MNRAS000,1–17(2016) Preprint10January2017 CompiledusingMNRASLATEXstylefilev3.0 xGASS: Gas-rich central galaxies in small groups and their connections to cosmic web gas feeding Steven Janowiecki,1(cid:63) Barbara Catinella,1 Luca Cortese,1 Am´elie Saintonge,2 Toby Brown,1,3 Jing Wang4 1InternationalCenterforRadioAstronomyResearch(ICRAR),M468,TheUniversityofWesternAustralia,35StirlingHighway,Crawley,WA,6009,Australia 2DepartmentofPhysicsandAstronomy,UniversityCollegeLondon,GowerPlace,LondonWC1E6BT,UK, 3CentreforAstrophysicsandSupercomputing,SwinburneUniversityofTechnology,Hawthorn,VIC3122,Australia, 7 4CSIROAstronomy&SpaceScience,AustraliaTelescopeNationalFacility,P.O.Box76,Epping,NSW1710,Australia 1 0 2 AcceptedtoMNRAS n a J ABSTRACT 6 ] We use deep Hi observations obtained as part of the extended GALEX Arecibo A SDSS survey (xGASS) to study the cold gas properties of central galaxies across G environments. We find that, below stellar masses of 1010.2 M(cid:12), central galaxies in groups have an average atomic hydrogen gas fraction ∼0.3dex higher than those in . h isolation at the same stellar mass. At these stellar masses, group central galaxies are p usually found in small groups of N=2 members. The higher Hi content in these low - massgroupcentralgalaxiesismirroredbytheirhigheraveragestarformationactivity o and molecular hydrogen content. At larger stellar masses, this difference disappears r t and central galaxies in groups have similar (or even smaller) gas reservoirs and star s a formation activity compared to those in isolation. We discuss possible scenarios able [ to explain our findings and suggest that the higher gas content in low mass group central galaxies is likely due to contributions from the cosmic web or Hi-rich minor 1 mergers, which also fuel their enhanced star formation activity. v 4 1 INTRODUCTION clusters. This pre-processing can occur via mergers (Mihos 5 2004), through gas interactions (Fujita 2004), or through 7 Studieshavelongshownarelationshipbetweengalaxymor- tidalinteractions(Mooreetal.1998),andhasbeenobserved 1 phology and environmental density (e.g., Hubble & Huma- ingalaxygroupsinthelocalUniverse(Corteseetal.2006). 0 son 1931; Dressler 1980; Postman & Geller 1984). At high 1. densities, galaxy clusters are predominantly inhabited by Even though pre-processing makes a significant contri- 0 gas-poor, red, passive galaxies, while increasingly low den- bution to galaxy evolution, it is difficult to study in small 7 sity areas are populated by galaxies which are increasingly groups. First, galaxy groups (with (cid:46)10 members) are dif- 1 blue, gas-rich, and actively star-forming. A strong relation ficult to consistently identify in optical galaxy surveys for v: has been shown between a galaxy’s morphology and its statistical reasons (see Section 4 and Berlind et al. 2006). i cluster-centricradius(Whitmoreetal.1993),whichdemon- Second, incompleteness in optically selected group catalogs X strates the connections between environmental density and is especially problematic for small groups, whose satellite r galaxy transformations. Galaxies falling into rich clusters members are often too faint for optical spectroscopy, but a are observed to experience rapid evolutionary transforma- canbeidentifiedbydeepHi observations(Kernetal.2008) tions through dramatic mechanisms including ram-pressure andblindHi surveys(e.g.,Hess&Wilcots2013;Odekonet stripping(Chungetal.2009)andstarbursts(seealsoBoselli al.2016).Third,sincegas-removalisoneofthehallmarksof & Gavazzi 2006). group pre-processing, the most-processed galaxies will also be the most difficult (and important) to detect in Hi and While striking and dramatic, these rapid transforma- H . tions in high-density environments are not the most impor- 2 tant environmental mechanism of galaxy evolution. Studies These challenges have lead to a wide variety of results have shown that cluster infall alone is insufficient to pro- in the literature. In recent optical studies of environment, cess field galaxies into cluster galaxies while still maintain- Bamford et al. (2008) found that at a fixed optical colour, ingobservedscalingrelationsacrossenvironments(Blanton the morphology-density relationship disappears, while Park & Moustakas 2009). In order to maintain both global scal- et al. (2007) found that at a fixed morphology and stellar ing relations and the morphology-density relation, galaxies mass,notrendswithenvironmentaldensityareobserved(in mustexperiencesignificantevolutionthroughpre-processing colour, concentration, size, star formation rate, etc.). Dif- in small groups before they eventually merge into larger ferent studies have found that a galaxy’s host dark matter ©2016TheAuthors 2 Janowiecki et al. halo mass is the primary driver behind environmental ef- Centralgalaxiesarethedominant(mostmassive)memberin fects (e.g., Blanton & Berlind 2007) while others conclude theirgrouporcluster,butaresometimesalsodefinedasthe thatthelocaldensityfielddrivesenvironmentaleffects(e.g., BrightestClusterGalaxy(BCG)orBrightestGroupGalaxy Kauffmann et al. 2004). (BGG) (as discussed further in Section 4). Central galaxies Hi studiesofotherwisesimilargalaxiesacrossdifferent usually reside at the center of the group’s dark matter halo environments have demonstrated that Hi-deficient galaxies butcanalsobefoundinisolation.Centralgalaxiesingroups are common in the high-density cluster environment (Gio- grow primarily by mergers and interactions, while isolated vanelli & Haynes 1985; Solanes et al. 2001) and also in the galaxies experience mostly secular evolution (e.g., Lacerna lowerdensitygroupenvironments(Verdes-Montenegroetal. et al. 2014, and references therein). 2001; Kilborn et al. 2009). However, observations have also Central and satellite galaxies are thought to follow dif- shown that Hi-rich galaxies in groups are more likely to be ferentevolutionarypathwaysastheyareaffectedbydifferent foundinHi-richenvironments(Wangetal.2015),analogous mechanisms. Satellite galaxies can experience a wide range to the conformity of galaxy colours in groups and clusters of environmental effects (e.g., ram pressure stripping, tidal foundbyKauffmannetal.(2010).Continuingtothesmall- interactions, etc.) while the evolution of central galaxies is estgroupscales,simulationsandobservationsofgalaxiesin more closely tied to their halo mass, involving fewer mech- pairshavefoundthattheyareenhancedinHi (Tonnesen& anisms, and central galaxies make a greater contribution to Cen2012)andSFR(Lambasetal.2003;Pattonetal.2013) the growth of stellar mass in galaxies (Rodr´ıguez-Puebla et compared to un-paired galaxies. al. 2011). The environmental effects on the Hi content of Taken together, most Hi studies of environment com- satellite galaxies are discussed in Brown et al. (2016) and prise a heterogeneous set of observations with a variety of will not be considered further in this work. sensitivities, sample selections, and multi-wavelength cov- Inthisworkwecomparecentralgalaxiesingroupsand erage. Blind Hi surveys such as the Arecibo Legacy Fast in isolation in order to identify possible environmental ef- ALFA(ALFALFA,Giovanellietal.2005;Haynesetal.2011) fects on their gas and star-formation properties. We also survey are providing large samples of galaxies, but cannot consider the effects that group size (i.e., total dark matter observe the gas-poor regime (i.e., those in group or cluster halo mass or multiplicity) and local environmental density environments)exceptforthemostnearbygalaxies(Gavazzi (i.e.,thedensityofnearbygalaxieswithin1Mpc)mayhave et al. 2013). on the properties of central galaxies in our sample. These Thegas-richpopulationofALFALFAgalaxieshasbeen environmentalmetricsaresomeofthemostcommonlyused used by Hess & Wilcots (2013) to study a sample of galaxy when studying the role of environment on galaxy evolution groups. They find that the fraction of Hi-detected group (Blanton & Berlind 2007). Finally, we make comparisons members decreases as group membership increases. AL- between galaxies in different environments at fixed stellar FALFA Hi data have also been used in stacking analyses mass, since many galaxy properties (e.g., star formation, (e.g., Fabello et al. 2011), which combine Hi spectra from size, luminosity) scale primarily with stellar mass (Kauff- non-detectedgalaxies, binnedby other properties (like stel- mann et al. 2003). lar mass). Brown et al. (2015) stack ALFALFA spectra in This paper is organised as follows. Section 2 describes a sample of ∼25,000 galaxies to study Hi scaling relations and characterizes the sample of galaxies used in this work. fully across the range of gas-rich to gas-poor galaxies. Still, Section 3 and Section 4 describe our determinations of star thestackingstudiesarelimitedtomakingstatisticalconclu- formation rates (SFRs) and environment metrics, respec- sions about the average properties of galaxies in each bin. tively. Section 5 describes our main results, and Section 6 To improve on the environmental coverage and depth discusses these results and their implications. We summa- of Hi surveys, the GALEX Arecibo SDSS Survey (GASS, rizeourmainconclusionsinSection7.Throughoutthiswork Catinella et al. 2010) observed a sample of ∼800 galaxies we use a ΛCDM cosmology with H0=70km s−1 Mpc−1 and with Arecibo until they were detected in Hi or reached ΩM=0.3. an upper limit of 0.015-0.05 in Hi gas fraction (MHi/M∗). Thissamplewasthefirsttosimultaneouslycoverasubstan- tial volume and measure Hi in galaxies across the gas-rich 2 xGASS SAMPLE and gas-poor regimes. One of GASS’s main environmental findings was that massive galaxies (M∗/M(cid:12)>1010) in large ThexGASSsurveyisanextensionofGASS(Catinellaetal. halos (1013<Mh/M(cid:12)<1014) have at least 0.4 dex lower Hi 2010) to include lower stellar mass galaxies (the GASS-low gas fractions than those with similar M∗ in smaller halos sample). (Catinella et al. 2013). The original GASS sample (of Catinella et al. 2013) Inthiswork,weusetheextendedGASSsample(xGASS was selected to have a flat distribution of stellar mass be- Catinella et al. 2017), which includes additional galaxies tween 1010≤M /M ≤1011.5 and redshifts 0.025≤z≤0.05. ∗ (cid:12) at lower stellar masses. Our Hi observations are exception- Each member of the GASS sample was observed in Hi un- allydeepandrepresentthelargestsampleofgalaxieswhich til detected or until an upper limit on the gas fraction probes the gas-poor regime across field and group environ- (MHi/M∗) of 0.01−0.05 was reached. Since GASS did not ments. These Hi measurements allow us to witness the full targetgalaxiesalreadydetectedbyALFALFA,theobserved range of environmental effects on a galaxy’s gas, from the sample lacked the most gas-rich objects, which needed to delicateeffectsofpre-processinginloosegroups,tothecon- be added back in proportions related to the ALFALFA de- spicuous transformative effects in large clusters. tection fractions in the GASS parent sample (see Catinella Inparticular,wefocusontheeffectsofenvironmenton et al. 2010, for complete details). This yielded the GASS the gas and star formation properties of“central”galaxies. representative sample (760 galaxies), which was based on MNRAS000,1–17(2016) Gas-rich low mass central galaxies 3 statisticsestimatedfromthe40%datareleaseofALFALFA are small. If these were observed to have extremely low Hi (Haynesetal.2011)andalsoincludedtheHi digitalarchive masses, our primary results would only weakly be affected, (Springob et al. 2005). With the recent 70% data release as our sample includes 55 low mass group central galaxies. (AA701)oftheALFALFAblindHi survey,werevisitedthe In this work we combine the xGASS and AA70gcent GASS representative sample to just include homogeneous samples, removing Hi-confused galaxies, those with no esti- AA70 observations and updated detection fractions. It is matesofSFR(seeSection3),andthosenotmatchedinthe important to remind the reader that, by construction, the group catalog (see Section 4). This leaves a final sample of representative sample still has as flat a stellar mass distri- N=1080 galaxies, of which there are 234 central galaxies in bution as the original GASS sample. The updated GASS groups and 525 in isolation. representative sample includes 781 galaxies. We also use CO(1-0) observations of a subset of the Galaxies in the GASS-low sample are selected from a xGASS sample to estimate their molecular hydrogen (H ) 2 parent sample extracted from SDSS DR7 (Abazajian et al. content.TheseobservationscomefromtheCOLegacyData 2009) having stellar masses 109≤M /M <1010.2 and red- base for the GASS survey (COLD GASS, Saintonge et al. ∗ (cid:12) shiftsbetween0.01≤z≤0.02.208galaxiesselectedrandomly 2011) and its low mass extension (COLD GASS-low, Sain- wereobservedwiththeAreciboradiotelescope.Wefollowed tonge et al. 2017). Analogously to COLD GASS, the low the same gas fraction limited strategy as GASS, but with- mass extension is a follow-up of a random subset of GASS- outimposingaflatstellarmassdistribution.Thisisbecause low, hence its M and z intervals are identical for xGASS ∗ at these masses the stellar mass function is flatter and we andxCOLDGASS.ThexCOLDGASSsampleprovidesH 2 sample almost equally all the stellar mass range of interest estimates for ∼400 of the galaxies in xGASS. Full details by construction. As in the case of GASS, for GASS-low we aboutxCOLDGASSanditspropertiesareincludedinSain- did not re-observe galaxies already detected by ALFALFA tonge et al. (2017). and we created a representative sample following an anal- ogous procedure. The final xGASS representative sample, whichincludesbothGASSandGASS-lowsamples,contains 3 STAR FORMATION RATE ∼1200 galaxies. DETERMINATION No environmental or other criteria are imposed on the GASS or GASS-low sample selections. Complete details of In addition to the observations of the atomic and molecu- the xGASS sample selection and its properties are included lar gas for galaxies in our sample, we are also interested in in Catinella et al. (2017). quantifying the star formation processes underway in these Withitslarge(3.5(cid:48))beam,theAreciboHi observations objects.Inanidealdust-freegalaxy,itsultra-violet(UV)lu- are susceptible to source confusion if multiple galaxies are minositywouldbeanexcellenttracerofrecent(<100Myr) nearby each other on the sky and have similar recession ve- star formation. However, dust absorbs up to ∼70% of the locities.EachoftheHi-detectedxGASStargetsarecarefully UVfluxandre-emitsitatmid-infrared(MIR)wavelengths, checked and flagged if they have significant confusion from requiring a correction to UV SFRs (Buat et al. 1999; Bur- sources within ∼2(cid:48) in projection (where the beam power garella et al. 2013). Dust emission and absorption vary as dropstohalfitspeak)andwithin∼200kms−1 inrecession a function of galaxy properties, so multi-wavelength obser- velocity.Wealsoflagtargetswithmoredistantcontaminants vations and corrections are required to determine the total ifthenearbysourcesareparticularlygas-richgalaxies.Non- SFRs in a sample of galaxies (e.g., Boquien et al. 2016). detections in xGASS are not checked for confusion. In all, Towardthatend,wegeneratetotalSFRsforallgalaxies weidentify∼10%ofxGASStargetsassignificantlyimpacted in our sample using both UV and MIR observations. While by confusion in Hi (for complete details see Catinella et al. thereareavarietyofwell-testedandstatisticallyrobustex- 2017). In this analysis we only consider the non-confused isting multi-wavelength star-formation indicators (e.g., the sample; Appendix B shows the small changes to our results recentUV+MIRSFRsfromSalimetal.2016),thegalaxies if these confused galaxies are not removed. inoursamplearetoonearbyandtooextendedtofullyrely As will be discussed in Section 4, the xGASS on automated MIR catalog photometry, which is typically sample only contains N=38 non-confused low mass best suited for measuring fluxes of point-sources. Our total (M∗/M(cid:12)<1010.2) central galaxies in groups. To improve SFRs are determined using standard SFR indicators from these statistics, we searched for additional group central UV(Schiminovichetal.2007)andMIR(Jarrettetal.2013) galaxies within the xGASS mass and redshift range in the luminosity conversions, and include a correction for stellar Yangetal.(2007)groupcatalog(seeSection4).Wematched MIRcontamination(Cieslaetal.2014).Allluminositiesare these galaxies to Hi observations from AA70, several of computed using luminosity distances determined from the which were already included in our xGASS representative SDSS redshifts for each source. sample.However,wefoundanadditional20lowmassgroup OurUVfluxescomefromtheGalaxyEvolutionExplorer centralgalaxieswhichwerenotincludedinxGASS,ofwhich (GALEX, Martin et al. 2005; Morrissey et al. 2007) which 17 are detected in Hi by AA70, and 3 are non-detections. collected UV images and spectroscopy from 2003 to 2012. Because this sample of central galaxies is nearly complete We find matches to our sources from catalogs available in in Hi, we decided to include these 17 detected sources in theGALEX CasJobsinterface2,includingboththeBianchi our analysis and refer to them as the“AA70gcent”popula- etal.(2014)Catalog(BSCAT3),theGALEX UniqueSource tion. The potential effects of the three un-detected galaxies 2 https://galex.stsci.edu/casjobs/ 1 obtainedfromhttp://egg.astro.cornell.edu/alfalfa/data/ 3 https://archive.stsci.edu/prepds/bcscat/ MNRAS000,1–17(2016) 4 Janowiecki et al. Catalog (GCAT4), and the GR6+7 data release5 to obtain in each of the WISE bands. We also apply a small correc- observationsfromtheMediumImagingSurvey(MIS,1500s tionforstellarMIRcontaminationbasedonw1 luminosity exposures) and All Sky Imaging Survey (AIS, 100s expo- (Cieslaetal.2014,calculatedinananalogouswaytow3 and sures).GivenmultipleNUVobservationsofthesametarget, w4), and the SFR estimates in w3 and w4 come from the we choose the GCAT measurements over the BSCAT mea- calibrationinJarrettetal.(2013),asshowninEquations2 surements, and the MIS observations over the AIS observa- and 3. tions.Weusethe“auto”fluxmeasurementswithinKron-like elliptical apertures which are suitable for extended objects. GCAT-MISprovidesfluxesfor∼60%ofoursample,GCAT- SFRw3[M(cid:12) yr−1]=4.91×10−10×(Lw3−0.201Lw1)[L(cid:12)] AISprovides∼30%,BSCAT-MISandBSCAT-AIStogether (2) provide∼1%,GR6+7provides∼2%,and14objectsdonot have any UV flux measurements from GALEX. TheseGALEX catalogsalsoprovideflagsoneachpho- SFR [M yr−1]=7.50×10−10×(L −0.044L )[L ] tometric measurement, to indicate whether the photome- w4 (cid:12) w4 w1 (cid:12) try may be contaminated by neighbors or if the object has (3) beendeblendedfromaneighbor.Approximately80%ofour sources have unflagged UV and are reliable. Even when For all of the galaxies in our sample with w4 detec- including the flagged sources, we find good agreement be- tions, the stellar MIR correction was never larger than the tween these UV fluxes and those measured by Wang et al. w4 SFR.For∼50%ofthe∼250galaxiesdetectedinw3 and (2011)forthegalaxiesincommonwiththissample.Wecon- not w4,thestellarcorrectionwaslargerthanthew3 SFR, verttheNUVfluxesintoSFRsusingtheobservedredshifts andsotheMIRcontributiontothetotalSFRwassettozero. of the sources, correcting for Galactic extinction (Schlafly These ∼125 galaxies are among the reddest in the sample & Finkbeiner 2011), and using the SFR calibration from (NUV-r>4.5)andaredistributeduniformlyacrossthesam- Schiminovich et al. (2007), as shown in Equation 1. ple volume (with w3 flux errors ≤3%). The w3 emission inobjectslikethesecanbeentirelyattributedtooldstellar populations, and not to recent star formation. SFRNUV[M(cid:12) yr−1]=10−28.165LNUV[erg/s] (1) We verified at this point that there were no system- aticdifferencesbetweenSFRw3 andSFRw4 estimatesforthe OurMIRfluxescomefromtheWide-field Infrared Sur- objects which were detected and unflagged in both bands. vey Explorer (WISE, Wright et al. 2010), which mapped SFRw4 is a more reliable tracer of the SFR; the 12µm lu- the whole sky between wavelengths of 3.4 and 22 µm. Its minosity is more affected by emission from polycyclic aro- large angular resolution (6(cid:48)(cid:48)/12(cid:48)(cid:48)) means that most of its matichydrocarbonsandoldstellarpopulations(Calzettiet detections are unresolved, and the AllWISE data release6 al. 2007; Engelbracht et al. 2008), and its stellar MIR con- includes only profile-fit flux measurements. While the All- tamination correction factor is correspondingly larger. WISEstackingprocessfurtherblurstheimages(to10(cid:48)(cid:48) and For galaxies with unflagged MIR and NUV observa- 17(cid:48)(cid:48)), most of our targets are still resolved at this scale, so tions,wegeneratetotalSFRsbysummingSFRw4(orSFRw3 we are unable to use the profile-fit measurements. Instead, if necessary) and SFR , as shown in Equation 4. NUV we perform aperture-photometry on the atlas images using SExtractor (Bertin & Arnouts 1996), and use “AUTO fluxesmeasuredinKron-likeellipticalapertures.Weusethe SFRNUV+MIR =SFRw4+SFRNUV (4) w3 (12µm) and w4 (22µm) images and find that ∼90% of our sources are detected in w3 and ∼60% are detected in Combined, this gives total SFRs for ∼70% of the xGASS w4, which has coarser resolution and is less sensitive. sample.Fortheremaining∼30%ofsourceswheregoodMIR and NUV observations are not both available, we use SFRs ToensurethatourMIRfluxmeasurementsarenotcon- taminatedbyneighbors,weflagallsourceswhichSExtrac- determined from the SED fits of Wang et al. (2011), when tor identifies as blended, and also those which have aper- available. For 7 central galaxies in our sample, neither ac- curate NUV flux measurements nor SED-fitting SFRs are tures overlapping by more than 25% with a neighbor that available. For the three galaxies with MIR-only detections, hasatleast25%asmuchfluxasthetarget(usingageomet- wecomputeMIR-onlySFRs,whicharelargerthantheSFRs ric algorithm from Hughes & Chraibi 2014). This identifies 46w3 sourcesand22w4 sourcesaspossiblycontaminated. fromMPA/JHUby0.1-0.4dex.Weexclude4galaxiesfrom our analysis for which none of the above methods can be We apply the standard aperture corrections (Jarrett applied, mainly as a result of blended sources. These four et al. 2013) to our SExtractor “AUTO” magnitudes of ±∼0.03 mag, and corrections for Galactic extinction in w1 central galaxies are shown in Appendix B and their exclu- andw2 (∼0.01mag,Schlafly&Finkbeiner2011),butnotin sion does not change our results. w3 and w4 as they are negligible. We also include a color We compared the UV+optical SED SFRs and the correction to w4 of ∼0.1 mag when w2-w3≥1.3 mag, as NUV+MIRSFRsforthesourcesincommonandfoundthat theSEDSFRsaresystematically1.49timeslargerthanthe recommended by Jarrett et al. (2013). NUV+MIRSFRs.WehaveappropriatelycorrectedtheSED The SDSS redshifts are used to calculate luminosities SFRs to be consistent with the full NUV+MIR SFRs. TofurtherverifytheSFRsdeterminedfromNUV+MIR 4 https://archive.stsci.edu/prepds/gcat/ photometry,weappliedthissamemethodtotheHi-selected 5 http://galex.stsci.edu/GR6/ sample of Van Sistine et al. (2016), who determined SFRs 6 http://wise2.ipac.caltech.edu/docs/release/allwise/ fromnarrow-bandHαimagingof∼1400nearbygalaxies.For MNRAS000,1–17(2016) Gas-rich low mass central galaxies 5 GASS(updated) Isolatedcentrals(N=1) Groupcentrals(N>1) GASS-low Groupcentrals(N>2) 100 AA70gcent N 10 1 9.0 9.5 10.0 10.5 11.0 11.5 11.0 11.5 12.0 12.5 13.0 13.5 14.0 1 10 100 logM [M ] logMhalo[M ] Ngal ∗ (cid:12) (cid:12) Figure 1. Left panel: histograms of stellar mass for the GASS (dashed) and GASS-low (dotted) representative samples are shown in black.Centralgalaxiesinourcombinedsampleareshowningroups(shadedgreen)andisolation(redlines).Notethatcentralgalaxies are more frequent at larger stellar masses. Also shown are the 17 group central galaxies from AA70 (“AA70gcent”). Center panel: halo mass histogram for central galaxies by environment (central galaxies without assigned halo masses are shown as dashed histograms at Mhalo=1011M(cid:12)).Rightpanel:histogramsofgroupmultiplicity(thenumberofgroupmembers,Ngal). the∼400galaxiesfromtheirsamplewhichhavereliableand low mass (109≤M /M <1010.2) group central galaxies in ∗ (cid:12) unflaggedMIR+NUVobservations,wefindgoodagreement xGASS we supplement the sample with additional galax- between the Hα and MIR+NUV SFR estimates (across 3 ies with Hi observations from AA70. The xGASS represen- orders of magnitude). The best-fit line (in log-space) be- tative sample already includes ALFALFA Hi observations tween these measurements has a slope of 0.95, an intercept in the correct proportions, but we here consider the entire of −0.01, and a scatter of ∼0.2 dex. AA70 footprint (within 0.01≤z≤0.02) and include the ad- ditional 17 low mass group central galaxies (AA70gcent) in our analysis. 4 ENVIRONMENT METRICS All but 24 of the galaxies in our sample (xGASS+AA70gcent) are matched to members of the We use multiple metrics to evaluate the environment of the DR7 group catalog“B”of Yang et al. (2007). These 24 are galaxiesinoursample.Differentmetricsaresensitivetodif- typically unmatched because of their proximity to bright ferent aspects of environment, each affecting galaxy evolu- stars or survey edges, and are not included in our analysis. tion in different ways. We also correct “false pairs” from this catalog, which are First,weusethegroupcatalogsofYangetal.(2007)to cases where a single galaxy is broken into multiple objects, identifywhethergalaxiesarecentral(mostmassive)intheir groups7,satellitemembersoftheirgroup,ornotinagroup. each separated by less than their own size (Skibba et al. 2009). After visual inspection, we find that a similar Yang et al. (2007) used a halo-based group finder (includ- threshold identifies ∼20 false pairs in the“B”catalog, and ing enhancements to typical friends-of-friends algorithms) weremovethesmallerobjectsofeachpair(seeAppendixA toidentifygroupsinSloanDigitalSkySurvey(SDSS)Data foralistofchangestocentralgalaxiesinoursample).Some Release 4 (DR4, Adelman-McCarthy et al. 2006). An up- ofthesewereidentifiedascentralgalaxiesingroupsofN=2, dated version based on Data Release 7 (DR7, Abazajian et al. 2009) has been made available online8. so were corrected to become groups of N=1. Others were satellite galaxies in groups, so their group sizes are reduced Yang et al. (2007) produce three versions of their DR7 by one. In one case (GASS 109081) a central galaxy in a group catalog, including increasingly more objects from de- group (of N=5) has been shredded into three galaxies, so creasingly reliable sources. Their“A”catalog includes only the group size is corrected to N=3. SDSS DR7 spectroscopic redshifts,“B”adds spectroscopic redshifts from other surveys,9 and “C” adds “nearest- Having made these corrections, we can now character- neighbor”redshifts, which are assigned to objects without ize thexGASS+AA70gcentsample(nowexcludinggalaxies spectra (due to fiber collisions) based on the redshifts of whichareconfusedinHi orhavenoSFRestimate)interms their nearest neighbors. Zehavi et al. (2002) find that in ofenvironmentalidentities.Wefindthat∼30%areclassified ∼40% of cases, these assigned redshifts are significantly in- as satellite galaxies in groups (and not discussed further in accurate.Weadoptthe“B”catalogasitislesscontaminated this study), ∼50% are identified as isolated central galax- by faulty redshifts than“C”, but more complete than“A” ies, and ∼20% are identified as group central galaxies (i.e. (see also Section 3.2 of Skibba et al. 2011). Our results are the most massive galaxy in their group), with at least two not strongly dependent on this choice (see Appendix B). groupmembers.Figure1showsstellarmass,halomass,and As mentioned in Section 2, due to the scarcity of group multiplicity for the xGASS and AA70gcent samples, afterremovingallconfusedgalaxiesandthosewithoutSFR estimates. Half of the groups which host central galaxies in 7 notethatweuse“group”torefertobothgroupsandclusters 8 http://gax.shao.ac.cn/data/Group.html our sample have multiplicities (total number of galaxies in 9 including2dFGRS(Collessetal.2001),IRASPSCz(Saunders the group) of N=2, and ∼80% are small groups with N≤4; etal.2000),RC3(deVaucouleursetal.1991),andKIAS-VAGC only above M∗=1011M(cid:12) are central galaxies found in large (Choietal.2010). groups with N>10 members. MNRAS000,1–17(2016) 6 Janowiecki et al. Atlowmasses(109<M /M <1010.2),ourgroupcentral ing affected by un-detected satellites around galaxies near ∗ (cid:12) galaxies are found exclusively in groups with 2-4 members: themagnitudelimitofourparentsample.Satellitegalaxies 89%arefoundingroupsofN=2;9%ingroupsofN=3,and are typically ∼2.5 mag optically fainter than their central 2% in groups of N=4. This distribution of multiplicities is galaxy, so we would be unable to detect any satellite galax- similar to that of the full Yang et al. (2007) DR7 group iesaroundacentralgalaxywhichisonly∼2.5magbrighter catalog,whichincludes∼1300groupcentralgalaxieswithin than our magnitude limit (Lacerna et al. 2014). This could this mass range. leadtoanartificialincreaseinthefractionofisolatedcentral Athighermasses(M /M ≥1010.2),∼75%liveingroups galaxies in the faintest 2.5 mag of the sample. ∗ (cid:12) ofN=2-4andtheremainderareinlargergroupsuptoN=62. To verify that this effect does not bias our results, we When we discuss results for group central galaxies, we in- createa“bright”sub-setofgalaxieswhichonlyincludesob- clude all groups regardless of multiplicity. At low stellar jects 2.5 mag brighter than the SDSS spectroscopic survey masses, our group centrals are dominated by N=2 groups, limit.Atlowmasses(M /M <1010.2),∼55%ofourcentral ∗ (cid:12) whileathighstellarmassestherearelargergroups.Wewill galaxies are included in this“bright”subset, as are ∼75% distinguish the N=2 and N>2 populations to show which of central galaxies at high masses (M /M >1010.2). The ∗ (cid:12) types of groups are driving the trends we see. isolated central galaxies in this “bright” sample are more Thegrouphalomassesareassignedwithanabundance- confidentlyisolatedgalaxies,andarenotartificiallyisolated matching method and are only available for massive halos because their satellites are too faint to be detected. This with M (cid:38)1011.5M (Yang et al. 2007); smaller halos do “bright”sub-set shows the same main relations and trends halo (cid:12) not have mass estimates. as the full sample, and is shown in detail in Appendix B. It is worth noting that identifying galaxy groups (with (cid:46)10 members) and assigning“central”or“satellite”identi- ties to galaxies is increasingly difficult for smaller groups 5 RESULTS (Berlind et al. 2006). Studies using mock catalogs have 5.1 Gas rich central galaxies in small groups shownthat itisespeciallydifficulttoidentifythegalaxyat the center of the halo (either as most massive or brightest) Our main goal is to understand the effects of the group en- inthedarkmatterhalosofsmallgroups.Skibbaetal.(2011) vironment on the Hi properties of central galaxies. Toward usedmockcatalogstoshowthattheBGGsin∼40%oflow that end, Figure 2 shows the Hi gas fraction (MHi/M∗) as mass halos (1012-1013M(cid:12)) are not located at the center of a function of stellar mass for central galaxies in our sam- the halo (i.e., the galaxy at the center of the group’s dark ple,separatedbetweenisolated(leftpanel)andgroup(right matterhaloisnotthebrightest).Thefractionoflargerhalos panel)environments.Acrossstellarmasses,galaxiesinboth (>1013M(cid:12))withthisdiscrepancydecreasesto∼25%.Simi- environments fully populate the ∼1.5 dex of Hi gas frac- larly,vonderLindenetal.(2007)usedSDSSobservationsof tion parameter space, but there are significant differences 625galaxyclusterstoshowthat∼50%oftheBCGsarenot betweenthedistributions.TheaveragevaluesofHi gasfrac- at the center of their cluster density fields. We bear these tion in each stellar mass bin show a general decrease as a challengesinmindwithoursimpledistinctionbetweencen- functionofstellarmass,withlowermassgalaxiesbeingmore tralandsatellitegalaxiesbasedontheirstellarmassranking gas-richinbothenvironments,ashasbeenpreviouslyfound within their group. (Kannappanetal.2009;Catinellaetal.2010;Corteseetal. Furthercomplicatingthispictureisthepossibilitythat 2011; Huang et al. 2012a; Brown et al. 2015, 2016). central galaxies in small groups may experience multiple However,atlowstellarmass(109≤M /M <1010.2),the ∗ (cid:12) transitionsbetweenisolationandgroupenvironments.Iftwo central galaxies in groups (shown as large green squares) small isolated galaxies interact and become gravitationally have 0.3 dex larger average Hi gas fractions than isolated bound, one will become a group central and the other a central galaxies of the same mass (red diamonds), and are satellite. If they later merge, the resulting galaxy will be- rarely found below the average value of the isolated galax- come“isolated”again. Park et al. (2008) find that a signif- ies. At these low masses, ∼90% of the groups have mul- icant fraction of isolated galaxies are actually the products tiplicity N=2. We include non-detected galaxies at their of recent mergers, and that recent mergers are even more upper limits averaging the Hi gas fractions. At moderate common among luminous isolated galaxies. These types of masses (1010.2≤M /M <1010.8) the group central galaxies ∗ (cid:12) difficultiesareinherentinanyattempttostudythesmallest have similar average Hi gas fractions to isolated galaxies, galaxy groups, and must be kept in mind. and for M /M ≥1010.8 they are more gas poor than iso- ∗ (cid:12) Inadditiontothegroupmembershipandenvironmental lated galaxies. identityofthegalaxiesinoursample,wehavealsoestimated In addition to the Hi relations, we can also test for thelocaldensityinfixedaperturesaroundeachobject.This differences between the specific SFR (sSFR) of the central calculation is made using a sample of galaxies from SDSS galaxiesingroupsandinisolation.Hi andstarformationare DR7 (Abazajian et al. 2009) with M /M ≥109 and which closelyrelated(e.g.,Kennicutt1998),andweexpectgas-rich ∗ (cid:12) fully encompasses the ALFALFA footprint (see Section 2 galaxies to have higher sSFRs. of Brown et al. 2016, for more details). The local density Figure3showstherelationshipbetweenoursSFResti- aroundeachtargetisdeterminedbycountingthenumberof mates(describedinSection3)andstellarmassforoursam- galaxieswithina1Mpc(projected)radiusand1000kms−1 ple, divided by environmental identity. The average trends velocitydifference.Theprojecteddensitiesarecalculatedin for sSFR in each environmental type are the same as those unitsofMpc−2 andincludethetargetgalaxyitself(sohave seen in the Hi gas fraction plots. When comparing group a minimum value of 1/π Mpc−2). centralgalaxieswithisolatedgalaxies,thosewithlowstellar As a final check, we also verify that we are not be- mass show larger sSFRs by 0.2−0.3 dex. MNRAS000,1–17(2016) Gas-rich low mass central galaxies 7 Isolated centrals (N=1) Group centrals (N>1) non-detections N=2 0.5 N>2 0.0 ∗ M 0.5 /I − H M 1.0 g o − l 1.5 − 2.0 − 120 168 179 58 22 33 48 131 9.0 9.5 10.0 10.5 11.0 11.5 9.0 9.5 10.0 10.5 11.0 11.5 log M [M ] log M [M ] ∗ (cid:12) ∗ (cid:12) Figure 2. On both panels, Hi gas fraction of our central galaxies is plotted as a function of stellar mass. For this and all subsequent figures,averagevalueswithinbinsareshownattheaveragex-andy-valuesofpointswithinthatbin,anderrorbarsshowstandarderror of the mean. Non-detections are included in averages at their upper limits. Left panel shows isolated central galaxies while right panel shows group central galaxies (N=2 in dark green and N>2 in light green); the average relations for isolated central galaxies are shown as large red diamonds in both panels and the averages for group centrals (at all multiplicities) are shown as large green squares. Open trianglesshowupperlimitsofnon-detections;noHi-confusedsourcesareincluded.Numbersatthebottomofbothpanelsindicatehow manygalaxieswereaveragedineachbin.HeavycolouredlinesconnectaveragesofthelogarithmoftheHi gasfraction((cid:104)logMHi/M∗(cid:105)) inbinsofstellarmass.Atlowstellarmasses(M∗/M(cid:12)<1010.2),centralgalaxiesingroups(greensymbols)arerarelygas-poorandhave Hi gasfractionswhichareonaverage∼0.3dexlargerthanthoseinisolation(inred,bothpanels). Isolated centrals (N=1) Group centrals (N>1) N=2 9 − N>2 ] 10 1 − − r y [ R F 11 S− s g o l 12 − 13 120 168 179 58 22 33 48 131 − 9.0 9.5 10.0 10.5 11.0 11.5 9.0 9.5 10.0 10.5 11.0 11.5 log M [M ] log M [M ] ∗ (cid:12) ∗ (cid:12) Figure 3. Specificstarformationrateisplottedagainststellarmassforcentralgalaxiesinbothenvironments.Thesamegalaxiesare shown with the same colour-coding and averaging as in Figure 2. The low mass group central galaxies (shown in green) have sSFRs whichareelevatedby∼0.2dexcomparedwithisolatedcentralgalaxies(showninred,withaveragerelationonbothpanels). MNRAS000,1–17(2016) 8 Janowiecki et al. 1010.2 M /M 1011.5 Isolatedcentrals(N=1) 1.0 ≤ ∗ (cid:12)≤ Groupcentrals(N>1) ) p=1E-09 p=2E-10 d0.5 e s i l a m0.0 r 109 M /M <1010.2 no1.0 ≤ ∗ (cid:12) ( N p=0.0008 p=0.007 0.5 0.0 -2.0 -1.5 -1.0 -0.5 0.0 0.5 -12.5 -12.0 -11.5 -11.0 -10.5 -10.0 -9.5 -9.0 log MHI/M log sSFR [yr−1] ∗ Figure 4. Each panel shows the Hi gas fraction or sSFR distributions in large bins of stellar mass (ranges shown at top left). Group centralgalaxiesareshadedinlightgreenandisolatedcentralgalaxiesareheavyredlines.Allhistogramsarenormalisedtohavethesame peak value. In the left column, non-detections in both environments are shown as shaded regions; no Hi-confused sources are included inanypanel.Alsoshownarethep-valuesfromatwo-sidedKolmogorov-Smirnovtestcomparingthegroupandisolatedcentralgalaxies (including Hi non-detections at their upper limits). These distributions quantify the sSFR and Hi differences between central galaxies ingroupsandinisolation. To better quantify the differences between group and 1010≤M /M ≤1011.5 and 0.025≤z≤0.05) find ∼2400 in ∗ (cid:12) isolated centrals, Figure 4 shows the distributions of Hi groups and ∼11,000 in isolation. gas fraction and sSFR in bins of stellar mass. In the low To avoid possible Hi confusion in the stacking process, massbin,theisolatedcentralgalaxieshavealargergas-poor galaxies are not included in stacks if they have a neighbor population than the group central galaxies. Also shown are within a projected separation of 2(cid:48) and velocity difference thep-valuesofatwo-sidedKolmogorov-Smirnovtestwhich smaller than ±200 km s−1regardless of their optical colour. show that the group and isolated central distributions are This threshold is quite conservative, as the Arecibo beam significantlydifferent.Atlowmasses,thedifferenceinaver- power is at half its peak at this radius and red galaxies age Hi gas fraction between the group and isolated central wouldbeunlikelytocontributeanyHi fluxtotheobserved galaxy populations is driven by a near-absence of gas-poor Hi signal. Nonetheless, this confusion criterion eliminates low mass group central galaxies. ∼1% of isolated central galaxies and ∼15% of galaxies in groups, but still gives a statistically robust sample. As an additional test, we used even more aggressive thresholds of 5.2 Consistency with Hi stacking and sSFR 3(cid:48)and300kms−1,andtheresultsareunchanged.Whilethe confusion-cleaned stacks include fewer objects, the results relations in larger samples are more reliable. Giventheinherentdifficultiesassociatedwithidentifyingthe AsdescribedinBrownetal.(2015),westartthestack- smallestgroupsofgalaxies(seeSection4)andtherelatively ing process by shifting individual HI spectra (both detec- small number of galaxies in our sample, we next explore tionsandnon-detections)toacommonrest-framefrequency. waystoverifythepropertiesoftheselowmassgroupcentral Nextweweighteachgalaxy’sspectrumbyitsstellarmass,to galaxies with larger statistical samples. stackinunitsofHi gasfraction(seeFabelloetal.2011).In First, to reach beyond the limits of our sample of Hi- eachstacktheresultingspectrumisastrongdetection,with detected galaxies, we use an Hi spectral stacking technique signal-to-noiseratios(calculatedasthepeakfluxdividedby on a much larger sample of galaxies drawn from the AL- the rms noise) between 12 and 74. FALFA blind Hi survey. While the survey depth is insuffi- cient to detect individual galaxies in the gas-poor regime, Figure5showstheHi gasfractionasafunctionofstel- stacking many Hi spectra can produce a statistical de- larmassforcentralgalaxiesinisolationandgroups,asmea- tection below its nominal sensitivity limit. We compare sured by stacking the Hi spectra of galaxies in each bin. with the sample of N∼25,000 Hi spectra from and fol- We stack the xGASS spectra in the same manner as the lowing the methodology of Brown et al. (2015) again us- ALFALFA sample (including the same 2(cid:48) and 200 km s−1 ing the Yang et al. (2007) DR7 group B catalog to test threshold cuts for confusion, which reduce the number of whether this same difference is observed. We include any xGASS objects in each bin compared with Figure 2). Be- central galaxies that match our sample selection (i.e., be- causestackingisinherentlyalinearprocess,Figure5shows tween 109≤M∗/M(cid:12)<1010.2 and 0.01≤z≤0.02, or between thelogarithmoftheaverageHi gasfraction(log(cid:104)MHi/M∗(cid:105)), MNRAS000,1–17(2016) Gas-rich low mass central galaxies 9 1.0 xGASScentrals(N=1) xGASScentrals(N=1) 112 152 173 55 9 xGASScentrals(N>1) xGASScentrals(N>1) − 0.5 17 26 32 86 ) A P 0.0 M 10 − ∗ ( M ] M/HI −0.5 1yr− 11 [− g R o 1.0 F l − S s 12 1.5 g − ALFALFAcentrals(N=1) o− l 5328 4864 3362 481 SDSScentrals(N=1) 901 5061 10602 1896 2.0 ALFALFAcentrals(N>1) − 185 438 918 753 13 SDSScentrals(N>1) − 47 296 1686 1974 9.0 9.5 10.0 10.5 11.0 11.5 9.0 9.5 10.0 10.5 11.0 11.5 log M [M ] log M [M ] ∗ (cid:12) ∗ (cid:12) Figure 5. The logarithm of the stacked Hi gas fraction Figure6. Averagevaluesofspecificstarformationrate(fromthe (log(cid:104)MHi/M∗(cid:105))isplottedasafunctionofstellarmassforcentral MPA/JHU catalog) are shown in bins of stellar mass for central galaxies in isolated (N=1, thick orange line) and group (N>2, galaxies.Relationsfromoursampleareshownwiththinlinesas light blue line) environments. Isolated (red line and points) and in Figure 5. The SDSS central galaxies are shown as light grey group(greenlineandpoints)centralgalaxiesfromoursampleare (isolated)anddarkgrey(group)points,theiraveragerelationsare also shown (confused galaxies are removed from both samples). shown as thick shaded lines, and the number of galaxies in each Jack-knifeestimatesofuncertaintiesarenotplottedonindividual bin are shown at the bottom. Heavy black error bars show the points,butarecomparabletothesizeofthesymbols.Thenum- typicalvaluesofthestandarderrorofthemeanforeachsample. ber of galaxies in each stacked bin is shown at the bottom. The Therelationsfromthelargecomparisonsampleareinagreement relations from the Hi stacking sample show the same difference withthosefromoursample,andagainthelowmassgroupcentral between the gas fractions of low mass central galaxies in groups galaxiesshowlargersSFRscomparedwiththoseinisolation. andinisolation. stellarmass,binnedinthesamewayasFigure3.Thebehav- whileourpreviousFigure2showedtheaverageoftheloga- ior of central galaxies in isolation and small groups is well- rithm of the Hi gas fraction ((cid:104)log MHi/M∗(cid:105)). matched between xGASS and the large MPA/JHU sample. Thereisgoodagreementbetweenthetrendsseeninthe Thislargesampleof∼32,000galaxiescontains∼3000group stackedxGASSHi gasfractionrelationsandthosefromthe centrals,∼300ofwhichpopulatethelowesttwobinsofstel- ALFALFA sample of Brown et al. (2015). In both samples, lar mass. the low mass (109≤M /M <1010.2) group central galaxies ∗ (cid:12) To summarise, the low mass group central galaxies ap- have Hi gas fractions which are ∼0.2 dex higher than iso- pearconsistentlyelevatedby0.2−0.3dexinHi gasfraction lated central galaxies of similar mass. In the highest mass andsSFR,whethermeasuredinoursample,intheHi stack- bin (1010.8≤M /M ≤1011.5), the group central galaxies in ∗ (cid:12) ing analysis, or in the MPA/JHU catalog. This widespread xGASS have a lower average Hi gas fraction by ∼0.25 dex agreementfurtherconfirmsthatthesegalaxiesareunusually thanthoseinthestackedsample.Thisoffsetresultsfromthe gas-rich and star-forming. difference in stellar mass distributions between xGASS (se- Next we explore whether other properties of the low lected to have a flat distribution of M ) and the ALFALFA ∗ massgroupcentralgalaxies(ortheirgroups)mighthelpex- sample (volume-limited, with a steeper power law decline plaintheirHi andsSFRproperties.Weconsidertheroleof atthesemasses).Withinthisbin,thexGASSgalaxieshave the group size (e.g., halo mass or multiplicity of members), ∼0.1dexlowersSFRsthanthoseintheALFALFAstacking theproximityofandstarformationintheirnearestsatellite sample.ThedisagreementbetweenstackedxGASSandAL- galaxy, and correlations with large-scale density measure- FALFA group central galaxies at high masses is a result of ments. different sample selection, but both samples show an offset between group and isolated central galaxies at low masses. To confirm our observed sSFR differences in a larger 5.3 Trends with group multiplicity sampleofgalaxies,weusetheMPA/JHUgalaxycatalogand theDR7groupcatalogofYangetal.(2007),selectedwithin Halo mass is an important property driving group evolu- the same stellar mass and redshift ranges as our sample. In tion, and is closely related to hydrodynamical feedback ef- thiscomparison,wealsousetheMPA/JHUsSFRestimates fectslikerampressurestripping.However,itisverydifficult forgalaxiesinoursample,forconsistencybetweenSFRcali- to estimate the total dark matter halo mass in the small brations.Figure6showstherelationshipbetweensSFRand groups of our low mass central galaxies. In particular, the MNRAS000,1–17(2016) 10 Janowiecki et al. Figure7. Hi gasfractionasafunctionofNgalforcentralgalax- Figure 8. Hi gas fraction is plotted as a function of projected ies.Isolatedcentralgalaxies(N=1)areshownasredhistograms separation between each central galaxy and its nearest satellite against the y-axis in dashed (low mass) and solid (high mass) galaxy, using the same colours and styles as in Figure 7. Aver- lines.TheiraverageHi gasfractionsareshownasopenpointsand ages and standard errors of the mean within bins are shown as horizontallinesforcomparison.Groupcentralgalaxies(N>1)are connectedpoints.Adecreasingupperenvelopeisapparentacross shownasfilledmagentatriangles(lowmass)andopengreentri- thefullpopulation;nostrongtrendsareevidentasafunctionof angles(highmass),andaveragetrendsareshownwithconnected satellitedistance. points.Adottedblacklineconnectstherelationsbetweenisolated andgroupenvironments.Thetrendamonghighmassgroupcen- trals seems to smoothly continue up to the value of the isolated lated low mass central galaxies have lower average Hi gas centrals.However,thelowmassgroupcentralsaremoreHi rich fractions than those in small groups, and the most gas-rich thancomparableisolatedgalaxies. galaxies in this population are in groups of N=2. Similar trends are also evident in sSFR as a function of group mul- tiplicity,butarenotplottedhere.Additionalexplanationis DR7 group catalog of Yang et al. (2007) does not provide requiredtoshowhowalowmasscentralgalaxywithasatel- any halo mass estimates for groups which have a central litecanbemoreHi-richandstar-formingthananotherwise galaxy below M /M <1010. ∗ (cid:12) similar isolated galaxy. Withoutgrouphalomassestimatesforallofthecentral galaxiesinoursample,weinsteadusegroupmultiplicityto comparebetweendifferentgroups.Atafixedcentralgalaxy 5.4 Characteristics of these small groups stellarmass,thereisacorrelationbetweengrouphalomass andgroupmultiplicity(forfurtherdiscussionseeFigureB2 Next,weexplorethepropertiesofthesmallgroupsthathost in Han et al. 2015). In this sub-section we compare central ourlowmasscentralgalaxiesthatareunusuallygasrichand galaxies in groups of different multiplicity. star-forming. We will explore whether they have any other Figure7showstheHi gasfractionasafunctionofgroup unusual group properties that could explain the scarcity of multiplicity(N )forthecentralgalaxiesinoursample.The Hi-poor group central galaxies. gal histogramsagainstthey-axisshowtheHi gasfractionsdis- Firstweconsidertheprojectedseparationbetweenour tributions for the isolated centrals at low (M /M <1010.2, group central galaxies and their nearest satellites (d ) to ∗ (cid:12) sat inpinkdashedlines)andhigh(M /M ≥1010.2,asredsolid explore whether recent or strong interactions from nearby ∗ (cid:12) lines) stellar masses. The average Hi gas fractions of these companions may be responsible for an enhancement in Hi twopopulationsareshownaslargedotsandhorizontallines. and SFR. Figure 8 shows the Hi gas fraction for central Our group central galaxies at high (green) and low (ma- galaxiesinoursampleasafunctionofd ,measuredinkpc sat genta) masses are plotted against their group multiplicity. (in projection). Central galaxies are binned in two intervals Averageswithinbinsofgroupmultiplicityareconnectedby of stellar mass, as in Figure 7. Isolated central galaxies are thick lines. shownasahistogramagainstthey-axis,separatedbymass. Thehighmasscentralgalaxies(ingreen)showacontin- The Hi gas fractions of the group central galaxies at uouslydecreasingaverageHi gasfractionwithgroupmulti- low and high masses show no strong trends with projected plicity from N=1 to 20+, such that the most gas-rich high- separationtotheirnearestsatellitegalaxy.Bestfittinglinear mass central galaxies are those in isolation. The low mass relations (not shown) to both the low and high mass pop- central galaxies do not follow this trend. Instead, the iso- ulations yield slopes consistent with zero (low mass slope MNRAS000,1–17(2016)