Accelerated galaxy growth and environmental quenching in a protocluster at z=3.24
Ke Shi, Jun Toshikawa, Kyoung-Soo Lee, Tao Wang, Zheng Cai, Taotao Fang
DDraft version February 15, 2021
Typeset using L A TEX twocolumn style in AASTeX63
Accelerated galaxy growth and environmental quenching in a protocluster at z=3.24
Ke Shi, Jun Toshikawa,
2, 3
Kyoung-Soo Lee, Tao Wang, Zheng Cai, and Taotao Fang Department of Astronomy, Xiamen University, Xiamen, Fujian 361005, China Institute for Cosmic Ray Research, The University of Tokyo, Kashiwa, Chiba 277-8582, Japan Department of Physics, University of Bath, Claverton Down, Bath, BA2 7AY, UK Department of Physics and Astronomy, Purdue University, 525 Northwestern Avenue, West Lafayette, IN 47907, USA Institute of Astronomy, University of Tokyo, 2-21-1 Osawa, Mitaka, Tokyo 181-0015, Japan Department of Astronomy, Tsinghua University, Beijing 100084, China (Received June 1, 2019; Revised January 10, 2019; Accepted February 15, 2021)
Submitted to ApJABSTRACTWe present a multiwavelength study of galaxies around D4UD01, a spectroscopically confirmedprotocluster at z = 3 .
24 to investigate environmental trends. 450 galaxies are selected based on K S band detection with photometric redshifts (photo- z ) at 3 . < z < .
4, among which ∼
12% are classifiedas quiescent galaxies. The quiescent galaxies are among the most massive and reddest ones in the entiresample. We identify a large photo- z galaxy overdensity in the field, which lies close to the previouslyspectroscopically confirmed sources of the protocluster. We find that the quiescent galaxies are largelyconcentrated in the overdense protocluster region with a higher quiescent fraction, showing a sign ofenvironmental quenching. Galaxies in the protocluster are forming faster than the field counterpartsas seen in the stellar mass function, suggesting early and accelerated mass assembly in the overdenseregions. Although weak evidence of suppressed star-formation is found in the protocluster, the statisticsare not significant enough to draw a definite conclusion. Our work shed light on how the formationof massive galaxies is affected in the dense region of a protocluster when the Universe was only 2 Gyrold. Keywords: cosmology: observations – galaxies: clusters: general – galaxies: evolution – galaxies:formation – galaxies: high-redshift INTRODUCTIONIt is well known that local environments have pro-found impacts on the formation and evolution of galax-ies. At low redshift (e.g., z (cid:46) z ∼ Corresponding author: Ke [email protected] ally experienced a short and intense star-formation andquenched very quickly afterwards (e.g., Stanford et al.1998; Thomas et al. 2005; Snyder et al. 2012; Mart´ın-Navarro et al. 2018).In order to further investigate the environmental im-pacts on galaxy formation and understand the detailedquenching mechanism of cluster galaxies, we need todirectly observe the progenitors of clusters (‘protoclus-ters’) and study their galaxy constituents at high red-shift. Many studies have shown that star formation ac-tivities in dense environments are enhanced relative tothe field at high redshift (e.g., Elbaz et al. 2007; Cooperet al. 2008; Tran et al. 2010; Koyama et al. 2013; Al-berts et al. 2014; Cai et al. 2017; Shimakawa et al. 2018;Lemaux et al. 2020), suggesting protocluster galaxiesmay have undergone accelerated mass assembly thanthe field counterparts. However, there are also some of a r X i v : . [ a s t r o - ph . GA ] F e b i Shi et al. protoclusters where no such differences are seen as com-pared to the field (e.g., Overzier et al. 2008; Toshikawaet al. 2014; Cucciati et al. 2014; Shi et al. 2019a). Thisdiscrepancy between different studies may rise from dif-ferent galaxy populations they use or the lack of statis-tics due to small sample size (Cucciati et al. 2014), or be-cause of different evolutionary stages in which these pro-toclusters are observed (Overzier et al. 2008; Toshikawaet al. 2014).Despite the above challenges in identifying the rever-sal of ‘star formation-density’ relation in protoclusters,a growing number of studies suggest that protoclustersoften host a larger fraction of massive red galaxies thathave already quenched their star-formation (e.g., Steidelet al. 2005; Kubo et al. 2013; Lemaux et al. 2014, 2018;Zavala et al. 2019; Shi et al. 2019a; Ando et al. 2020).These studies suggest that the cluster red sequence ob-served in the local Universe (Visvanathan & Sandage1977; Bower et al. 1992; Stott et al. 2009) may havealready been formed in protoclusters at earlier epochs.This is further supported by the numerical simulationof Chiang et al. (2017), who proposed an “inside-out”galaxy growth in protoclusters from z >
10 to z ∼ . < z < (cid:48) -30 (cid:48) in the sky (Chiang et al.2013; Muldrew et al. 2015), making it difficult and ob-servationally expensive to conduct a systematic search.While several tens of protoclusters have been spectro-scopically confirmed to date (see e.g., Overzier 2016;Harikane et al. 2019, for a summary), many were iden-tified by pre-selecting overdense regions traced by star-forming galaxies such as Lyman break galaxies (LBGs)or Lyman alpha emitters (LAEs), followed up by spec-troscopic confirmation.Toshikawa et al. (2016) discovered a protocluster inthe D4 field of the Canada-France-Hawaii-TelescopeLegacy Survey (CFHTLS) at z = 3 .
24. This proto-cluster, dubbed ‘D4UD01’ hereafter, was initially iden-tified using u -dropout selected LBGs at z ∼
3. A sig-nificant surface overdensity (4.4 σ ) of LBGs was foundin the field, implying the presence of a large structure.Follow-up spectroscopy has confirmed five galaxies at z = 3 .
24 within 2 Mpc (physical) with one another.Further comparison with simulation also suggested thatit will become a virialized cluster at z = 0 (Toshikawa et al. 2016). However, LBGs are star-forming galax-ies that severely suffered from projection effects due tolarge redshift uncertainties, therefore it is difficult toconduct a systematic study of environmental impacts inthis protocluster using only LBGs. To better character-ize the role of the environments, a detailed census of itsgalaxy constituents is needed. In this work we performa multiwavelength study of galaxies in and around thisprotocluster with the help of photometric redshift, aim-ing to further unveil the environmental trends in thisprotocluster.This paper is organized as follows. In Section 2 wedescribe the data and methods used to select the pro-tocluster galaxy candidates. Massive quiescent galaxiesare selected in Section 3. In Section 4 we investigate thespatial distributions of galaxies in the field and iden-tify an overdensity which is defined as the protoclus-ter region. The sky distribution of quiescent galaxies isalso studied to seek possible environmental trends. Wefurther compare the GSMFs and star-formation ratesof protocluster galaxies with the field counterparts inSection 5. We summarize our results in Section 6.Throughout this paper we use the WMAP9 cosmology(Ω M = 0 . , Ω Λ = 0 . , σ = 0 . , h = 0 .
69) from Hin-shaw et al. (2013). All magnitudes are given in the ABsystem (Oke & Gunn 1983). Distance scales are givenin comoving units unless noted otherwise. DATA AND ANALYSIS2.1.
Data and photometry
In this study, we make use of publicly available mul-tiwavelength data including the deep optical ugriz im-ages from the CFHTLS Deep Servey (Gwyn 2012) andthe near-IR
JHK S bands from WIRCam Deep Survey(WIRDS) (Bielby et al. 2012). We also use the Spitzer-IRAC 3 . µ m and 4 . µ m data from the NMBS-II IRACsurvey (Annunziatella et al. 2018). The pixel scale ofCFHTLS and WIRDS images is 0.186 (cid:48)(cid:48) while for IRACdata it is 0.558 (cid:48)(cid:48) . The photometric depths of CFHTLSand WIRDS data are measured from the sky fluctua-tions by placing 2 (cid:48)(cid:48) diameter apertures in random imagepositions while the depths of IRAC data are measuredwithin 3 (cid:48)(cid:48) aperture. Table 1 summarizes the data sen-sitivity and image quality in this paper. It is notedthat the IRAC data is fairly shallow (maximum expo-sure time of only ∼ u band data. To do iiiso, the radial profile of the PSF in each image is ap-proximated by a Moffat function with the measured see-ing FWHM. A noiseless convolution kernel between thelow and high-resolution images is then derived using theRichardson-Lucy deconvolution algorithm (Richardson1972). Each image is convolved with its respective ker-nel to match the PSF of the u band data.The WIRDS survey does not cover the entire 1 ◦ × ◦ D4 field (60% of the field has no data, see Bielby et al.(2012)), but fortunately D4UD01 is located within itscoverage. For source detection in this work, we use the K S band which samples the rest-frame optical emissionat z = 3 .
24, enabling the measurement of stellar massesof the galaxies. We trim the K S image to contain theregion receiving >
50% of the maximum exposure time,which result in a final area of 497 arcmin . All the otherbands are also trimmed to match the K S image.Source detection and photometric measurements inthe ugrizJHK S bands are carried out by running theSExtractor software (Bertin & Arnouts 1996) in dualmode on the PSF matched images with the K S im-age as the detection band. The SExtractor parameterMAG AUTO is used to estimate the total magnitude,while colors are computed from fluxes within a fixedisophotal area (i.e., FLUX ISO). As the images are PSFmatched, aperture correction in all bands is assumed tobe the difference between MAG AUTO and MAG ISOmeasured in the detection band. For sources not de-tected in certain bands, we use the 2 σ limiting magni-tude to give the upper limits.As for the IRAC images, since the PSFs of these im-ages are much broader ( ≈ (cid:48)(cid:48) ), source blending onthese images is a severe problem. In order to obtain ac-curate and unbiased measurement of fluxes and colors onthe IRAC images, we utilize the T-PHOT software (Mer-lin et al. 2015, 2016). T-PHOT performs “template-fitting” photometry on the low-resolution image usingthe information of high-resolution image and catalog. Inour case, the K S band image and catalog are used as theinput priors of T-PHOT while the low-resolution IRACimages are analysized to obtain corresponding photome-try. It is noticed that although T-PHOT is prior-based,the derived photometry of the low-resolution image doesnot strongly depend on which high-resolution image weuse as the input. For example, if we use the i band asthe input high-resolution prior to derive the 3.6 µ m pho-tometry, the resultant 3.6 µ m magnitude differences ascompared to the K S based have only an average valueof ∼ .
04. This confirms that our T-PHOT derived pho-tometry is not significantly biased by the prior.Finally, all photometric catalogs are combined to-gether to make a multiwavelength catalog. In this work, we focus on the sources with K S magnitudes smallerthan 24.29 (i.e., > σ detection limit). In the end 20,801sources are selected in the final catalog. Table 1.
Data SetBand Instrument Limiting magnitude (cid:63)
FWHM(5 σ ,AB) ( (cid:48)(cid:48) ) u MegaCam/CFHT 27.02 0.90 g MegaCam/CFHT 27.48 0.80 r MegaCam/CFHT 27.11 0.70 i MegaCam/CFHT 26.72 0.70 z MegaCam/CFHT 25.84 0.70 J WIRCam/CFHT 24.83 0.60 H WIRCam/CFHT 24.33 0.60 K S WIRCam/CFHT 24.29 0.603.6 µ m IRAC/Spitzer 22.27 1.864.5 µ m IRAC/Spitzer 22.29 1.75 (cid:63) σ limiting magnitude measured in a 2 (cid:48)(cid:48) diameter aperture forthe CFHT data, while for the Spitzer data the depths are mea-sured in a 3 (cid:48)(cid:48) aperture. Photometric Redshift and Spectral EnergyDistribution Fitting
The photometric redshift and the physical propertiesof each source in the catalog are derived via the spec-tral energy distribution (SED) fitting technique usingthe CIGALE software (Noll et al. 2009; Boquien et al.2019). Based on an energy balance principle (the en-ergy emitted by dust in the mid- and far-IR exactlycorresponds to the energy absorbed by dust in the UV-optical range), CIGALE builds composite stellar pop-ulation models from various single stellar populationmodels, star formation histories, dust attenuation laws,etc. The model templates are then fitted to the ob-served fluxes of galaxies from far-ultraviolet to the radiodomain, and photometric redshift as well as physicalproperties are estimated using a Bayesian analysis.For the SED templates, we use the stellar populationsynthesis models of Bruzual & Charlot (2003) , Calzettiet al. (2000) reddening law with E(B-V) values rangingfrom 0 to 2 in steps of 0.1 mag, the solar metallicity, andChabrier (2003) initial mass function. We use the de-layed star formation history (SFR ∝ t × exp[-t/ τ ]) withstar-forming time scale τ ranging from 0.1 to 10 Gyr.Nebular emission is also included and dust emission ismodeled by Dale et al. (2014). The input redshifts areset to be between 0.1 and 5.0 in steps of 0.1.v Shi et al.
To calibrate the photometric redshift (photo- z ), weuse a spectroscopic redshift (spec- z ) sample obtainedfrom the AAOmega instrument on the Anglo-AustralianTelescope (AAT) targeting X-ray point sources in theD4 field (Stalin et al. 2010). The sample has 1,809 spec- z sources with the majority of them lying at z < z sources with our photo- z catalog and find 191 counter-parts. The precision of the photo- z is measured usingthe normalised median absolute deviation (Hoaglin et al.1983) defined as σ z = 1 . × median( | ∆ z | /(1+ z spec )),where ∆ z = z spec − z phot . This scatter measurementcorresponds to the rms of a Gaussian distribution and isnot affected by catastrophic outliers (i.e., objects with | ∆ z | /(1+ z spec ) > .
15) (Ilbert et al. 2006; Laigle et al.2016). For these sources, we obtain σ z = 0 .
08. Thenumber of catastrophic failures take up to 10% of allthe sources.The mean photo- z error derived by CIGALE is ∆ z ∼ . z measurements of 3 . < z phot < . z = 3 .
24) lies within the coverage of thephoto- z error. Only three objects among these candi-dates have spec- z information. One ( z phot = 3 .
1) isin the AAOmega sample that indicates it is a QSO at z spec = 3 .
03, and we remove this object in our candi-date list. The other two ( z phot = 3 . .
0) are thespectroscopically confirmed LBGs at z spec = 3 .
24 and3 .
04 in Toshikawa et al. (2016). The remaining 14 spec- z sources in Toshikawa et al. (2016) are not detected in K S band, therefore not in our photo- z sample. For the782 sources, we remove the ones that have SExtractorparameter “CLASS STAR” greater than 0.9 to reducethe contamination of stars. We then visually inspect theremaining sources and remove those with potential con-tamination in the photometry, including those severelyblended with nearby bright sources. We also discardsources that are detected in less than four bands. In theend, 450 galaxies are selected as our photo- z galaxies.We fix the best-fit photo- z of the 450 galaxies andrefit their SEDs using CIGALE with the same config-uration to determine their physical properties such asstellar mass, star formation rate (SFR) and color excessof stellar continuum E(B-V), etc. The typical (median)error of stellar mass is ∼ . ∼ . z galaxies, we also estimate theirstellar mass completeness following an empirical ap-proach (Pozzetti et al. 2010; Ilbert et al. 2013; Laigleet al. 2016). For each galaxy, we compute the lowest stellar mass M lim it would need to be detected at thegiven K S magnitude limit K S lim = 24 . M lim ) = log( M ) − . K S lim − K S ) , and the stellar mass completeness limit corresponds tothe mass under which 90% of the galaxies lie. The cal-culated mass completeness limit is log( M lim ) = 10 . z sample. We also notice that only 193 (43%)of the photo- z galaxies satisfy the LBG criteria definedin Toshikawa et al. (2016). LBGs are believed to beyoung star-forming galaxies with typical stellar massesof 10 M (cid:12) (e.g., Giavalisco 2002). Therefore our sam-ple includes a large fraction of massive galaxies that arecommonly missed from the UV-selected LBGs, which ishelpful in studying the high-mass end of the stellar massfunction.We also consider possible contamination in our sam-ple. The photo- z galaxies lie around at z ≈ . K S band photometry could be potentially contami-nated by the [O iii ] λλ iii ] equivalent widths (EWs)for a sample of 3 . < z < . z ≈ .
2, this leads toan overestimate of K S band continuum flux density by0.3 magnitude. However, Malkan et al. (2017) noticedthere is an anti-correlation between the stellar mass and[O iii ] EW for LBGs at z ∼
3: galaxies with higherstellar masses usually have smaller EWs. According totheir relation, 98% galaxies in our sample with masses > M (cid:12) have typical EWs less than 100 ˚A, corre-sponding to a flux contamination smaller than 0.1 mag-nitude. More recently, Yuan et al. (2019) further inves-tigated the impact of including [O iii ] nebular emissiondata in the SED fitting analysis using a sample of LBGsat z ∼ .
5. They found an average discrepency of only ∼ . z sampleis also ∼ . SELECTION OF QUIESCENT GALAXYCANDIDATESThe presence of quiescent galaxies at high redshift cangive us valuable insight into how current-day massiveellipiticals obtain their masses. One of the main focusof this paper is to identify and study evolved galaxypopulations in the protocluster field. To do so, a reliable vmethod to separate quiescent galaxies from star-forminggalaxies is required.Various methods have been developed to classify qui-escent galaxy populations in the literature. Perhaps themost well-known method is using the rest-frame U − V vs. V − J color-color diagram ( U V J diagram), wherethe galaxy distributions are bimodal and a color cutcan be applied to separate the two populations (e.g.,Labb´e et al. 2005; Williams et al. 2009; Brammer et al.2011; Muzzin et al. 2013). Other color criteria have alsobeen proposed, such as the NUV − r vs. r − J (Ilbertet al. 2013) which can alleviate the confusion betweenred dusty star-forming galaxies and passive galaxies, and J − K S vs. [3.6]-[4.5] color in the observed frame thatselect galaxies with a strong Balmer/4000˚A break at2 < z < star )where a larger value typically corresponds to a lowersSFR (Brinchmann et al. 2004).In this work, D4000 index is inferred from the best-fittemplate of CIGALE. The typical error of D4000 valueis ∼ .
04. In the left panel of Figure 1, we show theD4000 distribution as a function of stellar mass. Ascan be seen in the figure, the distribution appears to bebimodal: most of the galaxies are located in the lowerleft corner whereas a fraction are concentrated in theupper right corner. This bimodality has been seen bothat low redshift ( z <
1) (Haines et al. 2017) and at highredshift (up to z ∼
3) (Johnston et al. 2015). Basedon this diagram, we apply a cut at D4000=1.2, whichroughly segregate the two populations, defining galaxies above this limit to be quiescent galaxy candidates. Intotal, 52 galaxies fall into the quiescent galaxy catalog.The right panel of Figure 1 shows the galaxies on theSFR–M star plane where the galaxies are color-coded bytheir D4000 values. There is a clear trend that galax-ies with larger D4000 tend to lie at the lower part ofthe plane. Indeed, all of our quiescent galaxies are lo-cated well below the star-forming main sequence rela-tions at z ∼ ∼ M (cid:12) , which ismost likely due to selection effect. The quiescent galaxycandidates is 90% complete above 10 . M (cid:12) , while theremaining star-forming galaxies has a completeness limitof 10 . M (cid:12) , as calculated using the method in Section2.2. This incompleteness issue also affects our resultsin Section 5. A sample images of the quiescent galaxycandidates can be found in Figure 2.As a final check, in the left panel of Figure 4 we showour quiescent galaxy candidates in the U V J diagram.The rest-frame colors of the galaxies are derived fromthe best-fit templates from CIGALE. For secure deter-mination of rest-frame J band magnitude, we plot onlythe IRAC 3.6 µ m and 4.5 µ m detected sources (i.e., > σ magnitude limits). Among the IRAC detected 29 candi-dates, 16 (55%) are within the quiescent region definedby Muzzin et al. (2013) while the remainder are alsoclose to the quiescent parameter space. This furtherjustifies our usage of D4000 index to select quiescentgalaxies.Figure 3 shows the SED-fitting results for a subsampleof the star-forming and quiescent galaxies defined usingthe above criterion. It can be seen that star-forminggalaxies are featured by their prominant emission lines,and they are less massive ( < M (cid:12) ) and younger thanthe quiescent galaxies which are lack of nebular emis-sions. On the other hand, 80% (42/52) of the quiescentgalaxies have masses greater than 10 M (cid:12) . Among thegalaxies of masses > M (cid:12) , 38% (42/112) are quies-cent, which is similar to the quiescent fraction observedin Kubo et al. (2013) and Ando et al. (2020) in high-redshift protoclusters.It is noteworthy that these quiescent galaxies are veryred with strong Balmer/4000˚A break between the J and K S bands, with a median J − K S = 2 .
0. In fact, 94%(49/52) of the quiescent galaxies have J − K S > . < z < Shi et al. (e.g., Labb´e et al. 2005; Kriek et al. 2006). The rightpanel of Figure 4 shows the color-mass relation for thephoto- z galaxies. The quiescent galaxy candidates areconcentrated in the top-right corner of the plane, sug-gesting they are among the most massive and reddestobjects in the entire photo- z sample.However, we caution that whether all of these galax-ies are truly “red and dead” remains uncertain. In lackof far-IR observations, especially the 24 µ m data, we areunable to quantify the possible emission features frompolycyclic aromatic hydrocarbons (Draine & Li 2007)at rest-frame ∼ . µ m at z = 3 .
24, which are heated byeither dust obscured star-formation or AGNs. What ismore, if some of these galaxies are dust-enshrouded star-forming galaxies, they could be detected at submillime-ter wavelength by ALMA/SCUBA-2 (e.g., Wang et al.2019). At the current stage, without submillimeter ob-servations it is difficult to further investigate this possi-bility. Nevertheless, we notice that Santini et al. (2019)recently analysed in detail 26 candidate quiescent galax-ies observed by ALMA at 3 < z < > σ ) inthe submillimeter wavelength. Given the upper limits ofthe detection and with a stacking analysis, they foundthe dust obsecured star-formation activity is lower thanthat inferred from UV-optical. Meanwhile, using theALMA-derived SFRs, ∼
50% of these galaxies are lo-cated at least 1 σ below the star-forming main sequence.They concluded that their sample is indeed quiescent ina statistical sense. Therefore, we argue that although wecannot completely rule out the contaminants of possibledusty star-forming galaxies, it is very unlikely that thered colors of all these candidates are caused by dust. SKY DISTRIBUTION OF GALAXIES4.1.
Sky Distribution of the Photo- z Galaxies
Protoclusters are usually discovered as overdensities ofgalaxies. In order to identify galaxy overdensities, manystudies smoothed the spatial distributions of galaxieswith a fixed or adaptive kernel to obtain the densitymaps (e.g., Hayashino et al. 2004; Matsuda et al. 2005;Yang et al. 2010; Lee et al. 2014; Shi et al. 2019a,b;Harikane et al. 2019). Alternatively, some studies utilizea scale independent method named Voronoi tessellation(Ramella et al. 2001; Kim et al. 2002; Cooper et al. 2005;Soares-Santos et al. 2011; Dey et al. 2016) to measurethe galaxy overdensity, and proved to be a good estima-tor of underlying density field (Darvish et al. 2015). Inthis study, we use Voronoi tessellation to estimate the 2D surface density of the galaxies, which is described inthe following.A Voronoi tessellation is a unique way of dividing atwo-dimensional distribution of points into convex cells,with each cell containing only one point and a set of ver-tices which are closer to that point than to any other inthe plane. It has the property that the local density ( f )of each cell is the inverse of the cell area ( a ). Thereforeto estimate the overdensity of each cell, one first needsto calculate the average density of the cells in the entireplane ( (cid:104) f (cid:105) = (cid:104) /a (cid:105) ), then the density contrast of eachcell is ˜ f = f / (cid:104) f (cid:105) .The Voronoi tessellation of our photo- z galaxies isshown in the top left panel of Figure 5. We can seethat there is a large overdensity of galaxies in the west-ern end of the field near the five spectroscopically con-firmed LBGs at z = 3 .
24 (Toshikawa et al. 2016). Asour sample is mass-limited and we may miss a lot oflow-to-medium ( ∼ − M (cid:12) ) mass galaxies in thefield that belong to the protocluster, therefore we de-fine the protocluster region in a conservative way, witha circle enclosing all the spec- z sources and most of thehigh-density ( ˜ f >
1) sources nearby. The average den-sity of all the cells in the circle is ˜ f = 1 .
3. The circlehas a surface area of 81 arcmin containing 96 galaxies,and its radius is ∼
10 Mpc, consistent with the typicalprotocluster size at z ∼ z sources within the circle. To verify this, we randomlyput 100 circles with the same radius into the field andcalculate the average density within. There are only fourrealizations in which the average density is comparableto the original one. The centers of these four circles areall very close to the one we used ( < (cid:48) ), and the aver-age densities of the remaining realizations are all below1.3. This confirms our visual impression and indicatesthe one near the spec- z sources is indeed the largestoverdensity in the field. The other small overdensitiescould be coincidental alignment of galaxies along theline of sight that have no physical associations, or theycould belong to smaller structures at 3 . < z < .
4. Atthe current stage, without spectroscopic observations,we cannot determine which case is true, therefore weleave it to future studies. In the remainder of this pa-per, we only regard the one we defined in Figure 5 asthe protocluster region at z = 3 .
24, and define the areaoutside the protocluster as the general field.For comparison and completeness, in the bottom panelof Figure 5 we show the smoothed surface density mapof the u -dropout LBGs as in Toshikawa et al. (2016). We vii log(M star )[M ] D log(M star )[M ] l o g ( S F R )[ M y r ] D Figure 1.
Left:
D4000 values of the photo- z galaxies as a function of stellar mass. A bimodality can be seen in the planefrom the 1 σ and 2 σ contour lines. The red line indicates our selection criterion for quiescent galaxies. The dashed vertical linedenotes the 90% mass completeness of the sample. Right:
SFR–M star relation for the photo- z galaxies color coded by theirD4000 values. The black line is the main-sequence relation calibrated from Speagle et al. (2014) and the green line is that froma semi-analytic model by Dutton et al. (2010). Galaxies with SFR=0 are indicated in the log(SFR)=0 location. u g r i z J H Ks 3.6 4.5 Figure 2.
Example postage-stamp images of quiescent galaxies. All images are 10 (cid:48)(cid:48) on each side. North is up and east is tothe left. do not use Voronoi tessellation for the LBGs since thereare nearly 6,000 sources which would make it difficult toidentify overdense structures. We see there are two sig-nificant overdensities in the field. The one in the northis roughly co-spatial with the five spectroscopic sourcesand the northeastern part of the photo- z overdensity.The southern one is largely outside of the photo- z over- density and the southwestern part of the photo- z over-density disppears in the LBG map. The discrepancybetween the photo- z distribution and LBG distributionis not surprising, as they both have large redshift un-certainties (∆ z ∼ . z ∼ Shi et al. -5 -4 -3 -2 -1 f ν [ m J y ] Star-forming
Quiescent -5 -4 -3 -2 -1 f ν [ m J y ] Observed Wavelength [ µ m ] -5 -4 -3 -2 -1 f ν [ m J y ] Observed Wavelength [ µ m ] Figure 3.
SED-fitting results for a subsample of the photo-z galaxies including star-forming galaxies and quiescent galaxies.The black solid lines are the best-fit model spectra. Filled circles represent the observed fluxes, while triangles denote 2 σ upperflux limits in the case of nondetection. In the inset of each panel, we also show the probability distribution function of thephotometric redshift for each galaxy, and the redshift of the protocluster is shown as a red vertical line. We also list the best-fitphoto- z , log(M star ) (in units of M (cid:12) ), SFR (in units of M (cid:12) yr − ), dust reddening parameter E(B–V), and age (in units of Myr)in the figure. they are selected in different ways (rest-frame optical vs.UV), and many of the UV selected LBGs are not de-tected in K S band (Section 2.2), photo- z galaxies mayrepresent more massive galaxy population that trace dif-ferent underlying large-scale structures than the LBGs.Only future spectroscopic observations can verify thesedifferent scenarios.The descendant mass of D4UD01 is calculated byToshikawa et al. (2016) using a set of lightcone models (Henriques et al. 2012) based on Millennium Simulation(Springel et al. 2005). They matched the observed sur-face density maps with those in the mock catalogs usingthe same selection method, finding a correlation betweenoverdensity of LBGs and its descendant halo mass. Theoverdensity value of D4UD01 is 4.4, which results in adescendant halo mass of 1 . ∼ × M (cid:12) (see theirFigure 7). According to Chiang et al. (2013), this struc- ix V-J U - V log(M star )[M ] J - K s Figure 4.
Left:
Rest-frame
UV J diagram for the IRAC detected sources. The red circles denote the quiescent galaxies definedby D4000 index while the blue circles are the star-forming galaxies. The solid lines are the cut used to define quiescent galaxiesin Muzzin et al. (2013). Note that some galaxies have the same colors as their best-fit galaxy templates are the same (only atdifferent redshifts).
Right:
Mass-color diagram for all the photo- z galaxies. The red circles denote the quiescent galaxies. Thehorizontal line represents the DRG criterion while the dashed vertical line is the stellar mass completeness limit of the sample. ture will evolve into a Fornax-like (1–3 × M (cid:12) ) orVirgo-like (3–10 × M (cid:12) ) cluster at z = 0.4.2. Sky Distribution of the Quiescent Galaxies
Above we have shown that the survey field contains alarge photo- z galaxy overdensity, which is most likely tobe a protocluster. In this section we investigate whetherthere is also presence of quiescent galaxies in this proto-cluster, which may shed light on how quenching of starformation depends on environment.The top right panel of Figure 5 shows the Voronoi tes-sellation of the quiescent galaxies selected in Section 3.It is clear that the quiescent galaxies tend to be con-centrated in the protocluster region: the galaxies withinthe circle have an average density of ˜ f = 1 . .
19 arcmin − . In comparison, thesurface density of all the quiescent galaxies in the entirefield is 0.10 arcmin − (52/497). Thus the number den-sity of quiescent galaxy candidates in the protoclusternearly doubles that in the average field. On the otherhand, the surface density of all galaxies in the protoclus-ter is 1.2 arcmin − , only mildly higher than the surfacedensity of all galaxies in the entire field which is 0.9arcmin − . Therefore the enhanced number of quiescentgalaxies in the protocluster cannot be simply explainedby the overall increased number of galaxies therein. Inaddition, the quiescent fraction in the protocluster is ∼ ∼ A large fraction of massive quiescent galaxies inthis protocluster strongly suggests that cluster galax-ies formed earlier than those in the field, that we maybe witnessing the environmental quenching that takesplace in the early stage of cluster formation long beforevirialization. These quiescent galaxies may have experi-enced an accelerated mass assembly in the high-densityprotocluster environment. We will discuss the environ-mental impacts on the galaxy stellar mass functions andstar-formation activities in the next section, to furtherreveal the possible differences of galaxies’ physical prop-erties caused by the environments. DISCUSSION5.1.
Environmental Dependence on the Galaxy StellarMass Function
We have shown that this protocluster appears to hosta higher fraction of massive quiescent galaxies than inthe field. To further investigate the possible environ- Although we use D4000 to select quiescent galaxies in this work,this conclusion does not depend on specific selection criteria. Forexample, if instead we use sSFR < − yr − as in Fontanotet al. (2009) to define quiescent galaxies, we would have 53 qui-escent galaxies among which 14 are inside the protocluster region.Therefore our results remain nearly the same. Shi et al. d e c Photo-z galaxies ra Quiescent galaxies ra d e c . . . . . . . . . . . . . . . . . . . . LBGs
Figure 5.
Top Left:
Voronoi tessellation of all the photo- z galaxies in the field. The different colored circles respresent galaxieswith different local density contrast (local density/mean density): ˜ f > . . < ˜ f < . . < ˜ f < . f < . z = 3 .
24. The large blue circle denotesthe protocluster region. The dashed lines are associated with boarder cells that have infinite area, which are excluded fromcalculation of the average density.
Top Right:
Voronoi tessellation of the quiescent galaxies.
Bottom left:
LBG density mapsmoothed using a FWHM=6 Mpc Gaussian kernel, and the contour labels show surface density levels relative to the field. mental trends in detail, we calculate the galaxy stellarmass function (GSMF) of the photo- z galaxies in thissection.As our survey volume is not large enough, cosmic vari-ance (CV) might be a severe issue that affects the un-certainty of the number count. We use the Cosmic Vari-ance Calculator (Trenti & Stiavelli 2008) to account forthe CV, and the final error of each mass bin includesboth Poission noise and the corresponding CV. To calcu-late the volume, we use the comoving volume at redshift z = 3 . − . z galax-ies including the protocluster and field galaxies. Forcomparison, we also plot the GSMFs from Caputi et al.(2011) and Davidzon et al. (2017) at the similar redshiftrange 3 . < z < . z galaxies.Above the mass completeness limit, our total GSMFagrees relatively well with Caputi et al. (2011). Whilethe GSMF of Davidzon et al. (2017) has a faster de-cline at high-mass end, which was also noticed in David-zon et al. (2017) and was attributed to cosmic varianceor difference in the photo- z calculation. If we com-pare galaxies within our own sample, the selection ef-fects could largely be ignored and our analysis would bemore robust. From Figure 6 we can see that protoclus-ter galaxies appear to have increased number densitythan the field both at low-mass end ( (cid:46) . M (cid:12) ) and xiat high-mass end ( (cid:38) M (cid:12) ). While at stellar massbetween 10 . M (cid:12) and 10 M (cid:12) there is a sudden num-ber drop of protocluster galaxies, making them almostindistinguishable from the field. This declined numberdensity of protocluster galaxies at medium mass rangemight be attributed to the incompleteness of the qui-escent galaxy population. In Section 4.2 we see thatthe protocluster region hosts a higher fraction of quies-cent galaxies than the average field. Under this circum-stance, we suspect that if quiescent galaxies are pre-ferrably located in the protocluster, many would not bedetected in our study below the mass completeness limitof 10 . M (cid:12) (Section 3), resulting a ‘dip’ in the mediummass range. If this is the case, it will imply an overallaccelerated galaxy growth in the protocluster.The above results suggest that we are witnessing anaccelerated mass assembly in D4UD01. The fact thatthe protocluster hosts a higher fraction of quiescentgalaxies also indicates a sign of environmental quench-ing. In addition, if the lack of medium mass galaxies inD4UD01 is due to the incompleteness of quiescent galax-ies, the quiescent fraction will be even higher than thatcalculated in Section 4.2, making the quenching moreeffecient in D4UD01 than the field.Similar trends have also been found in many otherprotocluster studies. For example, Lemaux et al. (2014,2018) discovered two protoclusters at z = 3 .
29 and z = 4 .
57 in the VIMOS Ultra-Deep Survey using spec-troscopic observations. They found that these proto-clusters tend to have an excess of more red and mas-sive galaxies ralative to the coeval field. Recently, Andoet al. (2020) searched for protocluster cores using pairsof massive galxies at z ∼ Environmental Impacts on Physical Properties ofGalaxies
In this section, we compare the physical properties ofgalaxies in and out of the protocluster, to further dis-cern possible environmental dependence on galaxy prop-erties.In Figure 7 we show the photo- z galaxies on the SFR-M star plane grouped by different environments. First,using the Kolmogorov-Smirnov (K–S) test, no signifi-cant differences ( p -values > star are found. It is possiblethat the large photo- z uncertainty dilute the signal ofpotential differences in galaxy properties. The lack ofmedium mass galaxies in the protocluster is likely dueto the combination of selection effect and environmentalquenching as discussed earlier, which results in a lowermedian mass than the field galaxies. On the other hand,we notice that the SFRs of protocluster galaxies appearto be skewed towards lower values than the field counter-parts, as can be seen in the histogram, suggesting pos-sible suppresion of star-formation activites. This is alsoconsistent with our previous findings of higher abun-dance of quiescent galaxies in this protocluster. Never-theless, overall this trend is too weak to be recognized inthe K-S test, and we tend to not give a definite conclu-sion here but leave it to future study when precise spec-troscopic observations on this protocluster are available.Many studies showed that the star-formation activ-ities are enhanced in dense protocluster environments(e.g., Koyama et al. 2013; Hayashi et al. 2016; Shi-makawa et al. 2018; Ito et al. 2020; Shi et al. 2020). Al-though these findings shed light on the possible reversalof “star formation–density” relation in some protoclus-ters, there are many other protoclusters where no suchdifferences are seen. For example, Cucciati et al. (2014)studied a protocluster at z = 2 . z = 3 .
29 and z = 4 .
57 in the VI-MOS Ultra-Deep Survey (Lemaux et al. 2014, 2018). Inaddition, Shi et al. (2019a) analyzed a protocluster at z = 3 .
78 using the similar photo- z technique as in thiswork, and found no significant environmental impactson star-formation activities.Although these different results sometimes appear tobe contradictory, we argue this could be likely due tothe different evolutionary stages and dynamical statesthese protoclusters are undergoing. For those proto-clusters where enhancement of star-formation activi-ties have been found, they could be experiencing anearly mass assembly and accelerated structure forma-ii Shi et al. tion (Steidel et al. 2005), resulting in the enhancement ofstar-formation we observed. While for those protoclus-ters where no such trends are spotted, they may havealready passed the peak era of their star-formation orstill in the early phase of formation before the emergenceof any environmental effects (Toshikawa et al. 2014). Inthis context, the differences observed in different pro-toclusters could originate from the so-called “halo as-sembly bias”, in a sense that the properties of galaxiesdepend not only on the mass of the halo they reside in,but also on the halo formation time (e.g., Gao et al.2005; Wechsler et al. 2006; Li et al. 2008; Zentner et al.2014). It is noteworthy that Shi et al. (2019b) and Shiet al. (2020) conducted a detailed study of a massiveprotocluster at z = 3 .
13, finding the protocluster con-sists of two disjoint structures where one contains mostlylow-mass star-forming galaxies while the other hosts alarge fraction of massive quiescent and/or dusty galax-ies. They also found that the former has a more en-hanced star-formation activity than the latter while thelatter is more similar to the field. All in all, these studiessuggest that D4UD01 may be a more evolved structurethat have already passed the peak of its star-formationera, so that we do not find any enhancement of star-formation but an excess of massive quiescent galaxieswithin.So far, we have not considered how galaxies’ dust con-tent may change with the environment. Many stud-ies have suggested that distant protoclusters often hostextremely dusty star-forming galaxies such as submil-limeter galaxies (SMGs) (e.g., Casey 2016; Miller et al.2018; Umehata et al. 2018; Cheng et al. 2019). TheseSMGs usually have extremely high star-formation rates( > (cid:12) /yr) and are generally invisible in rest-frameUV-NIR wavelengths due to heavily dust obscuration. Ifthese dusty star-forming galaxies exist in our protoclus-ter they would be totally missed in our selection. Futuresubmillimeter observations of D4UD01 may give us fur-ther insight into how massive galaxies have formed inthis protocluster. SUMMARY AND CONCLUSIONIn this paper, with the help of multiwavelength data,we study a protocluster D4UD01 at z = 3 .
24 by identi-fying its member galaxies using SED fitting and photo-metric redshift. In the 497 arcmin field which hosts theprotocluster, 450 K S band detected candidate galaxiesat 3 . < z phot < . z sample,reaching a mass completeness of 10 . M (cid:12) . We inves-tigate their distributions in the field and probe possi- ble environmental trends in the protocluster. Our mainconclusions are summarized below.1. Using D4000 index, 52 members are classified asquiescent galaxies in the photo- z sample. Among thesegalaxies, 80% have mass greater than 10 M (cid:12) and 94%have colors consistent with those of DRGs. Thereforethese galaxies are among the most massive and reddestones in the entire sample.2. A large galaxy overdensity is found in the field viaVoronoi tessellation, which contains 96 sources. Beingthe largest overdensity in the entire field, we define thisoverdensity as the protocluster region. Interestingly, wefind that the quiescent galaxies are mostly concentratedin the protocluster region with a higher quiescent frac-tion, suggesting potential environmental quenching ef-fect is taking place in this protocluster.3. The mass function of protocluster galaxies shows anenhancement in comparison to the field, suggesting anaccelerated mass assembly in the protocluster. Whenfurther studying the environmental impacts on galaxyproperties, a weak signal of suppressed star-formationactivities is found in the protocluster comparing with thefield, but the differences are not significant enough to beconclusive. It is argued that D4UD01 is a more evolvedstructure that already passed its peak star-formationera, than those younger protoclusters where enhancedstar-formation activities were found.We thank the anonymous referee for a careful reviewof this paper. K.S. is grateful to Richard Bielby andPatrick Petitjean for sharing the AAOmega spectro-scopic data in the D4 field for photo- z calibration. T.F.and K.S. acknowledge the funding from the NationalKey R&D Program of China No. 2017YFA0402600,and NSFC grants No. 11525312, 11890692. K.S. issupported by NSFC grants No. 12003023 and ChinaPostdoctoral Science Foundation (2020M680086). TheCFHTLS data used in this work are based on obser-vations obtained with MegaPrime/MegaCam, a jointproject of CFHT and CEA/IRFU, at the Canada-France-Hawaii Telescope (CFHT) which is operated bythe National Research Council (NRC) of Canada, theInstitut National des Science de l’Univers of the Cen-tre National de la Recherche Scientifique (CNRS) ofFrance, and the University of Hawaii. This work is basedin part on data products produced at Terapix avail-able at the Canadian Astronomy Data Centre as partof the Canada-France-Hawaii Telescope Legacy Survey,a collaborative project of NRC and CNRS. The IRACdata used in this study are based on data and cata-log products from NMBS-II IRAC, funded by the Na-tional Aeronautics and Space Administration (NASA)under grant number NNX16AN49G issued through theNNH15ZDA001N Astrophysics Data Analysis Program(ADAP). xiii M / M )6.05.55.04.54.03.53.0 l o g [ d e x M p c ] fieldprotoclustertotal Caputi et al. (2011)Davidzon et al. (2017)
Figure 6.
The GSMFs of different samples. The green and orange lines are the GSMFs from Davidzon et al. (2017) andCaputi et al. (2011) at 3 . < z < .
5. The dashed and dash-dotted lines are the corresponding 1 σ uncertainties. The dashedvertical line represents the completeness limit of the sample. REFERENCES
Alberts, S., Pope, A., Brodwin, M., et al. 2014, MNRAS,437, 437, doi: 10.1093/mnras/stt1897Ando, M., Shimasaku, K., & Momose, R. 2020, MNRAS,496, 3169, doi: 10.1093/mnras/staa1757Annunziatella, M., Marchesini, D., Stefanon, M., et al.2018, PASP, 130, 124501, doi: 10.1088/1538-3873/aae796Balogh, M. L., Morris, S. L., Yee, H. K. C., Carlberg, R. G.,& Ellingson, E. 1999, ApJ, 527, 54, doi: 10.1086/308056Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393,doi: 10.1051/aas:1996164Bielby, R., Hudelot, P., McCracken, H. J., et al. 2012,A&A, 545, A23, doi: 10.1051/0004-6361/201118547Boquien, M., Burgarella, D., Roehlly, Y., et al. 2019, A&A,622, A103, doi: 10.1051/0004-6361/201834156Bower, R. G., Lucey, J. R., & Ellis, R. S. 1992, MNRAS,254, 601, doi: 10.1093/mnras/254.4.601 Brammer, G. B., Whitaker, K. E., van Dokkum, P. G.,et al. 2011, ApJ, 739, 24,doi: 10.1088/0004-637X/739/1/24Brinchmann, J., Charlot, S., White, S. D. M., et al. 2004,MNRAS, 351, 1151,doi: 10.1111/j.1365-2966.2004.07881.xBruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000,doi: 10.1046/j.1365-8711.2003.06897.xCai, Z., Fan, X., Yang, Y., et al. 2017, ApJ, 837, 71,doi: 10.3847/1538-4357/aa5d14Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ,533, 682, doi: 10.1086/308692Caputi, K. I., Cirasuolo, M., Dunlop, J. S., et al. 2011,MNRAS, 413, 162, doi: 10.1111/j.1365-2966.2010.18118.xCasey, C. M. 2016, ApJ, 824, 36,doi: 10.3847/0004-637X/824/1/36Chabrier, G. 2003, PASP, 115, 763, doi: 10.1086/376392 iv Shi et al. log(M star )[M ] l o g ( S F R )[ M y r ] fieldprotocluster Figure 7.