A detailed X-ray investigation of ζ Puppis III. A spectral analysis of the whole RGS spectrum

0.8 MB·English

by  A. Hervé

Astronomy&Astrophysicsmanuscriptno.zetaPupIIIaa-correction3 c ESO2013
(cid:13)
January23,2013
A detailed X-ray investigation of ζ Puppis
III. A spectral analysis of the whole RGS spectrum ⋆
A.Herve´1,G.Rauw1,andY.Naze´1,⋆⋆
Grouped’AstrophysiquedesHautesEnergies,Institutd’AstrophysiqueetdeGe´ophysique,Universite´deLie`ge,Alle´edu6Aouˆt,Baˆt
B5c,4000Lie`ge,Belgium
ABSTRACT
3
1 Context.ζPup isthe X-ray brightest O-type star of the sky. This object was regularly observed with the RGS instrument aboard
0 XMM-Newtonforcalibrationpurposes,leadingtoanunprecedentedsetofhigh-qualityspectra.
2 Aims.Wehavepreviouslyreducedandextractedthisdatasetandcombineditintothemostdetailedhigh-resolutionX-rayspectrum
ofanyearly-typestarsofar.Herewepresenttheanalysisofthisspectrumaccountingforthepresenceofstructuresinthestellarwind.
n
Methods.Forthispurpose, weuseournew modelingtoolthatallowsfittingtheentirespectrumwithamulti-temperatureplasma.
a
J Weillustratetheimpactofapropertreatmentoftheradialdependence oftheX-rayopacityofthecoolwindonthebest-fitradial
distributionofthetemperatureoftheX-rayplasma.
2
Results.The best fit of the RGS spectrum of ζPup is obtained assuming no porosity. Four plasma components at temperatures
2
between0.10and0.69keVareneededtoadequatelyrepresent theobservedspectrum.Whilstthehardestemissionisconcentrated

between 3and4R ,thesofteremissionstartsalreadyat1.5R andextendstotheouterregionsofthewind.
R] Conclusio∼ns.Thein∗ferredradialdistributionoftheplasmatemp∗eraturesagreesratherwellwiththeoreticalexpectations.Themass-
lossrateandCNOabundancescorrespondingtoourbest-fitmodelalsoagreequitewellwiththeresultsofrecentstudiesofζPupin
S
theUVandopticaldomain.
.
h
Keywords.Stars:early-type–Stars:mass-loss–X-rays:stars–Stars:massive–Stars:individual:ζPup
p
-
o
r 1. Introduction Accountingforthe fragmentedstructureof the windsleads
t
s to a downward revision of the mass-loss rates by a factor be-
a Through their powerful stellar winds, and their violent deaths tween a few and ten. Owing to the differentsensitivities of the
[
as supernovae, massive stars have a tremendousimpact on the spectraovervariouswavelengthrangestothestructures,multi-

1 ecologyof theirgalaxies. The feedbackby the stellar winds of wavelengthanalysesarerequiredtodetermineconsistentmass-
v early-typestarsisindeedamajoringredientofthechemicalevo- lossratesofmassivestars.Forinstance,Hαemissionandfree-
0 lution of the interstellar medium. Moreover, the mass-loss has free radio emission are collisional processes which scale with
9 an important impact on the evolution of the star. The quantity thewind-densitysquaredandarethussensitivetotheclumping
0 of material released by the wind as well as its chemical com- factor regardless of the size of the clumps. In clumped winds
5
position dependon the evolutionarystage of the star. A proper withavoidinter-clumpmedium,thewind’svelocitylawmisses
.
1 understandingof the stellar winds of massive stars is therefore somevaluesalongagivensightline.Thisporosityintheveloc-
0 fundamentalforourknowledgeof the stellar evolutionand the ityfieldiscalled‘vorosity’(Owocki2011).Thelattermodifies
3 impacton the interstellar medium.However,the determination the Sobolev length and its impact is mainly visible on the UV
1 ofthemass-lossrateofastarisacomplicatedtask. resonance lines. In a fragmented wind with non-optically thin
:
v Overthelastdecade,theimportanceofthestructuresofthe clumps,X-raylinesaresensitivetoporosity,whichdependson
i stellarwindsandtheirimpactonthedeterminationofthemass- clumpcontinuumopticaldepth,andthusclumpsize as wellas
X
loss rates became more and more obvious. Observationally, theclumpingfactor.
r
a these structures manifest themselves through low-amplitude TheX-raylineprofilescanalsogiveinformationonthehy-
variabilityofemissionlinessuchasHα(Eversbergetal.1998), drodynamicalinstabilities in the wind. In fact, considering the
andthroughdifficultiestofittheobservedspectrawithmodelat- windembeddedshockscenario,theemission of X-rayphotons
mospherecodesthatassumehomogeneouswinds(Bouretetal. ariseswhentwofragmentswithdifferentvelocitiescollide(Lucy
2005).Thesesmall-scalestructuresareusuallyattributedtohy- &White1980).TheobservablemorphologyofanX-raylinede-
drodynamicinstabilitiesintrinsicto the winds(Feldmeieret al. pendsontheinterplayofemissionbythehotplasmaandphoto-
1997). electricabsorptionbythecoolwindcomponent.Thelattereffect
is obviouslyalso sensitive to the presence or absence of struc-
tures in the wind. On the other hand,the location of the X-ray
Sendoffprintrequeststo:A.Herve´
emission region as well as the temperature of the X-ray emit-
⋆ Based on observations collected with XMM-Newton, an ESA
ting plasma give importantinformationon the hydrodynamical
ScienceMissionwithinstrumentsandcontributionsdirectlyfundedby
propertiesofthewinds.
ESAMemberStatesandtheUSA(NASA).
⋆⋆ ResearchAssociateFRS-FNRS(Belgium) ζPup(=HD66811,O4Ief)hasbeenthetargetofnumerous
Correspondenceto:[email protected] studiesoverawiderangeofwavelengths.Thisstarisoftencon-
1
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
sideredasacornerstoneobjectfortheunderstandingofmassive 2. Observationaldata
stars.TheanalysisofthefarUV-UV-opticalspectrabyBouret
Naze´ et al. (2012) extracted and reduced 18 observations of
etal.(2012)confirmedasuper-solartotalCNOabundance,with
ζPupwithXMM-Newtonwhichyieldmorethan700ksofuse-
NbeinghighlyoverabundantandCbeingstronglydepleted.On
fulexposure.Forthe details ofthe data reduction,we refer the
theotherhand,Pauldrachetal.(2012)foundquitedifferentre-
reader to the paper of Naze´ et al. (2012, Paper I). Using this
sultsintheirfarUV-UVspectralmodeling,indicatinganover-
dataset, Naze´ et al. (2013, Paper II) performed an analysis of
abundance of both nitrogen and carbon and a strong depletion
thevariabilityoftheX-rayemission,findingnosignificanttime
ofoxygen.Also,theydetermineahomogeneousmass-lossrate
variabilityofthelineintensitiesandprofiles.Therefore,wecan
( 13 10 6M yr 1) aboutsix times largerthan the clumped
m∼ass-lo×ssra−tesd⊙ete−rminedbyBouretetal.(2012) andNajarro combinetheindividualRGS spectraintoasingle,high-quality,
fluxed spectrum. The calibration of the fluxed spectrum prop-
et al. (2011). This situation is embarrassing as one would aim
erlyaccountsforthedependenceoftheeffectiveareaasafunc-
for consistent parametersobtained from studies using different
tion of energy, but does not account for the redistribution of
wavebandsanddifferentmodelingtools.
monochromatic response into the dispersion channels. For in-
ζPup is also a bright X-ray source and has been observed trinsicallynarrowlines,thismightintroduceartifactsinspectral
both with Chandra and XMM-Newton at high spectral resolu- regionsof stronginstrumentalsensitivity changes.However,in
tion. The star was indeed regularly observed with the RGS in- the case of ζPup, the intrinsic width of the lines is large and
strument aboard XMM-Newton for calibration purposes, yield- thereforesucheffectsshouldbeofveryminorimpacthere.The
ing a combined data set of unprecedented quality (Naze´ et al. instrumental width of the RGS is accounted for in our fitting
2012). Previous studies of ζPup’s high-resolution X-ray spec- process,asthesyntheticspectraarebroadenedaccordingtothe
trum revealed new interesting results (Cassinelli et al. 2001, instrumentalprofile.
Kahn et al. 2001, Oskinova et al. 2006, Leutenegger et al. WhilsttheChandra-HETGspectrumofζPup(Cassinelliet
2006, 2007, Cohen et al. 2010). For instance, the analyses by al.2001)hasasuperiorspectralresolutionandextendstoshorter
Cassinelli et al. (2001) and Leutenegger et al. (2006) of the wavelengths than the RGS spectrum analysed in the present
forbiden,inter combination,resonance(fir)tripletsofHe-like work, it has a considerablylowersignal-to-noiseratio than the
−
ionsrevealedthattheinnerboundaryoftheX-rayemissionre- RGS spectrum. We therefore decided to focus our analysis on
gionisratherclosetothestar(atabout1.5R andpossiblyeven theRGSspectrumonly.
closer for the hottest plasma, Cassinelli et a∗l. 2001). However,
these studies were done on individual line profiles considered
separately.Eachspecific linewas insomesense assumedto be 3. X-rayspectralmodelingcode
formed by a plasma at the temperature of maximum emissiv-
3.1.Basicprinciplesofthecode
ity ofthe line.Thisapproachthusneglectsthe factthatseveral
plasmacomponents(eachonecharacterizedbyitsspecifictem- In order to analyse high-resolution X-ray spectra of massive
perature, emission measure and location inside the wind) may stars, we have developedourown dedicatedcode (Herve´ et al.
contributesignificantlytoagivenline.Tothebestofourknowl- 2012).Inourmodel,weassumethatthewindhostsseveralhot
edge,therehavebeennopreviousattemptstofittheentirehigh- plasma components. Each plasma componentextends between
resolutionX-rayspectrum(eitherRGSorHETG)ofanO-type an innerand an outer boundaryand has a specific kT which is
star atonce by properlycombiningthe contributionsfromsev- constantoverthatregion.Thechemicalcompositionistakento
eralhotplasmacomponentsandmodellingtheabsorptionbythe be the same for all components. We then discretize the X-ray
cool wind, accounting explicitly for the geometry of the prob- emissionregionofagivenplasmacomponentintoalargenum-
lem. berofcells. We computethe X-rayemissivityof eachcellas a
This paper thus explores, for the first time, the possibility function of its temperature and the quantity of material inside
the cell. For this purposewe use version2.0.1 of the AtomDB
to fitallthelinesofan X-rayspectrumbetween6 and35Å si-
database (Foster et al. 2012) with emissivities computed using
multaneouslyandcoherently,assumingadiscretedistributionof
the APEC code (Smith et al. 2001)1. This code calculates the
plasmacomponents,characterizedbytheirowntemperatureand
lineemissionarisingfrommanyphysicalprocessessuchascol-
location in the wind. To do so, we use the code developed by
lisionalexcitation,ionization,anddielectronicrecombinationin
Herve´etal.(2012)whichaccountsfortheintrinsicemissivityof
acollisionallyionized,opticallythinisothermalplasmainther-
thehotplasmaasafunctionoftemperatureandchemicalcom-
malequilibrium.Thecontinuumemissionisalsoincludedunder
position.Wefurtherintroduceamulti-temperatureplasmaspec-
theformofthebremsstrahlungemissionandthetwo-photonde-
tralfittingprocedure.Thegoalistodeterminethepropertiesof
cay.
theX-rayemissionregionaccountingforthemostrecentresults
We assume the plasma to be embedded in the wind and to
fromotherwavelengthdomains.Thesepropertiescanultimately
move along with the latter. The velocity of the wind is deter-
be comparedto the predictionsof existingand futurehydrody-
namicalsimulationsoftheinstabilitiesofstellarwinds. nmailnveedlobcyitayβo-flathweiw.ei.nvd(ra)n=dRv∞i(s1t−heRrr∗a)dβiuwshoefrethve∞stiasr.thFeotreeramcih-
This paper is organized as follows. In Sect2, we present a cell,weDoppler-shiftitscont∗ributionintotheobserver’sframe
shortsummaryoftheζPupRGSdataset.Thenwebrieflyrecall
of reference accountingfor the projectionof the wind velocity
themainassumptionsinourmodellingtool(Herve´ etal.2012)
onthelineofsight(MacFarlaneetal.1991).
concerning the intrinsic emissivity and the absorption formal-
Since the X-ray emittingplasma is assumed to be optically
ism (Sect.3). We also present, in Sect.4, the differentphysical
thin, an X-ray photon, once emitted, either escapes or is ab-
processeswe haveincludedin the code to betterreproducethe
sorbedbythecoolwindmaterialalongthelineofsight.Hence,
observationalspectrum of ζPup. Finally, in Sect.5, we discuss
the results that we obtain, especially as far as the composition 1 Theinterestedreaderisinvitedtovisitthesection’physics’ fora
andthelocationofthedifferentplasmacomponentsinthewind precise description of the calculation of the continuum and lines in a
areconcerned. hotplasmaontheAtomDBwebsiteattheurlhttp://www.atomdb.org/.
2
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
the radiative transfer problem reduces to the calculation of the Table 1. Chemical elements and their ionization levels used to
opticaldepthofthefragmentedcoolwindmaterial,τ,alongthe calculatethemassabsorptioncoefficientκandtheUVradiation
lineofsightfromtheemittingcelltotheobserver.Forthis pur- fieldinCMFGEN.
pose,weconsiderthat,atthemicroscopiclevel,thecrosssection
forphotoelectricabsorptionintheX-raydomainisdominatedby H I
thebound-freetransitionsfromtheKandLenergylevels. He I,II
Iftheclumpsareopticallythick,theeffectiveopacityisde- C II,III,IV
N II,III,IV,V
terminedbytheirgeometricalproperties(Feldmeieretal.2003).
O II,III,IV
Theeffectiveopacitycanbewritten(Owocki&Cohen2006):
Mg II,III,IV
Si III,IV
l2 κ
κ = = Ar III,IV,V
eff mc τc Fe III,IV,V,VI,VII
Ni III,IV,V
wherelisthescale-lengthoftheclump,m isitsmass,κ isthe
c P IV,V
massabsorptioncoefficientandτc theopticaldepthofaclump. Al III
Thislattercanbewritten: Na I
l κM˙ 1 l
τ =κ ρ = (1)
c h if 4πr2 v f
where ρ isthemeandensityofthestellarwindatpositionr, f 10 A 20 A
thevolhumiefillingfactorofthe clumps, M˙ isthemass-lossrate
ofthestar,andvisthevelocityofthewindattheradiusrwhich
isdeterminedbyaβ-law(seeabove). 2 2
Underthisassumption,itappearsthattheeffectivereduction
oftheopticaldepthasaresultofthefragmentationdependson
the ratio betweenthe clumpscale andfilling factor,also called
theporositylength(Owocki&Cohen2006).Thisresultcanbe 26.5 26.5
generalized from the optically thick to the optically thin limit
using a simplified scaled effective opacity (Owocki & Cohen 30 A 40 A
2006):
κ 1 2
eff 2
= (2)
κ 1+τ
c
Using this result in a steady state (i.e. non variable) wind, we
obtain the wind opticaldepth (Owocki& Cohen 2006) in con-
.5 26.5
ventional(p,z)coordinates:
Fig.1. Radial dependenceof the mass absorptioncoefficientat
R dz
τλ(p,z)=τ∗Zz∞ r′(r′−R∗)∗+τ′∗R∗h(r′) (3) dpiaffneerle).nTthweavmealesnsgatbhsso(rspptieocnificeodeffiinctiheentuspapreerclaelfctucloartnederwoiftheatchhe
elementslistedinTab.1withabundancesfoundinourstudy.
where r = p2+z2, τ κM˙ is the characteristic optical
depth of thepwind, and ∗h(≡r) 4πv∞lR∗is the porosity length. In the
≡ f
followingweadopth(r)=h r.
′
×
Aslightlydifferentformalismtodescribethefragmentation
of the wind is based on the concept of the fragmentation fre-
of temperature and mass-loss rate. In these ionization calcula-
quency (Oskinova et al. 2006). As we have shown in Herve´ et
tions, CMFGEN accountsfordifferentphysicalprocessessuch
al.(2012),bothformalismsyieldverysimilarX-raylineprofiles
ascollisionalexcitation,photo-ionizationbyphotosphericlight
andthecapabilitiesofthecurrentgenerationofhigh-resolution
andX-rays,aswellasradiativeanddielectronicrecombination.
X-ray spectrographs aboard XMM-Newton or Chandra do not
Onceweknowtheionizationstructure,wecanthencomputeκ.
allowustochoosebetweenthetwoformalisms.Hereweadopt
theporositylengthformalismbecauseitleadstoananalyticex- The results are illustrated in Fig.1 at four different wave-
lengths. This plot shows that the approximation of a mass ab-
pressionfortheintegralintheequationoftheopticaldepth.This
hastheadvantagetoallowafastercomputation.InHerve´ etal. sorption coefficient that is independent of the position in the
(2012),wehaveshownthattheinclusionofafewpercentofthe windisnotavalidoneinthiscase.Indeed,theionizationstruc-
tureof the windis notconstantthroughoutthe entirewind and
windmaterialina homogeneousinter-clumpmediumhaslittle
impactontheemergingX-rayspectrum.Inthepresentwork,we the impact on the value of the mass absorption coefficient is
thusconsiderthattheinter-clumpmediumisvoid. huge. The outer regions of the wind are less ionized and con-
sequently, the cool material produces a stronger absorption in
Ineitherformalism,thecharacteristicwindopticaldepthτ ,
andthusthemassabsorptioncoefficientκ,playakeyroleinth∗e theouterregions.
computationoftheabsorptionofX-raysbythestellarwind.To Comparedtoourpreviousversionofthe code(Herve´ etal.
estimate the values of κ, we use the non-LTE, line blanketed, 2012) where we adopted a value of κ that was independent of
model atmosphere code CMFGEN (Hillier & Miller 1998) to r, we havedecidedto renderthe treatmentof theabsorptionof
computethe ionizationstructureof thecoolwindas a function the wind more realistic. For this purpose, we approximate the
3
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
1.5 R* 3. R* Forinstance,theratiobetweenthefluxofNviiLyαat24.78Å
1-^m 1-^ ainndlinthee-bflyu-lxinoefatnhaelyNsevsiatrsipaletetsmapte2ra8t.u8re–i2n9d.i5caÅto,rw,choicuhldisbeusaefd-
g 2^m mg 2^ fectedbythestrongNivedge.
κc ni κmc ni
3.2.ThetreatmentofthefirtripletsofHe-likeions
Wavelength in Angstrom Wavelength in Angstrom In an opticallythin thermalplasma at hightemperatures,some
6. R* 15 R* elementsretain only two electronsand are thus presentas ions
1-^ 1-^m wiointhsparHodeu-lciekeaetlreipclterotnofcolinnfiegsucroantisoisnti(neg.go.fOsvoi-icaanlldedNfvoir)b.iSdudcehn
mg 2^ g 2^m 23S1 - 21S0 (f), inter-combination 23P1,2 - 21S0 (i) and res-
κmc ni κc ni odnenansictye 2en1vPi1ro-nm21eSnt0s(ro)rliinnetshe(Gaabbserinecle&ofJostrrdoanng1U9V69)r.adIniatlioown
fields,theupperleveloftheftransitionisonlydepopulatedvia
Wavelength in Angstrom Wavelength in Angstrom the forbiddenline which is then seen as a rather strong line in
Fig.2. Wavelength dependence of the mass absorption coeffi- thespectra(Blumenthaletal.1972).However,inthewindofa
cientatdifferentpositionsinsidethewind(1.5,3,6and15R as massivestar,thepresenceofanintenseUVradiationfieldalters
specifiedintheupperleftcornerofeachpanel). ∗ thissituation.Indeed,theUVphotonspumptheelectronsfrom
theupperleveloftheftransitiontotheupperlevelofthe iline
(Blumenthaletal.1972,Porquetetal.2001).Thereforetheratio,
= f isstronglymodifiedwhichprovidesaninterestingtoolto
radial dependence of κ by two simple relations that hold over cRonstirainthedilutionoftheUVradiationfieldatthepositionof
tworangesofvaluesofr.Alinearrelation the X-ray emission and hence to determine the location of the
hotplasmainthewind.
κ(r)=κ +γ(r R ) (4)
0 − ∗ SincethefluxofUVradiationseenbytheionsinthestellar
wind decreases in proportionto the geometricaldilution factor
forr <R ,andaconstant
lim
κ(r)=κmax (5) wra(dri)us=.T12h(e1ra−di[a1l−mo(dRri∗fi)2c]a21ti)o,nthdeueRtoratthioe ipshoaltsooexaciftuatnioctniofnroomf
themetastableupperleveloftheforbiddenlinecanbeexpressed
forr R .Hereκ andκ are,respectively,themassabsorp-
lim 0 max asfollows(Blumenthaletal.1972,Leuteneggeretal.2006):
≥
tioncoefficientintheinnermostpartandintheouterpartofthe
windandγistheslope.R isthepositioninthewindwhereκ 1
lim (r)= (7)
reachesitsplateauvalueκmax (seeFig.1).Theadvantageofthis R R01+2Pw(r)+Ne/Nc
parametrizationistoprovideamorerealisticdescription ofthe
variationof κ asa functionofthe radiusand,atthesame time, where P = φ and φ = 8ππe2 fHν with H the Eddingtonflux
to preservean analyticalsolution forthe integralof the optical φ∗c ∗ mev hν ν
and f isthesumoftheoscillatorstrengthsfor23S toallthree
depthequation.Indeed,foralinearlyincreasingporositylength 1
of23P .N isthefreeelectrondensity.φ andN arethecritical
(h(r)=h r),theopticaldepthisnowgivenby: J e c c
′× photoexcitation rate and the critical density (which depend on
atomicparametersandtheelectrontemperature,seeBlumenthal
τλ(p,z) = Z √(Rlim)2−p2r (r R )α+(κα0(−κγ+Rγ∗)(Rr∗dzR′ ))R h r etoteaxl.c1it9a7ti2o)n.Irnattehedowminidnsatoefshoovte,rlutmheincooullsismioansasilveexsctiatrast,iothnetperhmo-,
z ′ ′− ∗ 0 ′− ∗ ∗ ′ ′ and in the remainder of this work, we have thus neglected the
+ √(Rlim)2−p2 αγR∗dz′ Ne/Nc term.
Zz (r′−R∗)+α(κ0+γ(r′−R∗))R∗h′ The treatment of these effects is now included in our code
ακ R dz along the above prescriptions and using the Eddington fluxes
+ ∞ max ∗ ′ (6) computedwithourCMFGENmodel.InTable2,weindicatethe
Z√(Rlim)2−p2r′(r′−R∗)+ακmaxR∗h′r′ physicalinformationnecessarytoreproducethismodificationof
withα= M˙ . intensitiesoftheforbiddenandinter-combinationlinepresentin
4πRv thespectraofζPup.
Notethat∗w∞ehaveassumedinEq.6thatzisincludedinthe
interval[ R2 p2, R2 p2].Ifthelatterconditionisnot
−q lim− q lim− 4. Multi-temperatureplasmafittingprocedure
fulfilled, the boundariesof these integrals are adjusted, but the
principleremainsthesame. ThefullX-rayspectrumofamassivestarcanusuallynotbefit-
The first two integralshaveanalyticalsolutionsmakingthe tedwithasingletemperatureplasmamodel.Inthepresentcase
computation easier and the same is true for the last integral forinstance,themostprominentionizationlevelofiron(Fexvii)
which is formally identical to the one used in the original pre- seen in the RGS spectrum of ζPup, suggests the presence of a
scriptionofOwocki&Cohen(2006,seeourEq.3). hotplasmacomponentwithakT closeto0.50keV.Otherlines,
Figure2 illustrates the wavelength dependence of the mass suchastheSixiiilinesaround6.7Å,havetheirmaximumemis-
absorption coefficient at four different positions in the stellar sivityatevenhighertemperaturesaround0.90keV.Ontheother
wind.Thestrongestabsorptionedgeinthisfigure(atabout26Å) hand, the fir triplets of He-like carbon and nitrogen are rather
isdueto Niv.Thecomplexshapeofthisrelationcouldimpact indicativeofalowertemperatureplasma(kT around0.10keV).
someofthefrequentlyuseddiagnostictoolsbasedonlineratios. Fitting the observed spectrum hence requires the combination
4
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
Table2.Parametersforthetreatmentoftheforbidden,inter-combinationandresonancelines.
Ion λ λ λ λ λ φ H /(hν) H /(hν) f
r i f uv(i1) uv(i2) c ν1 ν2
Nvi 28.792 29.074,29.076 29.531 1907.26 1896.74 1.83 102 1.71 2.27 0.1136
Ovii 21.603 21.796,21.799 22.095 1638.28 1623.61 7.32×102 1.65 1.30 0.0975
Neix 13.447 13.548,13.551 13.697 1272.81 1248.28 7.73×103 1.95 1.87 0.0700
Mgxi 9.1681 9.2267,9.2298 9.3134 1034.31 997.46 4.86×104 1.19 1.10 0.0647
Sixiii 6.6471 6.6838,6.6869 6.7394 865.14 814.69 2.39×105 0.64 0.86 0.0562
×
Notes.λ andλ correspondtotheUVradiationwavelengthswhichdepopulatetheinter-combinationlevelofagivenelement.Allwave-
uv(i1) uv(i2)
lengthsaregiveninÅ.
φ aretakenfromBlumenthaletal.(1972).
c
H /(hν) are calculated withCMFGEN withour own abundances and mass-loss ratebut with T determined in the UV/optical domain. The
ν eff
valuesaregivenin108photoncm 2s 1Hz 1.
− − −
Theoscillatorstrengths, f,areextractedfromtheAtomDBwebsite.
ofseveralmodelswithdifferenttemperatures.Atthisstage,we Thefittingroutinethensystematicallyexploresthisdatabase,
have to stress that the HETG spectrum of ζPup (Cassinelli et combiningonlyplasmacomponentsthathavethesamechemical
al. 2001) as well as its EPIC spectra (Naze´ et al. 2012) reveal composition, h and M˙. The synthetic multi-temperature spec-
′
thepresenceofaweakSixivLyαlineat6.18Åandofaweak trumisbuiltasfollows:
Sxvtripletaround5.04–5.10Å.Theseionshavetheirstrongest n
emissionattemperaturesnear1.3–1.4keV.However,theselines S = f M +ǫ (8)
i hotgas,j j,i
areoutsidethesensitivityrangeoftheRGSinstrumentandare X
j=1
thusnotincludedinourfit.WewillreturntothispointinSect.5.
whereS istheobservedfluxatthewavelengthi, f isthe
Of course, each line is not only emitted at the temperature i hotgas,j
classicalvolumefillingfactorofthehotgascomponent j,M is
whereitsemissivityreachesitsmaximumvalue,butratherexists j,i
the synthetic flux of each X-ray emitting plasma componentat
oversome range of temperatures, although with a lower inten-
thewavelengthi,andǫ istheerrorofregression.
sity. Thereforemostlinesin the observedspectrumresultfrom
Foreachcombinationofplasmamodels,thebestfitvaluesof
thecombinationofseveralcontributionsassociatedwithplasma
f aredeterminedintheleast-squaressenseandtheresult-
componentsatdifferenttemperatures(seeFig.4).Themaindif- hotgas,j
ingχ2 isobtainedcomparingthemodeltothefluxedspectrum.
ficulty in the fitting procedure therefore consists in finding the ν
Thesenumbersarestoredforlateruse,includingtheidentifica-
bestvaluesoftheparametersthatdescribeoursyntheticspectra,
tion of the best-fit model based on the value of χ2. After sev-
includingthefillingfactor f (i.etheratiooftheX-rayemit- ν
hotgas eraltests,itappearsthatatleastfourcomponentsareneededto
tingvolumeandthetotalvolume)foreachplasmacomponent.
achieveagoodfitofthefluxedRGSspectrumofζPup.
We have thus developed a code which combines different
We have included an additional physical constraint to this
componentsat differenttemperaturesin order to reproducethe
fittingprocedure:weimposedtheglobalX-rayemittingregion
total spectrum of ζPup by a minimization of the χ2 of the fit.
aroundthestartobecontinuous.Inotherwords,fromthelowest
We have based ourfitting procedureon a multiregressionrou-
innerboundaryradiusoftheemittingregionstothelargestouter
tine2.Therearequiteanumberofparametersneededtodescribe
boundaryradius,theremustnotexistanyspatialgapwherethere
amodel.Thesearethegeneralwindandstellarparameters(M˙,
isnohotplasmaemissionatall.
v ,β,R andT ),theporosityparameterh′,theabundancesof Itisdifficulttomakeanystatementabouttheunicityofthe
th∞ediffe∗rentelem∗ entsthatmakeupthehotplasma,andforeach
solution.Theremightindeedexistothercombinationsof(more
plasma component,its temperature(kT), filling factor (f ),
hotgas than four) plasma componentswith values of the wind param-
andinnerandouterboundary(RinandRout).Notallofthemare eters (M˙, h,...) outside the range explored here that could fit
′
treated as free parameters in the present study: only the CNO
thespectrum.However,withintheassumptionsmadehere(four
abundancesare varied with respect to solar; v , R and T are
discrete plasma components,...),we are confidentthat our pro-
all taken from Bouret et al. (2012), whilst we∞ado∗pt β =∗1 to
cedurehasidentifiedthebestsolutions.
preserve the analytical solution of the optical depth integrals.
Nevertheless,weareleftwith4 n+5parameterswherenisthe
×
numberofplasmacomponents.In ourcase,we find that n = 4 5. Resultsanddiscussion
isrequiredtofittheobservedspectrumandwearethusdealing
To start our analysis, we used the stellar parameters of Bouret
with21freeparameters.Inordertoobtainthebestsolution we
etal.(2012).Thenweexplorealargerangeofvaluesforallthe
usedoursyntheticspectrasimulatorcode(Herve´ etal.2012)to
freeparametersinordertominimizetheχ2ofthefit.Ourbest-fit
build a database of individual plasma components (hence nine
modelofthefullRGSspectrumofζPupisillustratedinFig.3.
freeparametersperelementofthe database).Indoingso,each
parameter is sampled over a given range that we consider rea-
sonable(e.g.kT variesbetween0.08and0.80keVandM˙ varies 5.1.Temperaturedistributionandhotgasfillingfactor
intherange1.75 10 6to4.0 10 6M yr 1).
− − −
× × ⊙ The true temperaturedistribution of the X-ray emitting plasma
in the wind of ζPup is almost certainly a continuum of tem-
2 We use the IDL routine IMSL Multiregress. peratures between a minimum and a maximum temperature.
More information are available at Nevertheless, a discretization with four different temperatures
http://idlastro.gsfc.nasa.gov/idl html help/IMSL Multiregress.html is sufficientto reproducecorrectlythe observedRGS spectrum
5
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
Table3.Best-fitX-rayemittingplasmaparametersfoundinourstudy.
model h’=0.0 χ2=15.16 model h’=0.02 χ2=17.82
1 ν 2 ν
kT f R R f R R
hotgas in out hotgas in out
keV R R R R
0.10 0.012 7.5∗ 85∗. 0.013 2.∗8 95∗.
0.20 0.012 1.5 38. 0.012 1.6 40.
0.40 0.020 2.7 4.0 0.020 3.0 4.2
0.69 0.007 3.1 4.1 0.007 3.2 4.9
Notes. χ2isthereducedchi-squaredwith1222degreesoffreedom.
0.6
0.4
0.2
0
0.1
0
-0.1
10 15 20
0.6
0.4
0.2
0
0.1
0
-0.1
20 25 30 35
Fig.3. Ourbest-fitmodel(model ofTable3,indashedredline)oftheobservedRGSspectrumofζPup(insolidblackline).The
1
mostprominentemissionlinesarelabelled.Wenotethattherearemanymoreweakerlinesthatcontributetothespectrum andare
includedinourmodel.Theresiduals(inthesenseobservationminusmodel)areshowninthepanelsbelowthespectrum.
of ζPup. As pointed out above, these four components might triplet containsbetween 80 and 100 countsin the HETG spec-
not be sufficient to account for the presence of the highly ion- trum,asjudgedfromthe Chandratransmissiongratingcatalog
ized Sixiv and Sxv lines seen in the HETG and EPIC spec- (Huenemoerderetal.2011)3.Therefore,ifahotterplasmacom-
tra.However,althoughtheRGSsensitivityrangedoesnotcover ponentis neededto fully accountforthe flux of this triplet, its
the Sxv triplet, our model predicts an integrated line flux of emissionmeasureislikelyverylow.
0.64 10−13ergcm−2s−1, which is not too far away from the In this respect, it is interesting also to compare our results
value×0.81 10−13ergcm−2s−1 reported by Cassinelli et al. withthefitsoftheEPICspectrareportedinPaperI(Naze´ etal.
×
(2001), especially in view of the relative uncertainties on this
flux. The latter uncertainties are likely of order 10%, as this
3 seehttp://tgcat.mit.edu
6
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
Fig.4. Contributionsof the fourplasma componentsto the best-fit modelof the RGS spectrum of ζPup (model , in solid black
1
line):kT =0.10keV(indottedredline),kT =0.20keV(indashedgreenline),kT =0.40keV(indotteddashedblueline)and
1 2 3
kT =0.69keV(inlongdasheddottedpinkline).
4
2012).Inthatpaper,weusedafour-temperaturethermalplasma Thisisconsistentwiththefactthat,inthisspecificcase,thetwo
modelwith an overlying‘slab’ of absorbingmaterial. The best dominantcomponentshavetheirrelativeimportancerevertedin
fitplasmatemperatureswerefoundto be kT =0.09,0.27, thelinesoftheH-likeandHe-likeions.Insummary,whilstwe
1,2,3,4
0.56 and 2.18keV. Whilst the first three temperatures are very confirmthatlineratiosyieldsomeaverageestimateofthetem-
similarto-ormeanvaluesof-whatwefindhere,thefourthone perature,it is impossible to know,froma line-by-lineanalysis,
isclearlymuchhigherthanfoundinouranalysis.However,the whataretheweightsofthedifferentplasmacomponentsinthis
emissionmeasureofthe2.18keVcomponentinthefitsofPaper average.
I, is about a factor 100 below that of the 0.56keV component. The two plasma components with the lowest temperatures
Therefore,thisveryhotplasmashouldindeedhavea verylim- (kT of0.10keVand0.20keV)extendoverthewidestrangein
ited impact on the RGS spectrum analysed here. Nevertheless, radiusanddominatetheemission(model inTable3andFig.4).
1
this remark implies that there exists very hot material in the These emittingregionsextendfroma regionrelativelyclose to
windofζPup.Itwouldthusbeextremelyinterestingtocollect thesurfaceofthestar(7.5and1.5R forkT of0.10keVand0.20
aHETGspectrumofqualitycomparabletotheRGSspectrum, keV)toveryfaroutinthewind(85∗and38R ,respectively).The
andtorepeatthekindofanalysispresentedhere,toconstrainthe two hotterplasma components(kT of 0.40∗keV and 0.69 keV)
locationofthisveryhotplasma. startat2.7and3.1R andextendovera smallerrangeofradii,
In previous analyses, the line intensity ratios of hydrogen- outto4.0and4.1R ,∗respectively.
likeandhelium-likeionswasusedasadiagnosticoftheX-ray Withtheseresul∗ts,itappearsthattheregionbetween3.0and
source temperature (e.g. Waldron & Cassinelli 2007). Such ra- 4.0R containsX-ray emitting plasma with 3 differenttemper-
tiosareexpectedtoprovidesomeaveragetemperaturebetween atures∗. A priori,in an outwardsacceleratingwind, one expects
the various plasma components that contribute to the line for- thevelocityjumpsduetoinstabilitiestohavealoweramplitude
mation.However,Fig.4indicatesthattherelativeweightsofthe intheinnerpartsofthewind.Therefore,oneexpectsthattheen-
differentplasmacomponentsinthisaveragedependupontheel- ergyavailableintheshocksisonlysufficienttoheattheshocked
ementthatisconsidered.Forinstance,inthecaseofoxygen,the matter to rather low temperatures. Farther out in the wind, the
OviiiLyαlineisdominatedbythe0.20keVplasma,whilstthe differenceofvelocitybetweentwoconsecutiveclumpsofmatter
Ovii triplet includes contributions from the 0.10 and 0.20keV increases and the energy available in the shocks increases ac-
plasmacomponents.Waldron&Cassinelli(2007)derivedkT = cordingly. In the outermost parts of the wind, the density and
0.24keVfromtheoxygenlineratio,whichisconsistentwiththe hencetheemissionmeasuresbecomelowandthesameapplies
dominantroleofthe0.20keVplasmainthiscase.Ifweconsider to the velocity jumps since the efficiencyof radiative accelera-
neon instead, we see that the Nex and Neix lines are mostly tionprogressivelydropstozero.
formedinthe0.40and0.20keVcomponents,withminorcontri- From this simple reasoning, one would thus expect the
butionsfromthe 0.69keV component.Forthese ions,Waldron hottest plasma components to arise from intermediate radii in
&Cassinelli(2007)obtainedkT =0.33keVfromthelineratio. the wind, whilst the softer emission would be producedover a
7
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
widerrangerofradiistartingclosertothestellarsurface andex-
tendingfartherawayfromit. 1
s-
Withthecaveatthatwecurrentlyignorewherethe2.18keV m.
k
plasma, responsible for the high-energyextension of the EPIC V in
andHETGspectra,islocated,ourresultsareinqualitativeagree-
mentwiththeseexpectationsaswellaswiththe1-Dlinedriven
instability simulation of Feldmeieret al. (1997). These authors R in R*
show indeed that shocks can arise very close to the stellar sur-
Fig.5.Snapshotoftheradialvelocityinthestellarwindasob-
face (see Fig.5, A. Feldmeier, private communication). In our
tained in 1-D Line Driven Instabilities calculations (Feldmeier
bestfitmodel,thetemperatureofthehottestplasmacomponent
2012,privatecommunication).
that is present at a given radial position in the wind, increases
outwards from the innermost boundary of the emission region
(1.5R wherekT reaches0.20keV)toavalueofkT =0.69keV
ina re∗gionbetween3and4R .Furtherout,thetemperatureof 5.2.Abundances
the hottest plasma component∗decreases again. The maximum
Ourbest-fitmodeloftheRGSspectrumofζPupyieldsanover-
shocktemperaturehenceoccursinthe3–4R rangewherethe
velocity jumps predicted by the simulations r∗each indeed their abundance of nitrogen along with a depletion of oxygen and
carbon (see Table4). Our CNO abundances agree qualitatively
highest values. At positions closer to the stellar surface or far-
withthoseoftheUV/opticalstudiesofBouretetal.(2012),but
theroutinthewind,thesimulationspredictvelocitydifferences
are at odds with the result of Pauldrach et al. (2012), at least
betweentwoconsecutiveclumpsthataresmaller,in agreement
as far as the carbon abundance is concerned. The Lyα doublet
withthelowertemperaturesfoundintheseregions.
of Cvi at 33.73Å is the only isolated feature of carbon in the
ThelowvalueofthefillingfactorofthekT =0.69keVcom-
RGSwavelengthdomain.Nevertheless,itappearsimpossibleto
ponentindicatesthatonlyafractionoftheshocksintheregion
reproducethis line with an over-abundanceof carbon as advo-
between 3 and 4R actually heat the plasma to high tempera-
catedbyPauldrachetal.(2012).
tures.Toensuretha∗tmostofthekineticenergyofafastmoving
Theotherelements,thatarerelevantfortheX-rayspectrum,
clumpwhichhitsaslowerclump,movingaheadofit,isindeed
are assumed to have solar abundances (Andres & Grevesse,
convertedintoheat,thecollisionmustoccuralongtheradialdi-
1989). We cannot determine the abundances of hydrogen and
rection.Fromastatisticalpointofview,ina3-Dwind,theprob-
heliumdirectlywiththeRGSspectrumsincetherearenolines
abilitythatthevelocityvectorsoftwocollidingclumpsarenot
of these elements in the X-ray domain. However, they play a
collinear is most probably larger than the probability that they keyroleinthedeterminationofthemassabsorptioncoefficient.
arecollinear.Thusonlyafractionofthecollisionswillproduce
Therefore,bydefault,weusetheresultsofBouretetal.(2012).
therequiredheatingtoreachthehottesttemperature.
The particular properties of ζPup can explain the strongly
The overlapbetweenthehighesttemperatureandthelower non-solarCNO abundances.Indeed,ζPup is a rapidly rotating
temperature is somewhat inherentto our methodology.Indeed, runaway star and its kinematic properties and chemical abun-
whilstweworkedwithfourdifferenttemperaturesfortheX-ray dances could indicate that the star was part of a binary system
emitting plasma, we consider that the emission region of each that underwent either an episode of mass transfer and super-
plasmacomponentiscontinuous(i.e.therearenogaps)between novaexplosionofthecompanionoramergerevent(Vanbeveren
its inner and outer boundary. For instance, our results indicate 2011).
thepresenceofakT=0.10keVcomponentrelativelyclosetothe
star to fit the intensity of the lines at wavelengths above 20Å,
5.3.Mass-lossrateandfragmentation
butalso in the outerpartsof the windto fitthe broadwingsof
the line profiles. In ourcode we can only assume a continuous We find a slightly higher mass-loss rate ((3.5 0.25)
shellextendingallthewayfrom7.5to85R .Theintroduction 10 6M yr 1)thantheUV/opticalandIRstudies((2 ±0.2) 10×6
− − −
ofadiscontinuousregionwouldconsiderably∗increasethenum- and2.1⊙10 6M yr 1respectively,Bouretetal.2012±,Naja×rroet
− −
ber of free parameters in the regression routine. However, the al. 2011×) but the⊙difference is notimportant. Moreover,Bouret
scenarioofanextended,continuousshellforthecoolestplasma et al. (2012) indicate that their mass-loss rate is probably too
is supportedbytwo considerations.First, as stated above, non- low to be consistent with the line driving of the wind. Our de-
collinearcollisionsproducecoolerplasma,throughoutthewhole terminationofthemass-lossrateisalsoinreasonableagreement
wind. Second, in a wind embeddedplasma, the hottest plasma withthevalueof2.5 10 6M yr 1 obtainedindependentlyby
− −
cools down progressively as it moves outwards with the wind. Lamers & Leitherer (×1993) fro⊙m the radio fluxes of ζPup and
Forinstance, adoptingthe parametrizationof the coolingfunc- by Oskinova et al. (2007) froma fit of the Hα line and the Pv
tionusedbyFeldmeieretal.(1997),weestimateatypicalcool- resonancedoublet.
ingtimeof 180sforthe0.40keVplasmaattheinnerbound- Our results indicate that the wind structure of ζPup is un-
∼
aryof thatcomponent.In the same way,we estimate a cooling likely to be very porous. Indeed, we have tested various as-
timeof 600sforthe0.69keVcomponent.Ontheotherhand, sumptions on the porosity parameter: h = 0 (no porosity),
′
∼
the flow times required for the hot gas to move from the inner h = 0.07 (roughly equivalent to the fragmentation frequency
′
boundaryofitsproductionsite(near2.7and3.1R forthe0.40 n = 1.7 10 4s 1 derived by Oskinova et al. (2006) from a
0 − −
and 0.69keV plasma respectively) to a location at∗5R are es- line-by-lin×einvestigationof the HETG spectrum)and an inter-
timated at about 16500s and 13500s, respectivelyfor∗the 0.40 mediatevalueofh =0.02(model inTable3).Byfarthebestfit
′ 2
and0.69keVgas.HencetheX-rayphotonsemittedatenergies qualityisachievedforthemodelthathasnoporosity(seeTable
ofkT=0.10keVand0.20keVbeyondabout5 R existthrough 3).AsimilarconclusionwasreachedbyCohenetal.(2010)who
a combinationof direct emission from weaker sh∗ocks, and the foundthatporositywasnotrequiredtofitindividuallinesofthe
coolingofthehotplasmainitiallyheatedto0.40and0.69keV. HETGspectrum.
8
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
Table4.Stellarwindparameters
T v β R M˙ X(C) X(N) X(O)
kK km∞s 1 R∗ 10 6M yr 1
− − −
Thiswork 401 2300 1 17⊙.3 3.5⊙ 6.00 10 4 7.7 10 3 3.05 10 3
− − −
× × ×
Bouretetal.2012 40 2300 0.9 17.3 2.00 2.86 10 4 1.05 10 2 1.30 10 3
− − −
× × ×
Pauldrachetal2012 39 2100 - 19 13.7 8.2 10 3 9.13 10 3 1.14 10 3
− −
× × × −
Notes.1Wecannotdeterminetheeffectivetemperatureofthestarwithourcode.Asthisparameterisimportantforthemassabsorptioncoefficient
calculation,weindicatethevaluewehaveadoptedinouranalysis.TheCNOabundancesaregivenasmassfractions.NotethatPauldrachetal.
(2012)donotassumeaβ-lawforthewindvelocity,butperformafullhydrodynamicalcomputationofthewind.
Fromthestellarandwindparametersassumed(R ,v ,β)or NVI resonance line NVI intercombination line NVI forbidden line
derived (M˙) in this paper, we can estimate what wo∗uld∞be the
threshold value of the porosity parameter h for clumps to be-
′
comeopticallythick.FromFig.1,wefindthatκ ∼ 170cm2g−1 ^2 A)
atawavelengthof20Åandat4R .Withthevolumefillingfac- m
bhtwoe′er<nofiefn0edtd.h0eet2dh,cataotosoaalscucglhgauigsmeev(speftesτd=iczbe=0yo.0of15ul)art=dfie2tr∗s00i,.vÅ0iem5d(Rpabl∗niy,edsiB.aetoth.ua4hrt′eRτt∗=c)e.t=0Aa.2l1.v5(aa2wtlu02oe10u2olÅ)df, mission in photon /(s c
wwohuerldetohnelypobreosaitcyhiheavsedincinretahseedousitgenrmifiocsatnrtelygi(oansssuomfitnhge,awsinwde, Observed e
havedonehere,h(r)=h r).Therefore,inviewofourresults,
′
×
it seems that porosity does not play a role in the formation of
the X-ray spectrum of ζPup. This situation contrasts with the
results of Oskinova et al. (2012) for the WN5 Wolf-Rayet star Z/R*
WR6. The latter authors concluded that the wind of the WN5
Fig.6.Contributionto the observableemission forthe fircom-
star had to have a very porous structure to allow the observed ponentsof the Nvi He-like triplet. The productof the intrinsic
X-rayemissiontoescape.
emission per solid angle with exp( τ) is shown as a function
−
ofz(in(p,z)coordinateswith p = 0.5R ).Thetripletincludes
contributions from the 0.10 and 0.20keV∗ plasma components
5.4.Thefirtripletformation
startingrespectivelyatr= p2+z2 =7.5R andr=1.5R .
In previous studies (Waldron et al. 2001, Leutenegger et al. p ∗ ∗
2006,2007,Oskinovaetal.2006),theanalysisofthefirtriplets
of the different He-like ions was done line by line in order to
determine the hot plasma temperature and the dilution of the true in the innermost regions of the wind, where the 0.20keV
UV radiation field in the X-ray emission zone. For a single plasmacomponentisthesole contributortothe Nvitriplet.As
temperature plasma, this analysis constrains, in principle, the a result, = f+i i andthe ratiobetweenthe strengthofthe
positionoftheX-rayemittingplasmainthewind. inter-comGbinatiron≃linreandtheresonancelinetendstowardsthe
valueof predictedbytheatomicdata.
G
Inouranalysis,weclearlyshowthatmostspectrallines,in- Wefurthernotethepresenceofnumerousweaklinesinthe
cludingtheHe-likefirtriplets,aretheresultofthecombination AtomDB emissivities. Due to the Doppler broadening of the
of emission from several componentsat different temperatures lines and to the limited spectral resolution of the RGS instru-
(Fig.4). Thisimpliesthat the analysisof the fir triplets, assum- ment,theseweaklinescannotbeseenindividuallybuttheycon-
ingtheyareformedbyasingletemperatureplasmaandnottak- tributetothetotalemission.Thisisparticularlyimportantinthe
ing into account the abundances of the elements, could be bi- inter-combinationlineofNvi.
ased.LetusillustratethispointbytheOviiandNvitriplets.In Ascan beseen on Figs.3 and7, ourmodelhassomeof its
Fig.7, onecan see thatthe line profilesandintensitiesof these strongestdeficienciesinthe firtriplets. Forinstance,the model
triplets are reasonably well fitted. However, Fig.4 shows that predicts too large a flux in the red wind of the blends of the
bothtripletsareactuallyformedbycontributionsofthetwolow- MgxiandSixiiitriplets(Fig.7),whichlikelyindicatesanover-
esttemperatureplasmaswhichproducequitedifferentlinepro- estimate of the inter-combination(and possibly forbidden)line
files andintensities. Inthe case of the Nvitriplet,forinstance, withrespecttotheresonanceline.Likewise,theresonancelines
theinter-combinationlineisstrongerthantheresonancelinefor oftheNviandOviiionsarepredictedwithslightlytoostronga
the0.10keVplasmacomponent(inredinFig.4),whilstthere- blue-shiftcomparedto theobservations.Thereare several pos-
verse situation appliesto the 0.20keV component.The combi- siblereasonsforthesediscrepancies.
nationofbothcomponentsfitstheobservedtripletquitewell. Clearly, the sampling of the emitting plasma into four dis-
Figure6 illustrates the contributions to the observed emis- cretetemperaturecomponentsisasimplificationthataffectsthe
sion of eachcomponentofthe Nvi tripletfora line of sightof results.TheactualplasmainthewindofζPupismorelikelyto
impactparameter p = 0.5R .OwingtothepumpingbytheUV havearoughlycontinuoustemperaturedistribution.
photonsfromtheupperleve∗loftheftransition,thelatterlineis Anotherpointistheradialdependenceofthehotplasmafill-
stronglysurpressedatthebenefitoftheiline.Thisisespecially ing factor. In our model, we assume that the filling factor of
9
Herve´etal.:AdetailedX-rayinvestigationofζ-Puppis
a given plasma component remains constant over the full ex-
tentoftheemissionregionofthiscomponent.Thismightbean
over-simplification,asthefillingfactorcouldchange,especially
forthe 0.10and0.20keV plasmacomponentsthatspana wide
range of radii. Although implementing a radial dependence of
thefilling factorintoourcodeis in principlepossible,we have
refrainedfromdoingso,asitwouldintroduceadditionalfreepa-
rameters,therebyrenderingthesearchforabestfitmodeleven
moredifficult.
A third avenue to explore could be the velocity-law of the
hotgas.Throughoutthispaper,wehaveassumedthattheX-ray
emittingplasmafollowsthesameβ=1velocity-lawasthecool
gas.However,thehotplasmaishighlyionized,thuslackingline
opacityfortheradiativeaccelerationbythestellarUVradiation.
Although Coulomb forceshelp to carry the hot gas along with
thecoolwind,onecannotexcludethepossibilitythattheveloc-
ity law mightdifferfrom the one of the coolwind. Testing the
impactofthevelocitylawonourfitsisbeyondthescopeofthe
presentpaper.
Finally, more severe discrepancies,of similar type as those
observedhere, motivatedLeuteneggeret al. (2007) to incorpo-
rate the effect of resonant scattering into their line-by-line fits
of the Nvi and Ovii triplets4. Using the characteristic Sobolev
Fig.7. Ourbest-fit(dashedred line) of four fir triplets present
opticaldepthτ0, asafreeparameter,Leuteneggeretal.(2007) intheRGSspectrumofζPup(insolidblackline).
were indeed abl∗e to improve the quality, of their fits of these
twotriplets,butattheexpenseofverylarge,sometimesinfinite,
values of τ . Such large characteristic Sobolev optical depths
0,
areproblema∗ticthough,astheyrequireafillingfactorofthehot For instance, the Ovii resonance line has its observable emis-
plasmanearunity,farawayfromthevaluesderivedhereandby sionarisingmainlyfrombelow15R innottoobadagreement
other studies (Hillier et al. 1993). This problem was noted by withtheresultsofCassinellietal.(20∗01).
Leuteneggeretal. (2007), whoadvocatedthatsuchhighfilling FortheSxvtripletaround5.04-5.10Å,Leuteneggeretal.
factors were only required in small regions where line forma- (2006) obtained an inner radius of only (1.1+0.4)R , i.e. poten-
0.1
tiontakesplace.However,suchaconfigurationwouldthenalso tially much closer to the photosphere than a−ny of∗the emitting
have a strong impact on the formation of other lines, which is regions in our model. However, the HETG spectrum analysed
notseen.Thisiswhywedecidednottoincluderesonancescat- by these authorsis of ratherpoorqualityat these wavelengths.
tering into ourmodeland do not considerthis effect as a good Unfortunately, the RGS instrument does not cover this wave-
candidateforsolvingtheobserveddiscrepancies. length domain and we cannot verify whether or not the mor-
Usingthe ratiosevaluatedontheHETGspectraofζPup, phologyofthislineiswellpredictedbyourmodel.
R
Cassinellietal.(2001),estimatedaradialrangebetween1.9and
4.0R and between 1.7 and 2.25R for the formation regions
of the∗MgxiandSixiii tripletsrespe∗ctively.Based on the same 5.5.TheFexviiλ17.096Åline
dataset, Leutenegger et al. (2006) found an inner emitting ra-
TheforbiddenlinesoftheHe-likeionsarenottheonlyspectral
dius of the Mgxi and Sixiii triplets of (1.43 0.10)R , whilst
features arising from metastable levels which are depopulated
Oskinova et al. (2006) inferred an inner rad±ius of ∗ 1.5R . by the strong UV radiation field. Indeed, the transition Fexvii
At first sight, these results of Cassinelli et al. (2001≃) for th∗e
λ17.096Å also arises from a metastable level. Mauche et al.
Mgxi triplet are in reasonable agreement with the location of
(2001)haveshownthatthislevelcanbedepopulatedbyastrong
the 0.40keV plasma component, which dominates the forma-
extreme-UVcontinuumradiationbetween190and410Å.They
tion of these lines in our model (see Fig.4). This good agree-
also found that in the case of a stellar surface temperature
mentbetweenthelocalanalysisofCassinellietal.(2001)isdue ∼
55kK, the level could be totally depopulated. Since our code
to the narrow radial domain covered by the 0.40keV compo-
doesnotincludeafulltreatmentoftheatomictransitionsofthe
nent.Incontrast,Cassinellietal.(2001)foundthattheNeixand
Fexviiionandsincewehavenoobservationalconstraintsonthe
Oviitripletswerelikelyformedbelow4and10stellarradiire-
extremeUVemissionofζPup,wehaveincludedanadjustable
spectively. At first sight, this result disagrees with the location
reductionfactortofittheintensityofthisline.Ourbest-fitresults
of the 0.20keV plasma component which extends from 1.5 to
correspondto a division of the nominalline intensity by a fac-
38R . However, the actual location of the formation region of
tor4.5(seeFig.9).Therefore,itseemsthatthemetastablelevel
theo∗bservablelinesresultsfromtheinterplaybetweenintrinsic
is heavily depopulated,which is not surprising as the effective
emission and absorption, which is rather complex (see Fig.8).
temperatureofζPupis40kK.TheemissionseeninFig.9close
to17.1ÅactuallycorrespondstotheFexviiλ17.051Åline.
4 Leutenegger et al. (2007) further noted that models without res-
onant scattering produce an inter-combination line that was not suffi-
cientlyblue-shiftedcomparedtotheobservations(theirFigs.1and2). 5.6.Additionalunsolvedproblemsinthefit
The latterdiscrepancy isless severe in our model thanks to the com-
binationofseveralplasmacomponentswithdifferenttemperaturesand Beside the discrepancies in the fits of the fir triplets discussed
locations. above, a few additional regions in the spectra are not well fit-
10