The behavior of AB Aur during the MUSICOS 96 campaign is far more complex than expected. It is beyond the scope of the present paper to give a complete picture of the atmospheric and envelope structures leading to such a complex variability. However, we make below a series of remarks based on the results presented above, that may play a decisive role in building a more complete model for this star.
The features that need to be accounted for in the modelling are summarized below:
1a. the photospheric lines have a very asymmetric profile, with a blue absorption component variable in position, modulated with a 34 hr period, and a red component in emission, at a more or less constant velocity, variable in amplitude, with a possible periodicity near 43 hrs.
1b. the shape of the photospheric lines is not exactly the same for lines of different ions.
2a. the He I D3 line exhibits two components, evolving differently.
2b. the blue emission component of the He I D3 line occurs at a velocity which is modulated with a 45 hr period; the modulation appears as a double wave for the first part of the campaign, then as a single wave.
2c. the red component, sometimes in emission and sometimes in absorption, does not vary significantly in velocity, but its intensity shows strong variations, with indications of a periodicity near 45 hrs.
2d. a dramatic event occurs about 65 hours after the start of the campaign, during which a strong wide absorption appears, centered at rest wavelength, both in the He I and the photospheric lines. This event coincides with a strong widening of the H blue absorption component. It also coincides with the change from a double wave into a single wave modulation for the centroid velocity of the blue absorption component of the He I D3 line.
2e. the variations of the equivalent width of the the He I line are correlated with those of the photospheric lines (see Fig. 18); both sets of equivalent width variations are dominated by the behavior of the red half of the lines.
3a. the H line, presenting a P Cygni profile, is strongly variable, mostly in its blueshifted absorption component; on the other hand, the redshifted emission component is almost constant.
3b. there is no clear periodicity in the observed variations of H, although pseudo-periodic variations of its blue absorption component are suggested.
5.1. The photospheric lines
Strong perturbations of the photospheric layers are needed to explain the peculiar shape of the photospheric lines. In the following we discuss two of the three components appearing in the photospheric profile after subtracting a rotational profile (see Sect. 4.1): the blueshifted absorption component with a variable velocity modulated with a 34 hr period, and the low-velocity redshifted emission component with a mean velocity of +25 km s-1. There is an additional high-velocity redshifted absorption component with a mean velocity of +95 km s-1, but its existence is doubtful as argued in Sect. 4.1, and we shall not discuss it in detail here.
Since these two components behave differently, we first conclude that they originate from different parts of the stellar surface.
Let us first discuss the low-velocity redshifted component. The velocity of this component shows very little variation during the campaign. It is positive, but remains smaller than the rotation velocity. We conclude that this component probably originates from the polar region, since its velocity would be modulated by the star's rotation if it originated from lower latitudes, and that it is associated with a downflow occurring in this region. If the star is seen nearly edge-on, this implies large downward velocities. With the value of the inclination angle derived below, , we need velocities of the order of 70 km s-1.
This component is variable in intensity. Whether or not these intensity variations are periodic could not be ascertained on the basis of the data presented here, but we do have some indication of a possible periodicity near 43 hrs.
We are finally led to conclude that one or several downflows onto the pole of AB Aur must be present, with velocities of the order of 70 km s-1, and which must be hotter than the underlying unperturbed photosphere, in order to produce an emission in the average photospheric line. In order to account for the variations in intensity of this component, we must assume that these downflows are variable in filling factor, density and/or temperature.
The velocity of the blue component behaves differently, with a periodic modulation between -40 and -100 km s-1, with a 34 hr period. We may reasonably assume that this corresponds to the rotation period of the star's photosphere. Note that this period is very close to that observed in the Ca II K line by Catala et al. (1986b), 32 hrs, also interpreted as the rotation period of the star. More recently, Gahm et al. (1993) have also reported some indication for a 35 hr period in photometric data of AB Aur, although with a low confidence level. Thus the variations of the blue component may be due to a peculiar structure of limited area at the photospheric level or at the base of the wind.
If the rotation period of the star is indeed 34 hrs, and if its projected rotation velocity is 85 km s-1, as we have adopted earlier, then we determine an inclination angle i of the order of , assuming a radius of 2.5 , adequate for AB Aur (van den Ancker et al. 1997). However, this value is accompanied by a large error bar, considering the uncertainties on the projected rotation velocity, the modulation period and the stellar radius. We find that any inclination between and would be compatible with our data.
Important velocity fields must be present in the line formation region to account for our observations. Structures with no velocity fields (such as temperature or abundance inhomogeneities for instance) would produce perturbations crossing the line profile from blue to red, and extending as far to the red as to the blue. This is not what we observe here.
Similarly, horizontal velocity fields, such as e.g. meridional flows, cannot account for the observed blue component variations. Indeed, with the high inclination angle of the star's rotation axis implied by our data, such horizontal flows, if located at high latitude in the visible hemisphere or near the equator, will produce perturbations reaching almost as far to the red wing as they do to the blue wing, which is not observed; if located in the partly invisible hemisphere, such flows are seen only for a small fraction of the time, implying a very complex flow structure to account for the fact that the blue absorption component is seen in all of our spectra.
Conversely, we find that the observed blue component variations are compatible with radial flows in the photosphere. These radial flow structures must occur at high latitude, otherwise they would again produce perturbations extending significantly to the red for a significant amount of time. We have calculated that, for an inclination angle of the star's rotation axis of 70o as derived earlier, outward velocity fields of the order of 150 km s-1 at a latitude of 80o are necessary to explain the peculiar variability observed for the blue component.
Our observations can therefore be tentatively explained by:
In the outflowing region, we may expect drastic changes in the density compared to the unperturbed stellar photosphere. Such density inhomogeneities will strongly affect the line formation, which may explain the large differences in the profiles of the photospheric lines of various species. As far as the downflow is concerned, we expect large temperature enhancements where the material hits the stellar pole, which again will affect strongly the formation of the lines. This latter effect may explain why the low-velocity redshifted component is seen in emission against the unperturbed photospheric line.
The possible 43 hr period seen in the intensity variations of the
low-velocity redshifted component, if real, remains unexplained in the
framework of this interpretation, although we may argue that it is
somehow linked to the wind modulation which we describe in the next
sections. However, as argued earlier, this apparent period may be
simply due to the coincidental succession of two strong absorption
events, at t=65 hrs and t=150 hrs.
5.2. The He I D3 line
This line is normally very weak in the spectrum of an A0V star like AB Aur. Its presence as a strong line, whether in absorption or in emission, indicates the existence of heated layers above the photosphere.
As for the photospheric lines, the He I line shows two components behaving quite differently. Since the red component does not vary significantly in velocity (V 100 km s-1), we conclude that it may also correspond to material falling inward onto the stellar pole. We note in this respect that the equivalent width of the He I line is well correlated with that of photospheric lines, the equivalent width variations being dominated by the behavior of the red half of the lines. We therefore conclude that the red components of the He I and photospheric lines originate from the same phenomenon, probably an accretion to the pole affecting both the upper layers where it creates the red component of the He I line and the photospheric layers where it creates the low-velocity red component of the photospheric lines. For an inclination angle of 70o, we derive a velocity of the order of 300 km s-1 for the part of this downflow producing the He I red component.
This accretion must be variable to account for our observations. In particular, a strong modification of its characteristics must have occured near hrs, to give rise to the dramatic event that we have already described earlier. The exact nature of this modification is not known for the moment.
The blue component of the He I line seems clearly modulated with a 45 hr period. It is also very tempting to interpret this periodicity as due to the stellar rotation. In the framework of this interpretation, the puzzle is to understand the period difference between the photospheric (34 hrs) and He I (45 hrs) line modulations. This result is reminiscent of the behavior of the Ca II K and Mg II h&k lines reported by Praderie et al. (1986) and Catala et al. (1986b), with periods of 32 hrs (Ca II K) and 45 hrs (Mg II h&k). An interpretation of these previous results was given by these authors in terms of a wind structure of fast and slow streams controlled by a surface magnetic field. According to this model, the Ca II K line is modulated by this structure, which rotates with the rotation rate of the star. Further away in the wind, the structure is destroyed, and the lines originating from the outer regions of the wind, like the Mg II lines, are modulated by the rotation of the envelope at that distance.
This model was acceptable because the Mg II lines are formed at great distances from the stellar surface (up to 50 R*, see Catala et al. 1984), where indeed the stream structures are likely to have merged due to shocks at the interface between fast and slow streams. With the present set of data, we show that the He I line too is modulated with a longer period than the photosphere. The similarity between this period and that of the Mg II lines is striking. However, unlike the Mg II lines, this line cannot easily be formed at such great distances from the star, since it originates from very excited levels. The model presented in the past therefore cannot apply here. Moreover, since the period determined for the He I line is so close to that exhibited by the Mg II lines in the past, we may even argue that these new data contradict the previous model.
We are led to conclude that:
(i) either the wind is indeed structured into fast and slow streams rotating with the star, but in this case the different periods found in the data correspond to the rotation at different latitudes on the star; we shall call this "the equatorial wind" model; or
(ii) the wind does not originate from the stellar surface, but e.g. from a circumstellar disk, and the different periods exhibited by the different lines correspond to the rotation of the disk at different distances from the star; this model will be called "the disk wind" model.
We shall now examine these 2 different models in the light of the data presented in this paper. In both models, the photospheric lines are modulated by the rotation of a stream structure located at high latitude, and by a downflow of matter onto the stellar pole. The same downflow is also responsible for the red component of the He I D3 line.
5.3. The equatorial wind model
Clues for the existence of an equatorial wind in AB Aur and other Herbig stars were presented by Pogodin (1992). The best evidence is the repeated change from a P Cygni to a single emission profile at H, which is interpreted as a change in the latitude span of an equatorial wind: when the wind has a large opening angle, and extends to colatitudes below the inclination angle, the line of sight intercepts the expanding regions, and a P Cygni profile is formed. When this is not the case, the absorption component disappears, and H is left with a single emission.
If we adopt this idea, then we must consider that all wind lines are formed in a flow that originates from the equatorial regions of the star. Since we have many independent proofs of the presence of a strong stellar wind in AB Aur, through the various P Cygni profiles exhibited by some of its spectral lines, it is natural to assume that the blue component of the He I D3 line is formed in the wind, although pure accretion may also be capable of producing profiles like those of this line, as demonstrated by Hartmann et al. (1994) and Muzerolle et al. (1998). Bouret & Catala (1999) have recently produced a quantitative model in which this component is indeed formed in the inner region of the stellar wind. The H line is also obviously formed in the wind. Now this radial flow may be structured in fast and slow streams, which creates in the line profiles a modulation with the rotation period of the star's equatorial regions. This phenomenon would be at the origin of the modulation seen in the blue component of the He I D3 line. This component is likely formed near the base of the wind, where the density is high enough to form this high excitation line (Bouret & Catala, 1999). Much further out in the wind, where the H absorption is produced, the streams may have merged, and the line may vary due to the rotation of these remote regions. In that case, no clear periodicity remains, since the H line is formed in an extended region with a broad range of rotation periods. The resulting variations may thus exhibit a complex behavior, as observed, with several inhomogeneities affecting the line formation at various distances from the star and creating line profile disturbances with different time scales.
The fact that the absorption component of H varies much more than its emission component (relatively) is easy to understand in the framework of this model: the absorption component indeed is formed in the regions of the wind that are projected on the stellar disk from the observer's point of view, which corresponds to a small volume. Any variation of the physical characteristics of this small volume will translate into a variation of the absorption component. On the other hand, the emission component is formed over the whole wind, which encompasses a very large volume. Therefore, the wind inhomogeneities are averaged out in this component.
The details of the He I line variations also suggest that the stream structure in the wind includes twice as many streams at the beginning of the campaign than at the end of the campaign, the change occurring near t=65 hrs.
If this interpretation is correct, then it indicates that the equatorial regions rotate more slowly than the high latitude regions, by as much as 20%. This is not totally surprising since the stellar wind is exerting a braking torque at the stellar surface, more efficiently at the equator than at the pole. Depending whether the resulting horizonthal shear is capable or not of quickly redistributing the angular momentum at the stellar surface, a strong latitudinal differential rotation will exist or not, with the pole rotating faster than the equator. This point deserves further modelling.
Such a high level of differential rotation must lead to a strong shear of the magnetic field lines, which must be entirely wound up in a time frame of approximately 150 hrs. If the photospheric structure responsible for the observed modulation is controlled by a magnetic field, we therefore expect this structure to evolve significantly on that time frame. This may actually be supported by the obvious change in the variability pattern of the photospheric line blue component at about t=160 hrs (see Fig. 7). The same argument may be applied to the He I line, formed in the wind, and also presumably modulated by the rotation of azimuthal structures in the wind, and may provide a qualitative explanation for the high level of "intrinsic" variation on top of a simple periodic modulation stressed by Böhm et al. (1996), as well as for the abrupt change in the modulation pattern exhibited by this line near t=65 hrs (see Fig. 14).
Finally, we have checked that such a differential rotation would not affect significantly the shape of the photospheric profile.
5.4. The disk wind model
Alternatively, we may assume that the wind is originating from a circumstellar disk, rather than from the stellar surface. A model of this kind has been recently proposed for massive young stars (Drew et a. 1998), in which an accretion disk is reprocessing the stellar luminosity, and the radiation pressure produced by this reprocessed flux is sufficient to lift up disk matter which is subjected only to the reduced gravitation from the disk. Although AB Aur is probably not luminous enough for this model to be directly applicable, it may be argued that rotation and magnetohydrodynamical effects can be added to radiation pressure for driving a disk wind in this star.
Accretion disk models have been proposed for Herbig Ae/Be stars in order to reproduce the observed IR excesses (Hillenbrand et al. 1992; Lada & Adams 1992). The mass accretion rates required by these models to be consistent with the observed IR energy distributions are very high, typically in the range yr-1. This poses serious problems, and is in particular contradicted by the absence of optical veiling for the Herbig Ae/Be stars (Böhm & Catala 1993; Ghandour et al. 1994). On the other hand, the present photospheric data, showing a very distorted profile compared to a classical rotationally broadened profile, may hide a significant optical veiling. We must therefore consider this issue as still open.
The existence of optically thick accretion disks around most Herbig Ae/Be stars is also seriously questioned by the absence of asymmetry for the [O I] forbidden lines (Böhm & Catala 1994). Optically thick disks would indeed mask the receding parts of the stellar winds in which these lines are formed, which would result in a global blueshift of the line. Even though Corcoran & Ray (1997) have recently published a list of Herbig stars showing [O I] asymmetries, it remains true that many Herbig stars, including AB Aur, do not show any such asymmetries. On the other hand, Hirth et al. (1994) have suggested that the [O I] lines in Herbig Ae/Be stars may originate from a thin layer at the surface of an accretion disk, and therefore may not show any asymmetry even in the presence of disks. In this case, the emitting region must be very extended to account for the observed fluxes in the forbidden lines, and this prediction was recently contradicted by Böhm & Hirth (1997) on the basis of longslit spectra.
Recently, high angular resolution millimeter observations of Herbig Ae stars have revealed the presence of elongated structures around some of them, including AB Aur, on scales of several hundred AU (Mannings & Sargent, 1997). Such structures can be the signature of extended circumstellar disks around these stars. In particular, the evidence for keplerian rotation of molecular material in the elongated structure around AB Aur and HD 163296 makes a compelling case for circumstellar disks. However, these observations trace the stellar environment at several hundred AU from the star, and do not tell us much about the structure of the circumstellar material in the regions forming the lines we are studying here, which lie typically within 0.1 AU from the star.
All these results show that a high level of ambiguity still remains concerning the existence of thick accretion disks in the immediate environment of Herbig stars, and in particular around AB Aur. We also note that no observation precludes the existence of an optically thin disk.
The model we are discussing here therefore includes a rotating accretion disk around AB Aur, and a wind originating from this disk.
In the framework of this model, we do not need to invoke surface differential rotation to account for the different modulation periods derived from our data. The modulation of the photospheric lines is still due to the rotation of the inhomogeneous photosphere, with a period of 34 hrs. The periodic modulation of the lines formed in the wind is now due to the presence of inhomogeneities in the disk, producing local changes in the wind characteristics near its base (changes in velocity, density, and/or temperature). The spectroscopic signatures of these local wind changes are then modulated in velocity by the rotation of the disk structures to which they are anchored. We therefore expect to see a periodic modulation of the centroid velocity of the line components due to these inhomogeneities, with a period equal to the rotation period at the distance of formation of the line, , and with a total amplitude given by , where is the rotation velocity at a distance r from the star center, and i the angle between the line of sight and the normal to the disk plane.
For the blue component of the He I line, we find a period of 45.1 hrs, and a total modulation amplitude of about 200 km s-1. We have estimated i to about , consistent with the value of derived by Mannings & Sargent (1997) for the inclination of the circumstellar disk of AB Aur. These results would therefore imply that this component would be formed in the parts of the wind originating from the disk at about 1.6 R*, which may perhaps correspond to the inner boundary of the accretion disk.
The disk rotation implied by our data (rotation period of 45.1 hrs at 1.6 R*) is slower than keplerian, indicating that the accretion disk must lose a significant fraction of its angular momentum in the vicinity of the star, by a yet unknown mechanism. This conclusion is surprisingly opposite to the situation prevailing in the case of the classical T Tauri stars, where the angular momentum tends to be transferred from the disk to the star instead (Bouvier et al. 1993).
As in the case of the equatorial wind model, the H line is formed further out in the wind, and in regions which are more extended than those forming the He I line. The fact that the absorption component of H varies more than its emission component is easily understood, for the reason detailed in the previous section. We may also expect the observed complex behavior for the H line variations because the inhomogeneities responsible for them are distributed over regions of the disk with different rotation periods.
5.5. Downflows onto the stellar pole
In both types of models, we need to invoke downflows onto the pole of the star to account for the red components of the photospheric and He I lines. These downflows need to have velocities of 300 km s-1 where they produce the He I red component, and of 70 km s-1 in the region of formation of the photospheric red component. Since the region of formation of the He I red component is probably located above that of the photospheric red component, we conclude that the downflows must be strongly decelerated as they approach the star's surface, possibly in a shock.
The intensity of the red components of the photospheric and He I lines seems to be modulated with a period near 45 hrs, suggesting that the characteristics of the polar downflows (accretion rate, temperature, geometry) may be modulated with this period. However, as detailed at the end of Sect. 4.3, this conclusion is not very reliable, so that we do not propose below a detailed interpretation of this phenomenon.
The origin of the material falling inward onto the stellar pole is not clear. It may be linked to the expanding regions at high latitude that we evidenced from the blue photospheric component, the stellar photospheric material first rising from a localized high latitude region, then falling back down on the pole, presumably being channelled in a magnetic canopy.
The material falling down on the stellar pole may also originate from the equatorial wind, channelled by a magnetic field. On the other hand, in the disk wind interpretation, the matter falling on the stellar pole may also originate from the disk itself in a magnetic accretion column.
Whatever the origin of these downflows, they must give rise to shocks near the stellar pole, where they will dissipate large amounts of kinetic energy. Rough estimates of the densities required to explain the red emission component of the He I line indicate mass accretion rates of the order of at least a few yr-1 (Bouret & Catala, 1999). The red component of the He I line indicates that these downflows have a typical velocity near 300 km s-1 in the regions forming the He I D3 line, and therefore carry a flux of kinetic energy of typically = a few 1031 erg s-1. The dissipation of this kinetic energy may provide an adequate explanation to the emission seen in the photospheric lines, and also partly explain the presence of superionized species like N V observed in HST/GHRS spectra (Bouret et al. 1997).
© European Southern Observatory (ESO) 1999
Online publication: April 28, 1999