Table Of Content

DraftversionJanuary28,2015 PreprinttypesetusingLATEXstyleemulateapjv.5/2/11 EXTREMELYMETAL-POORGALAXIES:THEENVIRONMENT M.E.Filho1,2,3,4,5,J.SańchezAlmeida2,3,C.Munõz-Tunõń2,3,S.E.Nuza6,F.Kitaura6andS.Heß6 DraftversionJanuary28,2015 ABSTRACT Wehaveanalyzedbibliographicalobservationaldataandtheoreticalpredictions,inordertoprobetheenvi- 5 ronmentinwhichextremelymetal-poordwarfgalaxies(XMPs)reside. WehaveassessedtheH Icomponent 1 anditsrelationtotheopticalgalaxy,thecosmicwebtype(voids,sheets,filamentsandknots),theoverdensity 0 parameterandanalyzedthe nearestgalaxyneighbours. The aim isto understandthe role ofinteractionsand 2 cosmologicalaccretionflows in the XMP observationalproperties, particularlythe triggeringand feedingof n the star formation. We find that XMPs behave similarly to Blue Compact Dwarfs; they preferably populate a low-densityenvironmentsinthelocalUniverse: 60%occupyunderdenseregions,and 75%resideinvoids ∼ ∼ J andsheets. Thisismoreextremethanthedistributionofirregulargalaxies,andincontrasttothoseregionspre- 7 ferredbyellipticalgalaxies(knotsandfilaments).Wefurtherfindresultsconsistentwithpreviousobservations; 2 while theenvironmentdoesdeterminethe fractionof a certaingalaxytype, it doesnotdeterminethe overall observational properties. With the exception of five documented cases (four sources with companions and ] onerecentmerger),XMPs donotgenerallyshowsignaturesofmajormergersandinteractions;we findonly A oneXMPwithacompaniongalaxywithinadistanceof100kpc,andtheH I gasinXMPsistypicallywell- G behaved,demonstratingasymmetriesmostlyintheoutskirts.Weconcludethatmetal-pooraccretionflowsmay be drivingthe XMP evolution. Such cosmologicalaccretion could explainall the major XMP observational . h properties: isolation,lackofinteraction/mergersignatures,asymmetricopticalmorphology,largeamountsof p unsettled,metal-poorHIgas,metallicityinhomogeneities,andlargespecificstarformation. - Subjectheadings:galaxies:dwarf–galaxies:formation–galaxies:evolution–galaxies:interaction–galaxies: o r starburst t s a 1. INTRODUCTION Thetheorypredictsthatthesecosmologicalfilamentsarethe [ mainprocessbywhichstar formationisfedandtriggeredat Thestandardmodelof galaxyformationandevolutionin- 1 all cosmological times (Brooks et al. 2009; Verbeke et al. cludes hierarchical merging/clustering, where smaller struc- v 2014). Cosmological gas accretion should have been more tures are used as building blocks of more massive galaxies, 9 commonat high redshift, but in the local Universe, it is hy- and evolution through accretion from cosmological gas fila- 0 pothesized to intervene in the evolution of individual low- ments in the cosmic web (e.g., Combes 2004; Dekel et al. 7 mass galaxies in low-density environments, via small-scale 2009; Dekel, Sari & Ceverino 2009; Nuza et al. 2014a; 6 gasstreams,facilitatingthearrivalofgasfromlargedistances Sańchez Almeida et al. 2014b). Because merger events are 0 (Keresˇ et al. 2005; Dekel & Birnboim 2006; Dekel et al. . generallyresponsibleforthedisruptionofgasandstellarag- 1 gregates,theyhavebeentraditionallyassociatedwiththetrig- 2009;Dekel,Sari&Ceverino2009;Brooksetal. 2009;Cev- 0 erino,Dekel&Bournaud2010;Ceverinoetal.2014;Nuzaet gering of supermassive black hole accretion at the center of 5 al. 2014a). galaxies(ActiveGalacticNucleiactivity;Springeletal.2005; 1 There has been some observationalevidence for these ac- Hopkinsetal. 2008;DiMatteoetal. 2008),andwiththegen- : cretion flows at high redshift, from the detection of inverse v erationofstar formationepisodes(e.g.,Sanderset al. 1988; metallicity gradients (lower metallicity in the central star- i Robainaetal. 2009).However,thecontributionofcosmolog- X formingregionsthan in the peripheryof the sources; Cresci ical cold-accretionflows to galaxy evolution in general, and r star formation in particular, is recently gaining ground. In- et al. 2010) and from high-column density H I absorbers a (e.g., van de Voort et al. 2012). However, in the local Uni- deed,cosmologicalaccretionontogalaxiesofexternalmetal- verse,therehasonlybeenindirectevidencefortheseevents, poorgasispredictedbynumericalsimulationstobethemain mainly from the study of star formation in dwarf galaxies. modeofgalaxyanddiskassembly,athigh,andpossiblyalso For example, for galaxies of the same mass, the metallicity atlow,redshift(Birnboim&Dekel2003;Brooksetal. 2009). wasshown to decreasewhenstar formationincreases(Man- nucci et al. 2010; Lara-Lo´pez et al. 2010; Pe´rez-Montero email:mfi[email protected] et al. 2013). It was also found that metallicities in qui- 1Universidad deLasPalmas deGranCanaria-Universidad deLaLa- escent blue compact dwarf (BCD) galaxies are higher than guna, CIE Canarias: Tri-Continental Atlantic Campus, Canary Islands, Spain those of star-formingBCDs (SańchezAlmeida 2008, 2009). 2InstitutoAstrof´ısicadeCanarias,38200LaLaguna,Tenerife,Spain More recently, observations of tadpole galaxies in the Kiso 3DepartamentodeAstrof´ısica, UniversidadLaLaguna,38206LaLa- survey of UV-bright galaxies (Elmegreen et al. 2013) and guna,Tenerife,Spain extremely metal-poor dwarf star-forming galaxies (XMPs), 4CentrodeAstrof´ısicadaUniversidadedoPorto,RuadasEstrelass/n, haveshownthattheregionswherestarformationoccurspos- 4150-762Porto,Portugal sessalowermetallicitythantheunderlyinggalaxy(Sańchez 5InstitutodeAstrof´ısicaeCieˆnciasdoEspaço,UniversidadedeLisboa, OAL,TapadadaAjuda,1349-018Lisboa,Portugal Almeida2013,2014a). FromtheH Idata,ithasbeenshown 6Leibniz-Institutfu¨rAstrophysikPotsdam(AIP),AnderSternwarte16, thatlocalBCDsandXMPsaregenerallysurroundedbypools D-14482Potsdam,Germany and streams of H I gas, over three times larger than the op- 2 M.E.Filhoetal. tical extension, which frequently display, in the peripheries, DigitalSkySurvey(SDSS) DataRelease (DR)7 (Abazajian distorted and lopsided H I distributions and velocities (Be- etal.2009)andliterature.For19XMPs,archiveHIimaging, gum&Chengalur2003;Youngetal. 2003;Thuan,Hibbard at resolutions higher than 1, are available via interferomet- ′ & Le´vrier 2004; Ekta, Chengalur & Pustilnik 2006; Begum ric Very Large Array (VLA) and/or Giant Metrewave Radio etal. 2006;Chengaluretal. 2006;Walter etal. 2008;Ekta, Telescope(GMRT)observations;theseareusedtoprobethe Chengalur& Pustilnik2008,2009; Ekta&Chengalur2010; nearbyenvironment.Thecosmicwebenvironmentisinferred Ottetal.2012;Lellietal. 2012a,b;Lelli;Verheijan&Frater- fromthe constrainednumericalsimulationsby Heß, Kitaura nali2014a,b; Beccarietal. 2014). Furthermore,significant &Gottlo¨ber(2013)andNuzaetal. (2014b),andthecatalog quantitiesofmetal-poor,large-scaleHIgaswerefoundinthe ofSDSSgalaxyfilamentsassembledbyTempeletal. (2014). subset of XMPs (Filho et al 2013; hereinafter Paper I). Be- WehavealsosearchedforclosecompanionstotheXMPsin cause the gas in galaxy disks is diluted in a relatively small theSDSSdatabase. timescale(oftheorderofonegalaxyrotation;e.g.,deAvillez Thepaperisorganizedasfollows. InSection2wepresent & Mac Low 2002; Yang & Krumholz 2012), this cumula- a summary of the optical and H I properties of XMPs with tiveevidencesuggeststhatthegashadtohavebeenaccreted archivalhigh resolutionH I observations,whichincludesan recently,mostlikelyasvestigialstreamsofcosmologicalac- analysis of the small-scale environment. Section 3 contains cretion (Sańchez Almeida 2014b), like the streams modeled our investigationof the large-scale environmentof XMPs in inhigh-massgalaxies(Dekeletal2009;Dekel,Sari&Cev- lightofthreediagnostictools: cosmicwebtype,overdensity erino 2009; Ceverino, Dekel& Bournaud2010; Ceverinoet and nearest neighbours. In Section 4 we discuss the results al. 2014). Therationaleisthatthemetallicityoftheaccreted andpresenttheconclusions. TheAppendixcontainsaquali- HIgasislow. Asthisgasisfedtothestarformationregions tativedescriptionofthearchivalhighresolutionHImorphol- ofthesource,theoverallmetallicityinthisregionislowered, ogyandvelocityfieldofindividualXMPs. dueto dilutionofthegas,while elsewherein thegalaxy,the HIgashasbeenenrichedwithmetalsfromstellarwindsand 2. SMALL-SCALEENVIRONMENT supernovaefrompreviousstarformationepisodes. Of the 140 local XMPs in the parent sample, 29 sources Thus, XMPs may constitute the ideal local laboratoriesto withnoH Iinformationhavebeenobservedwiththesingle- probetheroleofcosmologicalaccretionflowsingalaxyevo- dishEffelsbergradiotelescope,resultingin11newdetections lution and star formation. XMPs comprise over 0.1% of (PaperI).Overall,53XMPspossessHIdatainliterature;but the galaxies in the local volume (Morales-Luiset al. 2011), of these, only 19 have been observed at high spatial resolu- andrepresentthelow-metallicityendofthemostmetal-poor tion(.1),witheithertheGMRTortheVLA.We usethese galaxies,belowanarbitrarilydefinedmetallicityofone-tenth ′ archivalHIobservationstoinvestigatethespatialanddynam- solar, or 12 + log(O/H) 7.65. Although most XMPs are ≤ icaldistributionofthe H I gas, andits relationto theoptical a subsample of star-forming BCDs (Sargent& Searle 1970; component, which can provide clues as to the origin of the Thuan & Martin 1981; Papaderos et al. 1996a; Papaderos observedXMPproperties. etal. 1996b;Telles&Terlevich1997;Kunth&O¨stlin2000; Cairo´setal. 2001;Bergvall&O¨stlin2002;Cairo´setal.2003; 2.1. XMPsWithArchivalHighResolutionHIObservations Noeskeetal. 2003;GildePaz&Madore2005;Amor´ınetal. 2007,2009;Michevaetal. 2013),someXMPsaremoredif- Lelli, Verheijan & Fraternali (2014a, b) have presented a fuseandcorrespondtothecategoryofdwarfirregulargalax- detailed quantitative assessment of the H I properties (from ies(e.g.,Skillmanetal.2013).Theycontaincopiousamounts newandarchivaldata)of18starburstdwarfgalaxies,fourof ofmetal-poor,unsettled,H Igas(Begum&Chengalur2003; whichare also in ourXMP sample(UGC 4483,UGC 6456, Young et al. 2003; Thuan, Hibbard & Le´vrier 2004; Ekta, IZw18,andSBS1415+437). Inthispaper,weattemptonly Chengalur& Pustilnik 2006; Begumet al. 2006; Chengalur a qualitative description of the high resolution H I data in etal. 2006;Walter etal. 2008;Ekta, Chengalur& Pustilnik literature, in order to obtain global statistical properties for 2008, 2009; Ekta & Chengalur 2010; Ott et al. 2012; Lelli theXMPsample. etal. 2012a,b; PaperI;Lelli, Verheijan&Fraternali2014a, InTable 1 we presenta summaryof the opticalproperties b; Beccari et al. 2014). Moreover, in 80% of the cases, ofthesampleof19localXMPswitharchivalhighresolution XMPsshowevidenceforopticalasymme∼try(Papaderosetal. interferometric H I observations. The data is adapted from 2008;Morales-Luisetal.2011;PaperI).Thelopsidedoptical Table3ofPaperI,withsomerevisionanddescriptionofthe morphology has been commonly interpreted as asymmetric opticalmorphology. Table 2 providesa summaryof the rel- starformation,occurringinagalaxy/diskthatisinatransient evantH I properties,atseveralscales, ofthe 19localXMPs phase(Papaderosetal. 2008;Elmegreen2009;Elmegreen& witharchivalhighresolutioninterferometricHIobservations. Elmegreen2010),arareoccurrenceatlowredshift. The propertiesare a compilationof interferometricH I data Theobjectiveofthispaper(PaperII)is toanalyzethe en- fromliterature,namelyVLAandGMRTdata.Lowresolution vironment of the XMPs, with the aim of understanding the mapsdisplayresolutions&30′′,whileintermediateresolution role of cosmological accretion flows in the galaxy/disk as- display 30 – 15′′ and high resolution display . 15′′. In the semblyandtriggering/feedingofthestarformationinXMPs. Appendix, we present a detailed description of the H I high Archival high resolution interferometric H I data are used resolutiondataforeachXMP. to investigate the nearby environment, whereas cosmologi- IfwecomparetheXMPHImorphologyandvelocityfield calsimulationsconstrainedbyobservations,andtheobserved with the classification given in Lelli, Verheijan & Fraternali distribution of galaxies, provideinformationon the environ- (2014b),weconcludethatalmostalloftheXMPsfallwithin mentonlargescales. ThesourcesarethesameasinPaperI, theirclassificationofTypeA/Bsources–galaxieswithregu- and haven been taken from the parent sample of 140 local larlyrotatingH Idisks/galaxieswithkinematicallydisturbed XMPs,selectedbyMorales-Luisetal.(2011),fromtheSloan H I disks. That is, the XMPs generally show an H I disk- type morphology, with a velocity field typical of a rotating TheEnvironmentofExtremelyMetal-PoorGalaxies 3 TABLE1 Summaryoftheopticaldataforthe19extremelymetal-poorgalaxiesinthelocalUniversewitharchivalhigh resolutioninterferometricHIobservations. Source RA(J2000) DEC(J2000) mg dopt PA O/H Optical Note Name hms mag Morphology ◦ ′ ′′ ′′ ◦ (1) (2) (3) (4) (5) (6) (7) (8) (9) J0119-0935 011914 -093546 19.5 21.8 115 7.31 cometary ... UM133 014442 +045342 15.4 54.7 105 7.63 cometary 1 SBS0335-052W 033738 -050237 19.0 5.0 ... 7.11 symmetric ... SBS0335-052E 033744 -050240 16.3 8.3 45 7.31 cometary ... UGC4305 081905 +704312 ... 26.5 115 7.65 multi-knotdisk 2 HS0822+03542 082555 +353231 17.8 13.9 45 7.35 cometary ... DD053 083407 +661054 20.3 71.2 55 7.62 multi-knotdisk 3 UGC4483 083703 +694631 15.1 61.8 80 7.58 cometary 4 IZw18 093402 +551425 16.4 18.5 45 7.17 two-knotcometary 5 LeoA 095926 +304447 19.0 272.0 10 7.30 multi-knotdisk ... SextansB 100000 +051956 20.5 269.0 25 7.53 multi-knotdisk 6 SextansA 101100 -044134 ... 264.0 45 7.54 multi-knotdisk/symmetric 7 UGC6456 112800 +785939 ... 93.0 115 7.35 multi-knotdisk 8 SBS1129+576 113202 +572246 16.7 45.6 75 7.36 cometary ... UGCA292 123840 +324601 18.9 57.0 165 7.28 multi-knotdisk 9 GR8 125840 +141303 17.9 78.0 130 7.65 multi-knotdisk 10 SBS1415+437 141701 +433005 17.8 55.5 120 7.43 two-knotcometary 11 SagDig 192959 -174041 ... 116.0 170 7.44 multi-knotdisk/symmetric 12 J2104-0035 210455 -003522 17.9 13.1 110 7.05 cometary ... Columns: Col.1:Sourcename. Col.2:RightAscension(J2000). Col.3:Declination(J2000). Col.4:SDSSDR7g-band(Petrosian)magnitude,fromMorales-Luisetal.(2011). Col.5:Opticaldiameteralongthemajoraxis,obtainedfromtheSDSSDR7r-bandorDigitizedSkySurvey(DSS)-II R-banddata,measuredat25magarcsec 2. − Col. 6: Approximateopticalpositionangle,alongthelargestextension, measuredcounterclockwise fromtheWest, fromtheSDSSDR10orDSSimages. Col.7:Metallicity,12 + log(O/H),fromMorales-Luisetal.(2011). Col.8:Opticalmorphology,obtainedfromtheSDSSDR10orDSSimages.Symmetric(overallsphericallysymmetric structure),disk(overalldisk-likestructure),two-knot(structurewithtwoprominentstarformationknots),multi-knot (structurewithmultiplestarformationknots)andcometary(overallhead-tail structure, withatleastoneprominent starformationknot),areused. Col.9:NoteontheopticalstructurefromtheSDSSDR10orDSSimages. Noteincolumn9: 1–TheextendedcometarytailiscurvedtotheWest,ina”boomerang-type”structure. 2 – There is a faint nodule of emission to the Northwest, a bright ”loop-shape” structure to the East, and a faint filament,beginningneartheloop,andcurvingtotheSouthwest. Thebrighteststarformationknotsarefoundonthe Easternedgeofdisk,neartheloopandfilament. 3–Thereisseconddisk-likeemissiontotheWest,atapositionangleofapproximately55 ,similartothemaindisk. ◦ ThebrighteststarformationknotsarefoundontheNorthwesterntipofthemaindisk. 4–Althoughtheoverallshapeiscometary, thereappearstobeafainterX-shapepattern, withthebrighterandthe fainterarmatapositionangleofapproximately95and60 ,respectively. ◦ 5–Thetwobrighteststarformationknotsareeachattheendofthecometarystructure. 6–ThereisafainttailofemissiontotheWest,atapositionangleofapproximately160 . ◦ 7–ThebrighteststarformationknotsarefoundontheEasterntipofthedisk/symmetricstructure. 8–Thereisfaintextendedemissionalongapositionangleofapproximately75 . Thebrighteststarformationknots ◦ arefoundontheEasternedgeofthedisk. 9–Thediskhasa”boomerang”shape,withthebrighteststarformationknotsinthecenterandontheEasterntipof thedisk. 10–Thediskshowscurvatureatbothends,ina”boomerang-type”morphology. 11–Thetwobrighteststarformationknotsareattheheadofthecometarystructure. Theextendedcometarytailis curvedtotheWest. 12–ThebrighteststarformationknotsarefoundontheEasterntipofthedisk/symmetricstructure. disk,whichmaypresentsomesmallasymmetriesandirregu- or”cradling”brightstar-formingregions,orshowingsignsof laritiesintheoutskirts,andwhichmaybe(onlyslightly)of- ”clearing” around the star-formationknots (Appendix). Sag fcenterrelativetotheopticalpeakand/orstar-formingknots; Dig has been suggested to be the result of a merger process statistically,wedonotobservesignificantperturbationsinthe (Beccari et al. 2014 and Appendix). Its H I morphology large-scale H I gas componentof the XMP population(Ap- is ring-like, with an offcenter depression, and the velocity pendix). High spatial resolutionobservationsof the H I gas field displays no sign of rotation. These features are gener- suggest feeding of the star formation regions; the gas is re- allynotfoundinXMPs, whichweinterpretasevidencethat solved into complexsubstructure, with the gas”pointingat” the XMPs, as a class, are currently not undergoing a major 4 M.E.Filhoetal. TABLE2 SummaryoftheHIdataofthe19extremelymetal-poorgalaxiesinthelocalUniversewitharchivalhigh resolutioninterferometricHIobservations. Source PAopt HI/optRatio HI/optOffset PAHI PAVG Note Ref Name ◦ ′′ ◦ ◦ (1) (2) (3) (4) (5) (6) (7) (8) J0119-935 115 4 10/10 120–110 120 ... 1 UM133 105 2 0/10 105 105 ... 1 SBS0335-052Wa ... 6 10/10 150 NE-SW–N-S P 2 SBS0335-052Ea 45 4 10/10 145–20 NE-SW P 2 UGC4305 115 5 0/0 ... 100 ... 3 HS0822+03542 45 10 10/10 70-80 70 P 4 DD053 55 3 10/10 150 45 ... 3,5 UGC4483 80 2 10/0 80–60–80 90–120 ... 6,7,8,9 IZw18 45 7 10/0 45 50–55 P 8,9,10 LeoA 10 3 ... 10 ... ... 5 SextansB 25 3 10/10 25 130 ... 6 SextansA 45 2 10/0 135 150 ... 6 UGC6456 115 2 10/0 75 90 ... 8,9,11 SBS1129+576 75 2 0/0 75–70 70 P 12 UGCA292 165 3 0/0 175 145 ... 6,13 GR8 130 2 10/10 165 80 ... 5,6,13,14 SBS1415+437 120 2 10/0 115 115 ... 8,9 SagDig 170 6 0/0 ... ... M 5,15 J2104-0035 110 5 10/10 90 90 ... 16 aSBS 0335-052E and SBS 0335-052W are an interacting pair. Due to lack of resolution, the PAVG is merelyindicative. Columns: Col.1:Sourcename. Col. 2:Approximatepositionangleofthelargestopticalextension,measuredcounterclockwisefromtheWest, fromtheSDSSDR10orDSSimages. Col. 3:ApproximateHI-to-opticalratio,wheretheHIextensionismeasuredinthelowestresolutioninterfero- metricmap. Col.4:ApproximateoffsetbetweentheHIpeakandtheopticalcenter/brighteststar-formingknot. Col.5:ApproximatepositionangleofthelargestHIextension,measuredcounterclockwisefromtheWest(from thelowesttohighestresolutionmap). Col.6:ApproximatepositionangleoftheHIvelocitygradient,measuredcounterclockwisefromtheWest(from thelowesttohighestresolutionmap). Col.7:Noteonthepossibleinteraction/mergeroftheXMP.”P”designatesagalaxythatbelongstoaconfirmed interactingpair.”M”designatesagalaxythatmayhaveundergonearecentmerger. Col.8:Reference,withinterferometerdataand/orsurvey. Instrument,interferometerandsurveyacronym: GMRT–GiantMetrewaveRadioTelescope VLA–VeryLargeArray THINGS–TheHINearbyGalaxySurvey ACS–AdvancedCameraforSurveys ANGST–ACSNearbyGalaxySurveyTreasury Referenceincolumn8: 1–Ekta&Chengalur(2010);GMRT 2–Ekta,Pustilnik&Chengalur(2009);GMRT 3–Walteretal.(2008);VLATHINGS 4–Chengaluretal.(2006);GMRT 5–Begumetal.(2006);GMRT 6–Ottetal.(2012);VLAANGST 7–Lellietal.(2012b);VLA 8–Lelli,Verheijan&Fraternali(2014a);VLA 9–Lelli;Verheijan&Fraternali(2014b);VLA 10–Lellietal.(2012a);VLA 11–Thuan,Hibbard&Le´vrier(2004);VLA 12–Ekta,Chengalur&Pustilnik(2006);GMRT 13–Youngetal.(2003);VLA 14–Begum&Chengalur(2003);VLA 15–Beccarietal.(2014);VLA 16–Ekta,Chengalur&Pustilnik(2008);GMRT TheEnvironmentofExtremelyMetal-PoorGalaxies 5 interactionormerger. Rather,theasymmetriesandirregular- environment(comicweb type and overdensity)in whichthe itiesoftheouterenvelopeareunderstoodassignsofweaker XMPsreside. Spectroscopicredshiftsfromthe SDSS DR10 interactions caused by minor mergers, gas accretion or gas (Ahnetal. 2012)areavailablefor 70%ofthe140XMPs. ∼ outflowsdrivenbygalacticwinds. For the remaining XMPs, photometric redshifts from the SDSSDR10orNASA/IPACExtragalacticDatabase7 (NED) 3. LARGE-SCALEENVIRONMENT wereused. InordertoinvestigatetheenvironmentinwhichXMPsoc- In Figure 1, we have compared the cosmic web type en- cur,andtheinfluencethatcosmologicalaccretionflowsmay vironment distribution of XMPs with other galaxy morpho- haveontheirevolutionandobservedproperties,wehavecom- logicaltypes,namely,ellipticals(E),lenticulares(S0),spirals puted the likelihood of an XMP to inhabit a certain type of (Sp) and irregular galaxies (Irrs), as assigned by the 2MRS environmentusingthreediagnostics: cosmicwebtype,over- team (Huchra et al. 2012). Although the Poisson errorbars densityandnearestneighbours. are largerfor the XMPsample, it is apparentthat the XMPs Theenvironmentcan beclassified accordingtolocalindi- follow similar trends as Irr galaxies, but with a more pro- cators, such as the local density, or according to non-local nounced behaviour (Fig. 1). The XMP prevalence in voids indicators,based,forexample,onthetidal-fieldtensor. Inthe andsheets,andtheavoidanceofknots,suggeststhat,statisti- firstcase,onecanestimatethelocaloverdensitybasedsolely callyandonalargescale( 1Mpc),theXMPsarerelatively ∼ on observations, defined as the number density of galaxies isolated(Fig.1). relativetothebackgroundnumberdensity,orbycharacteriz- Asaconsistencycheck,wehavealsoestimatedthecosmic ingthenearestneighbours(type,mass,numberanddistance). webtypedistributionforalargesampleofcontrolgalaxies( ∼ However, a classification of the environmentinto its cosmic 700BCDs,withZ>0.1Z )fromtheSDSSDR8(Aiharaet webcomponentsrequiresknowledgeofthelarge-scaleenvi- al. 2011). Wefindasimila⊙rbehaviouramongtheBCDsand ronment. Thesignoftheeigenvaluesofthetidal-fieldtensor XMPsample, which is markedlydifferentfrom the environ- indicate pure contracting regions, such as clusters or knots, mentsfoundforE,S0andSpgalaxies,andmoreextremethan or expanding regions in three, two, or one dimensions, rep- the behaviour of the Irr galaxies in the zero redshift 2MRS resentingvoids,sheets,orfilaments,respectively(Hahnetal. sample(tobeincludedinafuturepaper). 2007).Whilenumericalsimulationsprovideallthenecessary Figure2showstheprojectedspatialdistributionofthe140 information to perform a cosmic web analysis, observations local XMPs, color-coded for cosmic web type. The sizes requireapriorreconstructionofthelarge-scalematterdensity of the circles trace the overdensity (δ), such that, the larger field. Theparticularreconstructiontechniquewilldetermine the circle, the largerthe overdensity. The valuesrangefrom theaccuracyofthecosmicwebclassification. the most underdense (δ = 0.9, for PHL 293B) to the most overdense(δ=376,forVC−C0428,aBCDintheVirgoclus- 3.1. CosmicWebTypeandOverdensity ter; Meyeretal. 2014)sourceenvironment. Figure 3shows WehererelyonpreciseconstrainedN-bodysimulationsof the overdensity distribution (log(1+δ)), color-coded by cos- the Local Universe, using the Two-Micron All-Sky Galaxy micwebtype,fortheXMPsample. Arrows,color-codedby RedshiftSurvey(2MRS;Huchraetal. 2012),withinvolumes cosmicwebtype,showthemeanoverdensity(log(1+<δ>)) of 180h 3 Mpc3 (Heß, Kitaura & Gottlo¨ber 2013). Cosmo- for each distribution. Accordingto the overdensitydistribu- − logicalparametersonlyprovideastatisticaldescriptionofthe tion, 60% of the XMPs are found in underdense regions Universe, so that an infinite number of realizations are con- (δ .∼0) of the reconstruction (Fig. 2 and 3). On the other sistent with the same set of parameters. The procedurepro- hand, most of the XMP galaxies ( 75%) reside in the less ∼ videsacosmologicalsimulation,whichreproduces,indetail, dense cosmic web type environments ( 40% are in sheets ∼ thespatialdistributionofgalaxiesinthelocaluniverseasde- and 34%invoids),while 25%canbefoundinfilaments, scribedby2MRS.Themethodtoreconstructtheinitialfluctu- and.∼1%inknots(Fig.1,2∼and3). ationsusedin thesesimulationsisbasedonaself-consistent Intheanalysisbelow,becauseweareconstrainedbysmall Bayesian machine-learning algorithm, reaching an accuracy numberstatisticsandsystematicerrorsthatmayaffectthede- of 2Mpch 1 (Kitauraetal. 2012;Kitaura2013;Heß,Ki- terminationof the physicalparameters, the trends should be − ∼ taura&Gottlo¨ber2013). Thecosmicwebclassificationused takenwithcaution. inthisstudyhasbeenperformedfollowingForero-Romeroet WehaveinvestigatedwhethertheH Imass(M ),dynam- HI al. (2009), based on the best correlated constrained N-body ical mass (M ) or stellar mass (M ) of the XMPs depends Dyn simulation with the 2MRS galaxyfield (Nuza et al. 2014b). on the environment in which they r∗eside (Fig. 4). The H I The overdensity (δ) used in this study has been computed andstellarmassesweretakenfromTable5ofPaperI.Inthis from the same constrained simulation. The preference of a Paper II, we have estimated the dynamical masses with re- type of galaxy to be located in a particular environment is vised optical radius measurements. There is a large overlap quantifiedbyη(τ, ǫ), anexcessprobabilityratio, whereτ is in dynamical, H I and stellar mass for a range of overdensi- the galaxy type and ǫ is the cosmic web environment type ties (Fig. 4). However, XMPs in filaments and sheets show (voids, sheets, filamentsandknots; Nuza etal. 2014b). The thelargestrangeinH Imass,particularlyextendingtolower environment(overdensityorcosmicwebtype)ofasourcecan H I masses, while XMPs in voidsshow tendentiouslylarger beassignedaccordingtothecellinthereconstructionwhere H I masses, abovelog(MHI/M ) 8(Fig.4a). Fivesources thesourceislocated. Eachcellis characterizedbythreeco- in voids and sheets show the l⊙arg≃est dynamical masses (log ordinates: rightascension,declination,andredshift,withthe (M /M )&9;Fig.4b).Intermsofstellarmass,mostXMPs Dyn redshiftparameterizingthedistance,includingthepropermo- in filamen⊙ts foundin overdenseregions(δ & 0) show stellar tions caused by local tidal fields, taken consistently into ac- masses below log (M /M ) 7, with a tail of XMPs in less count. dense filamentary env∗iron⊙me≃nts towards high stellar masses BecausewepossesscoordinatesforourXMPgalaxies,we haveusedtheconstrainedsimulationtodeterminethetypeof 7http://ned.ipac.caltech.edu/ 6 M.E.Filhoetal. Fig.1.—TheexcessprobabilitydistributionofXMPs,andE,S0,SpandIrrgalaxiesofthezeroredshift2MRSsample(Huchraetal.2012;Nuzaetal.2014b). Thefourpanelscorrespondtosourcesfoundinvoids,sheets,filamentsandknots. Fig.2.—Theprojectedspatialdistributionofthe140localXMPs,color- Fig.3.—Theoverdensitydistributionforthe140localXMPsample,color- codedbycosmicwebtype(blueforvoids,redforsheets,greenforfilaments codedbycosmicwebtype(blueforvoids,redforsheets,greenforfilaments andcyanforknots). Thesizesofthecirclestracetheoverdensity,suchthat, andcyanforknots).Arrows,color-codedbycosmicwebtype,showthemean thelargerthecircle, thelargertheoverdensity. Thevalues rangefromthe overdensity(log(1+<δ>))foreachdistribution. mostunderdense(δ= 0.9,forPHL293B)tothemostoverdense(δ=376, forVCC0428,aBCD−intheVirgocluster)sourceenvironment. (Fig.5a), SFR (Fig. 5b)and gasconsumptionrate (Fig.5d). However, XMPs in voids show a slight tendency for larger (log (M /M ) & 8; Fig. 4c). The largest H I-to-stellar mass valuesofsSFR(logsSFR& 8yr 1;Fig.5c). Cometaryand − ratiobel∗ong⊙stoanXMPresidinginavoid(Fig.4d).Overall, multi-knotXMPsareroughly−equally(fractionally)prevalent we see no clear relation between the environment in which invoids,sheetsandfilaments(Fig.5eand5f),whiletwo-knot XMPsresideandtheirdynamical,stellarorHImass. sources are roughly equally (fractionally) prevalent in voids We have further investigated if there is an environment andsheets(Fig.5e and5f). Symmetricsourcesarefoundto trend with metallicity (12 + log(O/H)), star formation rate bemorecommoninvoids(Fig.5eand5f). Thetwosources (SFR),specificstarformationrate(sSFR SFR/M ),HIgas with morphologicalinformation,andcontainedin knots, are consumptionrate (1/t SFR/M ), an≡d optical∗morphol- bothcometary(Fig.5dand5e). MHI ≡ HI ogy(symmetric,cometary,two-knotandmulti-knot;Fig.5). In summary, we find that XMPs tend to reside in low- These values were taken from Table 3 and 4 of Paper I. We densityenvironments,andincosmicwebtypescharacterized findnoclearcorrelationbetweenenvironmentandmetallicity asvoidsandsheets. WefurtherfindthatXMPs,BCDsandIrr TheEnvironmentofExtremelyMetal-PoorGalaxies 7 (a) (b) (c) (d) Fig.4.—Theoverdensityasafunctionof(a)HImass,(b)dynamicalmass,(c)stellarmass,and(d)HI-to-stellarmassratio,color-codedbycosmicwebtype (blueforvoids,redforsheets,greenforfilamentsandcyanforknots). TheHIandstellarmassesweretakenfromTable5ofPaperI.InthisPaperII,wehave estimatedthedynamicalmasseswithrevisedopticalradiusmeasurements. galaxiestendtoinhabitthesametypeofenvironment,which Hence,thedistanceofthenearestgalaxyandmorphological isinstarkcontrastwiththebehaviourofE,S0,Spgalaxiesin filamentcanbeusedasaproxyforthedistanceoftheXMPs thezeroredshift2MRSsample. to morphologicalfilaments. In order to identify our sample XMPsintheA3,wesearchedtheA3forgalaxieswithinD ∼ 3.2. ProximitytoSDSSMorphologicalFilaments 10kpcoftheXMPpositions;nonewerefound.Weare,therefore,confidentthatthenearestgalaxyidentifiedforeachXMP Wehaveinvestigatedtheproximity(orinclusion)ofXMPs is truly a separate galaxy. We stress that A3 contains 500 tomorphologicalfilaments,i.e.,”chains”ofgalaxies,asem- ∼ 000 SDSS DR8 sources in a contiguous field, which could piricallydefinedusinggalaxiesinthespectroscopicsampleof miss the XMPs and some of their possible neighbours (see the SDSS DR8 dataset. Tempelet al. (2014)use a marked- below). pointprocesswithinteractions,calledtheBisousmodel(Sto- Figure 6 contains the XMP distance to the nearest galaxy ica,Gregori&Mateu2005),todetectgalaxyfilamentsinthe (D ) in A3, as a function of the distance between the spectroscopicsampleoftheSDSSDR8dataset(Tempel,Tago galaxy nearestgalaxyin A3andthe nearestmorphologicalfilament &Liivama¨gi2012). Aredshiftintervalof0.009 z 0.155 (D ),color-codedbycosmicwebtypeoftheXMP.The was used so that the lower redshift limit exclude≤s th≤e Local filament figure also contains solid black lines, indicating the radii Supercluster. The authors look for morphological filaments of the morphological filaments, so that XMPs in galaxy fil- witha radiusof0.5h−1 Mpc. Theyfindthattheshortestfil- aments should display D . 0.5 Mpc and D . amentsintheirfilamentcatalogare3–5h 1 Mpcinlength, galaxy filament − 0.5 Mpc. Only three of the XMPs are found within the whiletypicalfilamentlengthsareof60h−1Mpc. D . 0.5 Mpc and D . 0.5 Mpc region, and only We have mined the catalog of galaxies (hereinafter A3), galaxy filament marginally(Fig.6). Therefore,wefindthatXMPsaregener- used to generate the morphological filaments, to search for allynotcontainedwithinthemorphologicalfilamentsdefined XMP identifications, or, alternatively, to find the nearest byTempeletal. (2014). galaxy to each XMP. Distances between the XMPs and A3 galaxies were estimated neglecting galaxy proper motions, 3.3. NearestGalaxiesandPotentialPerturbers sincethelocalflowforpairsofnearbygalaxiesshouldbeof thesameorder. A3alsoincludesthedistanceofeachgalaxy AccordingtothegalaxiesintheA3catalog,onlytwoXMPs to the nearest morphologicalfilament (Tempel et al. 2014). have a galaxy within D . 100 kpc, with one of these galaxy 8 M.E.Filhoetal. (a) (b) (c) (d) (e) (f) Fig.5.—Theoverdensity asafunction of(a)metallicity, (b)SFR,(c)sSFR,(d)H Igasconsumption rate(SFR/MHI), and(eandf)optical morphology (symmetric,cometary,two-knotandmulti-knot),color-codedbycosmicwebtype(blueforvoids,redforsheets,greenforfilamentsandcyanforknots).These parametersweretakenfromTable3and4ofPaperI. XMPs residing in a filament and another in a void (Fig. 6). (D &10Mpc)distances(Fig.6). galaxy Indeed,therearefourdocumentedXMPsthatareknowntobe In order to complementthe results based on A3, we have interactingwithnearbycompanions(IZw18,SBS0335-052, also looked for nearest galaxy neighbours, as potential per- HS0822+03542andSBS1129+576;Sect.2andAppendix). turbers,usingthe fullSDSSDR9 (Ahnetal. 2012)catalog. Overall, the nearest galaxies in A3 show a large range in The search was performed within a 12 radius around each ′ distances to the XMPs (D 0.1 – 100 Mpc), which XMP,correspondingtoadistanceofD 200kpcatz=0.01. galaxy ∼ ∼ does not seem to depend strongly on the XMP environment Wefurtherconstrainedthegalaxiestohavespectroscopicred- (Fig.6). However,thereisatendencyforXMPsinfilaments shiftsof z < 0.1, resultingin 1 000galaxies, potentialper- ∼ to show a somewhat bimodalbehaviour: there is a group of turbersto104XMPs. Wehavealsoestimatedthestellarmass (nearest) galaxies at smaller (D . 1 Mpc) and larger ofthepotentialperturbersfromtheirSDSSDR9g rcolours, galaxy − TheEnvironmentofExtremelyMetal-PoorGalaxies 9 Fig.6.—ThedistancebetweentheXMPandthenearestgalaxyinA3,as Fig.7.—ThedistancetothenearestpotentialperturberintheSDSSDR9 afunctionofthedistancebetweenthenearestgalaxyinA3andthenearest (withinD.300MpcoftheXMPs),asafunctionofstellarmass,color-coded morphological filament, color-codedbycosmicwebtypeoftheXMP.The bythecosmicwebtypeoftheXMP(blueforvoids, redforsheets, green solid black lines indicate the radii of the morphological filaments, sothat forfilamentsandcyanforknots). Thestellarmassofthenearestpotential 0X.M5MPspicn.gTalhaexyblfiaclakmdeansthsesdholiunledidnidspiclaatyesDagarlaexpyre.se0n.t5atMivepcdaisntdanDcefilafmoerntth.e pmearstus-rtboe-rlighhastrbaetieonsefrsotimmaBteedllf&rodmeJthoengS(D2S0S01D).R9g−r colours, usingthe nearestgalaxy,inordertoexertsomegravitationalinfluenceovertheXMP (seeSect.3.3and3.4). usingthemass-to-lightratiosfromBell&deJong(2001). gerand/orfeedthe star formation. As a first approximation, In Figure 7 we plot the stellar mass (M ) and distance wehaveanalyzedtheconditionsforperturbationinasystem (D ) of the nearest potential perturbe∗r within D . perturber of a satellite galaxy(XMP) on a circular orbitarounda per- 300 Mpc of the XMPs, color-coded by cosmic web type of turber.Letusconsiderthatthesatellitegalaxyhasmassmand the XMP.Only oneXMPshowsa potentialperturberwithin radiusr,orbitingaperturbergalaxyofmass M,atadistance D . 100 kpc, with a stellar mass of M 107 M perturber ofD,andwherem<< M. (Fig.7). Wecannot,however,excludethepresen∗ce∼ofpoten⊙- Minormergers/interactions(whenm<< M)occurthrough tialperturberswithnomeasuredspectroscopicredshiftinthe dynamical friction, when the perturber and satellite share a SDSS, norless luminous(andless massive)than the XMPs, common dark matter halo. This dynamical friction causes dueto the magnitudecuttoffofthe SDSS (magnitudeinr & orbital decay of the satellite galaxy within the so-called dy- 17.7mag).ComparingFig.6andFig.7,weconcludethatthe namicalfrictiontimescale(e.g.,Pattonetal. 2000;Binney& potentialbiases in the A3 sample do not modifythe conclu- Tremaine2008): sionthatXMPsgenerallydonothaveclosecompaniongalax- ies. AboveaD 1MpcdistancetotheXMP,thestel- larmassesofthpeertnurebaerre∼stpotentialperturbersshowachange; T 2.64 105Gyr D2v , (1) the stellar masses not onlyspan a larger range, but potential fric ≃ · × mln Λ perturberswithstellarmassesofuptoM 1011 M canbe where D is the distance between the satellite galaxyand the found(Fig.7). Thenearest,leastmassive∗,∼andmost⊙massive perturber (in kpc), v is the circular speed of the satellite (in potential perturbers belong to XMPs in all three major cos- kms 1),misthemassofthesatellite(inM ),andlnΛisthe − micwebtypes(voids,sheetsandfilaments). Wefurthernote Coulomblogarithm. Thisexpressiondoesn⊙ottakeintocon- abreakinperturbermassdistributionat Dperturber 10Mpc; siderationanymassloss(fromtidalstripping)thatthesatellite ∼ above this distance, stellar masses are skewed towards the maysufferduringtheorbitaldecay. largerrange(M &109 M ),consistentwiththedetection,at If we assume a typical distance to the perturber of D largerdistances,∗ofmorel⊙uminous,andhence,moremassive 100 kpc (Fig. 6 and 7), a satellite mass of m 108 M≃ galaxies(Fig.7). (Fig.4c),aCoulomblogarithmof2(e.g.,Binney&≃Tremain⊙e Overall, with the exception of the already documented 2008;Filhoetal. 2014),anda circularvelocityofthesatel- casesofXMPswithcompanions(Sect.2andAppendix),sta- lite of v √GM/D, where G is the gravitational constant, tistically we do not find evidence for the presence of poten- ≃ this then yields a dynamical friction timescale of T fric tial perturbers (with measured spectroscopic redshifts in the ≃ 50 Gyr, for M m, or 150 Gyr, for M 10m. This esti- SDSS) within D . 100kpcoftheXMPs. We caution,how- ≃ ≃ mateprovidesatimescaleforthesatellitetospiralintowards ever,thatduetoalackofspectroscopicredshiftandthespec- the perturber. The resulting timescales are too long for the troscopicfluxlimitoftheSDSS,wemaybemissingpotential merger/interactionprocesstobesignificantinXMPs. perturbersofverylowmassandluminosity. Asthesatelliteapproachestheperturberonthisorbitalde- cay,thetidalforcesatworkonthesatellitebecomemoreim- 3.4. PerturbationsandInteractions portant. If we now consider a test particle at a distance ∆r Becausethenearestgalaxiesto XMPsaregenerallyfound from the surface of the satellite (where ∆r << r), then the to be at distances D & 100 kpc (Fig. 6 and 7), we have at- perturberwillexertaperturbativetidalforceonthetestparti- tempted to evaluate the gravitational influence these could cle(relativetotheforceexertedatthecenterofthesatellite), haveontheXMPs,particularlyiftheycouldpotentiallytrig- thatisgivenby(e.g.,Binney&Tremaine2008): 10 M.E.Filhoetal. webof E, S0 andSp galaxies, butinhabitingregionssimilar tothoseofIrrsandBCDs(Sect.3.1). We find results similar to those found for nearby (relatively) luminous galaxies (Blanton & Moustakas 2009) and Hα-emittinggalaxies(Darvishetal. 2014),wherebytheen- vironmentdoesnotdeterminetheoverallobservedproperties (mass,luminosity,SFR,etc.) ofacertaingalaxytype. How- ever,thefractionofacertaingalaxytype(inthiscase,XMPs), doesappeartobedeterminedbytheenvironmentinwhichit resides(Sect.3.1). The evidence for occupation of low-density environments is furtherstrengthenedwhen we investigatethe proximityto morphologicalfilaments(Sect.3.2),andgalaxiesinthevicin- ityofthe XMPs(Sect.3.3). We findnosignificantevidence forthepresenceofgalaxyfilamentswithinD 0.5Mpcofan ∼ XMP,andonlyonepotentialperturber(ofmassandluminos- Fig.8.—Thedisplacement producedbytidalforces, asafunctionofthe ityatleastequaltothatofanXMP,andwithameasuredspec- ratiobetweentheXMPradiusandthedistancetotheperturber,color-coded troscopicredshiftintheSDSS)isfoundwithin D 100kpc bytherelativeXMP/perturbermass: M/m=0.1(cyan),1(blue),10(red), ofanXMP.However,evenatthesedistances,wehav∼edemon- 100(green). Theblacklinecorrespondsto∆r/r=1,thelowerlimitcriteria stratedthatapotentialperturberisveryunlikelytocausegrav- fortidalstripping. itationaldisturbancecapableof drawingfreshgasor driving thestarformation(Sect.3.4). We have also analyzedthe small-scaleenvironment,prob- ing for signatures of galaxy merger/interactions in the pub- 2GM(r+∆r) 2GMr lished H I data of XMPs (Sect. 2 and Appendix). IZw 18 F . (2) M ≃ D3 ≃ D3 (Lelli et al. 2012a; Lelli, Verheijan & Fraternali 2014a, b), SBS 0335-052WandSBS 0335-052E(Ekta, Pustilnik& Thesatellitewillalsoexertarestoringforceonthetestparti- Chengalur 2009), HS 0822+03542 (Chengalur et al. 2006) cle, if tidalforcesdisplaceit fromits equilibriumconfigura- andSBS1129+576(Ekta,Chengalur&Pustilnik2006)have tionatr,namely,(e.g.,Binney&Tremaine2008): beenconfirmedtobeinteractingwithcompaniongalaxies,all within D .25kpc(Sect.2andAppendix). Moreover,ithas Gm∆r Fm = r3 , (3) been suggested that the H I properties observed in Sag Dig couldbetheresultofarecentmergerbetweendwarfgalaxies wherewehaveneglectedthepressuresupportinthesatellite (Beccariet al. 2014). However,if Sag Dig is the resultof a (Adamsetal. 2011). InequilibriumF F ;therefore,the merger(Beccarietal. 2014;Sect.2andAppendix),thenthe M m displacementproducedbytidalforcesisg≃ivenby: mergersignaturespresentin Sag Dig (ring-likeH I distribution versus disk-like optical morphology,clumpy H I distri- ∆r M r 3 bution with no optical correspondence, large offset between 2 . (4) r ≃ m (cid:18)D(cid:19) the optical center and the depression in the H I distribution, no clear rotationalmotion) are generally absent in the XMP The condition for tidal disruption of the test particle by the population.Indeed,wefindnostatisticalevidencefor(major) perturberisproducingalargeperturbation,i.e.,∆r/r&1. interactionswith (gas-rich)galaxy companionsin the recent In Figure 8 we show the displacement produced by tidal past (Sect. 2 and Appendix; see also Brosch, Almoznino & forces (Eq. 4), as a function of the ratio between the XMP Heller2004,Holwerdaetal. 2013andKreckeletal. 2015),a radius and the distance to the perturber, color-coded by the mechanismwhichistraditionallyinvokedasadriverforstar relativeXMP/perturbermass. AscanbeseenfromFigure8, formationandoverallgalaxyevolution(e.g.,Keeletal. 1985; inorderfortidalstrippingtooccur,theperturberisrequiredto Robainaetal. 2009).Althoughsuchaninteractioncouldpro- haveahighmassand/orbelocatedatasmalldistancetothe videlargequantitiesoffreshHIgas,andcreateconditionsfor satellite. For a perturberand satellite of similar mass (M triggering/feedingtheobservedstarformation,itcannotex- ≃ m),thedistancemustbeoftheorderofthesizeofthegalaxy plainwhythegasweobserveismetal-poor(PaperI),norcan (D r). Because XMPs have radii of a few kpc, D must it explain the metallicity inhomogeneitiesobserved in some ≃ also be a few kpc in order to exert appreciable gravitational XMPs(SańchezAlmeidaetal. 2013,2014a),anditdoesnot influenceontheXMP. explain why we do not systematically observe signatures of Therefore, although a significant number of XMPs show major disturbances/interactions in the H I gas, and particu- potential perturberswithin D 300 Mpc, it is unlikely that ∼ larlyintheH I-to-opticalmorphological/kinematicalrelation thesewillcausemajorgravitationaldisturbance(ortidalstrip- (Sect. 2 and Appendix). In fact, the central regions of the ping)oftheXMP,whichcouldinstigategastransferandsus- HIstructuresarerelativelywell-behaved;inmanycasesthey tainstarformation. areorganizedinadiskwithclearrotationalmotion,inwhich thekinematicalHIaxisisvirtuallyaligned(within15 )with 4. DISCUSSIONANDCONCLUSIONS ◦ theH I andopticalaxis, wherethe peakofthe H I emission XMPsarefoundtotendentiouslyresideinlow-densityen- ispositionallycoincident(within10 ) with thecenterofthe ′′ vironments ( 60%) and in cosmic web types characterized opticaldisk or the brighteststar formationknot(Appendix), asvoidsand∼sheets ( 75%; Sect. 3.1). Inthisrespect, their and where the velocity offsets between the stellar and H I behaviourisextreme,∼incontrasttothelocationinthecosmic components are absent or small (< 40 km s 1; of the order −