M. Deleuil1 - J.-C. Bouret1 - C. Catala2 - A. Lecavelier des Etangs3 - A. Vidal-Madjar3 - A. Roberge4 - P. D. Feldman5 - C. Martin1 - R. Ferlet3
1 - Laboratoire d'Astrophysique de Marseille, CNRS - Université de Provence, BP 8, 13376 Marseille Cedex 12, France
2 - Observatoire de Paris-Meudon, CNRS, Paris, France
3 - Institut d'Astrophysique de Paris, 98bis Bd Arago, 75014 Paris, France
4 - Dept. of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch Rd., Washington, DC 20015-1305, USA
5 - Dept. of Physics and Astronomy, Johns Hopkins University, 3400 N. Charles St., Baltimore, Maryland 21218, USA
Received 10 June 2004 / Accepted 28 July 2004
We present an analysis of the FUSE spectra of HD 163296, a young isolated Herbig Ae star. The spectra in the far UV spectral range are dominated by strong emission lines of C III and O VI, which exhibit broad and complex asymmetric profiles with wings extending over more than 1000 km s-1. The H I Ly and the C III emission profiles show evidence for an outflow with a terminal velocity of 270 km s-1. We also identified several narrower emission lines due to fluorescent decays in Fe II. Using an HST/STIS archive spectrum, we checked that the H I Ly line, which has a clear type III P Cygni emission profile, is broad enough to provide the photons required to photo-excite the Fe II upper levels. As previously explored for the well-known Herbig Ae star AB Aur, most of the spectral features including the fluorescent lines are consistent with the presence of a chromosphere above the photosphere. However, the over-ionized species cannot be produced in this region alone and we discuss two alternative mechanisms to explain such spectral signatures: magnetospheric accretion and a magnetically confined wind. Both interpretations involve a large-scale magnetic field whose detection is now a challenge for the new generation of spectropolarimeters such as ESPaDOnS.
Key words: stars: activity - stars: individual: HD 163296 - line: profiles - line: identification - stars: magnetic fields
HD 163296 belongs to the small number of the so-called isolated Herbig Ae/Be stars which present the usual characteristics of the classical HAeBes, but are not embedded in a nebulosity. First classified as a young Herbig Ae/Be star by Finkenzeller & Mundt (1984), the star (A1Ve, age = 4 +4-2 Myr, d = 122 pc; van den Ancker et al. 1998) is sometimes considered as prototypical of the Herbig Ae class.
Imaging techniques provide evidence for the distribution of the circumstellar material within a disk: a disk spatially resolved, with dimension AU, was first revealed by CO mm interferometry (Mannings & Sargent 1997) and more recently, coronagraphic observations mapped an extended disk-like structure in the visible (Grady et al. 2000). Both studies agree with an inclination of the system about 60. In addition, a subsequent study in the ultraviolet detected the presence of a bipolar outflow associated with the star (Devine et al. 2000). According to Infrared Space Observatory (ISO) observations, the CS dust presents a bi-modal mass over temperature distribution with a population of large grains (1 mm) (Bouwman et al. 2000). The composition of these grains appears to be dominated by amorphous silicates, iron oxide and water ice (Sitko et al. 1999; Bouwman et al. 2000; van den Ancker et al. 2000).
Warm CS H2 gas ( ) has also been observed toward HD 163296 (Lecavelier des Etangs et al. 2003). Although this detection is consistent with a location of the gas within a few AU of the star at most, the exact spatial distribution of this medium remains to be better constrained. More interestingly, as a member of the Herbig group, the spectrum of the star from the ultraviolet (UV) to the infrared (IR), shows a rich, mostly variable, emission-line spectrum. In the visible, the Balmer lines exhibit strong variations in intensity and position with an evolution of the line shape from a double-peak to a type III P Cygni profile on time scales from one hour to several days (Pogodin 1994; Thé et al. 1985). If the Balmer lines are the most intense emission lines, many other low-ionization lines are present in the UV and in the visible and most of them exhibit similar rapid line profile variability. However, only in a few cases has periodic behaviour been seen in the analysis of temporal series of optical spectra: Catala et al. (1989) found a periodic modulation in the variations of the Mg II and Ca II resonance lines, with however, different period values: h for Mg II and h for Ca II. More recently, Beskrovnaya et al. (1998) reported the presence of likely cyclic positional variations for the H and H emission lines as well as sinusoidal variations of the linear polarization parameters. However, depending the epoch of observations, this periodic behaviour can be masked by additional active flare-like phenomena that may dominate the variation of the line profile. Together with insufficient temporal coverage, this may explain the apparent lack of periodicity in time series analysis reported by other authors (Baade & Stahl 1989).
The observation of the star in the far ultraviolet (FUV) spectral range presented in this paper allows us to probe regions of higher temperature and complete the picture of the physical conditions of the close CS environment of HD 163296. We will discuss its characteristics and their implications in the context of two models, both involving a stellar magnetic field.
HD 163296 was observed with FUSE on 2001 April 27 and 29, for 15 920 s and 16 197 s respectively. The observations were made in the time-tagged mode, using the low-resolution aperture (LWRS). Each exposure was divided into 5 sub-exposures with exposure times between 2582 and 3558 s. These individual sub-exposures were processed with version 2.4.0 of the CALFUSE pipeline processing software (see Sahnow et al. 2000; Moos et al. 2000). The sub-exposures were then cross-correlated segment by segment prior to their co-addition to avoid the potential loss of co-alignment between the spectroscopic channels from one exposure to another and we did not co-add spectra from different channels. The final spectra were then re-sampled into larger bins: 8 initial pixels for the LiF segments and 7 pixels for the SiC one, as the FUSE spectra are highly oversampled.
To complement our FUV emission line analysis, we also used a HST/STIS high resolution observation of HD 163296 obtained from the archives. The STIS spectrum was recorded on 2000 July 22, 9 months before our FUSE observations. These observations cover two spectral ranges: one around H I Ly , from 1191 to 1245 Å, which includes the Si III and the N V doublet resonance lines; the second one around the C III* UV4 multiplet, from 1145 to 1199 Å.
We took advantage of the overlap between the FUSE and STIS spectral domains around the C III* multiplet to establish the absolute wavelength scale by cross-correlation techniques. We found that the FUSE LiF2a segments are shifted by km s-1. Note that according to the SIMBAD data base, the radial velocity of the star is -4 km s-1. The offsets of the absolute wavelength calibration for the other segments were established using the H2 absorption lines which are observed in all segments.
As expected for such a cool star, the photospheric spectrum of HD 163296 in the FUV spectral range is very faint, even in the long wavelength part of the FUSE domain. On the other hand, this low level continuum makes emission features easy to detect. The most prominent emission features are the resonance lines of C III 977, the O VI doublet (1032, 1037 Å), and the C III* multiplet at 1776 Å (UV4) (Fig. 1). The spectra also contain several narrower emission features observed mainly between 1110 Å and 1150 Å. We identified these lines as fluorescent decays of Fe II excited levels pumped by H I Ly (see Sect 3.2).
|Figure 1: FUV emission lines observed in the HD 163296 FUSE (2002/29/04) and HST/STIS spectra. The arrow points to the C II resonance line at 1036 Å, whereas the line labeled with the symbol, close to the 1037 O VI line, is an airglow emission line. The narrow H2 absorption lines and the O I fine structure lines, superimposed on the emission profiles are indicated with tic marks. The dotted lines correspond to the blue wing of each emission line reflected about the zero velocity in the stellar rest frame and overplotted on the red wing to highlight the asymmetry of the profile. Note that the emission peak of the second component of the N V doublet at 1242 Å is apparent at about +850 km s-1 from the first component.|
|Open with DEXTER|
All emission lines are wide with profile wings extending to at least 550 km s-1. Except for the C III resonance line, we noticed a clear absence of absorption features other than the superimposed narrow H2 circumstellar absorption lines. All the lines display a blue-red asymmetry about the line center in the stellar rest frame. This asymmetry is not obvious from a simple examination of the radial velocities of the line wings, but it is highlighted by reflecting each line profile with respect to the star's rest frame velocity, as illustrated in Fig. 1.
The O VI doublet shows the most regular profile although it cannot be fit by a single Gaussian. In this case, the lines appear with a flat-topped profile which is however compatible with an unresolved double-peak.
The N V emission line profiles in the STIS spectrum appear surprisingly different from those of the O VI doublet. The observed lines indeed seem compatible with an inverse P Cygni profile: a single broad emission peak with the centroid of the 1238.8 Å emission blueshifted by 110 km s-1 relative to the star's rest frame, and a strong absorption feature in the red wing of the line. However, both lines of the N V doublet are strongly blended: by the Mg II resonance line at 1240.4 Å, and by the lines of the N I multiplet UV 5 at 1243.1 Å. In the case of the well-known Herbig Ae star, AB Aur, which exhibits a similar line profile in the N V doublet, Bouret et al. (1997), using a detailed model of line formation, concluded that the origin of the Mg II and N I lines was purely photospheric. The similarities between AB Aur and HD 163296 together with the width of the absorption features, allow us to draw the same conclusion. The actual N V profile could thus be similar to that of O VI but a precise estimation of the N V line profile in HD 163296 would require detailed modeling of the physical regions where the different lines are formed, which is beyond the scope of the present paper.
The C III resonance line profile is very different from that of the C III* multiplet, which displays a clear double-peaked profile. The resonance line presents a strong, slightly redshifted emission peak with an extended red wing and a blueshifted emission peak of moderate intensity. It is surprisingly similar to the one we previously observed in the FUSE spectrum of the Herbig Be star, HD 100546 (Deleuil et al. 2004). Absorption features are superimposed on the profile: the narrowest ones are due to H2 electronic transitions and the broadest to O I resonance fine-structure transitions. A broad absorption feature appears in both exposures at a velocity of -58 km s-1 which can be related to an outflow in C III.
Two weak and broad emission features are also marginally detected: one in the blue wing of the O VI 1037 line, indicated by an arrow in Fig. 1, and a second one observed at 933.2 Å. Following the analysis of the FUSE spectrum of HD 100546 (Deleuil et al. 2004), the first emission feature is identified as the blue wing of an emission line due to the C II resonance line at 1036 Å. The line at 933.2 Å could be due to a resonance transition of S VI, but with a single observed line this identification still needs to be confirmed.
Finally, we also identified a few weak emission features at 1017.8 Å, 1021.6 Å, 1064.1 Å and 1072.1 Å for which we could not find any reliable identification.
Turning to the other emission lines observed with HST, we found that the Si III emission line displays a double-peaked profile, very similar to that of C III* (Fig. 1). The time-lag between the two spectra does not allow a direct comparison of the two line profiles. However, if in the C III* line the blue and red emission peaks have nearly the same intensity, the red peak in Si III line is clearly more intense than the blue one. The two profiles are however asymmetric, and in the Si III line especially the redshifted peak cannot be fit by a single Gaussian. The profile is rather compatible with a double-peaked emission with an additional redshifted peak. This suggests that the line could be formed in two different media.
Table 1: Emission line measurements. Note that the integrated fluxes are not corrected for reddening and the quoted errors are the 1 statistical errors.
|Figure 2: The H I Ly line in the STIS spectrum. The dotted line is the reflected profile which provides an estimate of the interstellar contamination to the emission line.|
|Open with DEXTER|
The H I Ly line is very broad with wings extending from -970 km s-1 to more than 1700 km s-1. The profile is a type III P Cygni, according to Beals (1951), with a strong red emission component and a secondary blueshifted emission peak separated by a deep and broad absorption feature (Fig. 2). At the time of the STIS observation no other absorption component was visible in the profile. We note that part of this broad absorption feature is due to interstellar gas along the line of sight which blurs the exact profile. According to the analysis of the local interstellar medium structure (Lallement & Bertin 1992), the predicted velocity of the local interstellar medium in the direction of the star is about -25 km s-1. This value is in agreement with the radial velocity of the reddest line of the N I triplet lines at 1134.98 Å which is observed in absorption in the FUSE spectra. A rough estimate of the contribution of the interstellar gas to the H I Ly profile can be obtained by reflecting the red wing of the profile about the zero velocity. As may be seen in Fig. 2, the interstellar absorption cannot be responsible for the whole absorption feature. The bluest part of the profile is thus evidence for the presence of wind. We measured a wind terminal velocity of km s-1 in this blueshifted feature. In spite of the high variability of the hydrogen emission lines, this result is consistent with the analysis of H and H Balmer lines made by Pogodin (1994). The author indeed reported night-to-night variations with a main absorption feature slightly blueshifted with a velocity between -50 to -80 km s-1, and a secondary one, blueshifted by about 200 km s-1.
Table 2: Fe II fluorescent lines in the FUSE spectrum of HD 163296. The lines labeled with (a) are the transitions identified in HD 104237 by Harper et al. (2001).
We suspect the presence of the H I Ly in the FUSE spectrum, as a weak and broad emission in the wings of the strong airglow emission. However, the geocoronal emission hides the stellar emission and prevents an analysis of the stellar line profile.
|Figure 3: Temporal variation of the C III* emission line profile: the FUSE spectrum rebinned by 8 pixels (2001/04/27, solid line) and the STIS spectrum (2000/07/22, dotted line).|
|Open with DEXTER|
To investigate the behaviour of each line, we measured the radial velocity of the emission line wings and the total line flux of the various lines (Table 1). For the FUSE observations, we derived the emission line parameters from the co-addition of the two segments of the same channel in order to increase the signal-to-noise ratios. During the short time-lag between the two FUSE observations, no clear variation of these FUV lines occurs, either in shape or in strength, within the instrumental flux accuracy. Indeed, the observed variations in the lines flux exceed the statistical error bars (Table 1); however, the absolute values of the flux given by FUSE are subject to uncertainties due to, for instance, channels misalignment and drift of the target within the spectrograph slits. As found from calculating the integrated flux in each individual sub-exposure, this uncertainty is about 15% and prevents us from concluding about the variation of the intensity of the emission lines on a time scale of 2 days.
However, the comparison of the FUSE and STIS spectra of the C III* multiplet shows significant changes in line profile: although the wings of the line remain identical and the global profile is double-peaked, the intensity of the blue peak changes (Fig. 3). In addition, a weak, slightly redshifted emission is clearly present in the 2001/04/27 FUSE observation indicating that at least part of this ion emission originates from the same region as the C III resonance line. Also, the similarities between the C III* and Si III line profiles show evidence of a common region of formation for these two ions.
|Figure 4: LiF2a FUSE spectrum of HD 163296 recorded on 2001/04/27. The main emission features identified as Fe II fluorescent decays following excitation by H I Ly are labeled at the top.|
|Open with DEXTER|
These strong and broad emission lines, with various profiles, probe layers above the photosphere with temperatures up to 350 000 K for O VI. All of the line profiles are asymmetric but appear less complex for increasing temperature. The type III P Cygni profile of the H I Ly line and, most likely, of the C III resonance line, are consistent with an outflow signature.
Conspicuous sharp emission features are observed in the FUSE spectra which correspond to transitions from high Fe II levels. Such phenomena have already been reported for HD 104237, another Herbig Ae star (Harper et al. 2001) and in fact the FUSE spectra of both stars are very similar in this spectral range. It is difficult to disentangle the emission and the absorption spectra in this spectral region where numerous absorption lines from the stellar photosphere or from the CS/IS gas occur, as seen in Fig. 4. The process of identification in our spectra was thus aided by comparison with the HD 104237 FUSE spectrum, in which fluorescent lines of Fe II were identified by Harper et al. (2001), who used an approximate line formation model. The wavelengths of the emission features with the proposed identifications are given in Table 2. Most of these lines correspond to Fe II transitions from the terms , and of the odd configuration . A few transitions occurring from upper levels which lie very close in energy as the term are also identified, as well as some from terms of the configuration which is close to 100 000 cm-1. All the identified lines correspond to decays to the term of the ground level or the sub levels of the first excited level: with energy ranging from 1872.56 to 3117.46 cm-1. According to Johansson & Jordan (1984), these upper levels can be populated from the levels by photons with wavelength within 3 Å of the rest wavelength of H I Ly . As seen in Fig. 2, the H I Ly wings extend from -3.9 to 6.9 Å and thus the flux available in the line must be sufficient to photo-excite levels around 10 eV in this star.
We observe that the average of the centroids for all the measured lines is redshifted by about 30 km s-1 with respect to the stellar velocity. Although the blending of the lines as well as the uncertainty in the exact position of the photospheric flux prevent us from deriving accurate line widths, we measured line widths from 70 km s-1 up to more than 170 km s-1. This suggests that the Fe II comes from a spatially extended medium where turbulence or inhomogeneities may be responsible for these widths.
Fluorescent excitation in O I leading to emission lines in the near IR and in the UV resonance triplet at 1304 Å have been reported for the star (Felenbok et al. 1988). The authors suggested pumping by H I Ly of the ground level to excited upper levels as a likely excitation mechanism for this ion. As described in Sect. 3.1, the H I Ly appears present in emission. Although the exact shape of the line could not be derived, we measured broad wings extending from -620 km s-1 up to roughly +380 km s-1, that is enough to allow the O I excitation to take place. It confirms that the H I Ly line is the likely pumping mechanism of O I.
Considering the spectral signatures observed in the HD 163296 spectrum, it is clear that the star presents a number of similarities with AB Aur. The latter is considered as the prototype of a subset of Herbig Ae stars called the P Cygni subclass (Catala et al. 1986) and has been the subject of a number of detailed analyses that give a good understanding of its close environment. HD 163296 shares with AB Aur several spectral characteristics which are interesting to consider in the analysis:
However, as in the case of AB Aur, such a model cannot account for all of the spectral characteristics, especially for the line profile of over-ionized species like O VI or N V (Bouret et al. 1997). The presence of such emission lines arising in transitions of highly ionized species indicates the presence of hot circumstellar matter, up to K, located close to the stellar surface. The large width of the lines and their asymmetry appear consistent with the presence of more than a single component. The variety of profile types also argues in favour of a composite origin. Another mechanism is thus required to explain at least part of the observed emission.
We thus propose and discuss in the following two plausible mechanisms able to generate the high temperature regions required to explain these new observed emissions.
The first mechanism is magnetic accretion onto the star which involves a circumstellar disk interacting with a magnetically active star. This model is frequently invoked to explain the main CTTS's observational characteristics: strong and broad emission lines, often with time-dependent variations, and an excess emission in the continuum (see Bouvier 2003). With a dipole-like configuration of the stellar magnetic field, the matter from the disk is channeled along the lines of force and reaches the stellar surface at supersonic velocity near the circumpolar regions. The impact of the accreting material onto the stellar surface produces a strong shock, which can appear as a hot spot or a ring, depending the magnetic field configuration. As a result, the gas is heated in the pre- and post-shock regions to temperature of a few 106 K (Lamzin 1998). Overionized species are created in these regions, close to the star and a strong continuum excess is also produced. Theoretical studies predict that the resonance lines of C IV, N V and O VI form in the pre-shock and post-shock regions (Lamzin 1998). The resulting line profiles display two emission peaks: one with its centroid velocity close to the stellar one and a second one at a velocity close to the free-fall velocity (Lamzin 2003b,a). The integrated profiles are thus asymmetric, redshifted, with a total width not exceeding the free-fall velocity, which for CTTSs is typically less than 400 km s-1.
For Herbig Ae stars, this mechanism is however not so well established due to the lack of clear observational constraints and specific theoretical developments. In particular, two points deserve more investigations: the observed width of the emission lines and the continuum excess expected from the impact of circumstellar material onto the star which veils the stellar spectrum. For HD 163296 using and (Bouwman et al. 2000), the free-fall velocity is about 630 km s-1, a value well below the total width of the observed emission lines, which is typically greater than 1000 km s-1 (Table 1). Note, in this context, that even in CTTSs the situation is similar for the well-studied Balmer lines (Muzerolle et al. 1998): the observed large width of the profiles is then explained by the presence of an additional mass outflow driven by open field lines, close to the corotation radius (Beristain et al. 2001).
Concerning the veiling phenomenon in Herbig Ae stars, we lack the necessary observational support as no systematic study has been carried out. AB Aur is the only star for which detailed investigations in the visible show an absence of a noticeable veiling effect (Böhm & Catala 1993). However, this non detection does not definitely rule out the magnetospheric accretion scenario for Herbig Ae stars. Indeed, on the one hand, the veiling effect is more difficult to highlight in stars hotter than the T Tauri stars as, due to the temperature inferred, it should be indistinguishable from the stellar photosphere. On the other hand, the configuration of the accreting column relative to the observer and the physical conditions of the gas could also make the detection difficult. Clearly, this point needs to be further investigated for this class of stars.
An appealing alternative process is the wind confined model proposed by Babel & Montmerle (1997) to explain X-ray emission from Ap and Bp stars. The advantages of this model are that it does not require a disk massive enough to accrete but a mass outflow from the stellar surface, together with, as in the previous scenario, a large scale intense magnetic field. Following the model developed by Babel & Montmerle (1997), the wind from the polar regions is confined and structured in the closed magnetosphere. The two stellar-wind streams from both magnetic hemispheres collide in the magnetic equatorial plane, producing a strong shock. As a result, a large and extended post-shock region where the gas is heated to temperatures up to a few million of Kelvin is created. In addition, the cooling of the material leads to the formation of an ionized disk plasma at the magnetic equator, in corotation with the star in the inner region. In such a scenario, the X-ray emission, but also the over-ionized species, should originate in the post-shock region or in the so-called equatorial disk. Turbulence in regions extending over a few stellar radii with also a large gradient in temperature may participate in the broadening of the line but rotation is the dominant process. In such a scenario, the variations in the mass-loss should naturally act on the post-shock region and consequently on the intensity of the emission lines. Note that such a model does not require a radiatively-driven wind.
In such a framework, we can provide a rough estimate of the magnetic
field intensity. Assuming a dipolar magnetic field, the Alfvén
is the distance, in stellar radius, where the kinetic energy
balances the magnetic pressure of the dipole:
|Figure 5: Observed profile of the Mg II k line (full line) from the IUE archives and a synthetic profile (dashed line) computed with ETLA, using = 7.3 yr-1.|
|Open with DEXTER|
The presence of a stellar magnetic field in Herbig Ae stars is controversial on both theoretical and observational point of views. Indeed, as these stars lie on radiative tracks, they are not expected to possess outer convective zones which is necessary to sustain a solar-like dynamo. However, alternative sources of magnetic field generation are possible and have been considered by different authors. For instance, the dynamo effect may result from a shear mechanism due to differential rotation which generates a magnetic field (Lignieres et al. 1996; Tout & Pringle 1995). In that case, the efficiency of the generating process is expected to decrease in time leading to the decay of activity signatures and X-ray emission in the corresponding main sequence stars. Another possibility is that some Herbig Ae stars are the progenitors of the Ap-type stars and therefore possess a fossil magnetic field, essentially dipolar. On the observational side, HD 104237 is the only Herbig Ae star for which a detection of magnetic field has been reported (Donati et al. 1997). However, with magnetic field intensity less than 1 kGauss associated with such highly rotating stars, the non-detection by previous surveys does not rule out this possibility and today this issue remains open. Clearly, a higher sensitivity is mandatory and in that field, new instruments such as ESPaDOnS (Manset & Donati 2003), the new generation of spectropolarimeter for CFHT, will provide new insights on this critical point. It is worth mentioning that a value such as the one we derived for the intensity of the magnetic field in HD 163296, is well within the expected performances of ESPaDOnS.
The magnetically confined wind hypothesis offers the appealing possibility to build a consistent picture of the close stellar environment. It provides a natural explanation for the presence of highly ionized species, the rotational modulation observed in some spectral lines, as well as the X-ray emission recently detected for HD 163296 (Catala et al., in preparation). Indeed, the variability of spectral lines may simply result from the rotational modulation of a magnetosphere when the magnetic axis is tilted with respect to the stellar rotation axis.
A better description of the structure of these hot regions above the photosphere requires a complete modeling of the lines which we intend to present in a forthcoming paper. Also a new set of observations in the FUV range would allow investigations of the temporal behaviour of the different lines and subsequently a better constraint on their formation processes.
We thank J. Linsky for his careful reading of an earlier version of the manuscript and his valuable comments and suggestions. This work is based on data obtained for the Guaranteed Time Team by the NASA-CNES-CSA FUSE mission operated by the Johns Hopkins University. Financial support to US participants has been provided by NASA contract NAS5-32985. J. C. Bouret is pleased to acknowledge the CNES for financial support.