Time, spatial, and spectral resolution of the Halpha line-formation region of Deneb and Rigel with the VEGA/CHARA interferometer
Olivier Chesneau, Luc Dessart, D. Mourard, Ph. Berio, Ch. Buil, D. Bonneau, M. Borges Fernandes, J.M. Clausse, O. Delaa, A. Marcotto, A. Meilland, F. Millour, N. Nardetto, K. Perraut, A. Roussel, A. Spang, Ph. Stee, I. Tallon-Bosc, Harold A. Mcalister, T.A. Ten Brummelaar, J. Sturmann, L. Sturmann, N. Turner, C. Farrington, P.J. Goldfinger
aa r X i v : . [ a s t r o - ph . S R ] J u l Astronomy&Astrophysicsmanuscript no. AB˙Supergiants˙vResubmission˙vEnglishCorr˙Printer c (cid:13)
ESO 2018October 11, 2018
Time, spatial, and spectral resolution of the H α line-formationregion of Deneb and Rigel with the VEGA/CHARA interferometer ⋆ O. Chesneau , L. Dessart D. Mourard , Ph. B´erio , Ch. Buil , D. Bonneau ,M. Borges Fernandes , , J.M. Clausse , O. Delaa , A. Marcotto , A. Meilland , F. Millour , N. Nardetto ,K. Perraut , A. Roussel , A. Spang , P. Stee , I. Tallon-Bosc , H. McAlister , , T. ten Brummelaar , J. Sturmann , L.Sturmann , N. Turner , C. Farrington and P.J. Goldfinger UMR 6525 H. Fizeau, Univ. Nice Sophia Antipolis, CNRS, Observatoire de la Cˆote d’Azur, Av. Copernic, F-06130 Grasse, France Laboratoire d’Astrophysique de Marseille, Universit´e de Provence, CNRS, 38 rue Fr´ed´eric Joliot-Curie, F-13388 Marseille Cedex 13, France Castanet Tolosan Observatory, 6 place Clemence Isaure, 31320 Castanet Tolosan, France Max-Planck Institut f¨ur Radioastronomie, Auf dem Hugel 69, 53121, Bonn, Germany Laboratoire d’Astrophysique de Grenoble (LAOG), Universit´e Joseph-Fourier, UMR 5571 CNRS, BP 53, 38041 Grenoble Cedex 09, France Univ. Lyon 1, Observatoire de Lyon, 9 avenue Charles Andr´e, Saint-Genis Laval, F-69230, France Georgia State University, P.O. Box 3969, Atlanta GA 30302-3969, USA CHARA Array, Mount Wilson Observatory, 91023 Mount Wilson CA, USA Observat´orio Nacional, Rua General Jos´e Cristino, 77, 20921-400, S˜ao Cristov˜ao, Rio de Janeiro, BrazilReceived, accepted.
ABSTRACT
Context.
BA-type supergiants are amongst the most optically-bright stars. They are observable in extragalactic environments, hence potentialaccurate distance indicators.
Aims.
An extensive record of emission activity in the H α line of the BA supergiants β Orionis (Rigel, B8Ia) and α Cygni (Deneb, A2Ia) isindicative of presence of localized time-dependent mass ejections. However, little is known about the spatial distribution of these apparentstructures. Here, we employ optical interferometry to study the H α line-formation region in these stellar environments. Methods.
High spatial- ( ∼ =
30 000) resolution observations of H α were obtained with the visible recombiner VEGAinstalled on the CHARA interferometer, using the S1S2 array-baseline (34m). Six independent observations were done on Deneb over theyears 2008 and 2009, and two on Rigel in 2009. We analyze this dataset with the 1D non-LTE radiative-transfer code CMFGEN, and assessthe impact of the wind on the visible and near-IR interferometric signatures, using both Balmer-line and continuum photons. Results.
We observe a visibility decrease in H α for both Rigel and Deneb, suggesting that the line-formation region is extended ( ∼ R ⋆ ). We observe a significant visibility decrease for Deneb in the Si ii ff erential phasefor Deneb, implying an inhomogeneous and unsteady circumstellar environment, while no such variability is seen in di ff erential visibilities.Radiative-transfer modeling of Deneb, with allowance for stellar-wind mass loss, accounts fairly well for the observed decrease in the H α visibility. Based on the observed di ff erential visibilities, we estimate that the mass-loss rate of Deneb has changed by less than 5%. Key words.
Techniques: high angular resolution – Techniques: interferometric – Stars: emission-line – Stars: mass-loss – Stars: individual(HD 197345, HD 34085) – Stars: circumstellar matter
1. Introduction
Supergiants of spectral types B and A (BA-type supergiants)are evolved massive stars of typical initial mass of 25-40 M ⊙ and high luminosity ( ∼ > L ⊙ ). Their luminosity and tem-perature place them among the visually brightest massivestars. Therefore, they are particularly interesting for extragalac-tic astronomy (Puls et al. 2008). Moreover, they represent at-tractive distance indicators by means of the identified wind- Send o ff print requests to : [email protected] ⋆ Based on observations made with the CHARA array momentum-luminosity relationship (WMLR) (Kudritzki et al.2008; Evans & Howarth 2003; Kudritzki et al. 1999). As a con-sequence, nearby BA supergiants have been analyzed withsophisticated radiative-transfer tools. Two subjects of intensescrutiny, both observationally and theoretically, have beenDeneb ( α Cygni, HD 197345, A2 Ia) and Rigel ( β Orionis,HD 34085, B8Ia) (Schiller & Przybilla 2008; Aufdenberg et al.2006b, 2002, for Deneb).Intensive spectroscopic monitoring of the activity of wind-forming lines, such as H α , suggests that the stellar-wind vari-ability of luminous hot stars is associated with localized and O. Chesneau et al.: The H α line-formation region of Deneb and Rigel co-rotating surface structures. The pioneering work of Lucy(1976) based on a period analysis of radial-velocity curvesof Deneb obtained in 1931 /
32 by Paddock (1935) revealedthe simultaneous excitation of multiple non-radial pulsations(NRPs). This e ff ort was followed by the extended opticalH eros campaigns (Kaufer 1998), monitoring for ∼
100 consec-utive nights late B-type / A-type supergiants (Kaufer et al. 1997,1996a,b) and early B-type hypergiants (Rivinius et al. 1997).Attempts to associate this activity with large-scale surfacemagnetic fields have so far been unsuccessful (Schnerr et al.2008; Verdugo et al. 2003; Henrichs et al. 2003; Bychkov et al.2003). However, the appearance of thin surface convectionzones at the supergiant stage may favor the formation and emer-gence of magnetic fields (Cantiello et al. 2009), and may alsohave a significant impact on potential NRPs.BA-type supergiants such as Rigel (B8 Ia) and Deneb(A2 Ia) exhibit observational evidence of the random andpseudo-cyclic activity of the stellar wind. A pulsation-drivingmechanism has often been proposed, although the associ-ated micro-variability is observed for a wide range of lumi-nosity and log T e ff . Gautschy (1992) argued for the presenceof strange modes. Gautschy (2009) tentatively proposed thatDeneb’s micro-variability is caused by a thin convection zonee ffi ciently trapping the non-radial oscillations, and significantprogress has been made in determining the mechanism leadingto mass ejection (Cranmer 2009). The collective e ff ect of multi-ple NRPs was also proposed as a promising means of explain-ing the large value of the macro-turbulence parameter; NRPsalso represent an attractive mechanism for the formation of Be-star disks (Aerts et al. 2009; Neiner et al. 2009).Based on intensive monitoring of spectroscopic lines, manyobservational campaigns have attempted to discriminate be-tween these models. Monitoring of Deneb and Rigel is still on-going (Danezis et al. 2009; Morrison et al. 2008; Rzaev 2008;Markova et al. 2008; Markova & Puls 2008; Morel et al. 2004;Rivinius et al. 1997; Kaufer et al. 1996b, for a non-exhaustivelist).The uniform-disk (UD) angular diameters of Deneb andRigel are 2.4 mas and 2.7 mas, respectively (Aufdenberg et al.2008; Mozurkewich et al. 2003), making them good targetsfor accurate long-baseline optical interferometry. An extensivestudy of Deneb was performed by Aufdenberg et al. (2002), us-ing di ff erent radiative-transfer models (with allowance for thepresence of a wind), and constraints from optical interferome-try data. Deneb’s wind should enhance the limb darkening rel-ative to hydrostatic models that neglect it. However, this canonly be constrained observationally by using long baselines inthe near-IR (longer than 250m) or by observing in the visible,thus relaxing the constraints on the baselines by a factor ofabout 3-4. The H α line is particularly interesting in that con-text. Its large opacity makes it very sensitive to changes at thestellar surface and above, in particular through modulations instellar-wind properties. Its extended line-formation region istherefore more easily resolved than the deeper-forming con-tinuum. Dessart & Chesneau (2002) investigated the possibil-ity of monitoring the mass-loss activity in wind-forming linesby means of optical interferometers equipped with spectral de-vices of su ffi cient dispersion. BA supergiants rotate slowly ( v sin i of about 25-40km s − ),at least relative to their terminal wind-velocity of ∼ − . Spectrally resolving the Doppler-broadenedBalmer lines thus requires a resolving power as high as R = α , which exhibitsthe distinctive P-Cygni profile morphology and has often beena crucial estimator of the mass-loss rate in spite of its doc-umented variability. While Aufdenberg et al. (2002) was suc-cessful in reproducing many observables obtained for Deneb,the modeling of the H α line proved di ffi cult. The most sig-nificant and persistent discrepancy between the synthetic andobserved profiles has been the depth of the absorption com-ponent, which is significantly weaker in the observed spec-trum (residual intensity ∼ ∼
50 km s − ,which is ∼
20% of the terminal velocity. Schiller & Przybilla(2008) used a di ff erent strategy, using the H α profile as a ref-erence for a detailed modeling of the wind characteristics, andreached a more satisfactory solution. Nevertheless, they alsoreported some remaining discrepancies in matching the H α ab-sorption, and proposed additional e ff ects that may need to beconsidered in the modeling, such as wind structure or the influ-ence of a weak magnetic field.The VEGA recombiner of the CHARA array is a recentlycommissioned facility that provides spectrally dispersed inter-ferometric observables, with a spectral resolution reaching R =
30 000, and a spatial resolution of less than one mas . The in-strument recombines currently the light from two telescopes,but 3-4 telescope recombination modes are foreseen in a nearfuture. The H α line of bright, slow rotators such as Denebor Rigel can be isolated from the continuum, and the spatialproperties of the line-forming region can thus be studied withunprecedented resolution. Using the smallest baseline of theCHARA array (baseline of 34m), we conducted a pioneeringtemporal monitoring of Deneb uncovering a high level of ac-tivity in the H α line-forming region. Rigel was also observed afew times.The paper is structured as follows. In Sect. 2 we presentthe optical interferometry data and their reduction, in additionto complementary spectra obtained by amateur astronomers.We then review in Sect. 3.1 the previous interferometric mea-surements aiming to more tightly constrain the diameter ofRigel and Deneb. We then apply this extensive record of mea-surements in the visible and near-IR continuum to interpretsemi-quantitatively the spectrally-dispersed measurements inSect. 3.2, and to elaborate a radiative-transfer model of the starsin Sect. 4.
2. Observations and data processing
Deneb and Rigel were observed with the Visible spEctroGraphAnd Polarimeter (VEGA instrument, Mourard et al. 2009) inte-grated within the CHARA array at Mount Wilson Observatory(California, USA, ten Brummelaar et al. 2005). Deneb obser- . Chesneau et al.: The H α line-formation region of Deneb and Rigel 3 Table 1.
VEGA / CHARA observing logs.
Date Time SCI / CAL Projected BaselineLength [m] PA [ ◦ ]2008.07.28 06:04 Deneb 33.4 6.22008.07.28 09:01 Deneb 32.9 -18.82008.07.28 10:51 Deneb 26.2 -46.72008.07.30 05:20 Deneb 33.3 10.32008.07.30 07:42 Deneb 32.3 -9.92008.07.30 11:32 Deneb 30.1 -37.62009.07.27 b a b b a a a a b a CAL1: HD 184006, A5V, V = ± b Observations performed remotely from France vations were carried out at regular intervals between July 2008and November 2009. Rigel observations were performed inOctober-November 2009. The VEGA control system can behandled from France remotely. A detailed description of thiscontrol system is presented in Clausse (2008).The red detector was centered around 656nm, and wemade use of the high-resolution mode (R =
30 000). For eachobservation, VEGA recombined the S1 and S2 telescopesforming the S1S2 baseline. This is the smallest baseline ofthe array, for which the extended H α emission is not over-resolved. The S1S2 baseline is almost aligned north-southand the projection of the baseline onto the sky does not varymuch during the night. This is considered an advantage in thecontext of our observations, which are designed to detect thetime variability of the H α line-forming region. Details of theobservations can be found in Table 2. On several occasions,more than one acquisition of Deneb was performed duringthe night, providing interesting, albeit limited informationabout the spatial asymmetry of the source. Since our goal wasto investigate the spectral and spatial properties of the H α line relative to the continuum, emphasis was not placed onobtaining very precise calibration of the absolute visibilities.The angular diameters of Rigel and Deneb are well known(see Sect. 3.1), so we relied on the published values to scalethe continuum absolute visibilities. Moreover, it is di ffi cultto find a suitable bright calibrator when observing at highspectral resolution, which is not well-suited to an accuratedetermination of the absolute visibility which implies a largecontinuum window. Improving upon the current determinationof the angular diameter in the visible would require a dedicatedobserving strategy using the medium-resolution mode, larger baselines coupled with a fringe tracking performed in thenear-IR. This possibility is foreseen in the future. Therefore,calibrator observations were not performed systematically, butonly to ensure that the instrument was behaving well.The data reduction method is fully described inMourard et al. (2009) and we only summarize it shortlyhere. The spectra are extracted using a classical scheme, ofcollapsing the 2D flux in one spectrum, wavelength calibrationusing a Th-Ar lamp, and normalization of the continuum bypolynomial fitting. We note that the photon-counting camerassaturate when the rate of photons is too high locally. Becauseof the brightness of the sources, neutral density filters of 0.6 to1 magnitudes had to be applied, depending on the atmosphericconditions and the spread of the speckle images on the slit.Spectra with a signal-to-noise ratio (SNR) of 300-400 wereroutinely obtained, although clear signs of saturation werefound in a few of them. After careful testing, we checkedthat the saturation had only a very limited e ff ect on theinterferometric measurements. When the quality of the spectrawas such that telluric lines are observable, these are used torefine the wavelength calibration, although most of the time theVEGA spectral calibration was checked against the referencespectra obtained by amateur astronomers. A series of spectrafor Deneb is shown in Fig. 1.The raw squared visibilities (V ) were estimated by com-puting the ratio of the high frequency energy to the low fre-quency energy of the averaged spectral density. The same treat-ment was applied to the calibrators, whose angular diameterwas computed using the software SearchCal (Bonneau et al.2006). The expected absolute visibilities of VEGA in the con-tinuum from the short baselines do not provide strong con-straints on the angular diameter and a calibrator was not sys-tematically observed. When such an observation is available,we carefully checked that the observed visibilities were com-patible with the expected ones. A contemporary measurementobtained with the medium resolution mode is presented inMourard et al. (2009).The information in a line was extracted di ff erentially bycomparing the properties of the fringe between a referencechannel centered on the continuum of the source, and a slid-ing science narrow channel, using the so-called cross-spectrummethod (Berio et al. 1999). The absolute orientation of the dif-ferential phase was established by considering the change in thespectral slope of the dispersed fringes (Koechlin et al. 1996)when changing the delay line position. Thus a positive valuecorresponds to a photocenter displacement along the S1S2 pro-jected baseline in the south direction .The width of the science channel was 0.02 nm in good at-mospheric conditions, and degraded to 0.08 nm in poor weatherconditions. The rms of the spectrally dispersed visibility in the654-655 nm continuum ranges from 3-4% at visibility 1 in thehighest quality nights (2008 / /
28, 2008 / /
30, 2009 / /
01) toabout 7-8% in medium nights (2009.07.27, 2009.08.26) andmore than 10% in poor nights (2009 / / ff erential phases follows the same trend ranging from 1-2 ◦ http: // / searchcal / O. Chesneau et al.: The H α line-formation region of Deneb and Rigel Fig. 1.
Top:
Montage of H α observations at various epochsfor Deneb and Rigel. We also include one observation of thecalibrator (HD17643). The dashed lines indicate the spectrarecorded by amateur astronomers. Bottom:
Normalized multi-epoch H α observations of Deneb. The color coding is the sameas that used in the top panel.in good conditions, to 3-4 ◦ in medium conditions, and 5-6 ◦ inpoor conditions. Figures. 2–3 illustrate the best observationsof Deneb secured in 2008. A Gaussian fit was applied to thedi ff erential visibilities and phase to retrieve accurate informa- tion about the spectral FWHM and position of the interfero-metric signal. The spectral location of the di ff erential visibil-ity and di ff erential phase dips are stable at a level of 0.005 nm( ∼ − ). Information from the blue camera was also used,as some important lines, e.g. Si ii Several H α spectra were obtained during the same periodwith the 0.28 m amateur telescope (Celestron 11) located inCastanet-Tolosan (France) equipped with the eShel spectro-graph and a QSI532 CCD camera (CCD KAF3200ME). Thesespectra were used in this study as an indication of the emissionlevel and variability of the stars. The typical resolution of thesespectra is ∼
11 000.The reduction of these data was performed using the stan-dard echelle pipeline (Reshel software V1.11). H2O telluriclines are removed by means of division by a synthetic H2Ospectrum using Vspec software - the telluric-line list was takenfrom GEISA database (LMD / CNRS). We corrected for thediurnal and annual earth velocity are corrected for (spectralwavelengths are given in an heliocentric reference for a stan-dard atmosphere). Systematic di ff erences were found in theVEGA / CHARA and amateur continuum correction producingthe di ff erences seen in Fig. 1 between the solid (VEGA) anddashed lines (eShel).
3. Results
For the most accurate estimates to date of the diametersof Rigel and Deneb, we refer to Aufdenberg et al. (2008,2006a), in which CHARA / FLUOR observations in the K bandwith baselines reaching 300 m are described. These obser-vations infer a UD angular diameter of 2.76 ± ± ± µ m,2.26 ± µ m, and 2.25 ± µ m. Thefew V measurements secured from the present observationswith a very limited spatial frequency range agree with theMarkIII measurements. We estimated the uniform disk diam-eter of Deneb using only the red camera to be 2.31 ± µ m, and using both cameras found the diameter to be2.34 ± = . Chesneau et al.: The H α line-formation region of Deneb and Rigel 5 Fig. 2.
Top row:
Normalized flux (upper curve) and dispersed visibility (lower curve; spectral band 0.02 nm) for the H α observa-tions of Deneb obtained on 2008.07.28 (solid line) and on 2008.07.30 (dotted line). The resolution is R = Bottom:
Sameas top row, but now for the di ff erential phase. A strong signal changing with baseline direction is observed, indicating that therewas a significant asymmetry in the line-forming region at this time. Table 2.
VEGA / CHARA V measurements performed on2009.07.27 and 2009.08.26. Date Wavelength Baseline V V error[nm] [m]2009.07.27 654 32.97 0.421 0.0192009.07.27 657 32.97 0.444 0.0242009.07.27 630 32.97 0.376 0.0162009.08.26 654 33.19 0.418 0.051 α line-formation region The H α line is one of the most optically thick of all lines seenin the optical and near-IR spectra of hot stars, hence representsan excellent tracer of their winds. This line has mostly been ob-served by means of spectroscopy at various spectral resolution,and in some dedicated campaigns with a very intensive timecoverage aimed at recovering the time variability of the line-forming region. We refer to Rzaev (2008) and Morrison et al.(2008) for the latest reports on Deneb and Rigel. This con-spicuous line-profile variability suggests that the wind itself,where the H α line forms, is variable in its properties. For ex-ample, variations may take place in the ionization, the mor- phology, or the density structure of the wind. The significantchanges in the line-profile shape with time indicates that theH α line-formation region is asymmetric, and that this asymme-try changes with time.The H α line observed in Deneb exhibited evidence of someactivity during the 2008-2009 observations, but not of any dra-matic change in mass-loss rate. The Rigel spectra were also ob-tained during what appears to have been quiet periods and di ff erfrom the “typical” spectrum shown in Fig. 1 of Morrison et al.(2008). As can be seen here in Fig. 1, the H α line is as deep asthe carbon C ii lines at 657.8 and 658.3 nm in Rigel. This is notan unusual state, as H α often appears in pure absorption andto be symmetric about the line center (in the rest-frame) about20% of the time (using Morrison et al. (2008) statistics).Both the Deneb and Rigel dispersed visibilities exhibit aprofound dip in the H α line, suggesting that the line forms overan extended region above the continuum. These dips are sym-metric about the line center (in the rest-frame) in both objects.The FWHM of the visibility dip was estimated by perform-ing a Gaussian fit to data of the high quality nights in 2008.These measurements yield a FWHM of 0.215 ± ± − for Deneb (Fig. 2). By comparison, the visibility O. Chesneau et al.: The H α line-formation region of Deneb and Rigel Fig. 3.
Same as Fig. 2, but now showing Deneb observations obtained with the blue camera, whose spectral range covers theSi ii ii ff set by 0.1 for clarity), marginally resolved by VEGA / CHARA. However, no phasesignal is observed.signal is narrower for Rigel, with a FWHM of 70 ± − .One can estimate the visibility in the line by using a contin-uum derived from published values of the angular diameter.The corresponding UD estimates for the highest extension ofthe H α line-forming regions are 4.1 ± ± R ⋆ , respectively.The di ff erential phases are a direct indication of the positionof the H α line-forming regions at di ff erent radial velocities, bycomparison to the continuum considered as the reference ofphase. For the 2008 data, the Gaussian fits provide FWHMsthat increase from the baselines oriented near + ◦ to -40 ◦ from61 ± − to 84 ± − . Sub-structures are also detected,with the highest peak near the zero velocity, and two satellitesat about 40 km s − . The peak level reached by the di ff erentialphases follows a trend from a large photo-center shift at P.A.close to 5-10 ◦ that decreases toward − ◦ .The di ff erential phases recorded for Deneb between 2008and 2009 are shown in Fig. 5. Dramatic changes are observedwith periods of large phase signal alternating with periods con-taining no detectable signal.The “calm” periods may potentially provide important in-formation about the rotation of the star. This characteristic sig- nal can be observed in the rotating circumstellar environmentof Be-star disks (see Delaa et al. in preparation; Berio et al.1999; Vakili et al. 1998; Stee 1996), and also directly onthe photosphere of a rotating star (Le Bouquin et al. 2009).Approaching and receding regions of a rotating star experiencedi ff erent Doppler shifts and are thus spectrally separated. Inthe sky, the natural consequence is that the line-absorbing re-gions are seen by the interferometer at one or the other sideof the continuum, generating the well-known S-shape signal inthe di ff erential phases. This signal might have been detectedin Rigel data of the 2009-10-01 (see Fig. 4). The rms of thephase of these data is 7.7 ◦ , and the maximum and minimumof the signal are at 26 ◦ and -17 ◦ , respectively, at the blue andred sides at ∼
15 km s − from line center. The signal is kept at asimilar level when extracted with a double binning of 0.04 nm,and the phase rms is decreased to 5.6 ◦ . This is on the order ofthe estimated v sin i of the source of about 36 km s − . We wereunable to detect a similar signal for Deneb probably becauseof the limited spectral resolution, insu ffi cient to resolve its lowestimated v sin i of 20 km s − . As the signal is anti-symmetric,it cancels out if the spectral resolution is too low. One must alsokeep in mind that the quality of the data fluctuated and that the . Chesneau et al.: The H α line-formation region of Deneb and Rigel 7 Fig. 4.
Top: H α observations of Rigel on 2009.10.01 (uppercurve), and the corresponding di ff erential visibility, scaled tothe expected absolute level of the continuum (lower curve). Bottom: Di ff erential phase curve exhibiting a clear S-shape sig-nal, indicative of rotating circumstellar material. A vertical dot-ted line is shown in both panels to indicate the location of theH α rest-wavelength determined by a Gaussian fitting of the vis-ibility curve.binning could not be kept identical at all epochs. It is not im-possible that weak signals are blurred (such a signal might bevisible the 2009 / /
17; Fig. 5). Given the large extension andthe activity observed in the di ff erential phases in H α , this line isprobably not ideally suited to inferring the rotation of the star.To first order, and for marginally resolved sources, the dif-ferential phases can be considered to linearly depend on thephoto-center position of the emitting source. This is no longerthe case when the source is significantly resolved as in ourcase, but we consider this as a rough estimate. Assuming thatthe line emission represent a fraction f ratio of typically 50%of the light in the H α line core, the following relation is used p = − ( φ/ λ/ B )(1 / f ratio ). We find that the astrometric shiftsinduced by the inhomogeneous H α emission reach 0.5 mas(0.2 R ⋆ ), but are lower than 0.3 mas most of the time. The eventobserved in 2008 is remarkable as the full line appeared to beo ff -center relative to the continuum. A similar event was ob-served on 2009 / /
01 to have an opposite direction. ii The Si ii α line in the blue camera (Fig.3) in 2008. The line profile showsa pure absorption, but is slightly asymmetric with an extendedblue wing due to extended absorption in the wind regions. TheVEGA / CHARA observations of Deneb obtained in 2008 indi-cate that the line-formation region of the Si ii Fig. 5.
Time evolution of the di ff erential phases of Deneb.The strong phase signal observed in 2008 is indicative of avery asymmetric environment, o ff -centerd from the continuumsource. No clear S-shape signal is observed, in agreement withthe low v sin i of Deneb of ∼
20 km s − , unresolved by the in-strument.extended than the continuum forming region, with di ff erential-visibility dips of 10%, 19%, and 15% for baselines in the range[5 ◦ :10 ◦ ], [-10 ◦ :-20 ◦ ], and [-35 ◦ :-50 ◦ ]. The rms of the dispersedvisibilities is 5%. The line FWHM estimated from Gaussianfitting is 55 ± − and the FWHM of the visibility dipis narrower with a FWHM decreasing from 33 ± − to24 ± − for P.A. from [5 ◦ :10 ◦ ] to [-35 ◦ :-50 ◦ ], respectively.One can roughly estimate the extension of the line-formingregion using the minimum of the visibility and UD approx-imation to be 2.6 mas, 2.7 and 2.75 mas using the value of2.34 mas from NPOI as the diameter at 630 nm. One may spec-ulate whether this trend in the visibilities is permanent or atransient event closely related to the asymmetries inferred fromthe di ff erential phases in the H α line. Because of a spectro-graph realignment carried out in 2009, the Si ii ii ff erential-phase signal was detected above an rms of ∼ < ◦ . This means thatthe imprint of the rotation of the star on the Si ii line-formationregion is not detected in the data, probably because of insu ffi -cient spectral resolution. Strong Si ii lines were also observedin Rigel’s spectrum, but neither di ff erential visibility nor anyphase signal was detected. The SNR of the data is poorer thanthe Deneb observations, with an rms in the dispersed visibilitiesand phases of 10%. O. Chesneau et al.: The H α line-formation region of Deneb and Rigel Fig. 6.
Left:
In the bottom panel, we show a contour plot of the quantity P × I ( P ) as a function of Doppler velocity and impact pa-rameter P for H α , computed with CMFGEN using the stellar parameters of Deneb and a mass-loss rate value of 3.1 × − M ⊙ yr − ,corresponding to the best-fit model. This figure illustrates the distribution of intensity with impact parameter, and in particularserves to infer the regions that contribute significant flux to the line, and hence the spatial extent of the H α line-formationregion. In the top panel, we show the corresponding normalized synthetic flux in H α . Right:
Same as left, but now for a mass-loss rate value of 6.2 × − M ⊙ yr − . Notice the sizable change in profile morphology, echoing the change in the extent of theline-formation region.
4. Comparison with radiative transfer models
The BA supergiants, especially nearby ones, have been exten-sively studied with sophisticated radiative-transfer codes. It isnot in the scope of this paper to determine more reliably thefundamental parameters of Deneb or Rigel. However, we wishto investigate several important issues related to the putative ef-fects of the wind on the interferometric observables. Deneb hasbeen observed by several northern-hemisphere interferometersin both the visible and the near-IR. Rigel is currently moni-tored with VEGA / CHARA in the northern hemisphere, as wellas in the southern hemisphere with the VLTI. Hence, in thisstudy, we used the stellar parameters obtained by Przybilla etal. (2006) and Schiller & Przybilla (2008) for Rigel and Deneb,respectively. After developing a convergent model for a refer-ence mass-loss rate value at which H α is predicted in absorp-tion, we then increased the mass-loss rate (keeping the otherstellar parameters fixed) until the H α line exhibited a well-developed P-Cygni profile. In this way, we explored the fol-lowings questions: – To what extent the angular diameter inferred from opticaland near-IR continua is sensitive to changes in mass-lossrates? – Are the H α , Pa β , and Br γ line-formation regions repro-duced by radiative-transfer simulations accounting for awind? These lines form in di ff erent parts of the wind andcan thus be used simultaneously to constrain its properties.In this paper, we focus on the H α line because it can be ob-served with VEGA / CHARA. A similar study for Pa β and Br γ line-formation regions is postponed to another paper in prepa-ration. Our radiative-transfer calculations were carried out with theline-blanketed non-LTE model-atmosphere code CMFGEN(Hillier & Miller 1998; Dessart & Hillier 2005), which solvesthe radiation-transfer equation for expanding media in the co-moving frame, assuming spherical symmetry and steady-state,and under the constraints set by the radiative-equilibrium and . Chesneau et al.: The H α line-formation region of Deneb and Rigel 9 Fig. 7. Top:
Theoretical visibility curves for Deneb, computed at selected wavelengths in the range ∼ α line, and for mass-loss rate values of 3.1 × − M ⊙ yr − (left) and 6.2 × − M ⊙ yr − (right). Bottom:
Same as top,but now for Rigel and using mass-loss rate values of 1 (left) and 2 × − M ⊙ yr − (right). The dotted and dashed lines correspondto the best UD curves fitting the visible and the near-IR continua, respectively. The range of baselines of the VEGA / CHARA isindicated by two vertical dash-dotted lines.statistical-equilibrium equations. It treats line and continuumprocesses, and regions of both small and high velocities (rel-ative to the thermal velocity of ions and electrons). Hence, itcan solve the radiative-transfer problem for both O stars, inwhich the formation regions of the lines and continuum extendfrom the hydrostatic layers out to the supersonic regions of thewind, and Wolf-Rayet stars, in which lines and continuum bothoriginate in regions of the wind that may have reached half itsasymptotic velocity.We used the stellar parameters inferred by Schiller &Przybilla (2008) for Deneb, and by Przybilla et al. (2006)and Markova et al. (2008) for Rigel. The Mg ii resonancelines suggest terminal wind speeds of ∼
240 km s − for Deneb(Schiller & Przybilla 2008) and ∼
230 km s − for Rigel (Kauferet al. 1996b). Their projected rotational velocities are low, i.e.20 ± − and 36 ± − respectively.Even for sources as “close” as Deneb and Rigel, the dis-tance estimates remain quite inaccurate. Schiller & Przybilla(2008) derived a luminosity of 1.96 × L ⊙ for Deneb, using adistance of 802 ±
66 pc (assuming that Deneb is a member of theCyg OB7 association), while the one derived from Hipparcos(van Leeuwen 2007) is significantly smaller (i.e., 432 ±
61 pc), which infers a luminosity estimate of 5.5 ± × L ⊙ . This re-assessment directly scales down by almost a factor two the lin-ear scales of the size parameters. Furthermore, recalling thatthe mass-loss rate scales with the luminosity, this should implya weaker steady-state wind mass loss for Deneb. The corre-sponding uncertainties are considerably lower for Rigel, whichis much closer than Deneb.By adopting values these parameters from previous works,we find that the synthetic spectra computed by CMFGEN agreefavorably with the observations. We thus adopt these stellar pa-rameters and vary the mass-loss rate value to assess the impacton the spectrum and in particular H α and Si ii ff ect of a wind mass-loss rate of1.55, 3.1, 6.2, and 12.4 × − M ⊙ yr − for Deneb, and valuesof 1, 2, 4, and 8 × − M ⊙ yr − for Rigel. A typical e ff ect thatappears in theoretical models is illustrated in the bottom pan-els of Fig. 6, where we show the distribution of the emergentintensity I (scaled by the impact parameter p ), as a functionof Doppler velocity and p . This type of illustration was in-troduced by Dessart & Hillier (2005) to study line formationin hydrogen-rich core-collapse supernova ejecta. Here, it pro-vides a measure of the extent of the line-formation region of α line-formation region of Deneb and Rigel Fig. 8.
Left:
In the upper panel, we compare the H α line profile obtained in July 2008 (black; rectified) of Deneb and thecorresponding model predictions for wind mass loss rate of 1.5 (orange), 3.1 (red), 6.2 (green) and 12.4 × − M ⊙ yr − (blue). Inthe lower panel, we give the corresponding dispersed visibility for a nominal baseline of 33 m. Right:
Same as before, but nowfor Si ii α relative to the neighboring continuum and the sites wheremost of the emergent photons originate. Comparing the left andright panels suggests that a variation by a factor of two in themass-loss rate value leads to sizable changes in the extent ofthe line-formation region, and correspondingly, large changesin the observed line profile. While spectroscopy is sensitiveto the latter, interferometry is sensitive to the former. Below,we describe the interferometric signals associated with theseintensity maps computed with CMFGEN and produced usingthe stellar parameters suitable for Rigel or Deneb, and variousmass-loss rates tuned to match observations. This is illustratedin Fig.7 in the case of Deneb and Rigel, using two di ff erentmass-loss rates. The visibility curves for various spectral chan-nels in the vicinity of H α are plotted as a function of the spatialfrequency. The di ff erential visibilities observed by VEGA aregenerating by plotting the di ff erent value for a given projectedbaseline (between 27m and 33m). In Fig.10, we show a zoom of the 0.6 µ m and 2.2 µ m squared-visibility curves of Deneb in the second lobe with the variousmass-loss rate values used in this paper. The visible contin-uum appears to be far more sensitive to such changes thanthe near-IR continuum. We note that these visibility curvesare computed for a spectral resolution of 30 000, and can-not be directly compared with the broadband measurementsof FLUOR / CHARA over the full K ′ -band (Aufdenberg et al.2008).Doubling the mass-loss rate from 3.1 to 6.2 × − M ⊙ yr − does not significantly alter the optical thickness of the wind. The angular diameter of the star in the continuum near H α de-termined by a UD fitting of the visibility curve does not there-fore change by more than 2%. However, we note that the sec-ond lobe is significantly a ff ected. This conclusion is also truein the near-IR. All models also show that the near-IR UD an-gular diameter is systematically larger than the visible one by ∼ / FLUOR valueof 2.363 mas, and this implies that the expected UD diameter inthe H α overlapping continuum is ∼ ± ± µ m and 2.26 ± µ m.For Rigel, the di ff erence between the near-IR UD angu-lar diameter and its visible counterpart is increased slightly toreach ∼ / CHARA measurement of2.758 mas as reference, this would imply a diameter of about2.64 mas in the visible. In the near-IR, some instruments suchas FLUOR / CHARA have an accuracy often better than 1% de-pending on the atmospheric conditions, and the ability to ob-serve with long baselines. The interpretation of the observa-tions obtained with this instrument might be a ff ected by a mass-loss rate variation in the form of localized inhomogeneities, butprobably on a smaller scale than an instrument with a similaraccuracy operating in the visible..The second lobe of the visibility curve is far more sen-sitive to any fluctuation of the mass-loss rate in the visiblethan the infrared. The reasons are twofold: a higher sensitiv-ity to limb-darkening e ff ects in the visible, and a more ex-tended continuum-formation region, despite the very limitedamount of flux involved (the wind remains in any case opti- . Chesneau et al.: The H α line-formation region of Deneb and Rigel 11 cally thin). This can be seen in Fig.10 as the second lobe of thevisibility at 2.2 µ m is closer to the uniform disk model than theone at 0.6 µ m. Balmer bound-free cross-sections increase from400 to 800nm. This causes the continuum photosphere to shiftweakly in radius across this wavelength range and also altersthe limb-darkening properties of the star. This e ff ect might alsobe discernible in the MarkIII data (Mozurkewich et al. 2003,Sect.3.1), Deneb appearing smaller at wavelengths close to theBalmer jump, although the baselines were too short to probethe second lobe of the visibility curve. An interferometer ableto resolve a hot star up to the second lobe in the visible isvery sensitive to the mass-loss rate, even in the case of a veryweak wind. In the near-IR, the free-free emission strengthens.Free-free opacities increase at longer wavelengths, increasingthe photospheric radius relative to that measured in the visible.Changing the mass-loss rate does not significantly a ff ect thelimb-darkening and the temperature scale near the star, thusthe location of the extended emission. As a consequence, thefree-free emission causes a larger diameter in the near-IR, in-dependently to first order of the mass-loss rate. α and Si ii lines on themass-loss rate We then computed the spectrum and the dispersed visibili-ties in the H α line at a spectral resolution R =
30 000. Wewere unable to perform a satisfactory fit to the H α profilefor either Deneb or Rigel, the absorption component beingsystematically too deep (Fig.8). This conclusion was reachedin many studies, and these profiles could not be reproducedin terms of spherically-symmetric smooth wind models (seefor instance Fig. 7 in Markova et al. (2008), Aufdenberg et al.(2002), and Schiller & Przybilla (2008)). In this temperatureregime, the models systematically predict profiles in absorp-tion partly filled-in by wind emission, hence only lower limitsto the mass-loss rate can be derived by fitting the H α line.However, this discrepancy is mitigated by the quality of thevisibility fit shown in the upper panel of Fig.8, which is by farless sensitive to small absorption e ff ects along the line-of-sight.Changing the mass-loss rate by a large factor of 2–4 has a dra-matic impact on the H α line-formation region, changing thespectrum appearance and the dispersed visibility curve. Whenusing a baseline in the range 26-33 m, there is a relationship be-tween the mass-loss rate and the FWHM of the visibility dropin the line that can be approximated by a second-order polyno-mial in the range considered (1.55 to 12.4 × − M ⊙ yr − ). Onecan see in Fig. 11 that the nominal model of Deneb with a mass-loss rate of 3.1 × − M ⊙ yr − fits the observed visibility curveswell. The model FWHM of the visibility is 89 km s − and theobserved FWHM is 98 ± − . Given the FWHM-mass-lossrelationship, this would correspond to a mass-loss rate in therange 3.7 ± × − M ⊙ yr − . For Rigel, a similar relationshipprovides a mass-loss rate in the range 1.5 ± × − M ⊙ yr − .We have studied the variations in 2008 and 2009 of the visi-bility profile without detecting any significant variation in thevisibility profile, which, translated in a mass-loss rate variation, Fig. 9.
Same as the upper panel of Fig. 8, but now for Rigel.suggests that the mass-loss rate has not changed by more than5%.By comparison, the Si ii − , comparedto the 2008 VEGA / CHARA measurement of 55 ± − .The FWHM of the corresponding model visibility dip areindependent of the mass-loss rate, being unchanged at thevalue 19.5 km s − and lower than the measured mean value of29 ± − for the 2008 observations. The depth of the visi-bility dip fits more accurately the curve with a mass-loss rateof 6.2 × − rather than the nominal value of 3.1 × − M ⊙ yr − inferred from the H α line. It is not possible to establish whetherthis issue is related to a particular event occurring during the2008 observations, or a permanent situation, or even a bias inthe radiative-transfer model.Despite the lower quality of the interferometric data onRigel, one can see in Fig.9 that the model provides a reason-able fit. We note that the C ii lines at 6578Å and 6583Å in thespectrum of Rigel are also slightly a ff ected by the wind. Theirvisibility for the VEGA baselines are about 0.5-1% lower thanthe nearby continuum (using our model predictions), depend-ing on the mass-loss rate. Detecting such a weak signal wouldrequire that the accuracy on the di ff erential visibility is about2-3 times better than the current instrument performance. The variability in the di ff erential phase signal observed is aclear sign of the stochastic activity of the wind of Deneb. Eventhough the data secured are partial, one can note that no dif-ferential phase signal exceeds 30 ◦ , which is impressively largeconsidering that the baseline is limited. In 2008 and in October α line-formation region of Deneb and Rigel Fig. 10.
Synthetic continuum squared-visibility curves at0.6 µ m (top) and 2.2 µ m (bottom) as a function of spatial fre-quency, and given for wind mass-loss rates of 1.5 (orange), 3.1(red), 6.2 (green), and 12.4 × − M ⊙ yr − (blue). The dottedblack line corresponds to a uniform disk of 3.32mas, and thedashed black line to a uniform disk of 3.36mas.2009, the di ff erential phase signal was able to be observedthroughout the full line, at a level above 30 ◦ . This signal didnot appear to be due to an increased level of stochastic ’noise’expected from the signature of small, localized clumps, but ex-hibited a well-structured signature. At other times, the di ff eren-tial signal was only observed close to ± vsini , and some patternswere reminiscent of the S-shape signature, originating in a ro-tating structure.The radiative transfer models demonstrate that at the in-ferred mass-loss rate of Deneb, significant emission originatesin a circumstellar region of up to 2-4 stellar radii, and that thelevel of this extended emission depends strongly on the mass-loss rate. We tried to investigate the di ff erential phases withad hoc approaches using the models as a basis to generate theperturbation. Di ff erential visibilities and phases provide con-tradictory and stringent constraints. On the one hand, a strongdi ff erential phase signal is observed, but on the other hand, thedi ff erential visibilities did not vary by more than 5% over thetwo years of sparse observations. This restricted considerablythe size, flux, and location of the perturbation. The emergentflux of the best-fit model was perturbed using a 2D Gaussian asa weighting function. The parameters of the perturbation wereits location, FWHM and flux ratio compared to the unperturbedmodel. The best parameter range was reached for a perturba-tion located at 2-3 stellar radii from the star, with a FWHMof 0.3-0.7 stellar radii, and a flux contrast of 10-25. One cansee in Fig.12 that the di ff erential visibilities of the perturbedand unperturbed models are similar, whereas a significant dif-ferential phase signal is observed. The baseline is aligned in thedirection of the perturbation observed in July 2008, i.e. northof the star. This model is also compatible with the di ff erential Fig. 11.
Variation in the FWHM measured from the H α vis-ibility curves computed with the intensity maps produced byCMGEN, shown here as a function of mass-loss rate and adopt-ing the stellar parameters suitable for Deneb. The adoptedbaseline is 33 m. For a high mass-loss rate value, the H α line-formation region is fully resolved, and the FWHM ofthe visibility curve is close to the wind terminal velocity.The dotted lines indicate the mass-loss rate inferred from theVEGA / CHARA measurements.phases observed with the two other baselines (Fig.2). Placingthe perturbation farther away leads to a phase signal that is farweaker than to the observations, since the line-forming regionof the unperturbed model naturally ends at 5-6 stellar radii.This is a limitation of this ad hoc approach. Making the per-turbation larger in size, even with a much lower flux contrastleads to a noticeable change in the di ff erential visibilities. Thesame consequence is reached when putting the perturbation tooclose to the star. Finally, if the perturbation is too small (i.e.FWHM ≤ β and H α lines in Deneb and Rigel, at a typical level of about0.1%. Hayes (1986) performed an intensive broadband moni-toring in B -band polarization of Rigel, revealing a variabilityat a typical level of 0.2%. The variability pattern in the Q − U plane suggested that the ejection of material was not limited toa plane, and non-radial pulsation were thus proposed as the rootcause of these localized ejections. This might be caused by thelimited amount of intensive observations. This might also bea consequence, in the case of Deneb, of a small inclination of . Chesneau et al.: The H α line-formation region of Deneb and Rigel 13 Fig. 12.
Comparison on the unperturbed model (dashed lines) with the perturbed model (blue lines, as explained in the text). Thethick lines are the observation of July 2008. The parameters of the perturbation are the followings: flux contrast of 20, FWHMof 0.5 stellar radius, and position at 2.8 stellar radius south of the star. The insert shows the square root of the perturbed model atthe wavelength corresponding to the core of the line, where the signal is maximum.the rotation axis on the sky. Any large-scale structures orbitingin the equatorial plane would generate a significant polariza-tion variability, without any preferential direction. In addition,no direction appears to be preferred by our observation, thoughwe emphase that such a conclusion is very limited given thestringent limitations in the uv coverage.
5. Deneb as a fast rotator
Aufdenberg et al. (2008, 2006a) presented evidence that Denebis a fast rotator, based on 24 high accuracy FLUOR / CHARAvisibilities in the K ′ -band with projected baselines rangingfrom 106 m to 310 m, sampling the first and second lobes ofthe visibility curve. They detected slight departures from apurely spherical model at a level of about 2% in the near-IR,a discovery that had not been anticipated for an AB super-giant. They tentatively fitted the data with a rotating model at-mosphere, and demonstrated that despite the low v sin i of thestar, a model at half critical speed, seen nearly pole-on mayaccount for the interferometric observations. A by-product ofthis fitting process is the determination of crucial parametersin this context, namely the inclination, estimated to be i = ◦ ,and the orientation on the sky of the rotation axis found to beat about P.A. = ◦ east from north. These findings have po-tentially important implications. In a general context, Denebwould be the first AB supergiant proven to be the descendantof a fast rotator. Support for this interpretation was provided by Schiller & Przybilla (2008). To more clearly interpret the pro-nounced mixing signature with nuclear-processed matter, theseauthors proposed that Deneb was probably a fast rotator ini-tially, and is currently evolving to the red-supergiant stage.In the context of our observations, the consequences onthe H α line-formation region must be evaluated. The H α line may be a ff ected by a moderate change in wind prop-erties, such as those that occur due to a latitudinal vari-ation in the e ff ective temperature of the star, estimated tobe ∼
700 K (Aufdenberg et al. 2006a). Moreover, as stated bySchiller & Przybilla (2008), hydrogen lines are mainly sensi-tive to variations in log g .Neither the VEGA / CHARA data nor the theoretical studypresented in this paper can provide definitive support for, orexclude, this interpretation. The H α variability observed byVEGA / CHARA is related to localized inhomogeneities in thewind of this star. Is not impossible that that these inhomo-geneities may a ff ect the continuum forming region in the Kband, although is has been shown in Sect. 4.2 that large varia-tions in mass-loss rate are required to significantly a ff ect thesecond lobe of the visibility curves. Yet, this conclusion isbased on the ideal case of a spherical model, although giventhat the wind is very optically thin in the continuum, those in-homogeneities should not significantly alter the properties ofthe continuum-formation regions.The 2008 data arguably provide some support for the fastrotator model. We note that the S1S2 projected baselines were α line-formation region of Deneb and Rigel roughly aligned (range of -45 /+
15 degrees) within the directionof the asymmetry found with CHARA / FLUOR at 150 ◦ , i.e. thedirection of the pole in the fast rotator model (Aufdenberg et al.2006a). On the other hand, the H α di ff erential phases are ob-served to increase from a baseline roughly aligned to the poledirection at PA = ◦ . This may imply that the asymmetry isgreater in a direction perpendicular to the pole. On the otherhand, the Si ii line dispersed visibilities seem deeper in the di-rection of the pole, suggesting a more extended line-formationregion. Moreover, the high mass-loss rate inferred from the fitto the visibilities through the SiII line may also be an indicationof a co-latitude dependence of the mass-loss rate. That no ev-idence of rotation was detected in the di ff erential phases is anadditionnal argument for a low v sin i , and therefore a nearlypole-on configuration for Deneb. These limited observationscannot provide definite conclusions, and the fast-rotator inter-pretation still needs to be investigated, both theoretically andobservationally. This could be done, for instance, by repeatingthe FLUOR / CHARA observations to check whether the near-IR interferometric signal has changed or not, or by devoting afull VEGA / CHARA run with more extensive coverage, prefer-ably by using the 3T mode. It would also be of interest to in-vestigate the impact of the fast rotation of Deneb on the H α and Si ii
6. Conclusion
We have presented the first high spatial and spectral observa-tions of two nearby supergiant stars Deneb and Rigel. The H α line-formation regions were resolved and their angular size wasfound to be in agreement with an up-to-date radiative transfermodel of these stars. The H α line-forming region appears to beasymmetric and time variable, as expected from the numerousintensive spectral monitoring of AB supergiants reported in theliterature. However, the time-monitoring of the dispersed vis-ibilities inferred from the H α line of Deneb did not provideevidence of mass-loss rate changes above 5% of the mean rate,and the activity observed can be considered most of the timeas a second-order perturbation of the wind characteristics. Theincreased coverage obtained in 2008 provides some evidenceof a latitudinal dependence of both the H α di ff erential phasesand the Si ii ff erential visibilities. This line was notsignificantly resolved subsequently. Only two observations ofRigel were secured. An S-shape signal was detected in the H α di ff erential phase of Rigel, suggesting that the rotation signalis detected. We note that the H α line profile was almost pho-tospheric at the time of our observations. Observations with alarge uv coverage may provide, in the case of Rigel, the direc-tion of the rotational axis on the sky and its inclination.Given the large angular size of Rigel and Deneb, the H α line-forming region is fully resolved by the interferometer withbaselines longer than 50m, and only the S1S2 baseline of theCHARA array is short enough to permit such an investigation.Therefore, it is not easy to significantly increase the observa-tional e ff ort performed on these stars with such stringent ob-serving restrictions. An extension of this work is possible forsources with an apparent angular diameter between 0.5 and 1.5 mas, that are large enough, but also bright enough to makeuse of the highest spectral dispersion of the VEGA instrument(limiting magnitude of about 3). This concerns the AB super-giants closer than 1.5 kpc, and the supergiants in the Orioncomplex ( d ∼
500 pc) are in this context of particular interest.Simultaneous 3 telescope recombination would provide muchbetter uv coverage than the one presented in this paper. Anotherinteresting possibility is to add to this study several diagnos-tic lines such as the Ca ii infrared triplet (849.8nm, 854.2nm,866.2nm). These resonance lines are highly sensitive to non-LTE e ff ects arising close to the photosphere and may shed somelight on the regions were the material is launched. Acknowledgements.
VEGA is a collaboration between CHARA andOCA / LAOG / CRAL / LESIA that has been supported by the Frenchprograms PNPS and ASHRA, by INSU and by the r`egion PACA.The project has obviously taken benefit from the strong support ofthe OCA and CHARA technical teams. The CHARA Array is oper-ated with support from the National Science Foundation and GeorgiaState University. We warmly thank Christian Hummel for having pro-vided the MarkIII data. The referee, Mike Ireland helped us by hisuseful comments to improve this paper significantly. This researchhas made use of the Jean-Marie Mariotti Center
SearchCal service co-developed by FIZEAU and LAOG, and of CDS AstronomicalDatabases SIMBAD and VIZIER . M.B.F. acknowledges ConselhoNacional de Desenvolvimento Cient´ıfico e Tecnol´ogico (CNPq-Brazil) for the post-doctoral grant. References
Aerts, C., Puls, J., Godart, M., & Dupret, M. 2009, A&A, 508,409Aufdenberg, J. P., Hauschildt, P. H., Baron, E., et al. 2002, ApJ,570, 344Aufdenberg, J. P., Ludwig, H., Kervella, P., et al. 2008, inThe Power of Optical / IR Interferometry: Recent ScientificResults and 2nd Generation, ed. A. Richichi, F. Delplancke,F. Paresce, & A. Chelli, 71Aufdenberg, J. P., M´erand, A., Ridgway, S. T., et al. 2006a,in Bulletin of the American Astronomical Society, Vol. 38,Bulletin of the American Astronomical Society, 84Aufdenberg, J. P., Morrison, N. D., Hauschildt, P. H., &Adelman, S. J. 2006b, in Astronomical Society of thePacific Conference Series, Vol. 348, Astrophysics in theFar Ultraviolet: Five Years of Discovery with FUSE, ed.G. Sonneborn, H. W. Moos, & B.-G. Andersson, 124Berio, P., Stee, P., Vakili, F., et al. 1999, A&A, 345, 203Bonneau, D., Clausse, J., Delfosse, X., et al. 2006, A&A, 456,789Bychkov, V. D., Bychkova, L. V., & Madej, J. 2003, A&A, 407,631Cantiello, M., Langer, N., Brott, I., et al. 2009, A&A, 499, 279Clarke, D. & McLean, I. S. 1976, MNRAS, 174, 335Clausse, J. 2008, in Presented at the Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference, Vol. 7019,Society of Photo-Optical Instrumentation Engineers (SPIE)Conference Series Available at http: // / searchcal Available at http: // cdsweb.u-strasbg.fr / . Chesneau et al.: The H α line-formation region of Deneb and Rigel 15line-formation region of Deneb and Rigel 15