2 Observations of the HAeBe binaries

2.1 The sample

The observed Herbig Ae/Be stars were extracted from Table 1 of Thé et al. (1994) catalogue. We obtained high resolution spectra for 42 objects with $m_V\,<\,11$, consisting of our principal HAeBe sample. This sample represents about 70% of the HAeBe candidates with $m_V\,<\,11$ from Table 1 of Thé et al. (1994) catalogue. Most of the remaining stars were not observed either because they were already well studied or because they are strong photometric variables and out of our limit of our observing capabilities at their minimum brightness.

A sub-group of other HAeBe candidates listed in Tables 2 to 5 from the catalogue of Thé et al. (1994) were also studied (sample T2-T5). They were included in our survey because they could be observed in parallel to our main program. Those stars were observed not only to test the presence of a companion, but also to precise, when possible, their spectral type or to test their belonging to the Herbig Ae/Be stellar class.

We address in this paragraph the possible bias by the so-called Branch effect . Branch (1976) pointed out that a magnitude-limited sample favors the detection of double-lined spectroscopic binary (SB2). In other words, some binary stars could have been selected because their total magnitude is below the magnitude limit, whereas individual stars are fainter than the m_V limit. Such systems should be removed when establishing the binary frequency of the sample. However, in our case with high mass primaries, unless the mass ratio is close to unity, the secondary component of a spectroscopic binary system contribute to few percents of the combined flux. We consider the example of the double-lined spectroscopic binary system TY CrA, adopting the most recent physical parameters determination by Casey et al. (1998): the primary (late-B type) has an effective temperature $T_\mathrm{I}$ of about 12000 K and a radius $R_\mathrm{I}=1.8$ $R_{\odot}$, while the secondary (late-G type) has $T_\mathrm{II} \approx 4\,900$ K and $R_\mathrm{II}=2.1$ $R_{\odot}$. The flux ratio between the two stars is

$$\frac{F_\mathrm{II}}{F_\mathrm{I}}= \frac{T_\mathrm{II}^4\times R_\mathrm{II}^2} {T_\mathrm{I}^4\times R_\mathrm{I}^2}=3.8\%.$$

Thus, in a HAeBe binary system with a primary of spectral type A or B and a lower-mass companion, the primary will be responsible for most of the observed flux. Therefore, the Branch effect is not significant in such cases. When both members of the spectroscopic HAeBe binary system have similar masses, and hence comparable luminosity, special attention should be paid to see whether or not one member of the system would have been in our sample or not.

Noticeably, we are faced with the large disparity in the location of the HAeBe type stars (some of them have a poorly determined distance). A distance limited criterion as considered by Duquennoy & Mayor (1991) is hardly conceivable here to select a sufficiently large sample and to determine a reliable binary frequency for Herbig Ae/Be stars.

2.2 Observing journal

The spectroscopic survey, initiated in 1994, was carried out in the two hemispheres, using three different instruments which characteristics and data reduction procedures are described hereafter.

2.2.1 Northern hemisphere

Spectra of northern hemisphere HAeBe stars were taken with ÉLODIE and AURÉLIE, two spectrographs of the Observatoire de Haute-Provence (OHP), south of France.

AURÉLIE is a high resolution spectrograph mounted at the Coudé focus of the 1.52m telescope. A detailed description is given in Gillet et al. (1994). The detector at the time of the observations was a Thomson double array and two different gratings were usually employed: #7 with a resolution $R \approx 38\,000$ at $\lambda=6\,500$Å and #5 (2nd order) with $R \approx 70\,000$ at $\lambda=6\,500$Å. Typical exposure time was 1 - 1.5 hour for our target stars, with a circular entrance hole of 3$^{\prime\prime}$ on the sky. A continuous light-source provided the flat field exposures to correct the instrumental response. The flat-field images were chosen to have a level similar to that of the science exposures and were repeated each night. After subtraction of the bias and dark current (measured routinely during each night), the science exposure was divided by an average of ten suited flat-field images. Thorium and argon lamps were used for wavelength calibration, and numerous exposures were taken each night to monitor the stability of the spectrograph. The final wavelength calibration accuracy is 2$\mathrm{km\,s}^{-1}$. The spectra were then transformed into the heliocentric rest frame and normalized to unity. Bad pixels or cosmic rays were removed by-hand. All the data reduction steps were performed with the ESO MIDAS software.

ÉLODIE, located at the 1.93m OHP telescope, is a fiber-fed échelle spectrograph. The detector is a 1024$\times$1024 Tektronics CCD. 67 orders are simultaneously recorded, giving in a single exposure a spectrum between 3906 and 6811Å, at a resolution of 42000. The optical layout as well as the reduction procedure can be found in Baranne et al. (1996). The fiber diameter is 2$^{\prime\prime}$ on the sky and thorium and science spectra were obtained separately. Sky background was estimated in the inter-order space. Typical exposure times range from 0.5 to 1 hour. The final velocity calibration is better than 1$\mathrm{km\,s}^{-1}$. Noticeably, a program at the telescope automatically performs the data reduction: no further work but heliocentric velocity correction and normalization to unity is required to have a set of homogeneous spectra.

2.2.2 Southern hemisphere

Southern HAeBe stars of our sample were observed with the CES (Coudé Échelle spectrograph) fed by the 1.4 m CAT telescope (La Silla Observatory, Chile). Most of the observations were made under remote control from the ESO headquarters in Garching bei München. From December 1994 to October 1995, the short camera configuration and the CCD#9 was used. Then we used the long camera and the new CCD#38 allowing a larger ($\approx\!60$Å) spectral coverage in the Li I 6708Å region. The resolving power was set to 60000. Typical integration time was 1 - 1.5 hour, the rectangular slit dimensions were ranging from 5 - 10$^{\prime\prime}$ $\times$ 1 - 2$^{\prime\prime}$. The data reduction procedure was identical to the one followed for the AURÉLIE data (except that the 2D- spectra were averaged perpendicularly to the dispersion).

The time distribution of the observations is a combination of high-frequency coverage (less than few days) and low-frequency coverage (greater than 1 year) over the final 3 years of the survey. Table 1 (only available in electronic form at the CDS) presents the observed stars.

Note: given the slit/fiber sizes and the extreme seeing values during the various observations (0.7 - 2$^{\prime\prime}$), we estimate that a binary separated by less than 1.5$^{\prime\prime}$ has been observed while integrating the flux from both components: such a system is thus considered "as a single target" in our spectroscopic observations. On the other hand, the primary HAeBe (brightest component in V) of a visual binary with separation greater than 1.5$^{\prime\prime}$ was independently observed (i.e. without integrating the flux of its companion), when seeing conditions allowed it.

2.3 Spectral analysis

We present here the two methods used to spectroscopically identify a HAeBe binary star.

2.3.1 Search for Li I 6708Å absorption

Martin (1994) quantitatively showed that the Li I 6707.8Å resonance doublet can be used to detect T Tauri companions of HAeBe stars. Indeed, in hot intermediate-mass stars, the Li I absorption line, extremely weak, is not detected, whereas in lower mass stars, Li I is detected (see Walter et al. 1988; Duncan & Rebull 1996; Jones et al. 1996). If the spectroscopic signature of this element is present in the spectrum of a HAeBe star, it reveals then the presence of a young lower mass and cooler companion.

2.3.2 Search for radial velocity variations

In order to monitor the radial velocity ($V_\mathrm{rad}$) variations, we mainly used the He I 5876 and 6678, Na I 5890 and 5895, Si II 6347 and 6371Å lines. The center of the lines was measured by fitting simple Gaussian functions: the errors of such measurements are of the order of 5 to 10$\mathrm{km\,s}^{-1}$, depending on the rotational velocity of the stars and its shape (if emission is also present and affects part of the photospheric line).

Note: although with ÉLODIE on-line cross-correlation spectroscopy is possible, we did not used this option: our stars, hot objects with $7\,000\,<\,T_\mathrm{eff}\,<\,30\,000$ K, display few lines in their visible spectrum, usually broadened by rapid rotation. Moreover, some lines may be filled-in by emission and show strong variations from night to night. Obviously, direct cross-correlation spectroscopy, as also proposed by Morse et al. (1991) for early-type stars, may not be appropriate for Ae/Be stars.