Durham Research Online Deposited in DRO: 25 July 2017 Version of attached (cid:28)le: Published Version Peer-review status of attached (cid:28)le: Peer-reviewed Citation for published item: Bruce, A. and Lawrence, A. and MacLeod, C. and Elvis, M. and Ward, M. J. and Collinson, J. S. and Gezari, S. and Marshall, P. J. and Lam, M. C. and Kotak, R. and Inserra, C. and Polshaw, J. and Kaiser, N. and Kudritzki, R-P. and Magnier, E. A. and Waters, C. (2017) ’Spectral analysis of four ‘hypervariable’ AGN : a micro-needle in the haystack?’, Monthly notices of the Royal Astronomical Society., 467 (2). pp. 1259-1280. Further information on publisher’s website: https://doi.org/10.1093/mnras/stx168 Publisher’s copyright statement: This article has been accepted for publication in Monthly notices of the Royal Astronomical Society (cid:13)c: 2017 The Authors Published by Oxford University Press on behalf of the Royal Astronomical Society. All rights reserved. Additional information: Use policy Thefull-textmaybeusedand/orreproduced,andgiventothirdpartiesinanyformatormedium,withoutpriorpermissionorcharge,for personalresearchorstudy,educational,ornot-for-pro(cid:28)tpurposesprovidedthat: • afullbibliographicreferenceismadetotheoriginalsource • alinkismadetothemetadatarecordinDRO • thefull-textisnotchangedinanyway Thefull-textmustnotbesoldinanyformatormediumwithouttheformalpermissionofthecopyrightholders. PleaseconsultthefullDROpolicyforfurtherdetails. DurhamUniversityLibrary,StocktonRoad,DurhamDH13LY,UnitedKingdom Tel:+44(0)1913343042|Fax:+44(0)1913342971 https://dro.dur.ac.uk MNRAS467,1259–1280(2017) doi:10.1093/mnras/stx168 AdvanceAccesspublication2017January20 Spectral analysis of four ‘hypervariable’ AGN: a microneedle in the haystack? A. Bruce,1‹ A. Lawrence,1‹ C. MacLeod,2 M. Elvis,2 M. J. Ward,3 J. S. Collinson,3 S. Gezari,4 P. J. Marshall,5 M. C. Lam,1 R. Kotak,6 C. Inserra,6 J. Polshaw,6 N. Kaiser,7 R-P. Kudritzki,7 E. A. Magnier7 and C. Waters7 1InstituteforAstronomy,SUPA(ScottishUniversitiesPhysicsAlliance),UniversityofEdinburgh,RoyalObservatory,BlackfordHill,EdinburghEH93HJ,UK 2Harvard–SmithsonianCenterforAstrophysics,60GardenSt,Cambridge,MA02138,USA 3DepartmentofPhysics,UniversityofDurham,SouthRoad,DurhamDH13LE,UK 4DepartmentofAstronomy,UniversityofMaryland,CollegePark,MD20742-2421,USA 5KavliInstituteforParticleAstrophysicsandCosmology,POBox20450,MS29,Stanford,CA94309,USA 6AstrophysicsResearchCentre,SchoolofMathematicsandPhysics,QueensUniversityBelfast,BelfastBT71NN,UK 7InstituteforAstronomy,UniversityofHawaii,2680WoodlawnDrive,Honolulu,HI96822,USA Accepted2017January18.Received2016December23;inoriginalform2016August22 ABSTRACT We analyse four extreme active galactic nuclei (AGN) transients to explore the possibility thattheyarecausedbyrare,high-amplitudemicrolensingevents.Thesepreviouslyunknown type-IAGNarelocatedintheredshiftrange0.6–1.1andshowchangesof>1.5magintheg bandonatime-scaleof∼years.Multi-epochopticalspectroscopy,fromtheWilliamHerschel Telescope, shows clear differential variability in the broad line fluxes with respect to the continuumchangesandalsoevolutioninthelineprofiles.Intwocases,asimplepoint-source, point-lensmicrolensingmodelprovidesanexcellentmatchtothelong-termvariabilityseenin theseobjects.Forbothmodels,theparameterconstraintsareconsistentwiththemicrolensing being due to an intervening stellar mass object but as yet there is no confirmation of the presenceofaninterveninggalaxy.Themodelspredictapeakamplificationof10.3/13.5and anEinsteintime-scaleof7.5/10.8yr,respectively.Inonecase,thedataalsoallowconstraints onthesizeoftheCIII]emittingregion,withsomesimplifyingassumptions,tobe∼1.0–6.5 light-daysandalowerlimitonthesizeoftheMgIIemittingregiontobe>9light-days(half- lightradii).ThisCIII]radiusisperhapssurprisinglysmall.Intheremainingtwoobjects,there isspectroscopicevidenceforaninterveningabsorberbuttheextrastructureseeninthelight curvesrequiresamorecomplexlensingscenariotoadequatelyexplain. Key words: accretion, accretion discs–gravitational lensing: micro–galaxies: active– galaxies:nuclei–quasars:absorptionlines–quasars:emissionlines. (TDEs)orextinctionevents.Afourthpossibility,andthefocusfor 1 INTRODUCTION thisparticularwork,isthatsomeoftheseAGNtransientsareac- Theidentificationofaclassofactivegalacticnuclei(AGN)tran- tuallyrare,high-amplitudemicrolensingevents.Ifthisisthecase, sients that are smoothly evolving, by factors of several, on year- through analysis of multi-epoch spectroscopy and simple lensing longtime-scales(Lawrence2012;Lawrenceetal.2016,hereafter models,wehavethepotentialtouncovervaluableinformationon L16) has raised a number of interesting questions regarding the theinnermostregionsoftheseenigmaticobjects. underlying cause. These ‘hypervariable’ AGN may simply be at Observational studies of the microlensing seen in multiply im- the extreme end of the tail of typical AGN variability (MacLeod aged AGN are well established and provide the means with etal.2010,2012)or,perhapsmoreinterestingly,theremaybean which one can ascertain properties intrinsic to those objects extrinsiccause.Plausiblemechanismsforthesetransientsinclude (Irwin et al. 1989; Eigenbrod et al. 2008; Morgan et al. 2010; thefollowing:accretionevents/instabilities;tidaldisruptionevents Blackburne et al. 2011; Mosquera & Kochanek 2011; Jime´nez- Vicenteetal.2012,2014;Sluseetal.2012;MacLeodetal.2015). ThisinvolvesamonitoringoftheAGNcomponentsinordertoas- (cid:2)E-mail:[email protected](AB);[email protected](AL) certainboththelevelofvariabilityintrinsictotheAGNandthatdue (cid:3)C 2017TheAuthors PublishedbyOxfordUniversityPressonbehalfoftheRoyalAstronomicalSociety 1260 A.Bruceetal. tomicrolensing,typicallyalow-level‘flickering’,inoneormore Table1. SummaryofthetargetselectioncriterianotedinSection2.1.2. of these components. In contrast, the AGN in this paper exhibit noclearsignofstronglensingeffectsandarevaryingsignificantly ShortID SDSSID Single- Double- Int- andsmoothly(relativetotypicalAGNbehaviour).Ifarare,high- peak? peak? gal? amplitude microlensing event is the cause of this variability, this J084305 J084305.54+550351.3 Yes No No placesthesepreviouslyunknownAGNinadifferentregimetotheir J094511 J094511.08+174544.7 Yes No No multiplyimagedcounterparts. J142232 J103837.08+021119.6 No Yes Yes In this paper, four promising AGN transients are selected as J103837 J142232.45+014026.8 No No Yes candidatesforanexplorationofthemicrolensingscenarioandits consequences. That is, the hypothesis that an intervening stellar Additionalfactorsthatfavouramicrolensingscenarioaresum- massobjectisresponsibleforthebulkvariabilityseenintheseob- marizedinTable1.Here,thepresenceofanear-symmetricsingle- jects.Itshouldbenotedthatmicrolensingeventswillnotexplainall or double-peaked structure in the light curve, or evidence for an hypervariableAGNactivity;rather,theylikelydescribeasubsetof intervening galaxy, and hence lens repository, are noted. At least thisinterestingpopulation.Intheinterestofbrevity,thispaperwill oneoftheseadditionalfactorswasrequiredinthetargetselection focusentirelyonthemicrolensingscenariobutinterestedreaders processleavingatotaloffourmicrolensingcandidatesforconsid- shouldrefertoL16andreferencesthereinformorediscussionon eration.Anyevidenceforaninterveninggalaxywillbediscussed thealternativevariabilityscenarios. aspartofthespectroscopicresultsforthattarget(Section3). Section 2 describes our sample selection, observational data and reduction pipeline. Spectroscopic results are detailed in Section 3. Section 4 outlines the procedures used for explor- 2.2 Photometry ing different lensing models with results detailed in Section 5. 2.2.1 SDSSdata In Section 6, the implications for future work and the advan- tages/disadvantages over existing microlensing studies involving SDSSdatausedinthispaperareDR9(Ahnetal.2012)cmodelg- multiply imaged AGN are discussed. Cosmological calculations bandmagnitudes.Theseshouldprovidethemostreliablemagnitude in this paper make use of Planck13 values (Planck Collaboration estimate for an extended object in a single band and should also XVI2014;H0=67.8,(cid:3)(cid:4)=0.693). agreewiththePointSpreadFunction(PSF)magnitudesforstars. Notallbandsarebestfitwithextendedmodelsfortheseobjects,as mightbeexpectedwiththepresenceofanAGNcomponent.Also, 2 OBSERVATIONS fromDR7toDR9,J103837hashadtheclassificationchangedto stellar.TheleastreliabledataisthatforJ142232,which,inaddition 2.1 Targetselection tobeingfaint,isflaggedashavingbeenaffectedbyacosmicray hit. 2.1.1 Parentsample Thefourobjectsinthispaperareselectedfromalargersampleof 2.2.2 Pan-STARRSdata highlyvariableAGN,discoveredaspartofawidertransientsearch initiallydesignedtolookforTDEsaroundquiescentblackholes.A The Pan-STARRS 1 3π g data used in this paper are from the P1 significantfractionofcandidateeventsvariedonlongertime-scales PV1.2datarelease(Schlaflyetal.2012;Tonryetal.2012;Magnier thanthatexpectedforTDEsorsupernovaeandweresubsequently et al. 2013) and magnitudes have been calculated from the PSF revealedtobeAGN.Thetransientswereidentifiedbylookingfor fluxesandassociatedzero-points. changesofgreaterthan1.5maginthePanoramicSurveyTelescope andRapidResponseSystem(Pan-STARRS)13πSteradianSurvey data when compared with the Sloan Digital Sky Survey (SDSS) 2.2.3 LiverpoolTelescopedata footprintfromaroundadecadeearlier.Inaddition,thetransienthad TheLiverpoolTelescope(LT)providesfinersamplingofeachtran- tobelocatedwithin0.5arcsecofanobjectclassifiedasa‘galaxy’in sientintheAGNsampleandwasalsoinstrumentalintheclassi- theDR7catalogue.Approximatelytwo-thirdsoftheflaggedobjects fication of the Pan-STARRS transients. The LT is a fully robotic havebeenspectroscopicallyconfirmedastype-IAGNanditisthese 2.0 m telescope situated on La Palma and operated by Liverpool thatmakeupthelarger‘hypervariable’AGNsample,currently63 JohnMoore’sUniversity(Steeleetal.2004).Thetransientswere objects.Forfurtherdetails,seeL16. initiallymonitoredeveryfewdaysinu ,g andr todetermine LT LT LT how fast they were evolving and then roughly every few weeks. Scheduling and weather constraints sometimes meant that there 2.1.2 Selectioncriteriaforthispaper wasnotalwaysthedesiredcadenceonsometargets.Theobserving programmewiththeLTisstillongoing.Onlytheg dataisutilized Toexplorethemicrolensinghypothesis,objectswereselectedfrom LT inthispaper. the hypervariable AGN sample based on the following criteria. First, that the photometry displays signs of smooth evolution on long time-scales with a change in magnitude of (cid:6)g > 0.5 mag 2.2.4 CRTSdata over the period of spectral observations. Secondly, that there are aminimumoftwospectralobservationsseparatedbyatleastone InadditiontotheSDSSandLTphotometry,therearedatafromthe year(observedframe).Theformerfavoursmicrolensingeventsover Catalina Real-Time Transient Survey (CRTS; Drake et al. 2009). intrinsicvariabilityandthelatterincreasesthechanceofobserving Thissurveymakesuseofthreedifferenttelescopesandallowsusto spectroscopic trends, allowing further testing of the microlensing recoverpre-Pan-STARRSeralightcurvesforsomeoftheobjects. hypothesis. A note of caution is that this survey uses clear filters calibrated MNRAS467,1259–1280(2017) Amicroneedleinthehaystack? 1261 to a v-band zero-point, so colour effects may be significant. An Exposuresweretakenin1800sincrementsandthenumberofshots approximate magnitude offset has been applied so that the light ontargetwasadjustedbasedonthelatestg photometry. LT curveappearsconsistentwiththeotherdatasetsforeachtargetand thedatahavealsobeenseasonallyaveragedforclarity.Duetothe uncertaintyregardingcoloureffects,theCRTSdataisnotusedin 2.3.2 MMTdata themodellingprocess. Asmallnumberofobservationsweremadewiththebluearmspec- trographonthe6.5mMMTsituatedonMountHopkins,Arizona. 2.2.5 Filterapproximation Here,the300gmm−1gratingwasusedwith2×binninginthespa- tialdirection.Afilterwheelissuemeantthatnoorder-sortingfilter Forthepurposesofthispaper,weassumethatthedifferencesbe- wasused.ThismayaffectthethirdepochfortargetJ094511and tweentheg ,g andg filterscanbeneglectedanddesignate SDSS P1 LT fourthepochforJ084305.Duetovariableconditionsandanearby magnitudesassimplyg.Thisshouldbereasonablegiventhatwe brightMoon,thesixthepochforJ142232wasverypoorandwill expect the photometric uncertainty to be dominated by intrinsic notbeusedintheanalysis. AGNvariability,typically0.1magorgreater.Forthemicrolensing models,magnitude-to-fluxconversionsareperformedassumingan effectivewavelengthof4770Å. 2.3.3 Spectroscopicreductionpipeline A reduction pipeline was created using custom PYRAF scripts 2.3 Spectroscopy and standard techniques. After bias-subtraction and flat-fielding 2.3.1 WHTdata the cosmic rays were removed using the LACOS_SPEC script (van Dokkum 2001). The spectra were then extracted and wavelength The majority of the spectral observations were performed on the calibratedusingthearcsobtainedthatnight.Inordertominimize 4.2 m William Herschel Telescope (WHT), La Palma, using the problemswiththecalibrationduetotemperaturevariationsorin- IntermediatedispersionSpectrograph andImagingSystem(ISIS) strument flexure, an additional step was to offset each red/blue long-slitspectrograph.The5300dichroicwasusedalongwiththe spectrum by a small amount, typically ∼±0–3 Å, to ensure the R158B/R300Bgratinginthered/bluearms,respectively,alongwith prominentatmosphericoxygenlineat5577.338Åhadthecorrect theGG495ordersortingfilterintheredarm.Typically2×binning wavelengthinallobservations.Fluxcalibrationwasperformedus- inthespatialdirectionwasusedtoimprovethesignal-to-noiseratio ing a single standard star and the mean extinction curve for the (SNR)alongwithanarrowCCDwindowtoreducediscusageand observatory.Wheretargetobservationswerebracketedbystandard readouttimes.Thisset-upgivesaspectralresolutionofR∼1500 starobservations,theoneleastaffectedbytransparencyissueswas at5200ÅintheblueandR∼1000at7200Åintheredforaslit used,i.e.thestandardwiththesuperiorsensfuncresponse.To widthof1arcsecandtotalcoverage∼3100–10600Å. better enable the combination of the red and blue data, the inner Typicallycalibrationimagesweretakenatthestartofeachnight wingofeachspectrum,wheretheresponseoftheinstrumentde- including bias frames, lamp flats and CuNe/Ar arc lamp images. clinessteeply,wasfluxcalibratedseparately.Thefinalstepwasto Spectroscopicstandardstarswereimagedat∼2hintervalsthrough- averageallcalibratedspectraandrebintoalinearwavelengthscale. outthenightthoughthiscadencewasnotalwayspossible.Obser- Long-slit spectroscopy can suffer from transparency problems, vationswerecarriedoutattheparallacticangleandfurtherobser- which affect the absolute flux calibration. In order to minimize vational details for each target in this paper are given in Table 2. theseeffects,eachtargetspectrumhasbeenrescaled.Forthreeof Table2. Detailsofthespectralobservationscarriedoutfortheobjectslistedinthispaper.Theairmassvaluesreflectthemeanairmassovertheduration oftheobservation(s).MagnitudesestimatesareapproximationsbasedontheLTphotometry.Seeingestimatesreflecttherangeinmeasuredfullwidth athalfmaximum(FWHM)ofthecentralregionofthebluearmtraceforeachimage. Target Scope Date Exposures gmag Slit Seeing Airmass Notes J084305−1 WHT 2013-02-09 1×1800s 20.0 2arcsec 2.2arcsec 1.74 Dark,patchycloud,variableseeing −2 WHT 2013-03-31 2×1800s 20.2 1arcsec 1.1–1.2arcsec 1.12–1.14 Moon73percent(sep122◦),stable −3 WHT 2014-12-17 4×1800s 21.1 2arcsec 2.0–2.7arcsec 1.51–1.24 Dark,variableseeing −4 MMT 2015-03-10 2×1800s 21.3 1.5arcsec 0.8–1.1arcsec 1.1–1.09 Moon82percent(sep96◦),variableseeing J094511−1 WHT 2013-05-15 2×1800s 20.7 1arcsec 1.3arcsec 1.18–1.28 Moon30percent(sep24◦),stable −2 WHT 2014-02-07 4×1800s 21.1 1arcsec 1.0–1.4arcsec 1.34–1.08 Moon63percent(sep80◦),lightcloud −3 MMT 2015-03-09 2×1800s 21.5 1.5arcsec 0.8–0.9arcsec 1.06–1.04 Moon89percent(sep63◦),variableseeing J142232−1 WHT 2013-02-11 2×1800s 20.0 1.5arcsec 1.8–2.0arcsec 1.24–1.18 Dark,variableseeing −2 WHT 2013-05-15 1×1800s 20.1 1arcsec 1.8arcsec 1.34 Dark,stable −3 WHT 2013-08-07 1×1800s 20.2 1arcsec 1.1arcsec 1.58 Dark,stable −4 WHT 2014-02-07 2×1800s 20.6 1arcsec 1.1–1.2arcsec 1.14–1.13 Dark,stable −5 WHT 2014-07-24 4×1800s 21.0 1arcsec 1.6–2.1arcsec 1.25–1.93 Dark,stable 1×900s (unused)−6 MMT 2015-03-09 2×1800s 21.0 1arcsec – 1.38–1.28 Moon89percent(sep14◦),variable,v.poor −7 WHT 2015-04-23 4×1800s 21.0 1arcsec 1.6–1.9arcsec 1.25–1.68 Dark,stable J103837−1 WHT 2013-02-11 1×1800s 19.7 1.5arcsec 1.9arcsec 1.13 Dark,variableseeing −2 WHT 2013-05-15 1×1800s 20.2 1arcsec 1.3arcsec 1.38 Moon30percent(sep40◦),stable −3 WHT 2015-04-21 2×1800s 19.8 1arcsec 0.8–0.9arcsec 1.13–1.12 Dark,stable MNRAS467,1259–1280(2017) 1262 A.Bruceetal. Table3. Informationonthetargetsinthispaper.TheAVvaluesare thoseobtainedfromSchlafly&Finkbeiner(2011)assuminganRV of3.1. Target RA Dec z AV J084305 084305.55 +550351.4 0.8955 0.0824 J094511 094511.08 +174544.7 0.758 0.0718 J103837 103837.09 +021119.7 0.620 0.0877 J142232 142232.45 +014026.7 1.076 0.09 thetargets,asmoothinterpolationthroughtheLTphotometrywas used to correct the spectra. In the case of J084305 and J094511, this was done by rescaling to the working microlensing model. ForJ142232,thiswasaccomplishedusingacubic-splinefittothe LT light curve. The fourth target, J103837, does not produce a Figure1. ExampleofthefittingprocessfortargetJ094511.Thecompo- satisfactorycubic-splinefitatallepochsduetoapoorerLTcadence. nentsincludethefollowing:powerlaw(green),Gaussianlinefits(magenta), Instead, the spectra for this target have been rescaled, so that the convolvedirontemplate(cyan)andoverallbestfit(red). measured[OIII]5007 flux,tiedtothethirdepoch,remainsconstant. Thismethodwillsufferiftherearesignificantseeingchanges,and 3 SPECTROSCOPIC RESULTS itmayevenbepossiblethatontheseyear-longtime-scales,there willbesomeintrinsicnarrowlinevariability(Petersonetal.2013). Theresultsfromthespectroscopicanalysisarepresentedinthree Spectral fluxes were measured using an LT g-band transmission keyfigures.Fig.2showsthetargetlightcurvesandlinefluxevo- curve.1ThescalingfactorsappliedarenotedinTableA1.Though lution.Thelinefluxeshavebeenplottedrelativetothefirstepoch theMgIIbroademissionlineispresentwithinthegbandforthese toallowquickcomparison.Fig.3showstheevolutionofthesys- objects(excludingJ142232),wehavenotattemptedtoaccountfor temic velocity offsets and emission line widths. Fig. 4 shows the thisintherescalingprocesssoastokeepthedataconsistentwith evolutionofthespectralprofilesaftercontinuumandFetemplate thephotometry. subtraction.Forclarity,amedianfilterhasbeenappliedtothebroad lines.AdditionalmeasurementscanalsobefoundinTableA1.The observedspectraareshowninFigsA1–A4. Threeofthefourtargets,J084305,J094511andJ142232,show 2.3.4 Spectralfitting clear evidence for a differential evolution of the continuum with respecttothelinefluxes.Ingeneral,thecontinuumdecreasesbya Forthespectralfittingprocess,aPYTHONpackage,LMFIT,wasused. Thisisanon-linearoptimizationandcurve-fittingtoolthatbuilds factor∼4and,toalesserdegree,theCIII]fluxtracksthischange. on a Levenberg–Marquardt algorithm. It was used to fit a single The MgII flux either tracks the continuum change weakly or is consistent with no change at all. The photometry indicates that Gaussiancomponenttotheemissionlinesandprovideapower-law thetargetshavebeenevolvingsmoothlyoverthisperiod.Amore fit to approximate the local continuum. The relatively low SNR inthebluearmforsomeepochsultimatelyruledouttheuseofa detailedsummaryofeachobjectnowfollowsandthereadershould referbacktothefiguresmentionedinthepreviousparagraph. multicomponentfitformostbroadlines.Thepowerlawusedinthe fittingroutinetakestheform Anyevidenceforaninterveninggalaxy,whichwouldlendweight tothemicrolensingscenario,willbepresentedonapertargetbasis. Fλ=A(λ/5100Å)β, (1) whereAisthenormalizationandβ isthepower-lawslope.Inad- 3.1 J084305 ditiontotheabovecomponents,atemplatefitwasusedtoestimate theFecontribution.IntheUV,theempiricaltemplateisthatfrom The light curve of this target shows a smooth decline of Vestergaard&Wilkes(2001),andintheoptical,wherepossible,that ∼0.6 mag yr−1 from a peak more than two magnitudes brighter ofVe´ron-Cetty,Joly&Ve´ron(2004).TheFetemplateisconvolved thantheSDSSobservations16yearsago.TherearenoCRTSdata withaGaussianinordertobetterapproximatethetrueblendedFe thatallowustoseethepre-Pan-STARRSevolution.Therearetwo emission.ThewidthoftheconvolvingGaussianwassettomatch spectral epochs near the observed maximum around 56100 MJD thatofthebroadMgII/Hβ componentsintheUV/optical,respec- andafurthertwoafteradeclineofatleastonemagnitude.Themost tively.Beforeperforminganyfitsthespectraarefirstcorrectedfor recentphotometrysuggeststhetargetisnowreturningtothelevelof MilkyWayextinctionusingtheA valuesinTable3andtheex- theSloanera.OfthethreeSDSSepochs,twoareatg(cid:5)22.25mag tinctionlawintheopticalfromCarVdelli,Clayton&Mathis(1989). and the third has the higher value of g = 21.56 mag, which may Noattempthasbeenmadetocorrectforhostgalaxyreddeningat indicateaproblemwiththeSloancmodelfit(s).Themicrolensing thisstage.Themostprominenttelluricfeaturesabove6860Åwere modelshown,towhichthespectrahavebeenrescaled,doesagood maskedoutduringthefittingprocess.WhenfittingHβ,thenarrow jobofreproducingthelargeamplitudechanges.Furtherdetailson linewidthsandcentresweretiedtothatof[OIII]5007.Anexample themodelareinSection5.1.1. fitisshowninFig.1. Astothespectralevolution,thecontinuumisseentodropbya factorof5overtheperiodspannedbytheobservations.TheCIII] flux shows a corresponding drop by a factor of ∼2. In contrast, the MgII and Fe fluxes are consistent with no change or perhaps 1http://telescope.livjm.ac.uk/TelInst/Inst/IOO/ alateincrease.BoththeMgIIandCIII]linesgetnarrowerduring MNRAS467,1259–1280(2017) Amicroneedleinthehaystack? 1263 Figure2. Left-handpanels:compositelightcurvesforeachtargetincludingspectralepochs.Theopencirclesshowtherawspectralmagnitudesbefore rescalingasperSection2.3.3andthegreencirclesshowtherescaledvalues.CRTSdatahaveoffsetsof+0.1/0.1/0.4/0.2mag,respectively.Theextended tailsontheerrorbarsforJ084305andJ094511reflecttheerrorsusedinthemicrolensinganalysis.Right-handpanels:relativefluxesofthevariousmeasured spectralcomponents.Forthefirstthreetargets,thefourthpanelshowstheCIII]/MgIIratio.ForJ103837,thetwoGaussiancomponentsusedinthefit(broad andwide)aredisplayed.Asterisksdenotelessreliablevaluesasnotedinthetext.ThegreendashedlinereflectsthecontinuumunderCIII]ineachcase. MNRAS467,1259–1280(2017) 1264 A.Bruceetal. Figure3. Left-handpanels:velocityoffsetsforthecentreofthemeasuredlineprofilesforeachtargetassumingrest-framewavelengths,inÅ,ofthefollowing: CIV(1549);CIII](1909);MgII(2800);[OII](3727);Hβ/[OIII](4861/5007).TheAGNsystemicredshift/velocityisdeterminedfromthemedianofthe[OII] centres.InthecaseofJ142232,duetoalackofobservednarrowlines,MgIIhasbeenused.Theerrorbarsinthiscaseonlyreflecttheuncertaintyinthe measuredlinecentreandomittheadditionalsystemicvelocityuncertainty.TheFWHMofCIVforJ142232isnotshownasthisvaluewastiedtothatofCIII] inthefittingprocess.Right-handpanels:FWHMofthemeasuredlineprofilesforeachtarget. MNRAS467,1259–1280(2017) Amicroneedleinthehaystack? 1265 Figure4. Left-handpanels:CIII]andMgIIprofilesforeachobject.ForJ103837,theprofileistruncatedduetothebluearmcut-off.Right-handpanels: additionalemissionlineprofilesmeasuredinthefittingprocess.Forclarity,allplots,barringthosefor[OII],havehadamedianfilterapplied. the decline. In addition, both lines appear blueshifted relative to apparentchangesintheredwingoftheMgIIprofileandbothwings thesystemicvelocity,CIII]moresothanMgII.Thereareintriguing of CIII], most notably around the expected position of the fainter but inconclusive signs of an evolution of the offsets, which shift AlIIIcomponent.Thescatterseeninthe[OII]fluxmaybedueto bluewards by 500 km s−1 or so and then recover. There are also slit-widthchanges. MNRAS467,1259–1280(2017) 1266 A.Bruceetal. Therearenoclearspectralfeaturessuggestiveofanintervening galaxybutthispossibilitycannotyetberuledout.Themicrolensing modelforthistargetshowsthatanyinterveninggalaxyfluxwillbe atleastafactorof2belowtheAGN/hostbaselinefluxinthegband, makingspectroscopicdetectionduringanongoingeventdifficult. In addition, two of the three SDSS epochs are flagged as having issueswiththePetrosianradii.Thismaysimplybeduetonoiseor isperhapsanindicatorofmorphologicalissues. 3.2 J094511 Since the SDSS epoch some 11 years ago, this target has been evolvingataround∼0.4magyr−1 anddisplaysanapproximately symmetric light curve about a peak around 56000 MJD, with an apparentdip/riseafterthe57000MJDmark.Therearethreespectral epochsforthisobjectspreadovertwoyears,afterthemaximum, Figure5. FourthspectralepochforJ142232highlightingthenarrowab- withafactor∼2decreaseinluminosityoverthisperiod.Aswith sorptionfeaturesseen.TheseareconsistentwithMgII/FeIIabsorptionat J084305,the spectra have been scaled tothe microlensing model a lower redshift or possibly an outflow intrinsic to the AGN. A similar (Section5),thoughtheaccuracyofthethirdepochscalingisless MgII/FeIIfeatureisseennearerthebluewingoftheMgIIbroadlinethough certain due to the additional structure in the light curve at this inthiscaseonlytheMgIIdoubletisreadilyvisible. point. islessreliableduetothelackofcalibrationpointsforthechosen AswithJ084305,thereisamarkeddifferencebetweentheevo- lutionofMgIIandCIII]fluxes.CIII]dropsto∼60percentofthe standardstars.2 ThespectroscopicresultsforJ142232arebroadlysimilartothat iannitiianlcrveaalsueeowfh∼ile30MpgerIIciesnctoinfsitshteentthiwrdithepnoochchsacnagleinagndispcoosrsriebclty. seen in J084305/J094511. There is a factor ∼4 drop in the con- Tusheidsiantcrtehaesethiisrdalseoposcehe.nAfotrt[hOisII]e,ptohcohu,gthhealMarMgeTrsslpitewctirdutmhwdaids ctilnousuelmy,adnrdopthpeinCgIbIIy]/CaIfVacfltuoxre∼s3a.ppTehaer tMogtrIaIcflkutxhiaslscohatnragceksqutihtee not have sufficient wavelength coverage to catch Hβ/[OIII], so it continuumtoalesserextent,droppingbyafactor∼2.TheFecom- is harder to draw firm conclusions on any trends here. The Hβ ponentbehavessimilarly,thoughthereisahigherdegreeofscatter. broadcomponentisredshiftedby∼1300kms−1 andMgIIshows TheCIII]centreshowsablueshiftof∼1000kms−1,whichdoes asmaller∼500kms−1 redshift.CIII]showsamodestblueshiftof notevolve,andthelinegetsnarrowerindecline.Incontrast,MgII ∼250kms−1.Thereisnosignofanevolutioninvelocityoffsets shows a higher degree of scatter in the velocity offset and an in- crease in line width of ∼1000 km s−1 over this period. There is asseenforJ084305.TheCIII]andMgIIlinewidthsshowthesame trendastheiramplitudes,asdothenarrowlines,thoughHβ only evidenceforevolutionofthelineprofiles,mostnotablyontheCIII] showsanincreaseinwidth.Lookingatthelineprofiles,CIII]shows bluewingMgII/CIVredwing. Inaddition,thereisalsoasetofstrongnarrowabsorptionfea- asignificantbluewingchangewhereasMgIIseesachangeonthe tures present, possibly evidence for an intervening galaxy. These rperdimwairnilgy.oIntitshepebrlhuaepwsiinngt.eNreostcinlegarthsapte[cOtraIIl]ssigeensataunreeonfhaanpcoesmsiebnlet features are consistent with MgII/Fe absorption at z = 0.855 and arehighlightedinFig.5.TheAGNisatz=1.076.Alternatively, interveninggalaxyhasbeendetected. thisfeaturemaybetheresultofanoutflowintrinsictotheAGN. However,this(cid:6)v iscloseto0.1c,muchgreaterthanthevalueof 3.3 J142232 0.01c typically used to distinguish between associated and inter- vening systems (York et al. 2006) making an outflow unlikely in The light curve for this target shows there has been a rise of ap- proximately three magnitudes since the SDSS epoch and |(cid:6)g| (cid:5) this case. There is also a second similar but fainter feature, seen 1.22magoverthesixspectralepochsasestimatedfrominterpola- nearerthebluewingoftheMgIIbroadline.Inaddition,theSDSS epochforthistargetisflaggedasapossibleblend,i.e.morethan tionoftheLTlightcurve.TheSDSSepochisfaintandshouldthere- one peak was detected in a given filter. Given that the target was forebetreatedwithcaution.Thephotometryshowsthistargetto faintatthisepoch,thismaynotbereliablebutlendsfurtherweight beevolvingsmoothlybutthereisanotablediparound55500MJD, tothepossiblepresenceofaninterveninggalaxy. thisisaftertheassumed‘peak’around55000MJD.Anearlierdip around54200MJDisalsovisiblethoughtheCRTSdatashowcon- siderablescatterhere.Afterthe56500mark,thereisarapiddrop 3.4 J103837 of ∼1.1 mag yr−1, the time of the spectral observations, but the Thisobjectexhibitsthemostcomplexlightcurveandthespectra mostrecentphotometryshowsthislevellingoff.Thecadenceofthe LTobservationsisnotalwaysideal,whichmayaffectthespectral havebeenscaledrelativetothe[OIII]5007flux(Section2.3.3).There isevidenceofatleastthreepeakswithoccasionalrapidchangesin scalingcorrections(Section2.3.3). For this target, outside of the rest-frame UV (λ > 6400 Å), brightness,particularlyaround56300MJD,whichshowsadropof obs ∼0.4magover2.5months.Here,thecadenceoftheLTphotometry nootherspectralfeaturesornarrowemissionlinesweredetected. Givenalackofdetectednarrowlines,thesystemicvelocity/redshift hasbeendeterminedfromthemedianMgIIlinecentre.Forclarity, 2ThefittingpipelinewouldoccasionallyfailtofitCIVandinsteadaddan thecorrespondingincreaseinuncertaintyhasbeenomittedfromthe additional continuum component. To prevent this, the width of CIV was plotinFig.3.Also,atthisredshifttheCIVlineisseen,albeitvery constrainedtobethesameasCIII],sotheCIVresultsshouldbetreatedwith nearthebluecut-off.Shortwardof∼3200Å,thefluxcalibration caution. MNRAS467,1259–1280(2017) Amicroneedleinthehaystack? 1267 &Fluke2013).Thisisjustifiedbecausewearedealingwithran- domsightlinesratherthanobjectspre-selectedasmultiplyimaged quasars.ForexampleasightlinethroughtheMilkyWayattheso- larradius,withstellardensity∼0.1pc−3andscaleheight300kpc, wouldproduceκ∼0.2ifplacedatz=0.2withthequasaratz=1.0. Infact,asexplainedinL16,wewillusuallybelookingthrougha galaxywellbelowL∗. On the other hand, the effect of shear caused by the overall potential of the lensing galaxy containing the microlensing star, willoftenbesignificant,sothatingeneralwewillbedealingwith ‘Chang–Refsdal’(CR)lenses(Chang&Refsdal1984).Fortwoof ourtargets,withdoublepeaks(J103837andJ142232),theshearis clearlyimportant.Theothertwo(J094511andJ084305)aresmooth and single peaked, so we start with the simplest model ignoring shear and using a simple point lens. In Sections 4.2 and 4.3, we Figure6. ThirdspectralepochforJ103837highlightingthenarrowab- examinetheeffectofvaryingthatassumption. sorptionfeature.ItisconsistentwithanMgIIdoubletatalowerredshift thantheAGN. 4.1 Point-sourcepoint-lensmodel islessthanideal.Itispossiblethedropwaslargerthanobserved. Itisusefultofirststartwiththesimplestofcases,thatofapoint- Afterthedrop,itappearstorisemoregraduallyat∼0.2magyr−1, masslensandpoint-sourcewithnoexternalshear.Thisislikelyan though there is a spike in the data (g = 19.79 mag) at the third oversimplification but will nevertheless prove useful, particularly spectralepochat57137MJD.Thereisapeak-to-peakchangefrom with regard to J084305 and J094511, both of which have light theSloaneraofatleasttwomagnitudes.Theincreaseincomplex- curves that are single-peaked and smoothly evolving (Fig. 2). In ity,coupledwiththelackofLTdataatthesecondspectralepoch thismodel,themagnificationisgivenby causesaproblemwhenattemptingtocorrectthespectraldatafor transparencyeffects(Section2.3.3).Forthisobject,the[OIII]flux μ= (cid:2)y2+2 , y :=β/θ . (2) wasusedtorescalethedataastheinterpolationusingtheLTdata y y2+4 E waspoor.Thismayintroduceadditionalerrorsifthelinefluxvaries either intrinsically over this time-scale or due to aperture/seeing Here,yisthenormalizedsourcepositioninunitsoftheEinstein effects. radiusofthelens,whichisgivenby (cid:3) (cid:4) Giventhatinthiscasethesecondspectralepochrepresentsthe 4GM D 1/2 minimum,itishardertodrawconclusionsregardinganyspectro- θ = ds . (3) E c2 D D scopictrends.TheSNRofthethirdepochwassufficienttoallow d s amorecomplexfitfortheHβ/MgIIlines.Inthiscase,twoGaus- Dd,Ds andDds aretheangulardiameterdistancesforthelens, sianprofileswereusedsimultaneously.TheCIII]resultsshouldbe sourceandbetweenthelensandsource,respectively.Tocompute treated with caution as the bulk of the line was beyond the blue thelightcurveF(t)=μ(t)F requirestheformulaforthetrajectory s cut-offrequiringthatthelinecentreandwidthbefixedrelativeto ofthesource,asinWambsganss(2006): thatofMgII.Itwasnotreliablydetectedatthesecondepoch.Both (cid:5) (cid:3) (cid:4) ofthewiderGaussiancomponentsforHβ/MgIIappeartoundergo y(t)= y2+ t−t0 2, t := DdθE, (4) ablueshiftovertheperiodofobservationsandHβshowsevidence 0 t E v E ⊥ ofanincreaseinFWHMalso. Forthistarget,anabsorptiondoubletwasseenthatmayindicate where y0 is the impact parameter at t0 and v⊥ is the transverse velocity of the lens relative to the observer/source line of sight. thepresenceofaninterveninggalaxy.ItisconsistentwithMgIIata TheEinsteintime-scale,t ,definesacharacteristictime-scalefor lowredshiftandishighlightedinFig.6.Thisfeaturehasonlybeen E the lensing event. If one includes an additional background flux reliablydetectedinthethirdepoch,likelyduetothehigherSNR achievedintheblue.Here,thesuspectedabsorberliesatz=0.18 contribution from the host/lens galaxy, this model has seven free andisfarlesslikelytobeanoutflowintrinsictotheAGN(z=0.62) parameters.ThesearelistedinTable4. asmightbethecasefortheabsorptionseeninJ142232.Ifinoutflow, Inordertoexplorethismodelinmoredetail,EMCEE,aBayesian thevelocitywouldbeinexcessof0.25c(i.e.(cid:6)0.01c).Thisisseen model fitting package that uses a Markov chain Monte Carlo insomehigh-ionizationspeciesbutisperhapsimplausiblyhighfor theselow-ionizationlines. Table4. Freeparametersinthesimplemi- crolensingmodel. 4 MICROLENSING MODELS: TECHNIQUES Parameter Description This section will summarize the methods used for exploring the Ml Lensmass suggestionthattheseAGNtransientsaremicrolensingevents.The v⊥ Transversevelocity resultswillbedetailedinSection5. t0 Mid-pointepoch Weconcentrateonmodellingeventsduetosingleisolatedlensing zd Lensredshift stars,ratherthanthecomplexmagnificationmapsduetomultiple y0 Impactparameter starsinthelineofsightnormallyconsideredinquasarmicrolensing Fs Sourceflux(pre-lensing) studies (e.g. Wambsganss 1992; Lewis & Irwin 1995; Vernardos Fb Backgroundflux(unlensed) MNRAS467,1259–1280(2017)