Mon.Not.R.Astron.Soc.000,1–5(2002) Printed5February2008 (MNLATEXstylefilev2.2) Asymmetric Wolf-Rayet winds: implications for GRB afterglows. ⋆ J. J. Eldridge . Astronomy Research Centre, School of Maths & Physics, Queen’sUniversity Belfast,Belfast, BT7 1NN, Northern Ireland, UK 7 0 0 2 n a ABSTRACT J RecentobservationsofWolf-Rayet(WR)binariesWR151andWR155inferthattheir 5 stellar winds are asymmetric. We show that such asymmetries can alter the stellar- 2 wind bubble structure, bringing the wind-termination shock closer to the WR star. If the wind asymmetry is caused by rotation, the wind density and distance to the 1 v wind-terminationshockarebothdecreasedalongtherotationaxisbyafactorofafew 7 fortheobservedequator-to-polewinddensityratioofWR151.Ifthisasymmetrylasts 0 until core-collapse the time taken to reach the wind-termination shock by supernova 7 ejecta or a gamma-ray burst jet is reduced. This leads to a distorted structure of the 1 supernova ejecta and makes it more likely a constant density environment is inferred 0 from gamma-ray burst afterglow observations. 7 0 Key words: stars: Wolf-Rayet – circumstellar matter – stars: individual: WR151 – / stars: winds, outflows – gamma-rays:bursts h p - o r t 1 INTRODUCTION some GRB are found to occur in constant density en- s a vironments rather than a freely expanding wind where : Wolf-Rayet (WR) stars are the most massive stars in the ρ ∝ 1/r2 (Chevalier, Li & Fransson 2004; Panaitescu 2005; v final stages of evolution and have lost (or in the process of van Marle et al. 2006). Stellar-wind bubbles do have a re- i losing the last remnants of) their hydrogen envelopes. Ob- X gion where the wind density structure is approximately a servationally they are very luminous, with broad emission r constantdensity.Theexpandingwindgoesthroughawind- lines indicating a fast and dense wind. WR stars are the a termination shock (a hydraulic jump) when the ram pres- preferred progenitors of type Ibc supernovae (SNe) due to sureofthefree-winddropsbelowthepressureofthestalled- the lack of a hydrogen envelope. They have also become windregion(Castor, McCray & Weaver1975).Thisstalled- theleading candidates to betheprogenitors of Gamma-ray wind region has a density profile that is roughly constant. bursts (GRB). A GRB is thought to occur at core-collapse Therefore if the free-wind region is small, the GRB jet can if a black hole forms and thecore material surrounding the traverse it before the afterglow emission begins and a con- blackholepossessesenoughangularmomentumtoresidein stant density environment would be inferred. Simulations an accretion disk. This disk can then feed material to the show it is very difficult to shrink the free-wind region to a blackholeresultinginahighlyrelativisticjet.Iftheprogen- smallenoughradiusforittoremainunobservedintheGRB itoriscompactthejetwillemergeatthesurfacetoproduce afterglow (van Marle et al. 2006; Eldridge et al. 2006). The thepromptemissionandlatertheafterglow.Thesupporting required radius is roughly 1018cm, a fraction of a parsec evidenceisthatGRBsoccurinstarformingregionsoftheir (Chevalier, Li & Fransson2004).Thereareanumberofpos- hostgalaxies,someGRBshaveassociatedtypeIbcSNeand sible physical processes that might shrink the free-wind re- someafterglowlightcurvesareconsistentwiththejetpropa- gionasoutlinedbyvan Marle et al.(2006)butfactorsmust gatinginafree-wind environmentexpectedaroundmassive becombinedtoproduceafree-windregionwiththerequired stars(forarecentreviewseeWoosley & Bloom(2006)).The radius. evidence for stellar progenitors from absorption lines in af- One possible process that may have an effect is asym- terglow spectra has weaken due to the recent findings of metry in WR star winds. The cause of asymmetric winds Chen et al. (2006). would typically be the result of stellar rotation or du- However when the circumburst environment (CBE) plicity, however, as we discuss below there is some uncer- density structure is inferred from afterglow observations tainty in predicting how rotation affects the WR winds. Wind asymmetry can be estimated by performing po- larimetry on stars as asymmetry induces strong polari- ⋆ E-mail:[email protected] sation (Brown, McLean & Emslie 1978). Polarimetry sur- 2 J.J. Eldridge veys of WR stars have been performed and for most the orbit shrank to its current size, the hydrogen envelope WR stars the polarisation is low with only 20 percent was lost and the helium core became a WN star. For the of the stars showing any noticeable intrinsic polarisation secondary,weassumethattheprimary’shydrogenenvelope (Harries, Hillier & Howarth 1998). The observations indi- waslostrapidlyandthesecondaryaccretedverylittlemate- cateamass-lossasymmetrybetweenthehighest andlowest rial from the primary. Therefore we assume that the initial wind density to be between a factor of 2 to 3. There is one mass of the secondary was similar to its current mass of important case of stronger polarisation from the close WR 30M⊙. When we compare the wind strengths we find the binaryWR151(Villar-Sbaffi et al.2006).Inthissystemthe primarywinddominatessoweignorethesecondary’swinds light is polarised by 4 percent and modelling indicates that when modelling the wind bubble. theequator-to-pole density ratio (α) is 5. Our stellar-wind bubblemodel is made using thesame Villar-Sbaffi et al. (2006) suggest that the difference is codeasdescribedinEldridge et al.(2006).Weusethemass- due to rotation redistributing the mass-loss from the pole loss rates and wind velocities from the solar metallicity to the equatorial region as in the wind-compressed zone 60M⊙model with an initial ISM density of n0 = 10cm−3. model of Ignace, Cassinelli & Bjorkman (1996). If this is We turn on the wind asymmetry when the star has lost its the case then there is a direct implication for the CBE of hydrogen envelope and assume that M˙(θ) ∝ 1+β(sinθ)2 GRB progenitors which are thought to be rapidly rotat- where beta is β = α−1. We normalise the mass-loss rate ing WR stars but the effect of stellar rotation on stellar so that thetotal mass-loss from thestar is unchangedfrom winds is somewhat uncertain. The effect on B star winds the symmetric case. Also we assume that the wind speed has been well studied. Bjorkman & Cassinelli (1993) and isunchanged.Weselect thestellar-wind bubblemodelwith Ignace, Cassinelli & Bjorkman (1996) first suggested that an age of 3.8 Myrs as the mass-loss rate and wind velocity inertial effects due to rotation reduce the polar wind den- are broadly in agreement with the observed parameters of sity and increase the equatorial wind density. A more rig- WR151. orous approach from Owocki, Cranmer & Blondin (1994), Figure1showsagreyscaleplotofthewindbubbleden- Owocki, Cranmer & Gayley (1996) and Petrenz & Puls sity structure with and without the asymmetric mass-loss. (2000) suggest the converse is true when the distortion of The dense shell in the stalled-wind region is the RSG wind thestellarsurfacebyrotationandtheradiativedrivingforce that has been swept up by the faster WR wind. The differ- are taken into account. ence between the symmetric and asymmetric cases is clear However only Ignace, Cassinelli & Bjorkman (1996) with the distance to the wind termination shock in the po- have considered WR winds. WR winds are different from lar direction reduced by nearly a factor of two. This is due winds of other hot stars as they are optically thick and hy- to the reduction in wind density in the polar direction and drodynamicmodelsofthewindsarequitecomplexandcom- therefore a reduction in the ram pressure compared to the putationalintensive(Grfener & Hamann2005).Theresults stalled-wind ambient pressure which remains constant. forBstarwindscannotbedirectlyappliedtoWRwindsbe- Figure 2 compares density profiles through this envi- cause although they are both line driven the optical thick- ronment. In this plot it is easier to compare the change ness may cause important differences and possibly result in the position of the wind termination shock in the polar in the inertial terms dominating and causing the winds to andequatorialdirections.Itisimportanttonotethatwhile be distorted as predicted by Ignace, Cassinelli & Bjorkman the polar distance is decreased the equatorial distance to (1996). thewindterminationshockisincreasedbecausetheslightly In this letter we use the observed asymmetry of higher density of the wind leads to a greater ram pressure. WR151’swindtoproduceamodelwindbubbleforthissys- Also in these plots the asymmetry we have introduced has tem and study the effect which the asymmetry has on the onlyjustbeenswitchedonwithinthesimulation andthere- bubble structure. However we do not consider what might fore thestalled wind region is similar between the two sim- causetheasymmetry.Thisavoidstheuncertaintyinthedif- ulations. ferent models for wind distortion byrotation. Wethen per- If we continue the simulations for another 0.34 Myrs form a small parameter survey varying the initial interstel- up until the end of the stellar models (about a year before lar medium (ISM) density and the size of the wind-density core-collapse) and we assume the asymmetry remains the asymmetry and estimate the time it would take a GRB jet distortionincreasesasthestalled-windregionisalsoaltered. to reach thestalled-wind region. Thisisimportantasthesupernovaorgamma-rayburstthat mayoccuratthattimewillpropagatethroughthisenviron- ment. The resulting wind-bubble structures are shown in Figures3and4.Bothwithandwithouttheasymmetrythe 2 THE CURRENT AND FUTURE STELLAR stalled-windregionhassomedenserswirls.Thesearedueto WIND BUBBLE OF WR151. instabilitiesinthewind(Garcia-Segura, Langer & Mac-Low Firsttomodelthestellar-windbubblewemustselectastel- 1996)thathavegrowntoproducegreatdensityvariationsas larmodeltouse.InWR151thecurrentmassoftheWRstar foundbyEldridge et al.(2006).Wefindwiththeasymmet- is estimated to be 23M⊙. To obtain the correct mass from ric wind acting for a long time the magnitude and number oursinglestarmodelweadoptasolarmetallicity(Z =0.02) of these swirls are greater. Overall we find the distance to starwithaninitialmassof60M⊙ thathasalongWNphase. the wind-termination shock, RSW is decreased in the polar WethereforeassumethatWR151wasinitiallyawidebinary direction by a factor of approximately 3. where the 60M⊙ star became a red supergiant, filling its If a SN was to occur in such a medium the ejecta RocheLobeandgrowingtoaradiusgreaterthanthebinary would reach the wind-termination shock in the polar direc- orbitsothatcommonenvelopeevolutionoccurs.Duringthis tion sooner than along the equatorial region and therefore Asymmetric Wolf-Rayet winds: implications for GRB afterglows. 3 Figure 1.Greyscaleplotsofthestellar-windbubblestructurewithin12.5parsecsoftheWRstar.TheWRstarhadaninitialmassof 60M⊙ and is pictured here at an age of 3.8Myrs, the initial ISM density was 10cm−3. The left panel shows the case for a spherically symmetric windwhile the right panel shows the case when the equator-to-pole wind density ratio, α=5. The polar direction is along thex-axisinbothfigures. deceleratemorequicklybyaccretingmorematerial.TheSN remnant would be distorted to an oblate shape rather than beingspherically symmetric.IfaGRBwastooccurinsuch astructuretheafterglowjetwouldreachthestalled-windre- gioninashortertimeandthereforetheafterglowwouldap- pear to be traversing a constant density environment. This is in comparison to the symmetric wind bubble case where a freely expandingwind would be inferred. There is another reason why the RSW is more un- even in Figure 3. Eldridge et al. (2006) found that the po- sition of the wind-termination shock is not alway directly related to the initial ISM density as the analytic relation of Castor, McCray & Weaver (1975). This is because if the wind speed and wind density of the source star are vary- ing on the same timescale as the time it takes for the wind toreachthewindterminationshockthentheboundarywill move.Asthattimenowvariesbetweenthepoleandequator so too does themagnitude of this timescale effect. 3 A PARAMETER SEARCH. Figure2.DensityprofilesthroughthesimulationsshowninFig- We have performed a parameter search varying the wind ure 1. The solid and dotted lines are through the spherically- symmetric wind while the dashed and dash-dotted lines are asymmetry, α, and the initial ISM density, n0, around the through the asymmetric simulation along the pole and equator star to investigate the resulting change in the polar dis- directionsrespectively. tance to the wind-termination shock. In table 1 we show the distance in parsecs to the wind termination shock and thewinddensityalongthepolardirection.Clearlytheeffect ofincreasingthewindasymmetryisthatbothRSW andA∗ similar timescale the wind-termination shock will respond decrease.TheeffectofincreasingtheinitialISMdensityde- to any change on a different timescales. creases RSW, as expected. There is one exception when at Withthesedistancesandwinddensitiesitispossibleto the highest ISM density, RSW increases. This is due to the estimatehowlongitwouldtakeforaGRBjettotraversethe wind-terminating shock becoming highly distorted so the distancetothebeginningofthestalled-windatRSW.Rather timeforthewindtotravelfromthestartotheshockvaries than perform adetailed afterglow calculation we usea sim- from pole to equator. Therefore if the wind is varying on a plemodelofisotropicrelativisticejectawithEiso=1053ergs 4 J.J. Eldridge Figure 3.Similarto Figure1 but 340,000 years later, a few years beforecore-collapse. The leftpanel shows the case for aspherically symmetricwindwhiletherightpanel shows thecasewhenthe equator-to-pole winddensityratiois5. Thepolar directionisalongthe x-axisinbothfigures. Table 1 show that the asymmetries greatly reduce the time for thejet to reach RSW. However a high density ISM is still required to reduce the size of the wind-bubbles. The timeofaroundanhourisshortenoughthattheearlyX-ray afterglow of GRBs may show some sign of occurring in a free-wind region while later emission may show a constant density profile. This agrees with Panaitescu et al. (2006) where for the first few hours of a sample of GRB after- glows a free-wind CBE is preferred while after a few days a constant density CBE become more common (Panaitescu 2005). Therefore we may see the transition from free-wind tostalled-windregionsinsomeGRBafterglowswithoutany bump in the lightcurve when the jet encounters the wind termination shock.Itisimportanttonotethatourtimees- timate is approximate, if for example the Lorentz factor or thekineticenergyoftheejecta ishigherthenthetimescan be reduced further. Although our results demonstrate that evenmoderateasymmetriesinthewindreducethetimesby an order of magnitude. 4 DISCUSSION AND CONCLUSIONS. Figure4.DensityprofilesthroughthesimulationsshowninFig- ure 3. The solid and dotted lines are through the spherically- While the asymmetry in WR winds of WR151 and other symmetric wind simulation while the dashed and dash-dotted WR stars could be dueto rotation, duplicity or some other lines are through the asymmetric simulation along the pole and factor, it is the asymmetry itself that has a large effect equator directionsrespectively. on the stellar-wind bubble structure. It may also affect the evolution of the star since angular momentum loss will depend on how much mass is lost at equatorial latitudes andΓinitial=100.Astheejectaexpandsitsweepsupmate- (Meynet & Maeder 2007). It will effect the evolution of a rial decreasing it’s velocity.Therefore shorterdistances and SNorGRBifitoccurswithinsuchadistortedwind-bubble. lower wind-densities both shorten the time for the jet to SphericallysymmetricSNejectawillbeforcedintoanoblate reach RSW. We note that we calculate the observer time form. For a GRB if we assume that the observed asymme- that includes the effect of time dilation by viewing along try is caused by rotation and that the direction of the low- the expanding ejecta direction of motion. However it does est wind density is the rotation axis then the GRB jet will notincludecosmological time-dilation thereforethesetimes propagateinthatdirection.Thereforethemoreasymmetric are a lower limit and will belonger for higher redshifts. the WR wind the more quickly it will encounter the wind- Asymmetric Wolf-Rayet winds: implications for GRB afterglows. 5 InconclusionsomeWRstarsareobservedtohaveasym- Table 1. Position of the wind-termination shock from the WR stars with different initial ISM densities and different equator- metric winds and this has a large effect on the position of to-pole density ratios during the WR phase. Also listed is the thewind-terminationshock.Thereductionofthewindden- winddensityinthepolardirectionforthedifferentdensityratios, sityinonedirectioncanleadtothewind-terminationshock A∗ =(M˙/10−5M⊙yr−1)/(vwind/1000kms−1). Also show is the movingcloser tothestar. IfaGRBjet was totravelinthis timetakenforejectatoreachthewind-terminationshockRSWfor direction then with the lower wind density and shorter dis- asimpleisotropicmodelwithEiso=1053ergsandΓinitial=100. tance, a constant density CBE is more likely to be inferred from theobservations of the afterglow. n0/cm−3= 103 102 101 RSW tSW RSW tSW RSW tSW α A∗ /pc /hrs /pc /hrs /pc /hrs ACKNOWLEDGEMENTS 1 0.66 0.76 24.4 2.67 794 5.25 5831 This work, conducted as part of the award “Understand- 2 0.33 0.47 7.2 2.02 356 3.80 2227 ing the lives of massive stars from birth to supernovae” 3 0.22 0.37 4.1 1.89 295 3.31 1485 made under European Science Foundation EURYI Awards 5 0.13 0.31 2.7 1.41 130 2.18 443 scheme. JJE would also like to thankAllard-Jan van Marle 7 0.09 0.25 1.6 1.08 62.2 2.18 443 andNorbertLangerforusefuldiscussion,SeppoMattilafor 10 0.07 0.31 2.7 1.08 62.2 1.70 219 thecarefulproofreadingthisletterandtherefereefortheir 15 0.04 0.22 1.3 0.70 19.7 1.05 57.6 comments that haveimproved this letter. 20 0.03 0.20 1.1 0.66 16.9 1.05 57.6 REFERENCES terminationshocktakingonlyafewhours.Furthermorebe- cause the wind density is also reduced the jet will also be Bjorkman J.E., Cassinelli J.P., 1993, ApJ,409, 429 decelerated less as it will sweep up less material. Brown J.C., McLean I.S., Emslie A.G., 1978, A&A, 68, These factors together indicate that the afterglow jet 415B is more likely to reach the constant density environment Castor J., McCray R., WeaverR., 1975, ApJ,200, 107 quickly so the free-wind environment does not affect the Chen,H.-W., ProchaskaJ.X., Bloom J., Ramirez-RuizE., afterglow. This is suggested by the times listed in Table Dessauges-Zavadsky M., Foley R.J., 2006, submitted to 1. Combined with the other possible mechanisms discussed ApJ. by van Marle et al. (2006) to reduce the wind-termination Chevalier R.A.,Li Z-Y., Fransson C., 2004, ApJ,606, 369 shock distance we should not be surprised that many GRB Eldridge J.J., Genet F., Daigne F., Mochkovitch R., 2006, afterglows apparently occur in constant density environ- MNRAS,367, 186E ments rather than free-wind environments. Also there is Garcia-SeguraG.,LangerN.,MacLowM.-M.,1996,A&A, some evidence that stellar wind mass-loss rates are lower 316, 133 thancurrentlyusedin stellarmodels with starslosing most Grfener G., Hamann W.-R., 2005, A&A,432, 633G oftheirmassinexplosiveoutbursts(Smith & Owocki2006). Harries T.J. et al., 2003, Nature, 423, 847 If mass-loss rates and therefore wind densities are reduced Ignace R., Cassinelli J.P., Bjorkman J.E., 1996, ApJ, 459, again a constant density environment becomes more likely. 671I We must also consider that GRB progenitors may re- LangerN.,Garca-SeguraG.,MacLowM.,1999,ApJ,520, quire very rapid rotation. If this is the case then the mod- L49 els of Ignace, Cassinelli & Bjorkman (1996) may not apply Meynet G., Maeder A., 2007, A&A in press, as pointed out by Owocki, Cranmer & Blondin (1994) and astro-ph/0701494 Owocki, Cranmer & Gayley (1996). If this is the case the OwockiS.P.,CranmerS.R.,BlondinJ.M.,1994,ApJ,424, reverse of our argument will be trueand the wind termina- 887 tion shock distance along the polar direction will be much Owocki S.P., Cranmer S.R., Gayley K.G., 1996, ApJ, 472, greaterandthewindmuchdenser.Itisdifficulttoknowex- L115 actly which model is correct, not becausethemodels of the Panaitescu A., 2005, MNRAS,363, 1409 wind acceleration are limited but rather because the stel- Panaitescu A., Mszros P., Burrows D., Nousek J., Gehrels lar evolution models are limited by being only 1D models. N.,O’Brien P., Willingale R.,2006, MNRAS,369, 2059P In the case of rapid rotation, the stellar structure will be- Petrenz P., Puls J., 2000, A&A,358, 956 comedistortedand1Dmodelsgivesomeindicationofwhat Smith N., Owocki S.P., 2006, ApJ,645L, 45S mightoccurbuttheycannotfully modelthisinherently3D van Marle A.J., Langer N., Achterberg A., Garcia-Segura problem. G., 2006, A&A,in press By using the observed results on WR winds of what Villar-Sbaffi A., St-Louis N., Moffat A.F.J., Piirola V., we assume is rotation (or duplicity) then we bypass this 2006, ApJ, 640, 995 uncertainty. One remaining problem is that the WR stars Villar-Sbaffi A., St-Louis N., Moffat A.F.J., Piirola V., observed to have asymmetries to date are solar metallicity 2005, ApJ, 623, 1092 WRstarsintheMilkyWayanditisthoughthatGRBpro- Woosley S.E., Bloom J.S., 2006, ARA&A,44, 507 genitors are lower metallicity WR stars. Therefore a study intothepolarisationandasymmetryoflowermetallicityWR stars in the LMC and SMC would be of great interest (e.g. ThispaperhasbeentypesetfromaTEX/LATEXfileprepared bythe author. Vink,in prep.).