<%BANNER%>

Properties of Quasar Broad Absorption Line Outflows

Permanent Link: http://ufdc.ufl.edu/UFE0044005/00001

Material Information

Title: Properties of Quasar Broad Absorption Line Outflows
Physical Description: 1 online resource (152 p.)
Language: english
Creator: Capellupo, Daniel M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: absorption -- outflows -- quasars
Astronomy -- Dissertations, Academic -- UF
Genre: Astronomy thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Broad absorption lines (BALs) in quasar spectra identify high velocity outflows that likely exist in all quasars and could play a major role in feedback to galaxy evolution. In this dissertation, I use two methods to illuminate important properties of these outflows with the goal of a better understanding of these outflow systems and ultimately of the connection between quasars and their host galaxies. The variability of BALs can help us understand the structure, evolution, and basic physical properties of the outflows. I report here on a BAL monitoring programme of a sample of 24 luminous quasars at redshifts (z) between 1.2 and 2.9. I first focus on CIV 1549A BAL variability in two different time intervals: 4 to 9 months (short-term) and 3.8 to 7.7 years (long-term) in the quasar rest-frame. I find that 39 per cent (7/18) of the quasars varied in the short-term, whereas 65 per cent (15/23) varied in the long-term, with a larger typical change in strength in the long-term data. The variability occurs typically in only portions of the BAL troughs. The components at higher outflow velocities are more likely to vary than those at lower velocities, and weaker BALs are more likely to vary than stronger BALs. I then directly compare the variabilities in the CIV and SiIV 1400A absorption to try to ascertain the cause(s) of the variability. I find that SiIV BALs are more likely to vary than CIV BALs. When both CIV and SiIV varied, those changes always occurred in the same sense (either getting weaker or stronger). The multi-epoch data, including up to 10 epochs of data per quasar, show that the BAL changes were not generally monotonic across the full ~5 to ~8 yr time span of our observations, suggesting that the characteristic time-scale for significant line variations, and (perhaps) for structural changes in the outflows, is less than a few years. The evidence presented here indicates that the cause of variability is likely a complex mixture of changing ionization in the outflowing gas and cloud movements across our lines-of-sight. Part of the BAL monitoring programme specifically focused on obtaining multiple observations at rest-frame time-scales less than 1 month in order to determine whether there is a minimum time-scale threshold below which there is no variability. The shortest variability time-scales help determine how close to the central SMBH this outflowing gas can be located. I detect variability down to a rest-frame time-scale of ~0.02 yr (8-10 days), which constrains the location of the outflowing gas from the central super-massive black hole in these systems down to sub-parsec scales. Finally, in order to determine the viability of quasar outflows as a feedback mechanism affecting galaxy evolution, we need estimates of their mass outflow rates and kinetic energy yields. These quantities depend on the column densities of the flows, which are difficult to obtain directly from spectra of the BALs. We turn to a low-abundance species, PV 1118, 1128A. Phosphorus is much less abundant than, for example, carbon (P/C = 0.001 in the Sun), so a detection of a PV BAL indicates that other lines, such as CIV, are saturated. We detect variability in a PV BAL in Q1413+1143, corresponding to variable SiIV and CIV BALs. The variability in the PV BAL confirms that the absorption is intrinsic to the quasar and provides a constraint on the location of the gas. Using the apparent optical depth of the PV BAL and photoionization models to constrain the true column density of the outflow, we estimate the kinetic energy yields and compare to simulations to find that this outflow could likely be a viable feedback mechanism.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Daniel M Capellupo.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Hamann, Fredrick.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-11-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044005:00001

Permanent Link: http://ufdc.ufl.edu/UFE0044005/00001

Material Information

Title: Properties of Quasar Broad Absorption Line Outflows
Physical Description: 1 online resource (152 p.)
Language: english
Creator: Capellupo, Daniel M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: absorption -- outflows -- quasars
Astronomy -- Dissertations, Academic -- UF
Genre: Astronomy thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Broad absorption lines (BALs) in quasar spectra identify high velocity outflows that likely exist in all quasars and could play a major role in feedback to galaxy evolution. In this dissertation, I use two methods to illuminate important properties of these outflows with the goal of a better understanding of these outflow systems and ultimately of the connection between quasars and their host galaxies. The variability of BALs can help us understand the structure, evolution, and basic physical properties of the outflows. I report here on a BAL monitoring programme of a sample of 24 luminous quasars at redshifts (z) between 1.2 and 2.9. I first focus on CIV 1549A BAL variability in two different time intervals: 4 to 9 months (short-term) and 3.8 to 7.7 years (long-term) in the quasar rest-frame. I find that 39 per cent (7/18) of the quasars varied in the short-term, whereas 65 per cent (15/23) varied in the long-term, with a larger typical change in strength in the long-term data. The variability occurs typically in only portions of the BAL troughs. The components at higher outflow velocities are more likely to vary than those at lower velocities, and weaker BALs are more likely to vary than stronger BALs. I then directly compare the variabilities in the CIV and SiIV 1400A absorption to try to ascertain the cause(s) of the variability. I find that SiIV BALs are more likely to vary than CIV BALs. When both CIV and SiIV varied, those changes always occurred in the same sense (either getting weaker or stronger). The multi-epoch data, including up to 10 epochs of data per quasar, show that the BAL changes were not generally monotonic across the full ~5 to ~8 yr time span of our observations, suggesting that the characteristic time-scale for significant line variations, and (perhaps) for structural changes in the outflows, is less than a few years. The evidence presented here indicates that the cause of variability is likely a complex mixture of changing ionization in the outflowing gas and cloud movements across our lines-of-sight. Part of the BAL monitoring programme specifically focused on obtaining multiple observations at rest-frame time-scales less than 1 month in order to determine whether there is a minimum time-scale threshold below which there is no variability. The shortest variability time-scales help determine how close to the central SMBH this outflowing gas can be located. I detect variability down to a rest-frame time-scale of ~0.02 yr (8-10 days), which constrains the location of the outflowing gas from the central super-massive black hole in these systems down to sub-parsec scales. Finally, in order to determine the viability of quasar outflows as a feedback mechanism affecting galaxy evolution, we need estimates of their mass outflow rates and kinetic energy yields. These quantities depend on the column densities of the flows, which are difficult to obtain directly from spectra of the BALs. We turn to a low-abundance species, PV 1118, 1128A. Phosphorus is much less abundant than, for example, carbon (P/C = 0.001 in the Sun), so a detection of a PV BAL indicates that other lines, such as CIV, are saturated. We detect variability in a PV BAL in Q1413+1143, corresponding to variable SiIV and CIV BALs. The variability in the PV BAL confirms that the absorption is intrinsic to the quasar and provides a constraint on the location of the gas. Using the apparent optical depth of the PV BAL and photoionization models to constrain the true column density of the outflow, we estimate the kinetic energy yields and compare to simulations to find that this outflow could likely be a viable feedback mechanism.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Daniel M Capellupo.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Hamann, Fredrick.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-11-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044005:00001


This item has the following downloads:


Full Text

PAGE 1

PROPERTIESOFQUASARBROADABSORPTIONLINEOUTFLOWS By DANIELMOSHINCAPELLUPO ADISSERTATIONPRESENTEDTOTHEGRADUATESCHOOL OFTHEUNIVERSITYOFFLORIDAINPARTIALFULFILLMENT OFTHEREQUIREMENTSFORTHEDEGREEOF DOCTOROFPHILOSOPHY UNIVERSITYOFFLORIDA 2012

PAGE 2

c 2012DanielMoshinCapellupo 2

PAGE 3

Tomyparents,JoeandSusan 3

PAGE 4

ACKNOWLEDGMENTS IthankmyPhDcommitteemembers,AnthonyGonzalez,VickiSarajedini,Jonathan Tan,andJamesFry,fortheirhelpfulcommentsandcarefulreadingofthisdissertation. Ialsoamverygratefultothosewhohelpedmakemydissertationpossible.TomBarlow madeavailableallthereduceddatafromhisdissertationinsupportofthisproject.Paola Rodr guezHidalgo,helpedmelearntoreducedataandnavigatetheworldofIRAF.Joe ShieldsspentmanynightsattheMDMObservatorytoacquiredataforthisproject.I amgratefulnotjustforthiscontribution,butalsoforhismanywordsofadviceaswellas helpfulcommentsonthemanuscriptsuponwhichmuchofthisdissertationisbased. Icertainlycannotforgetthechairofmycommittee,andadviser,FredHamann.His guidanceaswellashisnumerouscommentsonthemanuscriptsinthisworkhavebeen invaluable. Ihavebeenveryfortunatetohavemanygreatfriendsovertheyears.Duringmy rstcoupleyearsofgraduateschool,Igreatlybenetedfromtheadviceandfriendship ofmyofcematesinRoom309-A:DaveClark,MichelleEdwards,andAshleyEspy. KnicoleCol onhasbeenaclosefriendsinceshearrivedatUniversityofFlorida. Outsideoftheworldofastronomy,Ienjoyedmanyentertainingwalk-and-talkswith StaceySimonandScooter.MichaelTebbihasbeenmyclosestfriendfornearly10 yearsandtherstpersonIturntoforadviceanddiscussiononallthingsnotrelatedto astronomy. Ofcourse,myfamilyhasbeenastrong,supportivepresencethroughoutmylife.I mightneverhavepursuedastronomyifmyfather,JoeCapellupo,hadnotsubscribed toeverysciencemagazinehecouldnd.Bothheandmymother,SusanMoshin,have alwaysencouragedmetochoosemyowndirectionandpursuethelifethatIwant. Last,IwanttoremembertwowomenwhopassedawaywhileIworkedonthis dissertation.Ihavemanyfondmemoriesofspendingapartofmysummereveryyear duringmychildhoodwithmygrandmother,ElisaCapellupo.Ialsowillneverforgeta 4

PAGE 5

verygoodfriendwhotouchedmylifeinmanyways.Besidesbeingagreatconversation partner,DeborahHerbstmanintroducedmetotheworldofballroomdancingandwasa greatroadtriptravelpartner. 5

PAGE 6

TABLEOFCONTENTS page ACKNOWLEDGMENTS .................................. 4 LISTOFTABLES ...................................... 9 LISTOFFIGURES ..................................... 10 ABSTRACT ......................................... 12 CHAPTER 1INTRODUCTION ................................... 14 1.1QuasarsandtheirOutows .......................... 14 1.2VariabilityinQuasarAbsorptionLines .................... 16 1.3ColumnDensitiesandEnergeticsofQuasarOutows ........... 18 1.4OutlineofDissertation ............................. 18 2VARIABILITYINQUASARBROADABSORPTIONLINEOUTFLOWSI.TRENDS INTHESHORT-TERMVERSUSLONG-TERMDATA .............. 21 2.1DataandQuasarSample ........................... 21 2.2Analysis ..................................... 23 2.2.1Short-TermandLong-TermDataSets ................ 23 2.2.2CharacterizingtheQuasarSample .................. 25 2.2.3MeasuringtheBALVariability ..................... 31 2.3Results ..................................... 36 2.4SummaryandDiscussion ........................... 46 3VARIABILITYINQUASARBROADABSORPTIONLINEOUTFLOWSII. MULTI-EPOCHMONITORINGOFSi IV ANDC IV BALVARIABILITY ...... 52 3.1DataandAnalysis ............................... 53 3.1.1ObservationsandQuasarSample .................. 53 3.1.2MeasuringBALsandtheirVariability ................. 56 3.2Results ..................................... 65 3.2.1VariabilityinSi IV versusC IV BALs .................. 65 3.2.2Multi-EpochMonitoringofBALQSOs ................. 75 3.2.3NotesonIndividualQuasars ...................... 79 3.2.3.10119+0310 .......................... 80 3.2.3.20146+0142 .......................... 80 3.2.3.30842+3431 .......................... 82 3.2.3.40932+5006 .......................... 83 3.2.3.50946+3009 .......................... 84 3.2.3.61011+0906 .......................... 84 3.2.3.71232+1325 .......................... 84 6

PAGE 7

3.2.3.81303+3048 .......................... 85 3.2.3.91309 0536 .......................... 85 3.2.3.101331 0108 .......................... 86 3.2.3.111423+5000 .......................... 86 3.2.3.121435+5005 .......................... 86 3.3SummaryofResults .............................. 86 3.4Discussion ................................... 88 3.4.1ChangingIonization .......................... 89 3.4.2ChangingCoveringFraction ...................... 92 3.4.3Conclusions ............................... 93 4VARIABILITYINQUASARBROADABSORPTIONLINEOUTFLOWSIII.WHAT HAPPENSONTHESHORTESTTIME-SCALES? ................ 96 4.1DataandAnalysis ............................... 97 4.1.1ObservationsandQuasarSample .................. 97 4.1.2MeasuringBALsandtheirVariability ................. 100 4.2Results ..................................... 101 4.2.1AmplitudeofVariability ......................... 101 4.2.2ExamplesofVariabilityontheShortestTime-Scales ........ 103 4.2.2.11246 0542:Variabilityover8days ............ 104 4.2.2.20842+3431:Variabilityover10days ............ 106 4.2.2.31011+0906:Possiblevariabilityover17days ....... 109 4.2.2.40932+5006:Possiblevariabilityover31days ....... 110 4.2.2.50903+1734:Variabilityover72days ............ 112 4.2.3Time-scalesforVariability ....................... 114 4.3SummaryandDiscussion ........................... 122 4.3.1ImplicationsofVariabilityover < 0.20yr ................ 123 4.3.1.1Changingionization ..................... 124 4.3.1.2Changingcoveringfraction ................. 125 4.3.2OtherPotentialCausesofVariability ................. 128 5AVARIABLEP V BROADABSORPTIONLINEANDQUASAROUTFLOW ENERGETICS .................................... 130 5.1Data ....................................... 131 5.2AnalysisandResults .............................. 132 5.2.1VariabilityinaP V AbsorptionLine .................. 132 5.2.2LineOpticalDepthsandIonicColumnDensities ........... 135 5.2.3TotalColumnDensity .......................... 139 5.3Discussion ................................... 139 5.3.1LocationoftheOutow ......................... 139 5.3.2EnergeticsoftheOutow ....................... 140 5.4FutureWork ................................... 142 6SUMMARYANDCONCLUSIONS ......................... 144 7

PAGE 8

REFERENCES ....................................... 147 BIOGRAPHICALSKETCH ................................ 152 8

PAGE 9

LISTOFTABLES Table page 2-1QuasardataandC IV variabilityresults ....................... 24 3-1QuasardataforthefulldatasetthroughMarch2009 ............... 54 4-1FullBALquasardatasetandshortesttime-scaledata .............. 99 5-1Observationsummary ................................ 131 9

PAGE 10

LISTOFFIGURES Figure page 2-1Spectraofall24BALquasarsinthevariabilitysample .............. 27 2-2Averagenormalizedabsorptionstrength( # $ )versusoutowvelocity ...... 35 2-3FractionofBALsthatvariedversusoutowvelocity ................ 37 2-4Fractionofquasarsthatvariedversusabsorptionstrength ............ 40 2-5FractionofoccurrencesofBALabsorptionthatvariedversusabsorption strength ........................................ 41 2-6FractionalchangeinthedepthofBALabsorptionasafunctionofvelocity ... 43 2-7FractionalchangeinthedepthofBALabsorptionasafunctionofabsorption strength ........................................ 44 3-1SpectraoftheC IV andSi IV BALsforall24quasarsinthevariabilitysample 57 3-2Averagenormalizedabsorptionstrength, # $ ,inC IV versus # $ inSi IV .... 63 3-3FractionofSi IV BALsthatvariedversusoutowvelocity ............. 68 3-4FractionofoccurrencesofSi IV BALabsorptionthatvariedversusabsorption strength ........................................ 70 3-5FractionofoccurrencesofC IV absorptionthatvariedversus # $ inSi IV .... 71 3-6ChangeinstrengthofC IV BALabsorption, | | ,versus | | inSi IV ...... 73 3-7FractionalchangeinstrengthofC IV BALabsorption, | | / # $ ,versus | | in Si IV .......................................... 74 3-8Multi-epochspectraoffourwell-sampledBALquasars .............. 77 3-9Absorptionstrengthversustimeinfourwell-sampledBALquasars ....... 79 3-10Spectraofthetwolong-termepochsfor0146+0142 ............... 81 3-11Spectraofthetwoshort-termepochsfor0146+0142 ............... 82 4-1Changeinabsorptionstrength, ,versus t ................... 102 4-2Spectraof1246 0542 ................................ 105 4-3SpectraoftheC IV BALin1246 0542 ....................... 106 4-4Spectraof0842 3431 ................................ 107 4-5SpectraoftheC IV BALin0842 3431 ....................... 108 10

PAGE 11

4-6Spectraof1011 0906 ................................ 110 4-7SpectraoftheC IV andSi IV BALsin0932 5006 ................. 111 4-8Spectraof0903+1734 ................................ 112 4-9MeasuredprobabilityofC IV BALvariabilityversustime-scale .......... 113 4-10MeasuredprobabilityofC IV BALvariabilityversustime-scalewithtwomost frequentlyobservedquasarsremoved ....................... 116 4-11MeasuredprobabilityofC IV BALvariabilityversustime-scalewitheachquasar contributingoncetoeachbin ............................ 117 4-12MeasuredprobabilityofC IV BALvariabilityversustime-scaleatvelocities includedinthebalnicityindex ............................ 119 4-13AcumulativedistributionofthefractionofquasarswithC IV variabilitywith time-scale. ...................................... 121 4-14Schematicoftwomodelsforcrossingcloudscenario ............... 126 4-15Transversevelocityanddistanceoftheoutowversus t ............ 127 5-1FullspectrumofQ1413+1143 ............................ 133 5-2LineprolesforC IV ,Si IV ,andP V inQ1413+1143 ................ 134 5-3GuassiantstoC IV ,Si IV ,andP V BALproles .................. 136 5-4OpticaldepthprolesforC IV ,Si IV ,andP V .................... 137 11

PAGE 12

AbstractofDissertationPresentedtotheGraduateSchool oftheUniversityofFloridainPartialFulllmentofthe RequirementsfortheDegreeofDoctorofPhilosophy PROPERTIESOFQUASARBROADABSORPTIONLINEOUTFLOWS By DanielMoshinCapellupo May2012 Chair:FredHamann Major:Astronomy Broadabsorptionlines(BALs)inquasarspectraidentifyhighvelocityoutowsthat likelyexistinallquasarsandcouldplayamajorroleinfeedbacktogalaxyevolution.In thisdissertation,Iusetwomethodstoilluminateimportantpropertiesoftheseoutows withthegoalofabetterunderstandingoftheseoutowsystemsandultimatelyofthe connectionbetweenquasarsandtheirhostgalaxies.ThevariabilityofBALscanhelp usunderstandthestructure,evolution,andbasicphysicalpropertiesoftheoutows. IreporthereonaBALmonitoringprogrammeofasampleof24luminousquasars atredshifts1.2 < z < 2.9.IrstfocusonC IV 1549BALvariabilityintwodifferenttime intervals:4to9months(short-term)and3.8to7.7years(long-term)inthequasar rest-frame.Indthat39%(7/18)ofthequasarsvariedintheshort-term,whereas 65%(15/23)variedinthelong-term,withalargertypicalchangeinstrengthinthe long-termdata.ThevariabilityoccurstypicallyinonlyportionsoftheBALtroughs. Thecomponentsathigheroutowvelocitiesaremorelikelytovarythanthoseatlower velocities,andweakerBALsaremorelikelytovarythanstrongerBALs. IthendirectlycomparethevariabilitiesintheC IV andSi IV 1400absorptionto trytoascertainthecause(s)ofthevariability.IndthatSi IV BALsaremorelikelyto varythanC IV BALs.WhenbothC IV andSi IV varied,thosechangesalwaysoccurred inthesamesense(eithergettingweakerorstronger).Themulti-epochdata,including upto10epochsofdataperquasar,showthattheBALchangeswerenotgenerally 12

PAGE 13

monotonicacrossthefull % 5to % 8yrtimespanofourobservations,suggestingthat thecharacteristictime-scaleforsignicantlinevariations,and(perhaps)forstructural changesintheoutows,islessthanafewyears.Theevidencepresentedhereindicates thatthecauseofvariabilityislikelyacomplexmixtureofchangingionizationinthe outowinggasandcloudmovementsacrossourlines-of-sight. PartoftheBALmonitoringprogrammespecicallyfocusedonobtainingmultiple observationsatrest-frametime-scales < 1monthinordertodeterminewhetherthere isaminimumtime-scalethresholdbelowwhichthereisnovariability.Theshortest variabilitytime-scaleshelpdeterminehowclosetothecentralSMBHthisoutowing gascanbelocated.Idetectvariabilitydowntoarest-frametime-scaleof % 0.02yr (8 10days),whichconstrainsthelocationoftheoutowinggasfromthecentral super-massiveblackholeinthesesystemsdowntosub-parsecscales. Finally,inordertodeterminetheviabilityofquasaroutowsasafeedback mechanismaffectinggalaxyevolution,weneedestimatesoftheirmassoutow ratesandkineticenergyyields.Thesequantitiesdependonthecolumndensities oftheows,whicharedifculttoobtaindirectlyfromspectraoftheBALs.Weturn toalow-abundancespecies,P V 1118,1128.Phosphorusismuchlessabundant than,forexample,carbon(P/C % 0.001intheSun),soadetectionofaP V BAL indicatesthatotherlines,suchasC IV ,aresaturated.WedetectvariabilityinaP V BALinQ1413+1143,correspondingtovariableSi IV andC IV BALs.Thevariability intheP V BALconrmsthattheabsorptionisintrinsictothequasarandprovidesa constraintonthelocationofthegas.UsingtheapparentopticaldepthoftheP V BAL andphotoionizationmodelstoconstrainthetruecolumndensityoftheoutow,we estimatethekineticenergyyieldsandcomparetosimulationstondthatthisoutow couldlikelybeaviablefeedbackmechanism. 13

PAGE 14

CHAPTER1 INTRODUCTION 1.1QuasarsandtheirOutows Activegalacticnuclei(AGN)areextremelybright,compactsourcesatthecentersof manygalaxiesintheuniverse.AnAGNconsistsofasupermassiveblackholewithan accretiondiskofmaterialorbitingaboutit.Quasarsareaparticularlyluminoussubset ofAGN.WhereasthecentralsourceinSeyfertgalaxies,theotherlargesubsetofAGN, istypicallyasbright,invisiblewavelengths,asthecombinedlightfromallthestarsin thehostgalaxy,aquasartendstobebrighterthanthehostgalaxybyatleastafactor of100.Quasarsarethusbrighterthan % ,intheunitofabsolutemagnitude,or roughly30billiontimesbrighterthanourSun( Peterson 1997 ). Whilematerialisaccretingontotheblackhole,thereisevidenceforhigh-velocity owstravelingawayfromthequasarintothehostgalaxy.Thisevidencecomesin theformofblue-shiftedabsorptionlinesinquasarspectra.Therstofthesewide absorptionfeatureswerediscoveredinthelate1960sby Lynds ( 1967 )and Burbidge ( 1970 ). Weymannetal. ( 1991 )laterdenedabalnicityindex'whichclassiesquasars asbroadabsorptionline(BAL)quasarsifanabsorptionlinehasavelocitywidth > 2000 kms atabsorptiondepths > 10%belowthecontinuum.Aquasar'sstatusasaBAL quasaristypicallydenedbasedonthebalnicityoftheC IV 1548,1551 Aresonance line.Otherspecies,includingSi IV 1394,1403 A,canbepresentasBALsinquasar spectra. Roughly % 10-15%ofoptically-selectedquasarshaveC IV BALs( Hewett&Foltz 2003 ; Reichardetal. 2003 ; Trumpetal. 2006 ; Gibsonetal. 2009 Kniggeetal. 2008 ). Sinceonlyafractionofquasarspectradisplaythesefeatures,thepresenceofBALs couldrepresentaphaseintheevolutionofaquasar.Ontheotherhand,allquasars couldhaveintrinsicoutowsandthisfractioncouldindicatethecoveringfractionofthe outowaroundthequasaremittingsource.Variousstudieshavefoundsupportforthe 14

PAGE 15

latterscenario,givensimilaritiesbetweenBALquasarsandnon-BALquasars(e.g., Weymannetal. 1991 Shenetal. 2008 ). BALsarenottheonlyindicatorsofoutowsinquasars.Thereareother,narrower blue-shiftedabsorptionlinesthathavealsobeenidentiedasintrinsictothequasar system:narrowassociatedabsorptionlines(AALs),withvelocitywidthsof % 500kms andmini-BALs,whichhavewidthsbetweenthoseofAALsandBALs.Thisdissertation focusesjustonBALs,butsomecomparisonstostudiesofthesenarrowersystemswill bemadewhereappropriate. Sophisticatedmodelshavebeendevelopedthatenvisiontheseoutowsas arisingfromarotatingaccretiondisk,withaccelerationtohighspeedsbyradiative and/ormagneto-centrifugalforces( Murrayetal. 1995 Proga&Kallman 2004 Proga 2007 Everett 2005 ).Understandingtheseoutowscouldbeimportantforbetter understandingthephysicsofquasars.Theymayplayakeyroleintheaccretionprocess andthegrowthofthecentralSMBHs,byallowingtheaccretingmaterialtorelease angularmomentum. Furthermore,intheearly2000s,correlationsbetweenthemassofcentralSMBHs andpropertiesoftheirhostgalaxieswerediscovered(e.g., Gebhardtetal. 2000 and Ferrarese&Merritt 2000 ),indicatingthattheevolutionandgrowthofSMBHsare somehowconnectedtotheevolutionoftheirhostgalaxies.Quasaroutowsmayplay animportantroleinthisconnectionthrough"feedback"tothehostgalaxy.Theymight bepowerfulenoughtoprovideenoughkineticenergyfeedbacktohaveaneffecton starformationinthehostgalaxies.Theymightalsoaidinunveiling'dust-enshrouded quasarsandhelpdistributemetal-richgastotheintergalacticmedium(e.g., DiMatteo etal. 2005 Molletal. 2007 ). However,thereisstillmuchwedonotknowaboutquasaroutows.Thelocation andthree-dimensionalstuctureoftheoutowsarestillpoorlyunderstood.Improved observationalconstraintsarenecessarytoestimatemass-lossrates,kineticenergy 15

PAGE 16

yields,and,ultimately,theroleofquasaroutowsinfeedbacktothesurrounding environment.Inthisdissertation,IutilizetwodifferentmethodsofstudyingBALquasars, asdescribedbelow,tofurtherourunderstandingofthepropertiesoftheiroutows. 1.2VariabilityinQuasarAbsorptionLines Inthe1980s,repeatobservationsofBALquasarsproducedevidenceforchanges inthelineprolesovertime.Suchvariabilitywasrstreportedby Foltzetal. ( 1987 ), Turnshek ( 1988 ),and Smith&Penston ( 1988 ),allofwhichfocusedonsmallsamplesof 1to3objects. StudyingvariabilityinBALprolescanbeausefultoolforgainingnewinsightinto thestructureanddynamicsoftheoutows.Forexample,wecanuseinformationabout short-termvariabilitytoplaceconstraintsonthedistanceoftheabsorbingmaterialfrom thecentralSMBH.Themorequicklytheabsorptionisvarying,theclosertheabsorber istothecentralsource,basedonnominallyshortercrossingtimesformovingclouds ( Hamannetal. 2008 andSection 2.4 below)orthehigherdensitiesrequiredforshorter recombinationtimes( Hamannetal. 1997 ).Long-termvariabilitymeasurementsprovide informationonthehomogeneityandstabilityoftheoutow.Alackofvariabilityonlong time-scaleswouldimplyasmoothowwithapersistentstructure.Generalvariability resultsprovidedetailsonthesize,kinematics,andinternalmakeupofsub-structures withintheows.Variabilitystudiescanalsoaddresstheevolutionoftheseoutowsas theabsorptionlinesareseentocomeandgo( Hamannetal. 2008 ; Leighlyetal. 2009 ; Krongoldetal. 2010 ; Rodr guezHidalgoetal. 2011 ),oronetypeofoutowfeature evolvesintoanother(e.g.,fromamini-BALtoaBALorviceversa; Gibsonetal. 2010 ; Rodr guezHidalgoetal.inpreparation;alsothiswork). Barlow ( 1993 )istherststudyofBALvariabilityinalargesampleofBALquasars. AnotherlargestudyofBALvariabilitywasnotconducteduntilnearly15yearslater with Lundgrenetal. ( 2007 ),whichtookadvantageofrepeatspectralobservationsofa smallnumberofthequasarswithintheSloanDigitalSkySurvey(SDSS).Both Barlow 16

PAGE 17

( 1993 )and Lundgrenetal. ( 2007 )probetime-scalesof t & 1yr. 1 Gibsonetal. ( 2008 ) takesadvantageofoverlapbetweenthesamplesoftheLargeBrightQuasarSurvey (LBQS)andtheSDSStoinvestigatevariabilityonmulti-yeartime-scales.Both Lundgren etal. ( 2007 )and Gibsonetal. ( 2008 )focusonvariabilityinjusttheC IV linebetween 2epochs,and Gibsonetal. ( 2008 )detectedC IV BALvariabilityin12outof13BAL quasars(92percent). Gibsonetal. ( 2010 )goesfurtherbylookingat3 4epochsof datafor9BALquasarsandfoundtheBALsgenerallydonotvarymonotonicallyover time.TheyalsomakecomparisonsbetweenvariabilityinSi IV absorptionandvariability inC IV ,andtheirresultsincludeacorrelationbetweenfractionalchangeinequivalent width(EW)inSi IV andC IV .Noneofthesestudiesreportclearevidenceforacceleration intheBALows. Thesestudiesarelimitedbyrelativelysmallsamplesizesand/orasmallnumber ofepochs.Thesampleof Lundgrenetal. ( 2007 )isthelargestwith29BALquasars, butwithjusttwoepochsofdataperquasar. Gibsonetal. ( 2010 )hasupto4epochsof data,butforjust9BALquasars.Tobettercharacterizethevariabilityandputstronger constraintsonquasaroutows,largersamplesandmoreepochsofdataarerequired. InChapters 2 4 ,wereportonalarge,uniquemonitoringprogrammeofasample ofBALquasarsdenedin Barlow ( 1993 ).Byusingthesampleof Barlow ( 1993 )asour ducialsampleandobtainingnewer,morerecentdata,wecanlookatvariabilityonboth shortandlongtime-scaleswithinthesamesampleofobjects.Wealsoincludearchival spectrafromtheSDSS,whenavailable,toaugmentthetemporalsampling.Oursample sizeissimilartothatof Lundgrenetal. ( 2007 ),butweobtainupto13epochsofdataper quasar.ThefulldetailsofthesampleandobservationsaredescribedinSection 2.1 1 Throughoutthisdissertation,alltimeintervalsaremeasuredinyearsintherest frameofthequasar,unlessotherwisenoted. 17

PAGE 18

1.3ColumnDensitiesandEnergeticsofQuasarOutows Inthelate1980sand1990s,somequasarswerefoundtohaveunusualabundances intheiroutows.Inparticular,detectionsofP V BALsledtophosphorus-to-carbon abundanceratiosofP/C > % 0.04( Turnshek 1988 ; Junkkarinenetal. 1997 ; Hamann 1998 ).Forcomparison,inthesun,thisratiois0.001( Asplundetal. 2009 ).Such abundancesaredifculttoexplainandarelikelyincorrect.ItismorelikelythattheBALs areopticallythick,andthereisevidencethattheyjustpartiallycoverthebackground continuumsource( Hamannetal. 1993 ; Hamann 1998 ; Aravetal. 1999 ; Hamann etal. 2002 ).Inthiscase,alinecanbesaturatedwithoutgoingtozerointensity.This complicatesthecalculationofthecolumndensitiesintheow.Theapparentoptical depthsofthelinesonlygivealowerlimitonthecolumndensity,andthereforethe mass-lossratesandenergetics,ifthelinesaresaturatedandpartiallycoveringthe continuumsource. ToinvestigatethetruecolumndensitiesinBALoutows,itisnecessarytomake assumptionsaboutthetrueabundancesinthegas.Withtheseassumptionsandusing photoionizationmodels,theexistenceandtheopticaldepthsoflowabundancelines, suchasP V 1118,1128andHeI* 10380,constrainthetruecolumndensitiesinthe ows( Hamann 1998 ; Leighlyetal. 2011 ).InChapter 5 ,wetakethisapproachforone quasar(1413+1143)inourvariabilitysampleforwhichwedetectP V absorption.We thenusetheconstraintsweobtainonthecolumndensitiesoftheows,aswellasthe distanceestimatesfromthevariabilityintheBALs,toestimatetheenergeticsofthe owandtodiscussthepossibilityofthisowasanimportantcontributortofeedbackto galaxyevolution. 1.4OutlineofDissertation Chapter 2 focusesonvariabilityinjusttheC IV BALsinasubsetofthedatafrom theBALmonitoringprogrammeIintroducedabove.ThegoalofChapter 2 istoidentify basictrendsinthedatabetweenvariabilityandotherpropertiesoftheabsorbersandto 18

PAGE 19

directlycompareshort-termandlong-termvariabilitywithinthesamesampleofquasars. Iinvestigatethevariabilityinjusttwodifferenttimeintervals:ashort-termintervalof 0.35 0.75yrandalong-termintervalof3.8 7.7yr.Ialsointroduceanovelapproach tostudyingBALvariabilitybydevelopingameasureofBALstrengthwithinportionsofa trough,insteadofusingequivalentwidth(EW)measurements.Theworkpresentedin Chapter 2 hasbeenpublishedintheMonthlyNoticesoftheRoyalAstronomicalSociety ( Capellupoetal. 2011 ). InChapter 3 ,IextendtheanalysisofChapter 2 bylookingatvariabilityinSi IV and comparingittothevariabilityresultsforC IV (Section 3.2.1 ).Expandingthisstudyto includeSi IV absorptioncanhelpconstraintheoriesonthecause(s)ofBALvariability. CandSihavedifferentabundances,ifsolarabundancesareassumed,andtheyhave differentionizationproperties(e.g. Hamannetal. 2008 2011 ).ByexaminingifC IV andSi IV havedifferentvariabilityproperties,andhowtheydiffer,coupledwiththese differencesinabundancesandionizationproperties,wecangainnewinsightintothe cause(s)ofBALvariability.IalsoincludetheentiredatasetthroughMarch2009tolook atvariabilityinC IV andSi IV overmultipleepochs(Section 3.2.2 ).Thisprovidesupto 10epochsofdataperquasar,andincludingalloftheseepochsgivesbetterinsightinto thecharacteristicsofBALvariability.Increasingthenumberofepochsprovidesnew informationonhowstochasticthevariabilityis.Inotherwords,dotheBALschange monotonicallyovertimeorcantheyvaryandthenreturntoanearlierstate?Ialso highlightseveralindividualinterestingcasesofvariabilitythatcanfurtherhelpus understandBALoutows(Section 3.2.3 ).Thisworkhasbeenacceptedforpublicationin theMonthlyNoticesoftheRoyalAstronomicalSociety( Capellupoetal. 2012 ). WhilesomeoftheearlierworkonBALvariabilityhavedataontime-scales & 1 yr,informationonvariabilityontime-scales < 1monthisverylimited.Thisleaves openthequestionofwhetherthereisatime-scalebelowwhichthereisnovariability inanyobject.Asmentionedearlier,variabilityonshorttime-scalesconstrainsthe 19

PAGE 20

locationoftheoutowinggas.Thus,itisusefultohaveanestimateoftheshortest time-scalesofvariabilityinordertodeterminehowclosetothecentralSMBHthis outowinggascanbelocated.InChapter 4 ,Idescribethemostrecentpartofthe BALmonitoringcampaign,inwhichwetargetedroughlyhalfoftheBALquasarsinour samplemultipletimesover4monthsintheobservedframe(Section 4.1.1 ).Thisallows ustoprobetime-scales < 1monthinthequasarrest-frame,andIhighlightthecases ofBALvariabilitythatoccurredovertime-scalesof < 2months,includingonecaseof variabilityover8days(Section 4.2.2 )Afterincludingthenewdatainthefulldataset, IinvestigatetheincidenceofC IV variabilityversustime-scale(Section 4.2.3 ).Ithen discusstheimplicationsoftheseresults(Section 4.3 ).Thischapterisamanuscriptin preparationforpublication. Finally,inChapter 5 ,Ishowcaseoneobjectfromourvariabilitysampleforwhich wedetecteda(variable)P V line.WerstdiscusstheimportanceofdetectingP V absorptioninaquasarwithvariableabsorptionlines(Section 5.2.1 ).Wethendetermine theopticaldepthsoftheC IV ,Si IV ,andP V BALs(Section 5.2.2 )anduseresultsfrom thephotoionizationmodelof Hamann ( 1998 )todeterminethetotalcolumndensitiesin theows(Section 5.2.3 ).Wethencalculatethemass-lossratesandenergeticsofthe owsanddiscusstheirpotentialimportanceforfeedbacktothehostgalaxy(Section 5.3 ).Thischapterisalsoamanuscriptinpreparationforpublication. 20

PAGE 21

CHAPTER2 VARIABILITYINQUASARBROADABSORPTIONLINEOUTFLOWSI.TRENDSIN THESHORT-TERMVERSUSLONG-TERMDATA Inthischapter,whichiscurrentlypublishedin Capellupoetal. ( 2011 ),Ireportour resultsforC IV 1549variabilitywithmeasurementsselectedtoenablecomparison betweenshort-term(0.35to0.75yr)andlong-term(3.8to7.7yr)behavior.We identifytrendsinthedatawithvelocityandthedepthoftheabsorption.Weavoid usingequivalentwidth(EW)measurements,whichcanminimizeachangethatoccursin aportionofamuchwidertrough.Insubsequentpapers(i.e.chapters),wewilldiscuss variabilitypropertiesintheentiredataset,includingmoreepochsonindividualquasars, samplinginamuchshortertimedomain,andcomparisonsbetweentheC IV andSi IV 1400variabilitiestoplaceconstraintsontheoutowionizationsandcolumndensities. Section 2.1 belowdiscussestheobservationsandthequasarsample.Section 2.2 describesthestepswefollowedtoidentifytheC IV BALsandwheretheyvaried.Section 2.3 describesourresults,andSection 2.4 discussestheimplicationsoftheseresults withcomparisonstopreviouswork. 2.1DataandQuasarSample Between1988and1992, Barlow ( 1993 )andhiscollaborators(e.g., Barlow etal. 1992 )obtainedspectrafor28BALQSOswiththeLickObservatory3-mShane Telescope.Theyobtainedmulti-epochspectrafor23ofthesequasarsforastudyof short-termBALvariability.TheyselectedtheseobjectsfromalreadyknownBALQSOs, withredshiftsof1.2to2.9.Whenselectingobjectsfortheirmonitoringprogram,they gavepreferencetoquasarsknowntohaveeitherabsorptionlineorphotometric variability.However,inmostcases,thisinformationwasnotavailable,andthey selectedmorequasarsathigherredshiftsduetohigherdetectorsensitivityatlonger wavelengths.WhilethissampleofBALQSOsisnotrandomlyselected,itdoescovera widerangeofBALstrengths(Table 2-1 andFigure 2-1 ). 21

PAGE 22

Barlow ( 1993 )usedtheLickObservatoryKastspectrographtoobtainspectrawith settingsforhighresolution, / ( (230kms ),andmoderateresolution, ( (530kms ).Mostoftheobjectswereobservedatbothsettings.Forour analysisofthesedatadiscussedbelow,weusedthehigh-resolutionobservationsfor mostsources.Insomecases,onlyamoderate-resolutionobservationisavailableor choosingthehigh-resolutionobservationwouldcompromisewavelengthcoverage. SinceBALshaveawidthofatleast2000kms ,aresolutionof530kms issufcient todetectchangesintheirproles. StartinginJanuary2007,wehavebeenusingtheMDMObservatory2.4-mHiltner telescopetoreobservetheBALQSOsinthesampleof Barlow ( 1993 ),withthegoalof monitoringBALvariabilityoverawiderangeoftime-scales,from < 1monthtonearly 8years.WealsosupplementedourdatasetwithspectrafromtheSloanDigitalSky Survey(SDSS).Sofar,wehaveover120spectrafor24BALQSOs,andwecontinueto collectmoredata.WenotethattwooftheseobjectsarenotstrictlyBALQSOsbecause theyhaveabalnicityindexofzero(Section 2.2 below).Nonetheless,weincludethemin oursamplebecausetheydohavebroadabsorptionfeaturesatvelocitiesthatweinclude inourvariabilityanalysis(Section 2.3 ). WeusedtheMDMCCDSspectrographwitha350groovepermmgratingin rstorderanda "" slittoyieldaspectralresolutionof ( (250kms ),with wavelengthcoverageof % A.Thespectrographwasrotatedbetweenexposures tomaintainapproximatealignmentoftheslitwiththeparallacticangle,tominimize wavelength-dependentslitlosses.Thewavelengthrangeforeachobservationwas determinedbasedontheredshiftofthetargetquasarsuchthatthecoveragewasblue enoughtoincludetheLy emissionandredenoughtoincludetheC IV emissionfeature. Thecoaddedexposuretimesweretypically1-2hourspersource. 22

PAGE 23

WereducedthedatausingstandardreductiontechniqueswiththeIRAF 1 software.Thedataareuxcalibratedonarelativescaletoprovideaccuratespectral shapes.Absoluteuxcalibrationsweregenerallynotobtainedduetoweatherortime constraints. TheSloanDigitalSkySurveyisanimagingandspectroscopicsurveyofthesky atopticalwavelengths.Thedatawereacquiredbyadedicated2.5-mtelescopeat ApachePointObservatoryinNewMexico.Thespectracovertheobserver-frame opticalandnear-infrared,from3800to9200 A,andtheresolutionis ( (150km s ).Forspectra,typicallythreeexposuresweretakenfor15minuteseach,andmore exposuresweretakeninpoorconditionstoachieveatargetsignal-to-noiseratio.The multipleobservationswerethenco-added( Adelman-McCarthyetal. 2008 ).TheSDSS includesobservationsof11ofthequasarsintheLicksample.However,3oftheSDSS observationsareunusablebecausetheredshiftsofthosequasarsaretoolowforthe C IV BALstoappearintheSDSSspectra.Therefore,wehaveusableSDSSspectrafor 8oftheobjectsinoursample,andtheyweretakenbetween2000and2006. 2.2Analysis 2.2.1Short-TermandLong-TermDataSets FromthedatainourBALmonitoringcampaign,foreachobject,weselectonepair ofobservationsfortheshort-termanalysisandonepairforthelong-termanalysis.For thelong-termanalysis,wewantthelongesttimebaselinepossible.Therstobservation foreachobjectistheearliestLickobservationwiththehighestavailableresolutionas wellaswavelengthcoveragethatextendsfromtheLy emissionlinethroughtheC IV emissionline.Forthesecondobservationforeachobject,weusethemostrecentMDM 1 TheImageReductionandAnalysisFacility(IRAF)isdistributedbytheNational OpticalAstronomyObservatories(NOAO),whichisoperatedbytheAssociationof UniversitiesforResearchinAstronomy(AURA),Inc.,undercooperativeagreementwith theNationalScienceFoundation. 23

PAGE 24

Table2-1. QuasardataandC IV variabilityresults Coord.Short-TermLong-Term (1950.0)Name BIObs.1Obs.2 tVary?Obs.1Obs.2 tVary? 0019+0107UM2322.1322901989.841991.860.65N1989.842007.045.50N 0043+0048UM2752.1443302000.692001.790.35N1991.862008.035.15Y 0119+0310AD85D082.0951701989.85 1991.860.65Y1989.85 2007.075.57Y 0146+0142UM1412.9157801988.93 1991.860.75Y1988.93 2007.054.64Y 0226 1024WFM910226 10242.2577701991.872007.044.66Y 0302+1705HB890302+1702.8901989.84 2007.054.42N 0842+3431CSO2032.1544302007.042008.350.42Y1990.902008.355.54N 0846+1540H0846+15402.9301990.161992.190.52Y1992.192007.053.78Y 0903+1734HB890903+1752.77107002006.312008.280.52Y1989.26 2008.285.04Y 0932+5006SBS0932+5011.9379201989.841991.100.43N1989.842008.286.30Y 0946+3009PG0946+3011.2255501991.101992.230.51Y1991.102008.287.74Y 0957 0535HB890957 0551.8126701990.891992.320.51N1992.322008.355.70N 1011+0906HB891011+0912.2761002007.072008.350.39N1992.192008.354.94Y 1232+1325HB891232+1342.36110001989.26 2008.035.58N 1246 0542HB891246-0572.2448102007.042008.350.40N1992.192008.354.99Y 1303+3048HB891303+3081.7713902007.042008.280.45N1992.322008.285.76Y 1309 0536HB891309-0562.2246901992.192007.044.61N 1331 0108UM5871.88104002007.042008.280.43N1992.322008.285.55Y 1336+1335HB891336+1352.4571201989.26 2008.285.52N 1413+1143HB891413+1172.5668102006.312008.280.55Y1989.26 2008.285.35Y 1423+5000CSO6462.2530602007.072008.350.39N1992.322008.354.93N 1435+5005CSO6731.59115001992.192008.356.25Y 1524+5147CSO7552.8824901989.601991.520.49N1991.522008.284.32Y 2225 0534HB892225-0551.9879201988.461989.840.46N a TheseLickobservationsweretakenatthelowerresolutionsetting( ( ;Section 2.1 ). 24

PAGE 25

spectrum.Wehavelong-termdatafor23objects,withresulting tof3.8to7.7years. Wenotethatforcertainobjects,theredshiftislowenoughthattheLy emissionlineis notshiftedintothewavelengthcoverageofourdataandthereforeisnotpresentinsome ofthespectra(e.g.,0946+3009). Fortheshort-termanalysis,therangeinthevaluesof tamongtheobjects shouldbesmall,whileincludingasmanyobjectsaspossible.Thebestcompromise wefoundwasarangefor tof0.35to0.75yr,whichallowsustoinclude18objectsin theshort-termanalysis.Inordertoincludethismanyobjects,weusedataforsome objectsthatonlycoverstheregionfromtheSi IV emissiontotheC IV emission, whichissufcientfortheanalysisinthispaper.Theseobservationswereselected separatelyfromthelong-termepochsandwerechosentotintothis tinterval. Theseobservationsweretakenbeforeoratanytimeinbetweenthetwolong-term observations.Thereisoneobjectforwhichwehaveshort-termLickdata,butno long-termdata(2225-0534),sowehaveintotal24objectsinourfullsample. 2.2.2CharacterizingtheQuasarSample Table 2-1 summarizesourfullsampleoflong-termandshort-termdata,as describedabove.Therstfourcolumnsprovideinformationonall24quasars. istheemissionredshift 2 ,andBIisthe"balnicityindex,"asdescribedbelow.The remainingcolumnsareseparatedintotheshort-termanalysisandlong-termanalysis. Thecolumnswith"vary?"haveeitherYesorNovaluestoindicatevariabilityanywhere intheC IV troughs(refertoSection 2.3 ),and tisthetimedifference(inyears)inthe quasarframebetweenobservations.Theobservationsfrom1988to1992arefromLick, 2 Thevaluesof wereobtainedfromtheNASA/IPACExtragalacticDatabase (NED),whichisoperatedbytheJetPropulsionLaboratory,CaliforniaInstituteof Technology,undercontractwiththeNationalAeronauticsandSpaceAdministration. 25

PAGE 26

theobservationsfrom2000to2006arefromSDSS,andtheobservationsfrom2007to 2008arefromMDM. Figure 2-1 containsspectraforall24objects,showingthelong-termcomparisons betweenaLickspectrumandanMDMspectrum.Theoneexceptionis2225-0534,for whichweonlyhaveshort-termLickdata.Thevelocityscaleisbasedonthewavelength ofC IV intheobservedframecalculatedfromtheredshiftsgiveninTable 2-1 .Weusea weightedaverageoftheC IV doubletwavelengths,i.e.,1549.05 A. Thebalnicityindex(BI, Weymannetal. 1991 )isdesignedtodifferentiatebetween BALQSOsandnon-BALQSOsandprovideameasureoftheBALstrength,expressed asanequivalentwidthinunitsofvelocity.Itquantiesabsorptionthatisblueshifted between 3000and 25000kms fromtheC IV emissionlinecentre,isatleast % 10% belowthecontinuum,andiswiderthan2000kms .Bythisdenition,ifaquasarhasa BIgreaterthanzero,thenitisclassiedasaBALQSO.ThehighestpossiblevalueofBI is20,000. WemeasureBIinoneobservationperobjecttocharacterizethesampleandallow forcomparisonwithBALQSOsinotherstudies.WedonotuseBIinourvariability analysis.InordertonormalizethespectraandmeasureBI,andlaterabsorption strength(Section 2.2.3 ),wersthadtotapseudo-continuumtoonespectrumper quasar,whichincludesthequasar'scontinuumemissionplusthebroademission lines.WeadopttheLickobservationsusedinthelong-termanalysisasourducial epochforwhichwetthepseudo-continuaandmeasureBI.Weconstructedthis pseudo-continuumbyttingapower-lawtothecontinuumawayfromanyemission orabsorptionlines,andthenaddingtothatanestimateoftheC IV and,ifnecessary, theSi IV broademissionline(BEL)proles 3 .Totthecontinuum,wetapower-law functiontotworegionsonthespectrumthatarefreeofstrongemissionlines.The 3 NotethatthefeatureweattributetoSi IV mayalsoincludeO IV ] 1400emission. 26

PAGE 27

Figure2-1. Smoothedspectraofall24quasarsinoursample,showingthelong-term comparisonsbetweenaLickObservatoryspectrum(boldcurves)andthe mostrecentMDMdata(thincurves).Theoneexceptionis2225-0534,for whichweonlyhaveshort-termdata(Table 2-1 ).Theverticaluxscale appliestotheLickdata,andtheMDMspectrumhasbeenscaledtomatch theLickdatainthecontinuum.Thedashedcurvesshowour pseudo-continuumts.ThethinhorizontalbarsindicateintervalsofBAL absorptionincludedinthisstudy,andthethickhorizontalbarsindicate intervalsofvariationwithintheBALs.Theformal1 # errorsareshownnear thebottomofeachpanel. 27

PAGE 28

Figure2-1. Continued. 28

PAGE 29

Figure2-1. Continued. 29

PAGE 30

preferredwavelengthrangesforthetswere1270to1350 Aand1680to1800 A.These rangeswereadjustedtoavoidemissionandabsorptionfeaturesasmuchaspossible orduetothelimitsofthewavelengthcoverage.FittingtheC IV emissioniscomplicated by,forexample,potentialasymmetriesintheemissionprolesandabsorptioninthe emissionregion(e.g., Reichardetal. 2003 ).Therefore,weareexibleinourtting routine,usingbetween1and3gaussianstodenethelineprole.Thegaussians havenophysicalmeaning,andwesimplyusethemforobtainingareasonablettothe prole.Ifthereisanyabsorptionontopoftheemission,weignorethosewavelengths whendeningthet.Incaseswheretheabsorptioniswideenoughthatitobscuresup tohalfoftheemissionline(e.g.1413+1143),weassumetheemissionlineissymmetric. Incaseswherethegaussian,ormultiple-gaussian,tdidnotprovidetherightshape andwehaveenoughinformationabouttheprole(e.g.1423+5000),wemanually adjustedthettobettermatchtheC IV emissionline. ForquasarswheretheC IV absorptionisatavelocitythatoverlapstheSi IV emission,wealsottheSi IV emissionline.Ifmost,orall,oftheSi IV emissionfeature isabsorbed,wesynthesizedaSi IV prolebytakingtheC IV t,shiftingitscentre wavelengthtothelocationofSi IV ,adjustingtheFWHM(giventhewiderseparation betweentheSi IV doubletlines),andscalingtheemissionstrengthbyafactorof0.342. Thisscalefactorisbasedoncompositequasarspectrafromwhich VandenBerk&etal. ( 2001 )foundthetypicalstrengthofvariousemissionlines.Wereliedonthisscalefactor incasessuchas0146+0142,wherewehavelittleinformationabouttheSi IV emission line.Otherwise,inacasesuchas0846+1540,wetbetween1and3gaussianstothe Si IV line,aswedidfortheC IV emission,becausewehavemoreinformationaboutthe proleoftheSi IV emissionline.Thepseudo-continuumtsforeachobjectareshownby thesmoothdottedcurvesinFigure 2-1 30

PAGE 31

2.2.3MeasuringtheBALVariability Foreachquasar,wecomparethetwospectrainourshort-termandlong-term intervals(Table 2-1 )tosearchforvariabilityintheC IV BALs.First,wedeterminedthe velocityrangesoverwhichBALabsorptionoccurs.Theserangesmightdifferbetween thetwoobservingepochs.Wethereforesettherangetotheextremesobservedin eitherepoch.OurdenitionoftheserangesisguidedbythedenitionofBI,i.e.they mustpresentcontiguousabsorptionreaching ) 10%belowthecontinuumacross ) 2000 kms .Wedonotfollowtherequirementthattheabsorptionbeblueshiftedbetween 3000and 25000kms becausewewanttoincludeallC IV broadabsorptionin ouranalysis(asaresult,oursamplecontainstwoobjects,0302+1705and0846+1540, withaBIof0).Ingeneral,theC IV BALsoccurbetweentheSi IV andC IV emission lines,butsomeoftheC IV absorptionisblue-shiftedenoughthatitappearsblueward oftheSi IV emissionlinecenter(e.g.,0146+0142).Inthesecases,wecanconrmthat theabsorptionbluewardoftheSi IV emissionlineisduetoC IV ,andnotSi IV ,because anysignicantSi IV absorptionshouldbeaccompaniedbyC IV absorptionatthesame velocity.ThisisevidentfromnumerousobservationsofBALQSOswheretheSi IV and C IV absorptionareclearlyidentied(e.g.,gure1in Koristaetal. 1993 ),andfrom theoreticalconsiderationsoftherelativestrengthsoftheselines( Hamannetal. 2008 ). Oursampleincludesbroadabsorptiontroughsfrom0to 40000kms Inordertocompareanytwospectraforvariabilityinthedenedabsorptionranges, wescaledthespectrasothatregionsfreeofabsorptionandemissionineachspectrum match.Thisisnecessarybecauseofuncertaintiesintheuxcalibrationandpossible realchangesinthequasaremissions.Weappliedasimpleverticalscalingsothatthe spectrafromeachepochmatchtheducialLickspectrumalongthecontinuumredward oftheC IV emissionfeature(i.e.,from1560 Atothelimitofthewavelengthcoverage), betweentheSi IV andC IV emission( % 1425-1515 A),andbetweentheLy andthe Si IV emission( % 1305-1315 A).Innearlyallcases,asimplescalingproducedagood 31

PAGE 32

matchbetweenthespectrafromdifferentepochs.Theresultsforthetwolong-term epochsareshowninFigure 2-1 .Insomecases,however,thereweredisparitiesinthe overallspectralshapeofthetwospectra.Toremovethesedisparities,weteithera linearfunction(for0903+1734,1413+1143,and1423+5000)orquadraticfunction(for 0019+0107and1309-0536)tothespectralregionsthatavoidtheBALsinaratioofthe twospectra.Wethenmultipliedthisfunctionbythecomparisonspectrum. WeidentifyBALvariabilityinvelocityintervalsthatcorrespondtoatleast2 resolutionelementsinthelowerresolutionLickspectra,i.e.,1200kms .Our proceduretoidentifyvariabilityintervalsreliesmainlyonvisualinspectionbecause photonstatisticsalonearenotsufcientfordeningrealvariability.Mostofthe occurrencesofvariabilityrecordedforthisstudyareobviousbysimpleinspection. However,wealsoexaminedallcasesofweakorpossiblevariabilitytoensurethatno occurrencesofsignicantrealvariabilityaremissed.Wedeterminedthesignicance ofvariabilitybasedontwocriteria.First,wecalculatetheaverageuxandassociated errorwithineachvariabilityintervalandusethattoplaceanerrorontheuxdifferences betweenthetwospectra.Weincludeinthisstudyonlytheintervalsofvariabilitywhere theuxdifferencesareatleast4-sigma.Weemphasizethatalloftheintervalswhich variedby4-sigmaormorearereadilyidentiedbyourinitialinspectiontechnique. Therefore,thereisnodangerofvariableintervalsbeingmissed.However,the4-sigma thresholdalonecanincludespuriousorhighlyuncertainvariabilityresultsbecauseof uncertaintiesnotcapturedbythephotonstatistics.Theseadditionaluncertaintiescan arisefromtheuxcalibration,apoorlyconstrainedcontinuumplacement,underlying emissionlinevariabilityorourreconstructionofaseverelyabsorbedemissionline prole.Wethereforeapplyasecondsignicancecriterionbasedonourassessmentof theseadditionaluncertainties.Wetakeaconservativeapproachbyexcludingvariability intervalswheretheadditionaluncertaintiesmightbelarge(eveniftheuxdifference passestheformal4-sigmathresholddescribedabove).Thissecondcriterioncaused 32

PAGE 33

theexclusionofjustafewintervals.Seebelowforexamples.Thevelocityintervals nallyclassiedasvariableinthelong-termdataareindicatedbyboldhorizontallinesin Figure 2-1 Tohelpclarifyourassessmentsoftheadditionaluncertainties,wepointoutsome specicvelocityintervalsthatpassthe4-sigmathresholdandappearvariablebycasual inspection,butarenonethelessexcludedfromouranalysisbecausetheadditional uncertaintiesaresignicant.Forexample,in1246-0542,wedonotincludetheregion around 17,000kms becausethenoiselevelinthatportionofthetroughmakesit averytenuouscandidateforinclusioninthestudy.Theregionaround 21,000kms in1435+5005isasimilarcasebecauseheretheincreasednoiselevelofthespectra towardsgreateroutowvelocitiesmakesadeterminationofvariabilitymoreuncertain. Wenotethattheincreasedscatterinthedataforthesetwoobjectsislessnoticeable inthesmoothedspectraplottedinFigure 2-1 .Inacouplecases,aslightdifferencein theslopebetweentwocomparisonspectraaddssomeuncertainty.In1413+1143,the bluewingoftheC IV troughintheMDMspectrumextendsslightlyabovethedened continuum.Thismaybeduetosomeanomalyintheuxcalibration,sowedenethe blueendoftheBAL(andthevariabilityinterval)as 20,600kms .In2225-0534, thereisverylittlecontinuumavailableformatchingthetwocomparisonspectra,andthis addeduncertaintymakesthesmalluxdifferenceatthebluestportionoftheBALtrough tootenuoustobeincluded.Aslightdifferenceinslopebetweenthetwoepochscould beresponsibleforanyapparentchangesinthewingofthisBAL,andwedonothave enoughvisiblecontinuumtodiscountthispossibility. IdentifyingvariabilityinthespectralregionswhichoverlaptheC IV andSi IV emissionlinesisfurthercomplicatedbypossiblevariabilityintheemissionlines themselves(e.g., Wilhiteetal. 2005 ).WerecordBALvariabilityinthesespectralregions onlyifitislargecomparedtoanyreasonablechangeinthebroademissionlinesand/or theabsorptionchangesabruptlyoverjustportionsofaBELprole.Forexample,in 33

PAGE 34

0146+0142,theintervalfrom 38,800to 32,700kms clearlyshowsBALabsorption weakeningtoapproximatelycontinuumlevel.Inanothercase,0932+5006,weinclude theintervalsfrom 37,500to 35,400kms and 28,400to 27,000kms because thereisclearlyBALabsorptionthereintheMDMspectrum.Wedidnotattributethe changeinuxfromapproximately 34,000to 30,000kms toaBALbecauseitis unclearwhatistheprimarycauseofvariabilityinthisinterval.Whiletheidentications ofBALabsorptionandvariabilityintheregionsoverlappingC IV andSi IV emission aresecure,themeasurementsofabsorptionstrength( # $ ;seebelow)inthisregion haveaddeduncertaintysincettingtheemissionlinesismoreuncertainthanttingthe continuum.Whenlookingatcorrelationswithvelocityinthisstudy,weidentifytrends usingallthedata,butalsoconrmtheexistenceofthesetrendsinthevelocityrange 27,000to 8,000kms ,whichavoidstheemissionlines.Forcorrelationswith absorptionstrength(seebelow),sincetherearefewcasesofvariabilityattheextremes oftheexaminedvelocityrange,theirinclusiondoesnotsignicantlyaltertheplotsorthe nalresults. Inordertolookfortrendsinthedata,wecalculatedtheabsorptionstrengthas afunctionofvelocityineachspectrum.Werstnormalizedeachspectrumusingthe pseudo-continuumt(Section 2.2.2 )forthatquasar.Wecanapplythepseudo-continua derivedinSection 2.2.2 totheotherepochsforeachquasarbecausethoseepochs havealreadybeenscaledtomatchtheducialLickspectrum.However,weadjust theemissionlinetsincaseswheretheemissionlinevaried.Wedeneabsorption strength, ,asthefractionofthenormalizedcontinuumuxremovedbyabsorption (0 & & 1)withinaspeciedvelocityinterval.Asingle valueisadoptedforeach velocityintervalbasedontheaverageuxwithinthatinterval.Ifthereisvariability,then theabsorptionstrengthchangedandwecalculatetheaverageabsorptionstrength, # $ andthechangeinabsorptionstrength, ,betweenthetwoepochsbeingcompared. Thevelocityintervalsusedtocalculate # $ and aredenedtoi)includeonlyportions 34

PAGE 35

Figure2-2. Theaveragenormalizedabsorptionstrength( # $ ),calculatedthroughoutall oftheBALsinthequasarsinourlong-termsubsample,asafunctionofthe outowvelocityoftheabsorbingmaterial.Theindividualpointsineachbin (withwidth1000kms )representthe # $ forasinglequasar,andtheopen squaresrepresenttheaverageofallpointsinthatbin. ofthespectrumwithBALabsorptionandii)clearlyseparatethespectralregions withinBALtroughsthatdidanddidnotvary.Specically,thevariableandnon-variable absorptionregionsareeachdividedintoequal-sizedbins,withabinwidthof1000to 2000kms dependingonthevelocitywidthoftheregion.Thisstrategymakesthebin widthsusedtomeasure # $ and similartothenominalwidthof1200kms used above,whileprovidingtheexibilityneededtobeginandendattheboundariesdened abovebetweenabsorbed/unabsorbedandvariable/non-variablespectralregions.In eachofthesevelocitybins,wealsocalculatethefractionalchangeintheabsorption strength, | | / # $ ,whichcanhavevaluesfrom0to2.Spectralregionsthatdidnotvary accordingtothecriteriadescribedabovehave settozeroautomatically.Notethat, with # $ and determinedinthisway,eachquasarcancontributemorethanonce 35

PAGE 36

tothesequantitiesinourvariabilityanalysis,dependingontherangeofstrengthsand velocitiescoveredbyitsBALtroughs,becauseeachbinshowingvariabilityistreatedas aseparateoccurrence. Weplottherelationshipbetweenabsorptionstrength, # $ ,andoutowvelocity forallthequasarsinourlong-termsubsampleinFigure 2-2 .Acorrelationbetween thesetwoquantitieswillhaveimplicationsforourresultswhenwecomparevariability measurestoeachofthesequantitiesindividually(Section 2.3 andSection 2.4 ).Thereis atrendforlowervaluesof # $ atincreasingoutowvelocity,whichisespeciallyapparent from 27,000to 8,000kms ,awayfromtheregionsofSi IV andC IV emission. Theerrorsintheindividualpointsaremuchsmaller( # # $ & 0.01)thanthedispersion ofpointsineachbin.However,thebinsat < -30,000kms onlycontainoneortwo quasars,sotheaverage # $ inthosebinscanbeuncertainbyuptoafactorof2,given theadditionaluncertaintyofttingtheSi IV emissionline.Theweakerabsorptionfound athighervelocitiesinoursampleisconsistentwithabsorptioninthemeanBALspectra constructedby Koristaetal. ( 1993 ). 2.3Results Table 2-1 indicateswhethertherewassignicantvariabilityintheC IV BALsof eachquasarintheshort-termandlong-termdata.Wefoundthat39%(7/18)ofquasars withshort-termdatavaried,whereasinthelong-term,65%(15/23)varied.Onequasar variedonlyintheshort-term,socombiningbothshort-termandlong-termdata,67% (16/24)ofthequasarsvariedoverall. WeinvestigatedC IV BALvariabilityasafunctionofoutowvelocity.Usingthe velocityrangesforeachquasarwherewehaveidentiedabsorptionandvariability (Section 2.2.3 ),wecounthowmanyquasarshaveBALabsorptionandvariabilityin eachvelocitybin.InFigure 2-3 ,thetoptwopanelsshowthenumberofquasarswith C IV BALabsorptionandthenumberwithC IV BALvariabilityateachvelocity.Thethird panelisthesecondpaneldividedbythetopone,whichgivesthefractionofC IV BALs 36

PAGE 37

Figure2-3. ThetoptwopanelsshowthenumberofquasarswithC IV BALabsorption andthenumberwithC IV BALvariabilityateachvelocity.Thethirdpanelis thesecondpaneldividedbythetopone,whichgivesthefractionofC IV BALsthatvariedateachvelocity.Thedashed,dotted,andsolidcurves indicateresultsforthelong-term,short-term,andcombineddataset comparisons,respectively.Thebottompaneldisplaysthecombineddataset with1 # errorbars. 37

PAGE 38

thatvariedateachvelocity.Thedashed,dottedandsolidcurvesindicateresultsforthe long-term,short-termandcombineddatasetcomparisons,respectively.Theplotimplies thatBALsathighervelocitiesaremorelikelytovarythanthoseatlowervelocities. ThebottompanelinFigure 2-3 displays1 # errorbarsbasedon Wilson ( 1927 )and Agresti&Coull ( 1998 ).Theseerrorsarebasedoncountingstatisticsforthenumber ofquasarswithabsorptionandvariabilityateachvelocity.Weperformaleast-squares ttoconrmthatacorrelationexistsbetweenincidenceofvariabilityandvelocity.The slopeofthecombineddatais ,intheformalunit,fractionperkms Theslopeisnon-zeroata6-sigmasignicance.Ifwerepeattheleast-squarestwith onlythebinswiththemostdata(binscontainingatleast8quasarswithabsorption), theslopeis fractionperkms ,whichisnon-zeroata5-sigma signicance.Usingthisrestrictionlimitsthettoonlythedatafrom 29,000to 2,000 kms .Weperformtheleast-squarestathirdtime,wherewerestrictthettoonlythe dataintherange 27,000to 8,000kms thatcompletelyavoidsoverlapwiththeC IV andSi IV broademissionlines.Inthiscase,theresultingslopeis fractionperkms ,whichisnon-zeroata2.7-sigmasignicance.Therefore,thistrend isevidentevenwhenweremovethedataatthelowestandhighestvelocities,where additionaluncertaintyfromtheBELscanaffecttheresults. AnimportantquestioniswhetherthetrendseeninFigure 2-3 isacorrelationwith velocityinanabsolutesense,orthecumulativeresultofdifferencesbetweenthered versusblue-sidebehaviorsinindividualabsorptiontroughs.Toaddressthisquestion, wedividedthevelocityrangeofabsorptionforeachquasarintoaredandabluehalf andplottedtheincidenceofvariabilityseparatelyfortheredandblueportions.The resultinbothcasesisconsistentwiththebehaviorshowninFigure 2-3 .Wedonotsee ahigherincidenceofvariabilityintheblueportionsoftroughsversustheredportions. ThisindicatesthatthetrendinFigure 2-3 isatrendintheensembledependentonthe absolutevelocityratherthantherelativevelocitywithinagiventrough.Consistentwith 38

PAGE 39

thisndingisthefactthatnoneoftheBALsvariedatavelocitybetween 3500kms and0kms ,whilethetwoBALsinoursampleat < 33,000kms bothvaried. AnotherfactorthatmightaffecttheresultsinFigure 2-3 isthatBALsathigher velocitiestendtobeweaker(Figure 2-2 ).InFigure 2-4 ,weexamineC IV BALvariability asafunctionofabsorptionstrength.Usingthevaluesof # $ calculatedinSection 2.2 ,andabinsizeof0.05,wecounthowmanyquasarshaveabsorptionineachbin. Wethencounthowmanyquasarshaveatleastoneoccurrenceofvariabilityineach absorptionstrengthbin.Eachquasariscountedatmostonceineachbin,evenifithas thesameabsorptionstrengthatmorethanonelocationinthespectrum.AsinFigure 2-3 ,thethirdpanelisthesecondpaneldividedbythetopone,inthiscaseyieldingthe fractionofquasarswithaBALthatvariedineachabsorptionstrengthbin.Figure 2-4 indicatesthatquasarsaremorelikelytoexhibitBALvariabilityatweakerabsorption strengths.AsinFigure 2-3 ,wecalculateerrorsforeachabsorptionstrengthvalue.We againperformaleast-squarest,whichgivesaslopeof 0.487 0.095fractionper unitabsorptionstrength.Theslopeisnon-zeroata5-sigmasignicance.Werepeatthe least-squarestwithonlythebinswiththemostdata,asforFigure 2-3 ;thisrestrictsthe ttoallofthebinsexcepttheoneat # $ > 0.95.Theresultingslopeis 0.476 0.100 fractionperunitabsorptionstrength,whichisnon-zeroata4.8-sigmasignicance. Figure 2-4 ,however,doesnotaccountformultipleoccurrencesofthesame absorptionstrengthwithinanindividualquasar.InSection 2.2.3 ,wedeneamethod ofdividingeachBALinto % 1200kms bins,andwenowconsiderthesebinsas individualoccurrencesofabsorption.InFigure 2-5 ,weplotthenumberofthese occurrencesateachabsorptionstrengthvalueinthetoppanelandthenthenumberof theseoccurrencesthatvariedinthesecondpanel.Thethirdpanelisthesecondpanel dividedbythetopone,showingthefractionofoccurrencesateachabsorptionstrength thatvaried.AnindividualBALmaybecountedmorethanonceateachabsorption strengthvaluebecauseweconsiderportionsofeachBALtroughinthisgure.The 39

PAGE 40

Figure2-4. SameasFigure 2-3 exceptthequantitiesareplottedasafunctionofthe normalizedabsorptionstrength,averagedbetweentheepochsbeing compared. 40

PAGE 41

Figure2-5. Thetoptwopanelsshowthetotalnumberofoccurrencesofabsorption,and thenumberofthoseoccurrencesthatvaried,ateachstrength.Thethird panelisthesecondpaneldividedbythetopone.Thebottompaneldisplays thelong-termdatawith1 # errorbars. 41

PAGE 42

bottompanelisthesameasthethirdpanel,exceptitshowsjustthelong-termdatawith errorbars,calculatedasinFigures 2-3 and 2-4 .Aleast-squarestforthelong-term datainFigure 2-5 givesaslopeof 0.380 0.052fractionperunitabsorptionstrength, whichisnon-zeroata7-sigmasignicance.Figure 2-5 conrmsthatweakerportionsof BALtroughsaremorelikelytovarythanstrongerones. Figures 2-3 2-5 comparethe incidence ofvariabilitytopropertiesoftheBALs. Next,weusethevaluesof | | / # $ ,calculatedinSection 2.2 ,tocomparethe amplitude ofvariabilitytothesameBALproperties. InFigure 2-6 ,weplotthisfractionalchangeinabsorptionstrengthasafunction oftheoutowvelocity,usingbinsizesof1000kms width,asinFigure 2-3 .Ineach velocityinterval,theindividualpointsrepresentvariabilityinadifferentquasar.The averagevaluesof | | / # $ forjustthevariable(non-zero)pointsineachbinareshown bytheopensquares.Theasterisksshowtheaverageifthenon-varyingabsorption intervalsareincluded.Thisplotdoesnotindicateastatisticallysignicanttrendbetween fractionalchangeinabsorptionstrengthandtheoutowvelocity. InFigure 2-7 ,weplotthefractionalchangeasafunctionofaveragenormalized absorptionstrength,similartoFigure 2-6 .Inthisplot,however,individualquasarswill contributemultipletimestothemeasurementsatagivenabsorptionstrengthifthe quasarhasvariabilityinboththeredandbluesidesofanabsorptiontroughand/or inmultipletroughs.Theindividualpointsrepresentvariabilityatdifferentvelocitiesin eachquasar.Theaverageofallthevariable(non-zero)pointsineachbinareshown bytheopensquares,whiletheasterisksshowtheaverage | | / # $ ineachbinif non-varyingabsorptionregionsareincluded.Thisplotindicatesthatthefractional changeinstrengthisgreaterinweakerfeatures.However,theerrorsinthevaluesof | | / # $ arecorrelatedwiththevaluesof # $ .Atan # $ of0.15,thetypicalerrorvalues rangefrom % 0.11-0.33,whereasat # $ of0.60,theerrorvaluesare % 0.02-0.06.A largererroratsmallervaluesof # $ shouldcausealargerdispersioninthepointsinthat 42

PAGE 43

Figure2-6. FractionalchangeinthedepthofBALabsorptionasafunctionofvelocity. ThesymbolsfollowthesamepatternasFigure 2-2 ,withindividualpoints representingvariabilityinasinglequasar.Here,theopensquaresindicate theaverageofthesevariable,non-zeropointsineachvelocitybin.The asterisksshowtheaverageifthenon-varyingabsorptionintervalsare included.Thetoppanelshowstheshort-termresults,whilethebottompanel plotslong-termdata.Wenotethatthemaximumvalueof | | / # $ is2 (Figure 2-7 ),whichindicatesabsorptionappearingordisappearing completely. 43

PAGE 44

Figure2-7. SameasFigure 2-6 ,exceptthefractionalchangeisnowplottedasa functionofaveragenormalizedabsorptionstrength.Inthisplot,individual quasarscancontributemultipletimestothemeasurementsatagiven absorptionstrength.Thesolidcurverepresentsthemaximumpossiblevalue of | | / # $ regionofFigure 2-7 ,butinsteadthevaluesof | | / # $ convergetowardsthemaximum possiblevalueof | | / # $ .Thedispersionofpointsduetoerrorsin | | / # $ alone cannotexplainthetrendinthisgure.Instead,thisgureindicatesthatweseeweaker absorbers,orweakerportionsofBALs,appearordisappearcompletely,butstrong featuresdonottypicallyappearordisappearoverthetime-scalesthatourdatacover. 44

PAGE 45

Beyondlookingattrendswithvelocityandabsorptionstrength,itisinstructiveto examinetherelationshipbetweenthesetwoparameters.Figure 2-2 displays # $ asa functionofvelocity,revealingthatthesequantitiesarecorrelated.Thelackofaclear trendinFigure 2-6 issurprisinggiventhatweakersystemsaremorevariable(Figure 2-7 )andtheytendtoappearmoreoftenathighervelocities(Figure 2-2 ).Evidently, thecorrelationof | | / # $ withabsorptionstrengthisstrongerthanthecorrelation withvelocity.Wealsoinvestigatehow | | varieswithvelocityandwith # $ .Wedonot includetheplotshere,butwendnotrendbetween | | andvelocityorabsorption strength. Wedonotndanysignicantdifferencesinthetrendsdescribedabovebetween theshort-termandlong-termdata.Themaindifferenceisintheincidenceofvariability, 39%intheshort-termversus65%inthelong-termdata.Thelong-termdataalsoshow slightlylargerchangesinabsorptionstrength,i.e., | | ,comparedtotheshort-term data.Theaveragevalueof | | is0.16 0.08intheshort-termand0.22 0.10inthe long-term. Wealsonoteagainthatweincludetwoquasars(0302+1705and0846+1540) fromtheLicksample( Barlow 1993 )thatarenottechnicallyBALs.However,these twoobjectsdonotsignicantlyaffecttheresults.Ifweomittedthem,thenthefraction ofquasarswithvariabilitywouldbe35%(6/17)intheshort-termand67%(14/21)in thelong-term.Inthecaseof0302+1705,itdoesnotqualifyasaBALQSObecause itsbroadabsorptionfeatureisatlowvelocities(centeredat %" 2100kms ).Italso has # $ reaching > 0.90.Itisastrongfeatureatlowvelocitythatdoesnotvary,tting thetrendsdiscussedabove.In0846+1540,thereisbroadabsorptionatlowvelocity (centeredat %" 2900kms ),withaverage # $ of % 0.27,andathighvelocity(centered at %" 26,800kms ),withaverage # $ of % 0.18.Thisalsotstheoveralltrendfor higherincidenceofvariabilityathighervelocities. 45

PAGE 46

2.4SummaryandDiscussion Wehaveanalyzedshort-termandlong-termC IV 1549BALvariabilityinasample of24quasars,andwelookedfortrendswiththeoutowvelocityandabsorption strength.Wefoundthat39%(7/18)ofthequasarsvariedintheshort-term,intervals of0.35to0.75yr,whereas65%(15/23)variedinthelong-term,intervalsof3.8to7.7yr. Overall,67%(16/24)ofthequasarsvaried.Wendatrendforvariabilitytooccurmore oftenathighervelocities(Figure 2-3 )andinshallowerabsorptiontroughs(orshallower portionsofabsorptiontroughs;Figures 2-4 and 2-5 ).Whenlookingatthefractional changeinstrengthofthevaryingabsorptionfeatures,thereisnoapparentsignicant correlationwithvelocity(Figure 2-6 ),butthereisatrendforalargerfractionalchangein absorptionstrengthinshallowerfeatures(Figure 2-7 ).Wedoincludetwoobjectswith BI=0,butthevariabilityinthebroadabsorptioninthesetwoquasarsisconsistentwith thetrendsdescribedabove. Gibsonetal. ( 2008 )studiedC IV BALvariabilityontime-scalesof t % 3-6yrsin 13quasarswithtwoepochsofdataandfoundthat92%(12/13)varied.Thisislarger thanouroverallpercentageof67%.Thisissurprisingbecausetheselectioncriteria adoptedby Barlow ( 1993 )forthesampleusedheremightbebiasedtowardmore variablesources(Section 2.1 ).Evidently,thisbiasissmallornonexistent.Thedifference betweenourresultand Gibsonetal. ( 2008 )mightbeduetothesmallnumbersof objectsstudiedand/orthemoreconservativeapproachwetookforidentifyingvariable absorption. Lundgrenetal. ( 2007 ),whichisastudyofBALvariabilityontime-scalesof < 1yr, ndaninversecorrelationoffractionalchangeinequivalentwidth(EW)withaverage EW.WhileEWisadifferentcalculationthanourmeasureofabsorptionstrength, # $ encapsulatingtheentireBALproleinonenumberinsteadofconsideringseparately thedifferentproleregions(Section 2.2.3 ),thisresultisconsistentwithourndingsthat fractionalchangein # $ correlateswith # $ Lundgrenetal. ( 2007 )alsondnooverall 46

PAGE 47

trendbetweenfractionalchangeinEWandoutowvelocity,whichisagainconsistent withourresults.Whilenotplottedhere,wedonotndasignicantcorrelationbetween | | andvelocityorabsorptionstrength,whichisconsistentwiththeanalysisof Gibson etal. ( 2008 ). Gibsonetal. ( 2008 )alsocalculatethechangeinabsorptionstrength,with valuesconcentratedintherange0.15-0.25.Wendanaveragevalueof | | forthe long-termdataof0.22 0.10. Both Gibsonetal. ( 2008 )and Lundgrenetal. ( 2007 )commentonthelackof evidenceforaccelerationintheBALfeaturestheystudy.Wealsondnoevidence foraccelerationinoursample.TheBALvariationsonlyinvolvetheabsorptiontrough growingdeeperorbecomingshallower.Wenote,however,thatlimitsontheacceleration aredifculttoquantifyforBALsbecause,unlikenarrowerabsorptionlines,theydonot usuallyhavesharpfeaturestoprovideanaccuratevelocityreference.BALsalsoexhibit prolevariability,whichcanshiftthelinecentroidwithoutnecessarilycorrespondingtoa realshiftinthevelocityofmaterialintheoutow. Inthiswork,wecomparetwodifferenttimeintervals,andwedonotndasignicant differencebetweentheshort-termandlong-termdatainthetrendsdescribedinSection 2.3 .However,theincidenceofvariabilityandthetypicalchangeinstrengthisgreater inthelong-termthanintheshort-term.ThisindicatesthatBALsgenerallyexperience agradualchangeinstrengthovermulti-yeartimescales,asopposedtomanyrapid changesontime-scalesoflessthanayear.Thiswillbeinvestigatedmoreinalater paper(Chapters 3 & 4 )wherewewillincludealltheepochsfromourdataset. Variabilityhasbeeninvestigatedinothertypesofabsorbersthatarepotentially relatedtoBALs. Narayananetal. ( 2004 )and Wiseetal. ( 2004 )lookedforvariability ( t % 0.3-6yrs)inlow-velocitynarrowabsorptionlines(NALs)andfoundthat25%(2/8) and21%(4/19),respectively,oftheNALstheystudiedvaried. Rodr guezHidalgoetal. ( 2010 )examinedvariabilityinmini-BALs,whichareabsorptionlineswithwidthslarger thanthoseofNALsandsmallerthanthoseofBALs.TheyobservedC IV mini-BALsin 47

PAGE 48

26quasarsontime-scalesof % 1to3yearsandfoundvariabilityin57%ofthesources. ThisnumberissimilartoourresultforBALs,howeverthiscomparisoniscomplicatedby adifferingnumberofepochsanddifferenttimebaselines. Rodr guezHidalgoetal. ( 2010 )foundnocorrelationbetweenincidenceofmini-BAL variabilityandabsorptionstrengthoroutowvelocity.However,thiscomparisonmight beskewedbythefactthatmini-BALstendtobeweakerthanBALs.Forexample,none ofthemini-BALsstudiedby Rodr guezHidalgoetal. ( 2010 )havedepths > ,while someoftheBALfeaturesinoursamplehave % .Therearealsodifferencesin thevelocitydistributions.Themini-BALdistributionin Rodr guezHidalgoetal. ( 2010 ) peaksat %" 20,000kms ,whereasthedistributionofBALabsorptioninourdata peaksat 15,000to 12,000kms (Figure 2-3 ).Moreworkisneeded,e.g.,with largersamples,tocontrolforthesedifferencesintheabsorptioncharacteristicsand compareBALandmini-BALvariabilitiesonequalfooting.Comparisonslikethatcould providevaluableconstraintsonthephysicalsimilaritiesandrelationshipbetweenBAL andmini-BALoutows. TwopossibilitiesforthecauseofBALvariabilityarethemovementofgasacross ourlineofsightandchangesinionization.Wefoundatrendforlinesathighervelocity tovarymoreoften,whichsuggestsmovementsofcloudssincefastermovinggasmight varymore.Moreover,thisisaglobaltrend,dependentonvelocityinanabsolutesense, ratherthanrelativevelocitywithinagiventrough.However,itispossiblethattheroot causeofthevariabilityisassociatedwiththestrengthofthelines,andweakerlines happentoappearathighervelocities(Figure 2-2 and Koristaetal. 1993 ).Ineithercase, fastermovingmaterialtendstohaveloweropticaldepthsorcoverlessofthecontinuum source.Measuredvaluesof # $ thusprovideeitheradirectmeasureofopticaldepth,or coveringfractioninthecaseofsaturatedabsorption.Withthecurrentresults,wecannot yetdisentanglethetrendsinvelocityand # $ withcertainty. 48

PAGE 49

EvidencefromBALvariabilitystudiesnonethelesssupportsaninterpretationof varyingcoveringfactor.Inadditiontogreatervariabilityathighervelocities,wefound changesoversmallportionsoftheBALtroughs,ratherthanentireBALsvarying,which canbeunderstoodintermsofmovementsofsub-structuresintheows.Changesin theionizinguxshouldgloballychangetheionizationintheow,andwewouldexpect thesechangestobeapparentoverlargeportionsoftheBALprolesandnotinjust somesmallvelocityrange. Hamannetal. ( 2008 )lookedatthevariationintheC IV and Si IV BALsinonequasarandfoundthattheBALswerelockedatroughlyaconstant Si IV /C IV strengthratio.Theyattributethisresulttothemovementsofcloudsthatare opticallythickinbothlines.Ifthelevelofionizationisconstantbetweenobservationsor if $ + 1,thenthechangesinabsorptionstrengthinoursamplearedirectlyindicativeof changesinthecoveringfraction.WewillexaminetheSi IV /C IV lineratioinourBALdata inasubsequentpaper. Ontheotherhand,studiesofNALssupportthepossibilityofchangesinionization causingthelinevariability. Misawaetal. ( 2007 )and Hamannetal. ( 2011 )observed quasarswheremultipleNALsvariedinconcert,whichsuggestsglobalchangesinthe ionizationoftheoutowinggas.Ifachangeinionizationiscausingthevariabilityin theBALoutows,thentheopticaldepth( $ )oftheBALsischanging.Featureswith loweropticaldeptharemuchmoresensitivetochangesinionizationthanthosewith higheropticaldepth.Theoutowinggasispresumablyphotoionizedbytheuxfrom thecontinuumsource,andchangesinthecontinuumuxcouldcauseachangein theionizationoftheoutowinggas.However,studiestodatehavenotfoundastrong correlationbetweencontinuumvariabilityandBALvariability( Barlowetal. 1992 ; Barlow 1993 ; Lundgrenetal. 2007 ; Gibsonetal. 2008 ),castingdoubtonionizationchanges causingBALvariability. Finally,weusethebolometricluminositiesofthequasarsinoursampletoderive characteristictime-scalesoftheowsforcomparisontotheobservedtime-scales. 49

PAGE 50

Thebolometricluminositiesforthequasarsinoursamplerangefrom % to ergss ,andtheaverageis % ergss .Thesevaluesarebased ontheobserveduxat1450 A(rest-frame)inabsoluteux-calibratedspectrafrom Barlow ( 1993 ),acosmologywith kms Mpc, ,anda standardbolometriccorrectionfactor ( .Basedontheaveragevalue of % ergss ,acharacteristicdiameterforthecontinuumregionat1550 Ais % pcandfortheC IV BELregionis % pc,assuming / and % ( Petersonetal. 2004 Bentzetal. 2007 Gaskell 2008 ,Hamann& Simon,inpreparation).IftheoutowinggasislocatedjustbeyondtheC IV BELregion andithasatransversespeedequaltotheKepleriandiskrotationspeed,thenthegas cloudwouldcrosstheentirecontinuumsourcein % 1yr.Thetypicaltime-scaleinour short-termdatais % 0.5yrandthetypicalchangeinabsorptionstrengthisatleast10%, whichnominallyrequiresthemovingcloudstocrossatleast10%ofthecontinuum source.Thisimpliestransversespeeds > 1000kms andthusradialdistancesthat areconservatively < 6pcifthetransversespeedsdonotexceedtheKeplerianspeed. Atypicaltime-scaleinthelong-termdatais % 6yr,whichisenoughtimetocrossthe continuumsource % 7times(ifthegasislocatedjustbeyondtheC IV BELregion).This suggestsafairlyhomogeneousowforatleastthecaseswherethereisnolong-term variability. Toplacebetterconstraintsontheoutowpropertiesandunderstandtheunderlying cause(s)ofthelinevariabilities,ourfollowingpapers(chapters)willmakemorecomplete useofourentiredatasample.WewillexaminetheratioofSi IV toC IV absorptionto gainbetterinsightintothecause(s)ofthevariability,andwewillutilizemorecomplete time-samplinginsteadofjustlookingattwospecicrangesoftimeintervals.Wehave alsoobtainedmoredataonBALvariabilityattheshortesttime-scales, t % 1week to1month,inordertodeterminewhetherthereisaminimumtime-scaleoverwhich 50

PAGE 51

variabilityoccurs.Thiswillprovideabetterconstraintonthelocationoftheoutowing gas. 51

PAGE 52

CHAPTER3 VARIABILITYINQUASARBROADABSORPTIONLINEOUTFLOWSII.MULTI-EPOCH MONITORINGOFSi IV ANDC IV BALVARIABILITY MostoftheexistingworkonBALvariabilityhasfocusedonvariabilityinC IV 1549 overtwoepochs(e.g., Barlow 1993 ; Lundgrenetal. 2007 ; Gibsonetal. 2008 ). Gibson etal. ( 2010 )reportsonvariabilityonmulti-monthtomulti-yearrest-frametime-scales, using3-4epochsofdatafor9BALQSOsandfoundthatBALsgenerallydonotvary monotonicallyovertime.TheirstudyalsomakescomparisonsbetweenvariabilityinSi IV 1400absorptionandvariabilityinC IV ,andtheirresultsincludeacorrelationbetween fractionalchangeinEWinSi IV andC IV Thischapter,whichiscurrentlyacceptedforpublication( Capellupoetal. 2012 ), isthesecondpartofour3-partpaperseriesonBALvariability.Chapter 2 introduced ourongoingmonitoringprogrammeofasampleof24BALquasars.Webeganwith asampleofBALquasarsfrom Barlow ( 1993 ),whichincludesspectraoftheC IV absorptionregionand,inmostcases,coverageoftheSi IV absorptionregionaswell. Wehavere-observedthesequasarstoprovidealongertimebaselineoverwhichto studyvariability,aswellastoobtainmultipleepochsofdataperobject.Wecurrently haveupto10epochsofdataperquasaruptoMarch2009,coveringrest-frametime intervals( t)from15daysto8.2yr. Inthischapter,IextendtheanalysisofChapter 2 bylookingatvariabilityinSi IV andcomparingittothevariabilityresultsforC IV .ExpandingourstudytoincludeSi IV absorptioncanhelpconstraintheoriesonthecause(s)ofBALvariability.CandSi havedifferentabundances,ifsolarabundancesareassumed,andtheyhavedifferent ionizationproperties(e.g. Hamannetal. 2008 2011 ).ByexaminingifC IV andSi IV havedifferentvariabilityproperties,andhowtheydiffer,coupledwiththesedifferencesin abundancesandionizationproperties,wecangainnewinsightintothecause(s)ofBAL variability. 52

PAGE 53

OurdatasetisuniquelysuitedtothisstudybecausewehavecoverageoftheSi IV linefornearlyourentiresample(22outof24quasars).Inadditiontothelargersample size,wegobeyondexistingworkbyadoptingamethodofmeasuringtheabsorption strengthinportionsofBALs,insteadofEW.Equivalentwidthmeasurementsapplyto anentirefeatureandarelesssensitivetochangesinsmallportionsoftroughs.Our methodofmeasuringportionsoftroughsalsoallowsmoredirectcomparisonsbetween thebehaviorofC IV andSi IV variability. WealsoincludetheentiredatasetsofartolookatvariabilityinC IV andSi IV over multipleepochs.Thisworkcontainsupto10epochsofdataperquasar,andincluding alloftheseepochswillprovidebetterinsightintothecharacteristicsofBALvariability. IncreasingthenumberofepochsprovidesnewinformationonwhetherBALschange monotonicallyovertimeorwhethertheycanvaryandthenreturntoanearlierstate. InChapter 2 ,wereportedthattypicallyonlyportionsofBALsvaried.Multi-epochdata cantelluswhethervariabilityonlyoccursinthosespecicvelocityintervalsorifthe velocityrangeoverwhichvariabilityoccurscanchangeovertime.Wealsohighlight severalindividualinterestingcasesofvariabilitythatcanfurtherhelpusunderstand BALoutows.Section 3.1 belowreviewsthequasarsampleandanalysisintroducedin Chapter 2 ,Section 3.2 describesourresults,Section 3.3 summarizestheresultssofar fromChapter 2 andthecurrentwork,andSection 3.4 discussestheresultsandtheir implications. 3.1DataandAnalysis 3.1.1ObservationsandQuasarSample Inthiswork,weusethesamesampleof24BALquasarsintroducedinChapter 2 .ThissampleisbasedonthesetofBALQSOsstudiedin Barlow ( 1993 ).Thesample selectionandgeneralcharacteristicsaredescribedinChapter 2 .Thesedatawere obtainedfromtheLickObservatory3-mShaneTelescope,usingtheKastspectrograph. Mostofthespectraweusefromthatdatasethavearesolutionof / ( 53

PAGE 54

Table3-1. QuasarData Name BILickSDSSMDMTotal t 1988-922000-062007-09(yrs) 0019+01072.130229060170.08-5.79 0043+00482.137433022150.35-6.13 0119+03102.090607020130.65-5.57 0146+01422.909578020240.52-5.15 0226 10242.256777010124.66 0302+17052.890020130.27-4.42 0842+34312.1504430613100.06-5.87 0846+15402.928050380.04-4.93 0903+17342.7711070021470.04-5.29 0932+50061.926792041490.05-6.98 0946+30091.221555020350.11-8.16 0957 05351.810267020240.11-6.21 1011+09062.268610040370.10-5.94 1232+13252.3641100010230.35-5.93 1246 05422.236481020240.40-5.90 1303+30481.770139010450.05-6.10 1309 05362.224469020240.68-6.19 1331 01081.8761040021250.42-5.97 1336+13352.445712010340.07-5.79 1413+11432.558681021250.26-5.61 1423+50002.252306021250.39-5.87 1435+50051.5871150020240.34-7.72 1524+51472.883181031370.04-5.14 2225 05341.981792030030.27-0.73 (230kms ).Forepochswherethisresolutionisnotavailable,weusedatatakenat ( 600(530kms ).BALsaredenedtohaveawidthofatleast2000kms ,so eitheroftheseresolutionsissufcienttomeasurethelinesandstudytheirvariabilities. ThewavelengthcoverageofeachspectrumcoversatleasttheSi IV throughC IV emissionlines,andmostcoveratleasttheLy "! 1216throughC IV emissionlines. Wehavebeenre-observing23oftheBALQSOsfrom Barlow ( 1993 )attheMDM Observatory2.4-mHiltnertelescope,usingtheCCDSspectrographwitharesolutionof ( 1200(250kms ).TheobservationsusedinthisworkweretakeninJanuaryand February2007;January,April,andMay2008;andJanuaryandMarch2009.Weused thesamespectrographsetupeachtime,varyingonlythewavelengthrangeinorder 54

PAGE 55

toobserveeachquasaratroughlythesamerestwavelengthrange,fromLy through C IV emission.Oneexceptionis0946+3009,whichhasaredshifttoolowfortheSi IV emissiontoappearinourspectra. WesupplementourdatawithspectrafromtheSDSSDataRelease6 ( Adelman-McCarthyetal. 2008 )for8ofthequasarsinoursample,forwhichthe resolutionis ( (150kms ).Thesespectracovertheobservedwavelength range3800to9200 A,andweonlyincludespectrathatcoveratleasttheSi IV through C IV emission. Table 3-1 summarizesthefulldatasetpresentedinthiswork,includingtheemission redshift, 1 andthebalnicityindex'(BI)foreachobject(ascalculatedinChapter 2 ). AnyuncertaintyintheredshiftwillnotaffectourcomparisonsbetweenSi IV andC IV oranyofourotherresults.Thebalnicityindex,denedby Weymannetal. ( 1991 ),isa measureofthestrengthoftheBALabsorptionandiscalculatedasanEWinunitsof velocity.Itquantiesblue-shiftedC IV absorptionbetween 25000and 3000kms thatreachesatleast10percentbelowthecontinuumacrossaregionatleast2000km s inwidth.Thenextfourcolumnslistthenumberofobservationstakenforeachobject ateachobservatoryandthenthetotaloverallnumberofobservations.Thenalcolumn liststherangein tcoveredforeachquasar. TwoofthequasarsinoursamplehaveBI=0,sotheyarenotBALquasarsbasedon thebalnicityindex.Theybothcontainbroadabsorption,butthisabsorptionfallsoutside thevelocityrange, 25000to 3000kms ,usedtodeneBI.AsnotedinChapter 2 anddiscussedfurtherinSection 3.2 below,includingthesetwoobjectsinoursample doesnotaffectanyofourmainresults. 1 Thevaluesof wereobtainedfromtheNASA/IPACExtragalacticDatabase (NED),whichisoperatedbytheJetPropulsionLaboratory,CaliforniaInstituteof Technology,undercontractwiththeNationalAeronauticsandSpaceAdministration. 55

PAGE 56

3.1.2MeasuringBALsandtheirVariability InFigure 3-1 ,weplotspectraforall24objects,showingthelong-termcomparisons ( t=3.8 7.7yr)betweenaLickspectrumandanMDMspectrum.Theoneexception is2225-0534,forwhichweonlyhaveshort-termLickdata.Foreachobject,weplotthe C IV absorptionregioninthetoppanelandthecorrespondingSi IV absorptionregionin thebottompanel.ThevelocityscaleisbasedonthewavelengthsofC IV andSi IV inthe observedframecalculatedfromtheredshiftsgiveninTable 3-1 .Inordertocomparethe C IV andSi IV absorptionregions,weusethebluerlineinboththeC IV andSi IV doublet forthezero-pointsofthevelocityscales,i.e.,1548.20 AforC IV and1393.76 AforSi IV WeadoptthevelocityrangesoverwhichC IV BALabsorptionoccursdenedin Chapter 2 .TheseregionsweredenedbasedonthedenitionofBI,i.e.theymust containcontiguousabsorptionthatreaches ) 10percentbelowthecontinuumacross ) 2000kms .WeapplythesamedenitionwhendeningthevelocityrangesofSi IV BALabsorption. InChapter 2 ,wedenedapseudo-continuumtfortheducialLickobservation usedinthelong-termanalysisforeachquasarbyttingapower-lawtoregionsof thespectrumfreeofabsorptionandemission.Thepreferredspectralregionsforthe tswere1270 1350 Aand1680 1800 A,butwereadjustedtoavoidemissionand absorptionfeaturesasmuchaspossibleorduetothelimitsofthewavelengthcoverage. WethenttheC IV emissionlines,usingbetween1and3Gaussianstodenethe lineprole.InChapter 2 ,wealsottheSi IV emissionincaseswhereC IV absorption overlapswiththeSi IV emission.Forthiswork,weadditionallyttheSi IV emissionin theducialLickobservationforallthequasars.Forthelong-termcomparisonsbelow, wettheSi IV emissionfortheMDMspectrumincaseswheretheemissionlinevaried. SomespecialcaseswherethereweredifcultieswithttingtheSi IV emission,suchas 1011+0906and1309 0536,arediscussedfurtherinSection 3.2.3 below. 56

PAGE 57

Figure3-1. Spectraofall24quasarsinoursample,showingthelong-termcomparisons ( t=3.8 7.7yr)betweenaLickObservatoryspectrum(boldcurves)anda recentMDMspectrum(thincurves).Foreachquasar,theC IV regionis displayedinthetoppanelandthecorrespondingSi IV regionisshowninthe bottompanel.For2225-0534,weonlyhaveshort-termLickdata(Table 3-1 ). TheverticaluxscaleappliestotheLickdata,andtheMDMspectrumhas beenscaledtomatchtheLickdatainthecontinuum.Thedashedcurves showourpseudo-continuumts.Thehorizontalbarsindicateintervalsof BALabsorptionincludedinthisstudy,andtheshadedregionsindicate intervalsofvariationwithintheBALs.Weusedbinomialsmoothingto improvethepresentationofthespectra.Theformal1 # errorsareshown nearthebottomofeachpanel.(Thetwovariabilityintervalsdenedfor 1524+5147werelabeledasoneintervalinChapter 2 .) 57

PAGE 58

Figure3-1. Continued. 58

PAGE 59

Figure3-1. Continued. 59

PAGE 60

Figure3-1. Continued. 60

PAGE 61

Whencomparingmultipleepochs,wescaledallthespectratotheducialLick spectrumusedforthepseudo-continuumt.Weonlytthepower-lawcontinuumto onespectrumforeachobject,soanyerrorsinthiscontinuumtwillnoteffectourmain variabilityresults.Tomatchtheindividualepochsforeachquasar,weadoptasimple verticalscalingthatmatchesthespectraalongthecontinuumredwardsoftheC IV emissionline(i.e.from1560 Atothelimitofthewavelengthcoverage),betweenthe Si IV andC IV emission( % 1425-1515 A),andbetweentheLy andtheSi IV emission ( % 1305-1315 A).Forthefewcaseswhereasimplescalingdidnotproduceagood matchandthereweredisparitiesintheoverallspectralshapebetweenthecomparison spectra,weteitheralinearfunction(for0903+1734,1413+1143,1423+5000,and 1524+5147)orquadraticfunction(for0019+0107and1309-0536)totheratioofthe twospectraacrossregionsthatavoidtheBALs.Wethenmultipliedthisfunctionbythe SDSSorMDMspectrumtomatchtheducialLickdata. Withthespectraforeachobjectmatched,weusedvisualinspectiontoidentify velocityintervalswithawidthofatleast1200kms thatvaried.Weidentifyintervals ofvariabilityseparatelyforC IV andSi IV absorption.Wethencalculatetheaverage uxandassociatederrorforeachcandidatevariableintervalineachofthetwoepochs beingcompared.Theerrorontheaverageuxisgivenby # # where # istheerroronanindividualpixel,takenfromtheerrorarraysasdisplayedin Figure 3-1 ,and isthenumberofpixelsinthecandidatevariableinterval.Wethen calculatetheuxdifferencebetweenthetwospectraandplaceanerroronthisux differenceusingtheerrorontheaverageuxfromeachepoch.Weincludeallintervals ofvariabilitywheretheuxdifferencesareatleast4 # .Anyintervalwhichvariedbyat least4 # wasreadilyidentiedbyourinitialinspectionprocedure. 61

PAGE 62

However,photonstatisticsalonearenotsufcientfordeningrealvariability,sowe tookaconservativeapproach,describedinmoredetailinChapter 2 ,wherebyweomit ambiguouscasesofvariability,eveniftheymeetthe4 # threshold.Fluxcalibrations,a poorlyconstrainedcontinuumplacement,andunderlyingemission-linevariabilitycan alladdadditionaluncertaintytoidentifyingvariability.Forexample,in1011+0906and 1309 0536,theremightbeBALabsorption,andvariability,ontopoftheSi IV emission line(Figure 3-1 ).However,itistooambiguoustobeincludedinthisstudy.SeeChapter 2 forfurtherexamplesofintervalsofpotentialvariabilitythatwerenotincludedbecause ofadditionaluncertaintiesandSection 3.2.3 below,wherewecommentfurtheron certainindividualquasars.Weincludenarrowintervalsofvariabilitysuchastheshaded regioninSi IV in0043+0048andtheshadedregioninC IV in1246 0542inFigure 3-1 ,wheretheuxdifferencesare6.3 # and5.6 # ,respectively.Theseregionsmeetthe aforementionedthresholds,andthecomparisonspectramatchwellinregionsofthe continuumfreeofemissionandabsorptionandoneithersideofthevariabilityinterval. Wealsoincluderegionssuchasthoseshadedin1011+0906inFigure 3-1 because eventhoughtheerrorsareslightlyhigherintheMDMspectrashown,theuxdifferences arestill7.7 # forthevariableregioninC IV and6.3 # and7.9 # forthetworegionsof variabilityinSi IV .Overallourapproachisdesignedtobeconservative;wetrytoexclude marginalcasesofvariabilitytoavoidoverestimatingthetruevariabilityfractions. Wecalculatedtheabsorptionstrength, ,oftheBALsinoursample,where isthe fractionofthenormalizedcontinuumuxremovedbyabsorption(0 & A & 1)withina speciedvelocityinterval.ThesecalculationsaredescribedinChapter 2 .Verybriey, wedivideeachintervalofvariabilityandabsorption,asdenedabove,intoequal-sized binsofwidth1000to2000kms ,withthenalbinsizedependingonthetotalvelocity widthofthespeciedinterval.Then,foreachquasar,weadoptthesamebinsizeforthe epochsbeingcomparedandcalculate # $ and ineveryindividualbin. 62

PAGE 63

Figure3-2. Theaveragenormalizedabsorptionstrength, # $ ,inC IV versusthe # $ in Si IV foreachabsorptionbinineachquasarinthelong-termsubsample.The errorbarsrepresentthe1 # errorsbasedonphotonstatistics. OneofthedifcultiesindirectlycomparingC IV toSi IV absorptionisthewider doubletseparationinSi IV (500kms inC IV versus1900kms inSi IV ).Thiscan causetheSi IV absorptionintervalstobewiderthanthoseinC IV .Theonlyeffectthis shouldhaveonthevariabilityresultsinSection 3.2 belowisthattheremaybeportions ofSi IV absorptionthataredetectedasvariablebutthecorrespondingvelocityintervals inC IV maybetoonarrowtopassourvariabilitythreshold.Wecommentfurtheronthis inSection 3.2.1 .WemarktheregionsdenedasBALabsorption(horizontalbars)and variability(shadedrectangles)inFigure 3-1 .Theabsorptionandvariabilityregionsfor C IV weredenedinChapter 2 .Wedenedtheabsorptionandvariabilityregionsfor Si IV independentlyfromwhatwefoundforC IV .SeeSection 3.2.1 forafulldiscussion oftheseresults. 63

PAGE 64

InFigure 3-2 ,weplottherelationshipbetweenabsorptionstrength, # $ ,inC IV and # $ inSi IV forthelong-termspectrashowninFigure 3-1 .Eachpointrepresentsa differentabsorptionbininanindividualquasar,suchthatindividualquasarscontribute multiplepointstothisplot.Thediagonallinethroughtheplotrepresentsequalstrength inbothlines.SincewexthevelocityscalebasedonthebluerdoubletmemberinC IV andinSi IV andthedoubletseparationiswiderinSi IV ,theedgesoftheSi IV absorption troughstendtoextendredwardoftheedgesofthecorrespondingC IV troughs(see,for example,0302+1705inFigure 3-1 ).Asmentionedabove,tocalculatethe # $ values, wedivideeachBALintobins.Wethusremovetheredmostbinforeachabsorption troughbecauseSi IV mayhaveagreaterstrengththanC IV inthosebinsduetothe widerdoubletseparationandnotnecessarilybecausetheSi IV absorptionisactually strongerthanC IV .Wealsoplot1 # errorbars,calculatedbyusingtheerrorspectra showninFigure 3-1 andaveragingoverthevelocityintervalforeachbin.Wepointout thatnon-statisticalerrors,e.g.,fromthecontinuumtting,canincreasetheerrorin # $ measurementsbyupto0.05to0.1.ItisclearfromFigure 3-2 thatabsorptioninC IV is roughlyasstrongorstrongerthanthecorrespondingSi IV absorption.Insomecases, theC IV hasnodetectablecorrespondingSi IV absorptionatall. Therearealmostnowell-measuredcasesofC IV absorptionweakerthanSi IV .We didndafewabsorptionbinswhereSi IV appearedtobestrongerthanC IV ,butsome ofthesebinsmightnotactuallyhaveabsorptionduetoSi IV ,orduesolelytoSi IV .In about10%ofBALspectra,therearelowerionizationlines,suchasC II 1335andAl III 1855,1863( Trumpetal. 2006 ).ThewavelengthofC II placesanyC II absorptionline atavelocityof < 13500kms inSi IV velocityspace.Inordertolookforinterloping C II lines,wecheckedallofourobjectsforanotherlowionizationspecies,Al III ,which existsredwardofC IV ,inaregionofthespectrumrelativelyuncontaminatedbyother lines.Wefoundthatintwocases(1232+1325,1331 0108),thevelocityoftheAl III line putsthecorrespondingC II linewithinaSi IV BALtrough.Wethereforeremovedthebins 64

PAGE 65

affectedbyC II fromFigure 3-2 ,andweexcludethecontaminatedSi IV regionsinthese twoquasarsfromtheanalysisbelow.Therearestillafewpointsbelowtheone-to-one lineinFigure 3-2 .However,thesefewpointsaremostlywithin3 # oftheone-to-oneline andthereforeareconsistentwithequalstrengthchangesinC IV andSi IV .Theonepoint thatisjustbeyond3 # fromthelinecorrespondstotheinterval 3100to 1200kms in1303+3048. 3.2Results 3.2.1VariabilityinSi IV versusC IV BALs Inthissection,wedirectlycomparethevariabilityofSi IV toC IV inthe"long-term" datasetfromChapter 2 .Thisinvolves2epochsofdatafor23quasarsseparatedby 3.8to7.7yrs.OurmaingoalistodiscriminatebetweenthepossiblecausesofBAL variability.WebeginbylookingatwhatfractionsofquasarsexhibitedC IV andSi IV variabilitytodetermineifSi IV variesmoreorlessoftenthanC IV .InChapter 2 ,we foundacorrelationbetweentheincidenceofC IV BALvariabilityandoutowvelocity andabsorptionstrength.HereweinvestigatewhethersimilartrendsexistsforSi IV BALs.WethenlookattherelationshipbetweentheincidenceofC IV variabilityandSi IV strength.Last,wecomparethechangeinstrengthforthetwolineswhentheybothvary. Inthelong-termdatasetinChapter 2 ,15outof23quasars(65percent)exhibited C IV BALvariabilityand11outof19(58percent)exhibitedSi IV BALvariability,atany measuredvelocity.WedonothavedatacoveringtheSi IV regionfor2ofourquasars, andanother2quasarsdonothaveSi IV BALs.ThiscomparisonbetweenC IV and Si IV iscomplicatedbecausetheabsorptioninC IV isnotalwaysaccompaniedby correspondingabsorption(e.g.,atthesamevelocities)inSi IV (Figures 3-1 and 3-2 ). Inaddition,weareobservationallylesssensitivetoabsorptionandvariabilityathigh velocitiesinSi IV ,comparedtoC IV ,becausethosewavelengthscanhavepoorer signal-to-noiseratiosandlargeruncertaintiesinthecontinuumplacementcaused 65

PAGE 66

byblendswithunderlyingbroademissionlines.Altogetherthismeanswearemore sensitivetovariabilityinC IV thanSi IV inourdataset. Tomakeafaircomparisonbetweentheincidenceofvariabilityinthetwolines,we recalculatetheabovefractionswhileconsideringjusttheowspeedsat > 20000km s .Weadoptthisvelocityasthecutoffbecause,insomespectra,thereisemission duetoO I at %" 20500kms intheSi IV absorptionregion(seealso Gibsonetal. 2010 ).Withthisadditionalrestriction,wendthatSi IV ismorelikelytovarythanC IV .In particular,35percent(8/23)ofquasarsexhibitedC IV variabilityand47percent(9/19) exhibitedSi IV variability.ThedramaticreductionintheC IV variabilityrecordedthis way,comparedtothe65percentquotedabove,isduetoi)considerationofanarrower velocityrange,andii)thespecicexclusionofhighvelocities,whicharethemostlikely toshowvariability(Chapter 2 ).NearlyhalfoftheoccurrencesofC IV variabilitydetected inourdatasetareathighvelocities,i.e.,v < 20000km/s.TheincidenceofC IV variabilityfurtherreducesto31percent(6/19)ifweonlyincludethe19quasarswhich havecompletespectralcoverageacrossSi IV andhaveaSi IV BAL.Thefurtherdecline intheC IV variabilityinthiscaseprobablyoccursbecausethetwoquasarsexcludedfor havingnoSi IV BALhaveweakC IV lines,andweakC IV linesaremorelikelytovary thanstrongones(Chapter 2 ).Thus,weagainremovedC IV BALsthataremorelikelyto vary. Overall,itisimportanttorealizethatanumberoffactorscanaffectthemeasured incidenceofBALvariability.Ourcomparisonsshowthat,overmatchingvelocityranges, Si IV BALshaveasignicantlyhigherincidenceofvariabilitythanC IV BALs.This differenceisprobablyrelatedtothedifferentlinestrengths.Inparticular,theSi IV BALs aregenerallyweakerthanC IV BALs(Figure 3-2 ),andweakerlinestendtobemore variable(Chapter 2 andFigures 3-4 and 3-5 below). TomoredirectlycompareSi IV andC IV BALvariability,weexaminetheindividual velocityintervalsoverwhichthevariabilityoccurs.Weconsiderintervalsatallvelocities 66

PAGE 67

hereandintheremainderofthissection.VariabilityinC IV occurredinatotalof20 velocityintervals.TenoftheseintervalshavemeasurableSi IV absorption.Nineofthese 10intervals,or90percent,showedSi IV variationsinthesamesense(eithergetting strongerorweaker)astheC IV changes.Thereisonlyoneinterval(in0119+0310)that exhibitsvariabilityinC IV ,withoutcorrespondingvariabilityatthesamevelocitiesinSi IV Conversely,wendlong-termSi IV BALvariabilityinatotalof22velocityintervals.All ofthevariableSi IV intervalshavesignicantcorrespondingC IV absorption,and10of theseintervalsshowedC IV variationsinthesamesenseastheSi IV (45percent). AsmentionedinSection 3.1.2 ,Si IV hasawiderdoubletseparationthanC IV causingsomeoftheSi IV absorptionandvariabilityintervalstobewiderthanthe correspondingC IV intervals.FortheintervalsofSi IV variabilitywithoutcorresponding C IV variability,welookedforanyevidenceofvariabilityinC IV thatwasnotincluded becausethewidthofthecandidatevaryingregionwastoonarrowtomeetourvariability threshold(Section 3.1.2 ).Thereisonlyonecasewherewedetectamarginalnarrow variabilityregioninC IV correspondingtoaregioninSi IV classiedasvariable(in 0903 1734).EvenifweweretocountthisasavariableC IV interval,stillonly50per centofSi IV variabilityintervalswouldhavecorrespondingC IV variability.Therefore, while91percentoftheintervalsofC IV variabilityhadcorrespondingSi IV variability, only % 50percentoftheintervalsofSi IV variabilityhadcorrespondingC IV variability. Theseresultsreinforceourmainconclusionabove,thatSi IV BALsaremorelikelyto varythantheirC IV counterparts. NextweexaminethedependenceofSi IV variabilityonvelocityandabsorption strength,matchingouranalysisofC IV BALsinChapter 2 .Figure 3-3 showsthe incidenceofSi IV absorptionandSi IV variabilityversusvelocity.Forcomparison,this gurealsoshowsthecorrespondingdataforC IV takendirectlyfromFigure 2-3 (dashed curves).ThetoptwopanelsdisplaythenumberofquasarswithSi IV BALabsorption andwithSi IV BALvariabilityateachvelocity(solidcurves).Thethirdpanelisthe 67

PAGE 68

Figure3-3. ThetoptwopanelsshowthenumberofoccurrencesofSi IV (solidlines)and C IV (dashedlines)BALabsorptionandvariableabsorptionversusvelocity. Thethirdpanelisthesecondpaneldividedbytherst.Thebottompanel showsthesamecurvefromthethirdpanelforSi IV with1 # errorbars. 68

PAGE 69

secondpaneldividedbythetopone,whichgivesthefractionofSi IV BALsthatvaried ateachvelocity.ThetoppanelshowsclearlythattherearemoreC IV BALsthanSi IV BALsathighervelocities. InChapter 2 ,weshowedthattheincidenceofC IV variabilityincreasessignicantly withincreasingvelocity.ThistrendisnotevidentintheSi IV data.Figure 3-3 displays 1 # errorbarsbasedon Wilson ( 1927 )and Agresti&Coull ( 1998 ).Theseerrorsare basedoncountingstatisticsforthenumberofquasarswithabsorptionandvariability ateachvelocity.Weperformedaleast-squarest,andtheslopeoftheSi IV datais ,intheformalunit,fractionperkms .Theslopeisnon-zeroatjust a1.5 # signicance.Therefore,whilethere might beaweaktendencyformorevariability inSi IV athighervelocities(Figure 3-3 ),thetrendisnotstatisticallysignicant. TofurthermatchouranalysisfromChapter 2 ,welookedattherelationshipbetween theincidenceofSi IV variabilityandtheabsorptionstrength, # $ ,inSi IV .Asdescribed inChapter 2 (andSection 3.1 above),wedivideeachBALinto % 1200kms bins, andwetreateachbinasanindividualoccurrenceofabsorption.InFigure 3-4 ,thetop twopanelsshowthenumberoftheseoccurrencesofSi IV absorptionandthenumber oftheseoccurrencesthatvariedateachvalueofSi IV absorptionstrength, # $ .An individualquasarcancontributemorethanoncetoeachpointinthehistogram.The thirdpanelisthesecondpaneldividedbythetopone,andthebottompanelshowsthe samecurvefromthethirdpanelwitherrorbarsplotted,calculatedinthesamewayas forFigure 3-3 .WendonlyaweaktrendbetweenSi IV variabilityandSi IV strength. Theslopeoftheplotis fractionperunitabsorptionstrength,which isnon-zeroata3.5 # signicance.Thisismuchweakerthanthetrendbetweenthe incidenceofC IV variabilityandC IV strengthfoundinChapter 2 .Weoverplotthiscurve forC IV fromFigure 2-5 inthethirdpanelofFigure 3-4 (dashedcurve).Thisindicates thattheoccurrenceofvariabilityinSi IV islesssensitivetothestrengthofthelinethan theoccurrenceofvariabilityinC IV 69

PAGE 70

Figure3-4. ThetoptwopanelsshowthenumberofoccurrencesofSi IV BALabsorption andvariabilityversustheaveragenormalizedabsorptionstrength, # $ ,in Si IV .Thethirdpanelisthesecondpaneldividedbytherst.Thebottom panelshowsthesamecurvefromthethirdpanelwith1 # errorbars overplotted. 70

PAGE 71

Figure3-5. ThetoptwopanelsshowthenumberofoccurrencesofC IV BALabsorption andvariabilityversustheaveragenormalizedabsorptionstrength, # $ ,atthe samevelocitiesinSi IV .Thethirdpanelisthesecondpaneldividedbythe rst,andthebottompanelshowsthesamecurvefromthethirdpanelwith 1 # errorbars.Inthethirdpanel,weoverplotthefractionofoccurrencesof C IV absorptionthatvariedversus # $ inC IV 71

PAGE 72

Next,welookedattherelationshipbetweentheincidenceofC IV BALvariabilityand thestrengthoftheSi IV absorption, # $ ,atthesamevelocities.Siisknowntobeless abundantthanC,whensolarabundancesareassumed,andthehighionizationtypical inBALsfavorsC IV (Section 3.4 ).Therefore,theopticaldepthinC IV ishigherthanin Si IV ,sothestrongertheSi IV absorptionis,themorelikelyC IV istobesaturated.In Figure 3-5 ,weplotthenumberofoccurrencesofC IV absorptionthatoccuratthesame velocitiesinthesamespectraaseachSi IV # $ valueandthenthenumberofthese occurrencesthatvariedinthesecondpanel.AsinFigure 3-4 ,anindividualquasarcan contributemorethanoncetoeachpointinthehistogram.Thethirdpanelisthesecond paneldividedbythetopone,showingthefractionofoccurrencesofC IV absorptionthat variedateachSi IV absorptionstrengthvalue.AsinFigures 3-3 and 3-4 ,thebottom panelofFigure 3-5 showsthe1 # errorbars.Theslopeofthesepointsis fractionperunitabsorptionstrength,whichisnon-zeroata10 # signicance.Figure 3-5 thusindicatesthattheincidenceofC IV variabilitydecreaseswithincreasingSi IV absorptionstrength.Therefore,whentheSi IV absorptionisstrongerandtheC IV absorptionismorelikelytobesaturated,theincidenceofC IV variabilitydecreases.In fact,whenevertheSi IV absorptionstrengthisgreaterthan0.5,thecorrespondingC IV absorptionatthesamevelocitydoesnotvary. AsafurthercomparisontoChapter 2 ,weoverplotinthethirdpanelofFigure 3-5 thefractionofoccurrencesofC IV absorptionthatvariedversusC IV absorptionstrength (dashedcurve).InChapter 2 ,weconcludedthat,forC IV BALs,weakerlinesaremore likelytovarythanstrongerlines.Figure 3-5 showsthatC IV linesareevenmorelikelyto varywhenthecorrespondingSi IV linesarealsoweak. Wealsoinvestigatethechangeinstrength, | | ,inC IV comparedtothe correspondingchangesinSi IV inthesamevelocityinterval(Figure 3-6 ).Thereare 6quasarsinwhichthevelocityintervalsofC IV andSi IV variabilityoverlap,andweonly comparethevelocityintervalswhere both linesvaried.Eachpointrepresentsonebinin 72

PAGE 73

Figure3-6. ThechangeinstrengthofC IV BALabsorptionversusthechangeinstrength ofSi IV BALabsorptioninvelocityintervalswherebothlinesvaried.The errorbarsarecalculatedasinFigure 3-2 oneofthesequasars,asdescribedaboveforFigure 3-5 .Thediagonallinethroughthis plotcorrespondstoequalstrengthchangesinbothlines.Thereisclearlynocorrelation betweenthestrengthchangesinC IV versusSi IV evidentinthisgure.DespiteSi IV varyingmoreoftenthanC IV ,thestrengthchangesinSi IV arenotalwaysgreaterthanin C IV Finally,Figure 3-7 showsthefractionalchangeinstrength, | | / # $ ,inC IV comparedto | | / # $ inSi IV ,againforvelocityintervalswherebothlinesvaried.Asin Figure 3-6 ,therearepointslyingabovetheline,representinggreaterstrengthchanges inC IV thaninSi IV .However,thereisaweaktrendtowardsgreaterfractionalchangein strengthinSi IV ,whichisconsistentwiththecorrelationfoundbetweenfractionalchange inEWinC IV andSi IV in Gibsonetal. ( 2010 ). 73

PAGE 74

Figure3-7. ThefractionalchangeinstrengthofC IV BALabsorptionversusthe fractionalchangeinstrengthofSi IV BALabsorptioninvelocityintervals wherebothlinesvaried.TheerrorbarsarecalculatedasinFigure 3-2 AsmentionedinSection 3.1 ,weincludetwoquasarsthathaveBI=0becausethey dohavebroadabsorption,butitfallsoutsidethevelocityrange, 25000to 3000km s ,usedinthestrictdenitionofBI.Wealsoincludebroadabsorptioninotherquasars inoursamplethatfallsoutsideofthisvelocityrange.Theirinclusionhasminimalimpact onourresultsbecausemostoftheSi IV broadabsorptioninoursamplefallswithinthis velocityrange.ForthequasarswithBI=0,0846+1540doesnotcontainanySi IV broad absorptionatall,and0302+1705containsbroadabsorptionatlow-velocity,whichdid notvaryineitherC IV orSi IV WecansummarizeourcomparisonsbetweentheC IV andSi IV BALvariabilitiesas follows:1)Si IV BALsaremorelikelytovarythanC IV BALs.Thefractionsofquasars showingvariabilityinourlong-term2-epochdatasetare31%inC IV and47%inSi IV if weconsideronlythewell-measuredvelocityrange > 20000km/sandincludeonly 74

PAGE 75

quasarswithbothSi IV andC IV BALsdetected.2)Thevariabilitiesusuallyoccurinjust portionsoftheBALtroughs.3)WhenchangesoccurinbothSi IV andC IV ,theyalways occurinthesamesense(i.e.,withbothlinesgettingeitherweakerorstronger).They alsooccurinoverlappingbutnotnecessarilyidenticalvelocityranges.4)Thetrendfora higherincidenceofC IV variabilityathighervelocities,whichwereportedinChapter 2 ,is notclearlyevidentintheSi IV data.Finally,5)thereisnocorrelationbetweenabsorption strengthchangesinC IV versusSi IV whentheybothvary;although,thereisaweak trendtowardsgreaterfractionalchangeinstrengthinSi IV 3.2.2Multi-EpochMonitoringofBALQSOs Wenowexpandouranalysistothefulldataset,whichincludes2to10epochs ofdataforeachobject(Table 3-1 ).Includingalloftheseepochsandconsideringall measuredvelocities,thefractionofquasarsthatshowedC IV BALvariabilityis83per cent.Thisisasignicantincreasefromthe65percentwederivedconsideringonlytwo long-termepochs,orthe39percentderivedfromonlytwoshort-termepochs(Section 3.2.1 andChapter 2 ).Clearly,includingmoreepochsofdataincreasestheobserved variabilityfractions.Moreover,theselargervariabilityfractionsapplytoroughlythesame time-frameasour2-epochlong-termdataset.Therefore,themulti-epochdatadidnot ndnewoccurrencesofvariabilityatsomeothertime;theyidentiedvariabilitymissed bythe2-epochmeasurements. Toinvestigatethemulti-epochbehavioursofBALvariability,wecomparedall thespectraobtainedforeachquasar.Fromoneobjecttoanother,therearelarge differencesinthewidthsofthevaryingregionsandtheamplitudesofthechanges(e.g., Figure 3-1 and 3-6 ).However,therearecertaingeneraltrendsthatmost,ifnotall,the quasarsfollow.Inparticular,thevariabilityalmostalwaysoccurredwithinjustaportion ofaBALandnotintheentiretrough.Innearlyallthequasars,thevariabilityoccurred overthesamevelocityinterval(s)betweeneachepoch.Finally,whentherearemultiple velocityintervalsofvariabilitywithinthesamequasar,thechangesintheseseparate 75

PAGE 76

intervalsalmostalwaysoccurinthesamesense.Similarly,asdescribedinSection 3.2.1 ,whenthereisvariabilityinbothC IV andSi IV ,theyalsovaryinthesamesense. Wealsondnoclearevidenceforvelocityshiftsthatwouldbeindicativeof accelerationordecelerationintheows.Theconstraintsonvelocityshiftsaredifcultto quantifyinBALsbecausetherecanbecomplexprolevariabilities,butwespecically searchforanddidnotndcaseswhereadistinctabsorptionfeaturepreservedits identitywhileshiftinginvelocity.Despitethelargeoutowvelocities,thereisnoclear evidencetodateforaccelerationordecelerationinBALs,orinanyotheroutowlines (i.e.NALsandmini-BALs;e.g. Rodr guezHidalgoetal. 2011 ). Wehighlightbelowafewwell-sampledcasestoillustratethesegeneraltrendsinthe data.Figure 3-8 showsspecicallythequasarsforwhichwehaveatleast7measured epochsincludingonefromtheSDSS,whichhelpstospanthetimegapbetweenthe earlyLickdataandourrecentMDMobservations.Foreachobject,thetoppanelshows theC IV BAL(s)andthebottompanelshowstheSi IV BAL(s).Thebluecurvesshow theearlyLickdata,theredcurvesshowtheintermediateSDSSdata,andthegreen curvesshowtheMDMspectra.Wenotethatin1524+5147thereisanO I emissionline centeredat %" 20500kms intheSi IV panel. TheboldbarsinFigure 3-8 markintervalsofvariabilityidentiedforC IV .The velocityrangesareguidedbytheintervalsdenedforC IV inChapter 2 andareadjusted tocoverthecoreofthevaryingregionandavoidtheedgeswherethevariabilityisless pronounced.Thesesamevelocityrangesareappliedtoalltheepochsplottedhereand toSi IV forcomparison.Theoneexceptionisthethinbarmarkingaregionofvariability inSi IV in0842+3431withminimalcorrespondingvariabilityinC IV .Forallofthese varyingregions,wecalculatetheabsorptionstrength, ,forthedenedvelocityintervals ineachepochandthenplotthese valuesversustimeinFigure 3-9 Weplot insteadofEW,inFigure 3-9 ,inordertohighlighttheintervalsthatvaried. UsingEWwoulddilutethesechangesinstrength.Thedifferentcolorscorrespondto 76

PAGE 77

Figure3-8. SpectraoftheC IV (toppanel)andSi IV (bottompanel)BALsinfour well-sampledquasarsfromoursample,aftersmoothingthreetimeswitha binomialfunction.ThebluecurvesaretheLickspectra,redcurvesare SDSSspectra,andgreencurvesareMDMspectra.Theboldbarsmark varyingregionsidentiedforC IV .Theaverageerrorforeachspectrumis showninthetoppanelforeachquasar,wheretheheightoftheerrorbar represents # 77

PAGE 78

differentvelocityintervals.ThedashedanddottedlinesrepresentchangesinC IV and Si IV ,respectively.Wenotethattheselinesdonotrepresenthowthe valueschanged betweenepochs.Theysimplyconnectthemeasurementsfromdifferentepochsinorder toaidetheeye. InFigure 3-8 ,thespectraof0842+3431(intheC IV BAL),0903+1734,and 1524+5147showclearlyhowportionsofBALscanvary.In0842+3431,thereis signicantvariabilityinboththebluesideandtheredsideoftheC IV BAL;although inSi IV ,theentireBALvaries(seeSection 3.2.3.3 below).In0903+1734,thereis signicantvariabilityathigheroutowvelocities,butatlowervelocities,thereisno variability.ThisisconsistentwiththeresultfromChapter 2 thatthereisahigher incidenceofvariabilityathighervelocities.Thevariableregionsin1524+5147cover most,butnotall,oftheBAL.IncontrasttothegeneraltrendsinChapter 2 ,the blue-mostportionoftheBALdidnotvary.In0932+5006,however,theentireC IV BALsvaryatthehighestvelocities.IntermsofC IV toSi IV comparisons,1524+5147 isacasewherethereisweakcorrespondingabsorption,andvariability,inSi IV ,butno Si IV BAL.And,0932+5006showsclearlyacasewhereaSi IV BALvaried,buttheC IV BALdidnot. TheplotsinFigure 3-9 for0842+3431,0903+1734,and0932+5006showhow thechangein isnotalwaysmonotonic.ThesameBALcangrowdeeperfromone epochtoanother,thenbecomeshalloweragain.Thisisconsistentwiththeresultsof Gibsonetal. ( 2010 ).Furthermore,thechangeinthe valuefromoneepochtoanother generallyoccursinthesamedirection(eitherpositiveornegative)inbothC IV and Si IV ,whichisconsistentwiththeresultsofSection 3.2.1 .In0842+3431,0903+1734, and1524+5147,wheretherearetwoseparateintervalsofC IV variability,thechange instrengthoccursinthesamedirectionforbothintervals.Thehigh-velocityBALsin 0932+5006varyinconcertstartingwiththe1989.84epochthrough2008.03.Atthe earliestandlatestepochs,theydonotclearlyvaryinconcert,butthechangesin 78

PAGE 79

Figure3-9. TheabsorptionstrengthinbothC IV (dashedlines)andSi IV (dottedlines) asafunctionoftimeinthevelocityintervalsindicatedbybarsinFigure 3-8 aresmallandcouldbeaffectedbychangesintheunderlyingSi IV emissionline.We commentfurtherontheseBALsin0932+5006inSection 3.2.3.4 3.2.3NotesonIndividualQuasars Inthissection,wecommentonindividualquasarsthatarecasesofspecialscientic interest.Wealsocommentoncaseswheretherewerespecicissuesintheanalysisor measurementsthatresultinlargeruncertainties. 79

PAGE 80

3.2.3.10119+0310 0119+0310istheonlyquasarforwhichwerecordC IV BALvariationswithout correspondingchangesinSi IV inourlong-termsample(Figure 3-1 andSection 3.2.1 ).However,theseresultsareverytentativebecausetheSi IV absorptionispoorly measuredacrossthevelocitiesthatvariedinC IV .Thetwolong-termspectraforthis object,plottedinFigure 3-1 ,havealowersignal-to-noiselevelthanmostoftheother datainoursample.Furthermore,theSi IV absorptionatthevelocityofC IV variability ( %" 7500kms )isveryweak,andifthecontinuumtisoffbyeven % 5percent,this regioninSi IV mightnotbeconsideredpartoftheSi IV BAL.Ifthisregionisnotpart oftheBAL,thenwewouldnotincludeitinthecomparisonofC IV toSi IV .Therefore, whilewehaveseveralwell-measuredcasesofSi IV variabilitywithoutcorresponding C IV variability,weonlyhavethisonepoorly-measuredcaseofC IV variabilitywithno correspondingSi IV variability. Thisquasaralsoappearstodifferfrommostoftheotherquasarsinthesamplein thatthedifferentregionsofC IV variabilitydonotvaryinthesamesense(Figure 3-1 ). ThetwohighervelocityvariableregionsbothincreaseinstrengthbetweentheLickand MDMobservations,whilethelowervelocityvariableregiondecreasesinstrength.As mentionedabove,thisisoneofourleastwell-measuredquasars,sothisisatentative result.WendjusttwoothercaseswheretworegionsofC IV variabilityvaryinopposite directions(0146+0142and1423+5000). 3.2.3.20146+0142 Werstnotethatthisobjecthasahigh-velocityC IV BAL,sotheBALthatappears justredwardofthemarkedSi IV BALinFigure 3-1 isactuallyC IV .Asmentionedin Chapter 2 ,wecanconrmthatthisishigh-velocityC IV ,andnotSi IV ,becauseifit wereSi IV absorption,weshouldseecorrespondinglow-velocityC IV absorption(see alsogure1in Koristaetal. 1993 ).Wealsohavefurtherconrmationthatthisis high-velocityC IV absorptionbecausewendevidenceofcorrespondinghigh-velocity 80

PAGE 81

Figure3-10. Thetwolong-termepochsfor0146+0142,showingthefullspectrumfrom theLy emissionthroughtheC IV emission.Theshadingherediffersfrom Figure 3-1 ,withtheright-mostshadedregionsmarkingtheC IV variability andtheleft-mostshadedregionsshowingthecorrespondingvelocitiesin Si IV (butnottheexactregionsofSi IV variability).Thisgureshows evidenceofSi IV BALvariabilityontopoftheLy emissionline. Si IV absorptionontopoftheLy emissionline.Figure 3-10 showsthetwolong-term spectrafor0146+0142withthetworight-mostshadedregionsmarkingtheC IV variabilityandthetwoleft-mostshadedregionsmarkingthecorrespondingvelocities (butnotnecessarilytheentirevariableregions)inSi IV Anotherinterestingnoteabout0146+0142isthatinourshort-termdata,thereare twoseparateregionsofC IV BALvariability,buttheydonotvaryinthesamesense.In Figure 3-11 ,weplotthetwoshort-termepochswiththesetwovariableregionsshaded. Thereddervaryingintervalat %" 28300kms increasesinstrength,whilethebluer 81

PAGE 82

Figure3-11. Thetwoshort-termepochsfor0146+0142,withthetwoshadedregions markingC IV variability.Theabsorptioninthesetworegionsvaryin oppositedirections,withoneregionbecomingweakerwhiletheother becomesstronger. intervalat %" 36600kms decreasesinstrength.AsmentionedinSection 3.2.3.1 thisisoneofjustthreecasesshowingthisbehavior. 3.2.3.30842+3431 In0842+3431,thereissignicantvariabilityintwodistinctregionsoftheC IV BAL trough(markedbyboldbarsinFigure 3-8 ),whiletheentireSi IV troughvaries.However, thebluer,andmorevariable,regionintheC IV trough(theleft-mostboldbarinFigure 3-8 )correspondstoaregionofweakabsorptionandvariabilityinSi IV .Someofthe variabilityintheredmostportionoftheSi IV trough(theright-mostboldbarinFigure 3-8 )maybeconnectedwiththevariabilityintheredsideoftheC IV BALtrough,butthe variabilityinthecoreoftheSi IV trough,markedbythethinbarinFigure 3-8 ,cannot 82

PAGE 83

beexplainedbythewiderSi IV doubletseparationalone.SomeoftheSi IV variabilityis occurringatdifferentvelocitiesthantheC IV variability.Nonetheless,asseeninFigure 3-9 ,thechangesinstrengthinallthreemarkedregionsoftheSi IV troughoccurinthe samesenseasthechangesinstrengthinthetwodistinctvaryingregionsinC IV Anotherinterestingnoteabout0842+3431isthatweidentieditasvariablein theshort-term,butnotinthelong-term,inChapter 2 .Forthelong-termcomparison inChapter 2 ,weusedthe1990.90andthe2008.35observations.However,Figures 3-8 and 3-9 clearlyshowthattheC IV BALvaried.ThestrenghoftheBALisweakerin 2007.04thanin1990.90,buttheBALbecomesstrongeragainby2008.35.Therefore, whenlookingatjustthe1990.90and2008.35observations,itappearsasiftheBALdid notvaryatall.Thisshowshowvariableprolescanreturntoapreviousstateandthat variableBALscanbemissedin2-epochstudies. 3.2.3.40932+5006 Asin0146+0142,wenotethattheabsorptiontroughthatappearsjustredwardof theSi IV BALin0932+5006inFigure 3-1 isactuallyahigh-velocityC IV BAL(Figure 3-8 ). In0932+5006,thereisC IV absorptionoverlappingSi IV emission.Thevariability inthesetwodetectableBALtroughshasadifferentbehaviourthanthevariabilityin theotherquasarsinoursample.Whenlookingatthespectra(Figure 3-8 ),thetwo high-velocityBALsinthe2003.01spectrumappeartobeoffsetinvelocityfromthe BALsintheotherepochs.WhilethiscouldbeindicativeofashiftinvelocityoftheBALs, thisapparentoffsetcouldalsobeduetopartofthetroughweakeningwhiletheother partstrengthens.Thiscomplicatesthemeasurementof forFigure 3-9 becauseany measurementof inaxedvelocityintervalforeachofthesetwohigh-velocityBALs doesnotaccuratelyrepresenthowthelinechanged.Furthermore,theSi IV emission lineitselfcouldbevariable,whichcomplicatesanyanalysisofC IV BALvariabilityinthis velocityrange. 83

PAGE 84

3.2.3.50946+3009 0946+3009istheoneobjectinoursamplewitharedshifttoolowforourMDM spectratocovertheentireC IV regionouttotheSi IV emissionline.Thespectraonlygo asblueas %" 19000kms .Theabsorptionandvariabilityregionsthatwedenefor thisquasardonotextendallthewaytotheedgeoftheMDMspectruminordertoavoid anyuncertaintiesthere.Thedetectionofvariabilityinthisquasarissecurebecause wehaveanadditionalMDMspectrumofthisquasarthatmatchestheMDMspectrum showninFigure 3-1 3.2.3.61011+0906 AsmentionedinSection 3.1.2 ,someofthequasarsinoursamplehave low-ionizationBALs.WesearchedforAl III linesinoursampleandusedthevelocity oftheAl III linetodeterminethelocationofC II .1011+0906hasanAl III BAL,butthe velocityoftheabsorptionputsC II bluewardoftheSi IV absorption.Therefore,ifthereis anyC II absorptioninthisobject,itdoesnotaffectourmeasurementsoftheSi IV BAL. TheSi IV broademissionline(BEL)inthisquasarismostlyabsorbedbytheC IV BAL.WettheSi IV BELusingtheproceduredenedinChapter 2 forcaseslikethis. WetaketheC IV t,increasetheFWHMbasedonthegreaterdoubletseparationin Si IV ,andplaceitatthewavelengthwheretheSi IV emissionshouldbe.However,there isstillsomeslightemissionbluewardoftheSi IV BELt.Thisextraemissionmaybe partofthewingoftheSi IV BEL,ortheunderlyingpower-lawcontinuumtmightbe slightlytoolow.However,theSi IV BALislocatedatahighenoughvelocitythatany errorintheSi IV BELtshouldhaveanegligibleeffectonourmeasurementsoftheBAL anditsvariability. 3.2.3.71232+1325 1232+1325hasanAl III BALwhichputsC II withintheSi IV region.TheC II BAL isinthevelocityrange 25500to 19600kms (seealsoFigure 3-1 ),andweomit thesevelocitiesfromouranalysis. 84

PAGE 85

3.2.3.81303+3048 1303+3048isaBALquasarthatalsocontainedaC IV mini-BALat %" 18500km s whenrstobservedatLick(Figure 3-1 ).WeonlyhaveoneLickobservationofthis object,buttheMDMobservationsshowthatthemini-BALwidenedandincreasedin strengthtobecomeaBAL.ABALemergesatthesamevelocitiesinSi IV aswell.The lower-velocityBALin1303+3048isvisibleinC IV intheLickdata,butappearsasonly weakabsorptioninSi IV .However,betweentheLickandMDMepochs,aSi IV BAL emergesandthevariabilityintheSi IV absorptionextendstolowervelocitiesthanthe C IV variability.ThevariabilityatallvelocitiesinthisquasarinbothC IV andSi IV occurs inthesamesense;theabsorptionincreasesinstrength. 3.2.3.91309 0536 Asin1011+0906,theSi IV BELin1309 0536isheavilyabsorbedbyaC IV BAL. Weusedthesameprocedurethatweusedfor1011+0906tottheSi IV BEL,andwe foundwhatappearedtobesignicantemissionbluewardoftheSi IV BELt.Inthis case,theunderlyingpowerlawcontinuumtdidnothavethecorrectslope,sowemade aslightadjustmenttothecontinuumt.Adjustingthepowerlawcontinuumtcausedon averageanincreasein of % 6percentthroughoutmostoftheC IV trough,compared tothemeasurementsinChapter 2 .ThenewSi IV BELtincreased atthehighest velocitiesinC IV byuptoafactorof2.Evenwiththisadjustment,therestillappearsto besomeslightemissionbluewardoftheSi IV BELt,but,asin1011+0906,theSi IV BALin1309 0536isatahighenoughvelocitythaterrorsintheemissiontshouldnot havemucheffectonmeasurementsoftheBAL.Thisquasaralsodidnotvaryineither Si IV orC IV ,soanymeasurementerrorsforthisquasarwillnoteffectanyofourresults comparingSi IV andC IV variabilityproperties(e.g.,Figures 3-6 or 3-7 ). 85

PAGE 86

3.2.3.101331 0108 Asin1232+1325,1331 0108hasanAl III BALatavelocitythatplacesthe correspondingC II absorptionwithintheSi IV BAL.Wethereforeomitthevelocity range 23800to 15300kms intheSi IV regionfromouranalysis. WhileanalyzingtheSi IV regionin1331 0108,wenoticedthatthe pseudo-continuumtdenedinChapter 2 neededtoadjusted.Like1309 0536, 1331 0108hasverybroadBALs,whichmakesttingacontinuumdifcult.Theslope ofthetfor1331 0108isnowslightlysteeperthanthetusedinChapter 2 ,increasing themeasured valuesforC IV onaveragebyjust % 7percentthroughoutmostofthe troughandupto % 30percentatthehighestvelocities,wheretheabsorptionismuch weaker. 3.2.3.111423+5000 1423+5000isanotherquasarwherethereweretwoC IV BALsthatvaried,but oneBALincreasedinstrength,whiletheotherweakened.Thisquasarvariedbetween theLickandSDSSobservations,butwedidnotdetectanyvariabilityinthelong-term analysisinChapter 2 .0119+0310,0146+0142,and1423+5000aretheonlyquasarsin oursamplewhereweseetwoseparatevaryingregionsinC IV thatdidnotvaryinthe samesense. 3.2.3.121435+5005 1435+5005hasAl III absorptionthatiseitheraweakBALorstrongmini-BAL. However,thevelocityoftheabsorptionplacesC II bluewardoftheSi IV absorption. Wealsonotethatthesignal-to-noiselevelinthedatafor1435+5005decreases rapidlyatbluerwavelengths.Wethereforewedonotincludethespectralregion bluewardof 12100kms inSi IV inouranalysisinSection 3.2.1 3.3SummaryofResults Thisisthesecondpaperina3-partseriestoanalyzetheBALvariabilitiesina sampleof24BALquasarsmeasuredoriginallybyBarlow(1993)attheLickObservatory 86

PAGE 87

in1988-1992.WesupplementthosedatawithspectrafromtheSDSSarchives(for8 quasars)andourownmeasurementsobtainedattheMDMobservatory(Table 3-1 ).In Chapter 2 wediscussedthevariabilitypropertiesofC IV 1549measuredinjusttwo epochsthatspana"short-term"(0.35 0.75yr)anda"long-term"(3.8 7.7yrs)time interval.Herewebuilduponthatworkbyincludingourfullmulti-epochdatasetfor thesesamequasarsandmakingdetailedcomparisonsbetweentheSi IV andC IV BAL behaviors.Ourmainresultsarethefollowing: (1). BALvariabilityusuallyoccurredinonlyportionsoftheBALtroughs(Chapter 2 ; Section 3.2.2 ). (2). Inthelong-terminterval,65percentoftheBALquasarsinoursampleshowed C IV BALvariabilitywhileonly39percentvariedintheshort-term(Chapter 2 ). (3). C IV variabilityoccursmoreoftenathighervelocitiesandinshallowerabsorption troughs(orshallowerportionsofabsorptiontroughs;Chapter 2 ). (4). Inrarecases,BALfeaturesappear,disappear,orchangetoorfromnarrower mini-BALfeatures(Chapter 2 ;Section 3.2.3.8 ). (5). C IV BALsinourdataareasstrongorstrongerthanSi IV BALsatallvelocities(in allwell-measuredcases;Figure 3-2 ). (6). Si IV BALsaremorelikelytovarythanC IV BALs.Forexample,whenlookingat owspeeds > 20000kms ,47percentofthequasarsinoursampleexhibited Si IV variabilitywhile31percentexhibitedC IV variability(Section 3.2.1 ).The greatervariabilityinSi IV islikelyduetoacombinationofitems(3)and(5)above; weakerlinesaremorelikelytovary,andSi IV tendstobeweakerthanC IV (7). VariabilityinSi IV canoccurwithoutcorrespondingchangesinC IV atthesame velocities. % 50percentofthevariableSi IV regionsdidnothavecorresponding C IV variabilityatthesamevelocities.However,inonlyonepoorly-measuredcase werechangesinC IV notmatchedbySi IV (Sections 3.2.1 and 3.2.3.1 ). (8). AtBALvelocitieswherebothC IV andSi IV varied,thechangesalwaysoccurredin thesamesense(Section 3.2.1 ). (9). WedonotndanycorrelationbetweentheabsolutechangeinstrengthinC IV andinSi IV (Figure 3-6 ),butthefractionalchangeinstrengthtendstobegreaterin Si IV thaninC IV (Figure 3-7 ). 87

PAGE 88

(10). Whenadditionalobservingepochsareincluded(e.g.,ourfulldataset;Section 3.2.2 ),thefractionofC IV BALsthatvariedatanyvelocityincreasesfrom65per centto83percent.Thisincreasewascausedbyvariationsmissedinthe2-epoch comparisonsinChapter 2 (11). BALchangesatdifferentvelocitiesinthesameion almost alwaysoccurred inthesamesense(gettingweakerorstronger)butnotgenerallybythesame amount(Section 3.2.2 ).Wendjust3casesthatshowevidenceforoneC IV BAL weakeningwhileanotherstrengthenswithinthesameobject(Sections 3.2.3.1 3.2.3.2 ,and 3.2.3.11 ). (12). Themulti-epochdataalsoshowthattheBALchangesacross0.04 8.2years intherestframewerenotgenerallymonotonic(Section 3.2.2 ).Thus,the characteristictime-scaleforsignicantlinevariations,and(perhaps)forstructural changesintheoutows,islessthanafewyears. (13). Withmoreepochsadded,westilldonotndclearevidenceforaccelerationor decelerationintheBALoutows(Section 3.2.2 ). 3.4Discussion TheBALvariabilitydataprovideimportantconstraintsontheoutowphysical properties.However,theinformationwederivedependscriticallyonwhatcausesthe BALvariations.Inthissectionwediscussprosandconsoftwocompetingscenarios, namely,1)uctuationsinthefar-UVcontinuumuxthatcauseglobalchangesinthe outowionization,and2)outowcloudsmovingacrossourlines-of-sighttothequasar continuumsource. AnimportantpartofthisdiscussionistheBALopticaldepths,whichcanbe muchlargerthantheyappearinthespectrumiftheabsorberscoveronlypartofthe backgroundlightsource( Hamann 1998 ; Hamannetal. 2008 ).Comparisonsbetween theC IV 1549andSi IV 1400BALscanhelpbecausetheselinesprobeslightly differentionizationswithpotentiallyverydifferentlineopticaldepths.Forexample,ina simplesituationwithsolarabundancesandanionratioequaltotheabundanceratio, i.e.,Si IV /C IV =Si/C,theopticaldepthinSi IV 1400wouldbe % 3.4timeslessthan C IV 1549( Hamann 1997 ; Hamann&Ferland 1999 ; Asplundetal. 2009 ).Inactual BALows,therelativeSi IV opticaldepthshouldbeevenlowerbecauseBALionization 88

PAGE 89

tendstobehighandthusfavorsC IV .Wecannotmakespeciccomparisonswithout specicknowledgeoftheabsorberionizations.However,ifwereasonablyassumethat theionizationisatleastashighasthatneededforamaximumC IV /Cratio(e.g.,ina gasthatisphotoionizedbythequasarandopticallythinintheLymancontinuumgure A1inHamannetal.2011),thentheSi IV opticaldepthsshouldbe > 8timessmallerthan C IV 3.4.1ChangingIonization WhenthereisvariabilityindifferentvelocityintervalswithinthesameBALorwithin multipleBALsinthesamequasar,thechangesalmostalwaysoccurinthesamesense (e.g.,0842+3431and0903+1734Figures 3-8 and 3-9 ).Studiesofnarrowabsorption line(NAL)variabilityhaveobservedmultipleNALsinagivenquasarvaryinginconcert ( Misawaetal. 2007 ; Hamannetal. 2011 ). Hamannetal. ( 2011 )foundcoordinated linevariationsinveNALsystemsinasinglequasar.Theyarguethatthemostlikely explanationforthisisaglobalchangeinionization.Iftherearechangesintheionizing uxincidentontheentireoutow,thenglobalchangesinionizationshouldoccur.While theconnectionbetweenNALsandBALsisunclear,thisargumentcanbeappliedto BALsaswell.Absorbingregionsatdifferentvelocitieshavedifferentradialdistances fromthecentralSMBH.Theyarethereforespatiallydistinct,eveniftheyarepartof thesamelargeroutowstructure.Achangeincoveringfractionduetomovingclouds isunlikelyincasessuchas0842+3431and0903+1734becauseitwouldrequire coordinatedmovementsamongmultipleabsorbingstructuresatdifferentoutow velocitiesandradii. Tofurtherinvestigatethisscenario,forsimplicity,weconsiderasystemwith homogeneouscloudsoutowingfromtheaccretiondisk,withnotransversemotion acrossourline-of-sighttothequasar.Achangeinionizationwillcausetheoptical depthsinthelinestochange.Asmentionedabove,theopticaldepthsinC IV arehigher thaninSi IV ,soSi IV wouldbemoresusceptibletochangesinionization.Therefore,it 89

PAGE 90

ismorelikelyforSi IV tovarythanC IV inthisscenariobecauseC IV ismorelikelytobe saturated.ThisgenerallymatchesourresultssincewendthatSi IV ismorevariable thanC IV .WendonlyonecaseofC IV variabilityunaccompaniedbySi IV variability, anditisaverytentativeresult(Section 3.2.3.1 ). Thisscenariobecomesalittlemorecomplicatedwhenconsideringthattypically variabilityonlyoccursinportionsofBALtroughs,ratherthanentireBALtroughsvarying ( Gibsonetal. 2008 ;Chapter 2 ).Asmentionedabove,achangeinionizationshould causemoreglobalchanges,ratherthanchangesinsmall,discretevelocityintervals.It ispossiblethatthevariableregionsinthetroughshavemoderateorlowopticaldepths, whilethenon-variablesectionsaretoosaturatedtorespondtomodestchangesin theionizationandlineopticaldepths.WehaveevidencefromChapter 2 andFigure 3-5 thatweakerlines,orweakerportionsoflines,aremorelikelytovary.However,we alsofoundinChapter 2 thatvariabilityismorecommonathighervelocities,wherethe absorptiontendstobeweaker,soitisdifculttosaywhetheritisthehighervelocityor theweakerabsorptionstrengththatistherootcauseofthevariability.Ifitistruethat weakerportionsoflines,whichareleastlikelytobesaturated,aremorelikelytovary, regardlessofoutowvelocity,thenthiswouldsupportthechangingionizationscenario. However,ifitistruethatweakerportionsoflinesarelesssaturatedandthusmore likelytovary,itisunclearwhy,forexample,theweakbluewingoftheBALtroughin 0842+3431doesnotvary.Ifchangingionizationiscausingthevariabilityinthisquasar, thenthewingsofthelinemustbesaturatedwhiletheportionsofthelineadjacenttothe wingsarenotsaturated.Thereareotherexamplesofsimilarbehavior.In1524+5147, thestrongestvariabilityoccursinthedeepestsegmentoftheBAL,andthereisweakor novariabilityintheweakestsegmentsofthetrough(Figures 3-1 and 3-8 ).Similarly,the variabilityin1011+0906occurrednearthecoreoftheline,whilethewingsdidnotvary (Figure 3-1 ). 90

PAGE 91

Inorderforchangingionizationtocausevariabilityinjustthedeepestportions ofBALtroughs,asintheexamplesgivenabove,theremustbevelocity-dependent coveringfractionswithvelocity-dependentopticaldepths.Inthisway,eventheweak wingscanbehighlysaturated.Thereisevidenceintheliteraturethatbothoptical depthandcoveringfractionscanhavecomplexvelocity-dependentbehaviors( Barlow &Sargent 1997 ; Hamannetal. 1997 2001 ; Gangulyetal. 1999 ; deKooletal. 2002 ; Gabeletal. 2005 Gabeletal. 2006 ; Aravetal. 2008 ). ThetrueopticaldepthsandcoveringfractionsaredifculttomeasureforBALs.In ourdata,1413+1143providesdirectevidenceforvelocity-dependentopticaldepthsif thelinevariationsarecausedbyionizationchanges.AtthecoreoftheSi IV troughthere aretwodipsattheSi IV doubletseparation(Figure 3-1 ).Thedoubletratioisroughly one-to-one,indicatingsaturationatthesevelocitiesandthereforelittleornosensitivityto changesincontinuumux.Thispartofthetroughdidnotvary.However,thissaturated doubletissurroundedbyvariabilityathigherandlowervelocitiesinSi IV .InC IV ,which shouldhavegenerallylargeropticaldepths,variabilityoccursonlyathighervelocities. Thisbehaviorisatleastsuggestiveofloweropticaldepths(non-saturatedabsorption)at thevariablevelocities. Onefurtherpieceofevidenceforthechangingionizationscenariocomesfrom themulti-epochdatainSection 3.2.2 ,whichshowthatchangesinBALstrengthare notnecessarilymonotonic(seealso, Gibsonetal. 2010 ).Infact,in0842+3431,the absorptiontroughvaried,andtheninourlastMDMobservationitreturnedtothesame strengthithadinoneoftherstLickobservations.Inorderforachangeincovering fractiontohavecausedthevariabilityin0842+3431,thecloudmovementswouldhave toberepeatable,inadditiontobeingcoordinatedatdifferentvelocitiescorrespondingto differentspatiallocations.Achangeinionizationisamorelikelyexplanationbecause continuumuxvariationsarenotnecessarilymonotoniceither(e.g. Barlow 1993 ). 91

PAGE 92

Ifachangeinionizationdoesindeedcausethevariabilitywedetect,thenthere shouldbeaconnectionbetweenchangesincontinuumuxandBALvariability. However,theresultsfrompreviousstudieshavebeenmixed. Barlow 1993 foundsome evidenceforacorrelationbetweencontinuumvariabilityandBALvariations,atleast incertainindividualquasars,whileotherstudieshavenotfoundastrongcorrelation ( Barlowetal. 1992 ; Lundgrenetal. 2007 ; Gibsonetal. 2008 ).However,allofthese studieslookatnear-UVuxvariationsandlittleisknownaboutthefar-UVvariability propertiesofluminousquasars.Itisthefar-UVuxthatisthesourceoftheionizing radiation.Therefore,theseresultsdonotruleoutionizationchangesasacauseofBAL variability. 3.4.2ChangingCoveringFraction Whilemostoftheevidencepresentedsofarfavorsionizationchanges,previous BALvariabilitystudies,includingChapter 2 ,havefavoredchangingcoveringfractions overionizationchanges( Lundgrenetal. 2007 ; Gibsonetal. 2008 ; Hamannetal. 2008 ; Krongoldetal. 2010 ; Halletal. 2011 ).Toinvestigatethispossibility,weconsidera simplescenariowithcloudsthathaveconstantionizationandcolumndensity,butare movingacrossourline-of-sight.IfC IV andSi IV havethesamecoveringfraction,then C IV shouldbejustaslikelytovaryasSi IV ,whichisinconsistentwiththeresultsof Section 3.2.1 .Further,thechangeinstrengthinthetwolinesshouldbethesame,which iscontradictedbyourresultsinFigure 3-6 (also, Gibsonetal. 2010 ).Hence,thissimple scenarioclearlydoesnotmatchtheresultsofthisandpreviouswork. Amorerealisticscenarioinvolvescloudsthatcanhavedifferentcoveringfractions inC IV andSi IV ( Barlow&Sargent 1997 ; Hamannetal. 1997 2001 ; Gangulyetal. 1999 ; Gabeletal. 2005 Gabeletal. 2006 ; Aravetal. 2008 ). Hamannetal. ( 2001 )and Hamann&Sabra ( 2004 )discusssimpleschematicsofinhomogeneouscloudsthat couldleadtodifferentcoveringfractionsindifferentions(gure6in Hamannetal. 2001 andgure2in Hamann&Sabra 2004 ).Strongertransitionsinmoreabundantionscan 92

PAGE 93

havealargeropticaldepthoveralargerareaintheseschematicmodels.Thus,Si IV maytraceadifferentareaoftheoutowinggascloudsthanC IV IftheC IV andSi IV linesaresaturated,e.g.liketheBALsin Hamannetal. 2008 thenthestrengthsofthelineswouldbegovernedbythecoveringfractionsinthose lines.Inthiscase,asmallercoveringfractioninSi IV wouldbeconsistentwiththeresults ofFigure 3-2 ,whichshowsthatSi IV linesaregenerallyweakerthanC IV .IfSi IV is tracingasmallerareaofthegascloudthanC IV andthiscloudismovingacrossour line-of-sight,thenSi IV absorptionwouldgenerallybemorevariable.Furthermore,ifthe coveringfractionsaredifferentforeachion,thenthechangeincoveringfraction,aswell asthefractionalchangeinstrengthoftheabsorptionlines,foreachioncanalsodiffer. ThisisconsistentwithFigures 3-6 and 3-7 InSections 3.2.3.1 3.2.3.2 ,and 3.2.3.11 ,wereportonthreecasesthatshow evidenceofoneBAL,oroneportionofaBAL,strengtheningwhileanotherweakens withinthesamequasar.Thiscanbereadilyexplainedinamovingcloudscenario,for itispossibleforcloudsatdifferentvelocitiestoenter/leaveourline-of-sightatdifferent times.Ifonecloudentersourline-of-sight,whileanotherisexiting,wewouldseeone BALstrengtheningwhileanotherweakens. 3.4.3Conclusions ThehighervariabilityfractionsinSi IV versusC IV BALsandcoordinatedvariabilities betweenabsorptionregionsatdifferentvelocitiesinindividualquasarssupports thescenarioofglobalchangesintheionizationoftheoutowinggascausingthe observedBALvariability.Furthermore,velocity-dependentcoveringfractionsand opticaldepthscouldexplainwhyinmanycasesweseevariabilityinjustportionsofBAL troughs,ratherthanentiretroughsvarying.Ontheotherhand,variabilityinportions ofBALtroughstsnaturallyinascenariowheremovementsofindividualclouds,or substructuresintheow,arecausingchangesincoveringfractionsintheabsorption 93

PAGE 94

lines.ThisscenarioisalsoconsistentwiththemainresultsofSection 3.2.1 ,assuming thatSi IV hasasmallercoveringfractionthanC IV Inreality,changesinionizationandcoveringfractionscouldbothbecontributing toBALvariability.Inoursample,therearequasarsinwhichweobservednovariability; therearequasarsthatvariedinonlyoneortwonarrowvelocityintervals;and,thereare yetotherquasarswithvariationsoverawiderangeinvelocities.Itisunlikelythatone scenarioisgoverningthechangesinallofthesequasars.Perhapsinquasarswherethe linesaremoresaturated,thelinesarenotsusceptibletosmallchangesinionization,but caneasilyvaryduetocoveringfractionchanges.Inothercases,wherethelineshave loweropticaldepth,achangeinionizationcancauselargechangesoverawiderange invelocities,possiblymaskingvariationsduetochangesincoveringfraction. TherearestillsomeunansweredquestionsthatourresultsfromChapter 2 ,andthe currentwork,raise.Inparticular,inChapter 2 ,wendcorrelationsbetweenincidence ofC IV variabilityandbothvelocityandabsorptionstrength.However,velocityand absorptionstrengtharealsocorrelated.Ifthetrendisreallywithabsorptionstrength, indicatingthatweakerlines,whicharelesslikelytobesaturated,aremorevariable,then thisfavorsionizationchanges.Ifthetrendiswithvelocity,thentheimplicationsaremore ambiguous,butitwouldbemoreconsistentwiththecrossingcloudscenario.Clouds withhigheroutowvelocitiesaremorelikelytohavegreatertransversevelocitiesas well. TherearealsodocumentedcasesofBALsemergingwheretherehadhitherto beennoabsorption( Hamannetal. 2008 ; Krongoldetal. 2010 ),andinthiswork,we reportonaquasar(1303+3048;Section 3.2.3.8 )whereamini-BALbecameaBAL. ThesescenariosspeaktothegeneralcomplexityofBALvariability.Previousstudies havehypothesizedthatdifferentoutowlinesmayindicatedifferentinclinationsofour lines-of-sighttothequasarsandthatatcertaininclinationsnoabsorptionisseen( Elvis 2000 ; Gangulyetal. 2001 ).TheconnectionbetweenBALs,mini-BALs,andNALsisstill 94

PAGE 95

unclear.IfBALandmini-BALoutowsoccuratdifferentinclinations,thenperhapsour line-of-sightto1303+3048goesthrougharegionofoverlapbetweenthemini-BALand BALinclinations.OnemightexpectthisputativeborderregionbetweentheBALand mini-BALpartsoftheowtobethemostturbulentorunstable,andthusthemostprone toshowinglinevariabilitycausedbystructuralchanges/motionsintheow. Thisworkisonlythesecondpaperina3-partseriesonBALvariability.The nextpaper(Chapter 4 )willinclude1)amorethoroughexplorationofthevariability time-scales,withnewdataaddedtogiveextensivecoverageacrossweektomonth intervals;and2)amorecompletediscussionoftheimplicationsofvariability,e.g.,in termsofthelocationoftheoutowstructures. 95

PAGE 96

CHAPTER4 VARIABILITYINQUASARBROADABSORPTIONLINEOUTFLOWSIII.WHAT HAPPENSONTHESHORTESTTIME-SCALES? Thetime-scalesofvariabilitycanhelpconstrainthesizes,speeds,andlocationsof cloudsinthemoving-cloudscenarioorthelocationandunderlyingnatureofcontinuum changesinthecaseofchangingionization.Variabilityonshortertime-scalesplaces theoutowatsmallerdistancesfromthecentralSMBH,basedonnominallyshorter crossingtimesforcloudsmovingacrossourline-of-sight( Hamannetal. 2008 ;Chapter 2 )orthehigherdensitiesrequiredforshorterrecombinationtimes( Hamannetal. 1997 ). WhilesomeoftheearlierworkonBALvariabilityhavedataontime-scales & 1yr, informationonvariabilityontime-scales < 1monthisverylimited. 1 Isthereaminimum time-scalethresholdbelowwhichthereisnovariability?Informationonvariabilityon theshortesttime-scalesallowsustodeterminehowclosetothecentralSMBHthis outowinggascanbelocated. InPaper1(i.e.Chapter 2 ),wedirectlycomparedvariabilitypropertiesbetween a"short-term"intervalof0.35 0.75yranda"long-term"intervalof3.8 7.7yr.In particular,wereportedthat65percentofquasarsvariedinthelong-term,comparedto 39percentintheshort-termdata.Wealsofoundaslighttrendtowardsalargerchange instrengthinthelong-term.Weexpandonthisworkbyincludingallthedatafromour BALmonitoringprogramme,includingnewdataatshortertime-scalesthanwhatwe includedintheshort-termsubsampleinChapter 2 .Wenowhaveupto13epochsof dataperquasarovertime-scalesfrom0.02yr(7days)to8.7yr. ThisextensionofourBALmonitoringprogrammespecicallyprobestheshortest time-scales,andwehighlightsomeindividualcasesofrealandpotentialBALvariability 1 Throughoutthispaper,alltimeintervalsarereportedintherestframeofthequasar. 96

PAGE 97

thatoccurredovertime-scalesof < 2months.Inparticular,wedescribethecasesof 1246 0542and0842+3431,forwhichwefoundvariabilitydowntotime-scalesof8and 10days,respectively. Themulti-epochnatureofourdataset,spanningawiderangeoftimeintervals ( t),allowsustoinvestigatetheprobabilityofdetectingBALvariabilityversus t.We focusonjusttheC IV 1549lineagainhere.Bydeterminingtheincidenceofvariability versustime-scale,wecandeterminewhetherthereisadrop-offintheincidence ofvariabilityattheshortesttime-scales.Furthermore,lookingatthefullrangeof time-scalesgivesanestimateofhowlikelyitistondvariabilityoverdifferenttime intervalsandwhetherthereisatime-scalebeyondwhichallBALquasarswillvary.From Chapter 2 and Gibsonetal. ( 2008 ),weknowthatamajorityofBALquasarsvaryon multi-yeartime-scales.InChapter 3 ,wefoundthatbyaddingmoreepochsofdata, evenatsimilartime-scalesasthemeasurementsinChapter 2 ,thefractionofquasars thatvariedincreased.Byaddingevenmoreepochsofdatahere,includingsome measurementsatslightlylongertime-scalesthanpreviouslyreported,wewilldetermine whethertheincidenceofC IV variabilityapproachesunity. WegiveanoverviewoftheBALmonitoringprogrammeinSection 4.1.1 and describeouranalysisinSection 4.1.2 .Section 4.2 describesourresultsandSection 4.3 discussestheimplicationsoftheseresultsintermsofphysicalpropertiesofthe outowinggas. 4.1DataandAnalysis 4.1.1ObservationsandQuasarSample Thispapercontinuesouranalysisofthesame24BALquasarsstudiedinChapters 2 and 3 .Thissamplewasdenedin Barlow ( 1993 ),whichcontainsdatafromtheLick Observatory3-mShaneTelescope,usingtheKastspectrograph.Weincludedatafrom thisstudythathavearesolutionof / ( (230kms ),or ( 600(530 kms )ifthereisnohigherresolutiondataavailableforthatepoch.Eitherresolutionis 97

PAGE 98

sufcientforstudyingthevariabilityinBALssincetheyaredenedtohaveawidthofat least2000kms WehavebeenmonitoringthissampleofBALquasarsusingtheMDMObservatory 2.4-mHiltnertelescopewiththeCCDSspectrograph,witharesolutionof ( 1200 (250kms ).WealsoincludeSloanDigitalSkySurvey(SDSS)spectra,whichhavea resolutionof ( (150kms ),whenavailable( Adelman-McCarthyetal. 2008 ). WeincludeinthisworkalloftheobservationsdescribedinChapters 2 and 3 ,andmore detailsonthetelescopeandspectrographsetupscanbefoundtherein. Forthecurrentwork,weselectedasubsampleofourBALquasarsfromChapters 2 and 3 tore-observemultipletimesoverrest-frametimeintervalsof t % 1weekto1 monthduringthersthalfof2010.Wechoseobjectswitharightascensionof08 15 hrssothatwecouldobservethemfromJanuarythroughMay.Thisrestrictsoursample to17outof24quasars.Wewereabletoobserve13ofthesequasars2to4timeseach. Elevenoutofthese13quasarsthatwechosetomonitorpreviouslyexhibitedvariability intheanalysisinChapters 2 and 3 ,whichisslightlyhigherthanthevariabilityfractions weobtainedfromthesimple2-epochanalysisofChapter 2 .Therefore,intheanalysis below,wemightbeslightlybiasedtowardshighervariabilityfractions,especiallyatthe shortertime-scales. MostoftheseobservationsweretakenattheMDM2.4-minJanuary,February, March,andMay2010.WealsohavesomeobservationstakenattheKPNO2.1-min February2010.WeusedtheGoldCamspectrographwitha600groovepermmgrating inrstorderanda2arcsecslit,providingaspectralresolutionof ( 1100(275km s ).Theobservedwavelengthcoverageis % 3600to6100 Aforalltheobservations, whichcoversthefullSi IV andC IV absorptionregionsforthequasarsobserved.The co-addedexposuretimeswere2.5 3hpersource. 98

PAGE 99

Table4-1. FullBALquasardatasetandshortesttime-scaledata Name BINo.of tNo.ofMeas. t Obs.(yrs) < 72days(days) 0019+01072.13229070.08 5.79228 30 0043+00482.14433050.35 6.13...... 0119+03102.09607030.65 5.57...... 0146+01422.91578040.52 5.15...... 0226 10242.25777024.66...... 0302+17052.89030.27 4.42...... 0842+34312.154430130.03 6.51910 64 0846+15402.93080.04 4.93414 65 0903+17342.771070070.04 5.38215 32 0932+50061.937920130.02 7.3799 39 0946+30091.22555090.03 8.67712 51 0957 05351.81267080.02 6.9379 42 1011+09062.276100100.03 6.55510 37 1232+13252.361100030.35 5.93...... 1246 05422.24481070.02 6.5238 25 1303+30481.77139090.03 6.5179 41 1309 05362.22469070.02 6.5438 25 1331 01081.881040070.02 6.6419 1336+13352.45712040.07 5.79127 1413+11432.56681070.03 5.8919 1423+50002.25306080.02 6.4938 25 1435+50051.591150070.03 8.15310 31 1524+51472.88181090.02 5.3937 32 2225 05341.98792030.27 0.73...... Table 4-1 summarizesthefulldatasetfromourBALmonitoringprogramme, includingtheemissionredshift, 2 andthebalnicityindex'(BI)foreachobject(as calculatedinChapter 2 ).Thefourthcolumngivesthetotalnumberofobservations,and thefthcolumngivestherangein tvaluesforallthemeasurementsofeachobject, whereonemeasurementisapairofobservations.Thesixthcolumnliststhenumberof 2 Thevaluesof wereobtainedfromtheNASA/IPACExtragalacticDatabase (NED),whichisoperatedbytheJetPropulsionLaboratory,CaliforniaInstituteof Technology,undercontractwiththeNationalAeronauticsandSpaceAdministration. 99

PAGE 100

measurementsthatcovertime-scales < 0.20yr,or < 72days,andthenalcolumngives therangein tvalues,indays,forthesemeasurements. 4.1.2MeasuringBALsandtheirVariability ThispaperfocusesonvariabilityinC IV overdifferenttime-scales.Inorderto measurevariability,werstadoptthevelocityintervalsoverwhichC IV BALabsorption occursineachquasardenedinChapter 2 .WeusedthedenitionofBIasaguidefor deningtheseregions,i.e.theymustcontaincontiguousabsorptionthatreaches ) 10 percentbelowthecontinuumacrossatleast2000kms Forthecurrentwork,weadoptthepower-lawcontinuumtsdenedinChapters 2 and 3 .ApowerlawcontinuumisttooneducialLickspectrumperobject.For simplicity,weuseonlythepower-lawtfortheplotsinthispaperanddonottthebroad emissionlines(BELs).Anymeasurementsofstrengthinthispaper,asinSection 4.2.1 areperformedatvelocitiesthataremostlyunaffectedbytheBELs. Tocomparemultipleepochsforeachquasar,weusedasimpleverticalscaling, asdescribedinmoredetailinChapters 2 and 3 ,tomatchallthespectratotheducial LickspectrumdenedinChapter 2 .Insomecases,asimplescalingdoesnotproduce agoodmatch,andthereweredisparitiesintheoverallspectralshapebetweenthe comparisonspectra.Forthesecases,wetypicallytalinearfunction,orrarely,a quadraticfunction,totheratioofthetwospectraacrossregionsthatavoidtheBALsand thenmultipliedthisfunctionbytheMDMorKPNOspectrumtomatchtheducialLick data. InChapters 2 and 3 ,wehavealreadyanalyzedmostofthemeasurements presentedinthispaper.Forthenewdataincludedinthiswork,weusepreciselythe samecriteriaasdescribedinChapters 2 and 3 fordeterminingwhetheraquasarhas avariableBAL.Thisprocessincludesvisuallyinspectingeachpairofspectraforeach quasar.Forapairofobservationstoqualifyasanincidenceofvariability,thevarying region(s)mustmeettwothresholds.First,thecandidatevaryingregionmustbeatleast 100

PAGE 101

1200kms inwidth.Second,weusetheaverageuxandassociatederrorwithinthat velocityintervaltoplaceanerrorontheuxdifferencesbetweenthetwospectra(see Equation1inChapter 3 ).Theuxdifferenceinthisintervalmustbeatleast4 # tobe includedasavaryingregion. However,wetakeaconservativeapproachandomitasmallnumberofambiguous casesthatmeetthisthreshold.Photonstatisticsalonearenotsufcientfordeningreal variabilitybecauseuxcalibrations,apoorlyconstrainedcontinuumplacement,and underlyingemission-linevariabilitycanalladdadditionaluncertaintynotcapturedby photonstatistics.Chapters 2 and 3 haveamoredetailedexplanationofthisprocedure andidentifyspecicexamplesofsuchambiguouscases. Inthispaper,weaimtoidentifytheshortesttimeintervalsoverwhichvariability occurredinoursample.FromChapter 2 and Gibsonetal. ( 2008 ),thereisevidencethat boththeincidenceandtheamplitudeofvariabilityissmalleratshortertime-scales. Furthermore,whenidentifyingintervalsofvariabilityfortheshortesttime-scale measurementsinthiswork,thevelocitywidthsofthevaryingregionstendtobeat, orjustabove,thethresholddenedinChapter 2 .Wethereforeidentify"tentative"cases thatmeetthethresholdsaboveandareprobabledetectionsofvariability,butarenotas secureasthecaseswedeneasvariable.ThroughouttheanalysisofSection 4.2.3 ,we showresultsbothwithandwithoutthesetentativecases.Thisway,wecanbalanceour goalofonlyincludingsecurecasesofvariabilitywhileattemptingnottoexcludetenuous casesofvariabilitythatmayinfactbereal.Wediscussseveralexamplesofsecureand tentativedetectionsofvariabilityinSection 4.2.2 below. 4.2Results 4.2.1AmplitudeofVariability InChapter 2 ,wedirectlycomparedthevariabilitybetweena"short-term"time intervalof0.35 0.75yranda"long-term"intervalof3.8 7.7yr,andthechangein strengthwasgenerallyslightlyhigherinthe"long-term"data.Todetermineifthere 101

PAGE 102

Figure4-1. Thechangeinabsorptionstrength, ,forvariableportionsofabsorption troughsversusthetimeintervalbetweenobservations, t. isaclearcorrelationbetweentheamplitudeofstrengthchangesandthevariability time-scale,werstcalculatetheaveragestrength, ,foreachepochwithinthe velocityintervalsthatvariedandthendeterminethechangeinstrength, ,foreach measurementofvariability.Wedonotdividethevariableintervalsintobinsaswedo inChapters 2 and 3 ;instead,wecalculatetheaverage overtheentireinterval.In Chapters 2 and 3 ,wefoundthatvariabilitytendstooccurinjustportionsoftroughs,and sometimesinverynarrowportions(also Gibsonetal. 2008 ).Measurementsof and giveadirectmeasurementofthestrengthofthelines,andthechangeinstrength,in onlytheportionsthatvaried.Equivalentwidth(EW)measurementsofawidetroughare lesssensitivetostrengthchangesinanarrowportionofthetrough. 102

PAGE 103

Weplot versus tinFigure 4-1 .Inordertoavoidtheaddeduncertaintyof overlappingemissionlines,wecalculate onlyforabsorptionbetween 27000 and 8000kms (Chapter 2 ).Wendanincreasingenvelopeofvaluesof with increasing t.Thisisconsistentwiththendingsof Gibsonetal. ( 2008 ),whoshow thattheenvelopeofvaluesof EWexpandsas tincreases.Inourdataset, never exceeds % 0.3fortime-scalesuptoa tof % 4yr.Astime-scalesapproach tof10 years, Acanbeashighas % 0.46. Weincludeinthisplotmeasurementsofthetentativecasesofvariability(thegreen triangles;Section 4.1.2 ).Ontheshortesttime-scalesthatweareprobinginthispaper, with t < 0.20,the Avaluesareallbelow % 0.1.Therefore,whenwedeterminethe incidenceofvariabilityontheshortesttime-scales,thevalueswillbedominatedby weaklyvaryinglines.Wethereforearemorestringentwhenidentifyingmeasurementsof variabilityontheseshorttime-scales,andthatiswhyweclassifysomemeasurements astentativecasesofvariability. 4.2.2ExamplesofVariabilityontheShortestTime-Scales AsmentionedinSection 4.1 ,weobtainednewdataspecicallytoaugmentthe temporalsamplingofourdatasetattime-scales < 0.20yr( < % 72days)inthequasar rest-frame.Withthenewdata,wenowhave70measurementsfor17ofthequasarsin oursampleattheseshorttime-scales(Table 4-1 ).C IV BALvariabilityoccurredin2of thesequasars(12percent).AsmentionedinSection 4.1.2 above,weidentifytentative casesofvariabilitythatmeetourthresholdsforvariabilitybutappeartenuous.Ifwe includethesetentativecases,thenvariabilityoccurredin5quasars(29percent)over time-scales < 0.20yr.Evenwiththetentativecasesincluded,thisfractionisstilllower thantheincidenceofvariabilityinour"short-term"sub-samplefromChapter 2 ,where39 percentofthequasarsvariedovertimeintervalsof0.35 0.75yr. Determiningwhichmeasurementsexhibitrealvariabilityattheseshorttime-scales isimportantforSection 4.2.3 belowwhereweinvestigatetheincidenceofvariability 103

PAGE 104

versus t.Weneedtobemorestringentherebecausetherearefewermeasurements ofvariabilityontheseshorttime-scales,and,asdescribedinSection 4.2.1 above,the measurementsthatdoexhibitvariabilityallhavesmallamplitudesofvariability( ; Figure 4-1 ). Inthissection,wehighlightthetwoquasarswithsecuredetectionsofvariability ontheshortesttime-scalesinourdataset( < 0.20yr),aswellasexamplesofcases thatwereclassiedastentativedetectionsofvariability.Weemphasizethatallofthe casesofvariability,andtentativevariability,discussedheremeetthethresholddenedin Chapter 2 andusedthroughoutthispaper,i.e.thevariabilitymustoccuroveraregionat least1200kms wideandtheuxdifferencebetweenthetwospectramustbeatleast 4 # (Section 4.1.2 ).Wepresenttheseindividualcasesinorderofincreasingtime-scale. 4.2.2.11246 0542:Variabilityover8days Ourdatasetcontainsmeasurementsdownto tof0.02yr( % 7 8days),andwe foundvariabilitydowntothistime-scalein1246 0542.Wepresentthismeasurementin Figure 4-2 ,showingthefullspectrumforthe2010.13and2010.20observations,andin Figure 4-3 ,showingjusttheC IV BAL.AsindicatedbyshadedbarsinFigures 4-2 and 4-3 ,therearetwoseparatevelocityintervalsofvariabilitybetweenthe2010.13(black curve)and2010.20(redcurve)observations.Thetwovelocityintervalsarecentered at 24050kms and 17200kms ,withwidthsof1300kms and2600kms respectively.Theuxdifferenceinthesetwointervalsis8.0 # and13 # TherightmostshadedbarsinFigure 4-2 markthevariableintervalsinC IV ,and theycorrespondtotheshadedintervalsinFigure 4-3 .Wethenshiftedtheseintervals tothecorrespondingvelocitiesinN V ,whicharemarkedbytheleftmostshadedbars, andSi IV andwidenedtheintervalsbasedonthewiderdoubletseparationsinthesetwo lines.WendevidenceofvariabilityinN V atthesamevelocitiesastheC IV variability, andinparticular,intheredmostintervalofvariability.ThedetectionofvariabilityinN V issomewhattenuousonitsownbecausethereisaslightmismatchinthecontinuum 104

PAGE 105

Figure4-2. Thefullspectraof1246 0542,showing3epochscoveringa tof 0.02 0.07years.TherightmostshadedregionsmarktheintervalsofC IV variability,andtheotherpairsofshadedregionsmark,fromlefttoright,the correspondingvelocitiesinN V andSi IV nearthelimitofthewavelengthcoverage,butthevelocitiesofapparentvariability,inthe redmostshadedregionespecially,matchthevelocitiesatwhichC IV appearstovary. CorrespondingvariabilityinN V supportsthedetectionofvariabilityinC IV .However,we ndnocorrespondingchangesintheSi IV BAL.Thisisaninterestingresultbecause,in Chapter 3 ,weonlyfoundonetentativecasewhereC IV variabilitywasnotaccompanied bySi IV variabilityatthesamevelocity. Figures 4-2 and 4-3 alsoshowthattheBALin1246 0542againvariedbetween 2010.20and2010.35( t % 0.05yr,or17days),returningtoitspreviousstatefrom 25daysearlier.ThereisnodifferenceintheBALstrengthbetweentheearlier 2010.13spectrum(blackcurve)andthe2010.35spectrum(bluecurve).Wealso 105

PAGE 106

Figure4-3. SpectraoftheC IV BALin1246 0542fortheepochsshowninFigure 4-2 havemeasurementsofthisquasaratotherepochswithvariabilityinthesamevelocity intervalswherewedetectvariabilityinFigures 4-2 and 4-3 .Inparticular,inChapters 2 and 3 ,wendvariabilitybetweenthetwo"long-term"epochs,1992.19and2008.35,at thesamevelocitiesasthebluemostintervalhere.Thisgivesfurtherconrmationthat thedetectionofvariabilityover8daysbetween2010.13and2010.20issecure. 4.2.2.20842+3431:Variabilityover10days Wehaveanotherdetectionofvariabilityonasimilartime-scaleas1246 0542.We detectvariabilityover0.03yr(10days)betweenthe2010.11and2010.20observations of0842+3431.ThesespectraareshowninthetoppanelofFigures 4-4 and 4-5 ,andthe variableinterval,markedbyashadedbar,iscenteredat 17800kms andis1800 kms wide.Theuxdifferencebetweenthetwoepochsis8.5 # Thisvariableintervalisonthebluesideofthetroughatahighenoughvelocitythat thisdetectionofvariabilityisnotaffectedbyanyvariabilityintheC IV BEL.Thespectra 106

PAGE 107

Figure4-4. Spectraof0842+3431.Eachpanelshowsameasurementofvariabilityover atime-scaleof < 0.20yr,withthetoppanelshowingtwoepochsseparated byjust0.03yr(10days).Therightmostshadedregionsmarktheintervalof C IV variabilityforeachmeasurement,andtheothershadedregionsmark thecorrespondingvelocitiesinSi IV 107

PAGE 108

Figure4-5. SpectraoftheC IV BALin0842+3431.Thetoppanelshowsthesametwo epochsfromthetoppanelofFigure 4-4 ,andthebottompanelshowsallfour LickepochsfromthebottomtwopanelsofFigure 4-4 108

PAGE 109

alsomatchverywelloneithersideoftheshadedinterval.Furthermore,wemarkthe correspondingvelocitiesinSi IV withashadedbar,andthereispotentialvariability there. ThemiddlepanelofFigure 4-4 showtwoLickspectra,1990.90and1991.11, whichareseparatedby24days,andthebottompanelshowsthe1991.31and1991.86 spectra,whichareseparatedby64days.Inbothofthesemeasurements,thereis variabilityinSi IV atthesamevelocitiesasthevariabilityinC IV .Allfourspectrafrom 1990.90to1991.86areshowninthebottompanelofFigure 4-5 (afulldiscussionof themulti-epochnatureofthevariabilityin0842+3431canbefoundinChapter 3 ).The variabilitybetweentheseLickobservationshelpstoconrmthevariabilitybetween 2010.11and2010.20becausethevariabilityoccursatsimilarvelocities.Withoutthis cleardetectionofvariabilityintheLickdata,wewouldhaveclassiedthevariability between2010.11and2010.20asatentativedetection. 4.2.2.31011+0906:Possiblevariabilityover17days Inthissection,wedescribeameasurementthatweclassiedasatentative detectionofvariability.Thefullspectraforthe2010.20and2010.35observationsof 1011+0906,whichareseparatedbya tof0.05yr(17days),areplottedinFigure 4-6 ,showingtwopotentialregionsofvariabilitywithshadedbars.Theseintervals arecenteredat 26250and 20400kms andare1900and1400kms wide, respectively.Thesignicanceofthevariabilityis9.1 # inthewiderintervaland6.9 # inthenarrowerinterval.Thesevaluesarewellabovethethresholdof4 # ,andthere ispotentialvariabilityinSi IV atthesamevelocitiesasinC IV .However,thereappear tobemismatchesinotherregionsofthespectra.Weconsideredthepossibilitythat theslopeofoneofthespectraneedstobeadjustedtobettermatchtheother.We checkedtheotherdatacollectedonthesamenightsasthesetwospectra,butfoundno evidenceforanyissueswiththeuxcalibrationateitherepoch.However,wemaintain ourconservativeapproachhereandkeepthismeasurementinthe"tentative"category. 109

PAGE 110

Figure4-6. Thefullspectraof1011 0906,showing2epochscoveringa tof0.05yr. Weclassifythismeasurementasatentativecaseofvariability.The rightmostshadedregionsmarkthecandidateintervalsofC IV variability,and theotherpairofshadedregionsmarkthecorrespondingvelocitiesinSi IV 4.2.2.40932+5006:Possiblevariabilityover31days Wepresenthereanotherexampleofatentativedetectionofvariability.Figure 4-7 showstheC IV andSi IV BALsforthe2008.03and2008.28observationsof0932+5006, whichareseparatedby0.09yr(31days).Theshadedregionidentiesthecandidate variableinterval,withawidthof1200kms ,whichisjustatthethresholddenedin Chapter 2 .Theuxdifferencewithinthisintervalis8.3 # ,andthereisalsopossible variabilityatthesevelocitiesinSi IV .Asinthecaseof1011+0906,however,thereare mismatchesinotherregionsofthespectra,inparticulartowardstheblueendofthe spectrainthebottompanelofFigure 4-7 .Giventhesmallamplitudeofvariabilityin anintervalthatisjustwideenoughtomeetourthreshold,combinedwithmismatches 110

PAGE 111

Figure4-7. SpectraoftheC IV (toppanel)andSi IV (bottompanel)BALsin0932 5006 fortwoepochsseparatedby0.09yr(31days).Thisisanothertentative measurementofvariability. 111

PAGE 112

Figure4-8. Thefullspectraof0903+1734,showing2epochswitha tof0.20yr.The fourrightmostshadedregionsmarktheintervalsofC IV variability,andthe othershadedregionsmarkthecorrespondingvelocitiesinSi IV inotherregionsofthespectra,weclassifythismeasurementasatentativecaseof variability. 4.2.2.50903+1734:Variabilityover72days Thevariabilitybetween2006.31and2007.05in0903+1734isoneofthemost prominentcasesofvariabilityontime-scaleslessthanthoseanalyzedinChapter 2 ,i.e., < 0.35yr.Theseobservationsareseparatedby0.20yr(72days),whichisjustslightly higherthantherangeoftime-scalesincludedinthefourthpointofFigure 4-9 .Figure 4-8 showsthesetwospectrawiththevariableintervalsinC IV ,andthecorresponding velocitiesinSi IV ,shaded.Thevariabilityherecoversawiderangeinvelocities,unlike 112

PAGE 113

Figure4-9. ThetoppaneldisplaysthefractionofmeasurementswheretherewasC IV BALvariabilityversusthetimeintervalbetweenthetwoobservations.The greentrianglesshowthevariabilityfractionswithtentativecasesofvariability included.Themiddleandbottompanelsshowthenumberofmeasurements andthenumberofquasars,respectively,thatcontributetoeachbin. theotherexamplesatshortertime-scalesinthissection.Chapter 3 containsafull discussiononthemulti-epochbehaviorofthisquasar. 113

PAGE 114

4.2.3Time-scalesforVariability Inthissection,weexamineC IV variabilityacrossallmeasuredtime-scalesfrom 0.02to8.7yr.WelookedatC IV BALabsorptionatallmeasuredvelocitiesinourentire datasample,andwecomparedeachpairofobservationsforeachquasarandcounted theoccurrencesofC IV BALvariability,usingourdenitionofBALvariabilitydenedin Chapter 2 (Section 4.1.2 ).Wethencalculatedaprobabilitybydividingthenumberof occurrencesofvariabilitybythenumberofmeasurements,whereapairofobservations isonemeasurement,inlogarithmicbinsof t.Weplotthismeasuredprobabilityof detectingC IV BALvariabilityversus tinFigure 4-9 .The1 # errorbarsarebased oncountingstatisticsforthenumberofoccurrencesofvariabilityandthenumberof measurementsineachbin( Wilson 1927 ; Agresti&Coull 1998 ). Figure 4-9 representsprobabilitiesonlyifeachmeasurementisanindependent event.Eachquasarinoursampleactuallycontributesmultipletimestothisplot. ThemiddleandbottompanelsofFigure 4-9 displaythenumberofmeasurements andthenumberofquasars,respectively,thatcontributetoeach tbin.Eachquasar cancontributemultipletimestoasinglebin.Furthermore,theseareprobabilitiesfor detectingvariabilitybetweentwomeasurementsseparatedby t,andnotfordetecting variabilityatanytimeinthat ttime-frame.Inotherwords,ifaquasarvariedthen returnedtoitsinitialstateallwithinacertain t,thenameasurementatthatvalueof t wouldnotcountasvariableinthisplot. Figure 4-9 furtherconrmsthecomparisonoftheshort-termandlong-termdatasets inChapter 2 inthattheincidenceofC IV BALvariabilityishigherwhen tislarger.The probabilityofC IV variabilityapproachesunitywhen t ) 10yrs.Thisdecreasesto % 0.6 for t % 1yrand % 0.05 0.1for t % 0.08yr( % 1month). Below t % 0.08yr,themeasuredprobabilityofvariabilitybeginstoincreaseas tbecomesshorter.Thesepointsmaybeslightlybiasedtohighervaluesbecause wetypicallychosetomonitorquasarsknowntobevariable,especiallywhenobtaining 114

PAGE 115

thenewdataattheshortesttime-scalesforthiswork(Section 4.1.1 ).Forthefour shortesttime-scalebins,where t < 0.2yr,wealsocalculatethismeasuredprobability ofvariabilitywiththetentativecasesofvariability,asdescribedinSections 4.1.2 and 4.2.2 ,included.TheresultingvaluesareplottedasgreentrianglesinFigure 4-9 .This tendstoattenoutthetrendintheshortesttime-scalebins.Thegreenandblackdotted curvesshowtheleast-squaresttothepointswithandwithoutthetentativecases ofvariabilityincluded.Overall,inclusionofthetentativecasesofvariabilitydoesnot signicantlychangetheslopeofthepoints. Becausethequasarsthatwereobservedmostfrequentlythroughoutourmonitoring programmeweretypicallyonesthatwereknowntobevariable,themeasured probabilitiesofvariabilityinFigure 4-9 mayinfactbebiasedtohighervaluesacross theentirerangeoftime-scales,andnotjustattheshorttime-scaleend.Thosequasars withthemostepochsofdatawillbecountedmuchmoreofteninFigure 4-9 .Wehave 13epochsofdatafor0842+3431and0932+5006,anduptojust10epochsofdatafor eachoftheotherquasarsinthesample.WethereforeshowinFigure 4-10 aversion ofthisplotwhichomitsthesetwoquasarswiththemostdata.Weplotaleast-squares ttothedatapointsinFig 4-10 (dashedcurve)andoverplottheleast-squarestfrom Figure 4-9 (dottedcurve)tocomparetheirslopes.Thisshowsthatthemostfrequently observedquasarsarenotsignicantlyaffectingtheslopeofthepoints,buttheydoshift thelinetoslightlyhigherprobabilities.Forsimplicity,wedonotincludethetentative casesofvariabilityinthisgure,orinFigures 4-11 and 4-12 below. EventhoughtheslopedoesnotchangesignicantlybetweenFigures 4-9 and 4-10 ,thereisonepointat tof2.4yrthatdoesdecreasesignicantlyinFigure 4-10 .In Figure 4-9 ,thereareonly6quasarsthatcontributemeasurementstothisbinbecause aspectrumfromSDSSisnecessaryforhavingameasurementattheseintermediate time-scales.WeonlyhaveSDSSspectrafor8outofthe24quasarsinoursample.By removing0842+3431and0932+5006,weareleftwithjust4quasarsinthisbincentered 115

PAGE 116

Figure4-10. SameasFigure 4-9 ,butwiththetwomost-observedquasars,0842+3431 and0932+5006,removed.Thedashedlineinthetoppanelisattothe datapoints,withthetfromFigure 4-9 overplottedasadottedline. 116

PAGE 117

Figure4-11. AversionofFigure 4-9 whereeachquasaronlycontributesoncetothe incidenceofvariabilityateach t,andthenweiterate1000timestogeta rangeofleast-squarets,shownherebytheshadedregion.Thedotted lineistheleast-squarestfromFigure 4-9 at2.4yr,andmorethanhalfofthemeasurementsofvariabilityinthisbininFigure 4-9 aremeasurementsof0842+3431and0932+5006. Thislackofdataatintermediatetime-scalesmayintroducefurtherbiasesinto Figure 4-9 .ThemajorityofourobservationsweretakenatLickbetween1988and 1992andatMDMandKPNObetween2007and2010.Theresultofthisisthatinstead ofrandomlysamplingthequasarlightcurves,wehavemanymoremeasurements 117

PAGE 118

clusteredattheshorter( < 1yr)andlonger( > % 4yr)time-scalesthanatintermediate time-scales.Forexample,incaseswherewehavejustLickandMDM/KPNOdata, anyvariabilitythatwedetectbetweenthelastLickobservationandtherstMDM observationcouldhaveoccurredwithinanytimeintervalduringthose % 4years. Furthermore,inthisscenario,iftherewere,forexample,2Lickobservationsand5 MDMobservationsforanobject,andtheonlyvariabilitythatoccurredinthisquasar happenedbetweenthelastLickandrstMDMobservations,thenthisquasarwould contribute10measurementsofvariabilitytothelasttwo tbins.Wethereforemaybe biasedtowardsgreatervariabilityfractionsatlongertimeintervalsoverintermediateand shorttimeintervals. ToexploreanybiasthatourunevensamplingmayintroduceintoFigure 4-9 ,we createdaversionofthisplotwhereeachquasarcanonlycontributeonemeasurement toeach tbin.Ifaquasarhasmultiplemeasurementsinaparticularbin,thenone ofthosemeasurementsischosenatrandomtobeincludedinthisplot.Wethen recalculatethisplot1000times,eachtimerandomlychoosingwhichmeasurements areincludedandcreatingaleast-squarestforeachiteration.InFigure 4-11 ,weshow theresultingrangeofleast-squarestsweobtain.Byiteratingmanytimes,weare attemptingtouncoverarangeofslopesthatincludeswhattheslopewouldbeinthe idealcaseofevensamplingacrosstheentirerangeoftime-scalescoveredhere.The rangeinslopesthatweobtainoverlapstheslopeofthepointsinFigure 4-9 ,shown hereagainbythedottedcurve,buttheslopestendtobeshallowerand/orshiftedto slightlylowerprobabilitiesacrossthefullrangein t.Thisindicatesthatwedoindeed haveaslightbiastowardsgreatervariabilityfractionsatlongertime-scalesinFigure 4-9 .Attheshortesttime-scales,thecurvefromFigure 4-9 isroughlyinthemiddleof therangeofcurvesinFigure 4-11 .Wementionabovethatwetendedtomonitorthose quasarsknowntovary,whichmightintroduceabiastowardsgreatervariabilityfractions 118

PAGE 119

Figure4-12. SameasFigure 4-9 ,buthereweonlyconsiderabsorptionthatlies between 25000and 3000kms ,thussatisfyingallofthe requirementsofthebalnicityindex.Thedottedlineisthettothedata pointsinFigure 4-9 ,andthedashedlineisattothedatapointsinthe currentgure. attheshortesttime-scales.However,itisnotclearfromFigure 4-11 whethersuchabias exists,butifitdoes,itiscertainlyweakerthananybiasatlongertime-scales. ThenalversionoftheplotofmeasuredprobabilityofC IV BALvariabilityversus time-scalethatwepresenthereinFigure 4-12 isthesameasFigure 4-9 ,butweonly 119

PAGE 120

considerthevelocitiesthatcontributetothemeasurementofBI,i.e., 25000to 3000 kms .Therefore,weonlyincludeabsorptioninthisplotthatareBALsaccordingto thestrictdenitionof Weymannetal. ( 1991 ),sothe2quasarsinoursamplewithBI=0, 0302+1705and0846+1540,arenotincludedatallhere.Restrictingtothisvelocity rangealsoremovesuncertaintyinthevariabilitydetectionduetothepresenceofa potentiallyvariablebroademissionline(BEL). ThedottedlineinFigure 4-12 representstheslopefromFigure 4-9 andthedashed lineisattothepointsinthecurrentgure.Theselinesshowthattherestrictionin velocityrangechangestheslopeofthepointsslightly;theslopebecomesshallower.In Chapter 2 ,wefoundthattheincidenceofvariabilityincreasesathigheroutowspeeds. Here,weareremovingthehighest-velocitymeasurements,sothatwillshiftthepoints tolowervalues.Ifthefractionofvariablemeasurementsthathavevariabilityonlyat velocities < 25000kms isthesameacrosstheentirerangeoftime-scales,thenthe pointsatlongertime-scaleswillbeaffectedmorethanthepointsatsmallertime-scales, wheretheoverallincidenceofvariabilityissmaller.Forexample,if20percentof variablemeasurementsatanytime-scalehavevariabilityonlyatthehighestvelocities, thenremovingthesemeasurementswoulddecreasetheincidenceofvariabilityfrom % 1 to0.8atthelongertime-scalesandfrom % 0.1to0.08attime-scalesaround0.1yr.This wouldthereforecausetheslopetobecomeshallowerasshowninFigure 4-12 WealsoinvestigatethecumulativedistributionofvariableBALquasars.InFigure 4-13 ,eachpointrepresentsthefractionofBALquasarswithvariableabsorptionat any tvalueuptothatpoint.Bythelargest tvalues,allthequasarsinoursample, exceptfor2225 0534,arecontributing.Wedonotinclude2225 0534atallinthisplot becauseweonlyhavedataovertime-scalesof t < 1yrforthisobject.Thelastpointin theplotthusrepresentsthefractionofallthequasarsinoursamplewithmulti-yeardata thatvariedbetweenanytwoepochs.Overmulti-yeartime-scales,thefractionofquasars thatvariedreaches % 90percent. 120

PAGE 121

Figure4-13. AcumulativedistributionofthefractionofquasarswithC IV variabilitywith time-scale. FromFigure 4-13 ,weseethatoverhalfthequasarsexhibitedC IV BALvariability overtime-scalesupto1yr,butthefractiondropssteeplybelow t % 0.3yr( % 3.5 months).Thisisroughlythelowertime-scalelimitofthe"short-term"datainChapter 2 AsinFigure 4-9 ,thegreentrianglesincludetentativecasesofvariability,asdescribed inSections 4.1.2 and 4.2.2 above.Withoutthesetentativecasesincluded,theblack datapointsactuallydecreaseinvalueupto t % 0.3yr.Thisisbecause,despitethe increasednumberofquasarscontributingas tincreases,thenumberofquasarswith 121

PAGE 122

variabilitydoesnotincreasebetweenthebinat0.04yrandthebinat0.3yr.Including thetentativedetectionsofvariabiityadds3morevariablequasarsbelow t % 0.3yr, decreasingthedrop-offinFigure 4-13 below0.3yr. 4.3SummaryandDiscussion Figures 4-9 4-13 showclearlythatmanyBALquasarsexhibitvariableabsorption. Wedetectatrendtowardsgreatervariabilityfractionsoverlongertime-scales,butthis resultmaybebiasedduetounevensamplingofthequasarlightcurvesandatendency tomonitorquasarsthatalreadyvaried.AsdiscussedinSection 4.2.3 ,mostofour observationsareclusteredatshorter( < 1yr)andlonger( > % 4yr)time-scales.Ifwe tendtomonitorquasarsalreadyknowntobevariable,thenthosequasarswillcontribute manymoremeasurementsat t > % 4yrthanthosethatdidnotvary.However,itis possiblethatifwehadmoreepochsofdataonthosequasarsthatdidnotvaryatall, thenwemaydetectvariabilityintheirBALsthatwemissed. ItisclearfromSection 4.2.3 thatthereisastrongtrendtowardslessvariabilityat shortertime-scales.ThistrendremainsstrongeveninFigures 4-10 4-12 ,wherewe attempttoaccountforanybiasespresentinouroriginalplotofmeasuredprobabilitiesof C IV BALvariabilityversus t(Figure 4-9 ).ThistrendisalsoreadilyapparentinFigure 4-13 ,whereweplotthefractionofquasarsinsteadofthefractionofmeasurements.Ata time-scaleof0.3yr,thefractionofquasarsthatexhibitedvariabilityatany t & 0.30yris % 0.5.Thisfractiondecreasesrapidlyto % 0.1 0.2attheshortesttime-scalesweprobe (downto0.02yr).Forcomparison,thefractionofallquasarsinoursamplethatvaried atanytime-scaleis91percent. Gibsonetal. ( 2008 )foundthat92percent(12/13) quasarsvariedintheir2-epochmulti-yearsample. Notonlydoestheincidenceofvariabilitydecreaseatshortertime-scales,butthe amplitudeoftheabsorptionstrengthchanges( )alsodecreases(Section 4.2.1 ).Over multi-yeartime-scales, canreach % 0.46.Therstmeasurementwehavewhere exceeds0.2isata tof0.27yr(100days).Attheshortesttime-scalesweprobe( t < 122

PAGE 123

0.20yr), doesnotexceed0.1.Thisisgenerallyconsistentwith Gibsonetal. ( 2008 ), whondadecreasein EWvaluesatshortertime-scales. Onegoalofthisworkistodeterminewhetherthereisaminimumtime-scalebelow whichthereisnovariability.Wendthatboththeincidenceandamplitudeofvariability decreaseatshortertime-scales.However,westilldetectvariabilitydowntothelowest tvaluesthatweprobe( t % 0.02yr,or8 10days;Section 4.2.2 ).Theamplitude changesonthisshorttime-scalearesmall( < 0.1),andtheyappearinjustsmall portionsoftheBALtroughs(inintervalsof from1300to2600kms ;Figures 4-2 4-5 inSections 4.2.2.1 4.2.2.2 ).Therefore,ifaminimumtime-scalethreshold exists,itmustbelessthan0.02yr. 4.3.1ImplicationsofVariabilityover < 0.20yr Thetime-scalesofvariabilityareimportantforconstrainingthelocationofthe outowinggas,butthespecicconstraintsaredependentonthecause(s)ofthe variabilityandthestructureoftheoutows.InChapter 3 ,wediscusstwoscenarios thatcouldproducetheobservedBALvariability:1)changesinthefar-UVcontinuum uxcausingglobalchangesintheionizationoftheoutowinggas,and2)outow cloudsmovingacrossourlines-of-sighttothequasarcontinuumsource.Wefound evidencetosupportbothscenarios.Coordinatedvariabilitiesinabsorptionregions atdifferentvelocitiesinthesamequasarsupportsthechangingionizationscenario, whereasvariabilityinnarrowportionsofBALtroughstsmorenaturallyinthescenario ofcrossingclouds.ThislatterscenarioisfavoredbypreviousworkonBALvariability, including Lundgrenetal. ( 2007 ), Gibsonetal. ( 2008 ), Hamannetal. ( 2008 ), Krongold etal. ( 2010 ),and Halletal. ( 2011 ).WeconcludeinChapter 3 thatingeneraltheactual situationmaybeacomplexamalgamofchangingionizationandtransversecloud movements. 123

PAGE 124

4.3.1.1Changingionization InthecaseofchangingionizationcausingvariabilityintheBALproles,one recombinationtime,givenby % / ,isnecessaryforadjustmentsintheionization ofthegas,where istheelectrondensityand istherecombinationratecoefcient. TherecombinationrateforC IV C III is2.8x10 ,foranominaltemperatureof % 20000K( Hamannetal. 1995 ).Forthecasesof1246 0542and0842+3431,wecan usethevariabilitytime-scaleof8 10days(Sections 4.2.2.1 and 4.2.2.2 )tosetalower limitonthedensityof > % 4 5x10 cm .Furthermore,photoionizationcalculations provideanestimateofthemaximumdistanceforthegasbasedontheluminosityofthe quasar,theionizationparameter,andthedensityofthegas.Basedonthecalculations of Hamannetal. ( 1995 )andtheconstraintonthedensityfromtherecombinationtime, themaximumdistanceis < % 200pc,assumingtheionizationisatleastashighasthat neededforamaximumC IV /Cratio. However,variabilityduetochangingionizationislesslikelyontheseshortest time-scalespresentedhere.Oursamplecontainsbrightquasars,andbrighterquasars haveasmalleramplitudeofcontinuumvariabilitythanfainterAGN.Furthermore, theshorterthe tbetweenobservations,thesmallertheamplitudeofcontinuum variability(e.g. VandenBerketal. 2004 ). Misawaetal. ( 2007 )detectedvariabilityin amini-BALoveratime-scaleof16days,andtheyalsostatethatvariabilityonsucha shorttime-scaleismuchfasterthananyexpectedchangesinthecontinuumemission ofaluminousquasar.Theyproposethatascreenofgas,withvaryingopticaldepth,is locatedbetweenthecontinuumsourceandtheabsorbinggas.Thisscreencouldbe theionizedshieldinggasintheoutowmodelsof Murrayetal. ( 1995 ).Ifthisscreen isclumpy,thenasitrotates,theintensityoftheionizingcontinuumthatistransmitted throughthescreenvaries.Sincethescreenisclosertothecontinuumsource,itwould berotatingatafastervelocitythantheabsorberandcouldthencausechangesinthe ionizationoftheabsorbinggasmorequicklythancontinuumvariationsalone. 124

PAGE 125

4.3.1.2Changingcoveringfraction Wenowconsiderasimpliedscenariowheretheabsorbershaveaconstant ionizationandcolumndensityandthevariabilityisduetoasingleoutowcomponent movingacrossthecontinuumsource.Wecanusethetime-scaleoftheobserved variabilitytoestimatethecrossingspeedofthiscomponentacrossourline-of-sight. Inordertocalculatethiscrossingspeed,wecalculateacharacteristicdiameterforthe continuumregionat1550 A,asdescribedinChapter 2 .Usingtheobserveduxat1450 A(rest-frame)inabsoluteux-calibratedspectrafromBarlow(1993),whileadoptinga cosmologywith kms Mpc, andastandardbolometric correctionfactor, ( ,wecalculatearangeinbolometricluminositiesfor thequasarsinoursamplefrom % to ergss .Basedontheaverage bolometricluminosity, % ergss ,acharacteristicdiameterforthecontinuum regionat1550 Ais % pcandfortheC IV BELregionis % pc, assuming / and % ( Petersonetal. 2004 ; Bentzetal. 2007 ; Gaskell 2008 ;Hamann&Simon,inpreparation). Usingthecharacteristicdiameterforthecontinuumregion,wecalculatethe crossingspeedoftheoutowcomponentfortwomodels:1)acircularcloudmoving acrossacircularcontinuumsource(the"crossingdisks"model),and2)arectangular cloudcrossingarectangularcontinuumsource(the"knifeedge"model).Figure 4-14 showsasimpleschematicillustratingthesetwomodels.Weplotthetransversevelocity asafunctionoftime-scaleontherightordinateinFigure 4-15 forachangeinBAL strength, ,of0.05and0.2.A of0.05isatypicalvaluefordataontime-scales < 0.1yr,andvaluesbetween of0.05and0.2aretypicalontime-scalesof0.1 3 yr(Figure 4-1 ).Whencalculatingthetransversevelocityinthe"crossingdisk"model, weareassumingtheabsorberissmallerthanthecontinuumsource,whichgivesthe maximumtransversevelocity.Alargerabsorberwouldgiveasmallervelocityforthe 125

PAGE 126

Figure4-14. Asimpleschematicofthetwomodelsusedtocalculatecrossingspeedsof anoutowinthechangingcoveringfractionscenario:( a )the"crossing disk"and( b )the"knifeedge"model.Thedarkerregionsrepresentthe outowmovingacrossthecontinuumsource(lightgrayregion)alongour line-of-sighttothequasar.Thearrowsrepresentthedistancetheoutow musttraveltocoverthesamefractionofthecontinuumsourceinboth models a and b 126

PAGE 127

Figure4-15. Transversevelocityanddistanceoftheoutowinggasversusrest-frame time-scaleforthetwomodelsillustratedinFigure 4-14 andtwotypical valuesof (Figure 4-1 ). samevalueof t,downtotheminimumtransversevelocity,givenbythe"knife-edge" model. IfweassumethatthetransversespeedisroughlyequaltotheKeplerianrotation speed,thenwecanestimatethedistanceoftheoutowcomponentfromthecentral SMBH.ThedistanceisshownastheleftordinateinFigure 4-15 .Figure 4-15 clearly showsthattheshorterthevariabilitytime-scale,theclosertheoutowcomponentis locatedtotheSMBH.Thisgurehighlightstheimportanceofobtainingmeasurements ofBALvariabilityovershortertime-scalesforconstrainingthelocationoftheoutow. Theshortesttime-scalesoverwhichwedetectvariabilityis0.02 0.03yr(8and10 days)in1246 0542and0842+3431,with of0.05 0.06.Thetransversevelocityof theoutowinthesecasesis13000to58000kms for1246 0542and13000to58 000kms for0842+3431,withthelowerandupperlimitsgivenbytheknife-edgeand crossingdiskmodels,respectively.Thesetransversevelocitiesrangefrom % 0.53 3.4 127

PAGE 128

timestheradialowvelocity.Thedistanceoftheoutowis % 0.002 0.04pcforboth quasars.Forcomparison,theradiusofthecontinuumsourceis0.003pc(seeabove). Iftheoutowsarebeingdrivenbyradiationpressurealone,asinthemodels of Murrayetal. ( 1995 ),theradialowspeedsshouldbegreaterthanthecrossing speeds.Thehighcrossingspeedswedetectherefortheshortestvariabilitytime-scales, especiallyinthecrossingdisksmodel,thusappearunlikelyforthesetheoretical models.Suchhighcrossingspeedsaremorelikelyifthewindisbeingdrivenby magnetocentrifugalforces,asinthemodelsof Everett ( 2005 ).Furthermore,ifthehigh transversespeedsarebeingdrivenbymagneticforces,thenthecrossingspeedsmay behigherthantheKeplerianspeedattheradiusofthegas.Thismeansthattheoutow couldbelocatedatagreaterdistancethanindicatedbyFigure 4-15 PreviousstudieshavefoundevidenceforabsorberslocatedbeyondtheBELregion. Turnshek ( 1988 )describecaseswhereaBALoverlapsaBELinthespectraandthe absorptiongoesdeeperthancanbeexplainedbytheBALabsorbingjustthecontinuum underneaththeBEL.Thisindicatesthat,atleastfortheseparticularcases,theBAL regionoccultstheBELregion.TheC IV BELgasislocatedatadistanceof % 0.15pc inoursample,andtheshortestvariabilitytime-scalesinoursamplegiveradiiforthe absorberswithinthisradius(Figure 4-15 ).Forthetwocasesherewithvariabilityover % 0.02yr,1246 0542and0842+3431,theabsorptionlinesaredetachedfromthe emissionlines,sowecannotdeterminefromthespectrawhethertheBALregionoccults theBELregion.Wethereforecannotruleoutthatthedistanceoftheoutowsinthese casesareindeedwithintheBELradius. 4.3.2OtherPotentialCausesofVariability ThereareotherpossiblecausesforthevaribilityinquasarBALs.Forexample, achangeinthesizeofthecontinuumsourcewouldcauseachangeinthecovering fractionofsaturatedabsorptionlines.Themainproblemwiththisscenario,though,is 128

PAGE 129

thatitshouldchangethecoveringfractionofanentiretroughandnotjustsmallportions thereof.Mostofthecasesofvariabilityinoursampleoccurinjustportionsoftroughs. Asmentionedearlier,typicalvariationsinthecontinuumluminosityofabright quasarisunlikelytocausechangesintheabsorbinggasontheshortesttime-scaleswe probehere.Anotherpotentialsolutiontothisproblemishotspotsonaninhomogeneous accretiondisk(e.g. Dexter&Agol 2011 )whichcouldappearanddisappearonashorter time-scale.This,however,shouldalsocauseglobalchangesacrossentireBALtroughs. Onelastpossibilitythatweconsiderhereisinstabilitieswithintheowsthemselves, asinthesimulationsof Progaetal. ( 2000 ).Thetime-scaleforsuchinstabilities signicantlyaffectingtheoutowstructureissimilartotheoutowdynamicaltime. Ifaowhasaradialvelocityof15000kms andislocatednominallyaroundthe broad-emissionlineregion( % 0.15pc),itwouldtake % 10yrtoreachtwicethatradius. Thisdynamicaltime-scaleismuchlargerthanmanyofthemeasurementsofvariability time-scalesinthiswork. 129

PAGE 130

CHAPTER5 AVARIABLEP V BROADABSORPTIONLINEANDQUASAROUTFLOW ENERGETICS TherehavebeennumerousstudiesofBALsinquasarspectra,butinorderto determinetheviabilityofBALoutowsasafeedbackmechanism,weneedestimates oftheirmassoutowratesandkineticenergyyields.Thesequantitiesdependonthe columndensitiesoftheows,whicharedifculttocalculatedirectlyfromBALquasar spectrabecausetheabsorptionlinescanbesaturated(e.g. Aravetal. 1999 ). Onewaytoovercomethissaturationproblemistosearchforabsorptionin low-abundanceions,suchasP V .P V isagoodchoicebecauseithasareasonance doubletatacccessiblewavelengths,1118and1128 A,anditsionizationissimilar tomuchmoreabundantandcommonlymeasuredionssuchasC IV .P V absorption shouldbepresentifthecolumndensitiesintheoutowsarelargeenough.Inthe sun,theabundanceratioofphosphorusrelativetocarbonis0.001( Asplundetal. 2009 ),whereasthepresenceofaP V BALindicatesanabundanceratioofP/C > % 0.04 (e.g. Turnshek 1988 ; Junkkarinenetal. 1997 ; Hamann 1998 ).Theseabundancesare difculttoexplainandarelikelyincorrect.Instead,iftheBALsareopticallythickand justpartiallycoverthebackgroundcontinuumsource( Hamann 1998 Hamannetal. 2002 ,andreferencestherein),thentherelativeabundanceofphosphorustocarbon, forexample,canbemuchlowerthantheapparentopticaldepthsofthelineindicate. Ifassumptionsaremadeaboutthetrueabundancesintheoutow,thenconstraints canbeplacedonthetrueopticaldepthsandcolumndensities.Forexample,ifthe abundancesareapproximatelyequivalenttosolarabundances,thentheopticaldepth inP V shouldbeatleast % 50timeslessthanC IV ( Hamann 1998 ; Hamannetal. 2002 ). Thus,evenaveryweakP V lineindicatesaverysaturatedC IV BAL. Inthiswork,wedetectaP V BALinQ1413+1143.ThewavelengthofP V places itintheLy forest.Ourmulti-epochdatashowthattheabsorptionlinevaried,which conrmsthatthislineisindeedP V andnotacoincidentalbroadblendofunrelated, 130

PAGE 131

Table5-1. Observationsummary YearTelesc.Resol. t (kms )( A)(yr) 1989.26Lick3-m5303793 6382... 2006.31SDSS2.5-m1503803 92214.8yr 2010.11KPNO2.1-m2753482 62325.9yr interveningLy absorptionlines.ThevariabilityintheP V BALcorrespondstovariable Si IV 1400andC IV BALs,asdetectedinChapters 2 4 .TheP V detectionindicates thattheC IV lineissaturated,providingfurtherevidenceforcloudmovementsasthe causeofvariability,atleastinthisparticularquasar,asdiscussedbelow. Inthischapter,weusetheapparentopticaldepthoftheP V BALinQ1413+1143 andpreviousresultsfromphotoionizationmodelstoconstrainthetruecolumndensity oftheoutowinthisquasar.Withtheseconstraintsonthecolumndensityandonthe location(fromthevariabilityintheBALs)weestimatethekineticenergyinthisoutow andcomparetosimulationstoestimateitsviabilityasafeedbackmechanism. Section 5.1 describesthedataandpropertiesofQ1413+1143,Section 5.2 discussesthevariabilityintheP V BALanddescribesourmethodforconstrainingthe columndensityoftheoutowinthisquasar,andSection 5.3 discussesourestimatesof theenergeticsofthisowandtheirimplications. 5.1Data ThedataforQ1413+1143comesfromthevariabilitysampledescribedinChapters 2 4 .Wehave7epochsofdataforthisquasar,butonly3haveawavelengthcoverage thatextendsblueenoughtodetecttheP V BAL.Table 5-1 showsasummaryofthese data.Therstcolumngivesthefractionalyearsofthesethreeobservations.The nextthreecolumnslisttheobservatories,thespectralresolutionofthedata,andthe wavelengthcoverage.Thenalcolumnliststherest-frametime-scale, t,incomparison totherstobservation,from1989.26.Therstspectrumisfrom Barlow ( 1993 )andwas takeninApril1989.AsdescribedinChapter 2 (Section 2.1 ),wehavedatafromtheLick 131

PAGE 132

3-mattwodifferentresolutions.Thespectrumthatweuseherewastakenatthelower resolution,butitissufcientforstudyingbroadlines.Thesecondspectrumpresented hereisfromtheSloanDigitalSkySurvey( Adelman-McCarthyetal. 2008 )andwas takeninApril2006.ThemostrecentspectrumwastakeninFebruary2011attheKPNO 2.1m.SeeSection 2.1 formoredetailsontheLickandSDSSobservationsandSection 4.1.1 formoredetailsontheKPNOdata. Theemission-lineredshiftofQ1413+1143is =2.5630,asreportedbytheSDSS. Weestimatethebolometricluminosityandblackholemassusingthemethoddescribed inSection 2.4 .Thisquasarhasthehighestbolometricluminosityofthequasarsinour variabilitysampleat ergss ,whichisequivalenttothevaluecalculated by Lutzetal. ( 2007 ),andweapproximatetheblackholemassas ( M % 5.2AnalysisandResults 5.2.1VariabilityinaP V AbsorptionLine InthevariabilitystudydescribedinChapters 2 4 ,wefoundthatQ1413+1143 variedinbothC IV andSi IV .Uponfurtheranalysisofthespectrawehaveforthis quasar,wefoundaP V absorptionlinewithvariabilityatvelocitiescorrespondingtothe variabilityinbothC IV andSi IV .Outofthe7epochsofdatawehaveforthisobjectin ourBALmonitoringprogramme,3ofthespectraextendblueenoughtoshowthisP V absorption(Table 5-1 ). WeadopttheLickspectrumhereastheducialspectrum,forwhichweta power-lawtothecontinuuminregionsfreeofemissionandabsorption.Wealsot gaussianstotheC IV andSi IV broad-emissionlines(BELs)togetasmoothttothe proles.ThedottedcurveinFigure 5-1 showsthepseudo-continuumt,combining thepower-lawcontinuumtandthetstotheSi IV andC IV BELs.Moredetailsonthe procedurefordeningthepseudo-continuumtcanbefoundinChapter 2 .Tocompare thedifferentepochs,weuseasimplescalingtomatchthespectraalongregionsfree 132

PAGE 133

Figure5-1. FullspectrumofQ1413+1143,showingtheearlyLickepochandthemore recentKPNOspectrum.Themiddleshadedregionmarkstheintervalof variabilityinSi IV inthebluesideofthetrough,andtheleftmostand rightmostshadedregionsshowthecorrespondingvelocityintervalsinP V andC IV ,respectively.Theformal1 # errorsareplottedacrossthebottom. ofemissionandabsorption.Figure 5-1 showstheLickspectrumfrom1989.26(black curve)withtheKPNOspectrumfrom2010.11overplotted(bluecurve). Tocomparethevariabilityinthedifferentlines,wemarktheregionofvariabilityin thebluesideoftheSi IV troughwithashadedbar.Wethenmarkthecorresponding velocitiesintheP V andC IV BALswithshadedbars,withthewidthsofthebars adjustedforthedifferentdoubletseparationsinthelines.C IV hasthenarrowestdoublet separationat498kms ,andSi IV andP V havewiderdoubletseparationsof1930and 2670kms ,respectively.Figure 5-2 showstheC IV ,Si IV ,andP V prolesforallthree epochslistedinTable 5-1 .The1989.26and2010.11areagainshownwithblackand 133

PAGE 134

Figure5-2. TheBALlineprolesfor,fromtoptobottom,C IV ,Si IV ,andP V .Theshaded regionsheremarkthesameintervalsasinFigure 5-1 134

PAGE 135

bluecurves,respectively,andtheredcurveistheintermediate2006.31observation. Thevelocityscalesinthethreepanelsaredenedbasedonthewavelengthoftheblue doubletmemberforeachion. Figure 5-2 showsclearlyhowtheprolesofthesethreelinesmatchinvelocity space.ThemaindifferencehereisthatC IV hasalongerwingonthebluesideofthe proleandthevariabilityextendstohigheroutowvelocitiesthaninSi IV andP V .We alsonotethatinFigure 5-1 thereappearstobevariabilityinN V at % 1140 1160 A, correspondingtothevariabilityinthewingoftheC IV line. ThevelocitieswherevariabilityoccurredinthebluesideoftheSi IV troughandin theP V troughcloselymatch.TheP V linealsovariesinthesamesenseasC IV and Si IV ;allofthelinesgetweakerinthebluesideofthetroughs.ThechangesintheP V lineproleoccurbetweentheLickandtheSDSSobservations,whichareseparatedby 4.8yrintherest-frameofthequasar.TheP V linedidnotvarybetweentheSDSSand KPNOspectra,whichareseparatedby1.1yr,buttheC IV linedidvary.InChapter 4 ,we foundvariabilitydownto0.25yr(91days)inC IV between2009.22and2010.11. 5.2.2LineOpticalDepthsandIonicColumnDensities Fortheremainderoftheanalysis,wefocusonjusttheLick1989.26observation.In ordertoestimatethecolumndensitiesoftheoutowinggas,werstderivetheoptical depthversusvelocity, $ ,fortheC IV ,Si IV ,andP V BALs.Toderive $ ,wenormalize theLickspectrumusingthepseudo-continuumtshowninFigure 5-1 Withthespectrumnormalized,forsimplicity,wetgaussianstotheC IV andSi IV BALproles.Thesegaussiansdonotrepresentanyphysicalpropertiesofthegas; theyaresimplyusedtodeneasmoothttotheproles.Wethenusetheequation, ,toderive $ fortheC IV andSi IV lines.Thelinesweanalyzehere,C IV Si IV ,andalsoP V ,arealldoublets,soderiving $ inthiswayincludescontributions frombothtransitionsateachvelocity.WethereforeuseEquation(2)from Junkkarinen 135

PAGE 136

Figure5-3. ThenormalizedBALlineprolesfor,fromtoptobottom,C IV ,Si IV ,andP V fortheLick1989.26observation(boldcurves).Thethincurvesshowthe gaussiantstotheC IV andSi IV prolesinthetoptwopanels.Inthebottom panel,thethincurveisthettoP V usingtheopticaldepthproleforSi IV (Figure 5-4 ). 136

PAGE 137

Figure5-4. TheopticaldepthprolesforC IV ,Si IV ,andP V .TheC IV andSi IV proles arederivedfromthegaussiantstotheresidualintensitiesintheBALsas showninFigure 5-3 ,andtheP V proleisderivedbasedontheSi IV optical depthprole. etal. ( 1983 )toremovethedoubletstructure.ThesolidanddottedcurvesinFigure 5-4 showtherunofopticaldepthsversusvelocity, $ ,forC IV andSi IV ,respectively. Toestimate $ forP V ,wersttaketheopticaldepthprolesforC IV andSi IV andbroadentheprolesbasedontheincreaseddoubletseparationinP V .Wethen calculate fromthisadjusted $ ,andttheresultingC IV andSi IV intensityproles totheP V BAL.TheSi IV proleprovidedthebestttotheP V line,withasmall adjustmenttotheamplitude.WethencalculatedtheP V opticaldepthproleusingthis adjustedSi IV proleand .Weagainusedthemethodof Junkkarinenetal. ( 1983 )toremovethedoubletstructure,andtheresulting $ proleforP V isshownin Figure 5-4 asthedashed-dottedcurve. 137

PAGE 138

WederivecolumndensitiesandabundanceratiosforC IV ,Si IV ,andP V directly fromthe $ prolesshowninFigure 5-4 .Theioniccolumndensitiesthatwederive, atleastforC IV andSi IV ,mostlikelyunderestimatethetruecolumndensitiesbecause thelinesareprobablyverysaturated.Wecalculatethesevaluesforreferenceusingthe followingequation: % $ where istheoscillatorstrengthand isthelaboratorywavelength( Hamann& Ferland 1999 ).Weobtainthefollowingvaluesfortheioniccolumndensities:N % cm ,N % cm ,andN % cm IfweignoresaturationeffectsintheBALsandacceptthesecolumndensitiesat facevalue,wecouldusethemtoestimatetherelativeabundancesofC,Si,andP.The relativeelementalabundancesarecalculatedusing # $ % & % & % & % (51) % & (52) where / % isthesolarabundanceratio,and and arethecolumndensitiesand ionizationfractions,respectively,ofelements and inionstages and .ICisthe normalizedcorrectionfactor. Hamann ( 1998 )determinedthattheminimumvalueof ICforP V /C IV is % 3.1foranyionizationstateandanynominalcontinuumshape.P V existsinanarrowerrangeofionizationsthanC IV ,andthisminimumvalueofICisfor theionizationstatewhereP V ismostabundant.Atotherionizationstatesinthegas, P V /C IV islowerandthereforethecorrectionfactorislarger.Withtheioniccolumn densitieswecalculatedfromourdataandthisminimumvalueofIC,wecalculatea lowerlimitonP/Cof % 0.36.Thus,theabundanceratioofphosphorustocarbonthat wecalculatedirectlyfromthedataisenhancedbyafactorof > 360relativetosolar abundances.However,amuchmorelikelyinterpretationisthattheSi IV andC IV 138

PAGE 139

BALsarehighlysaturated,andallofthederivedcolumndensitiesarejustlowerlimits. Therefore,thetruephosphorus-to-carbonabundanceratiowouldbemuchlowerthan calculatedabove. Hamann ( 1998 )discussesthedetectionofP V absorptioninPG1254+047.They usethephotonionizationcodeCLOUDYtocalculatelineopticaldepthsand for constantdensity,photoionizedclouds,assumingsolarabundances.Theyadoptan opticaldepthof0.2fortheP V BALbasedontheirmeasurementsofthelineoptical depthsinPG1254+047.Inthismodel,theratioofthetrueopticaldepthinC IV versus P V is $ / $ > % 50.AveragingovertheopticaldepthproleforP V in1413+1143,as showninFigure 5-4 ,givesavaluefortheapparentopticaldepthof % 0.7,indicatingthat thetrueopticaldepthofC IV isatleast % 35,iftheabundancesareroughlysolar. 5.2.3TotalColumnDensity Toconverttheioniccolumndensitiesintoatotalcolumn, ,weadoptsolar abundancesandapplyionizationcorrectionsbasedonthephotoionizationcalculations in Hamann ( 1998 ). Hamann ( 1998 )calculatethecolumndensityversus forPG 1254+047.WedetectanO VI BALintheKPNO2010.11spectrum,withroughlythe samestrengthasC IV (theotherspectradonotextendtoshortenoughwavelengths tocoverO VI ).O VI isahigherionizationlinethantheotherlineswestudyhere,sothis givesanapproximatelowerlimiton .Furthermore,theopticaldepthinP V islarger for1413+1143byafactorof % 3.5thanthevalueusedinthecalculationsof Hamann ( 1998 ).Forvaluesoflog rangingfrom 1to1,thisputsalowerlimiton of % cm ,butthetruevalueislikelyashighas % cm ( Hamann 1998 ). 5.3Discussion 5.3.1LocationoftheOutow ThedetectionofaP V BALin1413+1143indicatesthattheC IV BALisvery saturated.InChapter 3 ,wediscussthetwomostcommonlycitedscenariosfor explainingthevariabilityinBALs.Therstscenarioisachangeinthequasar's 139

PAGE 140

continuumuxcausingglobalchangesintheionizationoftheoutowinggas.However, weestimateherethattheopticaldepthintheC IV lineisatleast35,soitistoosaturated tobesusceptibletochangesinionization. Instead,theresultspresentedheresupportthescenarioofcloudsmovingacross ourline-of-sighttothequasarcontinuumsource.Ifcrossingcloudsdidindeedcausethe variabilitywedetect,thenwecanconstrainthelocationofthegasusingthemethod describedinChapter 4 ,Section 4.3 .WerepeatthecalculationfromSection 4.3 usingtheblackholemass, ( M % ,for1413+1143(Section 5.1 ).The minimumvariabilitytime-scalefor1413+1143,basedonourmulti-epochdataoftheC IV BAL,is0.25yrandthevariableintervalhada % 0.09(Section 5.2.1 ).Thisgivesa distanceestimateof % 0.15 1.5pc(dependingonthesizeoftheabsorberrelativetothe continuumsource;Figure 4-14 ).Forcomparison,theradiusofthebroademission-line regionforthisquasaris % 0.3pc(Hamann&Simon,inpreparation). 5.3.2EnergeticsoftheOutow Theenergeticsofquasaroutows,andthustheirpotentialtosignicantlyaffectthe hostgalaxy,isstillpoorlyunderstood.Itisdifculttoconstrainthecolumndensitiesof BALoutowsbecausethebroadlines,particularlyinthemostwell-studiedBAL,C IV hidethedoubletstructure.Therefore,inmostcases,weobtaindirectlyfromthespectra justalowerlimitontheopticaldepthsandthecolumndensitiesoftheoutowinggas. For1413+1143,wehavebeenabletoconstrainthetruecolumndensityoftheow usingtheexistenceofastrongP V BALandthemodeldescribedin Hamann ( 1998 ), whichassumesthattheelementalabundancesintheowareroughlysolar(Section 5.2.3 ).Usingtheseconstraintson ,weestimatethemassoftheow,themass-loss rate,andthekineticenergyluminosity.Ifweestimatetheowgeometryaspartofathin sphericalshell,thenthetotalmassis ( % &% &% & % 140

PAGE 141

where isthedistanceestimatefromSection 5.3.1 and istheglobalcovering fractionoftheoutow,i.e.thefractionof4 % steradiansthattheowcoversfromthe point-of-viewofthecentralcontinuumsource( Hamann 2000 ).Weadoptavalueof % 15%basedonthefractionofquasarsthatexhibitBALsintheirspectra( Hewett&Foltz 2003 ; Reichardetal. 2003 ; Trumpetal. 2006 ; Kniggeetal. 2008 ; Gibsonetal. 2009 ). Forthelowerlimitsonboththecolumndensity( % cm ;Section 5.2.3 )and thedistance( % 0.15pc;Section 5.3.1 ),themassoftheoutowis3.8M % .Wefound thatthecolumndensitycouldbe > % cm ,andwecalculateanupperlimitonthe distanceof % 1.5pc.Together,thesevaluesimplyanapproximateupperlimitonthe massoftheowof3800M % Toestimatethemass-lossrates, ,wedividethemassoftheowbya characteristicowtime, % / .Weadoptanominalvalueof10000kms for thevelocityoftheowbasedonthevelocityofthevariableintervalidentiedinFigures 5-1 & 5-2 .Atdistancesfrom % 0.15 1.5pc, is % 15 150yr.Therefore,fortherange inmasseswecalculatefortheow, % 3.8 3800M % ,themass-lossrateis0.25 25 M % yr .Forcomparison,weestimatetheblackholemassaccretionratewiththe equation & ,where & ,theefciency,isnominally0.1( Peterson 1997 ).Forthe luminosity, ,weusethebolometricluminosityof % ergss (Section 5.1 ), whichgivesamassaccretionrate, ,of46M % yr .So,themass-lossratethrough theoutowis % 0.5 50%ofthemassaccretionrateontothecentralblackhole. Thekineticenergyoftheoutow,asdenedby / ,is ( % % & ( Dividingbythecharacteristicowtime, ,givesatime-averagedkineticluminosityof # $% to ergs forthefullrangeindistanceandoutowmass. Theratioofthiskineticenergyluminositytothequasarbolometricluminosityistherefore # $ / % .Thesevaluesaremorethananorderofmagnitudeless 141

PAGE 142

thantheratioof # $ / % thatistypicallycitedastherationecessaryforanoutow tobeimportantforfeedback( Scannapieco&Oh 2004 ; DiMatteoetal. 2005 ; Prochaska &Hennawi 2009 ).However, Hopkins&Elvis ( 2010 )determinethataratioof # $ / % 0.005couldbesufcienttoaffectthehostgalaxyevolutionviafeedback.Theupperlimit oncolumndensityanddistancethatweadoptheregivesavalueof # $ / of0.003, whichisroughlyconsistentwiththisratiofrom Hopkins&Elvis ( 2010 ). Forreference,iftheoutowisjustslightlyfurtherfromthecentralsource,at,for example,5pc,thenthemass-lossrateswouldbe8.9 89M % yr (20 200%ofthe massaccretionrate),dependingonthecolumndensity, ,oftheow.Themassand kineticenergyoftheowwouldbe % 4200 42000M % and % erg. Thisleadstoatime-averagedkineticenergyyieldof # $ / % 0.001 0.01.Therefore, theoutowin1413+1143isagoodcandidateforafeedbackmechanismbetweenthe quasarandthehostgalaxy. 5.4FutureWork Thecurrentstudyprovidesafoundationforfurtherworkontheenergeticsofquasar outows.WeprovideevidenceherethatthedetectionofP V absorptionindicates massiveowswithhighcolumndensitiesandpotentiallylargemassoutowratesand kineticenergyyields.Theevidencepresentedheresuggeststhat,atleastsome,BAL outowscouldbeimportantforfeedbacktothequasarhostgalaxy.Thenextstepinthis studyistoinvestigatetheoverallincidenceofP V absorptioninBALquasars. Weareplanningtoapproachthisquestionintwoways.First,wewilllookthrough existingarchivaldata,inparticularHubbleSpaceTelescope(HST)UVspectraofBAL quasars.ForthosethathavecoverageatleastasblueastheP V line,wewilllook foroccurrencesofP V absorptionandforclearnon-detectionsofP V .Becauseofthe locationofP V intheLy forest,therecouldbesomecasesthathaveaP V line,butit istooweaktodifferentiateitfrominterveningLy absorption.Wewouldthereforegeta lowerlimitontheincidenceofP V absorption. 142

PAGE 143

Second,weplantousetheongoingSDSS-IIIBaryonOscillationSpectroscopic Survey(BOSS),whichistargeting > quasarsatredshiftsof > 2.Aredshiftofz > 2.2isnecessarytoshiftP V intotheobservedvisiblespectrum.Thelargequasar sample,aswellasgreaterbluesensitivitythanSDSS-I/II,maketheBOSSdatasetideal forthisstudy.ThishigherredshiftrangemightmakedetectingP V moredifcultthanfor theHSTdataoflowerredshiftquasarsbecauseoftheincreasedamountofintervening Ly absorption,butthequasarinthiswork,1413+1143,hasaclearlydetectableP V line andhasaredshiftof % 2.6.Butagain,wewillbesettingalowerlimitonthetruefraction ofBALquasarswithP V absorption. WithasampleofP V detectionsandclearnon-detections,wecansearchfor correlationsbetweentheP V detections,whichindicateoutowswithlargecolumn densities,andpropertiesofthequasarsthatdrivetheseows,i.e.thequasarluminosity, theblackholemass,theeddingtonratio( / )andtheshapeoftheSEDs.Wewould alsolookforcorrelationsbetweenP V detectionsandotherpropertiesoftheows themselves,suchasX-rayabsorptionandtheprolesandvelocitiesintheC IV and Si IV BALs,andpropertiesofthebroademissionlinesthatmightformatthebaseof theseoutows(e.g.,thebroadC IV emissionprole,redshiftandstrength).Asampleof strong,deniteP V detectionsandrmP V non-detectionswillalsoallowustoapplyfor targetedfollow-upobservationsintheUVandX-raystoobtainevenbetterconstraints ontheoutowionizations,totalcolumndensities,energeticsanddrivingmechanisms. ThiswillfurtherourunderstandingofquasarBALoutowsandtheirpotentialroleasa feedbackmechanism. 143

PAGE 144

CHAPTER6 SUMMARYANDCONCLUSIONS Inthiswork,Ihavedescribedtheresultsofamonitoringprogrammeof24luminous broadabsorptionlinequasars.InChapter 2 ,IfocusedonjusttheC IV 1550lineintwo differentrest-frametimeintervals,andIfoundthat65percent(15/23)ofthequasars variedinthelong-termintervalof3.8 7.7yr,whilejust39percent(7/18)variedin theshort-termintervalof0.35 0.75yr.Ialsoreportasigncantcorrelationbetween incidenceofvariabilityandbothoutowvelocityandabsorptionstrength.Weaker featuresathighervelocitiesaremostlikelytovary. Chapter 3 containsadetailedstudyofthevariabilityinSi IV 1400BALsin comparisontothevariabilitypropertiesofC IV fromChapter 2 .IfoundthatSi IV BALs aremorelikelytovarythanC IV .Inthelong-termsubsampledenedinChapter 3 thereareclearcasesofSi IV variabilitywithnocorrespondingC IV variability.By comparison,thereisonlyonetentativecaseinthelong-termdatawhereC IV variability isnotmatchedbySi IV variabilityatthesamevelocity.Wedohoweverhaveoneclear caseofC IV variabilityinChapter 4 (Section 4.2.2.1 )withnocorrespondingSi IV variability.Despitethiscleardifferenceintheincidenceofvariability,thereisnoapparent correlationbetweenthechangeinstrengthinC IV andSi IV invelocityintervalswhere theybothvary.Although,thereisaweaktrendforgreaterfractionalchangeinstrength inSi IV Ialsohighlightthemulti-epochbehaviorofBALvariabilityinChapter 3 .Figures 3-8 and 3-9 showthebenetofourlarge,multi-epochmonitoringcampaign.Inparticular, thevariabilityinBALsisnotclearlymonotonic,butrathertheBALsinthesamequasar canbothweakenandstrengthenovertime.Theycanevenvaryandthenreturnto anearlierstate(Figures 3-8 and 4-2 ).Furthermore,Figures 3-8 and 3-9 showhow variabilitycanbeconnedtocertaindiscretevelocityintervals,evenwhenthevariability occursovermanyepochs.Finally,whenthereisvariabilityindifferentvelocityintervals 144

PAGE 145

inthesameion,theseintervalstypicallyvaryinthesamesense(theyeitherallget strongerorweaker).Furthermore,whenthereiscorrespondingvariabilityinC IV and Si IV ,thosechangesalwaysoccurinthesamesense. InChapter 4 ,Itakeacloserlookatthetime-scalesofvariability.Boththisand previousworkhavefoundmanycasesofBALsvarying,indicatingitisacommon occurrence.Inthiswork,weshowthatasthetimeintervalbetweentwoobservations approaches % 10years,theprobabilityofaBALquasarvaryingapproachesunity.Atthe shorttime-scaleend,despiteprobingtime-scalesdownto % 1weekinourmonitoring campaign,wedonotndaminimumtime-scalethresholdbelowwhichthereisno variability.Infact,wendcasesofvariabilitydowntotimeintervalsof8and10days. Themaingoalofstudyingthevariabilityinquasarabsorptionlinesistoobtain constraintsonpropertiesoftheoutowsthemselves.Suchconstraintsdependon thecause(s)ofthevariability.InChapter 3 ,wehaveadetaileddiscussioncomparing theevidenceforandagainstthetwomostlikelycausesofthevariability:achangein ionizationorachangeincoveringfractionduetocloudscrossingourline-of-sighttothe quasar.Neitherofthesescenarioscanberuledout,andtherealitymaybeacomplex combinationofthetwo. Ontheshortesttime-scalesdiscussedinChapter 4 ,achangeinionization becomeslesslikelybecausethemagnitudeofcontinuumvariabilitydecreaseswith decreasingtime-scale.Themagnitudeofcontinuumvariabilityisalsosmallerin luminousquasarsliketheonesweincludeinourstudy.Iftheabsorptionlinechanges attheshortesttime-scalesinourdatasetareduetocrossingclouds,thenwecan constrainthedistanceoftheseoutowcomponentsdowntosub-parsecscales,possibly withinthequasarbroademissionlineradius. InChapter 5 ,IgobeyondthevariabilitystudyofChapters 2 4 toinvestigatethe energeticsoftheseBALoutowsandwhethertheycouldbeimportantforfeedbackto thehostgalaxy.Weusethedetectionofalow-abundanceion,P V 1118,1128,inone 145

PAGE 146

quasarinoursample,1413+1143,toestimatethecolumndensityinitsoutow.The C IV andSi IV linescanbesaturated,butthedetectionofP V givesusanestimateof thetrueopticaldepthsintheselines,byassumingsolarabundances.Usingthemodel of Hamann ( 1998 ),weestimatethetotalcolumndensityintheow.Thehighoptical depthsinC IV andSi IV indicatethatcrossingcloudsisthelikelycauseofthevariability intheBALsin1413+1143,soweconstrainthedistanceoftheowusingthemethod wedescribedinChapter 4 .Withthetotalcolumndensityandthedistanceoftheow, wecalculatethemass,themass-lossrate,andthekineticenergyoftheow.Wend thatthekineticluminosityoftheowisupto % 0.003ofthebolometricluminosityofthe quasar.Thisisroughlyconsistentwiththeratioof0.005determinedby Hopkins&Elvis ( 2010 ),indicatingthattheoutowin1413+1143couldbeimportantforfeedbacktoits hostgalaxy. AttheendofChapter 5 ,Idescribehowweplantocontinuetheworkpresented here.SearchingformoredetectionsofP V withinBALspectraandstudyinghowP V detectionscorrelatewithotherpropertiesofthequasarsandtheoutowswillhelpusto betterunderstandtheseoutowsandtheirenergetics. 146

PAGE 147

REFERENCES Adelman-McCarthyJ.K.,etal.2008,ApJS,175,297 AgrestiA.,CoullB.A.,1998,TheAmericanStatistician,52,119 AravN.,BeckerR.H.,Laurent-MuehleisenS.A.,GreggM.D.,WhiteR.L.,Brotherton M.S.,deKoolM.,1999,ApJ,524,566 AravN.,KoristaK.T.,deKoolM.,JunkkarinenV.T.,BegelmanM.C.,1999,ApJ,516, 27 AravN.,MoeM.,CostantiniE.,KoristaK.T.,BennC.,EllisonS.,2008,ApJ,681,954 AsplundM.,GrevesseN.,SauvalA.J.,ScottP.,2009,ARA&A,47,481 BarlowT.A.,1993,PhDthesis,AA(CaliforniaUniv.) BarlowT.A.,JunkkarinenV.T.,BurbidgeE.M.,WeymannR.J.,MorrisS.L.,Korista K.T.,1992,ApJ,397,81 BarlowT.A.,SargentW.L.W.,1997,AJ,113,136 BentzM.C.,DenneyK.D.,PetersonB.M.,PoggeR.W.,2007,inHoL.C.,WangJ.-W., eds,TheCentralEngineofActiveGalacticNucleiVol.373ofAstronomicalSocietyof thePacicConferenceSeries,ReningtheRadius-LuminosityRelationshipforActive GalacticNuclei.p.380 BurbidgeE.M.,1970,ApJL,160,L33 CapellupoD.M.,HamannF.,ShieldsJ.C.,Rodr guezHidalgoP.,BarlowT.A.,2011, MNRAS,413,908 CapellupoD.M.,HamannF.,ShieldsJ.C.,Rodr guezHidalgoP.,BarlowT.A.,2012, ArXive-prints deKoolM.,KoristaK.T.,AravN.,2002,ApJ,580,54 DexterJ.,AgolE.,2011,ApJL,727,L24 DiMatteoT.,SpringelV.,HernquistL.,2005,Natur,433,604 ElvisM.,2000,ApJ,545,63 EverettJ.E.,2005,ApJ,631,689 FerrareseL.,MerrittD.,2000,ApJL,539,L9 FoltzC.B.,WeymannR.J.,MorrisS.L.,TurnshekD.A.,1987,ApJ,317,450 147

PAGE 148

GabelJ.R.,AravN.,KaastraJ.S.,KrissG.A.,BeharE.,CostantiniE.,GaskellC.M., KoristaK.T.,LaorA.,PaerelsF.,ProgaD.,QuijanoJ.K.,SakoM.,ScottJ.E., SteenbruggeK.C.,2005,ApJ,623,85 GabelJ.R.,AravN.,KimT.-S.,2006,ApJ,646,742 GangulyR.,BondN.A.,CharltonJ.C.,EracleousM.,BrandtW.N.,ChurchillC.W., 2001,ApJ,549,133 GangulyR.,EracleousM.,CharltonJ.C.,ChurchillC.W.,1999,AJ,117,2594 GaskellC.M.,2008,inRevistaMexicanadeAstronomiayAstrosicaConferenceSeries Vol.32ofRevistaMexicanadeAstronomiayAstrosicaConferenceSeries,Accretion DisksandtheNatureandOriginofAGNContinuumVariability.pp111 GebhardtK.,BenderR.,BowerG.,DresslerA.,FaberS.M.,FilippenkoA.V.,GreenR., GrillmairC.,HoL.C.,KormendyJ.,LauerT.R.,MagorrianJ.,PinkneyJ.,Richstone D.,TremaineS.,2000,ApJL,539,L13 GibsonR.R.,BrandtW.N.,GallagherS.C.,HewettP.C.,SchneiderD.P.,2010,ApJ, 713,220 GibsonR.R.,BrandtW.N.,SchneiderD.P.,GallagherS.C.,2008,ApJ,675,985 GibsonR.R.,JiangL.,BrandtW.N.,HallP.B.,ShenY.,WuJ.,AndersonS.F., SchneiderD.P.,VandenBerkD.,GallagherS.C.,FanX.,YorkD.G.,2009,ApJ, 692,758 HallP.B.,AnosovK.,WhiteR.L.,BrandtW.N.,GreggM.D.,GibsonR.R.,Becker R.H.,SchneiderD.P.,2011,MNRAS,411,2653 HamannF.,1997,ApJS,109,279 HamannF.,1998,ApJ,500,798 HamannF.,2000,QuasistellarObjects:IntrinsicAGNAbsorptionLines.BristolInstitute ofPhysicsPublishing,Bristol HamannF.,BarlowT.A.,BeaverE.A.,BurbidgeE.M.,CohenR.D.,JunkkarinenV., LyonsR.,1995,ApJ,443,606 HamannF.,BarlowT.A.,JunkkarinenV.,BurbidgeE.M.,1997,ApJ,478,80 HamannF.,FerlandG.,1999,ARA&A,37,487 HamannF.,KanekarN.,ProchaskaJ.X.,MurphyM.T.,EllisonS.,MalecA.L., MilutinovicN.,UbachsW.,2011,MNRAS,410,1957 HamannF.,KaplanK.F.,Rodr guezHidalgoP.,ProchaskaJ.X.,Herbert-FortS.,2008, MNRAS,391,L39 148

PAGE 149

HamannF.,KoristaK.T.,MorrisS.L.,1993,ApJ,415,541 HamannF.,SabraB.,2004,inRichardsG.T.,HallP.B.,eds,AGNPhysicswiththe SloanDigitalSkySurveyVol.311ofAstronomicalSocietyofthePacicConference Series,TheDiverseNatureofIntrinsicAbsorbersinAGNs.p.203 HamannF.,SabraB.,JunkkarinenV.,CohenR.,ShieldsG.,2002,inBollerT.,Komossa S.,KahnS.,KuniedaH.,GalloL.,eds,X-raySpectroscopyofAGNwithChandraand XMM-NewtonHowMassiveareBALQSOWinds?.p.121 HamannF.W.,BarlowT.A.,ChaffeeF.C.,FoltzC.B.,WeymannR.J.,2001,ApJ,550, 142 HewettP.C.,FoltzC.B.,2003,AJ,125,1784 HopkinsP.F.,ElvisM.,2010,MNRAS,401,7 JunkkarinenV.,BeaverE.A.,BurbidgeE.M.,CohenR.D.,HamannF.,LyonsR.W., 1997,inN.Arav,I.Shlosman,&R.J.Weymanned.,MassEjectionfromActive GalacticNucleiVol.128ofAstronomicalSocietyofthePacicConferenceSeries,On thePhosphorusOverabundanceintheBALQSOPG0946+301.p.220 JunkkarinenV.T.,BurbidgeE.M.,SmithH.E.,1983,ApJ,265,51 KniggeC.,ScaringiS.,GoadM.R.,CottisC.E.,2008,MNRAS,386,1426 KoristaK.T.,VoitG.M.,MorrisS.L.,WeymannR.J.,1993,ApJS,88,357 KrongoldY.,BinetteL.,Hern andez-IbarraF.,2010,ApJL,724,L203 LeighlyK.M.,DietrichM.,BarberS.,2011,ApJ,728,94 LeighlyK.M.,HamannF.,CasebeerD.A.,GrupeD.,2009,ApJ,701,176 LundgrenB.F.,WilhiteB.C.,BrunnerR.J.,HallP.B.,SchneiderD.P.,YorkD.G., VandenBerkD.E.,BrinkmannJ.,2007,ApJ,656,73 LutzD.,SturmE.,TacconiL.J.,ValianteE.,SchweitzerM.,NetzerH.,MaiolinoR., AndreaniP.,ShemmerO.,VeilleuxS.,2007,ApJL,661,L25 LyndsC.R.,1967,ApJ,147,396 MisawaT.,EracleousM.,CharltonJ.C.,KashikawaN.,2007,ApJ,660,152 MollR.,SchindlerS.,DomainkoW.,KapfererW.,MairM.,vanKampenE.,Kronberger T.,KimeswengerS.,RuffertM.,2007,A&A,463,513 MurrayN.,ChiangJ.,GrossmanS.A.,VoitG.M.,1995,ApJ,451,498 NarayananD.,HamannF.,BarlowT.,BurbidgeE.M.,CohenR.D.,JunkkarinenV., LyonsR.,2004,ApJ,601,715 149

PAGE 150

PetersonB.M.,1997,AnIntroductiontoActiveGalacticNuclei.CambridgeUniversity Press PetersonB.M.,FerrareseL.,GilbertK.M.,KaspiS.,MalkanM.A.,MaozD.,MerrittD., NetzerH.,OnkenC.A.,PoggeR.W.,VestergaardM.,WandelA.,2004,ApJ,613, 682 ProchaskaJ.X.,HennawiJ.F.,2009,ApJ,690,1558 ProgaD.,2007,ApJ,661,693 ProgaD.,KallmanT.R.,2004,ApJ,616,688 ProgaD.,StoneJ.M.,KallmanT.R.,2000,ApJ,543,686 ReichardT.A.,RichardsG.T.,HallP.B.,SchneiderD.P.,VandenBerkD.E.,FanX., YorkD.G.,KnappG.R.,BrinkmannJ.,2003,AJ,126,2594 ReichardT.A.,RichardsG.T.,SchneiderD.P.,HallP.B.,ToleaA.,KrolikJ.H., TsvetanovZ.,VandenBerkD.E.,YorkD.G.,KnappG.R.,GunnJ.E.,Brinkmann J.,2003,AJ,125,1711 Rodr guezHidalgoP.,HamannF.,HallP.,2011,MNRAS,411,247 Rodr guezHidalgoP.,HamannF.,NestorD.B.,ShieldsJ.,2010,inprep. ScannapiecoE.,OhS.P.,2004,ApJ,608,62 ShenY.,StraussM.A.,HallP.B.,SchneiderD.P.,YorkD.G.,BahcallN.A.,2008,ApJ, 677,858 SmithL.J.,PenstonM.V.,1988,MNRAS,235,551 TrumpJ.R.,HallP.B.,ReichardT.A.,RichardsG.T.,SchneiderD.P.,VandenBerk D.E.,KnappG.R.,AndersonS.F.,FanX.,BrinkmanJ.,KleinmanS.J.,NittaA., 2006,ApJS,165,1 TurnshekD.A.,1988,inBladesJ.C.,TurnshekD.A.,NormanC.A.,eds,Proceedings oftheQSOAbsorptionLineMeetingBALQSOs-Observations,modelsand implicationsfornarrowabsorptionlinesystems.pp1746 VandenBerkD.E.,etal.2001,AJ,122,549 VandenBerkD.E.,WilhiteB.C.,KronR.G.,AndersonS.F.,BrunnerR.J.,HallP.B., Ivezi c Z.,RichardsG.T.,SchneiderD.P.,YorkD.G.,BrinkmannJ.V.,LambD.Q., NicholR.C.,SchlegelD.J.,2004,ApJ,601,692 WeymannR.J.,MorrisS.L.,FoltzC.B.,HewettP.C.,1991,ApJ,373,23 WilhiteB.C.,VandenBerkD.E.,KronR.G.,SchneiderD.P.,PereyraN.,BrunnerR.J., RichardsG.T.,BrinkmannJ.V.,2005,ApJ,633,638 150

PAGE 151

WilsonE.B.,1927,JournaloftheAmericanStatisticalAssociation,22,209 WiseJ.H.,EracleousM.,CharltonJ.C.,GangulyR.,2004,ApJ,613,129 151

PAGE 152

BIOGRAPHICALSKETCH DanielCapellupowasborninNewYork,NY,andgrewupinParkRidge,NJ.He wasthesalutatorianoftheParkRidgeHighSchoolClassof2002.Hethenattended theUniversityofRochesterwherehemajoredinphysicsandastronomy,withaminor inmathematics,andgraduatedMagnaCumLaudein2006.Hewasalsoinductedinto thePhiBetaKappahonorsocietythatsameyear.HeleftthemildclimateofRochester, NY,topursuegraduatestudiesinastronomyattheUniversityofFlorida,wherehewas awardedaMasterofSciencein2008.HeintendstograduatewithhisPhDinMay2012. 152