7.8 ± 4.0 < 2.6 – – – –
10010 – – < 11.3 2.9 ± 1.5 6.9 ± 1.7 2.1 ± 2.1 13.6 ± 1.2 115.0 ± 6.6 < 1.5 45.9 ± 2.6
10018 < 4.1 < 5.9 5.1 ± 2.5 1.7 ± 1.4 < 3.7 < 4.1 7.6 ± 2.0 44.3 ± 8.6 < 3.4 26.5 ± 3.3
2 37.2 ± 4.1 29.9 ± 3.7 18.0 ± 3.0 13.2 ± 1.6 25.1 ± 2.4 9.9 ± 1.8 44.4 ± 4.8 561.1 ± 23.9 – –
4 < 2.3 2.5 ± 1.6 2.9 ± 2.1 2.2 ± 1.7 2.9 ± 1.7 < 1.7 6.1 ± 1.6 33.6 ± 6.5 – –
3 9.2 ± 2.6 15.3 ± 2.2 7.7 ± 2.2 11.0 ± 3.0 17.2 ± 2.9 6.1 ± 2.5 35.1 ± 4.4 340.6 ± 22.6 – –

JADES sample

18322 < 6.5 10.8 ± 6.1 16.5 ± 12.9 < 5.6 7.0 ± 4.4 4.5 ± 3.9 15.6 ± 4.1 115.3 ± 15.2 < 2.7 47.1 ± 3.8
19519 21.1 ± 5.6 9.6 ± 4.5 11.7 ± 7.3 < 5.7 5.7 ± 3.0 4.0 ± 3.6 16.3 ± 3.0 118.5 ± 11.5 < 5.7 48.9 ± 3.5
18090 18.0 ± 7.4 9.8 ± 4.3 < 6.7 < 10.3 9.2 ± 4.6 5.1 ± 4.9 25.3 ± 6.2 214.8 ± 35.2 < 9.6 69.2 ± 15.1
22251 11.3 ± 4.5 7.5 ± 3.8 < 4.7 < 7.2 < 6.4 5.6 ± 4.4 15.5 ± 3.2 127.0 ± 13.5 < 3.9 52.3 ± 4.0
18846 5.7 ± 3.8 4.7 ± 2.8 < 4.8 4.0 ± 3.6 4.8 ± 2.4 2.2 ± 2.3 14.6 ± 2.7 109.3 ± 11.6 < 5.5 56.9 ± 4.5
10058975 3.9 ± 3.0 4.6 ± 2.1 2.4 ± 2.0 < 2.0 5.8 ± 3.0 3.1 ± 3.0 13.8 ± 4.1 80.3 ± 15.9 – –

Note— Fluxes are in units of 10−19 erg/s/cm2 . Flux errors and upper limits are 3-σ. Lines without the spectroscopic data coverage are marked with “–”. Texts
in purple are measurements for G140M, blue for G235M, and red for G395M (and G395H for the JADES sample). The presented fluxes are not corrected for
magnification. Fluxes for the sources from Nakajima et al. (2023) are not shown, as we adopt their published values.
 Metallicity of Galaxies at 3 < z < 9.5 13

Table 2.

ID R.A. Decl. mF444W z log(O/H) Te AVneb. log M∗ log SFR |µ| ∆dMSA ∆fMSA

degree degree mag +12 ×104 K mag M⊙ M⊙ /yr kpc

GTO 1199 Sample

30010 177.387224 22.404619 25.7 3.013 > 7.28 < 2.51 0.00 8.81+0.02
−0.02 0.7+0.1
−0.1 2.1+0.9
−1.6 0.50 0.65
30055 177.388125 22.415668 24.9 3.214 7.91+0.18
−0.14 1.58 0.98 8.68+0.04
−0.04 0.6+0.1
−0.1 5.8+5.9
−1.8 0.24 0.66
20028 177.410476 22.419310 23.6 3.345 7.79+0.07
−0.06 1.51 0.84 9.79+0.01
−0.04 1.8+0.1
−0.1 1.4+0.1
−0.1 0.57 0.77
20036 177.392896 22.412893 28.1 3.450 – – – 7.67+0.04
−0.05 −0.2+0.1
−0.1 3.9+0.1
−0.1 0.27 0.00
150742 177.429260 22.415850 23.4 3.678 – – – 10.43+0.03
−0.03 1.9+0.2
−0.2 1.2+0.1
−0.1 0.66 0.78
20018 177.399186 22.384109 26.4 3.685 > 6.13 < 3.21 0.00 8.56+0.07
−0.05 0.1+0.1
−0.1 1.9+0.1
−0.1 0.36 0.20
20014 177.391600 22.384264 27.2 3.693 – – – 8.26+0.12
−0.12 0.2+0.1
−0.1 1.6+0.1
−0.1 0.49 0.07
20024 177.424202 22.386257 – 3.696 > 7.15 < 2.75 1.37 10.23+0.06
−0.05 1.8+0.1
−0.1 2.1+0.1
−0.1 0.17 –
20044 177.383725 22.406203 25.4 3.732 > 7.49 < 1.88 0.00 9.11+0.03
−0.03 0.7+0.1
−0.1 2.9+0.1
−0.1 0.54 0.63
30093 177.415100 22.418217 – 4.194 > 7.82 < 2.87 0.63 9.18+0.04
−0.11 1.0+0.1
−0.1 1.3+0.1
−0.1 0.40 –
150880 177.403656 22.368355 26.6 4.247 8.00+0.10
−0.08 1.49 0.50 9.28+0.02
−0.04 0.9+0.1
−0.1 1.3+0.1
−0.1 0.29 0.41
320108 177.402573 22.384069 23.7 4.257 8.36+0.05
−0.05 1.09 1.16 10.13+0.01
−0.01 1.6+0.1
−0.1 2.2+0.1
−0.1 0.42 0.46
80002 177.394550 22.394321 28.4 4.495 > 7.50 < 2.81 0.00 7.55+0.34
−0.25 −0.6+0.2
−0.1 5.7+0.2
−0.4 0.19 0.34
320002 177.428253 22.409452 27.5 4.658 7.41+0.19
−0.16 2.59 1.24 8.68+0.08
−0.05 0.7+0.1
−0.1 1.3+0.1
−0.1 0.19 0.39
320005 177.422580 22.414459 – 4.658 > 8.35 < 1.41 1.29 9.60+0.02
−0.02 1.4+0.1
−0.1 1.3+0.1
−0.1 0.55 0.65
150903 177.424210 22.408874 26.4 4.659 7.90+0.11
−0.09 1.68 0.00 9.25+0.01
−0.01 0.9+0.1
−0.1 1.4+0.1
−0.1 0.16 0.34
60003 177.375739 22.399017 25.0 5.465 – – – 9.19+0.07
−0.05 1.1+0.1
−0.1 1.7+0.1
−0.1 0.59 0.55
10007 177.420044 22.388666 – 5.583 – – – 7.48+0.75
−0.36 −0.1+0.2
−0.2 3.3+0.1
−0.1 0.22 –
60000 177.378968 22.402419 24.4 5.585 – – – 9.63+0.06
−0.06 1.3+0.1
−0.1 2.1+0.1
−0.1 0.23 0.47
60002 177.382540 22.395407 25.2 5.594 – – – 9.23+0.08
−0.04 1.0+0.1
−0.1 2.2+0.1
−0.1 0.62 0.69
60001 177.412014 22.415776 24.0 5.686 > 7.69 < 3.17 1.26 10.37+0.02
−0.02 1.7+0.1
−0.1 1.5+0.1
−0.1 0.58 0.72
80001 177.395860 22.412365 28.8 5.974 – – – 6.35+0.66
−0.73 −1.3+0.2
−0.1 11.3+12.9
−3.8 0.33 0.09
10001 177.388351 22.382391 26.4 6.119 – – – 9.41+0.01
−0.01 0.8+0.1
−0.1 1.5+0.1
−0.1 0.38 0.38
10010 177.412949 22.418858 25.7 6.311 7.66+0.21
−0.16 1.78 0.62 9.59+0.02
−0.03 1.0+0.1
−0.1 1.4+0.1
−0.1 0.17 0.44
10018 177.414642 22.417057 27.8 6.372 – – – 8.73+0.03
−0.07 0.4+0.1
−0.1 1.4+0.1
−0.1 0.41 0.51
2 177.417709 22.417431 23.8 7.232 7.92+0.04
−0.04 1.64 0.00 9.46+0.02
−0.03 1.9+0.1
−0.1 1.4+0.1
−0.1 0.26 0.54
4 177.382950 22.412039 27.6 7.238 > 7.21 < 3.35 0.00 7.76+0.12
−0.12 −0.1+0.1
−0.1 3.0+0.2
−0.1 0.16 0.03
3 177.389908 22.412710 25.6 9.114 7.82+0.07
−0.07 1.66 0.00 7.77+0.04
−0.02 0.5+0.1
−0.1 12.5+12.0
−5.0 0.31 0.57

JADES Sample

18090 53.167180 −27.774615 28.2 4.769 7.64+0.21

−0.17 1.93 0.00 8.58+0.05
−0.11 0.4+0.3
−0.5 – 0.37 0.00
18322 53.131707 −27.774119 28.4 3.153 7.26+0.20
−0.15 2.78 0.43 7.29+0.23
−0.16 −0.3+0.3
−0.3 – 0.20 0.36
18846 53.134917 −27.772709 27.3 6.336 7.55+0.22
−0.17 1.97 2.70 8.41+0.08
−0.07 0.7+0.1
−0.1 – 0.47 0.35
19519 53.166086 −27.771262 27.1 3.604 7.42+0.21
−0.16 2.47 0.39 8.64+0.08
−0.10 0.6+0.1
−0.1 – 0.22 0.34
22251 53.154069 −27.766072 27.0 5.803 7.22+0.18
−0.09 3.23 1.39 7.90+0.22
−0.16 0.5+0.1
−0.1 – 0.41 0.40
10058975 53.112434 −27.774609 27.5 9.435 7.11+0.25
−0.15 2.98 1.89 8.36+0.32
−0.19 0.9+0.3
−0.1 – 0.27 0.00

Nakajima et al. (2023) Sample

ERO 04590 110.859818 −73.449127 26.0 8.496 7.09+0.10

−0.08 2.49 1.57 9.05+0.23
−0.21 1.1+0.2
−0.1 8.7 0.31 0.24
ERO 5144 110.839674 −73.445357 27.9 6.378 7.89+0.09
−0.08 1.56 0.00 8.75+0.13
−0.15 1.2+0.8
−0.7 3.2 0.14 0.26
ERO 6355 110.844917 −73.435043 24.4 7.665 8.29+0.10
−0.08 1.19 0.00 9.07+0.16
−0.12 1.5+0.1
−0.1 1.8 0.29 0.49
ERO 10612 110.834351 −73.435059 25.8 7.660 7.66+0.06
−0.05 1.95 0.34 8.54+0.17
−0.16 1.0+0.1
−0.1 1.9 0.46 0.43
GLASS 100003 3.604509 −30.380444 25.1 7.877 7.67+0.15
−0.10 2.00 0.00 9.12+0.04
−0.05 1.4+0.1
−0.1 1.3 0.43 0.40
GLASS 10021 3.608511 −30.418540 24.7 7.286 7.88+0.10
−0.08 1.68 0.00 8.69+0.15
−0.04 1.6+0.1
−0.1 1.7 0.47 0.56
GLASS 150029 3.577166 −30.422576 26.2 4.584 7.77+0.06
−0.05 1.67 0.00 9.31+0.02
−0.03 0.9+0.1
−0.1 1.5 0.35 0.26
GLASS 160133 3.580275 −30.424404 26.2 4.015 8.00+0.05
−0.04 1.44 0.05 8.91+0.02
−0.04 0.7+0.3
−0.1 1.5 0.12 0.40
CEERS 01027 214.882994 52.840416 – 7.820 7.75+0.17
−0.12 1.79 3.18 9.02+0.13
−0.33 1.0+0.1
−0.1 – 0.38 –
CEERS 01536 214.977227 52.940782 – 5.034 7.64+0.13
−0.09 2.15 0.00 8.85+1.19
−1.19 1.2+0.2
−0.2 – 0.18 –

Note— †: Only those have direct metallicity measurements via [O iii]λ4363 are shown (i.e. Fig. 2). For those with upper limit on the [O iii]λ4363 flux, we
derive a lower limit on log (O/H). Errors and upper limits are 1 σ. Electron temperature (Te ) in units of 104 K. Stellar mass and star formation rate are
corrected for magnification. ∆fMSA : Fractional flux loss due to MSA aperture, estimated in F444W.
 14 Morishita, Stiavelli, et al.

Table 3. Mass–Metallicity Relations of Galaxies at 3 < z < 9.5. Kennicutt, Robert C., J. 1998, ApJ, 498, 541,
Linear Regression Best-Fit Coefficients doi: 10.1086/305588
Kirby, E. N., Cohen, J. G., Guhathakurta, P., et al. 2013,
N log M0 / M⊙ α βz αz σlog MZR
ApJ, 779, 102, doi: 10.1088/0004-637X/779/2/102
Langan, I., Ceverino, D., & Finlator, K. 2020, MNRAS,
Full Sample 494, 1988, doi: 10.1093/mnras/staa880
34 8.8 0.27+0.09 7.73+0.41 0.01+0.50 0.26+0.06 Laseter, I. H., Maseda, M. V., Curti, M., et al. 2023a, arXiv
−0.09 −0.42 −0.50 −0.04
e-prints, arXiv:2306.03120,
Loss-fraction < 0.5 doi: 10.48550/arXiv.2306.03120
—. 2023b, arXiv e-prints, arXiv:2306.03120,
21 8.8 0.39+0.11
−0.11 7.99+0.55
−0.54 −0.33+0.66
−0.67 0.27+0.07
−0.06
doi: 10.48550/arXiv.2306.03120
Note— Best-fit coefficients for the single slope regression, log (O/H) = Li, M., Cai, Z., Bian, F., et al. 2023, ApJL, 955, L18,
α log(M∗ /M0 ) + B(z), where B(z) = βz + αz log(1 + z).
doi: 10.3847/2041-8213/acf470
Lotz, J. M., Koekemoer, A., Coe, D., et al. 2017, ApJ, 837,
97, doi: 10.3847/1538-4357/837/1/97
Ma, X., Hopkins, P. F., Faucher-Giguère, C.-A., et al. 2016,
Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman,
MNRAS, 456, 2140, doi: 10.1093/mnras/stv2659
J. 2013, PASP, 125, 306, doi: 10.1086/670067
Maiolino, R., Nagao, T., Grazian, A., et al. 2008, A&A,
Foreman-Mackey, D., Sick, J., & Johnson, B. 2014, 488, 463, doi: 10.1051/0004-6361:200809678
doi: 10.5281/zenodo.12157 Marino, R. A., Rosales-Ortega, F. F., Sánchez, S. F., et al.
Fukugita, M., Ichikawa, T., Gunn, J. E., et al. 1996, AJ, 2013, A&A, 559, A114,
111, 1748, doi: 10.1086/117915 doi: 10.1051/0004-6361/201321956
Gallazzi, A., Charlot, S., Brinchmann, J., White, S. D. M., Morishita, T. 2023, mtakahiro/bbpn: v1.3, v1.3, Zenodo,
& Tremonti, C. A. 2005, MNRAS, 362, 41, doi: 10.5281/zenodo.10067906
doi: 10.1111/j.1365-2966.2005.09321.x Morishita, T., & Stiavelli, M. 2023, ApJL, 946, L35,
Gburek, T., Siana, B., Alavi, A., et al. 2019, ApJ, 887, 168, doi: 10.3847/2041-8213/acbf50
doi: 10.3847/1538-4357/ab5713 Morishita, T., Abramson, L. E., Treu, T., et al. 2019, ApJ,
Grillo, C., Karman, W., Suyu, S. H., et al. 2016, ApJ, 822, 877, 141, doi: 10.3847/1538-4357/ab1d53
78, doi: 10.3847/0004-637X/822/2/78 Morishita, T., Roberts-Borsani, G., Treu, T., et al. 2023,
Hainline, K. N., Johnson, B. D., Robertson, B., et al. 2023, ApJL, 947, L24, doi: 10.3847/2041-8213/acb99e
arXiv e-prints, arXiv:2306.02468, Nakajima, K., Ouchi, M., Isobe, Y., et al. 2023, arXiv
doi: 10.48550/arXiv.2306.02468 e-prints, arXiv:2301.12825,
Harris, C. R., Millman, K. J., van der Walt, S. J., et al. doi: 10.48550/arXiv.2301.12825
2020, Nature, 585, 357, doi: 10.1038/s41586-020-2649-2 Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713,
doi: 10.1086/160817
Hashimoto, T., Laporte, N., Mawatari, K., et al. 2018,
Osterbrock, D. E., & Ferland, G. J. 2006, Astrophysics of
Nature, 557, 392, doi: 10.1038/s41586-018-0117-z
gaseous nebulae and active galactic nuclei
Heintz, K. E., Giménez-Arteaga, C., Fujimoto, S., et al.
Pettini, M., & Pagel, B. E. J. 2004, MNRAS, 348, L59,
2023, ApJL, 944, L30, doi: 10.3847/2041-8213/acb2cf
doi: 10.1111/j.1365-2966.2004.07591.x
Henry, A., Scarlata, C., Domı́nguez, A., et al. 2013, ApJL,
Pontoppidan, K., Blome, C., Braun, H., et al. 2022, arXiv
776, L27, doi: 10.1088/2041-8205/776/2/L27
e-prints, arXiv:2207.13067.
Hoag, A., Bradač, M., Huang, K., et al. 2019, ApJ, 878, 12,
https://arxiv.org/abs/2207.13067
doi: 10.3847/1538-4357/ab1de7 Postman, M., Coe, D., Benı́tez, N., et al. 2012, ApJS, 199,
Izotov, Y. I., Chaffee, F. H., Foltz, C. B., et al. 1999, ApJ, 25, doi: 10.1088/0067-0049/199/2/25
527, 757, doi: 10.1086/308119 Sánchez-Menguiano, L., Sánchez Almeida, J., Sánchez,
Izotov, Y. I., Stasińska, G., Meynet, G., Guseva, N. G., & S. F., & Muñoz-Tuñón, C. 2023, arXiv e-prints,
Thuan, T. X. 2006, A&A, 448, 955, arXiv:2312.02046, doi: 10.48550/arXiv.2312.02046
doi: 10.1051/0004-6361:20053763 Sanders, R. L., Shapley, A. E., Topping, M. W., Reddy,
Kelly, P. L., Diego, J. M., Rodney, S., et al. 2018, Nature N. A., & Brammer, G. B. 2023, arXiv e-prints,
Astronomy, 2, 334, doi: 10.1038/s41550-018-0430-3 arXiv:2303.08149, doi: 10.48550/arXiv.2303.08149
 Metallicity of Galaxies at 3 < z < 9.5 15

Sanders, R. L., Shapley, A. E., Kriek, M., et al. 2016, Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al.
ApJL, 825, L23, doi: 10.3847/2041-8205/825/2/L23 2004, ApJ, 613, 898, doi: 10.1086/423264
Sanders, R. L., Shapley, A. E., Reddy, N. A., et al. 2020, Treu, T., Schmidt, K. B., Brammer, G. B., et al. 2015,
MNRAS, 491, 1427, doi: 10.1093/mnras/stz3032 ApJ, 812, 114, doi: 10.1088/0004-637X/812/2/114
Sanders, R. L., Shapley, A. E., Jones, T., et al. 2021, ApJ, Treu, T., Brammer, G., Diego, J. M., et al. 2016, ApJ, 817,
914, 19, doi: 10.3847/1538-4357/abf4c1 60, doi: 10.3847/0004-637X/817/1/60
Sawicki, M. 2012, PASP, 124, 1208, doi: 10.1086/668636 Treu, T., Calabrò, A., Castellano, M., et al. 2023, ApJL,
Shipley, H. V., Lange-Vagle, D., Marchesini, D., et al. 2018, 942, L28, doi: 10.3847/2041-8213/ac9283
ApJS, 235, 14, doi: 10.3847/1538-4365/aaacce Ucci, G., Dayal, P., Hutter, A., et al. 2023, MNRAS, 518,
Smit, R., Bouwens, R. J., Labbé, I., et al. 2016, ApJ, 833, 3557, doi: 10.1093/mnras/stac2654
254, doi: 10.3847/1538-4357/833/2/254 Vanzella, E., Loiacono, F., Bergamini, P., et al. 2023, A&A,
Steidel, C. C., Strom, A. L., Pettini, M., et al. 2016, ApJ, 678, A173, doi: 10.1051/0004-6361/202346981
826, 159, doi: 10.3847/0004-637X/826/2/159 Wang, X., Jones, T. A., Treu, T., et al. 2020, ApJ, 900,
Steinhardt, C. L., Jauzac, M., Acebron, A., et al. 2020, 183, doi: 10.3847/1538-4357/abacce
ApJS, 247, 64, doi: 10.3847/1538-4365/ab75ed Zahid, H. J., Kewley, L. J., & Bresolin, F. 2011, ApJ, 730,
Stiavelli, M., Morishita, T., Chiaberge, M., et al. 2023, 137, doi: 10.1088/0004-637X/730/2/137
ApJL, 957, L18, doi: 10.3847/2041-8213/ad0159 Zheng, W., Zitrin, A., Infante, L., et al. 2017, ApJ, 836,
Storey, P. J., & Zeippen, C. J. 2000, MNRAS, 312, 813, 210, doi: 10.3847/1538-4357/aa5d55
doi: 10.1046/j.1365-8711.2000.03184.x