The origin of hydrogen line emission for five Herbig Ae/Be stars spatially resolved by VLTI/AMBER

17 Pages·2008·1.11 MB·English

A&A489,1157–1173(2008) Astronomy
DOI:10.1051/0004-6361:200809946 &
(cid:2)c ESO2008 Astrophysics
The origin of hydrogen line emission for five Herbig Ae/Be stars
(cid:2)
spatially resolved by VLTI/AMBER spectro-interferometry
S.Kraus1,K.-H.Hofmann1,M.Benisty2,J.-P.Berger2,O.Chesneau3,A.Isella4,F.Malbet2,A.Meilland1,
N.Nardetto1,A.Natta5,T.Preibisch1,D.Schertl1,M.Smith6,P.Stee3,E.Tatulli2,L.Testi7,andG.Weigelt1
1 MaxPlanckInstitutfürRadioastronomie,AufdemHügel69,53121Bonn,Germany
e-mail:[email protected]
2 Laboratoired’AstrophysiquedeGrenoble,UMR5571UniversitéJosephFourier/CNRS,BP53,38041GrenobleCedex9,France
3 UMR6525H.Fizeau,Univ.NiceSophiaAntipolis,CNRS,ObservatoiredelaCôted’Azur,Av.Copernic,06130Grasse,France
4 Caltech,MC105-24,1200EastCaliforniaBlvd.,PasadenaCA91125,USA
5 INAF–OsservatorioAstrofisicodiArcetri,LargoFermi5,50125Firenze,Italy
6 CentreforAstrophysics&PlanetaryScience,UniversityofKent,CanterburyCT27NH,UK
7 EuropeanSouthernObservatory,Karl-Schwarzschild-Strasse2,85748Garching,Germany
Received10April2008/Accepted3July2008
ABSTRACT
Context.Accretionandoutflowprocessesareoffundamentalimportanceforourunderstandingoftheformationofstarsandplanetary
systems.Totracetheseprocesses,diagnosticspectrallinessuchastheBrγ2.166μmlinearewidelyused,althoughduetoalackof
spatialresolution,theoriginofthelineemissionisstillunclear.
Aims.EmployingtheAU-scalespatialresolutionwhichcanbeachievedwithinfraredlong-baselineinterferometry,weaimtodistin-
guishbetweentheoreticalmodelswhichassociatetheBrγlineemissionwithmassinfall(magnetosphericaccretion,gaseousinner
disks)ormassoutflowprocesses(stellarwinds,X-winds,ordiskwinds).
Methods.UsingtheVLTI/AMBERinstrument,wespatiallyandspectrally(λ/Δλ=1500)resolvedtheinner(<∼5AU)environmentof
fiveHerbigAe/Bestars(HD163296,HD104237,HD98922,MWC297,V921Sco)intheBrγemissionlineaswellasintheadjacent
continuum. Fromthemeasuredwavelength-dependent visibilities,wederivethecharacteristicsizeofthecontinuumandBrγline-
emittingregion.Additionalinformationisprovidedbytheclosurephase,whichwecouldmeasurebothinthecontinuumwavelength
regime(forfourobjects)aswellasinthespectrallyresolvedBrγemissionline(foroneobject).Thespectro-interferometricdatais
supplementedbyarchivalandnewVLT/ISAACspectroscopy.
Results.Forall objects (except MWC297), wemeasure anincrease of visibilitywithintheBrγ emission line, indicatingthat the
Brγ-emittingregionintheseobjectsismorecompactthanthedustsublimationradius.ForHD98922,ourquantitativeanalysisre-
vealsthattheline-emittingregioniscompactenoughtobeconsistentwiththemagnetosphericaccretionscenario.ForHD163296,
HD104237, MWC297,andV921Scoweidentifyanextendedstellarwindoradiskwindasthemostlikelyline-emittingmecha-
nism.Sincethestarsinoursamplecoverawiderangeofstellarparameters,wealsosearchforgeneraltrendsandfindthatthesizeof
theBrγ-emittingregiondoesnotseemtodependonthebasicstellarparameters(suchasthestellarluminosity),butcorrelateswith
spectroscopicproperties,inparticularwiththeHαlineprofileshape.
Conclusions.Byperformingthefirsthigh-resolutionspectro-interferometricsurveyonHerbigAe/Bestars,wefindevidenceforat
leasttwodistinctBrγline-formationmechanisms.Mostsignificant,starswithaP-CygniHαlineprofileandahighmass-accretion
rateseemtoshowparticularlycompactBrγ-emittingregions(R /R < 0.2),whilestarswithadouble-peaked orsingle-peaked
Brγ cont
Hα-lineprofileshowasignificantlymoreextendedBrγ-emittingregion(0.6<∼R /R <∼1.4),possiblytracingastellarwindora
Brγ cont
diskwind.
Keywords.stars:pre-mainsequence–stars:winds,outflows–planetarysystems:protoplanetarydisks–line:formation–
accretion,accretiondisks–techniques:interferometric
1. Introduction dust grains, providing (for the expected temperaturesand den-
sities) the dominant source of opacity. In recent years, there
Accretion disksaroundyoungstellar objects(YSOs) are atthe
hasbeensubstantialprogressinconstrainingthedetailedthree-
focus of astronomical research, not only because they play an
dimensionalgeometryofthedustdiskusing,forexample,com-
essentialroleinthestar-formationprocess,butalsobecausethey
bined modeling of the spectral energy distribution (SED) and
providethestagewhereplanetformationtakesplace.
spatially resolved infrared interferometric observations. While
Historically, these disks were discovered due to their char-
thethermalemissionfromthedustdiskislikelytobethedom-
acteristic infrared excess emission, which is believed to trace
inant contributor to the infrared excess emission observed to-
(cid:5) Based on observations made with ESO telescopes at the La Silla wardsYSOs, it isbelievedthatthe dustcontentmakesuponly
Paranal Observatory under open time programme IDs 077.C-0694, a small fraction of the total mass of the system. 99% of the
078.C-0360,and078.C-0680. mass is likely contributed by gas, in particular hydrogen, and
ArticlepublishedbyEDPSciences
1158 S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars
can mainlybetracedbythespectrallines,whichforpre-main- magnetic field lines of the accretion disk. In contrast to the X-
sequence stars are often found in emission. While some infor- windmodel,wherethewindislaunchedfromanarrowdiskan-
mationaboutthekinematicsofthegascanalreadybeextracted nulus around the inner truncation radius of the accretion disk,
from the line profile,the spatial originof the gasemission and diskwindsoriginatefromawiderangeofradii,likelyextending
thephysicalprocessestheytracearestillstronglydebated. fromtheco-rotationradiusouttoseveralAUs.
In this context, the Brackett-γ (Brγ) 2.1661 μm line is of Since mostof these processesare believedto take place on
special importance. It was found that the luminosity L(Brγ) (sub-)AU scales and were therefore not accessible with direct
of this line (as determined from the circumstellar component imaging techniques, most earlier studies tried to constrain the
of the line equivalent width) seems to correlate with the mass spatialdistributionandkinematicsoftheline-emittinggasfrom
accretionluminosityL ,asdeterminedfromUVveiling.This the shape of line profiles. However, since these line profile fit-
acc
empiricalL(Brγ)−L relationshiphasbeenestablishedforpre- ting techniquesare knownto be highly ambiguous(e.g.Catala
acc
main-sequencestarsofvariousmasses,includingbrowndwarfs etal.1999),spatiallyresolvedobservationsareurgentlyrequired
(Nattaetal.2004),TTauristars(Muzerolleetal.1998b),aswell toconstrainthephysicalmechanismandspatialoriginofimpor-
as intermediate-mass Herbig Ae/Be stars (Calvet et al. 2004; tant tracer lines like Brγ. Since the different emitting mecha-
vandenAncker2005).Based onthisempiricalcorrelation,ob- nisms noted above can be associated to distinct spatial regions
servers started to use the Brγ luminosity as an estimator for (see Fig. 1), spatially resolved observationsshould allow us to
the mass accretion rate (e.g. GarciaLopez et al. 2006), which identifyandphysicallycharacterizethetrueunderlyingemitting
followsfromL usingM˙ = L R /GM . mechanism.
acc acc acc (cid:5) (cid:5)
Therefore,it is now essential to identifythe process(es) in- The VLTI/AMBERinstrumentcombines,forthe first time,
volvedinthe formationoftheBrγ emissionline in YSOs. The the milli-arcsecond spatial resolution achievable with IR long-
main scenarios which have been proposed up to now include baseline interferometry with a good spectroscopic resolution
massinfall,aswellasmassoutflowmechanisms: (R=λ/Δλ=1500or12000)inthenear-infraredK-band.Inthis
study,weusetheuniquecapabilitiesofthisinstrumentinorder
to measure the geometry and position of the Brγ line-emitting
(a)Magnetosphericaccretion: thelineemissionmightemerge
region relative to the continuum-emittingregion, providing di-
from matter which is accreted onto the star through magneto-
rectinformationabouttheprocessesinvolved.
sphericaccretioncolumns(vandenAncker2005).Thisinfallis
A particularly well-studied class of YSOs are the
supposedtohappenveryclosetothestar,insidetheco-rotation Herbig Ae/Be stars (HAeBes). These are intermediate-mass,
radius,wheretheKeplerianangularvelocitymatchesthestellar
pre-main sequence stars. Over the last few years, the contin-
angularvelocity.
uumemissionfroma ratherlargenumberofHAeBescouldal-
readybestudiedwithbroad-bandinfraredinterferometry.These
(b) Gaseous inner disk: inside of the dust destruction radius, studiesrevealeda correlationbetweenthe size oftheNIRcon-
thegascontinuestoaccretetowardsthestar,formingagaseous tinuum emitting region and the stellar luminosity (Monnier &
inneraccretiondisk.Therecombinationlineemissionfromion- Millan-Gabet 2002), suggesting that for most Herbig Ae stars,
ized hydrogen in this disk might contribute to the line emis- the NIR emission likely traces dust at the dust sublimation
sion observed towards HAeBe stars. Muzerolle et al. (2004) radius.
estimated that especially for high mass-accretion rates M˙ > Only very little is known about the spatial origin of the
acc
10−7 M(cid:4) yr−1, the flux contribution of the inner gaseous disk hydrogen recombination line emission. First AMBER obser-
mightbeofimportance. vations of the Herbig Be star MWC297 have shown that the
Brγ-emitting region aroundthis star is more extended than the
continuum region (Malbet et al. 2007). Surprisingly, observa-
(c) Stellar wind: the P Cygni line profile observed in the hy- tions on the less luminous Herbig Ae star HD104237 did not
drogen lines of several HAeBe stars might also indicate mass show any change in visibility along the Brγ line (Tatulli et al.
loss through stellar winds (e.g. Mestel 1968;Catala & Kunasz 2007a). Eisner (2007) found a visibility increase within the
1987;Strafellaetal.1998).Whilemagneticallyacceleratedstel- Brγ line for the Herbig Ae star MWC480. This might sug-
lar winds seem plausible for massive stars rotating close to gestthattheBrγlinetracesfundamentallydifferentmechanisms
theirbreak-upvelocity,thisscenariodoesnotseemtoworkfor for Herbig Ae and Be stars. In this study, we present observa-
lower-massYSOs(Ferreiraetal.2006). tions on three HAeBes, which are resolved for the first time
withspectro-interferometry(HD163296,HD98922,V921Sco).
For HD104237, we present new observations which were ob-
(d) Stellar-field driven wind (X-wind): in the X-wind model
tained at longer baseline lengths, addingsubstantial new infor-
(Shuetal.1994),theoutflowsfromYSOsarelaunchedfroma
mation to the initial AMBER results presentedby Tatulli et al.
narrowinteractionregionwherethestellarmagnetospheretrun-
(2007a).Finally, we re-reducedthe archivaldata on MWC297
cates the accretion disk. Keplerian rotation leads to a winding
andHD104237inordertoprovideauniformlycalibratedsam-
up of the field lines and to the formationof magneticsurfaces.
ple of five stars for our interpretation. The stellar parameters
Chargedparticlesfromthestellarwindwillthengettrappedin
which we adopted for the stars in our sample are shown in
thesesurfaces,accelerated,andcollimatedintoabeam.
Table1.
Thispaperisstructuredasfollows:inSect.2wediscussthe
(e) Disk-field driven wind (disk wind): another mechanism spectro-interferometric and spectroscopic data and the applied
which has been proposed for the launching and collimation of datareductionprocedures,followedbyanoutlineoftheobtained
outflowsandjetsobservedtowardsYSOsaremagnetocentrifu- results(Sect.3).Then,wepresentthemodelswhichwerefitted
gally driven disk winds (Blandford & Payne 1982; Pudritz & tothedatainordertoconstrainthegeometryofthecontinuum-
Norman 1983; Ferreira 1997). Material from the disk surface (Sect.4)andBrγ-emittingregion(Sect.5).Afterinterpretingthe
of the rotating disk is centrifugally accelerated along the open individualobjects in detail (Sect. 6), we discuss generaltrends
S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars 1159
Fig.1.Illustrationoftheregionswhichhavebeenproposedastheoriginofthepermittedhydrogenrecombinationlineemissionobservedtowards
HAeBestars(thissketchisnottoscale;readSect.1fordetailsabouttheindividualmechanisms).
Table1.Targetstarsandadoptedstellarparameters.
Star Spectral L d T A R (a) logL(Brγ) (b) logM˙Brγ(c) vsini Hα(d) Ref.
(cid:5) (cid:5) V (cid:5) L(cid:4) acc
type [L(cid:4)] [pc] [K] [R(cid:4)] [M(cid:4)yr−1] [kms−1] profile
HD104237 A5 30 116(f) 8000 0.31 2.9 −2.74 −7.45 12±2(g) D(h) (i)
HD163296 A3 26 122(f) 8700 0.12 2.2 −2.78 −7.12 120+20(j) D(h) (k)
−30
HD98922 B9 890 540(e) 10600 0.3 9.1 −1.56 −5.76 – P(h) (l)
MWC297 B1.5 10600 250 23700 8 6.1 >−0.6 – 350±50(m) S(h) (m)
V921Sco B0 19950 800 30900 5.0 5.0 >−0.7 – – D(n) (n)
Notes–(a)ThestellarradiusR iscomputedusingthegiveneffectivetemperatureandbolometricluminosity;(b)luminosityoftheBrγemission
(cid:5)
line, as determined by GarciaLopez et al. (2006) from ISAAC spectra; (c) mass accretion rate, as determined from the L(Brγ)−L relation
acc
(GarciaLopez et al. 2006); (d) thiscolumn described the Hα-line profile shape of the starsin our sample using the classifications: D: double-
peaked profile; P: P-Cygni profile; S: single-peaked profile; (e) for HD98922, we assume the minimum distance given by GarciaLopez et al.
(2006).
References:(f)vandenAnckeretal.(1998);(g)Donatietal.(1997);(h)Ackeetal.(2005);(i)Tatullietal.(2007a); (j)Finkenzeller(1985);
(k)vandenAnckeretal.(2000);(l)GarciaLopezetal.(2006);(m)Drewetal.(1997);(n)Habartetal.(2003).
which we findin oursmall sample of HAeBe stars (Sect.7) in ofAMBER(R=1500)andfoundgoodagreement(seeFigs.3a
Sect.8. to7a,toppanel).
2.2.ArchivalISOandSpitzer/IRSspectroscopy
2. Observationsanddatareduction
InordertooptimallyconstraintheSEDforthestarsinoursam-
2.1.VLT/ISAACspectroscopy
ple, we obtained near- to mid-infraredspectroscopicdata from
The profile of the emission lines carries importantinformation the ISO and Spitzer Space Telescope archive. The Spitzer/IRS
about the kinematics of the emitting gas. Therefore, we com- spectra were pre-processed by the S13.2.0 pipeline version at
plement the spatially resolved AMBER spectro-interferometry the Spitzer Science Center (SSC) and then extracted with the
with high-spectralresolution (R ∼ 9000) spectra obtained with SMARTsoftware,Version6.2.5(Higdonetal.2004).
theVLT/ISAACinstrument.
Besides archival ISAAC data (ESO programme
2.3.VLTI/AMBERspectro-interferometry
073.C-0184, P.I. Habart), we also obtained new spectroscopic
data (for HD163296 and V921Sco, ESO programme 077.C- AMBER (Petrov et al. 2007) is the NIR beam-combiner of
0694, P.I. Kraus) in order to measure the line profile as close the Very Large Telescope Interferometer (VLTI), which is lo-
in time to the AMBER observations as possible. This new cated on Cerro Paranal/Chile and operated by the European
spectroscopic data covers not only the Brγ 2.1661 μm line Southern Observatory (ESO). Combining the light from up to
(Fig.2,left),butalsothePaβ1.2822μmline(Fig.2,right). threeofthefour8.4munittelescopessimultaneously,AMBER
The raw spectra were extracted using IRAF procedures measures not only visibility amplitudes, but also the closure
and then correctedfor atmosphericfeaturesusing telluric stan- phase (CP) relation.Inthe courseofthreeESO opentime pro-
dard star observations obtained during the same night. For grammes (077.C-0694, 078.C-0360, 078.C-0680, P.I. Kraus),
wavelength-calibration, we aligned the raw spectra to publicly we obtained spectrally dispersed interferograms in AMBER’s
available high-resolution(R = 40000)telluric spectra taken at medium resolution (MR) mode (R = 1500) on four HAeBes;
the NSO/Kitt Peak Observatory. To compare the final ISAAC namely, HD163296, HD104237, HD98922, and V921Sco.
spectra (Fig. 2) with the spectra extracted from the AMBER Thesenewdatasetswerecomplementedwitharchivaldatafrom
data,weconvolvedtheISAACspectratothespectralresolution MWC297 and HD104237, obtained earlier during AMBER
1160 S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars
Table2.ObservationlogofourVLTI/AMBERinterferometricobservationswithaspectralresolutionR=1500.
Targetstar Date Time Spectral DIT Telescope Projectedbaselines Calibrator Ref.
(UT) (UT) window [ms] triplet B PA B PA B PA
L L L
[μm] [m] [◦] [m] [◦] [m] [◦]
HD104237 2005-02-26 07:14 2.12–2.20 100 UT2-UT3-UT4 35.1 71 (60.8 120) (87.9 102) HD135382 (a)
HD104237 2007-01-09 08:03 1.94–2.26 200 UT1-UT3-UT4 (78.0 31) 58.5 84 (122.4 53) HD118934
HD104237 2007-01-09 07:41 1.94–2.26 500 UT1-UT3-UT4 (78.7 26) 58.0 78 (123.1 47) HD118934
HD163296 2006-04-13 06:19 2.12–2.20 50 UT2-UT3-UT4 43.9 19 52.3 99 74.1 63 HD171960
HD89682
HD98922 2007-02-04 07:19 2.12–2.20 50 UT2-UT3-UT4 42.9 46 (61.9 111) (88.9 85) HD101328
HD98922 2007-01-09 09:31 1.94–2.26 200 UT1-UT3-UT4 (90.5 42) 62.3 116 (122.8 71) HD118934
HD98922 2007-01-09 09:39 1.94–2.26 500 UT1-UT3-UT4 (89.9 43) 62.4 118 (122.0 73) HD118934
MWC297 2004-05-31 06:02 2.00–2.23 107 UT2-UT3 44.7 42 – – – – HD177756 (b)
V921Sco 2006-04-14 08:09 2.12–2.20 50 UT2-UT3-UT4 45.2 42 62.1 110 (89.3 82) HD159941
Notes–Formoredetailedinformationaboutthecalibratorstars,werefertoTable3.
References–(a)reprocessingofdatapresentedinTatullietal.(2007a);(b)reprocessingofdatapresentedinMalbetetal.(2007).
Table3.Calibratorstarinformationfortheinterferometricobservations Inafirstdatareductionstep,theAMBERrawinterferograms
presentedinTable2. wereclearedfromthecorrelateddetectornoiseeffectusingthe
AMDCsoftwaretool(Version1.1,Causietal.2008).Then,the
Star V K Spectral d rawdatawasreducedwith theamdlib2software1 (release2.1),
UD
type [mas] employingtheP2VMalgorithm(Tatullietal.2007b).Duetothe
HD101328 7.44 3.75 K4III 1.00±0.01(a) absenceofafringetracker,alargefractionoftheinterferograms
HD118934 7.92 4.02 K4III 0.89±0.01(a) is of rather low contrast (see discussion in Petrov et al. 2007).
HD135382 2.88 2.53 A1V 1.10±0.08(b) Therefore, we removed any frames from our dataset for which
HD159941 7.85 3.55 M0III 1.09±0.02(a) thelightinjectionfromthecontributingtelescopeswasunsatis-
HD171960 7.29 3.38 K3II 1.13±0.02(a) fying;i.e.,theintensityratiobetweenthephotometricchannels
HD177756 3.43 3.56 B9Vn 0.60±0.06(b) was larger than 4, or the fringe contrast was decreased due to
Notes – The V-band magnitudes were taken from SIMBAD and the instrumental jitter (the 20% best interferograms were selected
K-bandmagnitudesfromthe2MASSpointsourcecatalog. based on the Fringe SNR criteria, as defined in Tatulli et al.
References – (a) UD diameter taken from the CHARM2 catalog 2007b).
(Richichietal.2005).(b)UDdiametercomputedwithASPRO. Sincesomestarsinoursampleareclosetothecurrentsensi-
tivitylimitofAMBER’sMRmode,weappliedspectralbinning
commissioning and in guaranteed-time observations (Malbet to the raw data in order to increase the fringe SNR. For each
etal.2007;Tatullietal.2007a).Re-reducingthesedatasetswith data set, the width of the sliding window was chosen so that
thelatestsoftwareallowsustoensurehomogenitybothin data theresultingfringeSNRoftheinterferogramsexceedsthecrit-
reductionaswellasintheappliedmodelingprocedures.Theob- icalvalueof1.5,ensuringareliablevisibilityestimation.Since
servationsare summarizedin Table 2 and were obtainedunder the Brγ line for all our objects is spectrally resolved over sev-
goodatmosphericconditions(seeing0.5−0.9(cid:6)(cid:6),atmosphericco- eral spectral channels at R = 1500, the resulting decrease in
herencetime2−7ms).Inordertocalibratetheobtainedvisibil- spectralresolutiononlymarginallyreducesthecontrastbetween
itiesandCPsforinstrumentalandatmosphericeffects,weused the spectralline andthe underlyingcontinuum.Thisprocedure
interferometriccalibratorstars,takingtheintrinsicdiametersof increases the fringe SNR significantly, while the loss in spec-
thesestarsintoaccount(seeTable3).Byusingdifferentdetector tral resolution (from R = 1500 to 500 or 250) results only in
integrationtimes(DITs)forthedifferentobjects,weaimedfora aminordecreaseoftheline-to-continuumfluxratioFBrγ/Fcont.
compromisebetween collectinga sufficientnumberof photons BaselinesforwhichaspectralbinningtoR=250wasnotsuffi-
and minimizing the loss of fringe contrast due to atmospheric cienttoyieldanSNRof1.5wererejectedfromfurtheranalysis
piston.Ingeneral,observationswithlongerDITshouldprovide (inTable2,thesebaselinesareputinbrackets).Thefinalvisibil-
betterSNRinthedifferentialvisibilitymeasurement,whileob- ity curvesareshowninFigs.3ato7ainthesecondpanelfrom
servationswithshorterDITshouldprovideabetterabsolutecali- thetop.
bration.Sinceweaimedmainlyforaprecisemeasurementofthe In principle, two distinct phase quantities can be extracted
differentialquantitiesintheBrγandtheadjacentcontinuum,we fromAMBERinterferograms,namelydifferentialphases(mea-
usedratherlongDITformostobservations.Inordertoaccount suring the relative displacement of the line-emitting region
for the resultingerrorson the absolutecalibration,we estimate with respect to the continuum-emittingregion) and the closure
a rather large calibration error of 5%. The spectral calibration phase, which can indicate asymmetries in the source bright-
ofthedatawasdonebycomparingthespectrumextractedfrom ness distribution. Since for most of our datasets, the fringe
the AMBER data with ISAAC spectra of the same object (see SNR in individual frames is not sufficient to reliably correct
Sect. 2.1), which have been convolved to the spectral resolu- for disturbing atmospheric phase contributions (atmospheric
tion of AMBER. An additionalvery importantcalibration step piston), no scientifically meaningful differential phases could
inthecourseofAMBERdatareductionwastherelativespectral be extracted, unfortunately. However, for some data sets we
re-shifting of the photometric channels with respect to the in- could extract useful CP information, both in the continuum
terferometric channel. The appropriateshift was determinedto
sub-pixel accuracy by computing the auto-correlation between 1 Theamdlib2softwarepackageisavailablefromthewebsite
thespectraextractedfromthesechannels. http://www.jmmc.fr/data_processing_amber.htm
S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars 1161
Table4.Closurephasesextractedfromthedata.
Projectedbaselines Closurephase
u v u v continuum Brγline
12 12 23 23
[m] [m] [m] [m] [deg] [deg]
HD104237 33.1 11.6 52.9 –30.0 9.8±6.7 –
HD163296 14.5 41.4 51.6 –8.0 1.5±4.3 –
HD98922 30.9 29.7 57.6 –22.6 −9±16 –
V921Sco 30.2 33.7 58.3 –21.6 20±24 3.1±2.8
dereddenedthephotometricandspectroscopicdataassumingthe
valuesforA giveninTable1,R =3.1,andtheextinctionlaw
V V
byMathis(1990).ThederivedSEDs(Figs.3bto7b)allowusto
determinetheratiobetweenthestarfluxcontribution f andthe
star
fluxcontributionfromthecircumstellardisk f forthefitting
disk
ofgeometricmodels(Sect.4).
TheVLT/ISAACBrγlinespectraareshowninFig.2.With
the currently available resolution, all Brγ lines show a single-
peakedprofile.ForHD163296andV921Sco,measurementsat
twoorthreeepochsareavailable,indicatingthatthelineprofile
ofV921Scohasnotsignificantlychanged,whileforHD163296
wefindsomevariability.
3.2.Spectro-interferometricvisibilities
The wavelength-dependent visibilities extracted from our
AMBER data are shown in the second panel from the top in
Figs.3ato7a.Fourstarsinoursample(HD163296,HD104237,
HD98922, V921Sco) show a visibility increase within the
Brγ line. For MWC297,the visibility dropswithin the line, as
alreadyreportedbyMalbetetal.(2007).
Since the spectral channels which include the Brγ line
additionally contain flux contributions from the photosphere
and circumstellar material, the interpretation of the mea-
sured wavelength-dependent visibilities requires quantitative
modeling,aspresentedinSect.4.
Fig.2.Left:VLT/ISAACspectrashowingtheBrγlinewithaspectral
resolutionofR∼9000forthestarsinoursample.Right:VLT/ISAAC
3.3.Closurephases
spectraofthePaβlinewereobtainedforHD163296andV921Sco.
Closure phase measurements can provide unique information
aboutdeviationsfromcentro-symmetryinthesourcebrightness
(HD104237, HD163296, HD98922, V921Sco) as well as in distribution.ForYSOdiskgeometries,suchasymmetriesareex-
the line regime (V921Sco). It is expected that the continuum pected in particular for systems seen under intermediate incli-
CP does not change significantly over the small spectral win- nation. The strongest CP signals are predicted by disk models
dow covered by our observations, which allowed us to ap- with vertical puffed-upinner rim (e.g. Dullemond et al. 2001),
ply an optimized data reduction strategy for the extraction of while models with curved inner rims predict a smoother,more
the continuum CPs. First, we apply some spectral binning to symmetricbrightnessdistributioncorrespondingtosmallerCPs
the raw data (bracketing out the spectral channels containing (e.g.Isella&Natta2005).
the emission line), followed by an averaging of the complex Giventheimportanceofthisobservable,wepresentherethe
visibilities over various spectral channels (between 2.12−2.15 CPs extracted from our data, although the error bars on most
and 2.175−2.185 μm). The obtained CPs will be presented in measurements are rather large, typically due to the low fringe
Sect.3.3. contrastonthelongestbaselineinthetelescopetriplet.InTable4
we list the CPs measured from our datasets and give the cor-
responding (u,v) baseline vector for two baselines in the em-
3. Results ployed telescope triplet (the third (u,v)-vector is given by the
closurerelation).WefindthatallcontinuumCPsareconsistent
3.1.SEDandNIR-spectroscopy
with a zero CP on the 1σ (HD163296, HD98922, V921Sco)
In order to constrain the SED of the stars in our sample, we or 2σ level(HD104237).Thisaddssupportto the conclusions
collectedphotometricandspectroscopicdatafromtheliterature drawn by Monnier et al. (2006), who used the IOTA-3T inter-
andtheISOandSpitzerarchives(seeSect.2.2).Assumingthat ferometer(providingbaselinelengthsupto38m)andmeasured
theline-of-sightextinctioncanbemainlyaccountedtothelarge for 12 outof 13 sourcesCPs below ∼5◦ (excludingone binary
scaleenvelopes,inwhichsomeHAeBestarsareembedded,we source).SincetheCPsignalisexpectedtoincreaserapidlywith
1162 S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars
Fig.3.Spectroscopic, spectro-interferometric,andphotometricdataforHD104237: ina),weshowtheBrγ spectraextractedfromour ISAAC
andAMBERdata(toppanel),themeasuredwavelength-dependent visibilitiesV fordifferentbaselines(2ndpanel,thedatapointsareshown
tot
with statistical errors bars, whereas the estimated calibration errors are shown in the bottom-right corner), the continuum-corrected Brγ-line
visibilityV (3rdpanel),theUDdiameterΘ derivedfromV (4thpanel),andthediametersD andD derivedforthecontinuum-andthe
Brγ tot tot cont Brγ
Brγ-emittingregionusinga2-ringmodel(5thpanel).Inb),theSEDisshown,includingphotometricdatafromtheliterature(greypoints),archival
ISOspectra(magneta),andarchivalSpitzer/IRSspectra(blue).Inc),weplotthemeasuredK-bandcontinuum-visibilityasafunctionofbaseline
lengthandind)thecontinuum-correctedBrγ-linevisibility.Panele)showsthederivedringdiametersforthecontinuum-andline-emittingregion
plottedasfunctionofpositionangle.
baseline lengths, our new observations (with baselines up to 3.1±2.8◦,consistentagainwithacentro-symmetricbrightness
89 m) providenew constraintson this issue and favouragaina distribution.
centro-symmetricbrightnessdistribution.InSect.6,wediscuss
someofourCPmeasurementsqualitatively,butleaveittofuture
4. Modeling:continuum-emission
studiestoinvestigatewhetherourresultsarealsoinquantitative
agreementwiththecurrentgenerationofrimmodels. Over the last few years, ring models with uniform brightness
For V921Sco, the particularlystrongBrγ line flux allowed have emerged as the prototypical geometry for the interpre-
us also to measure an accurate CP measurement within the tation of YSO interferometric data. In most cases, this pref-
spectrally resolved emission line, yielding a CP signal of ered use of ring geometries instead of other simple geometries
S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars 1163
Fig.4.Spectroscopic,spectro-interferometric,andphotometricdataforHD163296(similartoFig.3).
(e.g. uniform disk, Gaussian, or disk geometries with constant luminosityscaling law (Monnier& Millan-Gabet2002).Inthe
temperature power-law) is not explicitly required to reproduce meantime, this finding was also interpreted in theoreticalwork
the details of the sampled visibility function (which typically (e.g. Natta et al. 2001; Dullemond et al. 2001; Isella & Natta
covers only the first lobe and is thus rather insensitive to the 2005),attributingthe near-infraredcontinuumemissionmainly
innergapinthebrightnessdistribution),butmainlybasedonin- tohotdustlocatedatthedustsublimationradius.Basedonthese
directevidenceortheoreticalarguments.Forexample,someof arguments, we also prefer to use ring-like geometries for the
the firstinfraredinterferometricmeasurementsonHAeBe stars interpretation of our visibility data (assuming a fractional ring
(e.g. Millan-Gabet et al. 2001) already showed that classical widthof20%,Monnieretal.2005),butalsogiveuniformdisk
geometrically thin accretion disk geometries extending down (UD)andGaussianFWHMdiameterstoallowcomparisonwith
to several stellar radii might be consistent with interferomet- other work. In order to estimate the contribution of the stellar
ric measurements for early-type Herbig Be stars, but result in photosphere to the total K-band flux, we estimate the flux ra-
too compact structures for Herbig Ae and late-type Herbig Be tiousingtheSEDandtheKuruczatmospheremodelsshownin
stars(Eisneretal.2004;Vinkovic´&Jurkic´2007).Ringgeome- Figs. 3b to 7b. For more details about the model fitting proce-
tries,ontheotherhand,yieldsizeswhichareconsistentwiththe dure,werefertooneofourearlierstudies(Krausetal.2008).
expecteddustsublimationradiiandfollowthepredictedstellar
1164 S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars
Fig.5.Spectroscopic,spectro-interferometric,andphotometricdataforHD98922(similartoFig.3).
Besides model fits to each individual spectral channel (see the visibilities derived for HD163296 provide sufficient posi-
fourthandfifthpanelofFigs.3ato7a),wealsofittedthesemod- tion angle coverage to investigate for a possible elongation of
els to continuum-visibilities,whichwere averagedoverseveral thecontinuum-emittingregion(e.g.,duetodiskinclination),we
spectral channels around 2.15 and 2.18 μm and plot the corre- also fitted inclined ring geometries to this data set and derived
spondingmeasurementsandmodelcurvesasafunctionofbase- aninclinationof68±10◦withaPAof144±9◦(corresponding
line length (Figs. 3c to 7c). In these plots, we also show the toaninclinedringdiameterof1.02×0.37AU).
upper limits, which we can put on the continuumvisibility us-
ingthebaselineswhichhadtoberejectedduetolowSNR(see 5. Modeling:Brγlineemission
Sect.2.3).ThefitteddiametersforGaussian,UD,andringpro-
filesarelistedinTable5.Toconvertthemeasuredangularsizeto FromeachspectralchannelofourAMBERinterferograms,we
physicalscales,weassumethedistanceslistedinTable1.Please can derive a value for the visibility amplitude, providing spa-
notethattheerrorsgivendonottakedistanceuncertaintiesinto tial information about the brightness distribution contributing
account.Byfittingthevisibilitiesmeasuredondifferentbaseline to this spectral channel. Within the Brγ line, the flux within a
orientations simultaneously, we assume that the intensity pro- spectral channel is composed of the line emission plus the un-
filedoesnotdependonpositionangle(e.g.face-ondisk).Since derlyingcontinuumcontribution.Tomodelsuchvisibilitydata,
S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars 1165
Fig.6.Spectroscopic,spectro-interferometric,andphotometricdataforMWC297(similartoFig.3).
either the composite (line and continuum) object can be mod- relation assumes a negligible (zero) differential phase (corre-
eled, or the measured continuum+line visibility has to be cor- sponding to coinciding photocenters between the continuum-
rected forthe continuumcontribution,yieldingthe visibility of and line-emitting regions). We compute the continuum-
thepureline-emittingregion. correctionforeachspectralchanneloftheAMBERdatawhere
Wecorrectthemeasuredvisibilitiesusingtherelation F is sufficientlylargeto applyEq.(1)reliablyand showthe
Brγ
resultingV valuesinthethirdpanelfromthetopofFigs.3a
Brγ
V = FtotVtot−FcontVcont, (1) to7a.
Brγ F
Brγ
The interpretation of the derived line visibilities is diffi-
where F and V are the flux and the visibility measured cult for several reasons. For instance, it is likely that the ge-
tot tot
within an arbitrary spectral channel. To compute the Brγ-line ometry of the line-emitting region is more complex than the
fluxF := F −F ,wedeterminethecontinuumfluxF continuum-emittingregion,possiblyextendingabovetheequa-
Brγ tot cont cont
from Kurucz atmosphere models, taking the underlyingphoto- torialdiskplane,perhapsintroducingstronginclinationeffects.
sphericBrγabsorptioncomponentintoaccount.Thenweinter- Furthermore,asindicatedbytheshort-periodlinevariabilityde-
polatethecontinuumvisibilityV overtheBrγlineandderive tectedtowardsYSOsofallmasses,thekinematicsandpossibly
cont
the continuum-correctedvisibility V of the Brγ-emitting re- also morphologyof the line-emitting gas might change signif-
Brγ
gion. As discussed in Weigelt et al. (2007, Appendix C), this icantly even on short time scales of days or weeks, involving,
1166 S.Krausetal.:OriginofhydrogenlineemissioninHerbigAe/Bestars
Fig.7.Spectroscopic,spectro-interferometric,andphotometricdataforV921Sco(similartoFig.3).
Table5.Best-fitparametersforourgeometricmodelfits.
Continuumemission Brγemission Brγemission
linecenter allspectralchannel
Gaussian uniformdisk ring ring ring
FWHM χ2 Θ χ2 D χ2 D χ2 D χ2
cont red cont red cont red Brγ red Brγ red
targetstar [AU] [AU] [AU] [AU] [AU]
HD104237 0.57±0.10 1.0 0.87±0.08 0.5 0.58±0.04 0.3 0.31±0.13 0.6 0.35±0.21 0.5
HD163296 0.39±0.05 0.9 0.61±0.08 1.5 0.41±0.06 1.9 0.35±0.15 0.1 0.25±0.19 0.1
HD98922 2.45±0.26 0.5 3.70±0.27 0.2 2.47±0.16 0.2 <0.5 – <0.5 –
MWC297 1.27±0.10 – 1.99±0.13 – 1.35±0.08 – 1.89±0.52 0.1 2.52±0.44 0.4
V921Sco 5.38±0.52 0.7 8.22±0.55 0.1 5.49±0.32 0.1 4.45±0.55 0.1 4.35±0.64 0.2
(cid:2)(cid:3) (cid:4)
Notes–χ2 isdefinedas (V −V )/σ 2/(N−1),whereN isthenumberofmeasurements,V isthemodelvisibility,andV
red meas model Vmeas model meas
andσ arethemeasuredvisibilityandtotalerror,respectively.
Vmeas