The star formation history of galaxies: the role of galaxy mass, morphology and environment
Valentina Guglielmo, Bianca M. Poggianti, Alessia Moretti, Jacopo Fritz, Rosa Calvi, Benedetta Vulcani, Giovanni Fasano, Angela Paccagnella
MMon. Not. R. Astron. Soc. , 1–17 () Printed June 30, 2015 (MN L A TEX style file v2.2)
The star formation history of galaxies: the role of galaxymass, morphology and environment
Valentina Guglielmo, , (cid:63) Bianca M. Poggianti, Alessia Moretti, , Jacopo Fritz, , Rosa Calvi, Benedetta Vulcani , Giovanni Fasano , Angela Paccagnella INAF-Astronomical Observatory of Padova, I-35122 Padova, Italy. Department of Physics and Astronomy, University of Padova, I-35122 Padova, Italy. Sterrenkundig Observatorium Vakgroep Fysica en Sterrenkunde Universiteit Gent, S9 9000 Gent, Belgium. Centro de Radioastronom´ıa y Astrof´ısica, CRyA, UNAM, Campus Morelia, A.P. 3-72, C.P. 58089, Michoac´an, Mexico. Instituto de Astrofisica de Canarias, Departamento de Astrofisica, Universidad de La Laguna, E-38200 La Laguna, Spain Kavli Institute for the Physics and Mathematics of the Universe (WPI), The University of Tokyo Institutes for Advanced Study (UTIAS),the University of Tokyo, Kashiwa, 277-8582, Japan.
Accepted: 2015 April 2. Received: 2015 March 20; in original form 2015 January 8.
ABSTRACT
We analyze the star formation history (SFH) of galaxies as a function of present-day environment, galaxy stellar mass and morphology. The SFH is derived by meansof a non-parametric spectrophotometric model applied to individual galaxies at z ∼ . − . z > ∼ . × higher thanthe high-z observed value. The slope of the SFRD decline with time gets progressivelysteeper going from low mass to high mass haloes. The decrease of the SFRD since z = 2 is due to 1) quenching – 50% of the SFRD in the field and 75% in clusters at z > Key words: galaxies: clusters: general galaxies: evolution galaxies: formation galax-ies: star formation.
In the quest to understand when galaxies formed their starsand assembled their mass, two complementary observationaltechniques can be employed: direct observations of galax-ies at different redshifts, and reconstruction of the previousgalaxy history from fossil records at a given epoch. The mainadvantage of the first method is that measuring the currentstar formation is less uncertain than estimating the past his-tory, especially in galaxies in which the light of young starsoutshines the older population, in particular at high redshift (cid:63)
E-mail: [email protected] (Papovich et al. 2001; Zibetti et al. 2009; Conroy 2013). Onthe other hand, the second method has the benefit of trac-ing the evolution of each individual galaxy, without havingto infer histories in a statistical sense with the problems in-volved in the identification of progenitors and descendants.Both methods heavily rely on spectrophotometric modeling,to calibrate the star formation rate (SFR) indicators and de-rive the star formation histories (SFHs), and are affected bythe choice of the initial mass function (IMF).On a cosmic scale, the collection of the star formationrate density (SFRD) measurements at different cosmic times(from z=8 to 0) give us an indication on the summa of theSFH of the Universe. (Madau et al. 1996; Lilly et al. 1996; c (cid:13) RAS a r X i v : . [ a s t r o - ph . GA ] J un V.Guglielmo et al.
Hopkins & Beacom 2006; Karim et al. 2011 (radio); Bur-garella et al. 2013 (FIR+UV); Sobral et al. 2013 (H-alpha);Bouwens et al. 2014 (UV); Madau & Dickinson 2014).It has emerged that the SFRD of the cosmos peaks at z ∼
2, following a rise after the Big Bang and before fallingby a factor about 10 to the current value. This picture isnow well established, though large uncertainties still existat high redshifts. The SFRD(z) has important implicationsfor the reionization of the Universe, the cosmic chemicalevolution, the transformation of gas into stars and the build-up of stellar mass.Ideally, however, one would want to go beyond the de-scription of cosmic global history, and trace galaxy evolutionon a galaxy-by-galaxy basis to understand the physical pro-cesses driving it. In this respect, great progress has beenmade by surveys at different redshifts that have establishedthe existence of a strong dependence of galaxy histories ongalaxy stellar mass. On average, more massive galaxies haveformed their stars and completed their star formation ac-tivity at higher z than less massive galaxies (the so calleddownsizing effect, Cowie et al. 1996; Gavazzi et al. 2006; DeLucia et al. 2007; S`anchez-Bl`azquez et al. 2009). The exis-tence of relations between star formation rate and galaxystellar mass (SFR-Mass) and specific star formation rateand mass (sSFR=SFR/Mass) have been established fromz=0 out to z >
M/R . is related to the dominance of the bulge componentwith respect to the disk, suggesting it might ultimately belinked with galaxy morphology (see also Driver et al. 2013).Even the sSFR-Mass relation might be due to the increaseof the bulge mass fractions with galaxy stellar mass, as theratio of SFR and stellar mass of the galaxy disk is virtuallyindependent of total stellar mass (Abramson et al. 2014).On the other hand, galaxy stellar population proper-ties have been known to vary strongly with galaxy environ-ment (Spitzer & Baade 1951; Oemler 1974; Davis & Geller1976; Dressler 1980). Galaxy clusters have seen an evolu-tion in their blue galaxy fractions that is even stronger thanin the field, and the evolution from blue star-forming tored passive takes place sooner in dense environments andmassive haloes (Poggianti et al. 2006; Wilman et al. 2005;Cooper et al. 2006; Cucciati et al. 2006; Iovino et al. 2010).Whether this environmental dependence is simply due to dif-ferent galaxy mass distributions and/or morphological dis-tributions with environment, or it reflects a stellar history that differs with environment at a given mass, is still a mat-ter of debate (Thomas et al. 2005,2010; Baldry et al. 2006;Peng et al. 2010, 2012; Poggianti et al. 2013). On a globalscale, the evolution of the SFRD in different environmentsat low redshift is not yet known, though the evolution of theblue galaxy fractions suggests a steeper decline in clustersthan in the field (Kodama & Bower 2001). The contribu-tion of haloes of different masses to the SFRD(z) has beenrecently quantified by Popesso et al. (2014a,b), who arguethat the process of structure formation, and the associatedquenching processes, play an important role in the drop ofthe SFRD(z) since z = 1. Overall, several lines of evidencesuggest that both galaxy mass and environment play a role,with environment being more relevant for lower mass galax-ies, at least as far as quenching is concerned (Haines et al.2007; Cooper et al. 2010; Pasquali et al. 2010; Peng et al.2010, 2012; McGee et al. 2011; Sobral et al. 2011; Muzzinet al. 2012; Smith et al. 2012; Wetzel et al. 2012; Lin et al.2014; La Barbera et al. 2014; Vulcani et al. 2015). However,while it is well established that the relative incidence of star-forming and passive galaxies changes with environment, itis still debated whether environment matters for the wholegalaxy stellar history, or it only causes it to end leading toquenching at some point.Turning to the reconstruction of galaxy SFHs from fos-sil records, this reaches high levels of precision in galaxieswith resolved stellar populations, such as our Milky Wayand the Local Group. Going to more distant galaxies, ithas to rely on the interpretation of the galaxy integratedspectrum, and is limited by our capability to discriminatebetween stars of different ages from the spectrum they emit.Spectrophotometric models capable of extracting SFHs fromintegrated spectra have been built by a number of groups:Heavens et al. (2000; MOPED), Cid-Fernandes et al. (2004;STARLIGHT), Ocvirk et al. (2006a,b; STECMAP), Fritzet al. (2007; now called SINOPSIS), MacArthur et al. 2009,Koleva et al. (2009; ULyss), Tojeiro et al. (2007; VESPA)and others (see sedfitting.org/SED08). They have been ap-plied to reconstruct the SFH of galaxies in large surveys (e.g.Panter et al. 2007 and Tojeiro et al. 2009 on Sloan DigitalSky Survey (SDSS); Fritz et al. 2011 on WINGS), and tostudy these histories for galaxy subsets of special interest(e.g. Tojeiro et al. 2013; Vulcani et al. 2015). Two studies inparticular (Heavens et al. 2004; Panter et al. 2007) derivedthe cosmic SFH from SDSS spectra, and were successful inreproducing the SFRD(z) and the downsizing effect.In this work we make use of a non-parametric spec-trophotometric model to derive the past history of star for-mation in five broad bins of age from integrated spectraof galaxies in clusters and the field and, within the field, ingroups and lower mass haloes. Searching for the origin of theoverall decline observed in the SFRD(z) since z = 2, we alsoconsider present-day star-forming galaxies separately fromthe rest, and quantify the relative role of their decline in starformation and that of galaxies that have been quenched. Ourgoal is to shed light on the history of galaxies of differentmasses and morphologies, and isolate any residual environ-mental trend. We stress that we look for SFH trends withgalaxy parameters today , that is as a function of the mass,morphology and environment that galaxies have at low red-shift, when the spectra we use to derive their past stellarhistory are taken. c (cid:13) RAS, MNRAS , 1–17 he star formation history of galaxies The outline of the paper is as follows: in Sect. 2 we de-scribe the datasets used, and in Sect. 3 the methods forassigning galaxy morphology and the spectrophotometricmodel used for galaxy stellar masses and SFHs. Sect. 4presents our results: in 4.1 the SFRD of the field sampleis compared with recent observational measurements at dif-ferent redshifts; in 4.2 we study the SFRD in different envi-ronments; in 4.3 we analyze the SFH of star-forming galaxiesboth in the field and in clusters, in 4.4 the contribution ofgalaxies of different mass and morphological type to the totalSFRD, and in 4.5 we present a global picture which consid-ers the mean SFH of galaxies in different environments, withthe same stellar mass but different morphology. Finally, wesummarize our findings in Sect. 5.The IMF adopted is a Salpeter one in the mass range0.1-100 M (cid:12) (Salpeter 1955), and the cosmological constantsassumed are Ω m =0.3, Ω Λ =0.7, H = 70 km s − Mpc − . The
Padova Millennium Galaxy and Group Catalogue (Calvi, Poggianti & Vulcani, 2011) is a database built on thebasis of the Millennium Galaxy Catalogue (MGC), a deepand wide B-imaging survey along an equatorial strip of ∼ deg obtained with the Isaac Newton Telescope (INT).The final catalogue is restricted to galaxies brighter thanM B = − . . (cid:54) z (cid:54) .
11, taken from the MGCz catalogue, thespectroscopic extension of the MGC, that has a 96% spec-troscopic completeness at these magnitudes (Driver et al.,2005). Most of the MGCz spectra of our sample come fromthe SDSS (Abazajian et al. 2003, ∼ . M B = − . (cid:63) / M (cid:12) = 10 .
44. Thecomoving volume of the PM2GC survey is 361424 h − Mpc .The image quality and the spectroscopic completenessof the PM2GC are superior to SDSS, and these qualities re-sult in more robust morphological classifications and bettersampling of dense regions. In particular, the MGC is basedon INT data (2.5m telescope) obtained with a median seeingof 1.3” and at least 750s of exposure, with a pixel scale of0.333”/pixel, while the SDSS (again, 2.5m telescope) has amedian seeing of 1.5” in g (the closest band to the PM2GC),an exposure time of 54.1s and 0.396”/pixel. As for spectro-scopic completeneness, 14% of all PM2GC galaxies do nothave an SDSS spectrum, and the SDSS incompleteness isparticularly severe in dense regions such as groups. More-over, the PM2GC data is very comparable in quality to ourcluster sample (WINGS) and the two samples have beenanalyzed in a homogenous way with the same tools.The characterization of the environment of the galaxieswas conducted by means of a Friends-of-Friends (FoF) algo-rithm. The methods and the presentation of the catalogues
Environment Number of galaxiesGroups 1033Single 1123Binary 486Mixed Sample 517General Field 3159
Table 1.
List of the number of galaxies in different environmentsin the PM2GC sample. are described in Calvi et al. (2011). Briefly, a catalogue of176 groups of galaxies with at least three members was builtin the redshift range 0 . (cid:54) z (cid:54) .
1, containing 43% of thetotal general field population at these redshifts. The meanredshift and velocity dispersion σ of the groups are respec-tively 0.0823 and 192 km s − . 88% of the selected groupsare composed by less than 10 members, and 63% by lessthan 5 members. Galaxies were assigned to a group if theywere within 3 σ from the group redshift and 1.5 R fromthe group geometrical center. We define as R the radiusof the sphere inside which the mean density is a factor 200 × the critical density of the Universe at that redshift. Thisparameter gives an approximation of the virial radius of acluster or group and for our structures it is computed fromthe velocity dispersions using the formula (Finn et al., 2005): R = 1 . σ km/s ) 1 (cid:112) Ω Λ + Ω (1 + z ) h − ( Mpc ) (1)with σ the group velocity dispersion and z its mean redshift.Galaxies that do not satisfy the group membership cri-teria have been placed either in the catalogue of single fieldgalaxies, that comprises the isolated galaxies, or in the cat-alogue of binary field galaxies, which comprises the systemswith two galaxies within 1500 km/s and 0.5 h − Mpc. Fi-nally, galaxies that were part of the trial groups in the FoFprocedure but did not fulfill the final group membership cri-teria are treated separately as “Mixed sample”.All galaxies in the environments described above arecollected in the ”general field” sample PM2GC.The number of galaxies in each sub-environment and inthe general field sample are shown in Table 1.In addition to the identification of PM2GC sub-environments, the masses of the dark matter haloes host-ing PM2GC galaxies were estimated by Paccagnella et al.(in preparation) exploiting a mock galaxy catalogue fromsemianalytic models (De Lucia & Blaizot 2007) run on theMillennium Simulation (MS, Springel et al. 2005), and mak-ing use of the already mentioned FoF algorithm (Calvi etal. 2011), as described in Vulcani et al. (2014). The massof a dark matter halo associated with a group (where inthis definition of group also singles and binaries are in-cluded) is tightly correlated with the total stellar mass of allmember galaxies (see e.g. Yang et al. 2007, 2008). Apply-ing this method to the PM2GC magnitude limited sample,Paccagnella et al. (in prep.) derived halo masses for 1141single galaxies, 245 binary systems and 92 groups. In thiscase not all PM2GC groups are considered but only 92 ofthe 176 in the complete catalogue, those in which the frac-tion of interlopers (i.e. the galaxies which are associated to c (cid:13) RAS, MNRAS , 1–17
V.Guglielmo et al. a groups by the FoF algorithm due to projection effects butdo not belong physically to them) is less than 30 %.
The WIde-field Nearby Galaxy-cluster Survey (WINGS)(Fasano et al., 2006) is a multi-wavelength survey of clustersat 0 . < z < .
07 in the local Universe.The complete sample contains 76 clusters selected fromthree X-ray flux limited samples compiled from ROSAT All-Sky Survey data (Ebeling et al., 1996, 1998, 2000), cov-ering a wide range in velocity dispersion, 500 km s − (cid:54) σ cl (cid:54) − and X-ray luminosity, typically 0 . − × erg s − . The survey is mainly based on optical imagingin B and V bands for all the 76 clusters taken with the WideField Camera (WFC) mounted at the corrected f/3.9 primefocus of the INT-2.5m in La Palma and from the Wide FieldImager (WFI) at the 2.2m MPG/ESO telescope in La Silla(Varela et al., 2009). The imaging survey covers a 34 (cid:48) × (cid:48) field, and this area corresponds to at least 0.6R for allclusters. In the following analysis all the cluster membersare used regardless of clustercentric distance since the frac-tion of galaxies that do not satisfy the 0.6 R criterion istiny compared to the entire distribution and does not affectsignificantly the sample.The optical imaging was complemented by a spectro-scopic survey of a subsample of about 6000 galaxies in 48 ofthe 76 clusters (Cava et al. 2009). The spectra were takenfrom August 2002 to October 2004 at the 4.2 m William Her-schel Telescope (WHT) using the AF2/WYFFOS multifiberspectrograph ( ∼ ∼ ± σ cl from the cluster mean redshift.The WINGS spectroscopic sample is affected by incom-pleteness. The completeness parameter, that is the ratio ofthe number of spectra yielding a redshift to the total numberof galaxies in the parent photometric catalogue, was com-puted using the V-band magnitude and turned out to beessentially independent from the distance to the center ofthe cluster (Cava et al., 2009). In the following, SFRs andstellar mass estimates in WINGS galaxies have always beencorrected for incompleteness.From the σ cl , by means of the virial theorem, the massof the dark matter halo in which the cluster resides wascalculated as follows (Poggianti et al. 2006) : R was computed from the cluster velocity dispersion σ cl (inkm s − ) using equation 1 (Cava et al. 2009). This relation yields reliable mass measurements for clusters,but not for groups where the σ is computed from a few redshifts, M halo = 1 . × ( σ km s − ) ) (cid:112) Ω Λ + Ω (1 + z ) h − ( M (cid:12) )(2)The latter equation was applied to all WINGS clustersusing the velocity dispersions given in Cava et al. (2009) for32 of the 48 clusters and for the remaining 16 clusters themost recent data from the OMEGAWINGS spectroscopiccatalogue (Moretti et al. in prep.).To compare different environments, we apply to theWINGS sample the same magnitude cut of the PM2GC.Therefore, in the following, for both WINGS and PM2GC,we use only galaxies brighter than M B = − .
7. In WINGS,this leaves 1249 galaxies ( ∼ (cid:63) / M (cid:12) = 10 .
44 (correspond-ing to M B = − . (cid:63) / M (cid:12) = 10 . h − Mpc . In order to convert this vol-ume into the comoving value it is multiplied for a factor(1 + z ) = 1 .
17, where z is the median redshift of the survey,z = 0.055.
All galaxies in both the PM2GC and WINGS samples havebeen morphologically classified using MORPHOT, an auto-matic non parametric tool designed to obtain morphologicaltype estimates of large galaxy samples (Fasano et al. 2007),which has been shown to be able to distinguish betweenellipticals and S0 galaxies with unprecedented accuracy. Itcombines a set of 11 diagnostics, directly and easily com-putable from the galaxy image and sensitive to some par-ticular morphological characteristic and/or feature of thegalaxies. It provides two independent estimates of the mor-phological type based on: (i) a Maximum Likelihood tech-nique; (ii) a Neural Network machine. The final morphologi-cal estimator combines the two techniques. The comparisonwith visual classifications provides an average difference inHubble type ∆T ( (cid:54) (cid:54) therefore for the groups we adopted the mass estimate methoddescribed in Sect. 2.1. c (cid:13)
RAS, MNRAS , 1–17 he star formation history of galaxies the V and B WINGS images (see Calvi et al. (2012) for moredetails). The morphological types we will consider are ellip-ticals, S0s (lenticulars) and late-types (any type later thanS0s). The SFHs and stellar masses of galaxies in the PM2GC andWINGS samples are derived using a model which is an im-proved and extended version of the spectrophotometric codedeveloped by Poggianti et al. (2001) to derive the SFHs froma galaxy integrated spectrum.The model and its application to WINGS are fully de-scribed in Fritz et al. (2007, 2011, 2014). It is based on astellar population synthesis technique that reproduces theobserved optical galaxy spectra.The code reproduces the main features of an observedspectrum: the equivalent widths of several lines - both inabsorption and in emission - and the fluxes emitted in givenbands of the continuum. This model assumes that an ob-served galactic spectrum is a combination of simple stellarpopulation spectra, and therefore a galaxy model spectrumis computed by adding the synthetic spectra of Single StellarPopulations (SSPs) of different ages.The model makes use of the Padova evolutionary tracks(Bertelli et al. 1994) with AGB treatment as in Bressanet al. (1998), and two different sets of observed stellar li-braries: for ages younger than 10 years Jacoby et al. (1984)was used, while for older SSPs spectra were taken from theMILES library (S`anchez-Bl`azquez et al., 2006). Both setswere degraded in spectral resolution, in order to match thatof the observed spectra. SSP spectra were then extended tothe ultra-violet and infrared using theoretical libraries fromKurucz (private communication), and gas emission was in-cluded by means of the photoionization code CLOUDY (Fer-land, 1996).The initial set of SSPs was composed of 108 theoreticalspectra referring to age intervals from 10 to 20 × years,that were binned into a final set of 12 SSPs used in thefitting.To treat dust extinction, the Galactic extinction curve( R v = 3.1, Cardelli et al. 1989) is adopted, but the value ofthe color excess, E(B-V) is let free to vary as a function ofSSP age: dust extinction will be higher for younger stellarpopulations.A single metallicity value is adopted and the model isrun for three metallicities: Z = 0.05, Z = 0.02, Z = 0.004,choosing as best fit model the one with the smallest χ . Fit-ting an observed spectrum with a single value of the metal-licity is equivalent to assuming that this value belongs tothe stellar population that is dominating its light. A checkon the reliability of the mass and SFHs derived using thismethod has been performed analysing synthetic spectra ofdifferent SFHs with metallicity that varies as a function ofstellar ages, so to simulate the chemical evolution of thegalaxy, and it turns out that the way metallicity is treateddoes not introduce any significant bias in the recovered stel-lar mass or SFH (Fritz et al. 2007).The SFH and mass estimates obtained from the fiberspectrum are scaled from the fiber magnitude to the totalmagnitude to recover galaxy-wide integrated properties as-suming a constant M/L. The differences in color between the fiber and the total magnitudes are however small forour cluster sample, as shown in Fritz et al. (2011), there-fore the assumption of a constant M/L ratio should notintroduce large uncertainties. It is worthwhile citing thatthe application of full spectral fitting techniques to integralfield spectroscopy data yields much more detailed informa-tion about the SFH per pixel (ATLAS3D: Cappellari et al.2012, CALIFA: S´anchez et al. 2012; Cid Fernandes et al.2013; Gonzalez-Delgado et al. 2014, SAMI: Allen et al. 2015,MaNGA: Bundy et al. 2015, CANDELS: Wuyts et al. 2012),however current Integral Field Unit (IFU) surveys are notsuited for a complete census of magnitude limited samplesin different environments. During the fitting, each one of the 12 SSP spectra is mul-tiplied by a value of SFR in that age interval. The fittingalgorithm searches the combination of SFR values that bestmatches the observed spectrum, calculating the differencesbetween the observed and model spectra, and evaluatingthem by means of a standard χ function. The 12 SFR val-ues are let free to vary completely independently from oneanother, without any a priori assumption on the form ofthe SFH. The observed features that are used to comparethe likelihood between the model and the observed spectraare chosen from the most significant emission and absorp-tion lines and continuum flux intervals, after the line equiv-alent widths are automatically measured (see Fritz et al.2007, 2014). The observed errors on the flux are computedby taking into account the local spectral signal-to-noise ra-tio, while uncertainties on the equivalent widths are derivedmainly from the measurements method. An Adaptive Sim-ulated Annealing algorithm randomly explores the param-eters space, searching for the absolute minimum of the χ function.The search of the combination of parameters that mini-mizes the differences between the observed and model spec-trum is a non-linear problem and it is also underdetermined,which means that the number of constraints is lower than thenumber of parameters. The solution given with this methodis non-unique, due to the limited wavelenght range underanalysis, together with the age-metallicity degeneracy andthe already mentioned non-linearity and underdetermina-tion. To account for this, error bars are associated to mass,extinction and age values, computed as follows. The pathperformed by the minimization algorithm towards the bestfit model (the minimum χ ) depends on the starting point,so, in general, starting from different initial positions canlead to different minimum points: 11 optimisations are per-formed, each time starting from a different point in the spaceparameter, obtaining 11 best fit models which are represen-tative of the space of the solutions. Among these, the modelwith the median value for the mass is considered, and er-ror bars are computed as the average difference between thevalues of the model with the highest and lowest total stellarmass formed in that age bin. In this way we are confidentthat the expected values are contained within the error barswe calculate. The values for the stellar masses have beenthoroughly compared both against other methods (Vulcaniet al. 2011) and other datasets (e.g. SDSS) having objects c (cid:13) RAS, MNRAS000
RAS, MNRAS , 1–17 he star formation history of galaxies the V and B WINGS images (see Calvi et al. (2012) for moredetails). The morphological types we will consider are ellip-ticals, S0s (lenticulars) and late-types (any type later thanS0s). The SFHs and stellar masses of galaxies in the PM2GC andWINGS samples are derived using a model which is an im-proved and extended version of the spectrophotometric codedeveloped by Poggianti et al. (2001) to derive the SFHs froma galaxy integrated spectrum.The model and its application to WINGS are fully de-scribed in Fritz et al. (2007, 2011, 2014). It is based on astellar population synthesis technique that reproduces theobserved optical galaxy spectra.The code reproduces the main features of an observedspectrum: the equivalent widths of several lines - both inabsorption and in emission - and the fluxes emitted in givenbands of the continuum. This model assumes that an ob-served galactic spectrum is a combination of simple stellarpopulation spectra, and therefore a galaxy model spectrumis computed by adding the synthetic spectra of Single StellarPopulations (SSPs) of different ages.The model makes use of the Padova evolutionary tracks(Bertelli et al. 1994) with AGB treatment as in Bressanet al. (1998), and two different sets of observed stellar li-braries: for ages younger than 10 years Jacoby et al. (1984)was used, while for older SSPs spectra were taken from theMILES library (S`anchez-Bl`azquez et al., 2006). Both setswere degraded in spectral resolution, in order to match thatof the observed spectra. SSP spectra were then extended tothe ultra-violet and infrared using theoretical libraries fromKurucz (private communication), and gas emission was in-cluded by means of the photoionization code CLOUDY (Fer-land, 1996).The initial set of SSPs was composed of 108 theoreticalspectra referring to age intervals from 10 to 20 × years,that were binned into a final set of 12 SSPs used in thefitting.To treat dust extinction, the Galactic extinction curve( R v = 3.1, Cardelli et al. 1989) is adopted, but the value ofthe color excess, E(B-V) is let free to vary as a function ofSSP age: dust extinction will be higher for younger stellarpopulations.A single metallicity value is adopted and the model isrun for three metallicities: Z = 0.05, Z = 0.02, Z = 0.004,choosing as best fit model the one with the smallest χ . Fit-ting an observed spectrum with a single value of the metal-licity is equivalent to assuming that this value belongs tothe stellar population that is dominating its light. A checkon the reliability of the mass and SFHs derived using thismethod has been performed analysing synthetic spectra ofdifferent SFHs with metallicity that varies as a function ofstellar ages, so to simulate the chemical evolution of thegalaxy, and it turns out that the way metallicity is treateddoes not introduce any significant bias in the recovered stel-lar mass or SFH (Fritz et al. 2007).The SFH and mass estimates obtained from the fiberspectrum are scaled from the fiber magnitude to the totalmagnitude to recover galaxy-wide integrated properties as-suming a constant M/L. The differences in color between the fiber and the total magnitudes are however small forour cluster sample, as shown in Fritz et al. (2011), there-fore the assumption of a constant M/L ratio should notintroduce large uncertainties. It is worthwhile citing thatthe application of full spectral fitting techniques to integralfield spectroscopy data yields much more detailed informa-tion about the SFH per pixel (ATLAS3D: Cappellari et al.2012, CALIFA: S´anchez et al. 2012; Cid Fernandes et al.2013; Gonzalez-Delgado et al. 2014, SAMI: Allen et al. 2015,MaNGA: Bundy et al. 2015, CANDELS: Wuyts et al. 2012),however current Integral Field Unit (IFU) surveys are notsuited for a complete census of magnitude limited samplesin different environments. During the fitting, each one of the 12 SSP spectra is mul-tiplied by a value of SFR in that age interval. The fittingalgorithm searches the combination of SFR values that bestmatches the observed spectrum, calculating the differencesbetween the observed and model spectra, and evaluatingthem by means of a standard χ function. The 12 SFR val-ues are let free to vary completely independently from oneanother, without any a priori assumption on the form ofthe SFH. The observed features that are used to comparethe likelihood between the model and the observed spectraare chosen from the most significant emission and absorp-tion lines and continuum flux intervals, after the line equiv-alent widths are automatically measured (see Fritz et al.2007, 2014). The observed errors on the flux are computedby taking into account the local spectral signal-to-noise ra-tio, while uncertainties on the equivalent widths are derivedmainly from the measurements method. An Adaptive Sim-ulated Annealing algorithm randomly explores the param-eters space, searching for the absolute minimum of the χ function.The search of the combination of parameters that mini-mizes the differences between the observed and model spec-trum is a non-linear problem and it is also underdetermined,which means that the number of constraints is lower than thenumber of parameters. The solution given with this methodis non-unique, due to the limited wavelenght range underanalysis, together with the age-metallicity degeneracy andthe already mentioned non-linearity and underdetermina-tion. To account for this, error bars are associated to mass,extinction and age values, computed as follows. The pathperformed by the minimization algorithm towards the bestfit model (the minimum χ ) depends on the starting point,so, in general, starting from different initial positions canlead to different minimum points: 11 optimisations are per-formed, each time starting from a different point in the spaceparameter, obtaining 11 best fit models which are represen-tative of the space of the solutions. Among these, the modelwith the median value for the mass is considered, and er-ror bars are computed as the average difference between thevalues of the model with the highest and lowest total stellarmass formed in that age bin. In this way we are confidentthat the expected values are contained within the error barswe calculate. The values for the stellar masses have beenthoroughly compared both against other methods (Vulcaniet al. 2011) and other datasets (e.g. SDSS) having objects c (cid:13) RAS, MNRAS000 , 1–17
V.Guglielmo et al. in common with WINGS, showing an excellent agreement(Fritz et al. 2011).The application of the spectrophotometric synthesismodel allows to derive the characteristics of the stellar popu-lations whose light constitutes the integrated spectrum: thetotal stellar mass, the mass of stars formed as a function ofage -i.e. the SFR within each time interval in the galaxy life-,the extinction and the single ”luminosity-weighted” metal-licity value. It is important to keep in mind that the modeloutputs describe the global history of all stars that at lowredshift are in the galaxy: the assembly of such stars in asingle galaxy, i.e. the galaxy merger history, is totally un-constrained with this method.All the galaxy stellar masses used in this paper aremasses locked into stars, including both those that are stillin the nuclear-burning phase, and remnants such as whitedwarfs, neutron stars and stellar black holes.The current SFR values are derived by fitting the fluxof emission lines, whose luminosities are entirely attributedto the star formation process, neglecting all other mecha-nisms that can produce ionising flux. In this way, for LIN-ERS and AGNs the SFR values can in principle be severelyoverestimated. The AGN identification for PM2GC galaxieswas done using the latest AGN catalog from SDSS . Theselection of AGNs in WINGS was performed with a verysimilar method (Marziani et al. 2013, in preparation). Wecalculate that the AGN contribution to the total SFR ofthe PM2GC sample is < ∼ . The reliability of the spectrophotometric technique wastested in two ways (Fritz et al. 2007, 2011). First, templatespectra - which resemble the characteristics and the qualityof the observed ones - spanning a wide range of SFHs werebuilt, to assess the capability of the model to recover theinput SFH. This test was done both on low and high S/Nspectra, in order to verify whether there was a dependenceof the quality of results from the spectral noise. This showedthat the error bars provided by our method for the physicalparameters reasonably account for the uncertainties, thatare dominated by the similarity of old SSP spectra and bythe limited spectral range at disposal for our analysis (Fritzet al. 2007).The second test-phase was done on WINGS spectra incommon with the SDSS project, to verify the reliability ofthe model in absolute terms, and the agreement with theresults on galaxy stellar masses obtained by other workswas very satisfactory (Fritz et al. 2007, 2011).There is an instrinsic degeneracy in the typical featuresof spectra of similar age, and this degeneracy increases forolder stellar population spectra. There is, hence, an intrinsiclimit to the precision of this method in determining the ageof the stellar populations that compose a spectrum. Thechoice of the time interval in which SFRs estimates can be mean z lower z upper δ t t mean t lower t upper Gyr Time from Big Bang (Gyr)0.06 0.04 0.09 0.6 12.7 12.9 12.30.10 0.09 0.12 0.4 11.9 12.3 11.90.40 0.12 0.67 4.6 9.6 11.9 7.31.44 0.67 2.21 4.4 5.4 7.3 2.96.49 2.21 10.71 2.5 1.4 2.9 0.4
Table 2.
Age and redshift intervals adopted. With the cosmo-logical parameters adopted, t
Universe =13.462 Gyr. z mean is themean redshift of the intervals, whose starting and ending valuesare given in z lower and z upper columns, respectively. δ t is thecorresponding time duration of the redshift bin, t mean , t lower and t upper are the age values corresponding to z mean , z lower andz upper , respectively. considered reliable accounts for this aspect, and the initial12 ages of the set of SSP spectra, i.e. the time intervalsover which the SFR is assumed to be constant, were furtherbinned into five intervals. These are the age intervals thatare used throughout this paper. Time and correspondingredshift intervals are listed in Table 2.To visually illustrate the reason for using a few age in-tervals, we plot in the lower panel of Figure 1 the spectraof stellar populations with ages reflecting the 5 age inter-vals adopted. The oldest spectrum, corresponding to a meanelapsed time from the Big Bang of ∼ (cid:62) When comparing model and observational SFR estimates,there are two sources of error: that associated to the SFRestimates from the spectrophotometric model and the typi-cal error for SFR estimates from observations.The first type of errors, computed as described in Sect.3.2.1, are considered symmetric with respect to the centralSFR value in the spectrophotometric fit. The observationalerrors are taken to be equal to the typical observational error(0.225 dex), defined as the mean deviation of star formationestimates obtained using different observables (i.e UV, IR,H α , etc) (Hao et al., 2011). In the following, when plottingSFR estimates for WINGS and PM2GC, these two estimatesare combined in quadrature. For errors on the sSFR we cal-culate the propagation of errors assuming a typical uncer-tainty on the stellar mass of 0.2 dex. The value obtained isthen combined in quadrature with the observational error c (cid:13) RAS, MNRAS , 1–17 he star formation history of galaxies Figure 1.
Bottom : Comparison between spectra of stellar popu-lations of the five age intervals corresponding to each of the fiveredshift intervals in table 2. The age of the populations (and theredshift) is decreasing from the bottom to the top of the panel.The oldest spectrum is plotted in red. Spectra are in arbitraryunits and are normalized at 8000 ˚A.
Top : Ratio between the spec-tra of the two oldest populations. for the SFR, normalized with the same procedure accordingto the mass.These estimates can be considered intrinsic errors anddo not take into account eventual systematic errors aris-ing from systematic uncertainties in the spectrophotomet-ric modelling, for example in the single stellar populationspectra due to isochrones and/or stellar libraries inaccuracy.Therefore, it is important to keep in mind that the errorsshown are lower limits.
In this section we present the methods and most significantresults of the SFH analysis conducted with our spectropho-tometric model on the PM2GC and WINGS.The reconstruction of the SFH of galaxies has been per-formed as follows: the twelve model SFRs are binned intothe five final age intervals as described in section 3.2.2 com-puting the mean constant value of SFR for the entire lengthof the corresponding redshift bin. These values are then di-vided by the comoving volume of the survey the galaxiesbelong to, to obtain the SFRDs.In all the plots five values of star formation are pre-sented, with horizontal bars indicating the redshift intervalthey refer to, and vertical error bars indicating the uncer-tainty computed as already described. The sub-division of galaxies according to their morphol-ogy and environments has been described in sections 2 and3. Below, we will consider the following galaxy stellar massbins: • (cid:54) logM star (M (cid:12) ) < .
44 - this bin is used only forWINGS galaxies, whose mass completeness limit is lowerthan in the PM2GC • M1: 10 . (cid:54) logM star (M (cid:12) ) < . • M2: 10 . (cid:54) logM star (M (cid:12) ) < . • M3: logM star (M (cid:12) ) (cid:62) .
2. The most massive galaxyin the PM2GC has logM star (M (cid:12) ) = 12 .
6, and in WINGSlogM star (M (cid:12) ) = 12 . In Fig. 2 we compare the SFRD of the PM2GC generalfield sample and the cosmic SFH derived from the most re-cent data at all redshifts (Madau & Dickinson 2014, MD14).These latter data are taken from galaxy surveys that provideSFR measurements from rest-frame far-UV (1500 ˚A) andmid- and far-infrared, and span the redshift range z = 0 − ∗ , L min = 0 . L ∗ .A Salpeter . − IMF was assumed in MD14. Togetherwith the data we also plot the best-fitting function givenby MD14, expressed by the analytical form:
SF RD ( z ) = 0 .
015 (1 + z ) . z ) / . . M (cid:12) yr − Mpc − . (3)The PM2GC values are shown as black circles. We notethat the Madau and PM2GC values refer to galaxy samplesselected with different criteria: the L min = 0 . L ∗ limit ateach redshift in MD14, as opposed to M B < − . The most no-ticeable discrepancy is in the highest redshift bin ( z > ∼ .
66 higher than themean SFRD obtained by integrating the MD14 best-fit func-tion at the same epoch. This behaviour can have severalreasons: a) the uncertainty in the two highest redshift bins To assess the effect of the different selection criteria on the totalSFRD estimate at low redshift, we compare the integral of thePM2GC SFR distribution function for the M B = − . L (cid:63) . Wefind that the M B = − . L (cid:63) cut, thus we conclude that the differentcriteria can lead to a ∼
10% difference.c (cid:13)
RAS, MNRAS , 1–17
V.Guglielmo et al.
Figure 2.
Comparison between the PM2GC cosmic SFH andobservational data from the literature (Table 1 in MD14). Theblack circles refer to the PM2GC field dataset. Error bars in or-dinate are smaller than the symbols, while the horizontal errorbars show the redshift intervals each circle is referring to. Thesolid curve is the best-fit SFRD shown in equation 3, as calculatedby MD14. The black empty triangle is the integral of the MD14curve between 10 Gyr and 13 Gyr, corresponding to the last red-shift bin in PM2GC. The data points refer to FUV+UV and mid-and far-IR rest-frame measurements and are taken from Table 1in MD14. Wyder et al. (2005), midnight blue hexagon. Schimi-novich et al. (2005), blue triangles. Robotham & Driver (2011),dark green pentagon. Cucciati et al. (2012), green squares. Dahlenet al. (2007), turquoise pentagons. Reddy & Steidel (2009), for-est green triangles. Bouwens et al. (2012a),(2012b), magenta pen-tagons. Schenker et al. (2013), black crosses. Sanders et al. (2003),brown circle. Takeuchi et al. (2003), dark orange square. Magnelliet al. (2011), red open hexagons. Magnelli et al. (2013), red filledhexagons. Gruppioni et al. (2013), coral hexagons. of the SFRD computed by our spectrophotometric modelalready discussed in section 3.2; b) an underestimation ofthe observed SFRD due to incompleteness of high redshiftdata from current surveys; c) the differences in the MD14vs. PM2GC selection criteria mentioned above.
In figure 3 we compare the PM2GC field SFRD (black cir-cles) with that of the WINGS cluster sample (red emptytriangles). The PM2GC sample has been also divided intosingle galaxies (blue squares), binaries (cyan diamonds) andgroups (green full triangles), according to the criteria de-scribed in Sect. 2.The SFRD is systematically higher in clusters than inthe field, of a factor >
100 at any redshift. This simplyreflects the difference in density (number of galaxies perunit volume) between the two environments, being clustersmuch denser environments than the field. Single galaxiescontribute to the total field SFRD by a factor 1.4 higher
Data sample Halo Mass Number of galaxiesPM2GC M halo < M (cid:12) M (cid:12) < M halo < M (cid:12) M (cid:12) < M halo < M (cid:12) M (cid:12) < M halo < M (cid:12) halo > M (cid:12) Table 3.
List of the number of galaxies with different halo massestimates both in the PM2GC and WINGS samples. than groups in the lowest redshift bin, while at z > . z (cid:62)
2, while more than half of all stars in field galaxiesformed at z <
2. The decreasing factor defined as the ratioof SFRD in the highest and the lowest redshift bin is roughly40 for WINGS, while is ∼ < M (cid:12) are considered, while in WINGS onlygalaxies in more massive systems are taken into account.The number of galaxies in each halo mass interval is listedin Table 3. The SFRs on the y-axis are normalized so to beequal to that of galaxies in the lowest mass haloes in thelowest redshift bin. As a consequence, only the redshift de-pendence of the SFHs of galaxies hosted in different halos iscompared, while absolute values of SFR are not.Globally, the decline in SFH gets progressively steepergoing from lower to higher mass haloes. The exact shapeof such decline seems to vary with the halo mass. In fact,the SFR ranking order at the highest redshift respects thehalo mass ranking, while at z ∼ . − − M (cid:12) groups and 10 − M (cid:12) clusters is not respected.Galaxies in haloes of mass < M (cid:12) show a very flat andwell separated SFH compared to all other masses. Yet, theystill display a SFR at z > c (cid:13) RAS, MNRAS , 1–17 he star formation history of galaxies Figure 3.
Comparison between the field (PM2GC, black circles)and clusters (WINGS, red empty triangles) SFRD. The field sam-ple has also been divided into groups (green triangles), binary(cyan diamonds) and single (blue squares) galaxies. Horizontalbars show the extension of time the circles are referring to. In thetop panel SFRD is given in M (cid:12) yr − Mpc − , in the bottom panelall samples are normalized to the PM2GC low-z value, indicatedby the black solid triangle. with both of our definitions of environment. In Sect. 4.5 wewill analyze in detail the origin of this effect. The SFH throughout the cosmic time in a given environmentincludes a large number of galaxies and at each epoch is theresult of star formation processes taking place in galaxiesthat are still actively forming stars. The decline of the SFRDfrom the past to the present age is in principle the cumu-lative result of declining star formation in galaxies that arestill star-forming today (i.e. at the redshift they are observedin the PM2GC or WINGS) together with the increase in thenumber of galaxies that at some point have stopped form-ing stars, i.e. have been quenched. The study of the SFH oftoday’s star-forming galaxies aims to disentangle these twoeffects.In the following, we consider as currently star-formingthose galaxies whose sSFR at the time they are observed (i.e.z = 0.03-0.11) is above a fixed threshold. For computing thesSFR, the current SFR is taken to be the average during thelast 20 Myr as obtained from the model.In figure 5 we report the sSFR-Mass relation from lowredshift measurements of SFR and galaxy stellar masses.The black dots in the figure are the PM2GC field galaxysample values (z = 0.03-0.11), the green dotted line is thefit from the star forming sequence from Salim et al. (2007)(z (cid:39) (cid:39)
Figure 4.
The SFH of galaxies divided according to the massof their host halo. Galaxies are from the PM2GC sample untilhalo masses of log ( M Halo ) /M (cid:12) =14 and from the WINGS clustersample for more massive halos. The SFRs are normalized so tocoincide in the lowest redshift bin, as indicated by the large blackfilled circle. The halo mass ranges considered are shown in thelegend. blue solid one. The threshold separating star-forming frompassive galaxies is chosen on the basis of this sSFR-massrelation and is taken to be equal to sSFR = 10 − yr − (seeFig. 5). This criterion selects 2094 star-forming galaxies inthe field and 612 in clusters.Figure 6 shows the mean SFH per star-forming galaxyin different galaxy mass bins, obtained dividing the sum ofall SFRs by the number of galaxies. The global decline inthe cosmic star formation is not only due to an increasingfraction of galaxies becoming quenched at lower redshifts,but also to the decrease with time of the average SFR oftoday’s star-forming galaxies.The trend depends on galaxy mass, as shown in Fig. 6: itis steeper in high-mass galaxies than in low-mass ones, bothin the field and in clusters. In clusters, the SFR drop betweenthe oldest and the second oldest time intervals is much morepronounced than in the field for all galaxy masses, in agree-ment with the fact that star formation in cluster galaxiesoccurred very early on.Figure 7 shows the redshift dependence of the ratio be-tween the total SFR of all galaxies at any given redshiftand the total SFR at the same redshift of galaxies thatare still forming stars today, for PM2GC (full circles) andWINGS (empty triangles) separately. The fractional contri-bution to the total SFR at any redshift of galaxies that arenow quenched is equal to (1 - 1/y), with y being the Y axis-value in Fig. 7. There is one extra redshift bin in this figure,because the first time interval of 600 Myr was splitted into20 Myr and 580 Myr, to isolate the present-day value accord-ing to our definition of star forming galaxies. The first pointplotted in the figure represents the ratio between the SFR c (cid:13) RAS, MNRAS , 1–17 V.Guglielmo et al.
Figure 5.
The sSFR-Mass relation. Black dots refer to thePM2GC galaxy sample. The main sequence of star forming galax-ies from Salim et al. (2007) is plotted with the green dotted lineand from Lara-Lopez et al (2013) with the blue solid line. Thetwo blue dashed lines are located at one sigma with respect tothe blue solid one. of today’s star forming galaxies and the current measuredSFR, and by definition it has a value equal to one.The resulting values have been interpolated using aleast squares method and the resulting interpolation linesare:PM2GC : y = (0 . ± . × log(z) + (1 . ± .
04) (4)rms = 0 . . ± . × log(z) + (3 . ± .
36) (5)rms = 0 . ∼
50% of the SFR in the field and ∼
75% in clusters at z > z ∼ .
5, these factors are 42%in the field and 73% in clusters.Moreover, in clusters the interpolation line is almostthree times steeper than in the field, meaning that the con-
Figure 7.
Ratio between the SFR from the complete sample andthe SFR of currently star-forming galaxies for PM2GC (full cir-cles) and WINGS (empty triangles). The solid lines are the linearinterpolation computed using an ordinary least square methodand whose equations are given in eqn. 6. Error bars have beencomputed from the errors relative to SFRs in both the completeand star-forming samples using error propagation. tribution of quenching to the SFRD(z) decline is much moresignificant in clusters than in the field.
In this section we focus on the comparison between clustersand field taking into consideration the contribution to theSFRD(z) of galaxies of different mass and morphology. Re-call that stellar masses and morphologies refer to galaxies asthey appear at low redshift, when we observe them. Theirmorphological type at higher z, at the moment they pos-sessed the SFRs we infer, might have been different, due tothe well known morphological evolution taking place bothin clusters and in the field (e.g. Dressler et al. 1997; Oeschet al. 2010; Vulcani et al. 2011).Figure 8 shows that the contribution to the SFRD(z)depends on the morphological type and, considering a giventype, on the environment. We note that the results for early-type galaxies (ellipticals andS0s) should be considered as upper limits, since the presence ofAGN could produce an overestimation of the SFR, therefore ofthe SFRD. Nonetheless, in our sample the total contribution fromAGN is negligible, as discussed in sec. 3.2.1. AGNs classified asearly-type in the PM2GC sample are 28 and their contributionto the z=0 early-type SFRD is ∼ . ∼ . (cid:13) RAS, MNRAS , 1–17 he star formation history of galaxies Figure 6.
The PM2GC (left) and WINGS (right) mean SFR of today’s star-forming galaxies. Galaxies are considered as star-forming ifthey have a sSFR higher than 10 − in the last 20 Myrs. The selected galaxies are divided into three mass bins, plotted with differentcolours and shapes as shown in the legend. Horizontal bars refer to the time interval over which the mean SFR is computed, while verticalones are associated with errors in the SFR determination. The main contribution to the SFRD in the field sam-ple (left panel in the figure) is given by today’s late-typegalaxies (marked with blue circles), which dominate at allredshifts. Compared to the total values estimated for thePM2GC, the SFRD of late-types is ∼
70% of the total atthe present epoch and ∼
40% at the highest z. The relativecontribution of different morphological types to the totalstar formation varies with time: (today’s) early-type galax-ies, which are composed mainly of old and red stars, gavea larger contribution to the SFRD at earlier epochs, whiletoday they contribute only for 30 % of the total SFRD. S0sand ellipticals have quite similar values at every epoch, withellipticals slightly dominating at all redshifts except the low-est bin.In principle, the analysis just performed depends onboth the stellar history of each type and the morphologi-cal distribution of galaxies within each environment, i.e. thenumber of galaxies populating each type. In our field sample59% of all galaxies are late-types, 21% are S0s and 19% areellipticals. At z = 0 on average the star formation activityin a late-type galaxy is 1.5 times higher than in an ellipticaland 1.6 times than in an S0, which is expected given thatearly-type galaxies today are on average more passive thanlate-types.In contrast with the field, early-type galaxies dominatethe total SFRD in clusters at all epochs, except in the lowestredshift bin. The difference in the fractional contribution oflate- and early-types, however, is much smaller than in thefield: in clusters, 40 % of the total today SFRD is due tolate-types, only slightly higher than the 32 % and 28 % ofS0s and ellipticals, respectively. This picture reverses goingback in time: between 0.1 (cid:46) z (cid:46)
1, today’s lenticular galaxiesproduce the majority of stars, and finally at the highest z el- lipticals dominate. The scenario just described is influencedby the significantly different distribution of morphologies incluster galaxies compared with that in the field: in clusters28% of all galaxies are ellipticals, 44% are S0s and 27% arelate-type.Overall, the trends in clusters and in the field are clearlyvery different as far as the relative roles of each type areconcerned.We now divide galaxies into mass bins, according tothe completeness limits of the two surveys: M (cid:62) . M (cid:12) for PM2GC and M (cid:62) M (cid:12) for WINGS. In figure 9 thefield and the cluster SFRDs are divided into respectivelythree and four mass bins. Qualitatively, the global SFRDin both environments is dominated by galaxies with M > . M (cid:12) . Going into more details, in the field, galaxies withmasses M (cid:62) . M (cid:12) give the main contribution to thetotal SFRD for z (cid:38) . M (cid:12) (cid:54) M < . M (cid:12) prevail. Low-mass galaxies (blue circles) have lower SFRDthan the intermediate mass galaxies, but still higher thanthe most massive galaxies for z (cid:46) (cid:62) . M (cid:12) galaxies c (cid:13) RAS, MNRAS , 1–17 V.Guglielmo et al. on average is (cid:118) (cid:118)
2, and at higher z the mostmassive ones prevail. Analysing again the mean SFRD pergalaxy within a certain range in mass it turns out thattoday the SFRD of the average low-mass galaxy becomesroughly equal to that of intermediate-mass ones, while high-mass galaxies have values higher of a factor (cid:118) (cid:118) (cid:118) (cid:118) (cid:118)
26 for high mass galaxies in the field. The same ratiosin WINGS galaxies are the following: (cid:118)
13 for low masses, (cid:118)
36 for intermediate ones and (cid:118)
94 for the highest masses.The numbers listed above demonstrate that the downsizingphenomenon acts in the field as well as in clusters, but inthe latter it is stronger. Even galaxies of the same mass arecharacterised by different timescales and SFHs dependingon their environment. The average decline of the star for-mation process in galaxies of a given mass is less steep inthe field than in clusters. Cluster galaxies form the bulk oftheir stars at earlier epochs with average high-z SFRD val-ues per galaxy systematically higher than those of the fieldat any given galaxy stellar mass today, as we will see in moredetails in the next section.
The last sequence of plots in figure 10 aims to answer thefollowing questions: on average, do galaxies of different mor-phological types but same masses have different histories?Do galaxies of the same mass and morphological type havedifferent histories depending on the environment?We divide galaxies according to their morphologicaltype, mass bin and environment. To avoid any possibleresidual mass dependence in each mass bin, we first veri-fied whether galaxies in each given mass bin and given mor-phological type had the same mass distribution in both en-vironments, performing a Kolmogorov Smirnov test (KS).The KS test found significantly different mass distributionsonly in three cases (elliptical galaxies in the lowest mass binand lenticular galaxies both in the lowest and in the high-est mass bins, plots not shown). For these, we constructedad hoc samples of randomly selected WINGS galaxies that matched the PM2GC mass distribution. Moreover, for thisplot, we limit the M3 bin to < × M (cid:12) , to have a sim-ilar upper mass limit for spirals, ellipticals and S0s in eachenvironment.Fig. 10 presents the SFH of galaxies in each mass bin,matched in mass when necessary, for different environmentsand morphologies. The total SFR is divided by the num-ber of galaxies of each subsample, in order to derive themean history of a galaxy of a given type, mass and envi-ronment. Environments are plotted in figures with differentsymbols (full circles, squares and triangles for field galaxiesand empty circles, squares and triangles for cluster galax-ies) and colours follow the same legend of the morphologicalanalysis in figure 8.Figure 10 highlights that, perhaps surprisingly, in agiven environment, galaxies of the same mass but differ-ent morphologies share the same history of star formation,except for the lowest redshift bin. In fact, the average SFRof late-type, S0 and elliptical galaxies of a given mass issimilar within the errors at all redshifts, except at z < . z > z > and environment, andis almost independent of its present-day morphology.Finally, computing from Fig. 10 the ratio of the aver-age SFR in the highest and lowest redshift bins for galax-ies of the same mass and morphology and comparing it fordifferent environments, we obtain cluster-to-field ratios typ-ically ranging from 4 to 7, with an average of 5. Thus, themuch steeper SFRD decline in clusters compared to the field(ratio=40/7=5.7) discussed in 4.2 can be explained by thesteeper history of cluster galaxies compared to the field, atfixed galaxy mass and morphology. We conclude that thedifferent slope in the SFRD(z) of clusters and field is notdriven by variations of the galaxy mass or morphologicaldistributions with environment, but by the fact that galaxystellar histories vary with galaxy location at each given massand morphology. c (cid:13) RAS, MNRAS , 1–17 he star formation history of galaxies Figure 8.
The SFH of PM2GC (in the left panel) and WINGS (in the right panel) whose galaxies have been divided according to theirmorphological type. Red triangles stand for ellipticals, cyan squares for lenticulars (S0) and blue circles for late-types. All SFRDs referto the same time intervals, here represented with horizontal bars. Vertical bars are associated to indetermination in SFRD values.
Figure 9.
The PM2GC (on the left) and WINGS (on the right) SFH for galaxies divided in mass bins, as shown in the legend. Massvalues here reported are calculated according to a Salpeter (0.1-100) IMF. All SFRDs are supposed to be constant in the same timeintervals, here represented with horizontal bars. Vertical bars are associated to indetermination in SFRD values.
Having derived the SFH of galaxies in clusters (WINGS) andthe field (PM2GC), we have investigated the SFRD evolu-tion with redshift as a function of environment, the historiesof galaxies that are still forming stars at the time they areobserved, and the role of galaxy masses, morphologies and environment in driving the differences of the SFRD(z) withenvironment. We have found that:(i) The PM2GC cumulative SFRD agrees quite well withthe SFRD observed at different redshifts (figure 2). Theonly discrepancy is seen at the highest z (z >
2) where the c (cid:13) RAS, MNRAS , 1–17 V.Guglielmo et al.
Figure 10.
The SFH of galaxies with different morphological type and mass: in the first plot galaxies have 10 . M (cid:12) (cid:54) M < . M (cid:12) ,in the second galaxies have 10 . M (cid:12) (cid:54) M < . M (cid:12) and in the third galaxies have masses M (cid:62) . M (cid:12) . Data reported with fullcircles, squares and triangles refer to the field sample PM2GC and empty ones refer to the cluster sample WINGS, with different coloursmeaning different morphological types, as shown in the legend. The average SFRs are assumed to be constant in the same temporalextension, here represented with horizontal bars. Vertical bars are associated to indetermination in SFR values. Symbols are horizontallyshifted by small arbitrary amounts within their redshift bin in order to avoid superpositions. PM2GC SFRD is a factor ∼ . c (cid:13) RAS, MNRAS , 1–17 he star formation history of galaxies this is true in particular for clusters. More than 50 % of theSFR in the field and more than 75 % in clusters at z > ∼ ACKNOWLEDGMENTS
We acknowledge the anonymous referee for her/his carefulreport, important suggestions and comments which helpedus improving our work. We thank Joe Liske, Simon Driverand the whole MGC team for making easily accessible agreat data set. We are grateful to the rest of the WINGSteam for help and useful discussions. VG and BMP ac-knowledge financial support from the Istituto Nazionale diAstrofisica through a PhD Cycle 30th grant. BV was sup-ported by the World Premier International Research CenterInitiative (WPI), MEXT, Japan and by the Kakenhi Grant-in-Aid for Young Scientists (B)(26870140) from the JapanSociety for the Promotion of Science (JSPS).We gratefully acknowledge Prof. Giuseppe Tormen forinspiring discussions, precious suggestions and support asinternal supervisor during all the work.
References
Abazajian K., et al., 2003, AJ, 126, 2081Abramson L. E., Gladders M. D., Dressler A., Oemler A.,Poggianti B., Vulcani B., 2014, arXiv, arXiv:1411.1431Allen J. T., et al., 2015, MNRAS, 446, 1567Baldry I. K., Balogh M. L., Bower R. G., Glazebrook K.,Nichol R. C., Bamford S. P., Budavari T., 2006, MNRAS,373, 469Bernardi M., et al., 2003, AJ, 125, 1849Bertelli G., Bressan A., Chiosi C., Fagotto F., Nasi E.,1994, A&AS, 106, 275Bothwell M. S., et al., 2011, MNRAS, 415, 1815Bouwens R. J., et al., 2012, ApJ, 752, L5Bouwens R. J., et al., 2012, ApJ, 754, 83Bouwens R. J., et al., 2014, ApJ, 795, 126Bressan A., Granato G. L., Silva L., 1998, A&A, 332, 135Brinchmann J., Charlot S., White S. D. M., Tremonti C.,Kauffmann G., Heckman T., Brinkmann J., 2004, MN-RAS, 351, 1151Bundy K., et al., 2015, ApJ, 798, 7Burgarella D., et al., 2013, A&A, 554, AA70 Calvi R., Poggianti B. M., Vulcani B., 2011, MNRAS, 416,727Calvi R., Poggianti B. M., Fasano G., Vulcani B., 2012,MNRAS, 419, L14Calvi R., Poggianti B. M., Vulcani B., Fasano G., 2013,MNRAS, 432, 3141Cappellari M., et al., 2012, Nature, 484, 485Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345,245Cava A., et al., 2009, A&A, 495, 707Cid Fernandes R., Gu Q., Melnick J., Terlevich E., Ter-levich R., Kunth D., Rodrigues Lacerda R., Joguet B.,2004, MNRAS, 355, 273Cid Fernandes R., et al., 2013, A&A, 557, A86Colless M., et al., 2001, MNRAS, 328, 1039Conroy C., 2013, ARA&A, 51, 393Cooper M. C., et al., 2006, MNRAS, 370, 198Cooper M. C., et al., 2010, MNRAS, 409, 337Cowie L. L., Songaila A., Hu E. M., Cohen J. G., 1996, AJ,112, 839Cucciati O., et al., 2006, A&A, 458, 39Cucciati O., et al., 2012, A&A, 539, AA31Daddi E., et al., 2007, ApJ, 670, 156Dahlen T., Mobasher B., Dickinson M., Ferguson H. C.,Giavalisco M., Kretchmer C., Ravindranath S., 2007, ApJ,654, 172Davis M., Geller M. J., 1976, ApJ, 208, 13De Lucia G., et al., 2007, MNRAS, 374, 809De Lucia G., Blaizot, J., 2007, MNRAS, 375, 214Dressler A., et al., 1997, ApJ, 490, 577Dressler A., 1980, ApJ, 236, 351Driver S. P., Liske J., Cross N. J. G., De Propris R., AllenP. D., 2005, MNRAS, 360, 81Driver S. P., Robotham A. S. G., Bland-Hawthorn J.,Brown M., Hopkins A., Liske J., Phillipps S., Wilkins S.,2013, MNRAS, 430, 2622Ebeling H., Voges W., Bohringer H., Edge A. C., HuchraJ. P., Briel U. G., 1996, MNRAS, 281, 799Ebeling H., Edge A. C., Bohringer H., Allen S. W., Craw-ford C. S., Fabian A. C., Voges W., Huchra J. P., 1998,MNRAS, 301, 881Ebeling H., Edge A. C., Allen S. W., Crawford C. S., FabianA. C., Huchra J. P., 2000, MNRAS, 318, 333Fasano G., et al., 2006, A&A, 445, 805Fasano G., Vanzella E., Wings Team, 2007, ASPC, 374,495Fasano G., et al., 2012, MNRAS, 420, 926Ferland G. J., 1996, Hazy, a Brief Introduction to Cloudy,Univ. Kentucky Phys. Dep. Int. RepFinn R. A., et al., 2005, ApJ, 630, 206Franx M., van Dokkum P. G., Schreiber N. M. F., WuytsS., Labb´e I., Toft S., 2008, ApJ, 688, 770Fritz J., et al., 2007, A&A, 470, 137Fritz J., et al., 2011, A&A, 526, AA45Fritz J., et al., 2014, A&A, 566, AA32Gavazzi G., Boselli A., Cortese L., Arosio I., Gallazzi A.,Pedotti P., Carrasco L., 2006, A&A, 446, 839Gonz´alez Delgado R. M., et al., 2014, A&A, 562, A47Gruppioni C., et al., 2013, MNRAS, 432, 23Haines C. P., Gargiulo A., La Barbera F., Mercurio A.,Merluzzi P., Busarello G., 2007, MNRAS, 381, 7 c (cid:13) RAS, MNRAS , 1–17 V.Guglielmo et al.
Hao C.-N., Kennicutt R. C., Johnson B. D., Calzetti D.,Dale D. A., Moustakas J., 2011, ApJ, 741, 124Heavens A. F., Jimenez R., Lahav O., 2000, MNRAS, 317,965Heavens A., Panter B., Jimenez R., Dunlop J., 2004, Na-ture, 428, 625Hopkins A. M., Beacom J. F., 2006, ApJ, 651, 142Iovino A., et al., 2010, A&A, 509, AA40Jacoby G. H., Hunter D. A., Christian C. A., 1984, ApJS,56, 257Karim A., et al., 2011, ApJ, 730, 61Kauffmann G., Heckman T. M., De Lucia G., BrinchmannJ., Charlot S., Tremonti C., White S. D. M., BrinkmannJ., 2006, MNRAS, 367, 1394Kodama T., Bower R. G., 2001, MNRAS, 321, 18Koleva M., Prugniel P., Bouchard A., Wu Y., 2009, A&A,501, 1269La Barbera F., Pasquali A., Ferreras I., Gallazzi A., deCarvalho R. R., de la Rosa I. G., 2014, MNRAS, 445,1977Lara-L´opez M. A., et al., 2013, MNRAS, 434, 451Lilly S. J., Le Fevre O., Hammer F., Crampton D., 1996,ApJ, 460, L1Lin L., et al., 2014, ApJ, 782, 33MacArthur L. A., Gonz´alez J. J., Courteau S., 2009, MN-RAS, 395, 28Madau P., Ferguson H. C., Dickinson M. E., Giavalisco M.,Steidel C. C., Fruchter A., 1996, MNRAS, 283, 1388Madau P., Dickinson M., 2014, ARA&A, 52, 415Magnelli B., Elbaz D., Chary R. R., Dickinson M., LeBorgne D., Frayer D. T., Willmer C. N. A., 2011, A&A,528, AA35Magnelli B., et al., 2013, A&A, 553, AA132Marziani P., D’Onofrio M., Bettoni D., Fasano G., Fritz J.,Poggianti B. M., Cava A., 2013, AN, 334, 412McGee S. L., Balogh M. L., Wilman D. J., Bower R. G.,Mulchaey J. S., Parker L. C., Oemler A., 2011, MNRAS,413, 996Muzzin A., et al., 2012, ApJ, 746, 188Noeske K. G., et al., 2007, ApJ, 660, L43Ocvirk P., Pichon C., Lan¸con A., Thi´ebaut E., 2006, MN-RAS, 365, 46Ocvirk P., Pichon C., Lan¸con A., Thi´ebaut E., 2006, MN-RAS, 365, 74Oemler A., Jr., 1974, ApJ, 194, 1Oesch P. A., et al., 2010, ApJ, 714, L47Omand C. M. B., Balogh M. L., Poggianti B. M., 2014,MNRAS, 440, 843Panter B., Jimenez R., Heavens A. F., Charlot S., 2007,MNRAS, 378, 1550Papovich C., Dickinson M., Ferguson H. C., 2001, ApJ,559, 620Pasquali A., Gallazzi A., Fontanot F., van den Bosch F. C.,De Lucia G., Mo H. J., Yang X., 2010, MNRAS, 407, 937Peng Y.-j., et al., 2010, ApJ, 721, 193Peng Y.-j., Lilly S. J., Renzini A., Carollo M., 2012, ApJ,757, 4Poggianti B. M., Bressan A., Franceschini A., 2001, ApJ,550, 195Poggianti B. M., et al., 2006, ApJ, 642, 188Poggianti B. M., et al., 2013, ApJ, 762, 77Popesso P., et al., 2014, arXiv, arXiv:1407.8214 Popesso P., et al., 2014, arXiv, arXiv:1407.8231Reddy N. A., Steidel C. C., 2009, ApJ, 692, 778Robotham A. S. G., Driver S. P., 2011, MNRAS, 413, 2570Rodighiero G., et al., 2011, ApJ, 739, L40Salim S., et al., 2007, ApJS, 173, 267Salpeter E. E., 1955, ApJ, 121, 161S´anchez S. F., et al., 2012, A&A, 538, A8S´anchez-Bl´azquez P., et al., 2006, MNRAS, 371, 703S´anchez-Bl´azquez P., et al., 2009, A&A, 499, 47Sanders D. B., Mazzarella J. M., Kim D.-C., Surace J. A.,Soifer B. T., 2003, AJ, 126, 1607Schenker M. A., et al., 2013, ApJ, 768, 196Schiminovich D., et al., 2005, ApJ, 619, L47Smith R. J., Lucey J. R., Hudson M. J., 2009, MNRAS,400, 1690Smith R. J., Lucey J. R., Price J., Hudson M. J., PhillippsS., 2012, MNRAS, 419, 3167Sobral D., Best P. N., Smail I., Geach J. E., Cirasuolo M.,Garn T., Dalton G. B., 2011, MNRAS, 411, 675Sobral D., Smail I., Best P. N., Geach J. E., Matsuda Y.,Stott J. P., Cirasuolo M., Kurk J., 2013, MNRAS, 428,1128Sobral D., Best P. N., Smail I., Mobasher B., Stott J., Nis-bet D., 2014, MNRAS, 437, 3516Speagle J. S., Steinhardt C. L., Capak P. L., SilvermanJ. D., 2014, ApJS, 214, 15Spitzer L., Jr., Baade W., 1951, ApJ, 113, 413Springel V., et al., 2005, Natur, 435, 629Takeuchi T. T., Yoshikawa K., Ishii T. T., 2003, ApJ, 587,L89Thomas D., Maraston C., Bender R., Mendes de OliveiraC., 2005, ApJ, 621, 673Thomas D., Maraston C., Schawinski K., Sarzi M., Silk J.,2010, MNRAS, 404, 1775Tojeiro R., Heavens A. F., Jimenez R., Panter B., 2007,MNRAS, 381, 1252Tojeiro R., Wilkins S., Heavens A. F., Panter B., JimenezR., 2009, ApJS, 185, 1Tojeiro R., et al., 2013, MNRAS, 432, 359van der Wel A., Bell E. F., van den Bosch F. C., GallazziA., Rix H.-W., 2009, ApJ, 698, 1232Varela J., et al., 2009, A&A, 497, 667Vulcani B., et al., 2011, MNRAS, 413, 921Vulcani B., De Lucia G., Poggianti B. M., Bundy K., MoreS., Calvi R., 2014, ApJ, 788, 57Vulcani B., Poggianti B. M., Fritz J., Fasano G., MorettiA., Calvi R., Paccagnella A., 2015, ApJ, 798, 52Wake D. A., van Dokkum P. G., Franx M., 2012, ApJ, 751,L44Wetzel A. R., Tinker J. L., Conroy C., 2012, MNRAS, 424,232Whitaker K. E., van Dokkum P. G., Brammer G., FranxM., 2012, ApJ, 754, L29Wilman D. J., et al., 2005, MNRAS, 358, 88Wyder T. K., et al., 2005, ApJ, 619, L15Wuyts S., et al., 2012, ApJ, 753, 114Yang X., Mo H. J., van den Bosch F. C., Pasquali A., LiC., Barden M., 2007, ApJ, 671, 153Yang X., Mo H. J., van den Bosch F. C., 2008, ApJ, 676,248Zibetti S., Charlot S., Rix H.-W., 2009, MNRAS, 400, 1181 c (cid:13) RAS, MNRAS , 1–17 he star formation history of galaxies This paper has been typeset from a TEX/ L A TEX file preparedby the author. c (cid:13) RAS, MNRAS000