A Comprehensive Analysis of the Magnetic Standard Star HD 94660: Host of a Massive Compact Companion?

0.45 MB·English

by  J. D. Bailey

Astronomy&Astrophysicsmanuscriptno.Bail-Grun-Land˙HD94660˙v8revised (cid:13)c ESO2015
January30,2015
A Comprehensive Analysis of the Magnetic Standard Star HD
⋆
94660: Host of a Massive Compact Companion?
J.D.Bailey1,J.Grunhut2,andJ.D.Landstreet3,4
1 MaxPlanckInsitutfu¨rExtraterrestrischePhysik,Giessenbachstrasse1,85748Garching,Germany
e-mail:[email protected]
2 ESO,Karl-Schwarzschild-Strasse2,85748Garching,Germany
3 ArmaghObservatory,CollegeHill,Armagh,BT619DG,NorthernIreland,UK
5 4 DepartmentofPhysicsandAstronomy,TheUniversityofWesternOntario,London,Ontario,N6A3K7,Canada
1
0 Received2014;accepted2014
2
ABSTRACT
n
a Aims.Detailedinformationaboutthemagneticgeometry,atmosphericabundancesandradialvelocityvariationshasbeenobtainedfor
J
themagneticstandardstarHD94660basedonhigh-dispersionspectroscopicandspectropolarimetricobservationsfromtheUVES,

9 HARPSpolandESPaDOnSinstruments.
2 Methods.Weperformadetailedchemicalabundanceanalysisusingthespectrumsynthesiscodezeemanforatotalof17elements.
Usingbothline-of-sightandsurfacemagneticfieldmeasurements,wederiveasimplemagneticfieldmodelthatconsistsofdipole,

] quadrupoleandoctupolecomponents.
R
Results.TheobservedmagneticfieldvariationsofHD94660arecomplexandsuggestaninhomogeneousdistributionofchemical
S elements over the stellar surface. This inhomogeneity is not reflected in the abundance analysis, from which all available spectra
. aremodelled,butonlyameanabundanceisreportedforeachelement.Thederivedabundancesaremostlynon-solar,withstriking
h
overabundancesofFe-peakandrare-earthelements.Ofnotearetheclearsignaturesofverticalchemicalstratificationthroughoutthe
p
stellaratmosphere,mostnotablyfortheFe-peakelements.Wealsoreportonthedetectionofradialvelocityvariationswithatotal
o- rangeof35kms−1inthespectraofHD94660.Apreliminaryanalysisshowsthemostlikelyperiodofthesevariationstobeoforder
r 840dand,basedonthederivedorbitalparametersofthisstar,suggeststhefirstdetectionofamassivecompactcompanionforamain
t sequencemagneticstar.
s
a Conclusions.HD94660exhibitsinterestinglycomplexmagneticfieldvariationsandremarkableradialvelocityvariations.Longterm
[ monitoringisnecessarytoprovidefurtherconstraintsonthenatureoftheseradialvelocityvariations.Detectionofacompanionwill
helpestablishtheroleofbinarityintheoriginofmagnetisminstarswithradiativeenvelopes.

1
v Keywords.stars:magneticfield–stars:chemicallypeculiar–stars:abundances–(Stars:)binaries-spectroscopic–
4
9
4 1. Introduction ofabout1kGandhBioforderseveralthousandstotensofthou-
7
sandsofgauss(e.g.Donati&Landstreet,2009).
0 Oforder10%ofmainsequenceAandBstarsaremagnetic.For
ThechemicalpeculiaritiesofApstarscanbequitestriking.
1. themajorityofthesestars,theirmagneticpropertiesarestudied For example, the Fe-peak element Cr may be as much as 102
0 usingcircularpolarisationspectra,which providea measureof
timesoverabundantcomparedtotheSun.Stillmoreimpressive
5 themeanmagneticfieldprojectedalongtheline-of-sight(hB i).
z aretheabundancesofrare-earthelements,whicharecommonly
1 Afractionofthesestarsrotateslowlyand/orhavestrongenough inexcessofsolarvaluesbyasmuchas104times.
: magnetic fields that Zeeman splitting is observed in individual
v Ap stars may be highly variable, with their magnetic field
i lines of the intensity (Stokes I) spectra (see Bailey, 2014, and
X references therein). The magnitude of this splitting provides a strengths and spectral line strengths and shapes varying with
therotationperiodofthestar. Thisvariabilityisbestexplained
r measure of the mean magnetic field modulusat the stellar sur-
a face(hBi).ThesemagneticAandBstarsexhibitanomalousat- via the rigid rotator model (see Stibbs, 1950). In this model,
the line-of-sight and magnetic axis are at angles i and β to
mosphericabundancesandarereferredtoasthechemicallype-
the rotation axis, respectively. Therefore, different magnetic
culiarA–type(Ap)stars. ThesestarsoftenhavehB iinexcess
z fieldmeasurementsthroughouttherotationcycleofthestarare
the result of observing the field at different orientations and
⋆ Based in part on our own observations made with the European sincechemicalelementsaredistributednon-uniformlyandnon-
Southern Observatory (ESO) telescopes under the ESO programme axisymmetricallyoverthestellarsurface,spectrumvariabilityis
093.D-0367(A) and programmes 076.D-0169(A), 088.D-0066(A), alsoobserved(Ryabchikova,1991).
087.D-0771(A), 084.D-0338(A), 083.D-1000(A) and 60.A-9036(A),
HD 94660( = HR 4263)is a bright(V = 6.11)chemically
obtained from the ESO/ST-ECF Science Archive Facility. It is also
peculiarmagneticApstarthatiscommonlyusedasamagnetic
basedinpartonobservationscarriedoutattheCanada-France-Hawaii
standardtotestpolarimetricsystemsinthesouthernhemisphere.
Telescope(CFHT)whichisoperatedbytheNationalResearchCouncil
ofCanada,theInstitutNationaldesSciencedel’UniversoftheCentre ThefieldwasfirstdiscoveredbyBorra&Landstreet(1975)us-
NationaldelaRechercheScientifiqueofFranceandtheUniversityof ingan Hαmagnetograph;theyreporteda valuefortheline-of-
Hawaii. sight magnetic field of hBzi = −3300±510 G. HD 94660 is a
2 J.D.Baileyetal.:TheMagneticStandardHD94660
Table 1. Summary of the stellar and magnetic properties of servationalsoincludinglinearpolarisationinbothStokesQand
HD94660. U),andtwo unpolarisedStokes I spectra,allacquiredbetween
May 2009 and April 2014. Each polarimetric observation (ex-
Spectraltype A0pEuSiCr Rensonetal.,1991 ceptfortheobservationtakenon31-05-2009)wasobtainedby
T (K) 11300±400 Thispaper
eff acquiring four successive individual spectra with the quarter-
logg(cgs) 4.18±0.20 Thispaper
wave plate rotated in such a way to acquire the desired Stokes
R(R ) 2.53±0.37 Thispaper
⊙ spectrum(seee.g.Rusomarovetal.,2013, forfurtherdetails).
vsini(kms−1) <2 Thispaper
P(d) 2800±250 Landstreetetal.,2014 The HARPSpol spectra were reduced using a modified
log(L⋆/L⊙) 2.02±0.10 Thispaper version of the reduce package (Piskunov&Valenti, 2002;
M(M ) 3.0±0.20 Thispaper
⊙ Makaganiuketal., 2011). Wavelength calibration was per-
π(milliarcsec) 6.67±0.80 vanLeeuwen2007
formed using the spectrum of a ThAr calibration lamp and
B (G) −7500 Thispaper
d then corrected to the heliocentric rest frame. Normalisation of
B (G) −2000 Thispaper
q the spectra was achieved by first dividing the spectra by the
B (G) 7500 Thispaper
oct
i(◦) 16 Thispaper optimally-extractedspectrum of the flat-field to correct for the
β(◦) 30 Thispaper blaze shape and fringing. The resulting spectra were then cor-
rected by the response function derived from observations of
theSun.Thelaststepinvolvedfittingasmooth,slowly-varying
functiontotheenvelopeoftheentirespectrum.Thefinaloutput
sharp-lined star with clearly resolved Zeeman splitting in sev- is a set of continuum-normalised Stokes I spectra, the Stokes
eralspectralines(Mathys,1990).Bohlenderetal.(1993)report parameter of interest (V,Q,U) and a diagnostic null spectrum
aprojectedrotationalvelocityofvsini<6kms−1andaroughly thatiscalculatedinsuchawaythatthepolarisationcancelsout,
constantline-of-sightmagneticfieldstrengthofhB i=−2520G, whichoftenallowsustoidentifyspurioussignalsthatarepresent
z
basedonmeasurementsusingHβ.Aperiodofrotationoforder in the processed data. The spectrum that was acquired on 31-
2700 d was first proposed by Hensberge (1993) and later dis- 05-2009onlycompletedhalf of the full polarimetricsequence,
cussed by Mathysetal. (1997) with respect to hBi data. More whichenablesthecancellationoffirstorder(linear)wavelength
recently, Landstreetetal. (2014) studied the field variations of drift, but does not correct for second order (quadratic) drift
HD94660from17FORS1observationstakenovera6yearpe- and does not allow the computation of a diagnostic null spec-
riod from 2002 to 2008. They find a peak-to-peak variation in trum. We also note that the retarder angles listed in the obser-
hB i of about800 G, rangingfrom about−2700and −1900G. vationsobtainedon05-01-2010wereinconsistentwiththeusu-
z
Fromthese variations,a rotationperiodof2800± 250d isde- ally adopted values for obtaining Stokes V measurements and
duced,whichagreeswithpreviousdeterminations. arelikelyerroneous.Wethereforeproceededtoreducethedata
Landstreet&Mathys (2000) were the first to report clear assuming the usual retarder angles for the sequence of obser-
hB i variations (between about −1800 and −2100 G) which, vations,andthenverifiedthattheresultingpolarisedandunpo-
z
together with hBi data from Mathys&Hubrig (1997), enabled larisedspectra werein goodagreementwiththe otherobserva-
them to model the magnetic field of this star using a colin- tions.
earmultipoleexpansion.Theyadoptedamodelthatconsistsof
InadditiontotheHARPSpolobservations,weidentifiedan-
dipole, quadrupole and octupole components with surface po-
otherarchival,circularlypolarised,high-resolution(R≃65000)
larfieldstrengthsof−8400,2700and6900G,respectively.As
spectropolarimetric observation acquired with ESPaDOnS on
seenabove,thevalueofhB iisalwaysnegative,indicatingthat
z January 9, 2006. ESPaDOnS, which is mounted at the 3.6 m
i+β<∼90◦i.e.onlythenegativemagneticpoleisobserved.This Canada-France-Hawaii Telescope (CFHT), is also a bench-
isconfirmedbythismodelwherethedeterminedvaluesforiand
mounted, cross-dispersed echelle spectropolarimeter, which
βare5◦ and47◦.We pointoutthat,ingeneral,thismodeldoes
coversabroaderspectralrangecomparedtoHARPSpolof3690
providearoughfirstapproximationtothefieldvariations.
–10481Å. Thepolarimetricobservationwasobtainedbytak-
Inthispaper,wediscusseffortstomodelthemagneticfield
ingasequenceoffoursub-exposureswithdifferentpositionsof
and chemical abundances of many elements based on high-
the Fresnel Rhomb to acquire a single Stokes V spectrum, ac-
dispersion,polarimetricspectra.Thefollowingsectiondiscusses
cordingtotheproceduredescribedbyDonatietal.(1997).The
the observations. Sect. 3 outlines the derived physical param-
ESPaDOnSspectrawereprocessedusingthe automatedreduc-
eters; Sect. 4 discusses the magnetic field measurements and tionpackagelibre-esprit, followingthedouble-ratioprocedure
model;Sect.5and6describethemodellingtechniqueandabun-
(Donatietal.,1997).
dance analysis, respectively;Sect. 7 reportsdetected radial ve-
locityvariations;andSect.8summarisestheresultsofthepaper. Lastly, we also found three nights of archival UVES data
consisting of a total of 11 spectra. UVES is a cross-dispersed
spectrographmountedontheESO8.2-mVeryLargeTelescope
2. Observations (VLT)locatedatParanal.Ithasbothablueandredarmoffering
differentspectralresolutionsandwavelengthcoverage.Theblue
We have acquired one spectropolarimetric observation of
armoffersR upto about80000anda spectralrangeof3100–
HD94660andhaveretrievedanadditionalsevenarchivalspec-
4900 Å, whereas the red arm covers 4800 – 10200 Å with R
tra that were utilised in this study, all of which were taken
uptoabout110000.Table2summarisesourentirecollectionof
with HARPSpol. This is a high-resolution (R ≃ 115 000),
spectraforHD94660.
cross-dispersedechellespectropolarimeterthatcoversaspectral
rangebetween3780–6910ÅandismountedontheEuropean
SouthernObservatory(ESO)3.6-mtelescopelocatedatLaSilla.
The high signal-to-noise ratio (SNR) observations consist of 6
circularly polarised Stokes V spectra (with the 01-04-2012ob-
J.D.Baileyetal.:TheMagneticStandardHD94660 3
Table2.LogofavailablespectraforHD94660.Successivecolumnslisttheinstrument,dateandJDofobservation,exposuretime,
estimatedsignal-to-noiseratio(SNR)per1.8kms−1velocitybinat5000Å,spectralresolution(R)andwavelengthcoverage.
Instrument Date JD t (s) SNR R λ(Å)
exp
(DD-MM-YYYY) (2450000+)
UVES 01-05-2001 2031.464 145 254 65030 3043–3916
2032.464 70 252 74450 4726–6808
2031.466 70 266 74450 4726–6808
UVES 01-08-2001 2038.441 100 379 65030 3731–4999
2038.443 100 383 65030 3731–4999
2038.441 100 101 74450 6650–10426
2038.443 100 101 74450 6650–10426
UVES 03-12-2005 3707.846 200 233 71050 3044–3917
3707.841 200 316 71050 3281–4563
3707.846 200 675 107200 4726–6835
3707.841 200 283 107200 5708–9464
ESPaDOnS 09-01-2006 3745.167 1200 270 65000 3690–10481
HARPSpol 24-05-2009 4975.546 600 311 115000 3780–6910
HARPSpol 25-05-2009 4976.536 1200 398 115000 3780–6910
HARPSpol 31-05-2009 4982.605 1200 476 115000 3780–6910
HARPSpol 05-01-2010 5201.833 808 397 115000 3780–6910
HARPSpol 19-05-2011 5701.450 800 392 115000 3780–6910
HARPSpol 20-05-2011 5702.449 800 427 115000 3780–6910
HARPSpol 01-04-2012 6018.559 1000 670 115000 3780–6910
HARPSpol 28-04-2014 6775.610 360 320 115000 3780–6910
3. PhysicalParameters 3.0 ± 0.20 M . This is in agreement to the mass of 3.51 ±
⊙
0.64M thatwecanestimatefromthestellarradiusandphoto-
⊙
3.1.EffectiveTemperatureandGravity
metricallydeterminedlog g.Further,thepositionintheHRdia-
gramsuggeststhatHD94660hascompletedlessthanhalfofits
Geneva and Stro¨mgren uvbyβ photometry are available for
mainsequencelifetime.ThisisconsistentwithLandstreetetal.
HD94660andbothwereutilisedtodeterminetheeffectivetem-
(2008),whofindstrongmagneticfieldsonlyinyoungApstars.
peratureT andgravitylog gforthestar.FortheGenevapho-
eff
tometry, we used the fortran code developed by Kunzlietal.
(1997).AmodifiedversionoftheNapiwotzkietal.(1993)code,
4. MagneticField
that corrects the effective temperature to the Ap temperature
scale (see Landstreetetal., 2007, for a complete discussion),
wasusedfortheStro¨mgrenphotometry.
From the Geneva and Stro¨mgren photometric systems, we
4.1.LongitudinalMagneticFieldMeasurements
foundT =11500K,log g=4.10andT =11100K,log g=
eff eff
4.26respectively.Weadoptthemeanofthesetwosetsofvalues ForeachoftheStokesVspectrawemeasuredthemean,surface-
for our analysis with Teff = 11300 ± 400 K and log g = 4.18 averaged longitudinal magnetic field (hBzi) from line-averaged
±0.2,withtheuncertaintiesestimatedfromthescatterbetween Least-SquaresDeconvolved(LSD; Donatietal.,1997)linepro-
the measurements,and by taking into accountthe intrinsic un- files. The technique involves obtaining mean line profiles by
certainties in computing these parameters for Ap stars. These combiningallspectrallinesinagivenlinelist(normallymetallic
parametersareusedforourabundanceanalysis(seeSect.6). andHelines).TheresultisamuchhigherSNRwiththeability
todetectweakerZeemansignaturesduetomagneticfields.hB i
z
was measured using the first-order moment method discussed
3.2.Luminosity,StellarRadiusandMass
by Rees&Semel (1979), and as implemented by Donatietal.
vanLeeuwen(2007)reportsaHipparcosparallaxof6.67±0.80 (1997)andWadeetal.(2000)accordingtotheequation:
milliarcseconds. From this value, a distance to HD 94660 of
about150pcisdeduced.Withawelldetermineddistance,anap- −2.14×1011R (v−v0)V(v)dv
hB i= . (1)
propriatebolometriccorrectionforApstarscanbeusedtodeter- z λzc R [1−I(v)]dv
minethestellarluminosity(seeLandstreetetal.,2007),whichis
foundtobelogL/L⊙ =1.97±0.25withuncertaintiespropagated In this equation, v is the velocity within the LSD profile, V(v)
intheusualway. is the continuum-normalised Stokes I profile and I(v) is the
Based on the T and luminosity determinations, it is continuum-normalised intensity profile. The wavelength λ (in
eff
straightforwardtocomputeastellarradiusofR=2.53±0.37R . nm)andLande´factorzcorrespondtotheweightingfactorsused
⊙
By further comparing the position of HD 94660 to theoreti- inthecalculationoftheLSD profiles(500-nmand1.2,respec-
calevolutionarytracks(Girardietal., 2000)in anHRdiagram, tively).
we are able to estimate the star’s evolutionarymass. Using the The LSD profiles were extracted using the iLSD code
adopteduncertaintiesinT andL,weproceedbycomparingthe of Kochukhovetal. (2010). As input the code requires a
eff
star’spositiontomultipleevolutionarytracksofvaryingmasses. line mask that was extracted from the Vienna Atomic Line
Inthismanner,weestimatethatHD94660hasamassofabout Database(VALD;Kupkaetal.,2000;Ryabchikovaetal.,1997;
4 J.D.Baileyetal.:TheMagneticStandardHD94660
Full Mask
Nd Mask
Balmer lines
-1750
-2000
G)
B ( Z-2250
-2500
-2750
6800 Fig.2. The variations in magnetic field strength hB i versus
z
EQWforindividualspectrallines.Differentsymbolsdenotedif-
6600
ferent elements, while different colours are used to represent
different Lande´ factor ranges. Note that the absolute value of
6400
thefieldstrengthisshownandeachmeasurementhasbeennor-
6200 malisedasdiscussedinthetext.Thedashedlineisalinearfitto
G)
B > (6000 thedata.
<
5800
of interest and the other containingall other lines. Both masks
5600 areusedsimultaneouslyasinputandallowustoextractarepre-
Fe II - 6149 Å sentativemean,unblendedprofileoftheelementofinterest.All
5400 CNrd IIII I- -6 863134 5Å Å LSDprofileswereextractedontoa1.8kms−1 velocitygridand
uncertaintieswerecomputedbypropagatingtheuncertaintiesin
5200
0 1 thefinalLSDprofiles.
Phase
OurfinalmeasurementsarelistedinTable3.Thetoppanel
Fig.1.hB iandhBivariationsofHD94660.Thesolidblacklines ofFigure1plotsthehB imeasurementsagainstrotationalphase
z z
aretheadoptedmagneticfield modelvariationsfromSect. 4.3. forthefullmetallicandNdmasksaswellasthemagneticfield
Top:hB iLSDmeasurementsusingthefullmask(blackcircles) modelderivedinSect.4.3.AlsoshownaretheBalmerlinemea-
z
and Nd mask (green triangles). Note that the open and filled surementsof Landstreetetal. (2014). Althoughthe HARPSpol
symbols denote measurements from the ESPaDOnS spectrum andESPaDOnSmagneticdatahavenotbeenintercalibratedfor
andHARPSpolspectra,respectively.Balmerlinemeasurements this star, we note that the consistency between the ESPaDOnS
fromLandstreetetal.(2014)arethefilledredsquares.Bottom: andHARPSpolmeasurementsissatisfactory.Furthermore,pre-
Phased hBi measurementsfromFe ii 6149(bluesquares), Cr ii vious studies have shown a good agreement between the po-
6833(redtriangles)andNd iii6145(greencircles).Thedotted larimetric spectra acquired with ESPaDOnS and HARPSpol
greenlineisthebest-fitsinusoidalvariationstotheNdmeasure- (Piskunovetal.,2011).
ments.
4.1.1. Measurementsfromindividualspectrallines
Piskunovetal.,1995;Kupkaetal.,1999)forthespectralrange
coveredby HARPS, using the mean abundancesdeterminedin
this work and discussed in Sect. 6. This linelist was used for The extremely sharp lines and strong magnetic field result in
both the HARPS and ESPaDOnS spectra to provide a consis- clean, strong Stokes V signatures for individual spectral lines.
tent hB i measurementbetween the two datasets. We then pro- This allows us to test the multi-line technique of LSD (see
z
ceededtoremovealllinesthatwereblendedwithlinesnotused Sect. 4.1)bymeasuringhB ifromindividuallinesofelements.
z
in this analysis, such as broad hydrogenlines or lines blended We performed these measurements for a total of about 80 un-
with strong telluric absorption bands. As a final step we then blendedlineswhichincludedtheelementsSi,Ti,Cr,FeandNd
automatically adjust the line depths from their theoretical pre- (thesameelementsforwhichweperformedLSDtocomputethe
dictionstoprovideabestfittotheobservedStokesI spectrum, field from individual elements). In general, the measurements
whilealsoremovinglinesthatarepoorlyfit(suchasthoselines agreewiththevalueswederiveandpresentinTable3,butwith
that show very strong Zeeman splitting). This final mask was large scatter between different lines of an individual element.
then used to extract mean line profiles for all spectra that we In an attempt to understand this scatter, we plot the measured
labelas‘fullmask’.Wealsoextractedmeanlineprofilesofindi- hB ivalueagainsttheequivalentwidth(EQW)forallthespec-
z
vidualelementsforSi,Ti,Cr,Fe,andNd,usingthemulti-profile tral lines (see Fig. 2). To account for any variation in the field
capabilitiesofiLSD.Ineachcaseweprovidedtwoinputmasks measurementsdue to the differentphases, all values have been
based on our final full mask, one made entirely of the element normalisedbythehB iandEQWvaluefortheFei4982lineob-
z
J.D.Baileyetal.:TheMagneticStandardHD94660 5
tainedforagivennight.Notethattheabsolutevalueofthefield where ∆λ denotesthe shift of the σ componentsfromthe zero
strengthisshownforclarity. fieldwavelength,λ istherestwavelengthinÅ andzistheef-
o
The results show a clear trend, namely a stronger field fectiveLande´ factorofthe line(1.00,1.34and1.33forλ6145,
strengthforlineswithsmallerEQWandaweakerfieldstrength 6149and6833,respectively).Theseparationoftheelementsof
forlineswithlargerEQW.Wenotethatalmostalloftheoutliers theZeemansplitlinesthatdetermine∆λwasmeasuredbyfitting
fromthistrendwerelineswithverysmallLande´ factors.These GaussianstoeachcomponentusingthesplotfunctioninIRAF.
resultsaredifficulttointerpret.Ononehand,thisresultcouldre- Theuncertaintieswere estimatedby consideringthe dispersion
flectdifferentialdesaturationi.e.thatstronglinesshouldbemore betweenaseriesofmeasurementsofhBifrommultipleGaussian
sensitivetothedesaturationofthelineduetomagneticsplitting fits to each component. Baileyetal. (2011) note that measure-
and so the relative change in the EQW of that line should be ments of hBi from Eqn. 2 are very sensitive to small changes
greaterthanforaweaklineinthepresenceofastrongmagnetic in∆λ.Becauseofthehighresolutionandessentiallyzerorota-
field. We attempted to test this hypothesisby producinga syn- tion rate of HD 94660, the dispersion in our measurements of
thetic spectrum using zeeman (see Sect. 5), which includes all ∆λislessthanabout0.005Åinmostcases.Table4summarises
ofthemeasuredspectrallines.Becausethefieldstrengthiskept thesehBimeasurementsandthebottompanelofFig.1showthe
constantforagivenphaseinzeemanforalllines,anysystematic phasedrotationalvariationsof the surface fieldsfromthe three
changeinthemeasuredhBzivalueasafunctionoftheEQWcan elements.
beattributedtodifferentialsaturation.Unfortunately,theresults The scatter of hBi values around the mean curve confirms
wederivedfromthistestarealsodifficulttointerpret,butitap- thatouruncertaintyestimatesarereasonable.Theconsistencyof
pears that the general trend of the synthetic tests do not agree measurementsfromeachindividualelementtakenonsuccessive
with the results fromthe observations.Anotherpossible expla- nights(24-05-2009and25-05-2009)andtheagreement,within
nation caused by differential saturation is that this result could estimated uncertainties, between hBi measurements of the Fe-
simplyreflectsmallsystematicchangesinthemeasuredcentre- peakelementsCriiandFeii(exceptforasinglemeasurement)
of-gravityofStokesV(thenumeratorinEqn.1).Thiswouldpre- further suggests that the quoted uncertainties are realistic. The
dominantly affect strong lines with complex splitting patterns, surfacefieldappearstovarysinusoidally;however,theaverage
resultinginablendedprofilethatdoesnothavethesamecentre- surfacefieldfromtheFe-peakelementsisoforder400Glarger
of-gravityastheunsaturatedlinewiththesamesplittingpattern. than that of Nd iii. Further, the peak-to-peak amplitude of the
Ontheotherhand,thisresultcouldbeunrelatedtosaturation variationsistwiceaslargeforNdiii(oforder800G)compared
effectsand couldbe a measureofthe changeofthe decreasing totheFe-peakelements(oforder400G).Thefactthatdifferent
fieldstrengthwithincreasingverticalheightintheatmosphere: elementssamplethefieldinsignificantlydifferentwayssuggest
thecoresofweaklinesareformeddeeperintheatmosphereand thattheabundancedistributionsofdifferentelements,inaddition
should have intrinsically stronger fields relative to strong lines tothefieldstructure,areinhomogeneousoverthestellarsurface.
with coresformedhigher.Using ourzeeman modelto estimate ThisisfurthersupportedbythediscrepanthB imeasurementsof
z
the depth of the atmosphere (∼0.6% of the stellar radius), we differentelementsdiscussedaboveandlistedinTable3.
compute an expected change of the order of ∼4% between the
measuredfieldstrengthoflinesformedatthetopandbottomof
theatmosphere.Fig. 2showsa typicalvariationinthe strength 4.3.MagneticFieldModel
of the magnetic field of order 30% from weak to strong lines
Given the long rotation period of HD 94660, we are fortunate
(closerto50%ifyouconsidertheextremevalues).Therefore,it
tohavesufficientphasecoverageforbothhB iandhBiinorder
seemsunlikelythatthiseffectisimportantandthenatureofthis z
toderiveasimplemagneticfieldmodel.Ingeneral,theFORS1
trendisstillunclear.
data show that the Balmer line and metallic line hB i measure-
z
mentsagreeclosely,exceptforonediscrepantmetalhB ivalue
z
(Landstreetetal.,2014).However,alargediscrepancyexistsbe-
tweenallFORS1measurementsandthosefromhigh-resolution
4.2.SurfaceMagneticFieldMeasurements spectra.Inparticular,theFORS1datashowamuchlargerrange
in hB i. This can be seen in the top panel of Fig. 1 where the
z
FORS1Balmerlinemeasurementsvarybyorder800Gwhereas
thefullLSDmaskfromthehighresolutionspectravarybyabout
The sharp-lined features and high resolving powers of the 300G. We optforusingthe Balmerline hB ivariationsasdis-
z
HARPSpol, UVES and ESPaDOnS instruments allowed us to cussedbyLandstreetetal.(2014)toderiveafieldgeometry.We
make measurements of the mean field moduli (hBi) from ob- point out that this is the main reason for the discrepancy be-
servedZeemansplittinginthelinesofNdiiiλ6145,Feiiλ6149 tween our field structure (see below) and the one derived by
andCriiλ6833.NotableexceptionsarefortheoneESPaDOnS Landstreet&Mathys(2000).
spectrumandtwonightsofUVESspectra.Thelowerresolving The fortran program fldsrch.f (Landstreet&Mathys,
powerofESPaDOnScomparedtoHARPSpoldidnotallowusto
2000)derivesamagneticfieldgeometrybasedontheobserved
seesufficientsplittinginNdiii,whereasinsufficientwavelength
hB iandhBivariations.Itproducesasimpleco-axialmultipole
z
coveragefortheUVESspectradidnotallowformeasurements
expansionthatconsistsofdipole(B ),quadrupole(B )andoc-
ofsplittinginCrii(on01-05-2011)orNdiiiandFeii(on01-08- d q
tupole (B ) components with the angles between the line-of-
oct
2011).
sightandrotationaxis(i)andthemagneticfieldandrotationaxes
To obtain values of hBi, the atomic data from the VALD ◦
(β),specified.ForHD94660,anewfieldgeometrywithi=16 ,
database was used and the field strength calculated from the
◦
equation β = 30 , Bd = −7500 G, Bq = −2000G, and Boct = 7500 G
isfound.ThephasedhB iandhBivariationsusingouradopted
z
∆λ geometry are shown in Fig. 1 and are in good agreement with
hBi = , (2)
4.67×10−13λ2z the longitudinal field variations measured by Landstreetetal.
o
6 J.D.Baileyetal.:TheMagneticStandardHD94660
Table3.Logoftheline-of-sightsurfacefield,hB i,measurementsofHD94660.Listedaretheinstrument,dateandJDofobserva-
z
tions,phaseandhB imeasurementsfromthefullstellarmaskandonlylinesofSi,Ti,Cr,FeandNd.
z
Instrument Date JD Phase hBi(G)
z
(DD-MM-YYYY) (2450000+) FullMask Si Ti Cr Fe Nd
ESPaDOnS 09-01-2006 3745.167 0.051 −2223±17 −2343±77 −1963±61 −2174±25 −2217±19 −2601±70
HARPSpol 31-05-2009 4982.605 0.493 −2287±14 −2231±71 −2341±51 −2114±22 −2329±16 −2381±64
HARPSpol 05-01-2010 5201.833 0.571 −2413±14 −2460±70 −2444±50 −2273±23 −2480±17 −2628±58
HARPSpol 19-05-2011 5701.450 0.750 −2304±14 −2515±68 −2281±51 −2106±22 −2348±17 −2739±53
HARPSpol 20-05-2011 5702.449 0.750 −2303±14 −2502±68 −2276±51 −2092±22 −2351±17 −2726±52
HARPSpol 01-04-2012 6018.559 0.863 −2252±13 −2399±65 −2352±48 −2151±22 −2280±16 −2625±49
HARPSpol 28-04-2014 6775.610 0.133 −2151±14 −2305±71 −1643±48 −2149±21 −2200±17 −2372±61
Table4.Logofsurfacefield,hBi,measurementsofHD94660.Foreachspectra,theinstrument,dateandJDofobservation,phase
andhBimeasurementsfromNdiiiλ6145,Feiiλ6149andCriiλ6833arelisted.
Instrument Date JD Phase hBi(G)
(DD-MM-YYYY) (2450000+) Ndiiiλ6145 Feiiλ6149 Criiλ6833
UVES 01-05-2001 2031.464 0.439 6124±184 6292±94 –*
UVES 01-08-2001 2038.441 0.441 –* –* 6253±120
UVES 03-12-2005 3707.841 0.038 5954±179 6711±105 6253±80
ESPaDOnS 09-01-2006 3745.167 0.051 – 6460±84 6315±102
HARPSpol 24-05-2009 4975.546 0.490 5877±129 6123±60 6208±117
HARPSpol 25-05-2009 4976.536 0.490 5868±141 6176±63 6191±77
HARPSpol 31-05-2009 4982.605 0.493 5900±113 6122±98 6221±119
HARPSpol 05-01-2010 5201.833 0.571 5881±114 6093±84 6194±147
HARPSpol 19-05-2011 5701.450 0.750 5395±71 6085±64 6044±89
HARPSpol 20-05-2011 5702.449 0.750 5549±93 6039±60 5993±88
HARPSpol 01-04-2012 6018.559 0.863 5708±83 6220±51 6244±78
HARPSpol 28-04-2014 6775.610 0.133 6209±85 6418±43 6426±60
Notes.(*)UVESspectradonotincludethesespectrallines(seeTable2)
(2014) from FORS1, as well as the measured Fe-peak surface it would require i to be within a small region of the visible
fieldvariations. hemisphere.However,theirgeometriesarebasedonlyoninten-
sityandcircularpolarisationspectrawhichmakesitdifficultto
Although our magnetic field model is based on the fit to
distinguish i and β. Fortunately, we have available two obser-
the large amplitude hB i FORS1 data, it is nevertheless useful
z
vations of the linear polarisation (Stokes Q and U) spectra of
tocompareourgeometrytotheonefromLandstreet&Mathys
HD94660takenwiththeHARPSpolinstrumenton01-04-2012
(2000) because it is not entirely clear which, if either, is bet-
(see Sect. 2). The linear polarisation data will allow us to dis-
ter to use. In general, our adopted field geometry is not dras-
tinguish which of the two angles is larger. Figure 4 compares
tically differentthanthe one proposedbyLandstreet&Mathys
the Stokes Q signature for our adopted magnetic field model,
(2000)(seeSect. 1).Theanglesi+βareroughlythe same,al-
assuming that i > β (top panel) and i < β (bottom panel). For
thoughthe individualvaluesof these anglesdiffer(recallthati
i > β, the model clearly predicts a Stokes Q signature that is
andβcanbeinterchanged).Thevaluesof B and B are sim-
d oct
toolargecomparedtotheobservations.Alternatively,fori < β,
ilar; however, the relatively small B , although comparable in
q
the model predicts a much more modest signature, in relative
magnitude,is negative in our derivedgeometry and positive in
agreementtowhatisobserved,althoughweadmitthatthemodel
thegeometryofLandstreet&Mathys(2000).Thecircularpolar-
doesnotaccuratelyrepresentthetruevariationsinlinearpolar-
isationsignaturesinStokesV arealsowellproducedwithboth
isation for this star (we note that similar results are also seen
geometries.Fig. 3 comparesthe observedand synthetic Stokes
I andV profiles,usingbothmodels,forNdiiiλ6145(notethat withStokesUandwiththeLandstreet&Mathys(2000)model).
Thisfigurestronglysuggeststhat,contrarytotheassumptionof
onlyonephaseisshownbutthesplittingisequallywellrepro-
Landstreet&Mathys (2000), for HD 94660 β is the larger of
ducedatallphases).Itisobviousfromthisfigurethatbothfields
thesetwoanglesforbothmodels.
doanadequatejobatreproducingtheobservedmagneticprop-
erties in the spectrum of HD 94660. Based on Figures 3, it is
unclearwhich modelis best and both modelsare too coarse to
describethecomplexfieldvariationsoverthevisiblestellarsur-
face.Althoughourfieldmodelispreferred,weperformanabun-
danceanalysisusingbothgeometries(seeSect.5).
Landstreet&Mathys(2000)suggestthatβislikelysmaller 5. SpectrumSynthesisProgram
thaniforaverylargefractionofslowrotators,arguingthatfor 5.1.zeeman.f
longperiodstars (P ≥ 25d)itis equallylikelyfortherotation
pole to be positioned anywhere on the visible hemisphere. For Thefortranprogramzeeman.fisaspectrumsynthesisprogram
this reason, smaller values of i are statistically unlikely, since developedbyLandstreet(1988).zeeman.fperformslineforma-
J.D.Baileyetal.:TheMagneticStandardHD94660 7
Table5.TheabundancesaremeasuredwithrespecttoH,andare
1.1
tabulated with their associated uncertainties. For reference, the
solarabundanceratioandthenumberofspectrallinesmodelled
1
arealsoshown.
0.9 Element log(N /N ) Solar #Lines
X H
y He <−2.20 −1.07 2
nsit0.8 O −3.77±0.20 −3.31 3
e
nt Mg −5.03±0.15 −4.40 1
ve I Si −3.21±0.30 −4.49 5
elati TLhMis 2 p0a0p0er Ca −6.19±0.33 −5.66 1
R0.1 Ti −5.55±0.50 −7.05 7
Cr −3.86±0.30 −6.36 30
0 Mn −5.67±0.20 −6.57 2
Fe −3.10±0.50 −4.50 28
-0.1 Co −4.51±0.25 −7.01 8
Ni −5.24±0.20 −5.78 1
-0.2 Sr −7.78±0.40 −9.13 2
6144.8 6145 6145.2 6145.4
La −7.68±0.16 −10.9 2
Wavelength (Å)
Ce −6.86±0.20 −10.42 3
Pr −6.80±0.25 −11.28 4
Fig.3. A comparison of magnetic field models. In black are
the observed Stokes I (top) and V (bottom) profiles for Nd iii Nd −6.54±0.30 −10.58 9
Eu −7.49±0.20 −11.48 3
λ6145 for the 28-08-2014 HARPSpol spectrum. We overplot
the computed line profiles for our adopted magnetic field ge-
ometry (red dashed line) and the magnetic field geometry of
Landstreet&Mathys(2000)(bluedottedline).
formabundanceisassumed.Foreachmodelledspectrum,abest-
0.04 i > β
fitvsiniandradialvelocityv arecalculated.
R
0.02
0 5.2.Choiceofmagneticfieldmodel
-0.02
Oneofthemajordifficultiesinperformingtheabundanceanal-
-0.04 ysisofamagneticstaristheinclusionoftheeffectsofthemag-
0.04 i < β neticfieldonspectrallineformation.Thisis,inpart,duetothe
limited number of tools that include this physics, but also be-
0.02 causeoftheinherentambiguityinselectinganappropriatemag-
netic field model. Section 4 highlighted this discrepancy with
0
two apparently adequate field models to explain the observed
-0.02 Stokes I andV profilesof HD 94660.Therefore,we compared
the abundances derived from these two magnetic field geome-
-0.04 tries.Wealsotestedtheabundanceswithasimpledipolarmodel
4616 4618 4620 4622 4624 4626 inwhichthe valueof B was chosento beapproximatelythree
d
Wavelength (Å) times the largest observedhB i value (this leads to roughlythe
z
sameextremaintheobservedhB iandhBivalues).
Fig.4. A comparison of the Stokes Q signatures of synthetic z
models to the HARPSpol observationon 01-04-2012.In black We notethattheagreementbetweenthedifferentmodelsis
is the observedStokes Q spectrum.Overplottedis our adopted impressive.Foranygivensetoffundamentalparameters(i.e.T
eff
magneticfield model(red dashedline). The top panelassumes andlog g),thedifferenceinthederivedabundancesbetweenthe
i>β,whereasthebottomassumesi<β. threemodelsaregenerallylessthanabout0.1dexandnoworse
than about 0.2 dex. It is encouraging that the choice of model
does not drastically influence the final abundances. In fact, a
tion and radiative transfer in a magnetic field. The input mag- verycoarsemodel,consistingonlyofadipole,cangiveaccurate
neticfieldgeometryconsistsoftheaxisymmetricsuperposition abundances,whichsuggeststhatthefinalabundancesmaynotbe
ofadipole,quadrupoleandoctupolewithspecifiedvaluesofthe verysensitivetothedifferencebetweenthesesimplemodelsand
anglesiandβ.zeeman.finterpolatesanappropriateatmospheric the real field distribution.Apparently,the inclusion a magnetic
model from a grid of solar abundance ATLAS9 models based field modelthatroughlyexplainsthe hB i and hBi variationsis
z
upontheT andlog gsupplied.Theatomicdataneededforline sufficienttoprovideanaccurateanalysis.Thislendssupportto
eff
synthesisaretakenfromtheVALDdatabase.zeeman.fassumes studies such as the one by Baileyetal. (2014) for which sim-
ahomogeneousabundancedistributionverticallythroughoutthe ple dipolar models are used to quantify the evolution of atmo-
stellaratmosphereandcomputesallfourStokesparameters.zee- spheric abundances in Ap stars. Since the abundances depend
man.fallowsuptosixabundanceringsthataresymmetricabout little onthemagneticfieldmodel,thefollowingsectionreports
themagneticaxistobespecifiedonthestellarsurface,witheach only on abundancesfoundusing the field geometrywe present
ring having equal extentin colatitude. Within each ring, a uni- inSect.4.3.
8 J.D.Baileyetal.:TheMagneticStandardHD94660
1
0.5 Cr II Cr II Cr II Fe II Fe IIFe II Fe IIFe II 6Man.1ad.nλyH5el8iln7iue6msaorfehperleiduimcteedxitsottboedtehreivsetraonngaebsutnhdealniucme.Hlineeisλin44t7h1e
Cr II Cr II spectrum,buttheyarenotdetectedintheavailablespectra.An
0 upper limit for helium is derived from these regions that is at
4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597
leastafactorof12belowthesolarabundance.Asexpectedofa
1 magneticApstar,HD94660isHe-weak.
Intensity0.5 Cr IICr II Cr II Fe II Fe II Si IISi II??Cr I Nd IIIPr IIIFe IIFe IICr I ??Nd III 6Tu.sh2ee.dOOtoxidymegrueilvnteipalemtaetan61a5b5u-n5d6a-n5c8eÅ.TihsepEreSsPenaDtiOnnaSllssppeeccttrruamanadlsios
0 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 includestheOilinesat7771-74-75Å,however,thistripletsuf-
fersfrom non-LTEeffects and is thereforenotconsidered.The
1
adoptedabundanceisabout2.5timeslessthanintheSun.
0.5 Fe I Fe IICo II Co II??Fe II Cr IIFe II Fe II Fe II Fe IIFe II Co IIFe II Cr II?? Co II Fe II??Co II 6.3.Magnesium
Fe II The only line suitable for modelling is Mg ii at 4481Å. At all
phases,thislineisreasonablywellfitwithanabundancethatis
0
5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 afactorof4belowthesolarratio.
Wavelength (Å)
Fig.5. Example of three synthesised spectral windows for the
6.4.Silicon
28-04-2014 HARPSpol spectrum of HD 94660. The observed
spectrumisinblack(solidline)andthemodelisinred(dashed SeveralcleanlinesofSiiiexistthroughoutthespectruminclud-
line).ThebottompanelillustratestheverticalstratificationofFe
ing strong lines at 5041 and 5055-56Å, as well as the weaker
intheatmosphereofHD94660,wherestronglinesofFeiiand
weaklinesofFeiarefitwellwiththeabundanceofTable5,but line at 4621 Å. When fit simultaneously, λ5041, 5055-56 re-
the weaker lines of Fe ii require an enhanced abundance to be quireanabundancethatisabout0.5dexlessthanthatofλ4621.
Table5reportstheaverageoftheseabundances.
wellmodelled.
WealsofoundalineofSiiiiat4552Å.Thislinerequiresan
abundancethatismorethan10timesthatofthemeanabundance
fromtheSiiilines.Thisisexactlywhatwasreportedforasam-
6. ElementalAbundances
ple ofmidto late B-type starsbyBailey&Landstreet(2013a),
whoarguethatthediscrepancybetweentheabundancesderived
Insignificantvariabilityinthestrengthsandshapesofmostspec-
fromthefirstandsecondionisationstatesofsiliconislikelythe
tral lines is observed among the twelve spectra available to
result of strong vertical stratification in the stellar atmosphere.
modelforHD94660.ForlinesthatexhibitstrongZeemansig-
Thedifferentabundancevaluesfoundusingstrongerandweaker
natures(forexampleNdiiiat6145Å)differencescanbeseenin lines of Si ii, and the line of Si iii, indicate that stratification is
the line structure (i.e. relative strengthsof the σ and π compo-
alsolikelyinHD94660.
nents)and,insomecases,marginalchangesinthelinestrength.
However,linedepthchangesarelimitedtolessthanabout0.05
inintensity.Therefore,althoughallspectraaremodelled,were-
6.5.Calcium
portonlyameanabundanceforeachelement.Uncertaintiesare
determinedintwoways,dependinguponthenumberofspectral Possible usefullines of calcium at 8498, 8542 and 8662 Å are
lines available to model. For elements with multiple lines, the presentintheESPaDOnSspectrum,however,theseareblended
uncertainties are estimated from the observed scatter between withthePaschenlines,whichcannotbecalculatedcorrectlywith
thebest-fitabundances.Whenonlyone(orfew)linesareavail- Zeemanatpresent.Therefore,onlyCaiiat3933Åwasusedto
abletomodel,weestimateuncertaintiesbychangingthebest-fit derive the final abundance of calcium. We note that distinctly
abundanceforeach elementuntilthe reducedχ2 deviatesfrom differentabundancesarerequiredtoadequatelyfitthewingsand
best-fitmodelsbyabout1(i.e.χ2 =χ2 +1).Theuncertaintyis core of this line (see Babel, 1992, where this effect is first ex-
best
thenthedifferencebetweenthetwoabundancesandcorresponds plained for Ca). An abundance derived from the wings of this
approximatelytoa1σuncertainty.Table5liststheabundances line, which are formed deeper in the atmosphere, suggest an
of each element, as well as the total number of lines modelled abundance that is approximately solar. On the other hand, the
for each element. For reference,the solar abundanceratios are core (formed higher up in the atmosphere) requires an abun-
provided(Asplundetal., 2009). Fig. 5 providesan example of dancethatisoforder0.8dexbelowthesolarabundance.Thisis
the qualityof fit for three spectralwindowsfor the HARPSpol asymptomofverticalstratificationanditwouldappearcalcium
observationtakenonApril28,2014.Below,eachelementisdis- is strongly stratified throughout the atmosphere of HD 94660.
cussedindividually. ThemeanabundanceofcalciumisreportedinTable5.
J.D.Baileyetal.:TheMagneticStandardHD94660 9
6.6.Titanium and 6137 Å were also modelled and required a similar abun-
dancetowhatisrequiredtofitthestrongerFeiilines.Itisclear
For derivingthe mean abundanceof titanium,there are several
that Fe is at least of order 30 times more abundantthan in the
linestochoosethroughoutthebluespectrum.Wehavemodelled
Sun.
thelinesofTiiiat4533,4549,4563,4568and4571Å,andalso
at4798and4805Å.Eachsetoflinesissynthesisedseparately
andwithin each set the lines are modelledsimultaneously.The 6.10.Cobalt
finalabundanceisthemeanofthesetwovalues(seeTable5) ThespectrumappearsextraordinarilyrichinCoii.Severallines
The abundancethatfitswell theweakerlinesofλ4568and of Co ii are scattered throughoutthe spectrum.A total of eight
4798,as wellasthe weak linesthatare blendedwith otherFe-
linesweremodelled:5016,5017,5023,5025,5027,5129,5135,
peakelementsatλ4533and4549,issystematicallytoolargefor 5131Å. All lines of Co ii are reasonablywellmodelledwith a
the strongerlines atλ4563,4571and 4805.Thisis a symptom
uniformabundanceandsuggestanabundancethatisatleast300
ofverticalstratification(Bagnuloetal.,2001).Nevertheless,Ti
timesgreaterthanthatoftheSun.
isclearlyoverabundantcomparedtotheSunbyaboutafactorof
ThedetectionofcobaltinthespectrumofahotApstarisrare
30.
(we note that it is more commonin the cooler roAp stars such
as10Aquilae;Nesvaciletal.,2013).Weareonlyawareoftwo
otherCo-stronghotmagneticApstars(HR1094andHR5049;
6.7.Chromium
Nielsen&Wahlgren, 2000; Dworetskyetal., 1980). The abun-
UnliketheotherFe-peakelements,theabundancederivedforCr dances for both these Co-strong stars were derived without in-
is not as drastically dependentupon the lines modelled, which cludingtheeffectsofthemagneticfieldandsuggestabundances
suggests that Cr may not be significantly stratified throughout ofthiselementgreaterthan1000timesthatoftheSun.Itisun-
the stellar atmosphere. This is illustrated in Fig. 5 where most clear how prevalent Co is in the atmospheres of magnetic Ap
lines of Cr ii (both strong and weak), as well as two lines of stars,whichemphasisestheimportanceoffurtherdetailedanal-
Cr i, are fit well with the same uniform abundance. Note that ysesofothermagneticApstars.
HD 94660 is relatively hot for the presence of neutral atoms,
however,theselinesareunambiguouslydetectedinallmodelled
spectraandarepredictedtobevisiblebyzeeman.f.Furthermore, 6.11.Nickel
testsofthedepthofcoreformationfortheselinesindicatethat Niiiat4067Åwasusedtoderivethemeanabundance.Theline
theyareformedintheupperpartofthestellaratmosphere,where is reasonably well modelled in all the spectra and Ni is appar-
the temperature is necessarily cooler and consistent with the ently overabundantcomparedto the solar abundanceratio by a
presence of neutral chromium. However, over 30 lines of Cr ii factorofabout3.
were tested throughout the spectrum with this uniform abun-
dance, and a notable fraction of weaker lines require a larger
abundance than is recorded in Table 5 to be adequately mod- 6.12.Strontium
elled.Althoughapparentlynotas significantasotherelements, Sr ii lines at 4077 and 4215 Å are present in the blue spec-
stratificationmaystillbepresentforCr.Thefinalabundanceof
trumatallphases.Theaccuratemodellingoftheselinesdepends
CrwasfoundbysimultaneouslyfittingCriilinesat4531,4539, stronglyonthe adoptedabundanceof Cr ii with which the two
4555,4558,4565,4587,4588,4590and4592Å.Atallphases, Srlinesareblended.SincetheblendingCrlinesareweaklines
Crismodelledwellwithanabundancethatisoforder300times in the wings of the Sr lines, we use a somewhat larger abun-
largerthanthesolarratio. danceforCrthanthevaluelistedinTable5whenmodellingthe
lines of strontium. In this way, we get more concordantresults
betweenthesetwolinesofSrii,buttheoverallagreementisstill
6.8.Manganese
poor.Nevertheless,it is clearthatSr isoverabundant,probably
TheabundanceofMnwasderivedfromthepairofMniilinesat byafactorofabout25comparedtotheSun.
6122Å.Thelinesarewellmodelledatallphaseswithauniform
abundance that is of order 8 times larger than the solar abun-
6.13.Lanthanum
dance.
TwolinesofLaiithataresuitableformodellingwerefoundin
the spectrum at 4605 and 6126 Å. The abundance was derived
6.9.Iron
fromthe formerand tested using the latter. Thisrare-earthele-
mentisover2000timesmoreabundantthanintheSun.
Onemeanabundanceisnotadequatetosatisfactorilymodelall
the spectrallines of Fe. It is clear from Fig. 5 that the adopted
abundance fits the weaker lines well, but is systematically too
6.14.Cerium
largeforthe strongerFe ii lines. Thisis evidentthroughoutthe
spectrum,withwellover20Felinesmodelledwiththeuniform SeverallinesofCeiiareavailabletomodelthroughoutthespec-
abundancelistedinTable5.Thisissymptomaticofverticalstrat- trumincluding4560,4562and4628Å.Thefinalabundancewas
ification in the atmosphere. Multiple lines of Fe ii are used to foundfrommodellingλ4628,whichsatisfactorilyfitstheother
compute the mean abundance of iron and the larger estimated lines of Ce. The models suggest that this rare-earth element is
uncertaintyis indicativeof the discrepanciesfoundwhenmod- morethan3000timesthesolarabundanceratio.
elling each line. The weaker Fe ii lines at 5029, 5031, 5032, Interestingly,unlikewhatwasfoundbyBailey&Landstreet
5034, 5035 and 5036 were fit simultaneously, and required an (2013b) for HD 147010, there is no discrepancy between the
abundancelargerby0.6-0.7dexthanthatdeducedusingstrong abundances derived from 4560-62 Å compared to 4628 Å and
linessuchas4583and5018Å.ThreelinesofFeiat5014,6136 suggeststhatperhapsthediscrepancytheyreportbetweenthese
10 J.D.Baileyetal.:TheMagneticStandardHD94660
lines for that star may not be due to inaccurate gf values, but shifts. Therefore, the polarimetric spectra afford us the oppor-
insteadpossiblydueto anunrecognisedblendorblendsinthis tunity to test for the presence of short-periodvariations. To do
morerapidlyrotatingmagneticstar(whichhasavsiniofabout so, we compared a synthetic model1, which takes into account
15kms−1). the predicted radial velocity variations according to the ∼0.5d
period solution, to the mean LSD profiles extracted from the
ESPaDOnSspectrumtotestifnullsignaturesshouldbepresent.
6.15.Praseodymium
The ESPaDOnSspectrum was obtained with the longest expo-
LinesofPriiiat4625,6160and6161Åareavailabletomodel sure time, and therefore should show the largest effect due to
in all spectra. For the ESPaDOnS spectrum,we were also able velocityshifts.Becauseofthestrongpolarisationsignal,ourre-
tomodelPriiiat7781Å.Littletonovariationisobservedinthe sultsshowthatwhenradialvelocityshiftsareaddedtoeachin-
dividualsub-exposure(inagreementwiththeexpectedvariation
strengthofthesespectrallinesandauniformabundancemodels
suggested by short-period variations), the resulting null profile
theobservedspectrumwellatallphases.Similarlytotheother
should also show a strongsignature, which is easily detectable
rare-earthelements,Prisdramaticallymoreabundantthaninthe
inthemeanLSDprofile.Whilethistestcannotdefinitivelyrule
Sun,byafactoroforder3000.
outthepossibilityofshort-termradialvelocityvariations,itdoes
providecompellingevidenceagainstit.Thisresultsuggeststhat
6.16.Neodymium the∼0.5dperiodis probablyanaliasofthe correctperiod.We
tested the plausibility of this hypothesis by subtracting a sinu-
Nd has the richest spectrum of all the rare-earthelements with
soidalfittotheRVvariationscorrespondingtothe∼840dperiod
multiple lines of Nd iii available for modelling: 4570, 4625,
plusitsfirstharmonic,andrecomputedtheperiodogramonthe
4627, 4911, 4912, 4914, 5050, 5127 and 6145 Å. In general,
residuals. The resulting periodogramno longer shows any sig-
abundancesderivedfromeachlineagreewellwithoneanother
nificantpowerabout0.5d,demonstratingittobeanaliasofthe
andarereasonablywellmodelledatallphases.Ndhasthehigh-
∼840dperiod.
est abundance of any rare-earth element studied here, being of
If the radial velocitiesdo vary with the ∼840d period,it is
order104timesthesolarvalue.
ourconclusionthatthesevariationsarelikelytheresultofbina-
rity.Ifthesevariationswereduetoshiftsinthecentre-of-gravity
6.17.Europium ofthelineprofileduetotheinhomogeneoussurfacedistribution
(e.g.spots)thatiscommonamongAp/Bpstars,thenwewould
Two suitablelines ofEuii are presentformodelling:4129and
expectmuchsmallerradialvelocityshifts,whichreflecttheline
6645Å.OnelineofEuiiiisalsoavailableat6666Å,however, distortion, with a maximum velocity range of the order of the
thislineisbadlyblendedwithFe-peakelements.Thisrare-earth linewidth.Aswell,wewouldexpectperiodicityconsistentwith
elementisabout1000timesmoreabundantthanintheSun. therotationalperiodoroneofitsharmonics.
If we adopt this period as the orbital period then we ob-
tain the orbitalsolution givenin Table 7 using version1.0.2of
7. RadialVelocityVariations
theprogramlosp(Lie`geOrbitalSolutionPackage;Rauwetal.,
In the process of performing our spectroscopic analysis, it be- 2000). Figure 7 displays the RV orbitalcurve for this long pe-
cameevidentthatHD94660exhibitssignificantradialvelocity riod.Of particularinterest, thissolutionsuggestsa higheccen-
variations.We thusreportsubstantialradialvelocity(RV) vari- tricity ∼0.4 and a high mass-function f(m) = m3sin3i/(M +
ationsinthemagneticstandardstarHD 94660,witha rangein m)2 ∼ 0.4 (where M is the mass of the observedstar, m is the
velocitiesoforder35kms−1.Thesevariationswerefirstreported mass of the unseen companionand i is the orbital inclination).
byMathysetal.(1997),whonotedthattheorbitalperiodshould Using the 1σ limits of our estimated mass of HD 94660, this
not be more than about 2 years, significantly shorter than the mass-functionimpliesa lowermasslimitforthecompanionof
rotation period. Some years later, Mathys (2013) determined a &2 M⊙. However,after co-aligningour spectra we find no evi-
value of the orbital period for HD 94660 of 848.96 days. Our dencetosuggestthepresenceofanyadditionalspectralfeatures
search for the best-fit period of these radial velocity variations thatarenotassociatedwithHD94660,whichshouldbevisible
was carried out using the Lomb-Scargle method (Pressetal., if HD 94660 hosted a ∼2 M⊙ MS companion. Given the rela-
1992).Themostsignificantfrequenciesin theperiodogramare tively high mass-function permitted by our preliminary orbital
located at periods ∼0.5d and ∼840d, as shown in Fig. 6. The solution, the most likely candidate is a high-mass neutron star
time series of observations is insufficient to provide a unique orblackhole;however,furtherdataisrequiredtoconstrainthe
periodandmoreobservationsarerequiredinordertobettercon- periodandverifytheorbitalsolutionbeforeanydefinitivecon-
strainanyperiodicbehaviour;however,weverifythatthemea- clusions can be made. Furthermore, our solution does not rule
surements phase well with these periods, and do not appear to outthe possibilityofa hierarchicalsystem,wherethe compan-
vary in a coherent way when phased with other periods corre- ion is a combination of several objects with a total combined
spondingtolowerpeaksintheperiodogram. massof&2M⊙ (suchastwowhitedwarfs).
Duetothelimitedtemporalsamplingofourdatasetandthe
precisionofourmeasurements,itisdifficulttodirectlymeasure
the expected radial velocity differences from spectra taken on
the same night. Therefore,it is not possible to verify the plau-
sibility of the ∼0.5d period solution. However, as shown by
Neineretal. (2012), rapid radial velocity variations that occur
overthetimescaleofasinglepolarimetricsequencecaninduce 1 This model is constructed by producing a sequence of individ-
detectablesignaturesinthediagnosticnullprofiles.Thesesigna- ual synthetic sub-exposures with different radial velocities and treat-
turesresultfromresidualpolarisationsignalsthatwerenotprop- ingtheminthesamefashionastheobservationsusingthedouble-ratio
erly cancelled during processing because of the radial velocity method(Donatietal.,1997).