Short-term spectroscopic variability of Plaskett's star
AAstronomy & Astrophysics manuscript no. Plaskett c (cid:13)
ESO 2018July 28, 2018
Short-term spectroscopic variability of Plaskett’s star (cid:63)
M. Palate and G. Rauw
Institut d’Astrophysique et de Géophysique, Université de Liège, Bât. B5c, Allée du 6 Août 17, 4000 Liège, BelgiumReceived 10 February 2014 / Accepted 30 September 2014
ABSTRACT
Context.
Plaskett’s star (HD 47129) is a very massive O-star binary in a post Roche-lobe overflow stage.
CoRoT observations of thissystem revealed photometric variability with a number of frequencies.
Aims.
The aim of this paper is to characterize the variations in spectroscopy and investigate their origin.
Methods.
To sample its short-term variability, HD 47129 was intensively monitored during two spectroscopic campaigns of six nightseach. The spectra were disentangled and Fourier analyses were performed to determine possible periodicities and to investigate thewavelength dependence of the phase constant and the amplitude of the periodicities.
Results.
Complex line profile variations are observed. Frequencies near 1.65, 0.82, and 0.37 d − are detected consistently in theHe i λ ii λ iii λ Conclusions.
Whilst all three scenarios have their strengths, none of them can currently account for all the observed properties of theline profile variations.
Key words. stars: early-type – stars: oscillations – stars: individual: HD 47129 – binaries: general
1. Introduction
The massive binary nature of HD 47129 was first reported byPlaskett (1922), who inferred a total mass of 138.9 M (cid:12) , the high-est ever observed at that time. He described the binary as con-sisting of two early-type stars: an Oe5 primary and a faintersecondary with weak and broad lines. For almost a century,HD 47129, also known as Plaskett’s star, has been the targetof many studies, with tremendous progress in the past twentyyears (e.g. Bagnuolo et al. 1992, Wiggs & Gies 1992, Linderet al. 2006, 2008, Mahy et al. 2011, Grunhut et al. 2013a).The masses of the components have been revised downwardsto m sin i = . m sin i = . M (cid:12) forthe secondary, leading to a mass ratio of about 1 . ± .
05 (Lin-der et al. 2008). Unfortunately, Plaskett’s star is a non-eclipsingbinary, preventing a precise determination of its orbital inclina-tion. The latter was estimated from polarimetry to 71 ± ◦ (Rudy& Herman 1978). Plaskett’s star is usually considered a memberof the Mon OB2 association. However, Linder et al. (2008) finda discrepancy between the luminosity and the dynamical massesof the two components of the system, which could be solved byassuming a greater distance for the star.In many respects, HD 47129 o ff ers a textbook example ofevolutionary e ff ects and the e ff ects of interactions in massive bi-naries. Linder et al. (2008) used a disentangling technique, basedon the method of González & Levato (2006), on high-resolutionoptical spectra to derive the spectral types O8 III / I and O7.5 V / IIIfor the primary and secondary stars, respectively. The broad andshallow absorption lines of the secondary star suggest that thisstar rotates rapidly ( v sin i ∼ − ), whereas the much (cid:63) Based on observations collected at the Observatoire de HauteProvence (OHP, France). sharper absorption lines of the primary indicate a projected rota-tional velocity of v sin i ∼ − (Linder et al. 2008). Usingthe CMFGEN model atmosphere code (Hillier & Miller 1998) toanalyse the disentangled spectra, Linder et al. (2008) show thatthe atmosphere of the primary has an enhanced N and He abun-dance and a depletion of C. The secondary atmosphere is possi-bly depleted in N. These results indicate that the binary system isin a post-Roche lobe-overflow evolutionary stage where matterand angular momentum have been transferred from the primaryto the secondary.Because of its high rotation speed, the wind of the secondarystar could be rotationally flattened. Such a situation would af-fect the properties of the wind interaction zone in this binary.This was confirmed by the studies of the H α emission regionby Wiggs & Gies (1992) and Linder et al. (2008), as well asby the study of the X-ray emission of the system by Linder et al.(2006). An alternative explanation for the origin of the equatorialwind of the secondary could be magnetic confinement. Indeed,Grunhut et al. (2013a) have recently reported the presence of amagnetic field in Plaskett’s star. In the framework of the mag-netism in massive stars (MiMeS) survey, they detected Zeemansignatures from the rapidly rotating secondary in high-resolutionStokes V spectra. The strength of the highly organized field hasto be at least 2850 ±
500 G, and variations compatible with rota-tional modulation of an oblique field were also found (Grunhutet al. 2013a).Mahy et al. (2011) studied the CoRoT (Convection, Rotation,and planetary Transits, Baglin et al. 2006; Auvergne et al. 2009)light curve of Plaskett’s star and extracted 43 significant fre-quencies. Among these frequencies there are three major groups:0.823 d − and six harmonics, the orbital frequency 0.069 d − andtwo harmonics, as well as two frequencies at 0.368 d − and 0.399 Article number, page 1 of 9 a r X i v : . [ a s t r o - ph . S R ] O c t & A proofs: manuscript no. Plaskett
Table 1.
Summary of the time sampling of the observing campaigns.
Campaign Starting and ending dates Starting and ending n < S / N > ∆ T < ∆ t > ∆ ν nat ν max (HJD-2450000) phases (days) (hours) (days) − (days) − .
40 1 .
90 10 − .
55 2 .
38 10 − Notes.
Note that for both campaigns, there are no observations for one night due to bad weather. n is the total number of spectra, < S / N > indicatesthe mean signal-to-noise ratio, ∆ T indicates the total time between the first and last observations, ∆ t provides the average time interval betweentwo consecutive exposures of a same night. The last two columns indicate the natural width of the peaks in the periodogram ∆ ν nat = ∆ T − and ν max = (2 < ∆ t > ) − provides an indication of the highest frequency that can possibly be sampled with our time series. d − . The latter two could be related to the rotation period of thesecondary.The discovery of the 0.823 d − frequency by Mahy et al.(2011) prompted us to perform a spectroscopic monitoring ofthe system, to constrain the origin of this variation. In this pa-per, we report the results of this study. The paper is organizedas follows. Sect. 2 describes the observations, data reduction anddisentangling treatment. In Sect. 3, the line profile variations arediscussed and the results of the Fourier analyses are given. Fi-nally, in Sect. 4, we discuss the strengths and weaknesses of threepossible scenarios and summarize our results and conclusions.
2. Observations and data reduction
Spectroscopic time series of HD 47129 were obtained dur-ing two six-nights observing campaigns in December 2009and December 2010 at the Observatoire de Haute Provence(OHP, France). We used the Aurélie spectrograph fed by the1.52 m telescope. The spectrograph was equipped with a 1200lines mm − grating blazed at 5000 Å and a CCD EEV42-20 de-tector with 2048 × µ m . Our set-up coveredthe wavelength domain from 4460 to 4670 Å with a resolvingpower of 20000. Typical exposure times were 20 – 30 minutes,depending on the weather conditions, and a total of 118 spectrawere collected. To achieve the most accurate wavelength cali-bration, Th-Ar lamp exposures were taken regularly over eachobserving night (typically once every 90 minutes).The data were reduced using the MIDAS software providedby ESO. The normalization was done self-consistently using aseries of continuum windows. Table 1 provides a summary ofour observing campaigns and the characteristics of the sampling. The goal of the present paper is to study and interpret the short-term line profile variability of Plaskett’s star. Because of theDoppler shifts and line blending induced by the orbital motionin such a binary system, it is not trivial to know which star trig-gers the variations. The first step in our analysis was thus to dis-entangle the spectra of the primary and secondary components.For this purpose, we applied our code based on the González &Levato (2006) technique to the data of the two observing cam-paigns simultaneously. This method allows to derive the radialvelocities as well as the individual spectra of the components.However, because of the very broad and shallow lines of the sec-ondary, its radial velocities (RVs) obtained with the disentan-gling were sometimes not well constrained. To avoid artefacts,we thus fixed the RVs of the secondary to the theoretical val-ues derived from the orbital solution of Linder et al. (2008). The situation for the primary was better, as we recovered RVs thatwere in good agreement with the orbital solution of Linder etal. (2008): the dispersions of the RVs about the latter orbital so-lution were 11 and 6 km s − for the He i λ ii λ ff erences between thedisentangled spectrum of the primary obtained with either theprimary RVs being fixed to the orbital solution or allowing themto vary. In the subsequent analysis of the line profile variability,we thus adopted the mean spectra as reconstructed keeping theRVs of the primary fixed to the orbital solution of Linder et al.(2008). The reconstructed spectra of the individual componentsare displayed in Fig. 1. These individual spectra are important tocorrectly interpret the results of our line profile variability anal-ysis. For instance, Fig. 1 reveals that the only secondary line thathas at least some parts of its profile that are not blended withprimary lines at some orbital phases is He i λ Fig. 1.
Mean disentangled spectra of the primary and secondary (shiftedupwards by 0.1 continuum units). Important spectral lines are identifiedby the labels.
3. Variability analysis He i λ The top panels of Fig. 2 display the dynamic spectra of theobserved data in the heliocentric frame of reference (hereaftercalled the raw dynamic spectra), whilst the bottom panels illus-trate the dynamic spectra after subtracting the mean disentan-gled primary spectrum and shifting the data into the secondary’s
Article number, page 2 of 9alate and Rauw: Plaskett’s star
Wavelength (˚A) T i m e ( H J D ) Wavelength (˚A)
Wavelength (˚A) T i m e ( H J D ) Wavelength (˚A)
Fig. 2.
Line-profile variations in the He i λ frame of reference (hereafter called the secondary dynamic spec-tra).The raw dynamic spectra are dominated by the orbital mo-tion of the relatively sharp primary line. These figures showwhich parts of the core of the secondary absorption are most a ff ected by blends with the primary. We note that the observedposition of the primary line slightly deviates from the positionexpected from the orbital solution. Another feature that can beseen on the first and third night of the 2009 campaign as wellas on the third night of the 2010 data is a discrete depression Article number, page 3 of 9 & A proofs: manuscript no. Plaskett
Wavelength (˚A) F r e q u e n c y ( d − ) Wavelength (˚A) F r e q u e n c y ( d − ) −3 Fig. 3.
Fourier power spectra of the2009 (left) and 2010 (right) time se-ries of the secondary’s He i λ − ), whilst the second and third pan-els illustrate the mean periodogram overthe emission humps, respectively beforeand after prewhitening (see text). Thespectral window is shown in the lowerpanel. that moves from the blue to the red wing of the broad secondaryabsorption.The secondary dynamic spectra reveal another feature thatis more di ffi cult to distinguish on the raw dynamic spectra. Thestrengths of the blue (4462 – 4465 Å) and red (4475 – 4480 Å)emission humps that flank the secondary’s absorption (see Fig. 1)undergo a strong modulation. The latter seems correlated withthe presence of the discrete depression pointed out hereabove.Indeed, when the redwards moving discrete depression is ob-served, the emission humps are at their low level. The deviationsbetween the observed and the theoretical positions and shapes ofthe primary profile lead to residuals when subtracting the meanprimary spectrum that contaminate the secondary dynamic spec-tra between about 4466 and 4475 Å.The dynamic spectra reveal that the variations extend over awider wavelength range than the widths of the primary lines. Wethus conclude that the secondary very probably displays varia-tions, although, at this stage, we cannot exclude that the primaryis also variable, especially in view of the residuals after subtract-ing the mean primary profile.To further characterize the variations in the He i λ / or subtracting the spectrum of the primary, re-sults in power spectra dominated by low frequencies due to theorbital motion of the primary and, to a lesser extent, secondarylines. To get rid of these orbital frequencies and analyse the gen-uine short-term line profile variations, we have subtracted the mean disentangled spectrum of the primary shifted to the ap-propriate RV from the observed spectra. The resulting di ff erencespectra were then further shifted into the secondary’s frame ofreference. Therefore, we obtained the “individual spectra” of thesecondary in the frame of reference of this star for each observa-tion. The resulting 2D periodograms of the corresponding 2009and 2010 time series are shown in Fig. 3.To start, we focus on the emission humps (4462 – 4465 &4475 – 4480 Å). We find that the mean periodogram of each ob-serving campaign is dominated by a frequency near 1.65 d − andits aliases. The error on the centre of a peak in the periodogramcan be estimated as ∆ ν = . / ∆ T . Here ∆ T is the time inter-val between the first and the last observations of the time-series(see Table 1) and we assume that the uncertainty on the peakfrequency amounts to 10% of the peak width. The uncertaintiesare of 0.02 d − for the campaigns separately and of 3 × − d − when both campaigns are considered simultaneously . However,in the present case, these estimates are quite optimistic becausethey neither account for the signal to noise ratio of the obser-vations nor for the impact of the treatment of the data prior tothe Fourier analysis. Therefore, we estimate an overall error of ± .
05 d − on the peak values.Because of this rather large error on the position of thepeak, some caution is needed when comparing the results of ourpresent analysis with those of the CoRoT light-curve analysis ofMahy et al. (2011). The latter authors reported a large number This corresponds to an uncertainty on the period of ∆ P = . × P ∆ T . We caution though that the sampling of the combined dataset is veryodd, leading to a large number of closely spaced aliases, and the e ff ec-tive uncertainties on the peak position are thus significantly larger thanestimated here.Article number, page 4 of 9alate and Rauw: Plaskett’s star of frequencies and some of the peaks in our periodograms couldbe blends of two or more frequencies. Indeed, our spectroscopictime series are not su ffi ciently long to distinguish closely spacedfrequencies. Moreover, our time series are heavily a ff ected by 1-day aliasing, which was absent in the CoRoT data. For instance,the 1-day aliasing introduces some coupling between frequen-cies near 0.4 d − and those near 1.6 d − . Nevertheless, we notethat our 1.65 d − peak nicely fits the 1.646 d − frequency of thestrongest signal found in the CoRoT data. Table 2 lists the fre-quencies from the
CoRoT data that are potentially consistentwith the frequencies found in the current analysis.To check whether the photometric frequencies can accountfor the periodograms of our data, we used a prewhiteningtechnique (Rauw et al. 2008). In this way, we found that the2009 periodogram can be prewhitened with a single frequency(1.64 d − ), whilst e ffi cient prewhitening of the 2010 data re-quires three frequencies (1.65, 0.82, and 0.37 d − ), which arethe three strongest, non-orbital, frequencies reported by Mahy etal. (2011). The prewhitened periodograms are shown in Fig. 3.We thus conclude that the variations in the emission humps ofthe secondary’s He i λ CoRoT photometry: 0.82 d − and itsfirst harmonic at 1.65 d − , with some possible contribution of the0.37 d − frequency.The power spectra over the absorption core are a ff ected bythe residuals from the subtraction of the primary spectrum (seeFig. 3). For the 2010 data, the corresponding mean power spec-trum features essentially the same set of frequencies as for theemission humps, although with larger residuals after prewhiten-ing. For the 2009 campaign, the power spectrum is more com-plex, showing considerable power at low frequencies. This is es-pecially the case at wavelengths between 4466 and 4471 Å be-cause the primary line remains in this wavelength domain overthe entire duration of the 2009 campaign (see Fig. 2). At longerwavelengths where the moving discrete depression is seen in thedynamic spectra, the dominant frequency is near 0.4 d − .As pointed out above, the observed wavelengths of the pri-mary line deviate from those expected from the orbital solutionof Linder et al. (2008). We thus performed a Fourier analysis ofthe He i λ − v sin i , v sin i ]. For both observing campaigns, the power spec-trum is completely explained by the 0.82 and 1.65 d − frequen-cies and their aliases. Since these are the same frequencies asthose found in the variations in the secondary’s emission humps,it seems quite possible that the variations are in fact due to thesecondary star. The variability of the He ii λ iii λλ i λ ii and N iii lines are dominated by the pri-mary’s orbital motion and do not reveal clear redwards-movingdiscrete depressions (see Fig. 4). The secondary dynamic spectraare qualitatively similar to those of He i λ ii λ − frequencies andtheir aliases (see Fig. 5). For the 2010 data, a slightly cleaner Wavelength (˚A) T i m e ( H J D ) Wavelength (˚A)
Wavelength (˚A) T i m e ( H J D ) Wavelength (˚A)
Fig. 4.
Same as Fig. 2, but for the He ii λ Table 2.
Correspondence between the best frequencies (in d − ) foundin this paper and those found by Mahy et al. (2011) in the CoRoT lightcurve.
Our Frequencies CoRoT Frequencies0 .
37 0.368 0.399 (0.650) (1.646)0 .
82 0.823 0.7991 .
65 1.646 (0.650) (0.368) (0.399)
Notes.
Photometric frequencies between brackets correspond to aliasesdue the 1-day aliasing that a ff ects our time series. prewhitening is achieved with 0.37 instead of 1.65 d − , and forthe 2009 time series, a more e ffi cient result is achieved includinganother harmonic frequency (2.47 d − ). Shifting the spectra intothe frame of reference of the primary star essentially yields thesame results.Given the nitrogen overabundance of the primary star and thepossible underabundance of the secondary reported by Linder etal. (2008), we expect a priori that the N iii lines should mainly beformed in the atmosphere of the primary star . We have analysedthe two strongest N iii lines ( λλ − frequencies and (in the 2009data) some contribution at 0.37 d − . We note however that Fig. 1 reveals some shallow N iii absorptionsalso in the secondary spectrum. Article number, page 5 of 9 & A proofs: manuscript no. Plaskett
Fig. 5.
Same as Fig. 3 but for the He ii λ − frequencies.
4. Discussion
The analysis of our spectroscopic time series of Plaskett’s star inthe previous section revealed three main frequencies 0.82, 1.65,and 0.37 d − , although the properties of the latter two are inter-dependent because of the aliasing problem. Before we discussthe possible interpretation of these results, let us first recall theingredients of Plaskett’s star as (we think) we know this system.The primary star is a rather slow rotator presenting chemical en-richment indicating that it is most probably in a post-RLOF sta-tus. The secondary is a fast rotator probably surrounded by aflattened wind. The orbit is circular and the stars currently donot fill their Roche lobes. A magnetic field, probably associatedwith the secondary, was detected. This magnetic field could beat the origin of the flattened wind structure.With these ingredients in mind, three scenarios could possi-bly explain the observed line profile variations: pulsations, tidalinteractions, and an oblique magnetic rotator. Mahy et al. (2011) tentatively assigned the 0.82 d − frequencyand its six harmonics to a low-order (2 ≤ l ≤
4) non-radialpulsation (NRP). Our Fourier analyses provide us with the am-plitude ( A ) and the phase constant ( ϕ ) of the variation as a func-tion of wavelength. These quantities can help us test the NRPscenario. Indeed, for NRPs with moderate degrees l the phaseconstant usually changes monotonically across the line profileand the di ff erence between the phase constant in the red wingand in the blue wing is directly related to the values of l and m ofthe pulsation (e.g. Schrijvers & Telting 1999, Zima et al. 2004).The amplitude and phase distributions of the 0.82 and1.64 d − frequencies of the He i λ ff ected by the residuals of the primary line, they arequite erratic and do not ressemble what one would expect fortypical NRPs.If we consider the amplitudes and phase constants in theframe of reference of the primary, the variations are less erraticand, at least in the case of the 0.82 d − frequency, are more rem-iniscent of genuine NRPs. This would imply that the primary isresponsible for the 0.82 and 1.65 d − frequencies. In the frameof reference of the primary, significant amplitude of variabilityis found far away from the core of the primary line, out to −
500 (-6.6 v sin i ) and +
400 km s − (5.3 v sin i ) away from the line cen-tre. Similar results are obtained for the He ii λ The analysis of the line profile variations in the slightly eccentricbinary system Spica (Palate et al. 2013 and references therein)has shown that tidal interactions could explain Spica’s line pro-file variations. Although Plaskett’s star is not an eccentric sys-tem, the non-synchronicity of the secondary rotation could leadto tidally induced variations such as those reported by Moreno etal. (2005). Therefore, we can wonder whether tidal interactionscould be responsible for the variations reported hereabove.We have thus tried to model Plaskett’s star with the TIDEScode (see Moreno et al. 1999, 2005, 2011) combined to CoM-BiSpeC (see Palate & Rauw 2012 and Palate et al. 2013). TheTIDES code computes the time-dependent shape of the stellarsurface for eccentric and / or asynchronous systems accountingfor centrifugal and Coriolis forces, gas pressure, viscous e ff ectsand gravitational interactions. From that, CoMBiSpeC computesthe synthetic spectra of the stars at several orbital phases.We stress that our goal here is not to fine tune the parametersthat reproduce perfectly the spectra of the components of Plas-kett’s star and their observed variations. Here, we rather wishto test whether or not tidal interactions can produce detectablevariations in the spectra of this system. Indeed, fitting the spec-tra rigorously is a challenging task because of a large number offree / unknown parameters and the non-solar composition of thestars (see Linder et al. 2008); CoMBiSpeC currently only workswith atmosphere models that have a solar composition. In thepresent paper, we have only adjusted the temperatures to repro-duce the observed strength of the He i λ ii λ − could correspond to the rotational frequency of the sec- TIDES: tidal interaction with dissipation of energy through shear.CoMBiSpeC: code of massive binary spectral computation.Article number, page 6 of 9alate and Rauw: Plaskett’s star
Fig. 6.
Amplitude and phase variations for the He i λ ν = .
82 and ν = .
65 d − . The top panels yield the mean secondary profile,whilst the amplitudes associated with the two frequencies are shown in the second panels. The two lower panels indicate the phase constants ofthese frequencies. Fig. 7.
Same as Fig. 6, but in the frame of reference of the primary. ondary. The projected equatorial rotational velocity of the sec-ondary has been evaluated to ∼
300 km s − by Linder et al. 2008.These authors also derived a log g equal to 3 . ± .
1. Adoptingan inclination of 67 ◦ as proposed by Mahy et al. (2011), andconsidering that the rotation frequency is 0.4 d − , the projectedrotational velocity of 300 km s − yields a radius R = v rot π ν rot of16.1 R (cid:12) for the secondary. The log g and minimum mass of thesecondary ( m S sin i = . M (cid:12) , Linder et al. 2008) along withthe adopted inclination of 67 ◦ (Mahy et al. 2011) rather suggesta radius of 20 . + . − . R (cid:12) . In this latter case, the projected rota-tional velocity would be equal to ∼
385 km s − . The discrepancybetween these rotational velocities can be explained by several The latter was determined from the relation g = GMR (1 − Γ ) where Γ accounts for the radiation pressure of the star. factors: the uncertainties on the inclination (Mahy et al. 2011indicated a range from 30 to 80 ◦ ), the error on the rotational ve-locity determination, on the log g , and on the frequency.Our simulations indicate that tidal interactions can producevisible variations in the lines of the secondary. Because the pri-mary is thought to be in synchronous rotation in a circular or-bit, it is expected that there is no variability of the primary dueto tidal interactions. The strength of the simulated variations iscomparable to the observed ones. Reducing the secondary radiusto 18 R (cid:12) slightly lowers the amplitude of the tidally induced vari-ations, although they remain well visible in the synthetic spec-tra. Fig. 8 displays the simulated line profile variations in thesecondary during the orbital cycle. Whilst our code cannot sim-ulate the impact of the tidal interactions on a co-rotating con-fined wind, it seems very plausible that the tidal interactions also Article number, page 7 of 9 & A proofs: manuscript no. Plaskett
Table 3.
Parameters used for simulation with the TIDES + CoMBiSpeCmodel.
Parameters Primary SecondaryPeriod (day) 14 . ◦ ) 67Mass ( M (cid:12) ) 58 . . R (cid:12) ) 20 . . (cid:39) . (cid:39) . rot (km s − ) 70 . i (km s − ) 64 . β (a) . . ν ( R (cid:12) d − ) 0 .
028 0 . ∆ R / R ) 0 .
07 0 . . . Notes. (a)
The β parameter measures the asynchronicity of the star atperiastron and is defined by β = . Pv rot R × (1 − e ) / (1 + e ) / , where v rot is theequatorial rotation velocity, R is the equilibrium radius, and e is theeccentricity. a ff ect the latter and thus extend into the emission humps. This re-sult brings up the interesting possibility that breaking tidal wavesnear the secondary’s equator could produce an enhanced mass-loss in this region (see Osaki 1999). Combined with wind con-finement due to the magnetic field (Grunhut et al. 2013a, 2013b),this e ff ect could produce the equatorial torrus of circumstellarmaterial that was advocated by Wiggs & Gies (1992) and Linderet al. (2006, 2008). The much smaller radius (10.7 R (cid:12) ) proposedby Grunhut et al. (2013a) not only fails to reproduce the log g value inferred from the model atmosphere code fits (see below),but would also considerably reduce the amplitude of surface os-cillations due to tidal interactions.We have performed a Fourier analysis of the individual sim-ulated spectra of the secondary stars. The power spectrum dis-plays a peak at the secondary’s rotation frequency imposed inthe model, but does not show peaks at the harmonic frequen-cies. Tidal interactions can thus induce profile variations, pro-vided the stellar radii are su ffi ciently large. However, the currentmodel does not explain the full frequency content of HD 47129’sspectroscopic and photometric variations. Grunhut et al. (2013a) reported on the discovery of a magneticfield in Plaskett’s star with a minimum surface dipolar strengthof 2.8 kG at the magnetic poles. Such a strong magnetic fieldcould confine the secondary’s stellar wind, thereby producingthe emission humps seen around the He i λ α emission. Using additional spectropolari-metric observations, Grunhut et al. (2013b) inferred a tilt angleof the magnetic axis of (80 ± ◦ with respect to the rotationaxis. This would result in a magnetic configuration very simi-lar to HD 57682 (Grunhut et al. 2012). One would then expect adouble-wave modulation of the strength of the emission features.Based on a periodicity search in the variations in the H α equiv-alent width and the longitudinal magnetic field, Grunhut et al. N o r m a li z e dflu x Wavelength (˚A)
Fig. 8.
Line profile variations in the synthetic spectra of the secondaryduring the orbital cycle. The spectra (separated by 0.05 in phase) havebeen shifted vertically by 0.1 continuum unit for clarity. (2013b) inferred a rotational period of 1.215 d for the secondary,corresponding to the 0.82 d − frequency.This scenario could thus explain our detection of the 0.82and 1.65 d − frequencies quite naturally in the emission humps.However, adopting 0.82 d − as the secondary’s rotational fre-quency would lead to a radius estimate from the projected rota-tional velocity that is twice smaller than what we have obtainedabove, thus enhancing the discrepancy between the radius esti-mated from v sin i and the one estimated from log g .Concerning the latter discrepancy, we note that Wiggs &Gies (1992) previously reported a 2.78 d period in the equiva-lent width of the H α emission wings . Our analysis of the He i λ iii λλ ii λ − . The strongest variations arefound for the He i line.Considering the current knowledge of the system, it is di ffi -cult to assign the spectroscopic variability to either the primaryor the secondary star. Indeed, our analysis yielded conflictingresults in this respect: whilst the velocity range over which vari-ations are detected clearly favours the secondary star, the patternof the amplitude and phase of the variations makes more sense ifthey originate in the primary star. As a result, none of the currentscenarios (pulsations, tidal interactions, magnetically confinedwinds) accounts for all the aspects of the observed variations. Acknowledgements.
We would like to thank Pr. G. Koenigsberger who providedus with the TIDES code and helped us to use it. We are grateful to the referee,Dr. Jason Grunhut who helped us improve this paper. We acknowledge supportthrough the XMM / INTEGRAL PRODEX contract (Belspo), from the Fonds deRecherche Scientifique (FRS / FNRS), as well as by the Communauté Françaisede Belgique - Action de recherche concertée - Académie Wallonie - Europe. Wiggs & Gies 1992 suggested an origin in the winds of the stars,corresponding to a frequency of 0.36 d − in good agreement with the0.37 d − rotation frequency suggested by Mahy et al. (2011). A corollarywould be that the variations in the longitudinal magnetic field presentedby Grunhut et al. (2013b) would probably have to be interpreted as dueto a more complex magnetic field than a simple dipole.Article number, page 8 of 9alate and Rauw: Plaskett’s star References