Evidence for gas-phase metal deficiency in massive protocluster galaxies at z~2.2
Zahra Sattari, Bahram Mobasher, Nima Chartab, Behnam Darvish, Irene Shivaei, Nick Scoville, David Sobral
DDraft version February 11, 2021
Typeset using L A TEX twocolumn style in AASTeX62
Evidence for gas-phase metal deficiency in massive protocluster galaxies at z ∼ . ∗ Zahra Sattari, Bahram Mobasher, Nima Chartab, Behnam Darvish, Irene Shivaei, † Nick Scoville, andDavid Sobral Department of Physics and Astronomy, University of California, Riverside, 900 University Ave, Riverside, CA 92521, USA Cahill Center for Astrophysics, California Institute of Technology, 1216 East California Boulevard, Pasadena, CA 91125, USA Steward Observatory, University of Arizona, 933 North Cherry Ave, Tucson, AZ 85721, USA Department of Physics, Lancaster University, Lancaster, LA1 4YB, UK
ABSTRACTWe study the mass-metallicity relation for 19 members of a spectroscopically-confirmed protoclusterin the COSMOS field at z = 2 . α emitting sources, chosen from the HiZELS narrow-bandsurvey, with metallicities derived from N2 ( [NII] λ α ) line ratio. For the mass-matched samples ofprotocluster and field galaxies, we find that protocluster galaxies with 10 . M (cid:12) ≤ M ∗ ≤ . M (cid:12) aremetal deficient by 0 . ± .
04 dex (2 . σ significance) compared to their coeval field galaxies. This metaldeficiency is absent for low mass galaxies, M ∗ < . M (cid:12) . Moreover, relying on both SED-derived andH α (corrected for dust extinction based on M ∗ ) SFRs, we find no strong environmental dependenceof SFR-M ∗ relation, however, we are not able to rule out the existence of small dependence due toinherent uncertainties in both SFR estimators. The existence of 2 . σ significant metal deficiency formassive protocluster galaxies favors a model in which funneling of the primordial cold gas throughfilaments dilutes the metal content of protoclusters at high redshifts ( z (cid:38) Keywords:
Metallicity (1031); Galaxy evolution (594); Protoclusters (1297); High-redshift galaxy clus-ters (2007); Large-scale structure of the universe (902) INTRODUCTIONIn the standard ΛCDM cosmological scenario, struc-tures form from the growth of small fluctuations throughgravitational instability. Dark matter structures growby two main processes: By merging small halos to formthe larger ones or by smooth accretion of dark matterfrom their immediate environment. Baryons fall into thepotential well of these dark matter structures and feedgalaxies within those structures with cold pristine gas,
Corresponding author: Zahra [email protected] ∗ Based on data obtained at the W. M. Keck Observatory, whichis operated as a scientific partnership among the California In-stitute of Technology, the University of California and the Na-tional Aeronautics and Space Administration. The Observatorywas made possible by the generous financial support of the W. M.Keck Foundation. † Hubble Fellow which allows them to form their stars. Galaxy clustersobserved in the present Universe are the largest virial-ized dark matter halos that are populated with massiveand evolved galaxies (e.g., Dressler 1980; Balogh et al.2004; Kauffmann et al. 2004; Peng et al. 2010). At highredshifts ( z (cid:38) z ∼ . a r X i v : . [ a s t r o - ph . GA ] F e b Sattari et al. vides the fuel for galaxies to form their stars, which arethe factories responsible for metal production in galax-ies. However, galaxies are not very efficient in formingstars as the cold gas accretes into their potential well.Behroozi et al. (2013a) found that the star formationefficiency of galaxies (the star formation rate (SFR) di-vided by the baryon accretion rate) can reach a max-imum of ∼ z ∼ z ∼ ∼ .
07 dexlower metallicity than the field galaxies at z ∼ .
3. Onthe other hand, Kacprzak et al. (2015) and Alcorn et al.(2019) found no significant environmental dependenceon the MZR of galaxies at z ∼ z ∼ . α emitting sources, selected from the narrow-band High-Z Emission Line Survey (HiZELS; Geach et al. 2012; Sobral et al. 2013, 2014). The paper isorganized as follows: In Section 2, we describe theKeckI/MOSFIRE observations and data reduction pro-cedure followed by stellar mass measurements and sam-ple selection. We then explain the stacking process andmeasurements of gas-phase metallicities in Section 3. InSection 4, we construct the MZR for both protoclusterand field samples and compare them to deduce the roleof the environment in MZR at z ∼ .
2. We discuss ourresults in Section 5.Throughout this paper, we assume a flat ΛCDM cos-mology with H = 70 kms − Mpc − , Ω m = 0 . Λ = 0 .
7. All the physical parameters are measuredassuming a Chabrier (2003) initial mass function (IMF). DATA2.1.
MOSFIRE Observation
In this paper, we use near-IR spectroscopy of galaxiesin a recently confirmed protocluster, CC2.2, at z ∼ . K ( ∼ . − . µm ) and H ( ∼ . − . µm ) bands, leading to 35 confirmed mem-bers. The primary spectroscopic sample comprises thenarrow-band H α emitting candidates from the HiZELSsurvey (Geach et al. 2012; Sobral et al. 2013, 2014). Fora full description of the protocluster identification andobservation, we refer readers to Darvish et al. (2020).As a control sample, we use KeckI/MOSFIRE spec-troscopic observations of 24 field galaxies in the redshiftrange 2 . ≤ z ≤ .
24, located in less-crowded regions ofthe COSMOS (16 galaxies) and UDS (8 galaxies) fields.All these field galaxies were observed over the observ-ing programs during 2018-2019 (PI: N. Z. Scoville) in K band (a few in H band as well) and are selected similarlyfrom the HiZELS survey to avoid any biases introducedby sample selection.The observations were performed the same way forboth field and protocluster galaxies, minimizing anysource of bias. All the observing nights were conductedunder clear conditions with the average seeing of 0 . (cid:48)(cid:48) ,and a typical exposure time of ∼
90 minutes per mask.2.2.
Data Reduction
The acquired data were reduced using the MOS-FIRE data reduction pipeline (DRP) . The outputs ofthe DRP are rectified, sky-subtracted, and wavelength- https://github.com/Keck-DataReductionPipelines/MosfireDRP ZR of a protocluster at z ∼ . wavelength(˚A) f λ ( a r b i tr a r y un i t) Field galaxyHiZELS ID= UDS-8254 H α [ N II ] λ wavelength(˚A) f λ ( a r b i tr a r y un i t) Protocluster galaxyHiZELS ID= COSMOS-9026 H α [ N II ] λ Figure 1.
The 2D and corresponding optimally extracted 1D spectra for two galaxies selected from our sample.
Left : Exampleof a field galaxy. This galaxy is at z = 2 .
223 and has a stellar mass 10 . M (cid:12) . Right : A galaxy member of the protocluster at z = 2 . . M (cid:12) . In both spectra, H α and [NII] λ calibrated 2D spectra and their associated uncertain-ties. We then extract optimally weighted 1D spectraand their errors using the optimal extraction algorithm(Horne 1986). We weight the flux of each pixel by theinverse of the flux variance and the spatial extent of the2D spectrum in an optimized window and then, sum theweighted fluxes along the wavelength axis. The size ofthe optimized window is determined such that the brightfeatures with the highest signal-to-noise (S/N) in the 2Dspectrum are surrounded by the window. The weightedsummation within the optimized window produces 1Dspectra of the sources along with their errors. Figure 1shows an example of the 2D and optimally extracted 1Dspectra for a field galaxy and a protocluster member.We fit a triple Gaussian function to the reduced 1Dspectra to extract [NII] λ α and [NII] λ α emission lines with S / N ≥ Stellar Masses and SFRs
We estimate the stellar masses (M ∗ ) and SFRs ofgalaxies by fitting synthetic spectral energy distribu-tions (SED) to their available photometric data (COS-MOS: Laigle et al. (2016); UDS: Mehta et al. (2018)). Toperform the SED fitting, we utilize Bayesian Analysis ofGalaxies for Physical Inference and Parameter EStima-tion (Bagpipes) code (Carnall et al. 2018), which uses2016 version of a library of Bruzual & Charlot (2003) synthetic spectra. We fix the redshifts of galaxies totheir spectroscopic values, considering delayed exponen-tially declining star formation history, te − t /τ . A rangeof 0 . −
10 Gyr with a uniform prior is assumed forthe star formation e-folding time-scale ( τ ). The neb-ular emission models which are constructed based onthe methodology of Byler et al. (2017) are added to theSEDs. Also a metallicity range 0 < Z / Z (cid:12) < . α emission line fluxes taken from HiZELS sur-vey (Sobral et al. 2013). Since the H band data arenot available for most of the sample, following Sobralet al. (2012) and Koyama et al. (2013), we utilize Garn& Best (2010) calibration to correct the H α luminosityfor the dust extinction based on the stellar masses ofthe galaxies. We then convert the dust-corrected H α luminosity to SFR using the calibration from Kennicutt(1998): SFR(M (cid:12) yr − ) = 7 . × − L H α (erg / s) . Oneshould note that the SFR derived from this method ishighly uncertain due to the existence of a large scatter indust attenuation calibration based on the stellar mass.2.4.
Sample Selection
We select galaxies with significant detection in H α emission lines (S/N ≥ ∗ < . M (cid:12) to construct a mass complete sample.We also remove potential mergers by visual inspectionof their spectra and images, as well as the active galac-tic nuclei (AGNs). The AGNs are identified throughtheir broad emission lines, X-ray flags included in theUDS and COSMOS catalogs, or IR emissions (Donleyet al. 2012). Moreover, optical AGNs are excluded by Sattari et al. requiring log( [NII] λ α ) < − . z = 2 . . ≤ z ≤ . COMPOSITE SPECTRA3.1.
Stacking Analysis
We measure the gas-phase metallicity of galaxies, us-ing the N2 ( [NII] λ α ) line ratio. Due to the prevalenceof [NII] λ / N < α luminos-ity. In each mass-bin, we bin the normalized spectrumwith a resolution of 0 . . f ( λ ) stacked = (cid:88) i ˜ f i ( λ ) σ i (cid:88) i σ i , (1)where ˜ f i ( λ ) is the flux density of each normalized spec-trum, σ i is its corresponding standard deviation, and˜ f ( λ ) stacked is the composite spectrum with the uncer-tainty of (cid:115) / ( (cid:88) i σ i ) in each mass-bin. The proto-cluster and field galaxies are also bootstrap resampled100 times to take into account the sample variance. Toperform the bootstrap resampling, we draw a randomsample of galaxies from the original sample consideringreplacement. The replacement allows us to have a ran-dom sample that may include some duplicate membersfrom the original sample, or may not contain some ofthe galaxies from the initial sample. This process is re-peated 100 times and each time we end up having newstacked spectra. The sample variance is calculated usingthe standard deviation of these 100 trials.3.2. Metallicities
Since only a few field galaxies are observed in the H band ( ∼
10% of the field galaxies), we utilize rest-frameoptical emission line in the observed K band to mea-sure the gas-phase metallicities. This is done by mea-suring the best-fit [NII] λ α emission line in- Table 1.
The stellar mass range of each mass bin in thestacking processSample log M ∗ M (cid:12) Number of galaxiesField [9.5,9.7) 8[9.7,10) 8[10,10.8] 8Protocluster [9.5,9.8) 7[9.8,10.5) 6[10.5,11.2] 6 tensities and estimating the gas-phase metallicity usingthe N2 ( [NII] λ α ) line ratio.To measure the line fluxes and their uncertainties forthe stacked spectra, we use the same methodology de-scribed in Section 2.2 for the individual spectra. Sincethe stacked spectra are normalized by the H α luminosity,the area underneath the [NII] λ (cid:104) [NII] λ α (cid:105) .In order to measure the gas-phase oxygen abundance(12 + log(O / H)) of galaxies, we employ the empiricalcalibration from Pettini & Pagel (2004), which is basedon electron temperature measurements in the local HIIregions, and is given by 8 . .
57 log(N2). RESULTS4.1.
SFR- M ∗ Relation
Different studies have shown that the fundamentalmetallicity relation (FMR) exists for galaxies at z ∼ α SFRs(corrected for extinction as mentioned in section 2.3) asa function of stellar mass.An important issue in the SFR estimation using thesetwo methods is that there are large uncertainties in bothmeasurements. This will not allow us to rule out anysmall environmental dependence of the main sequence,if exists. For a better comparison, we also add the best-fit line from Koyama et al. (2013) in the right panel ofFigure 2. Our result is in agreement with theirs, show-ing no significant environmental dependence of the mainsequence galaxies for our star forming sample.4.2.
Mass-Metallicity Relation
ZR of a protocluster at z ∼ . . . . . log(M ∗ / M (cid:12) ) . . . . l og ( S F R S E D [ M (cid:12) y r − ] ) Best-fit(Field)Best-fit(Protocluster)FieldProtocluster . . . . log(M ∗ / M (cid:12) ) . . . . l og ( S F R H α [ M (cid:12) y r − ] ) Koyama+13 (Field)Koyama+13 (Cluster)
Figure 2.
SFR of the field (blue circles) and protocluster (red circles) galaxies as a function of their stellar masses.
Left : TheSFRs are calculated from SED fitting.
Right : The SFRs are determined using the H α luminosity of galaxies corrected for thedust extinction based on their stellar masses. The solid lines in each panel show the best-fit to the data points. The slope of theprotocluster best-fit line (solid red lines) is fixed at the slope of the best-fit line for field galaxies (solid blue lines). The dashedblue (red) line shows the best-fit for the field (cluster) galaxies from Koyama et al. (2013). In this section, we study the MZR for the field andprotocluster samples to investigate the role of the envi-ronment in the gas-phase metallicity of galaxies at fixedM ∗ . As discussed in Section 3.1, we divide the sampleinto three stellar mass bins with equal number of galax-ies in each bin. Figure 3 shows the MZR for the stackedspectra and individual galaxies. The metallicities of thestacked spectra in three stellar mass bins are shown inblue (red) squares for field (protocluster) galaxies. Thestellar mass uncertainty in the stacked data points shows1 σ scatter in each bin.As shown in Figure 3, in the stellar mass range10 . M (cid:12) (cid:46) M ∗ (cid:46) . M (cid:12) , the average protoclustergalaxies have a relatively lower metallicity than the aver-age field galaxies by ∼ . ∗ < . M (cid:12) , do notsignificantly depend on the environment. Moreover, inthe massive end of the MZR (M ∗ > . M (cid:12) ), due tothe small number of field galaxies, we cannot draw ro-bust conclusions on the MZR variation between the fieldand protocluster galaxies. In the following section, wematch the stellar mass distributions of protocluster andfield samples to properly isolate the effect of stellar massfrom galaxy environment.4.3. The Mass-Matched Samples
It is known that protoclusters often host more mas-sive galaxies compared to the field. Therefore, to havea reliable comparison between the metallicity of proto-cluster and field galaxies at fixed stellar mass, the twosamples should have similar stellar mass distributions.Otherwise, any change in the MZR may be attributedto the differences in stellar mass distributions. The left panel in Figure 4 shows the stellar massdistribution for the field and protocluster galaxies. Itis clear that, for protocluster galaxies in the massiveend of the distribution, there is no analog of the fieldgalaxy, resulting in a biased comparison between fieldand protocluster samples. We resolve this by construct-ing the mass-matched samples. Similar to Chartabet al. (2021), we match the stellar mass distributionsof field and protocluster galaxies with the resolution oflog(M ∗ / M (cid:12) ) = 0 . ∗ / M (cid:12) ) = 0 . . M (cid:12) ≤ M ∗ ≤ . M (cid:12) , but just two field galaxiesare in this range of stellar mass. Thus, to take into ac-count all the galaxies that reside in each mass-matchedregion, we randomly subsample galaxies 500 times andeach time, we construct the stacked spectra and perturbthem based on their uncertainties. The stacking pro-cess is the same as described in Section 3.1, where weconsider both measurement errors and sample variance.The average and standard deviation of 500 trials corre-spond to the composite spectrum and its error, respec-tively. The right panels in Figure 4 show the compos-ite spectra for the mass-matched sample of protoclusterand the field galaxies in two stellar mass bins 10 . M (cid:12) ≤ M ∗ < . M (cid:12) , and 10 . M (cid:12) ≤ M ∗ ≤ . M (cid:12) . Sattari et al. . . . . . . . ∗ / M (cid:12) )7 . . . . . + l og ( O / H ) N ProtoclusterFieldN2 Upper Limit ProtoclusterN2 Upper Limit FieldStacked ProtoclusterStacked Field
Figure 3.
MZR for the field (blue circles) and protocluster (red circles) galaxies at z ∼ .
2, without controlling for the stellarmass distribution (mass-matching). The 1 σ upper-limits for galaxies with undetected [NII] λ σ scatter in the stellar mass of each bin. We report the gas-phase metallicities of the field andprotocluster galaxies for the mass-matched samples inTable 2. Red (blue) squares in Figure 5 (top panel)show the MZR for the stacked protocluster (field) sam-ple in two stellar mass bins 10 . M (cid:12) ≤ M ∗ < . M (cid:12) ,and 10 . M (cid:12) ≤ M ∗ ≤ . M (cid:12) . The metallicity esti-mates for all the galaxies in the mass-matched samplesare also included in the figure. At z ∼ .
2, the proto-cluster galaxies in the massive end of the MZR are metaldeficient by 0 . ± .
04 (2 . σ significance) dex comparedto those residing in the field. However, this deficiency isnot significant ( < σ ) in the lower mass bin (0 . ± . Table 2.
Gas-phase metallicities of the stacked spectra forthe mass-matched samples. The second and third columnsshow stellar mass ranges and the mean value of mass in eachrange, respectively.Environment log M ∗ M (cid:12) (cid:104) log M ∗ M (cid:12) (cid:105) (cid:104)
12 + log(O / H) (cid:105) Field [9.5,9.9) 9 . ± .
04 8 . ± . . ± .
12 8 . ± . . ± .
03 8 . ± . . ± .
12 8 . ± . Comparison with Literature
To compare our results with those in the literature, weshow the offset between the average gas-phase metallic-ity of the protocluster (galaxies in overdense regions)and the field galaxies from different studies (includingthe present work) as a function of stellar mass in Figure5 (bottom panel).We emphasize that gas-phase metallicity is calibratedlocally and its absolute value at high redshift could beuncertain (Steidel et al. 2014; Shapley et al. 2019). How-ever, when we study relative metallicity and estimate thedifference of metallicities between field and protoclustergalaxies, the calibration effect is not a concern.Moreover, consideration of selection biases is essen-tial in measuring the MZR (Stott et al. 2013). Differentselection criteria for protocluster and field samples re-sult in an unreliable comparison between their respectivemetallicities. As both protocluster and field samples inthe present work are H α -selected, the metallicity offsetdoes not suffer from such selection biases.Kulas et al. (2013) studied the MZR of a protoclustersample at z = 2 . ∗ ∼ M (cid:12) ), they found an ZR of a protocluster at z ∼ . Figure 4.
Left : The stellar mass distributions for the field (blue) and protocluster (red) galaxies. The green hatched regionshows the matched stellar mass distributions of protocluster and field galaxies.
Right : Composite spectra for the mass-matchedsamples in two bins of stellar mass (10 . M (cid:12) ≤ M ∗ < . M (cid:12) , and 10 . M (cid:12) ≤ M ∗ ≤ . M (cid:12) ). The stacked spectra ofprotocluster galaxies are shown in red and the stacked spectra of field galaxies are shown in blue. offset of 0 .
15 dex between the metallicity of protoclusterand field galaxies, i.e., their field sample is more metaldeficient than the protocluster. Our result is in contrastwith their findings, possibly due to the fact that theydid not employ mass-matched samples.Also, in the stellar mass range covered in this paper,Shimakawa et al. (2015) found metallicity enhancement( ∼ .
15 dex) for two protoclusters at z = 2 . z = 2 . z = 2 .
2. A part of this enhancement can be causedby different selection criteria they used for protoclusterand field samples (Their protoclusters are narrow-bandselected, but the field sample from Erb et al. (2006) isUV-selected). Comparing our field sample with the UV-selected sample of Erb et al. (2006), we notice that inthe low-mass end of the MZR, the narrow-band selectedsample has systematically higher metallicity comparedto the UV-selected sample. Thus, the disagreement be-tween the present work and Shimakawa et al. (2015) canbe originated from selection biases.Kacprzak et al. (2015) found no significant differencebetween the MZR of a protocluster at z = 2 and afield sample at the same redshift. On the other hand,Valentino et al. (2015) found that a protocluster sam-ple at z ∼ ∗ ∼ . M (cid:12) is 0 .
25 dex metaldeficient compared to field galaxies at the same red-shift. Their results are in qualitative agreement withthe results in this study; however, we find ∼ . z ∼ .
2. Moreover, Chartab et al. (2021) recently stud-ied the environmental dependence of the MZR for a sam-ple of H-band selected galaxies in the MOSDEF surveyat 1 . ≤ z ≤ .
61. For a mass-matched sample in theredshift range 2 . ≤ z ≤ .
61, they found ∼ .
07 dex metal deficiency for galaxies in overdense regions com-pared to field galaxies. As shown in the bottom panelof Figure 5, our results are in agreement with their find-ings. Additionally, they found that this metal deficiencyincreases by the stellar mass, which is also seen in ourresult in Figure 5 (top panel). SUMMARY AND DISCUSSIONIn this paper, we studied the mass-metallicity relationfor 19 galaxies in a spectroscopically confirmed proto-cluster at z ∼ . [NII] λ α ratio to measure thegas-phase metallicity of these galaxies. After matchingthe stellar mass distributions of field and protoclustersamples, we found that the protocluster galaxies with10 . M (cid:12) ≤ M ∗ ≤ . M (cid:12) are 0 . ± .
04 dex (2 . σ sig-nificance) metal deficient in comparison to field galaxiesat the same redshift. However, this metal deficiency isnot significant for low-mass galaxies.Darvish et al. (2020) predicted that this protoclusterwill grow to a Coma-type cluster with ∼ × M (cid:12) at z = 0. Dekel & Birnboim (2006) found that at z (cid:46) halo (cid:38) M (cid:12) will be dominated with hot-mode accretion. However, as we go to higher redshifts,cold streams can penetrate massive halos from filamentshosting dense pristine gas (Kereˇs et al. 2005). Basedon halo mass evolution trajectories of Behroozi et al.(2013b), we estimate that the progenitor of this proto-cluster at z (cid:38) . halo (cid:46) . M (cid:12) , where theprotocluster is in a phase that hosts cold streams inhot media (Dekel & Birnboim 2006). Therefore, the ob-served protocluster at z = 2 . ∼ . z ≥ . Sattari et al. . . . . . log(M ∗ / M (cid:12) ) . . . . . . + l og ( O / H ) N ProtoclusterFieldN2 Upper Limit ProtoclusterN2 Upper Limit FieldStacked ProtoclusterStacked Field . . . . log(M ∗ / M (cid:12) ) − . − . − . . . . . ∆ l og ( O / H ) c l u s t e r − fi e l d This work ( z = 2 . z = 2 . z = 2 . z = 2 . z = 2 . . . ≤ z ≤ . Figure 5.
Top : The MZR for the mass-matched samples at z ∼ . λ . M (cid:12) ≤ M ∗ < . M (cid:12) , and 10 . M (cid:12) ≤ M ∗ ≤ . M (cid:12) are also shown in red and bluesquares, respectively. The horizontal error bars in the stacked data show 1 σ scatter in stellar mass of each bin. Bottom : Thecompilation of the difference (offset) between gas-phase metallicity of protocluster and field galaxies as a function of stellar massfrom literature. The red data points show the offset for the mass-matched samples used in this work in two stellar mass bins.The uncertainties in the offsets include errors from gas-phase metallicities of both protocluster and field samples.