DRAFTVERSIONMAY25,2016 PreprinttypesetusingLATEXstyleemulateapjv.5/2/11 MOLECULARHYDROGENABSORPTIONFROMTHEHALOOFAZ 0.4GALAXY ∼ SOWGATMUZAHID1,GLENNG.KACPRZAK2,JANEC.CHARLTON1,ANDCHRISTOPHERW.CHURCHILL3 1ThePennsylvaniaStateUniversity,StateCollege,PA16802,USA;([email protected]) 2SwinburneUniversityofTechnology,Victoria3122,Australia 3NewMexicoStateUniversity,LasCruces,NM88003,USA 6 1 DraftversionMay25,2016 0 2 ABSTRACT y Lyman-andWerner-bandabsorptionofmolecularhydrogen(H2)isdetectedin 50%oflow-redshift(z <1) a DLAs/sub-DLAs with N(H ) > 1014.4 cm−2. However, the true origin(s) o∼f the H -bearing gas remain 2 2 M elusive.HerewereportanewdetectionofanH absorberatz =0.4298intheHST/COSspectraofquasar 2 abs PKS2128–123.ThetotalN(HI)of1019.50±0.15cm−2classifiestheabsorberasasub-DLA.H2absorptionis 3 detecteduptotheJ = 3rotationallevelwithatotallogN(H ) =16.36 0.08correspondingtoamolecular 2 2 fractionoflogf = 2.84 0.17. TheexcitationtemperatureofT =±206 6Kindicatesthepresenceof coldgas. UsingHd2etail−edioni±zationmodelingweobtainanear-solarmexetallicity±(i.e.,[O/H] = 0.26 0.19) ] andadust-to-gasratiooflogκ 0.45fortheH -absorbinggas. Thehostgalaxyofthesub-D−LAisd±etected A atanimpactparameterofρ 4∼8−kpcwithanincl2inationangleofi 48◦ andanazimuthalangleofΦ 15◦ G withrespecttotheQSOsight∼line.Weshowthatco-rotatinggasinane∼xtendeddiskcannotexplaintheobs∼erved . kinematicsofMgII absorption. Moreover,theinferredhighmetallicityisnotconsistentwiththescenarioof h gasaccretion. Anoutflowfromthecentralregionofthehostgalaxy,ontheotherhand,wouldrequirealarge p opening angle (i.e., 2θ >150◦), much larger than the observed outflow opening angles in Seyfert galaxies, - o in order to intercept the QSO sightline. We thus favor a scenario in which the H2-bearing gas is stemming r fromadwarf-satellitegalaxy,presumablyviatidaland/orrampressurestripping.Detectionofadwarfgalaxy st candidateintheHST/WFPC2imageatanimpactparameterof 12kpcreinforcessuchanidea. ∼ a Subjectheadings:galaxies:haloes–galaxies:ISM–quasars:absorptionlines–quasar:individual(PKS2128– [ 123) 3 v 2 1. INTRODUCTION Crightonetal.2013;Oliveiraetal.2014;Srianandetal.2014; 8 Molecularhydrogen(H )isthemostabundantmoleculein Duttaetal.2015;Muzahidetal.2015b). 7 theuniverse,anditplaysa2crucialroleinstarformationinthe UsingarchivalHST/COSspectra,Muzahidetal.(2015b) 6 interstellar medium (ISM; see Shull & Beckwith 1982, for have surveyed H2 absorption in 27 low-z DLAs/sub-DLAs 0 for the first time and have reported detections of 10 H ab- a review). H exhibitsnumeroustransitionsfrom its ground 2 1. electronicsta2tetotheLymanandWernerbandsatultraviolet sorbersintotal. TheH2 detectionrateatlow-z,i.e. 50+−2152%, 0 (UV) wavelengths (900–1130 A˚). The Lyman- and Werner- for systems with N(H2) > 1014.4 cm−2, is a factor of & 2 6 band absorption of H is detected in diverse astronomical higherthanthatfoundathigh-zbyNoterdaemeetal.(2008). 2 1 environments, e.g., in the Galactic disk (Spitzer & Jenkins The increase of the cosmic mean metallicity of DLAs/sub- : 1975;Savageetal.1977),Galactichalo(Gillmonetal.2006; DLAs (e.g., Rafelski et al. 2012; Som et al. 2013) and the v i Wakker 2006), Magellanic Clouds (Tumlinson et al. 2002; dimmingoftheambientradiationfieldduetothedecreaseof X Welty et al. 2012), Magellanic Stream (MS; Sembach et al. thecosmicstarformationratedensitywithcosmictime(e.g., r 2001),MagellanicBridge(MB;Lehner2002),high-velocity Bouwensetal.2011;Haardt&Madau2012)arethoughttobe a clouds (HVCs; Richter et al. 1999, 2001), intermediate- responsibleforsuchanenhancedH2detectionrateatlow-z. velocity clouds (IVCs; Gringel et al. 2000; Richter et al. The median N(HI) value of 1019.5 cm−2 for the low- 2003), and in high-redshift damped Lyα absorbers (DLAs; z sample of Muzahid et al. (2015b) is an order of mag- Ledouxetal.2003;Noterdaemeetal.2008). nitude lower than that of the high-z sample. Neverthe- Forhigh-redshift(z >1.8)absorbers,theUVtransitionsof less, the low-z H2 absorbers show molecular fractions, H2 get redshifted to optical wavelengths. Absorption lines fH2 = 2N(H2)/[N(HI)+2N(H2)], thatarecomparableto of high-z H2 absorbers, thus, are well studied using high- the high-z H2 absorbers (e.g., median logfH2 = −1.93± resolution optical spectra obtained from ground-based tele- 0.63). High molecular fractions at lower N(HI) values in- scopes (e.g., Srianand et al. 2005, 2008, 2012). Due to the dicate that (a) the density is extremely high and/or (b) the lack of high-resolution, UV-sensitive, space-based spectro- prevailingradiationfieldisveryweakintheabsorbinggas. graph, H was never studied in absorptionat low-z (z < 1) Usingsimplephotoionizationmodelsandfromthelackof 2 beyondthe LocalGroupuntil the last decade. However, the highJ (i.e.,J > 3)excitations,Muzahidetal.(2015b)have high throughputof the Cosmic Origins Spectrograph(COS) inferredthattheradiationfieldismuchweakerthanthatseen onboardtheHubbleSpaceTelescope(HST)hasrecentlyen- intheMilkyWaydiffuseISM(Blacketal.1987),incontrast abled such observation. Detection and analysis of low-z H2 tohigh-zH2 absorbers(e.g.,Srianandetal.2005). Theden- absorbershave been presented in severalrecentstudies (i.e., sityoftheH2-absorbinggasispredictedtobeintherangeof 10–100cm−3 (orlower),whichisconsistentwiththatofthe 2 S.Muzahidetal. Galactic diffuse ISM. The median rotational excitation tem- √3. All ourmeasurementsand analyseswere performedon peratureofT =133 55K1isalsosuggestiveofadiffuse thebinneddata. 01 ± atomicgas-phase(Snow&McCall2006). InadditiontotheFUVobservations,near-UV(NUV)spec- Detectinghostgalaxiesofhigh-z H absorbersisdifficult, traldataarealsoavailablefortheQSO.TheseNUVobserva- 2 andsearcheshavenotbeensuccessfulso far. Theadvantage tionsconsistofG185MandG225Mgratingintegrations.The of studying low-z H absorbers is that it is relatively easier G185MgratingdatawereobtainedusingHST/COSCycle- 2 to identify the host galaxies responsible for absorption. For 19 observations under program ID GO-12536, whereas the example,8ofthe10H absorbersinthesampleofMuzahid G225Mgratingdata were obtainedby our ownprogram,ID 2 etal.(2015b)havehostgalaxiesidentifiedfromtheliterature. GO-13398.IndividualG185MandG225Mintegrationswere Interestingly,5ofthe8H absorbershavehostgalaxiesatan co-addedusingcustom-writtenIDLcodefollowingsimilaral- 2 impactparameterofρ > 20kpc. Moreover,thetwosystems gorithms to the “coadd x1d”. Both G185M and G225 data with the highestmolecularfractionsshow the largestimpact consist of three 35 A˚ stripes separated by two 65 A˚ gaps. parameters (i.e. ρ > 50 kpc; see Fig. 9 of Muzahid et al. G185M data cover 1670–1705A˚, 1770–1805A˚, and 1870– 2015b). Evenforthelowestimpactparameter(ρ < 10)sys- 1905 A˚, whereas G225M data cover 2100–2135 A˚, 2200– tem the QSO sightline is well beyondthe stellar disk of the 2235 A˚, and 2300–2335A˚. The co-added G185M spectrum hostgalaxy(see,e.g.,Petitjeanetal.1996;Chenetal.2005). has an S/N of 5–13 per resolution element. However, the Itis,therefore,conjecturedthatthelow-zH2absorbersarenot G225MspectrumhasanS/N<3perresolutionelement. relatedtostar-formingdisks,butemergefromtidallystripped orejecteddiskmaterialinthehalo. 2.1.2. Keck/HIRES HerewepresentadetailedanalysisofanewH absorberin 2 asub-DLAatz =0.4298detectedtowardQSOPKS2128– TheopticalspectrumofPKS2128–123wasobtainedwith abs 123. The host galaxy of the H absorber is detected at theHIRES mountedonthe Keck-I10m telecopeonMauna 2 ρ 48kpc. Themainfocusofthisworkistounderstandthe Kea in Hawaii. The QSO was observed under program IDs or∼iginofthedetectedH absorption. Thearticleisorganized C54H (PI: W. Sargent), U51H (PI: S. Vogt), and C99H (C. 2 asfollows:InSection2wepresenttheavailableobservations. Steidel).WeusethespectrumobtainedfromtheKeckObser- Analysisrelatedto absorptionspectraand ionizationmodels vatory Archive’s automated reduction pipeline software for are presented in Section 3. In Section 4 analysis of the host our analysis. The pipeline-reduced spectrum was not flux galaxyispresented.OurresultsarediscussedinSection5fol- calibrated; however, the wavelength was both vacuum and lowedbyasummaryinSection6. Throughoutthispaperwe heliocentric velocity corrected. The final reduced spectrum adoptanH =70kms−1Mpc−1,Ω =0.3,andΩ =0.7 covered3200–6100A˚ ata spectralresolutionofR 45,000 0 M Λ cosmology. Solar abundances of heavy elements are taken (FWHM 6.6kms−1)withaspectralS/Nof 20∼and 45 ∼ ∼ ∼ fromAsplundetal.(2009).Allthedistancesgivenareproper per pixel in the blue and red parts, respectively. We note (physical)distances. that the pipeline-reduced spectrum used here is identical to the one independentlyreducedand kindlyprovidedto us by HadiRahmaniusing MAKEE3 datareductionsoftware. Con- tinuum normalizations of all the spectra (COS/FUV, NUV 2. OBSERVATIONS andKeck/HIRES)weredoneby fitting the line-freeregions 2.1. AbsorptionData withsmoothlow-orderpolynomials. 2.1.1. HST/COS 2.2. GalaxyData Far-ultraviolet (FUV) spectra of the background QSO PKS2128–123wereobtainedusingHST/COSCycle-21ob- A 600 s HST/WFPC2 F702W image of the PKS 2128– servationsunderprogramIDGO-13398asapartofour“Mul- 123fieldwasobtainedunderprogramID5143. Thereduced tiphaseGalaxyHalos”survey. ThepropertiesofCOSandits and calibrated image was obtained from the WFPC2 Asso- in-flight operations are discussed by Osterman et al. (2011) ciations Science Products Pipeline. Apparent Vega magni- andGreenetal.(2012). TheobservationsconsistofG130M tudesweredeterminedfrom1.5σisophotesusingSExtractor and G160M FUV grating integrations at a medium resolu- (Bertin&Arnouts1996). Galaxyskyorientationparameters, tionofR 20,000(FWHM 18kms−1)overthewavelength suchasinclinationangle(i)andazimuthalangle(Φ),werede- rangeof1∼140–1800A˚.How∼ever,duetothecompleteLyman- terminedusingGIM2Dmodels(Simardetal.2002)following limitbreakfromtheabsorberofinterestatz =0.4298,no themethodsofKacprzaketal.(2011).Theimagehasalimit- abs QSOfluxisrecordedforλ<1310A˚.Thedatawereretrieved ingmagnitudeof25.5,whichtranslatestoaB-bandabsolute fromtheHST archiveandreducedusingtheSTScICALCOS magnitudeofMB = 15.2andL=0.004L∗atz =0.43. − v2.21 pipeline software. The reduced data were flux cali- Theopticalspectrumofthehostgalaxywasobtainedusing brated. To increase the spectral signal-to-noise ratio (S/N), theKeckEchelleSpectrographandImager(ESI;Sheinisetal. individual exposures were aligned and co-added using the 2002)on2015July16withanexposuretimeof6000s. The IDLcode“coadd x1d”developedbyDanforthetal.(2010)2. mean seeing was 0.9′′ (FWHM) with variable cloud cover. TheESIslitis20′′ longand1′′ wideandweused2 2on- ThecombinedspectrumhasaS/N 9–20perresolutionele- × ∼ chip CCD binning. Binning by two in the spatial direction ment.AstheCOSFUVspectraaresignificantlyoversampled resultsinpixelsizesof0.27′′ 0.34′′overtheechelleorders (i.e. 6rawpixelsperresolutionelement),webinnedthedata − of interest. Binning by two in the spectral direction results by 3 pixels. This improvesthe S/N per pixel by a factor of in a velocity dispersion of 22 km s−1 pixel−1 (FWHM 1 Defined as NN10 = gg01 exp(−E01/kT01), where Ni and gi are the 90kms−1). ∼ columndensityandstatisticalweightoftheithrotationallevel,respectively. 2http://casa.colorado.edu/∼danforth/science/cos/costools.html 3http://www.astro.caltech.edu/∼tb/makee/About.html H absorptionfromagalaxyhalo 3 2 Figure1. Velocityplotofthesub-DLAsystemstudiedhere. Thezerovelocitycorrespondstotheredshiftofthestrongestmetal-linecomponent, i.e.,zabs =0.429805.Theverticaldottedlineineachpanelcorrespondstothehostgalaxyredshiftofzgal=0.42966.Bottom:Sub-dampedLyαprofileoftheabsorber. Theblackhistogramsrepresentthedata. Thesmooth(red)anddashed(blue)curvesrepresentthebest-fittingVoigtprofileandits1σuncertainty,respectively. Top:Variousmetallinesarisingfromthesamesystem.TheabsorptionlinesplottedintheleftmostpanelaredetectedintheHIRESspectra.Allotherabsorption linesarefromCOSspectra. Therelevantpartofeachlineismarkedinblueinordertodistinguishitfromanunrelatedabsorption. NotethatNVandCIVare notdetectedinthissystem. Thespectrumwasreducedusingthestandardechellepack- methodsofKacprzaketal.(2010)whereweextractindividual ageinIRAF4alongwithstandardcalibrationsandwasvacuum spectrabysumming3-pixel-wideaperturesat1-pixelspatial and heliocentric velocity corrected. The data were not flux incrementsalongtheslit. Eachspatialspectrumisthenwave- calibrated due to the variable conditions of the sky. We use lengthcalibratedbyextractingspectraofthearclinelampsat our own fitting software FITTER (see Churchill et al. 2000), the same spatial pixels as the extracted galaxy spectra. The which computes best-fit Gaussian amplitudes, line centers, centroid of each emission line in each spectrum was deter- and widths to obtain emission-line redshifts and equivalent minedwithaGaussianfitusingFITTER. widths. The galaxy rotation curve was extracted following the 3. ABSORPTIONANALYSIS 3.1. DescriptionoftheAbsorber 4 IRAF is distributed by the National Optical Astronomy Observatory, In Figure 1 we display the absorptionprofiles of Lyα and which is operated by the Association of Universities for Research in As- tronomy(AURA)underacooperativeagreementwiththeNationalScience differentmetallinesarisingfromthezabs=0.4298absorber. Foundation. The Lyα absorptionclearly shows a dampingwing. Strong, 4 S.Muzahidetal. Figure2. VelocityplotofnumerousH2absorptionlinesfromdifferentJlevelsarisingfromthesub-DLA.Thezerovelocitycorrespondstozabs=0.429805. Theredandbluesmoothcurvesarethebest-fittingVoigtprofilestothedata(blackhistograms)correspondingtothetwodifferentVPFITsolutionsasdiscussed inthetext. Thebluecurvescorrespondtoalowb-parameter(2.8±0.3kms−1)solution. Theredcurvesrepresentasolutionwithb(H2)=7.1±0.3kms−1. Thelowb-parametersolutionresultsinanN(H2)of∼2ordersofmagnitudehigherascomparedtotheother. Ab-parameterof2.8kms−1issignificantly lowerthanthespectralresolutionofCOS.Moreover,suchalowb-valueisnotobservedforthemetallinesdetectedatthesamevelocityinthehigh-resolution Keck/HIRESdata. Wethereforepreferthesolutionwithb=7.1kms−1,i.e.,theredcurves. Ourphotionizationmodel(seeSection3.3)furthersuggeststhat theN(H2)correspondingtothelowb-parameterrequiresadensityinconsistentwiththeobservedN(MgI)/N(MgII)ratio. saturated absorption lines from MgIIλλ2796,2803 doublets other unblendedneutral species (i.e., OI6, NI, and ArI) ex- aredetectedin thehigh-resolutionKeckspectrum. Thevery hibitasingle-component,simpleabsorptionkinematics. Ex- weak componentat v +90 km s−1 is notdetected in any ceptfortheSiIIλ1020,PIIλ963,andNIIλ10837,presenceof ∼ otherneutral/singlyionizedmetaltransitions. Thetwoother multicomponentstructure is seen in all other singly ionized weak components at v 40 and 25 km s−1 are, how- and higher-ionization absorption lines8. Very strong, satu- ∼ − − ever,presentintheMgI andFeII absorptionlines. Thenew ratedCIIIabsorptionisdetectedovertheentirevelocityrange COS observations cover several metal lines stemming from different elements (e.g., C, N, O, S, Si) and from different ionizationstates(e.g.,OI,CIII,OVI).ExceptfortheCI5,all line6sTheOIλ988.7lineisself-blendedwiththeweakerOIλ988.6andλ988.5 7WenotethattheNIIlineispartiallyblendedwiththeGalacticCIVλ1550 line. 8NotethattheredwingoftheCIIλ1036iscontaminatedwiththeL5R0 5TheredwingoftheCIλ945lineispartiallyblended,possiblywiththe H2linedetectedfromthissub-DLA.Moreover,theNIIIλ989lineispartly Lyαabsorptionfromthezabs=0.1118system. blebdedwiththeSiIIλ989.8absorption. H absorptionfromagalaxyhalo 5 2 Figure3. Voigtprofilefitstothemetallinesarisingfromthesub-DLA.Thesmoothredcurvesarethemodelprofilesoverplottedontopofdata(blackhistograms). Thezerovelocitycorrespondstozabs=0.429805.Thelinecentroidsofindividualcomponentsusedtofitagivenlinearemarkedbytheblueticks. forwhichotherlow-ionizationspecies,MgIIinparticular,are 3.2. AbsorptionLineMeasurements detected. We primarily use the VPFIT9 software for measuring ab- The details of the high-ionizationspecies are presentedin sorption line parameters (i.e., N, b, and z). Additionally, Appendix A since the main focus of this paper is the low- wheneverpossible,wehaveusedthecurve-of-growth(COG) ionization gas-phase giving rise to H absorption. In brief, 2 andtheapparentopticaldepth(AOD)techniques.First,wefit OVIλλ1031,1037 absorption lines are detected in both the allthemetallinesdetectedinthehigh-resolutionKeckspec- doubletsinthevelocityrangeof 100to+200kms−1witha trum simultaneously, i.e., by keepingthe absorptionredshift totallinespreadof∆v =246−kms−1 (seeFigureA1). No 90 and b-parameterof a given componenttied to each other. A CIV and/orNV linesaredetectedinthissystem. TheVoigt minimumoffivecomponentsarerequiredtofittheMgIIdou- profile fit parameters for the OVI absorption and 3σ upper bletssatisfactorily. As noted earlier, the weak componentat limitsonN(NV)andN(CIV)aregiveninTableA1. v +90kms−1 isnotpresentinanyotherlinesdetectedin Besides several metal absorption lines, numerous Lyman- the∼Keck spectrum. Except for the componentat 0 km s−1, and Werner-band absorption lines of H , arising from the 2 we tied the b-parameter of different ions via thermal broad- J = 0, 1, 2, and 3 rotational levels, are detected from this ening with a gas temperature of 104 K. We note that a gas absorber. The H absorption lines that are not self-blended 2 cloudphotoionizedbytheextragalacticUVbackgroundradi- or contaminated by other unrelated absorption are shown in ationcannotcoolefficientlybelow104Kviametal-linecool- Figure 2. Similar to neutral metal-line transitions detected ing (Wiersma et al. 2009). Therefore, it is expectedthat the inthe COSspectra, theH linesfromdifferentJ levelsalso 2 componentswithout detected H will have gas temperatures showasingle-component,simpleabsorptionkinematics. Im- ontheorderoforhigherthan1024K.Forthestrongestmetal- portantly, the absorptionredshifts of H transitions coincide 2 withthestrongestmetal-linecomponent. 9http://www.ast.cam.ac.uk/∼rfc/vpfit.html 6 S.Muzahidetal. line component at 0 km s−1, we set the temperature to be Table1 156K,i.e.,therotationalexcitationtemperature(T01)weob- Component-by-componentVoigtProfileFitParametersfortheMetalLines tain from the H2 column densities in the J = 0 and J = 1 DetectedintheHigh-resolutionKeckSpectrum. levels as discussed below. The Voigt profile fit parameters aresummarizedinTable1,andthemodelprofilesareshown Ion zabs b(kms−1) logN/cm−2 T(K)1 in Figure 3. Since there is no Lyα forest crowdingat low-z MgI 0.429616±0.0000013 5.9±0.4 10.82±0.16 104 MgII ... 5.9±0.0 12.28±0.02 ... and the absorptionlines of interest do notfall on top of any FeII ... 5.6±0.0 12.34±0.03 ... emission line, the continuum fitting uncertainty is generally small and is not taken into accountin the absorption-linefit MgI 0.429685±0.0000015 5.0±0.5 10.76±0.17 104 parameters. MgII ... 5.0±0.0 12.10±0.03 ... FeII ... 4.6±0.0 12.31±0.03 ... Next, we fit the sub-DLA profile detected in the medium- resolutionCOS spectrum. Notethattheline-spreadfunction MgI 0.429805±0.0000009 6.3±0.2 12.08±0.02 156a (LSF)oftheCOSisnotaGaussian. ForourVoigtprofilefit MgII ... 6.3±0.0 13.91±0.04 ... analysisweadoptthelatestLSFgivenbyKriss(2011). The FeII ... 6.3±0.0 13.68±0.02 ... CaII ... 6.3±0.0 11.90±0.02 ... LSF was obtainedby interpolatingthe LSF tables at the ob- MnII ... 6.3±0.0 11.72±0.09 ... served central wavelength for each absorption line and was convolvedwith the model Voigt profile while fitting absorp- MgI 0.429862±0.0000011 4.4±0.2 11.40±0.06 104 tionlinesusingVPFIT.ThetotalHIcolumndensityweobtain MFegIIII ...... 44..40±±00..00 1133..7439±±00..0052 ...... byfittingthedampingwing(Figure1)oftheLyαabsorption CaII ... 4.1±0.0 11.22±0.07 ... isN(HI)=1019.50±0.15cm−2. MnII ... 4.0±0.0 11.20±0.26 ... Following Muzahidet al. (2015b), we choose a set of un- contaminatedH linesforVoigtprofilefitting(seeFigure2). MgII 0.430245±0.0000014 1.9±0.8 11.75±0.04 ...b 2 FeII ... ... <11.00 We fitthem simultaneouslybykeepingthe b-parametertied. MgI ... ... <10.60 Additionally,weconstrainthecolumndensitytobethesame forallthetransitionsfroma givenJ levelandallow itto be Notes–1AssumedgastemperatureintheVoigtprofilemodel. aConstrained different for different J levels. The COS wavelength cal- fromtheT01measuredfromthecolumndensitiesofJ =0andJ =1levels ibration is known to have uncertainties at the level of 10– ofH2.bNotassumedsincenoionsotherthanMgIIaredetected. 15 km s−1 (Savage et al. 2011). We use the numerous H 2 absorptionlinesasguidestoimprovethewavelengthcalibra- componentdetected in the Keck spectrum. The b-parameter tionuncertainty. Ourfinalcorrectedspectrumhasa velocity of 16.8 4.0 km s−1 required for a free fit to the weak SII accuracyof 5kms−1. Asa consequence,we didallow absorpti±onlines is higher than that is observed for the other ∼ ± theredshiftofindividualtransitionstobedifferentfromeach lines. Nevertheless, the estimated column density is robust otherbyamaximumof 5kms−1. since the lines are weak and fall on the linear part of COG. ± Wenoticedthat,dependingontheinitial-guessb-parameter, Finally,duetotheblendintheredwingoftheCIabsorption, VPFIT converges at two different solutions with signifi- weestimatethemaximumcolumndensitythatcanbeaccom- cantly different total N(H ) values. The fit parameters are modatedbykeepingtheb-parameterandcomponentvelocity 2 summarized in Table 2. For an initial-guess b-value of (i.e., zabs) the same as the MgI line. We treat the N(CI) 8kms−1, VPFIT convergeswitha b(H2)of7.1 0.3anda obtainedbythismethodasaconservativeupperlimit. ≥ ± totallogN(H )=16.36 0.08.Ontheotherhand,aninitial- Next, we perform COG analysis for the ions detected in 2 guessb-valueof<8km±s−1 leadstoasolutionwithab(H ) themedium-resolutionCOSspectrawithatleastthreeavail- 2 of2.8 0.3andatotallogN(H ) =18.27 0.03. Suchade- ableunblendedtransitions. TheresultsofourCOG analyses 2 genera±cyisalsopresentwhenweusethest±andardCOGtech- are shown in Figure 4. The best-fitting COG column densi- nique. The model profiles corresponding to the two differ- ties and b-parameters are given in each panel of the figure. entsolutionsareshowninFigure2. Fromthemodelprofiles The b-parameters obtained for both OI and NI, using the and/orfromthereducedχ2values,itisdifficulttochooseone COG method, are consistent with the b-parameter we mea- modelasopposedtotheother. However,wenotethattheb- sureforthestrongestabsorptioncomponentintheKeckspec- parameterinthelattercaseissignificantlylowerascompared trum. Moreover,theyare also consistentwith the valueswe to the velocityresolutionofthe COS spectra( 18kms−1). obtain using VPFIT within 1σ allowed uncertainties (see Ta- Such a narrowb-value is notmeasured in the c∼orresponding ble3). Theinferredb-parameterforSiII issomewhathigher metal-linecomponent,eveninthehigh-resolutionKeckdata. than those for OI and NI. This is possibly due to the pres- Furthermore,wedemonstrateinthenextsectionthattheob- ence of multiple components in the strong SiII transitions. served N(MgI) to N(MgII) ratio does not allow N(H2) to Theb-parameterforSIIisnotwellconstrainedsinceallthree beashighas1018.27 cm−2. Assuch, we excludethehigher transitionsareonthelinearpartoftheCOG.Itisinteresting N(H )solution. to note that at least one line in each case falls on the linear 2 The unblendedlines of NI, OI, ArI, and SII, detected in partoftheCOG,ensuringthatthecolumndensitiesweobtain theCOSspectra,arefittedwithasingle-componentVoigtpro- usingCOGarerobust. filekeepingbothredshiftandb-parameterfree(seeTable3). The AOD technique (Savage & Sembach 1991) is known TheweakArIabsorptionisdetectedonlyintheλ1048transi- to provideaccuratecolumndensitymeasurementsforunsat- tion10. AfreefittotheArIλ1048leadstoahighlyerroneous urated absorption lines. For saturated absorption the AOD solution. We, therefore, fix the b-parameter at 6.3 km s−1, methodyieldsalowerlimitonthecolumndensity. Thecol- umndensitiesfromAODmeasurementsarealsolistedinTa- which we obtained from the fit to the strongest metal-line ble4. NotethatAODcolumndensitiesareconsistent,within 10TheexpectedwavelengthrangefortheotherweakertransitionofArI 1σmeasurementuncertainties,withbothVPFITandCOGcol- (i.e.,λ1066)isnoisyandpartlyblendedwiththeL3P2lineofH2. umn densities, whenever available. For our ionization mod- H absorptionfromagalaxyhalo 7 2 Table2 VoigtprofilefitparametersofH2. zabs (kbm(Hs2−)1) J =0 J =1 logNJ(H=22) J =3 J =4 log(Ncm(−H22))tot logfH2 T(K01) χ2red (1) (2) (3) (4) (5) (6) (7) 0.429807 7.1±0.3 15.74±0.08 16.22±0.09 15.23±0.03 14.83±0.03 <13.8 16.36±0.08 −2.84±0.17 156±47 2.1 0.429807 2.8±0.3 17.65±0.04 18.09±0.02 17.23±0.07 15.93±0.15 <13.5 18.27±0.03 −0.98±0.14 143±17 2.0 Notes–(1)Medianabsorptionredshift. AllH2 absorptionlinesfromdifferentJ levelsareconsistentwithin±5kms−1ofthemedianredshift. (2)Doppler parametersobtainedfromVoigtprofilefitting. Notethat VPFITconverges attwodifferentsolutionsdependingontheinitialb(H2)values(seetext). (3)H2 columndensitiesfordifferentJ levels. Standard3σupperlimits,obtainedfromthenondetectionoftheL10R4transition(fλ =30.76),areprovidedforthe J =4level. Limitsarecalculated assumingthattheb-parameteroftheundetected L10R4lineisthesameasthecorrespondingb(H2)listedincolumn2. (4)TotalH2columndensity,i.e.,thesumofcolumndensitiesobtainedfromdifferentJ levels. (5)Logarithmicmolecularfraction. (6)Rotationalexcitation temperature.(7)Reducedχ2valueforthefitasreturnedbyVPFIT. Figure4. ResultsofCOGanalysisforOI,NI,SiII,andSIIinpanelsA,B,C,andD,respectively.Notethatforeachcaseatleastonetransitionisonthelinear partoftheCOG,whichenablesustoestimatethetotalcolumndensityrobustly. ThisisfurthersupportedbytheindependentAODmeasurementsoftheweak lines(seeTable4)thatareonthelinearpartoftheCOG. Table3 Table4 Single-componentVoigtProfileFitParametersfortheMetalLinesDetected Summaryoftotalcolumndensitymeasurements. intheMedium-resolutionCOSSpectra. Ion VPFIT AODa COG AdoptedValueb Ion zabs b(kms−1) logN/cm−2 OI 15.87±0.06 15.90±0.19 16.02±0.10 15.93±0.12 NI 0.429815±0.000003 6.8±0.6 14.78±0.06 SII 14.33±0.06 14.21±0.20 14.25±0.25 14.26±0.16 OI 0.429811±0.000003 7.7±0.3 15.87±0.06 SiII ... 14.61±0.30 14.70±0.09 14.66±0.18 ArI 0.429828±0.000010 6.3±0.0 13.14±0.08 CI <13.9 ... ... <13.9 CI 0.429805±0.000000 6.3±0.0 13.88±0.13a NI 14.78±0.06 14.55±0.19 14.58±0.06 14.65±0.09 SII 0.429818±0.000012 16.8±4.0 14.33±0.06 ArI 13.14±0.08 12.99±0.28 ... 13.10±0.16 MgI 12.20±0.04 12.20±0.14 ... 12.20±0.09 Notes– Zero error indicates that the corresponding parameter was kept MgII 14.14±0.04 >13.7 ... 14.14±0.04 tied/fixedwhilefitting.aThisvalueshouldbetreatedasanupperlimit. FeII 13.92±0.02 13.90±0.09 ... 13.91±0.05 MnII 11.83±0.13 11.93±0.64 ... 11.88±0.41 CaII 11.98±0.03 11.97±0.17 ... 11.98±0.10 eling in the next section we adopt the column densities that are listed in the last columnof Table 4. Fora givenion, the Notes–Measurementisnottakenusingtheparticularmethodwhenempty. adoptedcolumndensityissimplytheaverageoftotalcolumn aExceptfortheSIIandFeII theweakestavailabletransitionsareusedfor estimating AOD columndensities. Theweakest transitions ofSII(λ1250 densitiesweobtainedusingthedifferentmethods(i.e. VPFIT, andλ1253)andFeII(λ2374)arenoisyandhencearenotused. bAdopted COG,andAOD). columndensitiesforourphotoionizationmodel. 3.3. IonizationModels izationrateof logΓ = 17.3(Williamsetal. 1998). We CR − Inthissectionweexplorethechemical/physicalconditions do not consider the effect of a galaxy/stellar radiation field of the cool gas-phase, giving rise to a plethora of H and sincewe donotknowtheSFR ofthehostgalaxy. However, 2 neutral/singlyionizedmetalabsorptionlines, usingthepho- wenotethattheeffectofagalaxyradiationfieldwillbenegli- toionizationsimulationcodeCLOUDY(v13.03,lastdescribed gibleatthisredshiftforanimpactparameterof 50kpcfrom ∼ by Ferland et al. 2013). We use the observed total column a sub-L∗ ( 0.5L∗) host galaxy (see, e.g. Narayanan et al. ∼ densities of different species to constrain the model param- 2010;Werketal.2014).Thelackofhigher-J(i.e. J >3)ex- eters and assume that all the N(HI) is associated with the citationsalsosuggeststhattheprevailingradiationfieldisnot H component (but see Srianand et al. 2012). In order to strong(Shull&Beckwith1982;Tumlinsonetal.2002). The 2 obtain an accurate H equilibriumabundance,we use full J absorbinggasisassumedtobeaplane-parallelslabirradiated 2 resolvedcalculations,asdescribedinShawetal.(2005),us- bytheionizingradiationfromoneside. ing the ‘ATOM H2’ command. The ionizing radiation used The total HI column density, i.e., logN(HI) = 19.50 ± inourmodelistheextragalacticUVbackgroundradiationat 0.15, measured in this sub-DLA indicates that the absorb- z = 0.42 contributedby both QSOs and galaxies(Haardt& ing gas is optically thick. Therefore, unlike the optically Madau2012).Inaddition,cosmicraysareaddedwithanion- thin medium, the ionization correctionsfor differentspecies 8 S.Muzahidetal. Figure6. Ionization-corrected abundances of different elements for our adoptedphotoionizationmodel. Notethatthex-axisheredoesnotrepresent Figure5. Photoionization-model-predictedmolecularfraction(bottom)and anything.WefindthatMg,Fe,Mn,andCaaresignificantlydepleted. MgI-to-MgIIcolumndensityratio(top)asafunctionofgasdensity.Ineach panelthehorizontaldashedanddottedlinesrepresentthecorrespondingob- servedvalueandits1σuncertainty,respectively. Theverticalbandindicates thealloweddensityrangewithinwhichthemodel-predicted valuesofboth Table5 logfH2andlogN(MgI)/N(MgII)matchwiththeobservations.Weadopt Abundancesofdifferentelements. lognH=−0.2asourdensitysolution. Element Ion [X/H]a ǫb [X/H]c uncorr X corr O OI −0.26±0.19 −0.003 −0.26±0.19 S SII −0.36±0.22 −0.002 −0.36±0.22 arenolongerindependentoftheassumedmetallicityand/or Si SiII −0.35±0.23 −0.009 −0.36±0.23 N(HI). Moreover,thedust-to-gasratioisanotherimportant C CI −2.03 +1.542 <−0.49 inputmodelparametersinceH ispresent.Theobservedtotal N NI −0.68±0.17 −0.004 −0.68±0.17 2 Ar ArI +0.04±0.22 −0.800 −0.76±0.22 N(OI) corresponds to an uncorrected gas-phase metallicity Mg MgI −2.90±0.17 +1.927 −0.97±0.17 of[O/H] = 0.26 0.19. Thus, we useaninputmetallicity Fe FeII −1.09±0.16 −0.010 −1.10±0.16 of[X/H] = −0.3d±exforourmodel. Assumingtheintrinsic Mn MnII −1.05±0.44 −0.009 −1.06±0.44 [Fe/X] to be−solar, the dust-to-gasratio, relative to the solar Ca CaII −1.86±0.18 +0.085 −1.78±0.18 neighborhood,isdefinedas Notes–aAbundancewithoutionizationcorrection. bIonizationcorrectionas κ=10[X/H](1 10[Fe/X]), (1) definedinthetext.cAbundanceafterionizationcorrection. − whereXisanelementthatdoesnotdepleteontodustsignif- ceptionof OI andNI, the ionizationcorrectionis important icantly(seeWolfeetal.2003). AssumingX S, weobtain (i.e., ǫ >0.1 dex) for all neutral species. On the other X ≡ | | logκ = 0.45. Hence, we input logκ = 0.5, with dust hand, except for CaII, ionization corrections are negligibly − − composition similar to Milky Way, in our model. The sim- smallforallsinglyionizedspecies. Theionization-corrected ulationis run on a density grid. For each density, the calcu- abundanceofoxygen[O/H] = 0.26 0.19isthesameas − ± lationisstoppedwhenN(HI)inthesimulationmatchesthe theuncorrectedvalue. Sulfurandsiliconhave0.1dexlower observedvalue. abundancesascomparedtooxygen. Nevertheless,theabun- The variations of logfH2 and the N(MgI)/N(MgII) ra- dances of O, S, and Si are consistent within 1σ. Estimated tio with density in our model are shown in the bottom [C/O]< 0.23and[N/O]= 0.42 0.25suggestthatboth − − ± and top panels of Figure 5, respectively. Parameter f C and N are underabundantwith respect to O. Additionally, H2 shows a sharp increase with density. The observed f wemeasure[Ar/H]= 0.76 0.22dex.Similarabundance H2 − ± value and its 1σ uncertainty correspond to a density in the forthe“inertgas”ArisseenintheMilkyWay“coolclouds” range logn 0.15 to 0.25. Interestingly, the model- (Savage&Sembach1996).Adetaileddiscussiononelemen- H ≃ − − predictedN(MgI)/N(MgII) ratiosalso matchthe observed talabundancesanddepletionsispresentedinSection5.2. valuewithin1σalloweduncertaintyinthisdensityrange.We Althoughtheionizationmodelpresentedaboveisbasedon thereforeadoptlogn = 0.2, correspondingto an ioniza- the observed total column densities, the dominating compo- H tionparameteroflogU =−5.6,asthedensitysolution. The nent is the one at 0 km s−1 in which H is present. This 2 − ∼ predictedgastemperatureat thisdensity (i.e. 135K) is also component comprises about 60% of the total MgII and ∼ consistentwiththeobservedT01. FeII, and 75% of the total MgI column densities. Thus, ∼ We use CLOUDY-computedionizationfractionsof HI and the model parameters essentially correspond to this compo- different metal ions, at lognH = 0.2, for estimating nent. Due to the lack of N(HI) information in individual − ionization-corrected abundances. Abundances of different components, we could not build component-by-component metals,afterionizationcorrections,areshowninFigure6and models. This is because all the higher-order Lyman se- summarizedinTable5. Abundancesarecalculatedusingthe ries lines are heavily saturated. Nonetheless, from the mea- standard convention: [X/H] = log(X/H)gas log(X/H)⊙. sured N(MgI)/N(MgII) and N(MgII)/N(FeII) ratios we − Ionization correction, in this context, is defined as ǫ = infer that the density in the two weak components at v X log(f /f ), where, f and f are the ionization frac- 30 km s−1 (see Figure 3) could be as high as seen in th∼e HI Xi Xi HI − tions of the (i 1)th ionization state of element X and HI, H2 component, provided that the metallicities are not very respectively. I−t is apparent from Table 5, that with the ex- low. TheweakestMgIIcomponentatv 90kms−1 ismore ∼ H absorptionfromagalaxyhalo 9 2 QSO N E 1" 3" Figure7. Left:HST/WFPC2imageoftheQSOfield.ThehostgalaxyislocatedtowardthenortheastwithrespecttotheQSOatthecenter. Middle:Hαand NIIemissionlinesfromthehostgalaxyareshowninthebottompanel;thepresenceofanothergalaxy,possiblyadwarf-satelliteofthehostgalaxy,isclearly visibleintheQSOPSF-subtractedimageofthefieldinthetoppanel.Right:Therotationcurve,MgIIabsorptionkinematics,andthehalomodelofSteideletal. (2002).ThehalomodelisinconsistentwiththeMgIIkinematicsforthemostpart(seetext).Thevelocitiesbelow−50kms−1intherotationcurveareaffected byaskylinecontamination. Theopencirclesintherotationcurvearethemirrorofthecorrespondingmeasurementsat+vevelocities. likely to be similar to the “weak systems” studied by Rigby [NII] emission lines (see Figure 7), allowed us to measure etal.(2002). a redshift of z = 0.42966 0.00016 and a metallicity of gal ± If we use the higher N(H ) solution corresponding to 12+log(O/H)=8.68 0.09forthehostgalaxy.Themetal- 2 ± logf = 0.98 then the requireddensity is logn = 0.8, licity is estimated using the N2 relation of Pettini & Pagel H2 − H i.e., an order of magnitude higher than our adopted density. (2004),i.e.,12+log(O/H)=8.90+0.57 N2,whereN2 × ≡ Atsuchahighdensitythe N(MgI)/N(MgII)ratioisa fac- NII/Hα. Wenotethat(i)thezgal isconsistentwiththezabs tor of 5 higher than the observed value. Additionally, the within1σuncertaintyand(ii)thehostgalaxyhasametallicity ∼ predictedphotoionizationequilibriumtemperatureatthisden- consistentwiththesolarvalue(i.e.,12+log(O/H) = 8.69, sity (i.e. 40 K) is significantlylower thanthe corresponding Asplundetal.2009). T value (143 17 K; see Table 2). Next, Srianand et al. The pointspread function(PSF) subtractedand magnified 01 ± (2012) have found that H components in high-z DLAs are image of the QSO field is shown in the middle column of 2 compact(withsizesof.15pc)andcontainonlyasmallfrac- Figure 7. Interestingly, we found another galaxy at a much tion (.10%) of the total N(HI). If this is also true for the lower impact parameter (ρ = 2.06′′) in this image. In fact, presentsystem,thentheH -bearinggaswouldrequireamuch the presence of this dwarf galaxy is noticeable even in the 2 higher density (> 10 cm−3), which is inconsistent with the left column of Figure 7. This galaxy could be a satellite of density range allowed by the N(MgI)/N(MgII) ratio (see the bigger galaxy at ρ = 48.4 kpc. However, we do not Figure5). Finally,wenotethatanappreciablestellarcontri- have a spectrum of this galaxy to constrain its redshift. As- butiontotheionizingbackgroundontopoftheextragalactic sumingthatthedwarfgalaxyisatz , weobtainanimpact abs UV backgroundwould require a higher gas density in order parameter of ρ = 11.6 kpc and an apparent magnitude of to reproducethe same amountof N(H2) fora givenN(HI) mF702W = 23.20 0.06(Vega). Thedwarf-satellitegalaxy (see,e.g.,Duttaetal.2015;Muzahidetal.2015b). ishighlyinclinedw±ithi= 70.0+15.0 deg. Similartothebig- −21.1 gergalaxy,theprojectedmajoraxisofthesatellitegalaxyalso 4. GALAXYANALYSIS hasasmallazimuthalangle,i.e.,Φ=10.6+−1190..46deg,withre- specttotheQSOsightline. The HST/WFPC2 F702W image of the field is shown The rotation curve of the spectroscopically confirmed in the left column of Figure 7. The host galaxy is located host galaxy is constructed using the [OIII] λ5008 emission at an impact parameter of ρ = 48.4 kpc toward the north- line and is shown in the right column of Figure 7. The east with respect to the QSO. The host galaxy has an incli- zero velocity here corresponds to the z and not to z . nation angle of i = 48.3+3.5 deg. The angle between the gal abs −3.7 The galaxy has a maximum projected rotation velocity of projected major axis of the host galaxy and the QSO sight- v 170 km s−1. Note that the rotation curve at the line (i.e. the azimuthal angle, Φ) is 14.9+−04..69 deg. The az- | mveaxv|el∼ocities is affected by a skyline contamination to the imuthalangleindicatesthattheabsorbinggasislocatednear −[OIII]λ5008line. theprojectedmajoraxis. TheB K colorof2.06(Nielsen WenowconstructthehalomodelofSteideletal.(2002)us- − etal.2013),suggestsamoderatestarformationrate(SFR)in ingi,Φ,ρ,andv ofthehostgalaxy. FollowingKacprzak max the host galaxy. Due to the unavailability of flux-calibrated et al. (2010)we use a gas scale height, h , a free parameter v spectra (see Section 2.2), we could not constrain the SFR ofthe model, of1000kpc (fora non-lagginghaloabovethe of the host galaxy. Nonetheless, the presence of Hα and 10 S.Muzahidetal. absorption or have logf < 4.5. In contrast, a signifi- H2 − cantfraction ( 65%)of the Galactic halo systems (Gillmon ∼ etal.2006),withdetectedH ,havef andN(H )valuesin 2 H2 2 the ranges similar to those seen for the low-z systems. The physical conditions in the H -absorbing gas, therefore, are 2 more similar to those found for diffuse molecular clouds in the Galactic halo than to those for molecular clouds in the Galacticdisk. TherelativelevelpopulationsofH provideasensitivedi- 2 agnosticofgastemperature(Tumlinsonetal.2002;Srianand etal.2005).Theexcitationdiagramforthissystem,shownin Figure 9, indicates thatall the J-levelsare consistentwith a singleexcitationtemperatureofT = 206 6K.Notethat ex ± theT isconsistentwiththeT withinthe1σallowedrange. ex 01 Such gas temperatures, along with the observed N(HI) and f values,areconsistentwiththediffuseatomicgas-phaseof H2 theGalacticISMascharacterizedbySnow&McCall(2006). WhenH issufficientlyshelf-shieldedandcollisionalpro- 2 cessesdominatetheJ =0and1levelpopulations,T repre- 01 sentsthekinetictemperatureofthegas(Royetal.2006;Snow & McCall 2006). Using HI 21 cm observations, Roy et al. (2006)haveshownthattheHIspintemperatures,Ts,arecor- relatedwiththeT forsystemswithlogN(H )&16.0.Note 01 2 thatKanekar&Chengalur(2003)observedbutdidnotdetect HI21cmabsorptionfromthissub-DLA.Fromthenondetec- Figure8. Molecular fraction versus the total hydrogen column density, tion,theauthorsplaceda 3σ lowerlimitonthespintemper- N(HI + H2), for different astrophysical environments: low-z (Muzahid et al. 2015b) and high-z (Noterdaeme et al. 2008) DLAs/sub-DLAs, the atureofTs > 980K, assuminga coveringfactoroffc 1. Galactic disk(Savage etal.1977)andhalo (Gillmonetal.2006), andthe Clearly, such a high temperature is not consistent with∼our Magellanic Clouds (Tumlinson et al. 2002; Welty et al. 2012). For each T or T measurements. The apparent mismatch between sitasmaprele,shsoywstenmbsywoipthendestyemctbeodlsH.2Tahreesshyostwemnbsytufidlileeddhsyermebioslsshaonwdnthbeylitmhe- TesxandT0ex1(orT01)couldbeduetoasignificantlylowercov- red star. The magenta lines delineate the region (logfH2 > −4.5 and eringfactoroftheH2-bearinggas-phase.Denseandcompact logN(HI +H2) < 20.7) within which all the low-z H2 systems and a structures of H2-absorbing gas with small covering fraction majority(i.e.,∼65%)oftheGalactichalosystemsreside. Thedashedline arecommonlyobserved(e.g.,Srianandetal.2012;Duttaetal. correspondstologN(H2)=14.0,whichcanbetreatedasatypicaldetection 2015)andarepredictedinnumericalsimulationsaswell(e.g., threshold. Hirashitaetal.2003). galaxyplane). Such a modelexaminesthe rangeof allowed Thepopulationsofhigherrotationallevels(i.e.,J >3)are determinedbyUVpumpingandformationpumping(Shull& velocities along the line of sight that are consistent with ex- Beckwith1982;Tumlinsonetal.2002). Thissystem,consis- tended galaxy rotation. The model is shown in the bottom paneloftherightcolumnofFigure7.TheD isthedistance tent with other low-z H2 absorbers, do not show absorption los along the line of sight relative to the point where the sight- from J > 3 levels. The lack of high-J excitations clearly indicatesanabsenceofa localUV radiationfield. Thelarge lineinterceptstheprojectedmid-planeofthegalaxy. Ahalo modelisconsideredsuccessfulwhentheD curvespansthe impactparameter(ρ 50 kpc)of the identifiedhostgalaxy los ∼ isconsistentwithsuchanobservation. samevelocityrangeasthatoftheobservedMgIIabsorption. ItisimportanttonoticethattheobservedMgIIkinematicsis 5.2. Elementalabundancesanddust notconsistentwith such a modelof a co-rotatingdisk but is MeasurementsofelementalabundancesinDLAsandsub- consistentwithacounter-rotatinglagginghalo. DLAs provide essential clues on cosmic history of gas and 5. DISCUSSION dust in host galaxies. The main challenge in derivingabun- dance,particularlyinmetal-richsystems,istodisentanglethe 5.1. Physicalconditions nucleosyntheticcontributionsfromdustdepletioneffects.As- As mentioned in Section 1, the Lyman- and Werner-band sessment of intrinsic abundance, from gas-phaseabundance, absorptionofH iscommonlyobservedina widevarietyof of an element becomes complicated when a fraction of that 2 astrophysicalenvironments.InFigure8wehavesummarized elementdepletesontodust. Anon-negligibleamountofdust the H observations from the literature. It is evident from with a dust-to-gas ratio of logκ = 0.45 is present in the 2 − the figurethatthe system studiedhere, consistentwith other sub-DLAstudiedhere. Nevertheless,wecouldestimategas- low-z H systems,showsanunusuallylargef foritstotal phase abundances of several α-elements (i.e., O, S, Si, Ar, 2 H2 (atomic+molecular)hydrogencolumndensity,N . Asnoted Mg,andCa)andFe-peakelements(i.e.,FeandMn),asshown H by Muzahid et al. (2015b), the low-z H systems populate inFigure6. 2 theupperleftcornerofthef –N plane,asmarkedbythe In general, metallicity is derived from [O/H] due to the H2 H magenta lines in the plot. H2 systems in the Galactic disk charge transfer reaction between HI and OI. The inferred (Savage et al. 1977), Magellanic Clouds (Tumlinson et al. metallicity from our photoionization model is [O/H] = 2002; Welty et al. 2012), and high-z samples (Noterdaeme 0.26 0.19. Petitjean et al. (2006) noted a correlation − ± etal.2008)aremostlyDLAswithlogN >20.7. Addition- between gas-phase metallicity and H detection for high-z H 2 ally, the systemswith logN < 20.7eitherdonotshow H DLAs in which higher-metallicity DLAs tend to show H H 2 2