Table Of Content

Mon.Not.R.Astron.Soc.000,1–12(2004) Printed2February2008 (MNLATEXstylefilev2.2) Star formation and the environment of nearby field galaxies Ab´ılio Mateus Jr.1 ⋆ and Laerte Sodré Jr.1 ⋆ 1Departamento de Astronomia, IAG-USP, Rua do Matão 1226, 05508-090, São Paulo, Brazil 4 2February2008 0 0 2 ABSTRACT n We investigate the environmental dependence of galaxies with star formation from a a volume-limited sample of 4782 nearby field galaxy spectra extracted from the 2dF J Galaxy Redshift Survey final data release. The environment is characterized by the 6 localspatialdensity ofgalaxies,estimatedfromthe distance to the 5thnearestneigh- bour. Extensive simulations have been made to estimate correction factors for the 3 local density due to sample incompleteness. We discriminate the galaxies in distinct v spectral classes – passive, star-forming, and short starburst galaxies – by the use of 9 the equivalent widths of [O ii]λ3727 and Hδ. The frequency of galaxies of different 4 classes are then evaluated as a function of the environment. We show that the frac- 3 tionofstar-forminggalaxiesdecreaseswithincreasingdensity,whereaspassivegalaxies 7 presenttheoppositebehaviour.Thefractionofshortstarburstgalaxies–thatsuffered 0 astarburstat∼200Myrago–donotpresentstrongenvironmentaldependence.The 3 0 fractionofthisclassofgalaxiesisalsoapproximatelyconstantwithgalaxyluminosity, / except for the faintest bins in the sample, where their fraction seems to increase. We h find that the star-formation properties are affected in all range of densities present p in our sample (that excludes clusters), what supports the idea that star-formation - o in galaxies is affected by the environment everywhere. We suggest that mechanisms r like tidal interactions, which act in all environments, do play a relevant role on the t s star-formationin galaxies. a : Keywords: star:formation–galaxies:stellarcontent–galaxies:starburst–galaxies: v evolution i X r a 1 INTRODUCTION galaxies (Giovanelli & Haynes 1985; Solanes et al. 1996; Bravo-Alfaro et al. 2000; Goto et al. 2003), that could ex- It has become clear along thelast decades that theen- plain the interruption of star formation in terms of the de- vironment a galaxy inhabits has a profound impact on its pletion of the galaxy gaseous content, either by its con- star-formationproperties.Indeed,thestudyofthevariations sumption or suppression (Gisler 1978; Kennicutt 1983; of star formation in galaxies as a function of their environ- Kenney& Young1989).Ontheotherside,someworksshow ment has a great importance for extragalactic astronomy, that spirals in clusters may have a star formation rate sim- sinceitmakespossibletounderstandsomepropertiesofthe ilar or even larger than field spirals (Gavazzi & Jaffe 1985; formation andevolutionofgalaxiesintheUniverse,helping Moss & Whittle 1993; Gavazzi et al. 1998; Biviano et al. to discriminate between ‘nature’ and ‘nurture’effects. 1997), although, as suggested by Gavazzi & Jaffe, the ob- It is well known that at low redshifts star-forming servedincrementinthestar-formationrateofthesegalaxies galaxies constitute a small population in high-density re- may be a transient phenomenon, which occurs while the gions, especially in galaxy clusters. Osterbrock (1960) galaxysuffersthesuppressionofitsgaseouscontent.Inlow- observed that the frequency of emission-line galaxies density environments, tidal interactions are the most rel- is lower in clusters than in the field. Later, this evant process. Ordinary star-forming galaxies have an ex- trend was verified in a quantitative way by many ternal gaseous reservoir that is critically important to the works (Gisler 1978; Dressler, Thompson & Shectman 1985; continuousgassupplyofthegalaxydisk.Numericalsimula- Abraham et al. 1996; Balogh et al. 1999; Poggianti et al. tions(Bekki, Couch & Shioya2001)showthatthisreservoir 1999; Loveday,Tresse & Maddox 1999; Ellingson et al. isfragileand,consequently,moresusceptibletotidaleffects. 2001, and others). Furthermore, it is also known the trend for H i deficiency among cluster spirals Effects of the environment on the star-formation in galaxies, then,must involve at least two typesof processes: i) those that decrease the gaseous content and, therefore, ⋆ E-mail:[email protected];[email protected] reduce the potential of star formation in galaxies, and ii) 2 A. Mateus Jr. & L. Sodré Jr. processes that trigger bursts of star formation. Among the 2.1 Limited volume sample first class of process there are: interactions between the in- The data used in this work were extracted from the tragalacticandintergalacticmedium,includinggasremoval 2dFGRS spectra recently available on the final data re- and evaporation (Gunn & Gott 1972; Fujita & Nagashima lease (Colless et al. 2003, see also Colless et al. 2001 for 1999); tidal interactions, that remove the gas of the further and detailed information about the survey)1. The disk of spiral galaxies (Byrd & Valtonen 1990); suppres- 2dFGRSobtainedspectrafor245591objects,mainlygalax- sion of the accretion of gas-rich materials in the neigh- ies, brighter than an extinction-corrected magnitude limit bourhood of the galaxy (Larson, Tinsley & Caldwell 1980; of b = 19.45. We have constructed a limited volume sam- Bekki, Couch & Shioya2001).Thesecondtypeofprocesses J ple with galaxies within the strips located in the Northern include: gas compression by ram-pressure, that induces (NGP; 2.5◦ >δ>−7.5◦, 9h50m <α<14h50m) and in the star formation (Dressler & Gunn 1983; Bothun & Dressler SouthernGalactichemispheres(SGP;−22.5◦ >δ>−37.5◦, 1986; Vollmer et al. 2001); fusion with other systems 21h40m < α < 3h40m). The sample comprises galax- (Barnes & Hernquist 1991; Lavery & Henry 1994; Bekki ies with radial velocities between 600 and 15000 km s−1 2001);tidalinteractions(Moss & Whittle2000).Thus,pro- brighter than an extinction corrected absolute magnitude cessliketidalinteractionsmaytriggerstar-formationaswell Mlim = −17.38 +5logh, corresponding to corrected ap- as may contributeto its end. bJ parentmagnitudesb 618.50. Theadvantageofthisap- Recent galaxy surveys are allowing to investigate the Jlim proachisthattheradialselectionfunctionisuniform(inthe relationbetweenstarformationandenvironmentwithlarge caseofcomplete sampling) andvariations inthespaceden- samplesofdatacollectedinauniformway.Thisincludesthe sityofgalaxieswithinthevolumeareduetoclusteringonly LasCampanasRedshiftSurvey(Hashimoto et al.1998),the (Norberg et al. 2002). Considering only galaxies with spec- 15R-NorthGalaxyRedshiftSurvey(Carter et al.2001),the tra with quality parameter Q>3 (Colless et al. 2001), this 2dF Galaxy Redshift Survey (2dFGRS, Lewis et al. 2002), initial sample contains 8040 galaxies. As will be shown in and the Sloan Digital Sky Survey (SDSS, Gómez et al. Section 3.2, the fraction of star-forming galaxies increases 2003). Lewis et al. (2002) have analyzed the environmen- with the limiting absolute magnitude of the sample, and tal dependence of star formation rates (SFRs) near galaxy hence our results regarding the fraction of spectral classes clustersinthe2dFGRSregion,findingthatthereisacorre- are also dependenton this parameter. lation between SFR and projected density, that disappears below ∼ 1 h70 galaxy Mpc−2 (for galaxies brighter than Since we are interested in field galaxies, it is necessary to remove from the sample galaxies that are members M =−19). The SDSS data shows a strong correlation be- b of galaxy clusters. The clusters within the 2dFGRS region tween SFR and local projected density, with a “break” in were studied by De Propris et al. (2002), and their cluster this relation at the same density found near the 2dFGRS catalogue may be considered complete up to z = 0.1, well clusters; below this threshold, corresponding to a cluster- above the limit of our volume limited sample. The stud- centricradiusof∼3virialradii,theSFRvariesslowlywith ies of Gómez et al. (2003) and Lewis et al. (2002) for the local projected density. SDSS and 2dFGRS, respectively, indicate that a represen- In this work, instead of focusing on the SFR, we anal- tative sample of field galaxies cannot be obtained within ysetheenvironmentaldependenceofthepopulationofstar- ∼ 3 to 4 virial radius of the cluster core. The virial radius forming nearbyfieldgalaxies (z <0.05) based onavolume- of a cluster can be calculated as R = 0.002 σ h−1 Mpc, limited sample extracted from the 2dFGRS final data re- V r where σ is the radial velocity dispersion in units of km lease (Colless et al. 2003). The star formation is character- s−1 (Girrardi et al. 1998), and for the clusters studied by ized by spectral classes defined using the equivalent widths Gómez et al. (2003) its mean value is ∼ 1 h−1 Mpc. Con- of [O ii]λ3727 and Hδ.The environmentalparameter is the sequently,weexcludedfromthesample(aswellasfromthe local spatial density of galaxies. The paper is organized as simulations described below) all galaxies inside a sphere of follows. Section2presentsthesampleselection,themethod 4 h−1 Mpc radius around thecluster centers catalogued by adoptedforthedensityestimation, andthespectralindices DePropris et al.(2002).Theresultingsampleoffieldgalax- that will be analysed here. Section 3 describes the spectral ies contains 6768 galaxies. classification and presentstherelation between thefraction ofdistinctgalaxytypesandenvironmentinoursample.Our results are discussed in Section 4. Finally, in Section 5 we 2.2 Simulations summarize ourconclusions. Despite the fact that the 2dFGRS final data release has a high completeness, this is not true for some regions coveredbythesurvey,andweneedtotakeintoaccountthe possibility of surveyincompletenessin thedetermination of the environment associated to each object in our sample. 2 THE SAMPLE AND THE ESTIMATION OF With this aim, we generated mock catalogues to simulate LOCAL GALAXY DENSITY the distribution of galaxies in our selected volume and the distributionof theobjects selected bythe2dFGRS.Firstly, In this section we describe the sample of galaxies that wecomputedthemeangalaxynumberdensitybrighterthan willbeanalysedandpresenttheparameterthatwillbeused theadopted luminosity limit: to describe a galaxy environment, the local galaxy density. We also present the spectral indices that will be used to describe star-formation properties of the galaxies. Finally, 1 The2dFGRSdatabaseandfulldocumentationareavailableon we discuss the selection function of thesample. theWWWathttp://www.mso.anu.edu.au/2dFGRS/ Star formation and environment 3 Figure1.Redshiftcompletenessasafunctionofpositionforthesimulatedgalaxies.ThetoppanelisfortheNGPstripandthebottom panel is for the SGP strip. The horizontal axis represents the right ascension (in hours) and the vertical axis is the declination (in degrees). The completeness, between 0 and 1, is represented by different levels of intensity, with darker regions representing regions of highercompleteness,followingthescaleshownbelowthestrips. ∞ (2001)presentadescriptionofthemagnitudeandcomplete- ρ¯(>L )= Φ(L)dL (1) min Z nessmasksthatwereconstructedforthispurpose.Thered- Lmin shift completeness is given by a parameter called R, which where Φ(L) is the Schechter luminosity function with pa- dependson theposition at thestrips covered bythesurvey rameters (in b band) M∗ = −19.66+5logh, α = −1.21 andistheratiobetweentheobservednumberofobjectsand J and Φ∗ =1.66×10−2h3Mpc−3 (Norberg et al. 2001), that the total numberof objects in the parent catalogue at that are appropriate for the 2dFGRS. Then, for the NGP and position.Moreover,themaskssupplyamagnitudelimitand SGPregions,weevaluatedthemeangalaxynumberineach a µ parameter that also depends on the position. The red- region,N¯,thatwouldbeexpectedforauniformgalaxydis- shift completeness is magnitude dependent and, as shown tribution in each volume. After, for each simulation, we as- byNorberg et al. (2001), it can bewritten as sumed a Gaussian distribution (with mean N¯ and disper- c (b ,µ )=0.99[1−exp(b −µ )]. (2) sion N¯1/2) to determine the actual number of galaxies, N, z J i J i includedin each volume.Finally,for theN galaxies ineach Thus, in a second stage of the simulations, we haveapplied region we have randomly selected a position (α, δ), a lumi- the completeness masks to our simulated data to select a nosity (based on the Schechter luminosity function) and a sample of simulated galaxies with spectra within each re- redshift, assuming a uniform distribution within each vol- gion.Inthisway,eachsimulationresults,foreachregion,in ume. twovolume-limitedsamples,onethatrepresentsthe‘parent’ Atthispointweneedtoverifyhowtheincompleteness galaxydistributionintherangeofredshiftsandmagnitudes ofthesurveyaffects thesimulated catalogues. Colless et al. considered here, and another simulating the sub-sample of 4 A. Mateus Jr. & L. Sodré Jr. galaxies with 2dFGRS spectra. Thisapproach allows tode- fineacorrectionfactortothemeasuredlocalgalaxydensity, thatwillbediscussedinthenextsection.InFig.1weshow thecompletenessmapresultingfromthesimulationsforthe two strips covered by the survey (for comparison, see maps shown in Colless et al. 2003). 2.3 Local spatial density of galaxies In recent studies, the environment has been characterized either by the local number galaxy density (e.g., Hashimoto et al. 1998; Carter et al. 2001), or by the projected galaxy density (e.g., Lewis et al. 2002; Gómez et al. 2003). Weadoptedanon-parametricmethodtodeterminethe local number density of galaxies, based on the kth nearest neighbourdensityestimator(kNN).Thismethodfixavalue for k and let the volume V, centred on a given object and extending to its kth nearest neighbour, be a random vari- able. This volume is large in low density regions and small in highdensityregions. The kNNdensityestimator may be written as (Casertano & Hut1985; Fukunaga1990): k−1 Figure2.Distributionofthelocalgalaxydensitycorrectionfac- ρ= (3) tor(C).Thefilledcirclerepresentsthemedianvalueandtheerror V(r) bars are the respective quartiles of the distribution of C values. with V(r)=4πr3/3, where r is the distance to kth nearest The arrow indicates the upper limit in C adopted to select the sample. neighbour. For our purposes we have used the value k =5. Wehavemade tests for several values of k, concluding that 5 is an adequate choice, given the shape of the survey re- and contains 5463 galaxies. In Fig. 3 we present thecorrec- gions, that does not favour larger values. Additionally, we tionfactorasafunctionofpositionforthetwostripscovered have prevented an incorrect density estimate due to bor- by the survey, where we only show the regions comprising dereffectsbyexcludinggalaxieswhosekthneighbourshave theselected sample that follows the constraints on C. projected distances greater than the distance of the galaxy to the closest border of the survey region or of the sample volume. 2.4 Measurements of spectral indices Tocorrecttheestimateddensityforsampleincomplete- For each galaxy in our sample we measured the total ness, we have generated 800 pairs of simulated catalogues, (i.e, emission plus absorption) equivalent widths (EWs) of each pair comprising a ‘parent’ and an ‘observed’ sample. [O ii]λ3727, Hδ, Hβ, [O iii]λ5007, [N ii]λ6548, Hα and This procedure allowed us to compute a mean local correction factor for thedensity,C, given by [N ii]λ6583 directly from the 2dFGRS spectra; hereafter these quantities will be called ‘spectral indices’. We adopt ρ C =< p >, (4) herepositivevaluesforemission lineEWs,and negativefor ρ o absorption EWs. The indices were computed by fitting the whereρ isthenumberdensityassociated withthe‘parent’ continuumdefinedin tworegions around theline (blueand p sample and ρ is thesame for the‘observed’ sample. red continuum) and measuring the line fluxnormalized rel- o ThelocalvaluesofC werethenusedtocorrecttheden- ativetothiscontinuum.TheEWerrorswerecomputedfol- sity associated to each galaxy in our initial volume-limited lowing the prescription of Cid Fernandeset al. (2001), that sample. Fig. 2 shows the distribution of C for this sample. takesintoaccountthenoisewithinthelinewindowandthe The filled circle represents the median value of the distri- uncertaintyin thepositioning of thecontinuum. bution and the error bars are therespective quartiles. Note InSection3wewillmakeaspectralclassificationbased that,bydefinition,C >1.Toavoiduncertaintiesintroduced onthe[O ii]λ3727andHδ EWs.Thispairofindicesiscon- in the density by large corrections, we have restricted the venient because it can be measured even at high redshifts sampletothoseobjectslocatedinregionswhere16C 61.7; and, thus, it can be used directly to probe evolutionary ef- in Fig. 2 this limit is indicated byan arrow. fects in samples of more distant galaxies. The median val- Since we are neglecting peculiar velocities in distance ues of these indices are very similar to those obtained by estimates, galaxies in dense environments, where the veloc- Balogh et al. (1999) for a sample of field galaxies extracted itydispersionishigh,willprobablyhavetheirlocaldensities from the CNOC1 sample; our EW[O ii] has also mean and underestimated. However, since we have removed from the medianvaluescomparabletothoseobtainedbyGómez et al. samplegalaxies within∼4virialradiusaroundtheclusters (2003)foraSDSSsub-sample.Themedianerrors(withthe inthe2dFGRSsurveyarea(c.f.Sect.2.1),ourresultsshould quartilesoftheerrordistribution)are3.3+1.7 and1.8+1.0 ˚A −0.9 −0.6 not be strongly affected by peculiar velocities. for [O ii]λ3727 and Hδ, respectively. The error distribution At this point, the sample volume is ∼ 14400h−3Mpc3 has a long tail towards large errors and, thus, we excluded Star formation and environment 5 Figure3.Correctionfactor,C,asafunctionofpositionfortheselectedregions.ThetoppanelisfortheNGPstripandthebottompanel isfortheSGPstrip.Thehorizontalaxisrepresentstherightascension(inhours)andtheverticalaxisisthedeclination(indegrees).The values of C are represented by different levels of intensity, with darker regions representing regions of lower corrections, accordingly to thescaleshownbelowthestrips. from the analysis 456 objects with EW uncertainties larger accordingly to their spectral types. We will return to this than10˚AforEW([Oii])and6˚AforEW(Hδ).Additionally, point in Section 4. following Lewis et al.(2002),weexcluded225galaxieswith EW(Hα) > 10 ˚A and EW([N ii]λ6583) > 0.55EW(Hα). 2.5 The selected sample Theseobjectsareclassifiedasactivegalacticnuclei(AGNs) andhavebeenremovedfromthesamplebecausetheyhavea After the exclusion of the objects with high uncertain- significant non-thermal component (Veilleux & Osterbrock tiesinthespectralindicesandthoseidentifiedasAGNs,the 1987),contrarily totypical star-forming H ii regions. final sample comprises 4782 galaxies, whose properties will beanalysed and discussed in the following sections. The sample completeness may be described by a selec- Apointthatdeservesmentionhereisthebiasthatmay tion function S(b )that wepresent in Fig. 5.This function be introduced in the analysis due to the use of small fibers J is defined as to measure the galaxy spectra. This effect, known as aper- N turebias,isdiscussedindetailbyKochanek,Pahre & Falco S(b )= sel, (5) (2000), who demonstrated that it can lead to an under- J NT estimate of EW values, and consequently, to an overes- where, for a given magnitude bin, N is the number of sel timate of the fraction of early-type galaxies in a survey. selected galaxies and N is the total number in the vol- T Madgwick et al. (2002) discuss the presence of this effect ume. Fig. 4 shows that the selection function remains con- in 2dFGRS spectra, concluding that it does not introduce stant,andwithmaximumvalue,untilb ∼14.0.For14.0. J anysignificantbiasinthefractionsofgalaxiesdistinguished b .14.5thefunctionpresentsavariablebehaviourand,for J 6 A. Mateus Jr. & L. Sodré Jr. 3 SPECTRAL CLASSIFICATION AND ANALYSIS 3.1 Spectral classification Several authors have used EWs in galaxy classification (e.g., Hamilton 1985; Poggianti et al. 1999; Balogh et al. 1999; Bekki, Shioya& Couch 2001; Poggianti, Bressan & Fransceschini 2001). One advan- tage of EWs is that they are not affected by extinction, although they are sensitive to differences in the extinction of the regions that produce the emission lines and those producing the continuum (the selective extinction; e.g., Stasin´ska & Sodré 2001). Moreover, EWs are relatively insensitive to changes in instrumental resolution and they can be measured in non-flux calibrated spectra, like the 2dFGRS data. Here we adopt the [O ii] and Hδ EWs to classify the galaxies in the sample. These lines can be measured up to large redshifts, and thus our results may be considered a z = 0 calibration for evolutionary studies of the environmental dependence of the fraction of star forming galaxies. The EW of the [O ii]λ3727 doublet is associated with the presence of young, massive stars, and is a useful tracer Figure4.Selectionfunctionasafunctionofapparentmagnitude (bJ)fortheselectedsample. of star formation at the blue side of an optical spectrum. For this reason it has been adopted in many studies of starformationpropertiesingalaxies.Infact,worksasthose ofGallagher, Hunter& Bushouse(1989),Kennicutt(1992), and Tresse et al. (1999) have shown that the [O ii] line presentsagoodcorrelation withprimarytracersofstarfor- mation, like Hβ and Hα emission lines. The other spectral index we adopted for the spectral classification is the EW of the Balmer line Hδ. When in emission, this feature is produced in objects with high increment in the star formation rate. On the other hand, strong absorption in Hδ is associated togalaxies whichhadanintensestarburstended about 1–2 Gyr ago (Barbaro & Poggianti 1997). Following Balogh et al. (1999), in this work we will classify galaxies in spectral classes accordingly to their position in the EW([O ii])–EW(Hδ) plane, which relates an index linked to star formation activity to another associated with star- burstage.Additionally,wedefineanobjectasastar-forming galaxy if it has EW([O ii]) >5˚A, that is about 1.5σ above our detection limit. It is also important to note again that weusepositivevaluestoindicateemissionlinesandnegative valuesforabsorptionlines.Weadopthere3spectralclasses (see Balogh et al. 1999): • passivegalaxies(P:EW([Oii])65˚A):galaxieswithout evidence of significant current star formation (in general E or S0); Figure 5. Magnitude distribution for the initial sample (with • short starburst galaxies (SSB: EW([O ii]) > 5 ˚A, NT =5463 galaxies), shown as a solidline, and for the selected EW(Hδ) > 0): galaxies where a large fraction of the light galaxies(NSel=4782), shownasadashedline. comes from a starburst that started less than ∼ 200 Myr ago; • ordinary star-forming galaxies (SF: EW([O ii]) > 5 ˚A, EW(Hδ) < 0): includes most normal spirals and irregulars, faintermagnitudes,itisapproximatelyconstant,decreasing that have been forming stars for several hundred million for b & 17.0 and reaching the final value S(b ) ∼ 0.88. J J years. Fig. 5shows thedistribution of magnitudesfortheselected objects (dashed line) and for the initial sample (solid line). Fig. 6 shows the EW([O ii])–EW(Hδ) plane with the This figure suggests that our selection procedures did not distributionof galaxies in theregions ofeach spectralclass. introduceanysignificantbiasinthemagnitudedistribution ThetotalnumberofgalaxieswithstarformationisN = TSF of our final sample. 2750, ofwhichN =285andN =2465, andthenum- SSB SF Star formation and environment 7 Figure6.TheEW([Oii])–EW(Hδ)plane,indicatingtheregions Figure7.Fractionofgalaxiesineachspectralclassasafunction that define the three spectral classes (P, SF, SSB) discussed in of the absolute magnitude. For each class, each magnitude bin thiswork. contains about the same number of objects; the error bars were computedassumingaPoissonianstatistics. ber of those without evidence of ongoing star formation is NP = 2032. Considering only galaxies with star formation, emission line, has been calculated for the Las Campanas SSBs correspond to ∼10% of the total. This fraction com- RedshiftSurvey(Lin et al.1996)andtheESOSliceProject pares very well with that obtained by Balogh et al. for a (Zuccaet al.1997),aswellasforthe2dFGRS(Folkes et al. sample of field galaxies (∼ 9% of star-forming galaxies). 1999;Madgwick et al.2002).Thesestudieshavefoundthat Note that the classification of Balogh et al. also includes star-forming galaxies tend to be less luminous than pas- other two classes: K+A e A+em. These classes comprise sive galaxies. These results are also confirmed by studies galaxies with strong Hδ absorption (EW(Hδ) < −5 ˚A). In of luminosity function of samples selected by morphologi- our classification, the K+A galaxies belongs to the P class cal (Marzke et al. 1998) and spectral types (Bromley et al. and the A+em are in the SFclass. 1998; Folkes et al. 1999), which have shown that ellipticals It isworth mentioning thatthe2dFGRSdatabase pro- andlenticularstendtobebrighterthanlate-typespiralsand vides for each galaxy a spectral type, derived from a Prin- irregulars. cipal Component Analysis of the spectra (Madgwick et al. Fig.8showstheluminositydependenceofHαand[Oii] 2002). These spectral types are strongly correlated with EWs.ThereisacleartrendindicatingthattheseEWstend EW(Hα) and, consequently, they follow the same trends to decrease with increasing luminosity, in a fashion analo- than EW(Hα). gous to Fig. 7. The EW of Hα is related to the ratio be- tweentheUVfluxemittedbyyoungstarsandthefluxfrom theold stellarpopulation thatproducesmostofthecontin- 3.2 Correlations with luminosity uumat thelinewavelength.Thus,alarge EW isdueeither to a large UV flux and/or to a small continuum from the The fraction of galaxies in each spectral class (relative old stars. Consequently, galaxies with high EW(Hα) form to the total number of objects in the selected sample), is a blue population, while those without Hα emission are shown in Fig. 7 as a function of the absolute magnitude. preferentially redder (Kennicutt& Kent 1983). The same The fractions of SF and P classes show a strong correla- trendisobservedinthecaseof[Oii]λ3727.Assuggestedby tion with luminosity: low-luminosity objects are essentially Tresse et al.(1999),theionizationsourceswhicharerespon- galaxiesthatpresentevidenceofstarformationactivity,gen- sibleforHαemission arealso related tothe[O ii]emission, erally late-type spirals and irregulars; the opposite is seen as well as to other metallic emission lines, like those from for brighter objects, which include mainly passive galaxies. [S ii], indicating that low-luminosity star-forming galaxies The population of SSB galaxies present an excess of ob- tend to have large emission line EWs in both the blue and jects with low luminosity, with their fraction increasing for M −5logh&−18; SSB galaxies brighter than this value thered side of theoptical spectrum. bJ do not show trends with luminosity. Relations between star formation and luminosity as 3.3 Environmental distribution of spectral classes those shown in Fig. 7 have been detected in many surveys. Indeed, the luminosity function of nearby galaxies, The normalized cumulative distribution of all galaxies divided accordingly to the presence or lack of the [O ii] in the sample (N = 4782), of all star-forming galaxies ALL 8 A. Mateus Jr. & L. Sodré Jr. Figure 9.Cumulative distributionof the number of galaxies as a function of the density for all galaxies in the sample (thicker solidline), for all galaxies with star formation(SF + SSB; solid line),andforpassivegalaxies(dotted line). Figure 8. EW(Hα) and EW([O ii]) as a function of absolute isnotasignificantcorrelationbetweenfractionandthelocal magnitude. Thetriangles aremedianvalues evaluated inmagni- density for this class. These results are robust, even for the tudebinscontainingaboutthesamenumberofobjects.Theerror barsarethequartilesoftheEWdistributionwithineachbin. less populated SSB class, as we verified by repeating this analysisconsidering onlygalaxies with errorsin EW([O ii]) and EW(Hδ) below the median. All the trends have been (SF+SSB; N = 2750), and of passive galaxies (N = confirmed,at similar levels of significance. TSF P 2032), as a function of the local galaxy density ρ, is shown Since the fraction of star-forming galaxies decreases inFig.9.Passivegalaxieshaveadistribution,relativetothe with increasing density whereas that of SSBs is relatively curve for all galaxies, skewed towards high density values, densityinsensitive,indenseenvironmentsSSBsbecomerel- whereastheoppositetendencyisseenforstar-forminggalax- ativelymorefrequentamongstar-forminggalaxies.Thisbe- ies. This indicates that there is, indeed, a star formation– haviourhadbeen noticed already in a sampleof galaxies in environment relation, in the sense that the population of the Shapley supercluster (Cuevas 2000). From a study of 8 star-forming galaxies decreases in denser environments. Abell clusters, Moss & Whittle (2000) have also concluded Fig. 10 presents the relation between the fractions of that the fraction of SSBs among spirals increases from re- galaxies in different spectral classes – SSB, SF and P – as gions of lower tohigher local galaxy surface density. a function of ρ. Each density bin contains about the same It is also interesting to examine in this sample the en- number of galaxies (ranging from 32, for SSB, to 274, for vironmental dependence of the EWs of emission lines that the SF class); the errors were computed assuming a Pois- are sensitive to star formation. In Fig. 11 we show the dis- son statistics. The figure shows that the fraction of passive tributionsofEW(Hα)andEW([O ii])asafunctionoflocal galaxies increases with increasing density; in contrast, the galaxydensity,forthe4782objectsinoursample.Thefilled SF fraction decreases with ρ. This figure also shows that circles linked by solid lines represent the median values for theincidenceofSSBisessentiallyenvironmentindependent: each density bin (containing ∼ 399 galaxies), whereas the their fraction is ∼6 per cent everywhere. The behaviour of error barsare therespectivequartiles. The EWsof Hα and thepopulationswithlocaldensityissmoothandanythresh- [Oii]tendtodecreaseregularlywithincreasingdensity.The old is seen overthis range of densities. declineoftheseEWswithρispresentintheupperandlower Assuming that the trends shown in Fig. 10 are linear, quartiles, as well as in the median values. Assuming again wemayexaminetheirsignificancethroughthePearsoncor- linear correlations, we obtain in this case r equal to −0.94 relation coefficient, r (e.g., Press et al. 1992). The P and and −0.96, and p equal to 6.2×10−6 and 1.2×10−6, for SFclasses haver equalto0.97and−0.96,respectively,and EW(Hα) and EW([O ii]), respectively. Thus, we may con- theprobabilitypofthenullhypothesisofzerocorrelationis clude the mean star formation properties of field galaxies, equal to 1.7×10−5 and 4.0×10−5, for P and SF. For the probed by these EWs, varies smoothly with the environ- SSB we have r=−0.39 and p=0.30, indicating that there ment,even at low densities. Star formation and environment 9 Figure10.Fractionofgalaxiesineachspectralclassasafunction of the local galaxy density. Each density bin contains about the samenumberofgalaxies.Theerrorbarsarecomputedassuming aPoissonstatistics. 3.4 Aperture bias At this point it is convenient to revisit the prob- Figure 11. Hα and [O ii] equivalent width distributions as a lem of aperture bias (see 2.4). This effect was not con- function of local density of galaxies. The filled circles connected sidered in the sample selection nor in the previous anal- by a solid line represent the median values of each distribution ysis. However, it may introduce a redshift dependence in andthe errorbars arethe respective quartiles. Each density bin the measured galaxy spectra, since the fraction of galaxy contains aboutthesamenumberofgalaxies(399). light received by a fiber increases with increasing distance. Zaritsky,Zabludoff & Willick (1995), based on an analysis of LCRS spectra, suggest that for reshifts z > 0.05 the ef- Surveydatatostudytherelationbetweenlocaldensityand fectisminimized.Thisredshiftisexactlytheupperlimit of star formation in galaxies with same concentration index. our sample and, thus, it is necessary to verify whether our Their results show that star-forming galaxies are preferen- results are significantly affected by this bias. tially found in low-density environments, and the authors For that, we investigated the behaviour of the EWs of point out that the SFRof galaxies with similar structureis [O ii],Hδ,andHαasafunctionofredshift.Wedividedthe sensitivetothelocalgalaxydensity.Similarresultswerealso sample in several redshift bins containing thesame number obtained by Carter et al. (2001), using the spectroscopic of objects, and computed the median value and the quar- data from the 15R-North galaxy redshift survey. Here we tiles of each spectral indice in each bin. Fig. 12 shows as haveconsidered asample offieldgalaxies, showingthat the solid lines the median values of each distribution, as well fractionsofthePandSFspectralclassesvarysmoothlywith as their respective quartiles. Each redshift bin contains the thelocal density. same number of objects. Clearly, the median values of the Lewis et al. (2002) have used 2dFGRS spectra to in- equivalent widths of the three lines analysed here do not vestigate thestar formation rate of galaxies in different en- ∗ show any trend with the redshift, meaning that our mea- vironments around clusters using a parameter µ which is surements seem to beunaffected by theaperture bias. proportional to EW(Hα).They found that thereis, indeed, acorrelationbetweenSFRandlocalprojecteddensitywhich is significant only for projected densities above ∼ 1 galaxy Mpc−2 (for galaxies brighter than M = −19, assuming 4 DISCUSSION b H0 = 70 km s−1 Mpc−1), that corresponds approximately The environmental behaviour of the fractions of SF to the mean density at the cluster virial radius. Using data and P spectral classes, shown in Fig. 10, complements re- from the first release of the SDSS project (Early Data Re- sults previously obtained by other authors mainly for the lease, EDR), Gómez et al. (2003) also obtain a relation be- SFR. For example, Balogh et al. (1998, 1999) found that tweenSFRandprojecteddensity.Theyshowthatthereisa cluster galaxies have lower star-formation rates than field characteristic break in the SFR distribution at a local pro- galaxies with similar disk-to-bulge ratio and luminosity. jected galaxy density of ∼ 1h−2Mpc−2, corresponding to a 75 Hashimoto et al. (1998) used the Las Campanas Redshift clustercentricradiusof∼3–4virialradii.Belowthisdensity 10 A. Mateus Jr. & L. Sodré Jr. act everywhere, like tidal interactions, should be playing a relevant role. At this point it is convenient to discuss the behaviour of SSB galaxies. We have shown in Section 3 that the fraction of these galaxies is essentially independent of their luminosity and local number density, at least for the range of values of these quantities present in our sample. We have also shown that an increase in density tends to decrease the fraction of star-forming galaxies, and then the observed independence of the SSB fraction with ρ is somewhat unexpected. In fact, such a behaviour may be an indication that interactions, either between galaxies, or between a galaxy and its environment, may trigger a starburst in a galaxy. Moss & Whittle (2000) suggest that tidal effects due to gravitational interactions (ei- thergalaxy−galaxy, galaxy−group orgalaxy−cluster) orig- inateburstsofstarformationandmorphologicaldistortions in spiral galaxies, with a frequency increasing within high- densityenvironments.Similarresultshavebeenreportedby Hashimoto & Oemler(2000).Davis et al.(1997),inastudy of poor groups of galaxies, also suggest that theinteraction of a galaxy with their neighbours increases its star forma- tionrate;however,theypointoutthatmorphological pecu- liarities produced by tidal effects may leave the H i disks more vulnerable to external hydrodynamic forces, turning subsequent gas removal more efficient. Consequently, any star-formation burst in high density environments should beshort-lived in this scenario. In order to verify whether tidal interactions are actu- ally inducing thestarbursts that characterize SSBgalaxies, Figure 12.DistributionofEWsasafunctionoftheredshift:a) we tried to verify whether these galaxies in our sample do EW([O ii]), b) EW(Hδ), c)EW(Hα). It is also shown, for each showanyevidenceofmorphologicaldistortions.Forthis,we distribution,theirmedianvaluesandtherespectivequartiles(as have examined their images, using the SuperCOSMOS Sky solidlines). Survey2(Hambly et al. 2001). We divided the SSB galaxy sample in two groups containing 50 objects in each one, comprisinggalaxies located inthebinoflowest galaxy den- theSFRincreasesonlyslightly,whereasatdenserregionsit sity(−2.44<logρ(Mpc−3 h−3)<−1.52) andinthebinof isstronglysuppressed.Thus,ifoneassumesthattheequiva- highestdensity(−0.40<logρ(Mpc−3h−3)<0.46),respec- lentwidthofHα,forinstance,isproportionaltothestarfor- tively. In low-density regions, the fraction of SSBs showing mationrateofagalaxy,likeassumedbyLewis et al.(2002), any kind of peculiarity is ∼ 10 per cent. When we analyse our results are similar to those obtained by Gómez et al. imagesofthegalaxiesinthedensestenvironments,thisfrac- (2003)forregionsfartherthan3virialradiiofclustercores. tion increases to ∼ 30 per cent. As a conclusion, although Thus, the correlation between theHα and [O ii]equivalent tidal interactions may not be the only mechanism produc- widths and the local density shown in Fig. 11 reinforce the ing the starburst that characterizes SSB galaxies, they are idea that even low-density environmentsplay an important probably having a major role in inducing the large rates of roleinstablishingthefractionofpassiveand/orstarforming star-formation observed in these objects, even in the low- galaxies. density regions of thefield. Our results also show that the fraction of star-forming galaxiesvariessmoothlyalongallrangeofdensitiescovered by our sample, suggesting that the population of galaxies showing evidences of star formation activity seems to be 5 SUMMARY AND CONCLUSIONS affected by the environment everywhere. This is an impor- In this paper we have investigated the environmental tantissuethatneedstobeclarified,becauseitmayprovide dependence of the population of star-forming galaxies with importanthintsaboutthephysicalmechanismsthatarebe- avolume-limitedsampleoffieldgalaxiesextractedfromthe hind the relation between star formation and galaxy envi- finalreleaseofthe2dFRedshiftGalaxySurveyandcontain- ronment. Indeed, if only environments with relatively large ing 4782 objects. We have adopted a spectral classification densitiesaffectsstarformation,processesthatoccurmainly thatcharacterizesthestarformationthroughtheequivalent inclusters,likeram-pressurestrippingorgalaxyharassment widthsof[Oii]λ3727andHδ.Theenvironmentisdescribed may be considered the drivers of the trends observed between star formation and environment; on the other hand, if this trend is present for galaxies in clusters as well as for those in the field, as our results support, mechanisms that 2 http://www-wfau.roe.ac.uk/sss/