As we have seen dramatic changes in the photometry of HD 45677 on different time scales one could expect to see such changes in the spectroscopic behaviour as well. Indeed, spectral variations have been observed for the lines of MgII, HeI and slightly in CaII (Swings et al. 1980). Because of the high radial velocities measured for the HeI and MgII lines, 150-200 km s, those lines are thought to be formed close to the stellar surface. For the H and H lines more dramatic variations are seen as well, most significantly in the violet parts of the profiles and on time scales ranging from hours up to days (Merrill 1952; Swings et al. 1980). Similar variations in the H profiles were already reported on a longer time scale on plates taken between 1923 and 1947 (Swings & Struve 1943; Merrill 1952). Although Swings & Allen (1971) did not report any correlations of line variations with detected photometric changes, we can give a more exact view at the moment since more spectroscopic observations overlap with the different time scales of the photometric variations.
To study the long time scale variations we use as a starting point the spectral descriptions of Merrill (1928) and of Swings et al. (1980). We have searched for additional spectra which are listed in Table 1 (click here). This summary is certainly not complete as probably many plate and CCD spectra are reported but not discussed, e.g. Jaschek et al. (1992), or unpublished, e.g. Swings (1995). Several of such spectra, as well as many others listed in Table 1 (click here), will be discussed in detail by Israelian et al. (1996). We refer to the spectra in the following sections by their number, the year of observation, as listed in Col. 1 of Table 1 (click here). Most historical known spectra are of low resolution. It is therefore only possible to make use of our own low and intermediate resolution spectra for these purposes. As our low resolution spectra are quite similar we have only presented the 1992e and 1992g in Fig. 4 (click here), together with the line identifications which will be discussed. For a similar reason the 1992b intermediate resolution spectrum is not shown in Fig. 5 (click here). However, from the presented H profiles clear changes are seen.
Investigating such short time scale variations, as also known from the literature, we use our monitoring data. We have obtained time series with a duration of 6 days by high resolution spectroscopy in two wavelength regions. Besides spectra centered on the H laboratory wavelength, with a resolving power of 50 000, we have monitored the NaID lines, with a resolving power of 55 000. The choice for this line is inspired by the clear detection of accreting cool material in the case of the Herbig Ae star UX Ori (Grinin et al. 1994). Furthermore, our setup includes the HeI line at 5876 Å and the [FeII] line at 5870 Å. The NaID and H time series were taken in October 1993. One other spectrum, in both regions, was obtained in January 1995. The spectra are plotted in Figs. 9a and 10a.
There is no spectral data that would illustrate changes on intermediate time scales in either low or high resolution. This will therefore not be discussed.
The existence of an extended gaseous shell of considerable density was first inferred from the appearance of nebular lines of [OI] at 6300 Å and 6363 Å and of [SII] at 4068 Å (Merrill & Burwell 1933; Swings & Allen 1971).
Contrary to spectroscopic variations as mentioned above, the forbidden lines, like [FeII], [OI], [NII], are stable; no clear changes are seen since the first detections. Probably those lines are formed in an extended region far away from the central star, see Fig. 13 (click here) of Swings (1973). The red [SII] lines were not detected and the [NII] lines very weak. As reported by Swings (1973), the FeII lines are double peaked and more dispersed than the [FeII] lines.
Because of the wide variety of the spectroscopic data in wavelength coverage and resolution we will discuss them by each group separately. In some cases a more detailed description will be given by Israelian et al. (1996).
Figure 4: December 1992 low-resolution spectra of HD 45677 with the most prominent lines identified
Figure 4: continued
Table: Identified emission lines in the low- and intermediate-resolution spectra of HD 45677, Figs. 4 (click here) and 5 (click here). Uncertain identifications are given between brackets
In the 1971 and 1992 spectra, H is seen in emission accompanied at the red side by the [NII] line. In the intermediate resolution 1992a and b spectra H is also accompanied at the blue side by the weaker [NII] line of the same multiplet. Both are in emission. It is also seen in most other spectra, like in 1989, in which H shows self-absorption. The velocities and line profiles measured in the 1989 spectrum are similar to those reported before, e.g. Merrill (1952). However, the H profile is very complex. We will discuss this further in Sect. 4.2.2.
In the 1992c-e spectra, as in the previous ones, the OI lines at 7774 Å and 8446 Å are very strong. The [SII] lines, which were not detected before, are weakly in emission (spectra 1992a and b). No differences can be seen for the FeII lines in the 1992 spectra and those observed before. The H and other blue hydrogen line profiles confirm the detections given by Swings (1973) and Swings et al. (1980). From 1964 to 1966 emission components appeared in the blue hydrogen lines, from H10 up to H5 (= H). In 1979 H shows central emission. In 1974, Swings (1974) reported that H exhibits a very strong P Cygni profile. In 1966 no emission is seen. Furthermore, the known strong iron emission lines were observed double peaked, separated by 30 km s, being the same value as reported above. In the 1992a and b spectra H is seen in emission with a central emission embedded in a photospheric absorption line, as the profile was observed in the 1971 spectra. In the 1992 intermediate resolution spectra we also note that the FeII lines are relatively broad, when compared to typical nebular lines as [NII], [SII] and [OI] in the same spectra. The FeII lines also hide traces of double peaked profiles with a separation of about 40-60 km s, which is somewhat higher than in the spectra taken before the photometric minimum. Although the FeII lines are relatively broad and the resolution of the 1992 spectra are too low to determine good velocities, most identified lines seem to be centered on a velocity of about 20 km s.
Figure 5: May 1992 intermediate-resolution spectra of HD 45677. The full wavelength range of the 1992a spectrum is shown, the 1992b spectrum is very similar and therefore not included. However, line profiles can be very different as shown in the lower plots for H
The properties of the described spectra are similar to those observed and discussed by Swings (1973) and Swings et al. (1980). Even the short time scale variations reported in these papers are seen in the 1992 intermediate resolution spectra. We have compared both of the spectra, 1992a and 1992b. The difference spectrum (not shown here) clearly indicates significant changes in the H line while all other lines seem to be stable. In the 1992b spectrum we clearly have a broad extra absorption at the blue side of the H spectrum, as well as two small absorption components at +4 and -60 km s. In the next section we will discuss the H profile in more detail.
High resolution spectra of HD 45677 are rare. A comparison of our data with other spectra, besides the hydrogen lines, is therefore not possible. However, we will discuss here the spectra mentioned in Table 1 (click here) of CaIIK (Fig. 6 (click here)) and [OI]6300 (Fig. 7 (click here)) including the MgI 6318.23/.75 Å doublet (Fig. 8 (click here)). The other high resolution spectra will be discussed in the next section.
A H spectrum from Doazan et al. (1991) shows a strong double peaked emission line with the red component much stronger than the violet peak similar to the 1989a spectrum and the ones reported before (e.g. Merrill 1928). However, the peak separation in the H spectra by Doazan et al. (1991) and by Merrill (1928) is much bigger (150 and 170 km s, respectively) than in all the H spectra ( km s). This difference supports our interpretation of the double-peaked structure of the hydrogen lines as being due to self-absorption rather than being due to two separate regions of hydrogen emission. For our 1993 H spectra the two peaks differ less in strength than in the 1988b H spectrum, while in the 1992a and b H spectra we even have . This indicates a strong variability of the hydrogen lines as noted previously by Swings et al. (1980).
The CaIIK line is seen with the absorption center at 20 km s. At the blue and red side of this narrow line (the maximum velocity dispersion, at the continuum level, is less than 50 km s) weak emission is seen. This profile is identical to the ones from 1932 to 1947 (Merrill 1952) and similar to the one mentioned by Swings (1973). The [OI]6300 and the MgI doublet are in emission with the central peak at about 20 km s. Although the velocity dispersion of the [OI] line is higher, about 90 km s, than the CaII line, it is much lower than the 250 km s of the MgI doublet. Note also that the latter is asymmetric with extra emission at the blue side of its profile extending to 150 km s, whereas in the red part the velocity goes up to about 100 km s.
From the velocities we conclude that the MgI line is formed closer to the central star than the [OI]6300, which must be formed in an extended low density region. Although MgI will be destroyed close to the central star there must be some outflow of this material. Also the emission in the CaII resonance line shows the presence of some cool material. When distributed in a disk this material must be very dense as seen by the strong self-absorption line.
In comparing the red spectra of 1971b and 1973, although the 1971b spectrum is of poor quality, we see in the 1971b spectrum HeI 6678 Å in emission but not in the 1973 spectrum. In 1973 FeII 7712 Å and OI 7774 Å are in emission but in 1971b weak or probably not. In the red part of 1992c and in 1992d the FeII 7712 Å, [FeII] 7732 Å and OI 7774 Å are seen in emission, similar to the 1973 situation. In 1993 these lines are also visible in emission, only the OI line was weaker than it was in 1992d. In the 1989b spectrum the OI 8446 Å is strong in emission, also the Paschen lines and many FeII lines are in emission, very similar to the 1973 situation. The same wavelength region of the 1992d spectrum looks exactly like in the previous ones, with the exception of the one detecting the decreasing strength of OI 7774 Å.
The 1988 spectra look exactly as the ones of HD 87643 and CD-4211721, two other well known B[e]-stars. Only H-lines are in emission, no other features are seen. No other corresponding spectra are available for comparison.
The high resolution sodium and H line profiles are presented in Fig. 9 (click here)a and Fig. 10 (click here)a, respectively. Their characteristics will be presented separately. A more qualitatively modeling on these spectra will be presented in another paper (Israelian et al. 1996).
It is striking to see that in all spectra the NaID lines are significantly in emission with a strong absorption core at 20 km s, similar to the observed CaIIK line. The emission components of the lines is somewhat stronger than the line, as usual, but combined with the relatively strong self-absorption component the profile looks somewhat different. The interstellar component is not visible and is probably blended in the broad circumstellar absorption feature.
Between the spectra taken in 1993 and the one of 1995, no dramatic changes are seen, although the strong telluric waterlines could give a wrong impression. However, comparing the time series spectra to each other, using the difference between a spectrum and the one of 09/10/1993, it is clear that day by day variations are significant in both the NaID and the HeI lines (Fig. 9 (click here)b). Note the changes of the emission strength at the blue sides of the NaID lines. However, the wings of the red sides of these lines are remarkably stable as they all extent up to 147 5 km s for the and up to 140 5 km s for the components. Similar profiles, but with more dramatic changes, are observed by Israelian et al. (1996) on a somewhat longer time scale.
Figure 6: High resolution CaIIK spectrum of HD 45677, rebinned to heliocentric velocities
Figure 7: High resolution [OI]6300 spectrum of HD 45677, rebinned to heliocentric velocities
Figure 8: High resolution spectrum of the MgI 6318.23 + 6318.75 Å doublet of HD 45677, rebinned to heliocentric velocities
In the spectra of 13 and 14/10/1993 the blue emission wings are quite strong and extend up to -99 km s and -65 5 km s, respectively, while on 10 and 11/10/1993 these blue parts are depressed, and show even weak narrow absorption components at -18 1 km s (not to be confused with the telluric waterline as seen in the 16/1/1995 spectrum). On 09 and 12/10/1993 these weak lines are located at -11 1 km s. An extra wide but very weak absorption is also seen on 11/10/1993 at -75 2 km s.
Figure 9: a) High resolution NaID spectra of HD 45677, rebinned to heliocentric velocities. b) The same spectra but taken as the difference compared to the one of 09/10/1993
The blueshifted weak absorption components seem to be dynamical on a daily time scale, but can also appear for a few days, with relatively large changes in their velocities. In our time series this absorption slowly disappears and in the last two spectra, 13 and 14/10/1993, only extra emission is seen.
Several H spectra were taken in the same nights as the NaID spectra. The one taken on 11/01/1994 is somewhat saturated, as is the 09/10/1993 spectrum slightly overexposed. As we noticed only small daily changes between the NaID spectra, the same seems to be the case for the quasi-simultaneously taken H profiles, see Fig. 10 (click here)a. The H profiles are quite complex. A strong self-absorption component at +11 1 km s is visible in all the spectra. However, it is not located at the center of the emission profile.
We have determined the center of the emission profile by the fitting of a Gaussian to the observed profiles and to modified ones, in which the absorption components are deleted by interpolation. The results of both fitting procedures are almost identical. The emission peak seems to shift from +37.5 1 km s for the 09/10/1993 profile to +40.5 1 km s for the 12/10/1993 spectrum, with a daily step of +1 km s. This is almost 10 km s higher than the radial velocity of the H lines before 1928 (Merrill 1928). The velocity offset between the H emission and systemic velocity from +17.5 to +20.5 km s indicates gaseous movements and acceleration towards the central star. It would be interesting to monitor HD 45677 daily for a longer time-interval to reconfirm such a mechanism.
Figure 10: a) High resolution H spectra of HD 45677, rebinned to heliocentric velocities. b) The same spectra but taken as the difference compared to the one of 09/10/1993. Because of saturation effects the spectrum of 11/01/1994 is not shown here
In a plot of the differences between the spectra (Fig. 10 (click here)b), the daily movement is seen as the sharp extra emission on the red side. Furthermore it is clear that an extremely broad and very weak component in the H profile changes from a slightly redshifted absorption (10/10/1993), in fact less emission as it is the difference spectrum with the one taken on 09/10/1993, to extra blueshifted emission (11 and 12/10/1993), note the velocity dispersion close to the continuum level in these spectra. Again, this daily change visible as small V/R-variations, with an extremely broad velocity dispersion, is either due to the underlying photospheric line or to variations in the wings of the emission profile. Anyway, it seems that infalling gaseous material is followed by extra outflow. This process must be studied further, but it will be related to a region close to the stellar surface.
Another important detection in the H profiles are the weaker absorption components and very weak dips, besides the very strong absorption line at +11 km s. We detected these components at -80, -37, +49 and +90 km s. For the weaker ones the error is about 5 km s. The redshifted components are more prominent.
Table 4: SED fit results for HD 45677 at different V magnitudes
Figure 11: Typical observed (squares) and extinction-corrected (circles) SEDs for HD 45677. Also shown is the Kurucz model for a B2 V star, fitted to the extinction-free SED