Mon.Not.R.Astron.Soc.000,1–15(2015) Printed13January2016 (MNLATEXstylefilev2.2) Modelling the circumstellar medium in RS Ophiuchi and its link to Type Ia supernovae R. A. Booth,1,2∗ S. Mohamed3 and Ph. Podsiadlowski2 1InstituteofAstronomy,UniversityofCambridge,MadingleyRoadCambridge,CB30HA,UK 2DepartmentofAstrophysics,UniversityofOxford,KebleRoad,OxfordOX13RH,UK 3SouthAfricanAstronomicalObservatory,POBox9,Observatory7935,CapeTown,WesternCape,SouthAfrica 6 1 0 13January2016 2 n a ABSTRACT J Recent interpretations of narrow, variable absorption lines detected in some Type Ia super- 1 novaesuggestthattheirprogenitorsaresurroundedbydense,circumstellarmaterial.Similar 1 variations detected in the symbiotic recurrent nova system RS Oph, which undergoesther- monuclear outbursts every 20 years, making it an ideal candidate to investigate the origin ] of these lines. To this end, we present simulations of multiple mass transfer-novacycles in E RS Oph.We findthatthe quiescentmasstransfer producesa dense, equatorialoutflow,i.e., H concentratedtowardsthe binaryorbitalplane,andanaccretiondisc formsaroundthe white . dwarf.Theinteractionofasphericalnovaoutburstwiththeseasphericalcircumstellarstruc- h p turesproducesabipolaroutflow,similartothatseeninHSTimagingofthe2006outburst.In - ordertoproduceanionizationstructurethatisconsistentwithobservations,amass-lossrate o of 5×10−7M⊙yr−1 from the red giant is required.The simulations also producea polar r accretionflow,whichmayexplainthebroadwingsofthequiescentH lineandhardX-rays. t s Bycomparingsimulatedabsorptionlineprofilestoobservationsofthe2006outburst,weare a abletodeterminewhichcomponentsariseinthewindandwhichareduetothenovae.Weex- [ plorethepossiblebehaviourofabsorptionlineprofilesastheymayappearshouldasupernova 1 occurinasystemlikeRSOph.OurmodelsshowsimilaritiestosupernovaelikeSN2006X, v butrequireahighmass-lossrate,M˙ ∼ 10−6 to10−5M yr−1,toexplainthevariabilityin 5 ⊙ SN2006X. 3 6 Keywords: binaries:symbiotic–circumstellarmatter–stars:individual(RSOph)–stars: 2 winds,outflows–stars:novae,cataclysmicvariables–supernovae. 0 . 1 0 6 1 INTRODUCTION is close to the Chandrasekhar mass (Yaronetal. 2005). Com- 1 bined with the fact that the white dwarf mass may be increasing : RSOphiuchi (RSOph) isasymbioticrecurrent nova; thesystem v (Hachisu,Kato&Luna 2007; Hernanz&Jose´ 2008), this makes Xi consists of a red giant and a white-dwarf companion that under- RSOphanidealcandidateTypeIasupernovaprogenitor. goes nova outbursts approximately every 20 years. The most re- r centoutburstwasin2006, inwhichasustainedmulti-wavelength A further possible connection between RS Oph and Type Ia a campaign resulted in improved estimates of the explosion pa- supernovae (SNeIa) isthedetection of narrow, time-variable ab- rameters and geometry. Resolved VLBI and HST images con- sorptionlines,whichmayberelatedtothepresenceof acircum- firmedtheasphericalnatureofthenovashell(O’Brienetal.2006; stellarmediumaroundthesupernovaprogenitor.Time-variableab- Bodeetal. 2007; Rupen,Mioduszewski&Sokoloski 2008), and sorption lines were first detected in the Type Ia supernova SN early X-ray emission from the outburst showed evidence of in- 2006X (Patatetal. 2007), in which the NaI D lines were first teraction between the nova and a dense circumstellar medium observed to strengthen around maximum light, and then weaken (CSM) (Sokoloskietal. 2006; Vaytetetal. 2011). The outburst by 60 days after the explosion. This behaviour was suggested to was characterised by low ejecta mass, Mej ∼ 10−7 to originate in surrounding nova shells, which first recombine and 10−6 M⊙ (Sokoloskietal.2006),andhighejectavelocity,vej ∼ subsequently are swept up by the supernova. The RS Oph – su- 4000kms−1(Das,Banerjee&Ashok2006;Anupama2008).The pernova connection is further supported by the 2006 outburst of ejecta mass and velocity estimates, together with the fast decline RS Oph, in which similar behaviours were seen in the NaI D of the nova (Hounselletal. 2010), indicate that the white dwarf and CaII H & K lines (Patatetal. 2011). Time-variable absorp- tionhassincebeendetectedinseveralother SNeIa(Simonetal. 2009; Blondinetal. 2009; Stritzingeretal. 2010) and statistical ∗ E-mail:[email protected] studies suggest that 20 to 25 per cent of SNe Ia may have pre- (cid:13)c 2015RAS 2 Booth,Mohamed& Podsiadlowski supernova outflows (Sternbergetal. 2011; Maguireetal. 2013). ing (Mohamed 2010; Mohamed,Mackey&Langer 2012). For Theoriginof theselineshasbeen thesubject of arange of theo- temperatures above 104K, the NEI [Fe/H] = −0.5 cooling reticalstudies,fromredgiant windsandrecurrentnovae (Chugai curves from Sutherland&Dopita (1993) were used. Following 2008; Moore&Bildsten 2012) totidal tailsejected from double- Mohamed,Mackey&Langer(2012),fine-structureandmolecular degeneratesystems(Raskin&Kasen2013)orevennovaeinHe+ coolingprocessesincludingH ,COandH Owereusedattemper- 2 2 COsystems(Shen,Guillochon&Foley2013). aturesbelow104K. Spectroscopic measurements of the velocities of both stars For the binary parameters, we assumed the values derived has enabled an accurate determination of the orbital period, from the spectroscopic orbit for RS Oph (Brandietal. 2009) in 453.6 days, mass ratio, q = 0.6 and an eccentricity, e = 0 whichthebinaryorbitalperiodwas453.6days.Theinternalstruc- (Dobrzycka&Kenyon 1994; Shoreetal. 1996; Fekeletal. 2000; tures of the white dwarf and red giant were not modelled in de- Brandietal. 2009). Combined with the estimate for the white- tail,insteadtheirgravitywastreatedassumingpointpotentialswith dwarfmass,thisallowsanaccuratedeterminationofthered-giant massesof1.38M⊙and0.8M⊙,respectively.Themasslossfrom (RG)mass,MRG ≈0.8M⊙,inclination,i≈50◦andseparation, thered giant was modelled by injecting the particles closeto the a ≈ 1.48AU. Measurements of the rotation period of the giant surfacetogiveamass-lossrateofapproximately10−7M⊙yr−1. suggest that the giant is close to fillingitsRoche lobe unless the Theparticles were injected with alow velocity, v = 20kms−1, redgiantissuper-synchronous,i.e.therotationperiodofthegiant much less than the escape speed of the RG, vesc ≈ 55kms−1. is less than that of the binary (Zamanovetal. 2007; Brandietal. While this is sufficient to lift the material beyond the red giant’s 2009).SuchtightconstraintsonthephysicalparametersmakeRS Rochelobe,thematerialremainsboundtothebinary.Theescape OphanidealcandidatefortestingtherecurrentnovamodelofSNe velocityisfinallyachievedviacontinuedtidalaccelerationofcir- Ia. cumbinarymaterialbythebinary. Observations provide a number of constraints on the CSM Since our simulations are aimed at studying the large scale in RS Oph. Direct estimates of the quiescent mass-loss rates are structureofthecircumstellarenvironmentofrecurrentnovae,itis somewhatuncertain,varyingfrombetweenM˙ ≈10−8 M⊙yr−1, notfeasibletoalsosimulateindetailtheflowstructureclosetothe as estimated by dust emission (Evansetal. 2007), to M˙ ≈ whitedwarfduetotheverydifferenttime-scalesinvolved(tensto 10−6 M⊙yr−1 from NaI D absorption lines during the 2006 hundredsofyearsforthenovaeandhourstodaysfortheflowac- outburst (Iijima 2008). The NaI D lines provide a measure of cretingontothewhitedwarf).Forthisreasonweuseasimplepre- the wind velocity, from which Iijima (2008) obtained a velocity scriptiontotreataccretionontothewhitedwarf,anddonotinclude of 33kms−1. High-resolution spectroscopy shows a range blue- any explicit mass-lossfromthe whitedwarf or itsaccretion disc, shifted(out-flowing)componentsfrom10to37kms−1.Whilethe andtreat thewhitedwarf asasink. FollowingTheuns&Jorissen slower components were weaker after the nova, the fastest com- (1993)andMohamed&Podsiadlowski (2012),themassofparti- ponentsstrengthened(Patatetal.2011).Thecomplexstructureof cleswithinh˜ofthewhitedwarfisdecreasedbyafactor(r/h˜)2in thewindisfurther demonstrated by thepresence at allepochs of eachtime-step,whereh˜istheaveragesmoothinglengthofparticles ared-shifted(in-flowing)absorptioncomponentwithavelocityof within0.1auofthewhitedwarf.Particlesareremovedoncetheir v.20kms−1. massdropsbelowonepercentoftheirinitialmass.Thiseffectively Understanding how and where the absorption lines form in ensuresnopressurebuildupinthesinkregion. theCSMiskeytounderstanding whether recurrent novae arere- Whilethisundoubtedlyaffectsthestructureoftheflowclose sponsible for the origin of the lines in SN Ia. Simulations have tothewhitedwarf,wehavefound thatitmakesverylittlediffer- already shown strong asymmetry in the mass loss from binary encetothelargescalestructureofthewind(exceptingthecooling systems (Theuns&Jorissen 1993; Mastrodemos&Morris 1998; andtemperatureofthewind,discussedbelow).Despitetheuncer- Mohamed&Podsiadlowski 2007; Walder,Folini&Shore 2008), taintiesassociatedwithourtreatment,theflowclosetothebinary whichneedstobetakenintoaccountinrecurrentnovamodelsand hassomeinterestingfeaturesthatmaybeabletoexplainanumber theirrelationtoSNeIa. of interesting observational properties of RS Oph. Therefore, we In section 2 we present simulations of mass transfer in RS taketheapproachofdescribingthesefeatures,alongwiththeap- Oph. We include multiple novae and build a model of the CSM propriatecaveatsandfurtherpossibleevidenceforthesefeatures. around RS Oph. Using photoionization calculations we constrain The circumstellar wind structure formed after 13 orbits is thequiescentmass-lossrateinRSOph.Insection4.1wecalculate shown in Fig. 1. The structure formed by the interaction of a theoreticallineprofilesbeforeandafterthenovae. Bycomparing slow wind with a binary companion has been discussed in de- themtoobservations,wedeterminetheoriginofthesecomponents tailbyTheuns&Jorissen(1993).Wethereforeonlyrepeatthede- in RS Oph. In section 3 we investigate a polar inflow on to the tailsneededinthepresentwork.Thebinarypotentialconfinesthe whitedwarfasapossibleexplanationforthebroadwingsobserved windtobinaryorbitalplane,formingatwo-armedspiralstructure. inhydrogenlinesduringquiescence.Finally,wediscusstheresults One spiral forms in the flow passing the white dwarf and escap- intermsoftheprospectsforSNeIainsection5. ingthroughoneoftheouterLagrangepoints(seeFig.1,label3). ThesecondformsduetotheinteractionoftheRochelobeoverflow streamthatpassesthroughtheinnerLagrangepoint,withthepart ofthewindpassingthewhitedwarf(flowslabelled1and2,Fig.1) 2 CIRCUMSTELLARMODEL .Theseflowscanbeseeninthetopleftpanel ofFig.1,withthe Rochelobeoverflowstreamflowingfromlefttoright,andthewind 2.1 QuiescentMass-lossPhase arrivingfromthetop. The RS Oph binary system, quiescent mass loss and nova out- The two spirals run into each other, forming a shock. The bursts were modelled using the Smoothed Particle Hydrodynam- streamsmergeafterapproximately1.25orbits.Behindtheshock, ics(SPH)codeGADGET-2(Springel2005),whichhasbeenmod- thespiralbecomesclumpy, drivenbycooling,whichisthermally ified to include binary motion, stellar winds, accretion and cool- unstable around 103K. The clumps form at the resolution scale. (cid:13)c 2015RAS,MNRAS000,1–15 CSM in RSOph 3 v [50 km/s] v [50 km/s] 1011 1011 2 2 1012 1012 1 1 1013 AU] 0 2 3 1013-3cm] AU] 0 1014-3cm] y[ 1 1014 [g y[ 1015 [g r r -1 -1 1016 1015 1017 -2 -2 1016 1018 -2 -1 0 1 2 -2 -1 0 1 2 x [AU] x [AU] v [50 km/s] 102 102 101 101 5 5 100 100 ] ] 2 2 AU] 0 101-cm AU] 0 101-cm [ g [ g y [ z [ S S 102 102 -5 -5 103 103 -5 0 5 104 -5 0 5 104 x [AU] x [AU] Figure1.SpiralstructureformedduringtheRSOphquiescentmass-lossphase.Theredgiantistotheleftofthewhitedwarfandaccretiondisc.UpperPanel: Densitycross-sectionsthroughtheflowstructureandthevelocityrelativetothecentreofmass.Partofthewindfallsbackontothewhitedwarffromabove andbelowtheplaneofthebinary.ThenumbersshowtheRochelobeoverflowstream(1),whichinteractswiththewindsweptupduringtheorbit(2)toform aspiralwind,andthewindescapingthroughtheL1Lagrangepoint(3),whichformsanotherspiral.ThesolidwhitelinesshowthesizesoftheRochelobefor theredgiantandwhitedwarf.LowerPanel:Large-scalestructure,showingthemergerofthetwospiralstreamsandthestratificationofthewind.Thecolour barshowscolumndensityintegratedoverz/y.Atdistancesgreaterthanafeworbitalradii,thewindbecomesballisticandthevelocityisapproximatelyradial andperpendiculartothespirals. (whiletheresolution scaleistypicallylessthanthesizeofstruc- lutionsinceitiscontrolledbythemassflowintothewhitedwarf’s tures in the flow, the physical size of the clumps is likely much Rochelobe,whichsuggestsanexplanationforthevaryingaccre- smaller). tiondiscmass.Sincetheartificialviscosityislowerinhigherreso- An accretion disc forms which extends to the edge of the lutionsimulationsahigherdiscmassisneededtomaintainasimilar white dwarf’s Roche lobe, approximately 1013cm. The forma- accretionrateandthereforethediscmassadjustsitselftogivethe tion of the accretion disc is enabled by the cooling; similarly to same accretion rate in steady state (1–2×10−8M⊙yr−1). Due Theuns&Jorissen (1993), we found no accretion disc forms in totheuncertaintyintheaccretiondiscpropertiesfromthesimula- simulationswherecoolingisneglected.Theouterpartsoftheac- tions,thediscussionwillbeguidedbyobservationswherepossible. cretiondiscareonanellipticalorbitthatremainsalignedalongthe The passage of the whitedwarf through the windof the red linebetween theredgiantand whitedwarf. Theaccretiondiscis giant as the binary rotates gives rise to a component of the wind onlyresolvedverticallybyafewsmoothinglengthsinthesimula- thatfallsontothewhitedwarfandaccretiondisc,aswasoriginally tionsandsincewehaveneglectedsourcesofheatingthethickness seeninthesimulationsofTheuns&Jorissen(1993).Whilethede- of the accretion disc is likely an underestimate. We also find the tails of this flow are dependent on the structure of the accretion massintheaccretiondiscisresolutiondependent(higherinhigher disc, the presence of some material falling onto the white dwarf resolutionsimulations).Theaccretionrateisnotsensitivetoreso- andaccretiondiscreflectsthefactthatthepressurescaleheightin (cid:13)c 2015RAS,MNRAS000,1–15 4 Booth,Mohamed& Podsiadlowski 102 Sincethegasremainingfromthepreviousnovaeismuchlessdense thanthemateriallostduringquiescence,anyerrormadeinthelarge 20 101 scalewindstructureshouldbesmall. Withinthe first day the nova ejecta interacts withthe accre- 10 tiondisc. Sincethe gasinthe accretiondiscismuch denser than 100 the nova ejecta, it immediately distorts the nova shell, restricting AU] 0 101-2cm] tbheecoflmowebiinpothlaer.bTinhaeryreosrubltiitnalgpsltaruncetuarnedsfaorrecsinhgowthneinnoFviag.sh3e.lTlhtoe [ g large wind mass close to the orbital plane causes the nova shells z [ S todeceleratetovelocitiesof100to200kms−1astheentirewind 102 -10 duringthepreviousquiescentphaseissweptupoverthecourseof afewyears.Thisresultsintheformationofanequatorialring-like 103 structure(componentsN1andN2,Fig.3)thatisconsiderablymore -20 densethantherestofthenovashell.Whilethebipolar lobespri- marilyconsistofnovaejecta,theequatorialringconsistsmainlyof 104 -20 -10 0 10 20 sweptupwindmaterialandwillthereforehaveasimilarcomposi- x[AU] tiontotheredgiant.Inthepolardirections,thenovashellrapidly escapestheCSMandremainsclosetoitsinitialvelocity.Asthefor- Figure2.Columndensityinamodelwithnocooling, whichproducesa wardshocktravelsdownthedensitygradientinthepolardirection, lessstronglyconfinedwindstructure.Forcomparison,thecolourbarshows it accelerates, and the swept-up circumstellar gas reaches veloci- thesamecolumndensityrangeasFig.1,butthespatialscaleis2.5times tiesof4000kms−1to4500kms−1.Thehighvelocityofejectain largertodemonstratethelargerverticalextentoftheflow. thepolardirectionsmeansthattheejectacatchupanddriveshocks throughtheequatorialringsofthepreviousnovae.Thisinteraction theaccretiondiscissmallerthanthescaleheightinthepartofthe leadstoclumpformationastheshellcoolsandfurtherenhancesthe windthatisclosethebinary.Thereforethewindthatpassesabove clumpingintheoldshells.Whiletherepeatedshockingandcool- thewhitedwarfisunabletosupportitselfandfallsontothewhite ingoftheoldernovashellscausestheshellstopartiallybreakup, dwarfandaccretiondisc.Thepolarstreamthatarisesinoursimu- multipleequatorialringsmaybevisible. lationscontributesanaccretionrateontothewhitedwarfoforder Wefindthattheaccretiondiscsurvivesthepassageofthenova M˙ ∼5×10−9M⊙yr−1.Insections3and4.1wediscusstheev- sinceitisconsiderably moredense thanthenova ejecta. Thisre- idenceforapolaraccretioncomponent. sultissomewhatsensitivetoourassumptions,sinceinsomeofthe Althoughthewindbecomesballisticbeyondafewbinarysep- lowerresolutionsimulationstheaccretiondiscdidnotsurvive,due arations,thetemperatureofthewindclosetothebinaryisimpor- to the lower mass in the accretion disc at lower resolutions. The tantforcontrollingtheverticalscaleheightinthewindanddepends survivaloftheaccretiondiscwillalsodependonthedetailsofits ontheheatingandcoolingmodelsused.Whilecoolinghasbeenin- structure,suchasitsscaleheight.However,our simulationsmost cludedinthemodelspresented,nocontributiontoheatingfromthe likelyunderestimatethemassintheaccretiondiscandobservations whitedwarf,accretiondiscorredgianthasbeenincluded.Thisre- duringthe2006outburstofRSOphsuggestthatthediscsurvives sultsinlowtemperaturesinthewindclosetothebinaryandawind (Hachisuetal. 2006). The interaction of the nova ejecta with the thatishighlyconfinedtothebinaryplane.Inmodelsinwhichthe accretion disc immediately strips the weakly bound gas from the coolinghasbeensuppressedthewindislessstronglyconfinedto surface of the disc, which gets entrained in thenova shell. Addi- binaryplane,resultinginlowerdensitiesneartheplanebuthigher tionally,oncethenovaejectahaspassedtheaccretiondisc(within densities at higher latitudes, although clear asymmetries are still hoursinoursimulations),weseeevidenceforanadditionalphase present. This is demonstrated in Fig. 2, which shows a model in ofmassloss.Thismasslossoccursbecauseinadditiontostripping which the flow is adiabatic. Even in this extreme case, the wind thegas,thenovadrivesshocksthroughtheaccretiondisc,heatingit remainsasymmetric.ThewindinRSOphislikelytoliebetween up.Thehighpressureintheshockedgasthendrivesanadditional thesetwoextremes–theeffectthishasonrelationshiptoobserva- ‘evaporative wind’ from the surface of the accretion disc, which tionsofRSOphisdiscussedfurtherinsection4.1. lastsforseveraldaysandappearsasanadditionalshell-likefeature (componentA,Fig.3).Asimilareffectprovidesanimportantcon- tributiontothe total mass lost when a supernova interacts witha 2.2 NovaeOutbursts binarycompanion(Wheeler,Lecar&McKee1975). Afterthequiescentmasstransferhasevolvedfor18years,anova Thehighestdensityinthenovashellsisintheequatorialring, shellwasinjectedintothesimulation.Subsequentlytheinteraction which is nH ∼ 106cm−3 2000 days after the nova. This den- betweenmultiplenovaeandthecircumstellarmediumweresimu- sityislower thanthe107cm−3 predicted by 1D analytical mod- lated.Thenovaeweremodelledusingauniformshellfortheejecta, els(Moore&Bildsten2012)scaledtoacomparablemass-lossrate insertedwithinthesinkregionaround thewhitedwarf. Thenova totheeffectivemass-lossrateinthebinaryplane.Thediscrepancy masswastakentobeMej ≈ 2×10−7M⊙,andthevelocitywas maybeduetoavarietyofreasons.Onepossibilityisthatthe3D settoproduceaterminalvelocity,vej ≈ 3500kms−1.Theinter- geometry allowsthe shell tospread out in thevertical directions, nalenergyoftheejectawasinitiallysetto25percentofthekinetic whichisimpossiblein1D.Another possibilityisthatthesimula- energy,ensuringthedynamicsaredominatedbytherampressure. tionsmaynotsufficientlyresolvethedenseshell.Thisintroduces Threenovae,eachseparatedby18years,havebeenmodelled, anuncertaintyintotherecombinationtime-scale,whichdependson eachusingthecircumstellarstructureformedbythepreviousno- thedensitythrought = (n α)−1,whereαistherecombination r e vae.Ratherthanre-computethequiescentmassloss,weassumed coefficientandn istheelectrondensity.Althoughthevelocityof e the presence of the previous novae had no effect on this phase. theabsorption lines formed in thenova shells should bereliable, (cid:13)c 2015RAS,MNRAS000,1–15 CSM in RSOph 5 400 1018 400 1018 1019 1019 200 200 ] ] AU] 0 N1 N2 1020-3cm AU] 0 1020-3cm y[ [g y[ N1 N2 A [g 1021 r 1021 r -200 -200 1022 1022 -400 -400 1023 1023 -400 -200 0 200 400 -400 -200 0 200 400 x[AU] x[AU] 102 102 4000 4000 103 103 2000 104 2000 104 AU] 0 105-2cm] AU] 0 105-2cm] y[ 106 [g z[ 106 [g S S -2000 107 -2000 107 108 108 -4000 -4000 109 109 -4000 -2000 0 2000 4000 -4000 -2000 0 2000 4000 x[AU] x[AU] Figure3.Density structure 2000days after thesecond nova outburst. Upper Panel: Zoom-inoftheequatorial ring. Thelabels show thelocation ofthe equatorialringsfromthefirst(N1)andsecondnovae(N2),alongwiththe‘evaporative’windfromtheaccretiondisc(A).Thefirstnovashell(N1)ismore clumpy,reflectingtheadditionalcoolingphasethatitunderwentafterbeingshockedbytheejectafromthesecondnova.LowerPanel:Largescalestructure ofthenovae,showingthebipolarstructureshapedbytheinteractionwiththeaccretiondiscandthewind.Thevelocityofthebipolarlobesismuchhigher (4000kms−1)thantheslowmovingequatorialring(200kms−1). since it iscontrolled by momentum conservation, theuncertainty Dobrzyckaetal. (1996) used UV observations to derive a hot intherecombination time-scalemeansthestrengthofthesecom- component luminosity of 100 − 300L⊙, and a temperature ponentsremainsonlyqualitative(seeSec.2.3). T ≈ 105K. Although this is the luminosity after the emis- sionhas been reprocessed by theregion close tothe whitedwarf (Anupama&Mikołajewska 1999), it is the appropriate luminos- 2.3 Mass-lossrateestimatefromtheionizationstateofthe ityforcalculatingtheionizationstateofmaterialinthewind.Soft circumstellarmedium X-rays emitted as a result of the novae outbursts also contribute to the ionizing flux. During quiescence an X-ray luminosity of Theionizationconditionsofstellarwindscanbeaneffectivetracer 2×1032ergs−1wasmeasuredbyRXTEin1997,roughly12years ofthemass-lossrate,since,neglectingopticaldepth,thephotoion- afterthe1985outburst(Mukai2008). ization rate from a point source scales as r−2, and the recombi- nation ratescales asr−4 ina one-dimensional wind. This means Thequiescenthydrogenionizationstatewasmodelledusinga thatunlesstheopticaldepthissufficienttoabsorbtheionizingflux Stro¨mgrensphereapproximationforlinesofsightfromeachpar- theentirewindwillbeionized.InRSOph,hydrogenisonlyion- ticle to the white dwarf, with each line of sight treated indepen- izedrelativelyclosetothewhitedwarf(Anupama&Mikołajewska dently and the ionization of each particlefound by balancing the 1999), an idea supported by our NaI D line profile modelling, UV photoionization with Case B recombination, taking into ac- therefore we can use this to estimate the mass-loss rate from the count the optical depth along the line of sight. Case B recombi- redgiant. nation assumes that any photons emitted with E > 13.6eV are During quiescence the dominant ionizing flux in RS Oph reabsorbedlocallybyanotherhydrogenatomandcanbeneglected comes from the hot component, although whether this flux is from the overall recombination rate. The effective photoioniza- dominated by the white dwarf or accretion disc is unknown. tioncross-section iscalculated byaveraging thecross-section for (cid:13)c 2015RAS,MNRAS000,1–15 6 Booth,Mohamed& Podsiadlowski ❍ recombination rate, which depends on the local electron density. ❳ ✥ Wealso include collisional ionization which can dominate in the ✌✌ ✠ ✵✡☞ ❳❍ novae; where thegasbehind theforward shock can reach 106K. Thephotoionization, recombination andcollisionalionizationco- efficients from Verner&Ferland (1996), Verneretal. (1996) and ✵✡☛ ✞ Arnaud&Rothenflug(1985)wereused. ✝ ☎✆ ✵ ③ Since sodium is fully ionized in regions where hydrogen is ✵✡✟ ionized we focused on the ionization state of sodium in the neu- tral hydrogen regions. Since in this region the optical depth to ✲✠ ✵✡✠ UV photons is large, we modelled the ionization of sodium ne- glecting the contribution from the UV continuum short-ward of ✵ 13.6eV.Includingthesephotonserroneouslyionizeshydrogen,but ✲✟ ✲✠ ✵ ✠ ✟ ①(cid:0)✁✂✄ makesonlyaminordifferencetotheionizationrateofsodium.For Figure4.Structureofthehydrogenionizationfractionduringquiescence, photons with energies less than 13.6eV, we assumed the optical foracross-sectionthroughtheredgiantandwhitedwarf.Thewindcloseto depth issmall. Theionization fraction of hydrogen isalso calcu- theredgiantandtheaccretiondiscprovidesufficientopticaldepthtoshield latedintheneutral region, since, atapart fromthehighest densi- thegasbehindthem,elsewherethegasisionized. Thethree-dimensional ties, hydrogen provides the dominant contribution to the electron ionization structureisthereforeconical, centredonthewhitedwarf,with density.Sincetheionizationfractioncanbeverylowinthedens- theadditionalneutralregionbehindtheredgiant.Contoursshowtheden- est parts of the wind we applied a floor to the electron fraction, sciotym.pNaoreteththeemspdairteiacltlsyc,atlheeisdeanpspirtyoxcimonattoeulyrstwshiocweacslewairdlyetahseinedFgieg.o1f.tThoe Xe =ne/nH =10−5,torepresentthecontributionfrommetals. densewindregionneartheRG. Away from the white dwarf, the dominant contribution to the hydrogen ionization rate comes from X-ray ionization, whichwasalsoincludedintheionizationcalculationforsodium. The X-ray ionizing rate was computed using the fits given by photoionizationfromthegroundstateoverablack-bodyspectrum at 105K. The Case B recombination coefficient at 104K from Kozma&Fransson(1992),whichshowthatapproximately40per cent of the X-ray energy deposited goes into ionizing the gas. Ferguson&Ferland(1997)isusedandthephotoionizationcross- The rate of X-ray energy deposited can be estimated as F σ , sectionisfromVerneretal.(1996). X X whereσ istheeffectiveX-rayabsorptioncross-sectionandF TheresultsoftheStro¨mgrencalculationareshowninFig.4. X X istheX-rayflux.Thiswasestimatedusingthecross-sectionfrom Only the accretion disc and wind closest to the red giant pro- Morrison&McCammon(1983),averagedoverablack-bodywith vide sufficient optical depth that the gas becomes neutral, result- temperature2×106K.Theobservedluminosityistranslatedintoa ing in a conical ionization structure, with a larger neutral re- fluxtakingaccountofthefactthattheX-rayemissioncomesfrom gion in the gas shielded by the red giant. The low UV optical theshockednovashells,whichareexternaltotheredgiantwind. depth is in disagreement with Shore&Aufdenberg (1993) and Inthiscase,treatingtheemissionasapointsourcewouldoveres- Anupama&Mikołajewska (1999), who suggest a large optical timate the flux near the binary. Instead the flux is approximated depth isrequired to explain thelineratios, and that thehot com- viaF = L /4πmax(r,r )2,whichgivesthecorrectlimitsfor ponentisembeddedinaneutralwind.Thestrongabsorptionseen X X s r≪r ,wherer istheradiusofthenovashell.Duringquiescence intheBalmerseries(Bulla2013)andthepresenceofmetalabsorp- s s weusedr = 1015cm.Duringoutburstr evolvessincethenova tionlinessuchastheNaIDlinesthatforminthewind(Patatetal. s s shells expand – we used the mean radius of X-ray emitting gas, 2011)alsosupportalargeopticaldepth. whichistakentobethegaswithT >106K. Thelargerdensityclosetothewhitedwarfmustmeanahigher mass-loss rate, since the simulations over-estimate the density in During quiescence we calculated the ionization state in a thewindbecausetheyneglectedphotoionizationheating.Recalcu- steady-stateapproximation,usingthesameluminositiesasforthe latingtheionizationstructureformodelsinwhichthedensityhas Stro¨mgrenspherecalculations.Forthelineprofileevolutionafter been scaled to mimic an increased mass-loss rate, we find that a thenovathequiescentUVfluxwasused,sinceweconsideredthe mass-lossrateM˙ =5×10−7M⊙yr−1issufficienttoproducean evolutionofthelinesaftertheendofthesuper-softsourcephase. ionized bubble around thewhitedwarf withaneutral region out- By this time accretion had resumed in RS Oph (Wortersetal. sideit.Therelativelymodestincreaseinmass-lossrateissufficient 2007),sothequiescentUVfluxisappropriate.ForthesoftX-rays, sincetheoptical depthissensitivetodensity; inahighlyionized we used the observed luminosity from Bodeetal. (2008). At the region, theneutral fractionY ≈ nα/Fσ, whereF isthefluxof end of the super-soft phase, 100 days after the explosion, the lu- ionizingphotons,nisthenumberdensityofhydrogenandσisthe minosity was 5× 1035ergs−1, decaying to the quiescent value ionizationcross-section.Thisresultsintheopticaldepth,τ,obey- by 1200 days after theexplosion. During thistimetheionization ing τ ∝ M˙2. The optical depths for the simulated NaID lines statewascalculatedtime-dependently,assumingthatbothsodium discussedinsection4.1alsosupportahighermass-lossrateinRS andhydrogenwerefullyionizedattheendofthesuper-softsource Oph. phase andthetemperatures weretaken fromtheSPHsimulation. In order to calculate the NaID line profiles, the ionization We found the ionization fraction of sodium in the regions where state of sodium must also be modelled, because its low ioniza- hydrogen is neutral was in the range 10−3 to 10−4 during qui- tionenergy(5.1eV)meanssodiumcanremainphotoionizedinre- escence. Forthenovae, initiallythe hightemperature inthenova gions where hydrogen is neutral. A detailed, time-dependent and shellpreventssodiumfromrecombining.Astheshellsexpandand multi-species photoionization model in 3D is beyond the scope cool,sodiumintheequatorialringisabletorecombine.However, of this work, instead we use a model that captures the impor- atlowerinclinationsthenovaejectaremainsionized.Theeffectsof tant quantities for sodium, i.e., the photoionization rate and the ionizationcanbeseeninthelineprofilespresentedinsection4.1. (cid:13)c 2015RAS,MNRAS000,1–15 CSM in RSOph 7 3 POLARINFLOW riving at a location r in the disc from a volume element d3r′ is Thesimulationsshowthepresenceofapolarinflowinthesystem, with gas falling back onto the binary from 1014cm. The accre- η(r′,ν′)d3r′ F = . (2) tionrateinthiscomponent isapproximately 5×10−9M⊙yr−1 4π|r−r′|2 (Sec.2.1).Polaraccretionflowsareaubiquitousfeatureofbinary Theprobabilitythatthefluxisscatteredthroughanangleθ tothe s windmodels,arisingduetothepassageofthewhitedwarfthrough theobserveris(1−exp(−τ))I(θ ),whereI(θ )∝1+cos2(θ ) s s s the red giant wind (this can be seen in the upper right panel of istheangularpartoftheThomsonscatteringcross-sectionandthe Fig.1). factor(1−exp(−τ))takesintoaccounttheopticaldepththrough The occurrence of such a flow is supported, at least at dis- thedisc.Thetotalluminositycanbewrittenas tances of 1014cm, by thedetection of red-shifted components in absorption lines of Na, Ca and K in RS Oph (Patatetal. 2011). However, thestructureof theinflow within1012cm of thewhite L (ν)=(1−exp(−τ)) η(r′,ν′) I(θ (r′,r))d3r′d2r, dwarfisaffectedbytheuseofasinkinthesimulationandisthere- s 4π|r−r′|2 s Z forenot well known. It isalsopossible that thestructure may be (3) furthercomplicatedbytheoccurrenceofwindsfromtheaccretion wheretheintegraloverd3r′isoverthevolumeoftheemittingre- disc(notincludedinourmodels),since,althoughRSOphisbelow the Eddington luminosity, L/L ∼ 10−2, line opacity maybe gionandtheintegrald2risoverthesurfaceofthedisc.Forcoher- Edd ent scattering the relationship between the emitted and observed sufficientforRSOphtodriveadiscwind(Drew&Proga2000). frequencies, ν′ = ν′(ν,r′,r),wascalculatedby considering the X-ray observations of RSOph during quiescence tentatively Doppler shift between the frame of the emitting gas and the rest supportthepresenceofapolaraccretionflow.Nelsonetal.(2011) frameofthediscatr′,followedbytheshiftbetweentheframeof foundthatthehardX-rayemissioncanbeexplainedbyemission thediscandtheobserver. from shocked gas as it settles onto the white dwarf, and mea- sured an accretion rate 2×10−9M⊙yr−1. They found a maxi- The resulting line profiles for a range of cone opening an- gles and optical depths are shown in Fig. 5. In the conical mod- mumtemperatureof6keV,whichcorrespondstoashockvelocity v≈1.7×103kms−1.ThisvelocityisclosetothewidthoftheHα els, the density in the flow has been adjusted so that each model has the same accretion rate and the observer is at an inclination emissionlinewingsinRSOph(Zamanovetal.2005),suggesting of45◦.Modelswithnegligibleopticaldepthintheaccretiondisc, theHαemissioncouldariseinaninflow. τ =0,producesymmetriclineprofilesmakingitimpossibletotell WeestimatedtheHαemissionfromtheinflowusingasim- whethertheemissionisduetoanoutfloworinflow.Thepresence plified model in which the emission comes from gas in free-fall ofscatteringbreaksthesymmetry,producingprofileswithstronger ontothewhitedwarf.Weneglecttheeffectsofangularmomentum red-shiftedemission. ontheflow,exceptthatwedonotincludeemissionfromtheparts offlowwithv > 1700kms−1,sinceithasbeenshockedtohigh Models with low conical opening angles produce double- peakedprofilesandintermediatemodelscanproducetriplepeaked temperaturesandformstheX-rayemittingcoolingflow.Theinflow emission,althoughscatteringbytheaccretiondiscresultsinasin- istakentobesphericallysymmetric;however,wealsoconsidered glepeakedprofile.Wenotethatwhilethemodelwithanopening anaxis-symmetricconicalstructure,sincetheaccretiondiscmust angleθ = 30◦ andτ = 1mimicsaP-Cygniprofile,theHαpro- blockpartoftheinflow. fileinRSOphcannotbeproducedbythismechanismalone.The Theluminosityisassumedtobeduetorecombination, reason for thisisthat in thecentre of higher Balmer serieslines, L(ν)=hνfα(T) n(r)2φ(ν′)d3r. (1) such as Hβ and Hγ, show P-Cygni profiles in which the flux is ZV belowthelevelofthecontinuum,andabsorptionmustthereforebe The line profile, φ(ν), was taken to be the Doppler profile, the present(Bulla2013). density set viamass conservation, n = n0r02v0/(r2v(r)), where TheP-CygniabsorptioninRSOphcannotbeassociatedwith v2(r) = −2GM(r−1−r−1)+v2.Thefrequencyν′ istherest- apolarinflow;sinceitisblue-shifted,itmustariseinanoutflow. 0 0 frame frequency and φ(ν′) is the Doppler emission profile. The Themostlikelyexplanationisthattheabsorptionarisesfurtherout subscriptzerodenotesquantitiesmeasuredfromthesimulationsat inthewindwherethetemperatureislower.Thisiscompatiblewith r0 = 2.5×1012cm,whichgivesn0 = 5×10−16gcm−3 and observationsofthe2006novainRSOph,whichshowthenarrow v0 = 80kms−1. For the temperature, we use T = 30000K, P-Cygniprofilewasstillpresent2daysafterthenova(Patatetal. which only affect the total luminosity. The emissivity from re- 2011).Fortheabsorbingmaterialtobeoutsidetheshockfrontat combination has been approximated via hνfα(T)n(r)2, where this time the distance to the absorbing material must be at least α is the recombination coefficient and f average number of Hα 5×1013cm. photons per recombination. The recombination coefficient used Furthermore, unliketheHαlinewings, thevelocityabsorp- is that for Case B, in which recombination photons emitted with tion component does not correlate with the motion of either the hv > 13.6eVareassumedtobeimmediatelyreabsorbed. Inthis whitedwarfortheredgiant(Brandietal.2009).Incontrasttothis, case,f ≈0.5(Storey&Hummer1995). thecentralvelocity,asmeasuredbytheHαlinewings,doescorre- Self-absorption in the inflow has been neglected, however, latewiththemotionofthewhitedwarf.Thiscorrelationwouldbe electron scattering froma razor thin accretion disc wasincluded. expectedforapolarinflowmodelsincethelinewingsdependonly Emissionfrommaterialonthefarsideofthediscisreducedbya ontheaccretionrateandwhite-dwarfmass,notthedensityandve- factorexp(−τ),whereτ istheopticaldepththroughthediscand locityawayfromthewhitedwarf.Therefore,wesuggestthepolar scatteringwastreatedinthesinglescatteringapproximation. The inflow as an explanation for the line wings. However, the central scatteredluminosity,L (ν),isderivedbyconsideringthescatter- componentmaybedominatedbyanoutflow. s ingofphotonsemittedatapointr′bytheaccretiondiscatapoint Whiletheassociationof Hαlinewingswithan inflow may r. For an emissivity, η(r′,ν′) = hνfαn2(r′)φ(ν′), the flux ar- be compelling, we note that accretion disc wind models may be (cid:13)c 2015RAS,MNRAS000,1–15 8 Booth,Mohamed& Podsiadlowski 1e28 1e28 1e27 4.5 1.0 5 4.0 0 3.5 0.1 0.8 4 1] 3.0 1 1] 1] - - - z 10 z 0.6 z 3 H 2.5 H H 1 1 1 - 2.0 - - s s 0.4 s 2 g g g r 1.5 r r e e e [ [ [ L L L 1.0 0.2 1 0.5 0.0 0.0 0 -700-350 0 350 700 -700-350 0 350 700 -700-350 0 350 700 v[km s-1] v[km s-1] v[km s-1] Figure5.Hαemissionfromthepolarinflowmodel.Thekeydenotesthescatteringopticaldepthintheaccretiondiscandtheobserverisataninclinationof 45◦.Ineachmodeltheaccretionrateisthesame,buttheopeningangleofthepolarinflowincreaseslefttorightacrossthepanelswithθ=30◦,60◦and90◦ respectively.Notethedifferentluminosityscalesforthedifferentopeningangles. abletoproduce similarlineprofiles.Calculationsoflinesformed higherthanthewidthoftheabsorptionlines,andislikelyassoci- in disc-winds from white dwarf stars have been limited to res- atedwithemissionfromtheionizedregion.Thelineprofileswere onance lines tend to produce absorption profiles for inclinations derived by calculating the optical depth tothe photosphere along i & 70◦ (Progaetal. 2002). However, disc wind models for T linesof sight through theCSM.Thephotosphere wassubdivided Tauri stars can produce similar Hα line profiles to those seen into lines of sight to ensure that the sub-structure along the line in RS Oph (Kurosawa,Harries&Symington 2006). Polarization of sight is resolved. Finally, the flux from each line of sight was measurementsmaybeabletodistinguishbetweeninflowandout- combinedandweightedbyarea. flow models since, intheinflow models, the fractionof scattered Fig.6showstheresultingsyntheticlineprofiles.Forinclina- lightcontributingtothebluewingislarger. tions i & 70◦, the line profile shows both red- and blue-shifted components. Thered-shiftedcomponents ariseinthewindthatis fallingbackontothewhitedwarfandaccretiondisc,andisthere- forestrongestwhenthewhitedwarfisclosesttotheobserverand 4 SODIUMABSORPTIONLINES thelineofsightintersectsthemostmaterial.Theblue-shiftedcom- ponentsarisefurtheroutintheCSM,beyond1014cm,whichhas 4.1 NaIDlinesinRSOph aterminalvelocityof10to20kms−1.Athigherinclinationsthis Sodiumabsorptionlineshavebeensuggestedasausefuldiagnostic component isrelativelystablewithphasebecausethedistantma- ofthecircumstellarenvironmentinbothsupernovaeandtheirpro- terialdominatesthelineprofileandcontributestoalllinesofsight genitorsystems(Patatetal.2007;Chugai2008;Patatetal.2011). independent of phase. At low inclinations the contribution from TheNaIDlines,togetherwithotherstrongmetallinessuchasthe materialatlargedistancesissmaller,duetothelowerdensityfar CaII H &K lines,probe thedensityand ionizationconditions in above the mid-plane. Therefore, the high inclination line profiles the circumstellar environment. We modelled the line profiles ex- aremoresensitivetothematerialclosetothebinary,whichvaries pectedfromthesimulations,withtheaimofunderstandingwhere withphase. theobservedfeaturesinRSOphformintheCSMandwhetherthe The inclination at which the behaviour of the blue-shifted same circumstellar components can be responsible for metal line component changes from stable to phase-dependent depends on variationsobservedinsomeSNIa. howconcentratedtheoutflowis.Forourcoolingmodel(Fig.6)this We have modelled NaI D absorption line profiles directly occursati ∼ 70◦,butformodelswithoutcoolingthelineprofile fromthe simulations, using the resultsfromthe ionization calcu- isstableforalargerrangeofinclinations,withthetransitionat30◦ lation described above. All neutral sodium atoms are assumed to to45◦.InRSOphtheblue-shiftedlinesarerelativelystable,which be in the ground state and contribute to the absorption lines. We suggeststhattheoutflowinRSOpharelessstronglyconcentrated presentlineprofilesfromcalculationsinwhichthemass-lossrate thaninourcoolingmodel,sincetheinclinationisi≈50◦. hasbeenscaledto5×10−7M⊙yr−1,sincethisgivesbetteragree- Theabsorption linesformed inthenova shells areshown in mentwithobservationsbothfortheionizationstructureandtheline Fig. 7. These lineswere modelled fromthe end of the super-soft profiles.WedidnotincludeadetailedtreatmentfortheHIIregion source phase and assume that the central system had reached a intheNaI Dlineionizationcalculation;insteadweneglectedthe quiescent steady-state once more. The evolution of the line pro- contributionfromallmaterialwithin3×1013cm.Theresultsare filesisgovernedbythecompetingeffectsofrecombination,which notparticularlysensitivetothischoiceofradius. strengthen the lines, and expansion which weakens them. Since WeassumethatthecontinuumfluxaroundtheNaIDlinesis the nova shell expands rapidly in the polar directions this com- dominated bytheredgiant. Thepresence ofanemissioncompo- ponent is not dense enough to form strong sodium lines (at 3 to nent inRSOphlimitstheextenttowhichthisistrue(Patatetal. 4×103kms−1).Instead,theabsorptionlineprofilesati.60are 2011). The emission component has a width v ∼ 100kms−1, dominatedbythe‘evaporativewind’,whichhasv .100kms−1. (cid:13)c 2015RAS,MNRAS000,1–15 CSM in RSOph 9 1.0 x u 0.8 l F d 0.6 e z i a l 0.4 a m or 0.2 11.85 12.10 12.35 12.60 N i=60(cid:0) 12.10 12.35 12.60 12.85 1.0 x u 0.8 l F d 0.6 e z i a l 0.4 a m r o 0.2 N i=75(cid:0) 1.0 x u 0.8 l F d 0.6 e z i a l 0.4 a m r o 0.2 N i=90(cid:0) 0.0 -30-20-10 0 10 20 30-20-10 0 10 20 30-20-10 0 10 20 30-20-10 0 10 20 30 V [km/s] V [km/s] V [km/s] V [km/s] Figure6.ModelNaIDabsorptionlinesproducedduringthequiescentmass-lossphase,atinclinationsof60◦(top),75◦(middle)and90◦(bottom).Thekey denotesthenumberoforbitssincemasstransferbeganwheretheredgiantisnearesttotheobserveratphase0.Sincetheblue-shiftedcomponentsarisein thecircumstellaroutflow,theydisappearatlowerinclinationsduetothelowercolumndepth.Thered-shiftedcomponentvarieslessstronglywithinclination angle. Thelargewidthofthelinefori=30◦isaresultoftheevaporative isconsiderablylowerthanthefastestobservedcomponent, which windlyingalongthelineofsighttotheobserver,whileathigher hasavelocityv=−37kms−1.Whileadditionalsourcesofaccel- inclinationslinesofsightcutacrosstheevaporativewind. erationcouldbeinvokedtoexplainthedifference,themostnatural Forinclinationsofi & 60◦,theinteractionofthenovashell explanationcomesfromthenovae,whichshowtheevaporationof with the wind is strong enough to produce an additional compo- gasfromtheaccretiondiscafterthenovahaspassedatvelocities nenttothelineprofile,athighervelocitythanthematerialfromthe ofv∼−50kms−1. accretiondisc.Initiallythecontinuedinteractionofthenovashell Observations before and after the 2006 outburst support withthewindkeeps theshellhot,withcollisionalionizationpre- the suggestion of differing origin for the 10 to 20kms−1 and ventingrecombinationinthesodium.Astheshelleventuallycools 37kms−1components.Sinceaftertheoutburstthelowervelocity andrecombinesstrongabsorptionlinesappear.Whilethecompo- componentwasweaker,whilethehighervelocitycomponent was nentsfromtheevaporativewindweakenovertime,theycanstillbe stronger(Patatetal.2011).Thestrengtheningofthehigh-velocity seen2000daysafterthenova. componentisexpectedifitisassociatedwiththeevaporativewind, The different absorption components in our models can be asitexpands and thelines weaken continually asthe nova shells usedtodeterminetheoriginoftheabsorptioncomponentsobserved evolve. Similarly, lines arising in the wind should be weaker af- inRSOph.Thegasthatfallsbackontothebinaryfromaboveand ter the nova as the nova shells sweep up the wind, which slowly below the plane naturally explain the otherwise unexplained red- replenishesonceaccretionresumes.ObservationsofHαandHβ shifted absorption components seen in RS Oph. Additionally we absorptionspriortothe2006novasupportthis,sincetheyhadthe findtheterminalvelocityintheoutflowis10to20kms−1,which samevelocityastheslowNaIDcomponent,v∼−10kms−1,and (cid:13)c 2015RAS,MNRAS000,1–15 10 Booth,Mohamed& Podsiadlowski 1.0 x u 0.8 l i=30(cid:0) F d 0.6 e z i a l 0.4 a m or 0.2 150 310 530 860 N 310 530 860 2090 1.0 x u 0.8 l i=45(cid:0) F d 0.6 e z i a l 0.4 a m r o 0.2 N 1.0 x u 0.8 l i=60(cid:0) F d 0.6 e z i a l 0.4 a m r o 0.2 N 0.0 -225 -150 -75 0 -225 -150 -75 0 -225 -150 -75 0 -225 -150 -75 0 V [km/s] V [km/s] V [km/s] V [km/s] Figure7.ContributionofthenovaejectatotheNaIDlinesforinclinationsof30◦(top),45◦(middle)and60◦(bottom).Thekeydenotestimeindayssince thestartofthenovaoutburst.Thegreylinemarksthewindvelocityv ≈ −12kms−1,columndensitiesatvelocitiesgreaterthanthiswillbeaffectedby interactionswiththeRGwind.Forinclinationsi.45◦,thehighvelocitycomponentsweakenasthenovashellevolves.Athigherinclinations,forlinesof sightintersectingtheequatorialring,acomponentat150kms−1appearsoncetheringhascooledbelow103K.Thecomponentsat100kms−1fori=60◦ ariseinthegasevaporatedfromtheaccretiondisc.Theevaporativecomponentweakenscontinuallyafterthenova,andmayexplaintheadditionalmaterial seenafterthe2006outburstofRSOph. wereobserved tocontinuously strengthen during theperiod prior weuseasimpleapproximationduetoChevalier(1981)isused tothe2006outburst(Brandietal.2009).Therefore,wesuggesta 2.61×1014t2/3cm t <40.7, cexopmlabninaetidonqufoierstcheentlinmeassos-blsoesrsveadndinnRovSaOmpohd.el providesanatural Rp =(8.86×1013 d1− 1t0d8 2 tdcm tdd >40.7, (4) wheret isthetimeinhdays.(cid:0)This(cid:1)exipressionoverestimatesthepho- d 4.2 RSOphandTypeIasupernovae tosphericvelocityvp = Rp/tbyapproximately25percentwhen compared with photospheric velocities from abundance tomogra- InordertocompareoursimulationsofRSOphwithobservations phy (Stehleetal. 2005; Mazzalietal. 2008). Thesupernova pho- ofTypeIasupernovae,wecalculatedthelineprofilesaswouldbe tospherebeginstorecedeafterapproximately60days,andthesu- seenagainstasupernovaphotosphere.Fortheionizationconditions pernovastartstobecomeopticallythintocontinuumradiationand inthesupernova,wefollowthehypothesissuggestedbyPatatetal. a model for the emissivity should be used instead of an approxi- (2007) for explaining theobservations of SN2006X –thesuper- matephotosphere.Howeverthisdoesnotoccuruntilaftertheline novainitiallyionizesthecircumstellarmedium,whichthenbegins profileshavefinishedevolving. torecombine20daysaftertheexplosion.Thesupernovaistakento Weassumetheexplosionoccursaftertheendofthesimula- beanexpandingspherewithavelocityof20,000kms−1,which tions,oncethreerecurrencetimesoccur,whichcorrespondstothe sweepsupthecircumstellarmedium,andtheforthephotosphere lowest densityof, andlargest thedistancetotheinner-most nova (cid:13)c 2015RAS,MNRAS000,1–15