A Massive Quiescent Galaxy in a Group Environment at z = 4.53

We report on the spectroscopic confirmation of a massive quiescent galaxy at z spec = 4.53 in the COSMOS field. The object was first identified as a galaxy with suppressed star formation at z phot ∼ 4.65 from the COSMOS2020 catalog. The follow-up spectroscopy with Keck/MOSFIRE in the K band reveals faint [O ii] emission and the Balmer break, indicative of evolved stellar populations. We fit the spectral energy distribution using photometry and a spectrum to infer physical properties. The obtained stellar mass is high (M * ∼ 1010.8 M ⊙) and the current star formation rate is more than 1 dex below that of main-sequence galaxies at z = 4.5. Its star formation history suggests that this galaxy experienced rapid quenching from z ∼ 5. The galaxy is among the youngest quiescent galaxies confirmed so far at z spec > 3 with z form ∼ 5.2 (200 Myr ago), which is the epoch when 50% of the total stellar mass was formed. A unique aspect of the galaxy is that it is in an extremely dense region; there are four massive star-forming galaxies at 4.4 < z phot < 4.7 located within 150 physical kpc from the galaxy. Interestingly, three of them have virial radii that strongly overlap with that of the central quiescent galaxy (∼70 kpc), suggesting that the overdensity region is likely the highest-redshift candidate of a dense group with a spectroscopically confirmed quiescent galaxy at the center. The group provides us with a unique opportunity to gain insights into the role of the group environment in quenching at z ∼ 5, which corresponds to the formation epoch of massive elliptical galaxies in the local Universe.


INTRODUCTION
Massive elliptical galaxies in the local Universe are known to be dominated by old stellar populations with little cold gas and dust, exhibiting no active ongoing star formation (e.g., Young & Scoville 1991;Gallazzi et al. 2005;Bellstedt et al. 2020).Since these massive galaxies likely have played a major role in cosmic star formation activities in the early Universe, it is very important to study their formation mechanisms.The star formation histories (SFHs) of massive elliptical galaxies inferred from spectroscopic observations show that they experienced an intense burst of star formation in the early Universe and then rapidly quenched (stopped their star formation, Nelan et al. 2005;Thomas et al. 2005Thomas et al. , 2010;;Renzini 2006).However, the physical drivers of their starburst and subsequent quenching remain unclear, which is one of the most outstanding issues in the field of galaxy formation and evolution (Man & Belli 2018).
To elucidate these problems, a lot of efforts have been put into the search for progenitors of local elliptical galaxies at high redshifts and examine them closer to the time of quenching.Over the last two decades, massive quiescent galaxies with no significant ongoing star formation at z ∼ 2, which are likely the progenitors of local ellipticals, have been extensively studied (e.g., Glazebrook et al. 2004;Cimatti et al. 2004Cimatti et al. , 2008;;Daddi et al. 2005;Kriek et al. 2009Kriek et al. , 2019;;Whitaker et al. 2012;Toft et al. 2012;Onodera et al. 2012Onodera et al. , 2015;;Belli et al. 2017Belli et al. , 2019;;Morishita et al. 2019;Stockmann et al. 2020).These researches infer SFHs of the galaxies and confirm the evidence of starburst at z > 3 and short quenching timescale.These galaxies also have very large stellar masses (M * > 10 11 M ⊙ ) and compact sizes (R e < 2 kpc) compared to their counterparts in the local Universe (e.g., van Dokkum et al. 2008;Cassata et al. 2011;Newman et al. 2012;van der Wel et al. 2014;Lustig et al. 2021).
Thanks to the recent development of sensitive nearinfrared (NIR) spectrographs, massive quiescent galaxies at up to z ∼ 4 are spectroscopically confirmed (e.g., Glazebrook et al. 2017;Schreiber et al. 2018;Forrest et al. 2020a,b;Valentino et al. 2020;D'Eugenio et al. 2020D'Eugenio et al. , 2021;;Nanayakkara et al. 2022;Antwi-Danso et al. 2023;Tanaka et al. 2023).This epoch is consistent * JSPS Research Fellow (PD) with the formation epoch inferred from the local elliptical galaxies (Thomas et al. 2010), and the identification of massive quiescent galaxies at z > 4 is critical to understand their quenching mechanism.The number of sources confirmed at z > 4 is still relatively limited (Tanaka et al. 2019;Carnall et al. 2023a), but deep ground-based observations and new JWST spectroscopy are about to change the landscape.Also, recent cosmological simulations have difficulties in reproducing these galaxies (e.g., Schreiber et al. 2018;Guarnieri et al. 2019;Cecchi et al. 2019;Merlin et al. 2019;Valentino et al. 2020;Gould et al. 2023).Observational confirmations of massive quiescent galaxies at z > 4 are an important step to constrain the galaxy formation models.
In terms of the environment in which galaxies reside, it is known that elliptical galaxies dominate clusters of galaxies in the nearby Universe (e.g., Dressler 1980).This morphology-density relation shows that the surrounding environment may affect physical properties of galaxies.Actually, massive quiescent galaxies dominate high-density environments at z < 1 (e.g., Gómez et al. 2003;Peng et al. 2010;Kawinwanichakij et al. 2017).These environments are basic components of large-scale structures in the nearby Universe, and examining the formation process by observing the highredshift Universe is crucial.At z > 2, previous studies confirm "proto-clusters" which are over-density regions of galaxies and are considered progenitors of local clusters (e.g., Capak et al. 2011;Toshikawa et al. 2014;Lemaux et al. 2018;Oteo et al. 2018;Harikane et al. 2019;Hu et al. 2021;Brinch et al. 2024).However, these proto-clusters are often identified using starforming galaxies as a tracer, and there are only a few systems with (confirmed) quiescent galaxies (e.g., Kubo et al. 2021Kubo et al. , 2022;;Kalita et al. 2021;McConachie et al. 2022;Ito et al. 2023).To understand the role of the environment in galaxy quenching at high redshift, it is crucial to examine more (proto-)clusters and look for a sign of environmental effects in the act.
In this paper, we report the confirmation of a massive quiescent galaxy at z = 4.53 in the COSMOS field using the Keck/MOSFIRE spectrograph.The galaxy is located in a group environment, and we discuss a possible role of environment for quenching at this high redshift for the first time.This paper is structured as follows.First, we introduce the target selection and spectroscopic follow-up observation in Section 2. In Section 3, physical properties of the quiescent galaxy in-ferred from spectral energy distribution (SED) fitting are summarized.Then, we discuss the galaxy's surrounding environment as well as its possible formation scenarios in Section 4. Finally, we conclude the paper in Section 5. We assume a Chabrier (2003) initial mass function (IMF) and a flat ΛCDM cosmology with H 0 = 70 km s −1 Mpc −1 , Ω m = 0.3, and Ω Λ = 0.7.All magnitudes are in the AB system (Oke & Gunn 1983).

Target Selection
Following the successful confirmation of the z = 4.01 quiescent galaxy in the Subaru/XMM-Newton Deep Field (Tanaka et al. 2019), we searched for massive quiescent galaxy candidates at z > 4 in the COSMOS field (Scoville et al. 2007).The COSMOS field is ∼ 2 deg 2 and has been extensively observed from the X-ray to the radio wavelengths, which allows us to obtain highly accurate photometric redshifts (photo-z's).As summarized in Ito et al. (2022), we applied our photo-z code (Tanaka 2015) to the COSMOS2020 Classic catalog (Weaver et al. 2022).The code adopts the Bruzual & Charlot (2003) models, assuming exponentially declining star formation histories (SFHs, SF R(t) ∝ e −t/τ ), solar metallicities, the Calzetti et al. (2000) dust attenuation curve, and the Chabrier (2003) IMF.From the comparisons between photo-z's and spectroscopic redshifts from the literature, our photo-z's are confirmed to be accurate for massive quiescent galaxies at 2 < z < 4 in the COSMOS field (σ[∆z/(1+z spec )] = 0.03, Median(∆z) = 0.011: Ito et al. 2022, and references therein).
Based on this photo-z catalog, we applied the following cuts to select quiescent galaxy candidates; (1) z phot > 4 with little probability of being at lower redshifts to ensure that the Balmer break is in the K-band, (2) M * ,c ≳ 5 × 10 10 M ⊙ so that the target is sufficiently bright for us to trace its continuum (M * ,c refers to a stellar mass in the photo-z catalog), and (3) star formation rate (SFR) inferred from the SED fitting is more than 1 dex below the star-forming main sequence (Schreiber et al. 2015).The last criterion is effectively similar to the rest-frame U V J selection (Williams et al. 2009).
After careful visual screening, we identify a massive quiescent galaxy candidate at z phot ∼ 4.65 with these cuts.Its SED exhibits a strong Balmer break, which suggests that the galaxy experienced an intense burst of star formation in the recent past and that its SFR has been declining since then (post-starburst phase).This object (COSMOS-1047519) is the subject of the paper.We note that there are other quiescent galaxy candidates, but most of them are very faint in the K s -band (too high redshifts where the Balmer break is outside of the K s -band or suffer from the dust extinction).The selected candidate is the brightest candidate after the selection.

Spectroscopic follow-up
We performed spectroscopic follow-up in the K-band using the MOSFIRE spectrograph (McLean et al. 2010(McLean et al. , 2012) ) on the Keck I telescope.We expected to fully cover the entire Balmer break from the photo-z.The target is rather faint with a K s -band magnitude of 23.2 mag, and a long integration was required to identify absorption features.However, the weather did not fully cooperate and we could observe roughly half of the allocated time (∼ 9 hour on-source exposure).The observations were performed on 3 half nights in 2022 March and April.We took ∼ 100 pairs of the ABBA nodding with a 3 min exposure at each position.We nod ±2 ′′ along the slit for good sky subtraction.We included two bright stars in the mask to monitor the sky conditions (see below).The mean seeing was around 0 ′′ .7.
We reduced the data in the standard manner using the MOSFIRE Data Reduction Pipeline (DRP) 1 as well as custom-designed code.We reduced each A/B pair using DRP (flat-fielding, wavelength calibration, and background subtraction).The flux and the FWHM of the bright stars in each exposure were used to evaluate the observing conditions.We used the weight of flux/FWHM in the coaddition of the spectra to give more weight to exposures taken under better conditions.Four exposures were taken through thick clouds and were thus excluded from the coadd.
We observed the A0V star (HD50046) as a spectrophotometric standard star on the first night of the observation.We applied the flux calibration and correction for the telluric absorption by comparing the alpha-Lyr spectrum from HST/CALSPEC (Bohlin et al. 2014) and the observed standard star spectrum.The CALSPEC spectrum was scaled to the standard star using the K s -band photometry from 2MASS (Skrutskie et al. 2006).
We performed the optimal extraction (Horne 1986) to extract the 1D spectra from the 2D spectra.The spatial profile of the bright star spectrum was fit by Gaussian, and it was used to extract the flux of the target galaxy.The error spectrum was estimated using the standard deviation of the background pixels in the same slit.Finally, the flux of the 1D spectrum is scaled to match the observed photometry to correct for the slit loss.Figure 1 shows the observed 2D and 1D spectra rebinned over 5 pixels (10.8 Å).Due to the low S/N of the spectrum, we could not securely identify absorption lines.For example, the median S/N at 21000 Å < λ < 22000 Å is only 1.4 per 2.2 Å (original resolution).However, the spectrum exhibits a faint [O II] λλ3727, 3729 emission with S/N peak ∼ 5.A blow-up of [O II] is shown in the right panel of Figure 1.This [O II] emission line provides a secure spectroscopic redshift for this object.We perform a Monte-Carlo run to fit the [O II] doublet with two Gaussians and adopt the median as a point estimate and the 68% interval as an uncertainty, z spec = 4.5313 ± 0.0005.This redshift is consistent with the photo-z (z phot = 4.654 +0.134  −0.104 ).Based on this redshift estimate, we discuss detailed properties of the galaxy in the next Section.

SED Fitting
Figure 2 shows the SED of the galaxy.The black points are the photometry from the COSMOS2020 Classic catalog with 40 bands from u-band to IRAC ch4 (Weaver et al. 2022).
The blue spectrum is the MOSIFRE spectrum, which shows the prominent Balmer break, suggesting the post-starburst nature of this galaxy.The spectrum fully covers the break feature as expected, and we can accurately infer its star formation history as shown below.
We fit the observed SED with prospector (Johnson et al. 2021).The MOSIFRE spectrum is re-binned over 3 pixels to gain S/N (e.g., the median S/N ∼ 2.5 per 6.5 Å at 21000 Å < λ < 22000 Å).The prospector code uses the Flexible Stellar Population Synthesis model (Conroy et al. 2009;Conroy & Gunn 2010) to infer phys-  We assume two models of star formation history (SFH).One is a parametric SFH model with a linear rise of SFR followed by an exponential decrease ("delayed tau-model", SF R(t) ∝ te −t/τ ).The parameter τ represents decay timescale of star formation.We choose this model because it is widely used and can compare with other quiescent galaxies from the literature.The other model is a non-parametric SFH model, which al-lows the SFR in each age bin to vary, thus gives more flexibility in the SFH modeling.There are several priors to use this non-parametric model, and we use "log-mass prior" and "Dirichlet prior" (Leja et al. 2017(Leja et al. , 2019)).The log-mass prior is the most flexible model, but it could lead to a non-physical SFH such as causing intense burst star formation just after the Big Bang.Thus, we also use the Dirichlet prior, which forces the SFH to be smoother.As discussed below, the estimated SFHs are similar regardless of the SFH model used.
For all models, we apply the Markov Chain Monte Carlo (MCMC) using emcee (Foreman-Mackey et al. 2013) to infer the best-fitting parameters and their associated errors.We use the log-uniform prior to the input parameters such as τ , age, and stellar mass.To be specific, in the parametric SFH model, the parameters are allowed to vary between 0.001 Gyr < τ < 10 Gyr, 0.13 Gyr < age < 1.30 Gyr, which is the age of the Universe at z = 4.53, and 10 9 M ⊙ < M * < 10 12 M ⊙ .For dust optical depth, we use the top-hat prior between 0 < A V < 6.In the non-parametric SFH model, we use the uniform prior to constrain the stellar masses of each age-bin.
The best-fit model spectrum is shown in red in Figs. 1  and 2. The model clearly shows the strong Balmer break in the K-band, which indicates that the galaxy recently experienced a starburst and is now in the post-starburst phase.There is relatively strong rest-frame UV continuum, implying that there are residual star formation activities in the galaxy.This is thus likely a young poststarburst and is likely being quenched.We will further discuss this point using physical properties inferred from the SED fit in the following subsection.

Inferred physical properties and SFH
From the SED fit, we infer physical properties of the galaxy such as stellar mass, star formation rate (SFR), dust optical depth, and star formation history.These parameters are summarized in Table 1.We estimate the SFR from the SED fits and also from [O II] fit using two Gaussians.We use the following equation to convert the [O II] emission flux to the SFR (Kewley et al. 2004;Schreiber et al. 2018).

SF R
(1) We correct for dust attenuation of L [O II] using the dust optical depth from the best-fit model.Figure 3 shows the location of the object on a SFR vs. stellar mass diagram.We also plot the normal starforming galaxies from ALPINE survey (Le Fèvre et al. emission corrected for dust are located around or below the orange line.The gray points are the normal star-forming galaxies at zspec ∼ 4.5 from the ALPINE survey (Le Fèvre et al. 2020;Béthermin et al. 2020;Faisst et al. 2020).There are two ALPINE galaxies below the orange line, but the one at log (M * /M⊙) ∼ 10.3 has a contaminating galaxy nearby, and the other one is possibly undergoing dust-obscured star formation (Faisst et al. 2020).We also include massive star forming galaxies around the quiescent galaxy (the black diamonds, see discussion in Section 4).2020; Béthermin et al. 2020;Faisst et al. 2020) for reference.The galaxy is massive (M * ≈ 6.0 × 10 10 M ⊙ ), and has a very low SFR (SF R ≈ 10 M ⊙ yr −1 ) averaged over 10 7.5 yr.The estimated SFR is more than 1 dex below the star-forming main sequence (Schreiber et al. 2015).This supports our argument in the previous subsection that the galaxy is a massive quiescent galaxy based on Schreiber et al. (2018) criteria.If we perform the SED fitting without using the spectrum, uncertainties of physical properties are about 2 times larger.In addition, the height of the Balmer break cannot be fully determined by K s -band photometry only, since the K sbend is on the break.The spectrum, on the other hand, can firmly determine the break strength, resulting in a more precise age determination.Figure 3 also shows that the [O II]-based SFRs are higher than those obtained by SED fitting, which we interpret as a contribution from active galactic nuclei (AGN) as we discuss in Section 3.4 (the model spectrum cannot reproduce [O II] either in Figure 1).
Figure 4 shows the inferred star formation history of the galaxy.All of the SFH models are broadly consistent with each other particularly during the quenching epoch; the galaxy experienced a rapid quenching process   beginning around z ∼ 5, which closely aligns with the SFH of massive elliptical galaxies (Thomas et al. 2010).
The galaxy is young with a stellar age of 198 +44 −32 Myr from delayed tau-model, and the observed low SFR is consistent with the SFHs.Inferred physical parameters and SFH are not significantly changed by the age bins of the non-parametric SFH model.We conclude that this is a massive quiescent (or being quenched) galaxy and is likely a progenitor of massive elliptical galaxies in the local Universe.
The quiescence of the galaxy is also supported by the multi-wavelength data.We look at the infrared to radio wavelengths to see a sign of dust-obscured star formation activity.However, the object is not detected in FIR (S/N IR ∼ 1.25, SF R IR < 300 M ⊙ yr −1 in "Super-deblended" catalog: Jin et al. 2018) and radio (f 10 cm = 6.3 × 10 −4 ± 2.8 µJy in VLA 3GHz: Smolčić et al. 2017).The object has not been observed by ALMA, and we do not have a strong constraint on dust emission at this point.

Comparison to massive quiescent galaxies at z > 3
An interesting aspect of our quiescent galaxy is that it is currently experiencing ongoing suppression of star formation.The prominent Balmer break suggests an evolved stellar population, while the remaining UV continuum as well as the weak [O II] emission suggests that the galaxy is not completely quenched yet (Figs. 1 and  2).The galaxy has a young luminosity-weighted age, and we must be observing the galaxy close to its primary formation epoch.
Figure 5 shows the relationship between the formation redshift (z form ) and observed redshift (z spec ) of massive quiescent galaxies at z spec > 3 collected from the literature.The formation redshift is defined by the epoch at which the galaxy formed 50% of total stellar mass.This figure shows that massive quiescent galaxies at z > 3 have a wide range of formation epochs (also confirmed at z < 3: summarized in Kalita et al. 2021 The young age of our galaxy is likely due to a selection effect; we intentionally targeted the brightest candidate in the K s -band (which covers the Balmer break), and thus we tend to select galaxies with a prominent Balmer break, leading us to young post-starbursts.It is interesting that old quiescent galaxies show little dust attenuation (A V ∼ 0: e.g., Carnall et al. 2023a;Nanayakkara et al. 2022), while young ones suffer from a larger amount of attenuation (A V ∼ 0.7: This work, Valentino et al. 2020).This indicates that post-starburst galaxies have residual dust.
Based on the non-parametric star formation history (Figure 4), the peak SFR of COSMOS-1047519 is SF R ∼ 200−400 M ⊙ yr −1 .This is about 2-3 times higher than the SFR of star-forming main sequence galaxies at z = 5 (Schreiber et al. 2015).If the galaxy experienced an intense starburst only once, starbursting galaxies are possible progenitors as suggested by previous research (e.g., Forrest et al. 2020a,b;Valentino et al. 2020;Carnall et al. 2023a).Massive quiescent galaxies at z ∼ 2 show similar properties to submillimeter galaxies (SMGs) at higher redshifts in terms of masses, effective radii, and number densities (Toft et al. 2014;Gómez-Guijarro et al. 2018).SMGs tend to have a larger star formation rate than the star-forming main sequence.By comparing the physical properties, the peak SFR in SFH of our galaxy is consistent with that of SMGs which have similar stellar mass at z ∼ 5 (e.g., da Cunha et al. 2015;Miettinen et al. 2017).Moreover, the quenching timescale for our quiescent galaxies is roughly 100 Myr, which is comparable to the gas depletion timescales of SMGs with similar specific star formation rates at 1 < z < 5 (Gómez-Guijarro et al. 2022).This quenching timescale is defined as the period from when the galaxy reaches its peak SFR in the nonparametric SFH until the SFR decreases to 10% of that peak.Also, Zavala et al. (2022) summarize the redshift evolution of gas depletion timescale of both starburst galaxies and normal star-forming galaxies and show that the timescale is shorter at the high-redshift Universe.Our galaxy's quenching timescale is consistent with the starburst galaxies' gas depletion timescale at z ∼ 5. Thus, our work supports the scenario in which SMGs evolve into quiescent galaxies.

Comparisons with simulations and the role of AGNs
The Illustris TNG300 simulation (Pillepich et al. 2018;Nelson et al. 2019) is a large cosmological hydrodynamical simulation and reproduces the cosmic evolution from the early Universe to the local Universe.Thus, the simulation is a very useful tool to help us understand how massive galaxies form in the early Universe.We search for similar objects to COSMOS-1047519 in a snapshot at z = 4.43.However, there are no massive (M * > 10 10.5 M ⊙ ) quiescent galaxies with sSF R < 10 −9.5 yr −1 .Hartley et al. (2023) confirm the first quiescent galaxy in the TNG300 is at z ∼ 4.2 although the definition of the quiescent galaxy is slightly different from ours (sSF R < 10 −10 yr −1 ).They conclude that kinetic feedback from AGN has a significant effect on the quenching process of quiescent galaxies reconstructed at z ∼ 4.2.We note that COSMOS-1047519 is not detected in the deep X-ray observation (L 2−10 keV < 5.1 × 10 43 erg s −1 in Chandra COSMOS Legacy: Civano et al. 2016), but this upper limit of X-ray luminosity does not rule out the possibility of low-luminosity AGN in this galaxy.
Figure 1 shows that the model spectrum cannot fully reproduce the [O II] flux.Table 1 and Figure 3 also show that [O II]-based SFR is higher than that from SED fitting.The observed excess suggests that AGN may reside in the galaxy because [O II] is sensitive to ionization by (low-luminosity) AGN as well as star formation (e.g., Yan et al. 2006).Maseda et al. (2021) find that more than half of massive quiescent galaxies at z < 1 exhibit [O II] lines and conclude that this feature is due to AGN activities.At z > 2, there are massive quiescent galaxies with AGN (e.g., Kriek et al. 2009;Marsan et al. 2015;Schreiber et al. 2018;Kubo et al. 2022).Our galaxy may be a similar case, although it is located at a higher redshift.Future deep X-ray or JWST NIR observations of AGN activity (confirming [O III], [N II], and Balmer lines) are necessary to identify or rule out low-luminosity AGN.To gain deeper insights into quenching physics, we investigate the environment of our galaxy.We select massive (M * ,c > 10 10 M ⊙ ) galaxies regardless of SFRs at 4.4 < z phot < 4.7 from the photo-z catalog in the COSMOS field (Section 2.1: Ito et al. 2022).The inset of Figure 6 shows a blow-up around the quiescent galaxy (QG).There is a companion galaxy (z phot = 4.535 +0.168  −0.105 ) right next to it with a separation of 2 ′′ .Furthermore, there are 3 massive starforming galaxies (SFGs) (z phot = 4.57−4.69)within 23 ′′ or 151 kpc in physical scale.We recall that the quiescent galaxy has a photo-z of 4.65.We summarize the SEDs and probability distributions of photo-z inferred from MIZUKI (Tanaka 2015) in Figure 7, and the coordinates in J2000 are summarized in Table 2.Each SED shows that all member candidates have bluer SEDs compared to the quiescent galaxy.The probability distributions show that all galaxies have the highest probability at z phot ∼ 4.5, and most of them have consistent photometric redshifts within 1σ of the quiescent galaxy.These photo-z's are also consistent with the COSMOS2020 catalog.
To evaluate the physical properties of these SFGs and the QG homogeneously, we perform the SED fitting analysis of 4 SFGs by prospector using a parametric SFH model (delayed-tau model with the same priors as those used for the QG).The inferred physical properties are summarized in Table 3.The SFGs are normal starforming galaxies (their locations on a SFR vs. stellar mass diagram are shown in Figure 3 as black diamonds) but their stellar masses are smaller than our original estimates from our photo-z code.Yet, these values are still satisfied the 70% completeness limit of SFGs at z = 4.5 (M * > 10 9.37 M ⊙ , Weaver et al. 2023).The quiescent galaxy has the largest stellar mass, and this environment may be the highest redshift galaxy group with a spectroscopically confirmed quiescent galaxy at the center.
To evaluate the significance of this over-density of massive galaxies, we make a density plot in Figure 6.This plot is obtained by the Gaussian kernel density estimation method.The galaxy distribution in the entire COSMOS field is smoothed over 30 ′′ (200 physical kpc at z = 4.5, which roughly corresponds to the scale of this over-density region).Our sample here is complete for SFGs but not for QGs (Weaver et al. 2023), and we may be missing low-mass QGs if any.SFGs significantly outnumber QGs at these redshifts and missing low-mass QGs will not strongly affect the density field.We find that our galaxy is located in a highly significant (12σ) over-density region, and this structure turns out to be the densest region of massive galaxies at 4.4 < z phot < 4.7 in the COSMOS field.On the other hand, if the galaxy distribution is smoothed over 2.5 ′ (1 physical Mpc at z = 4.5), the region is a less significant (2.2σ) over-density.These kernel density does not follow the Gaussian distribution, and the absolute significance is not very important.We also estimate the excess of galaxy density using the 5th Nearest Neighbor method, which yields a density distribution more similar to Gaussian and obtains 180σ.The group still remains the densest region in the entire COSMOS field.Thus, the group is very compact.For comparison, we indicate in Figure 6 a massive proto-cluster at z ∼ 4.57 with the blue circle (Lemaux et al. 2018).As can be seen, the proto-cluster is much more extended than our group and on the scale of ∼ Mpc.The structure and size of our group is really different from typical proto-clusters at this high redshift (Overzier 2016 for a review).

Halo mass and virial radius
While we have not spectroscopically confirmed the redshifts of the SFGs yet, we make an attempt to infer the halo mass of each member and see if they can be physically associated.We use the stellar-to-halo mass relation (SHMR) observed in the COSMOS field at 4.5 < z < 5.5 (Shuntov et al. 2022) to estimate the halo mass from the stellar mass based on prospector assuming that each member is a central galaxy.Estimated   values are summarized in Table 3 ( 8).The most massive member (quiescent galaxy) has M h ∼ 10 12.3 M ⊙ , which is slightly smaller than the galaxy group at z ≳ 3 from previous research (M h ≳ 10 13 M ⊙ : e.g., Miller et al. 2018;Hill et al. 2020;Daddi et al. 2021).Next, we translate the halo mass into the virial radius.Here, we use the radius r 200 (Carlberg et al. 1997) as the virial radius.Table 3 ( 9) summarizes the virial radii of the member candidates.The dotted circles in the inset of Figure 6 show the virial radii of the member candidates significantly overlap.Two SFGs are just outside of the virial radius of the quiescent galaxy in Figure 6, but they are actually consistent with being inside the virial radius within the uncertainties (see Table 3).The right bottom galaxy is outside of it, but there is intrinsic scatter in the SHMR.These 3 -4 galaxies likely form a group with the quiescent galaxy at the center.Spectroscopic confirmations of these member candidates are required for further characterization of the group environment.and then rapid quenching started at z ∼ 5.The peak SFR and quenching timescale from the SFH are similar to the SMG's SFR and gas depletion timescale at z ∼ 5, respectively.Also, the K-band spectrum of the quiescent galaxy shows the excess of [O II] emission compared to the model spectrum from prospector, and this galaxy may harbor a low-luminosity AGN.Combining these points, the star formation and AGN activities induced by the galaxy-galaxy interactions and mergers drive the quenching of this quiescent galaxy.For example, observational and theoretical studies argue that minor mergers and galaxy interactions increase star formation efficiency and AGN activity (e.g., Hopkins et al. 2008;Brodwin et al. 2013;Gómez-Guijarro et al. 2018;Coogan et al. 2018).Also, previous studies confirm that gas mass fractions of QGs are significantly lower than those of SFGs (e.g., Sargent et al. 2015;Bezanson et al. 2019;Williams et al. 2021;Whitaker et al. 2021;Suzuki et al. 2022).Thus, it is possible that the quick gas depletion due to starburst is responsible for quenching.Shimakawa et al. (2018) suggest that group/cluster environments have evolved stellar populations, and AGN feedback suppresses the star formation of member galaxies at high redshift (2 < z < 3).Chiang et al. (2017) argue that significant group-or cluster-sized cores (with the scale of r 200 ) would be the first region to show the galaxy quenching.Our group candidate has a comparable scale of these cores, and the quenching observed in this group environment at z > 4 is probably consistent with the core-scale quenching.Our work suggests this core-scale quenching has already occurred at z > 4 by gas consumption due to the starburst and/or AGN feedback.Furthermore, recent James Webb Space Telescope (JWST) observations confirm compact galaxy groups at 5 < z < 8 (e.g., Sun et al. 2023;Jin et al. 2023;Helton et al. 2023;Morishita et al. 2023;Hashimoto et al. 2023).Member galaxies are mostly low-mass SFGs, but Helton et al. (2023) show that more massive galaxies with larger SFR populate in the over-density region compared to the field.Therefore, the formation of galaxy groups with a QG at z = 4.5 could result from the evolution of such very high-redshift compact galaxy groups.

CONCLUSION
We present the spectroscopic confirmation of a massive quiescent galaxy at z = 4.53 based on the Keck/MOSFIRE observation.We confirm a very large stellar mass and a very low star formation rate from the extensive SED fitting using both the photometry and spectrum.The estimated SFR is more than 1 dex below the star-forming main sequence at z = 4.5.The star formation history of the galaxy inferred from the SED fitting indicates rapid quenching with a timescale of ∼ 100 Myr from z ∼ 5.
To constrain physical drivers of the starburst and subsequent quenching, we compare the properties of this object with those of quiescent galaxies from the literature.The galaxy is one of the youngest quiescent galax-ies from the formation redshift; this is an interesting object to understand the physical processes responsible for the suppression of star formation.The fact that the quenching timescale of the galaxy is comparable to the gas depletion timescale of SMGs at similar redshift indicates that the quenching is simply due to gas consumption.Also, the observational excess of [O II] indicates the existence of low-luminosity AGN.In addition, the galaxy is found to be located in the group environment including a companion galaxy at 2 ′′ (13 physical kpc).These findings suggest that gas depletion due to the starburst and/or AGN feedback triggered by galaxy-galaxy interactions or mergers may be responsible for quenching.
In Figure 5, we can clearly say that we are approaching the formation epoch of the massive quiescent galaxies.This is an exciting redshift regime to directly unveil the quenching physics.Recent JWST observations confirm quiescent galaxies at these redshifts (Carnall et al. 2023a) as well as a large number of the quiescent galaxy candidates (Valentino et al. 2023;Carnall et al. 2023b;Long et al. 2023).Upcoming imaging and spectroscopic observations will place a more stringent constraint on the physical mechanism of quenching and detailed star formation history in the near future.
We thank the anonymous referee for constructive comments.This study was supported by JSPS KAKENHI Grant Number JP21K03622 (MO).PFW acknowledges funding through the National Science and Technology Council grant 111-2112-M-002-048-MY2 and National Taiwan University grant NTU 112L7318 and NTU 112L7439.The data presented herein were obtained at the W. M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration.The Observatory was made possible by the generous financial support of the W. M. Keck Foundation.The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Maunakea has always had within the indigenous Hawaiian community.We are most fortunate to have the opportunity to conduct observations from this mountain.Based on observations collected at the European Southern Observatory under ESO programme ID 179.A-2005 and on data products produced by CALET and the Cambridge Astronomy Survey Unit on behalf of the UltraVISTA consortium.

Figure 1 .
Figure 1.Left: Keck/MOSFIRE K-band spectrum of COSMOS-1047519 smoothed over 5 pixels (10.8 Å).The top panel shows the two-dimensional spectrum, and the bottom panel shows the one-dimensional spectrum (the blue line) and the associated noise spectrum (the yellow line).The red line shows the best-fit model spectrum from prospector using spectrum and photometry (see Section 3).Right: Blow-up of [O II] doublet lines smoothed over 2 pixels (4.3 Å).The red line represents the best-fit Gaussians.

Figure 2 .
Figure 2. Spectral energy distribution of the target.The black points and error bars show the photometry from the COSMOS2020 catalog (Weaver et al. 2022), and the blue spectrum at ∼ 2 µm is the MOSFIRE spectrum.The red spectrum is the best-fit model spectrum from prospector.

Figure 3 .
Figure 3. Stellar mass and SFR relation inferred from the three SFH models (color-coded symbols) compared to the star-forming main sequence (the blue line: Schreiber et al. 2015).The orange line shows the star-forming main sequence shifted by -1 dex.Inferred SFRs from both SED and [O II]emission corrected for dust are located around or below the orange line.The gray points are the normal star-forming galaxies at zspec ∼ 4.5 from the ALPINE survey(Le Fèvre et al. 2020;Béthermin et al. 2020;Faisst et al. 2020).There are two ALPINE galaxies below the orange line, but the one at log (M * /M⊙) ∼ 10.3 has a contaminating galaxy nearby, and the other one is possibly undergoing dust-obscured star formation(Faisst et al. 2020).We also include massive star forming galaxies around the quiescent galaxy (the black diamonds, see discussion in Section 4).

Figure 4 .
Figure 4. Inferred SFHs of the target; the black, blue, and red lines are the non-parametric SFH model with logmass prior, non-parametric model with Dirichlet prior, and the delayed-tau model, respectively.The filled circles and diamonds with the error bars indicate the epochs when the galaxy formed 50% and 80% of the stars, respectively (the colors correspond to the SFH models).

4 .
GROUP ENVIRONMENT AROUND THE QUIESCENT GALAXY 4.1.Over-density in the surrounding structure

Figure 6 .
Figure 6.Distribution of massive (M * ,c > 10 10 M⊙ inferred from the photo-z catalog) galaxies at 4.4 < z phot < 4.7 in the entire COSMOS field.The red color shows the density excess of the massive galaxy candidates smoothed over 30 ′′ which corresponds to the scale of the galaxy group candidate using kernel density estimation.The blue circle shows the massive proto-cluster at zspec ∼ 4.57 (Rproj = 2 Mpc: Lemaux et al. 2018).The right panel shows a pseudo-color image of the area around the quiescent galaxy (Red: VISTA/Ks-band, Green: VISTA/H-band, Blue: HSC/i-band, McCracken et al. 2012; Aihara et al. 2022).The five galaxies indicated with the white solid circles are the member candidates of this group centered on the quiescent galaxy.Our massive quiescent galaxy at z = 4.53 (the red point in the RGB image) exists in the densest region (12σ in the COSMOS field).The red and white dotted circles show the estimated virial radii of the quiescent galaxy and star-forming member candidates, respectively.The virial radii are estimated based on the stellar masses inferred from prospector.

4. 3 .Figure 7 .
Figure 7. SEDs of the group member candidates.The black points show the photometry from the COSMOS2020 catalog(Weaver et al. 2022), and the red lines and blue squares indicate the model SEDs from MIZUKI(Tanaka 2015).The bottom panels show the probability distribution of photometric redshift.All galaxies have a large probability at z phot ∼ 4.5.

Table 1 .
Inferred physical properties of the quiescent galaxy (COSMOS-1047519) Note-1.Photometric redshift inferred from MIZUKI (Tanaka 2015).2. Spectroscopic redshift from the Keck/MOSFIRE spectrum.3. Name of the star formation history model used in SED fitting (Johnson et al. 2021).4. Inferred stellar mass. 5. Star formation rate from the SED fitting averaged over 10 7.5 yr. 6. V -band magnitude of dust attenuation from SED fitting.7. Formation redshift at which the galaxy formed 50% of total stellar mass.Here, the lookback times from the SFH are converted into redshifts assuming the adopted cosmology (see Section 1).8. Quenching timescale only in the delayed tau-model.9. Star formation rate from the flux of the [O II] emission corrected for dust attenuation.

Table 3 .
Inferred physical properties of massive galaxy group members