2. Spectroscopic analysis
V 3903 Sgr was observed between 1989 and 1992 with the 1.6 m telescope and coudé spectrograph at Pico dos Dias Observatory (PDO), National Laboratory for Astrophysics (LNA-CNPq) at Brasópolis-MG, Brazil. From 1989 to 1991 an EEV CCD (386 576 m square pixels) P8603S chip was used with a model #1 (serial no. 48) camera and acquisition program (AT1, version 3.1) from Wright Instruments Ltd. The mean dispersion was 18.1 Å/mm (0.397 Å/pix, corresponding to 230 Å covered in each frame), with a projected slit width of 1.40 pixel. Average exposure times were from approximately 10 min at 6600 Å to 20 min at 4400 Å. The grating (Milton Roy Co.) used in this observing period has 600 lines/mm blazed for 8000 Å (). The observations in 1992 were done with a 1800 lines/mm holographic grating (Jobin Yvon) blazed from 3000 Å to 7500 Å and with an EEV CCD 05-20-0-202 (770 1152 m square pixels, grade 0) chip with UV coating at a mean dispersion of 5.84 Å/mm (0.131 Å/pix, 150 Å covered in a frame), with a projected slit of 1.37 pixel. All spectra of 1992 were exposed for 20 min. We reduced all spectra by using a C computer program implemented by Vieira (1991, 1993) using the optimum extraction algorithm by Horne (1986).
After analysis of the results from several individual lines, as described by Andersen (1975), the He lines listed in Table 2 were found to give consistent velocities. The diffuse He I lines have been found to give systematic errors of 10% in mass for other OBA-type binaries (e.g. Andersen et al. 1980), but the lack of lines in the short-wavelength region and the orbital phase distribution of the spectra forced us to use them. However, as we are using a significant number of sharp lines, the effect mentioned above should be small and will be neglected. As a matter of fact, starting from the adopted solution (Table 4) and using only the present observations obtained with the sharp lines, the final elements remained practically unchanged, with only the errors becoming larger (by approximately 20%) due to the smaller number of points.
Table 2. He lines measured in V 3903 Sgr
2.1. Radial-velocity curves and the mass ratio
After reduction and calibration of the spectra, radial-velocity measurements were made by (least-squares) fitting double lorentzian and/or gaussian curves as described by Myrrha (1991, procedure used in LZ Cen, Vaz et al. 1995, and V906 Sco, Alencar et al. 1997, also) in order to find the observed center of the lines of Table 2. Our (PDO) spectra do not cover more than 230Å in each frame and it is difficult to use cross-correlation functions (e.g. Hill & Khalesseh 1991), as would be preferable. However, we did use cross-correlation techniques to obtain the last point of Table 3 (CTIO echelle spectra), which is in very good agreement with the rest of our data, showing that we did take correct account of the effects of blending. The fits were done on the normalized spectra, after determination of the continuum. Fig. 1 shows part of the normalized spectrum of V 3903 Sgr for the phases and .
Many interstellar lines are seen in the spectra, as shown in Fig. 1; the single pair of Na I interstellar lines appear at a velocity of -1.8 0.1 km/s. V 3903 Sgr is possibly the illuminating star of the bright nebula IC 4685 (Hirshfeld & Sinnot 1982, 1985), and it is not unexpected that we detected so many interstellar lines. On the other hand, it is surprising not to have found more than one resolved component of the Na I pair, due to the position of the system in the galactic plane (Table 1) and its distance to the Sun ( 1500 pc, see Table 10). Probably this is due to the fact that in the direction of the Galactic Centre, all the motion is transverse and not radial, and so any interstellar velocity will be small (Stickland 1996).
The measured radial velocities are shown in Fig. 2 and in Table 3, where the last point is one measurement obtained with the 1.5 m telescope at CTIO equipped with the bench-mounted echelle spectrograph (see Sect. 2.2). The columns 4 and 7 of Table 3 (labelled "w") indicate if the observations were used (1) or not (0) in the solutions. Excepting the last point, which was determined using a different instrument and procedures, all the observations of Table 3 were obtained through a homogeneous process and those which received zero weight were blended and/or underexposed. The proximity effects on the measurements are modeled as described by Wilson and Sofia (1976) in the WD code used to analyse the light curves (Sect. 3.3). Velocity corrections for the observed phases have been computed with the final model parameters adopted for V 3903 Sgr (Sect. 3.3.2) and were included in the (O-C) columns of Table 3 (columns 5 and 8, but are not included in the observed data, columns 3 and 6).
Table 3. Radial velocity observations of V 3903 Sgr.
The adopted spectroscopic solution is given in the last column of Table 4 and is shown as a solid line in Fig. 2. The solutions of the spectroscopic elements were performed both with a C program (written by NCS) based on the harmonic "Wilsing-Russell" method (Wolfe Jr. et al. 1967) and using the FORTRAN program SBOP (Etzel 1985) which allows solutions with the methods of Lehmann-Filhes (1894) and Sterne (1941), also. The solutions were performed with four different procedures: the Wilsing-Russell and Sterne methods applied to each component separately, and solving for both components simultaneously with Lehmann-Filhes method; the fourth is the WD method, used in Sect. 3.3.2 to solve simultaneously the uvby colours and the radial-velocity curve of each component. Our results with these different methods agree very well with each other and we adopt the simultaneous solution for both components with the Lehmann-Filhes method, as described below.
Table 4. Spectroscopic solutions for V 3903 Sgr. Adopted final orbital elements are in the last column.
In Table 4 we present solutions (Lehmann-Filhes applied to both components simultaneously) performed only with the non-zero weight observations of Table 3 (solutions 1 and 2). We applyed this method to Niemela & Morrison observations alone (also shown in Fig. 2) and reproduced their published solution (1988), which agrees well with solution 2 of Table 4. Therefore we decided to combine our data (Table 3) with those by Niemela & Morrison (1988), generating the solutions 3 and final. In solutions 1 and 3 we left e, the orbital eccentricity and , the angle of periastron passage, free to be adjusted. Table 4 gives for solutions 1 and 3 (free ) as the time of periastron passage, while in the circular orbit solutions (2 and final) is the time of conjunction, given by the adopted ephemeris (Sect. 3.1). Even though the formal errors of both e and are small enough to make the value significant, we notice that the values of in solutions 1 and 3 are very different, indicating that the orbital eccentricity may be an artifact of the data. Therefore we assumed a circular orbit, what is supported by the photometric observations, as described in Sect. 3, which give no indication of an eccentric orbit. Niemela & Morrison (1988) also assume a circular orbit in their solution for the system. Although the (1 obs.) for the adopted solution is the largest of Table 4, it is compatible with the intrinsic error of our observations (20 km/s).
The adopted solution of Table 4 yields the mass ratio , essentially the same result given by the WD model in the simultaneous solutions of Sect. 3.3.2. This value agrees with and improves Niemela & Morrison's (1988) determination, .
2.2. Rotation rates and luminosity ratios
We are in debt to Dr. R.D. Mathieu, who very kindly took some spectra of V 3903 Sgr and rotational standard stars with the 1.5 m telescope of Cerro Tololo Interamerican Observatory (CTIO) and the bench-mounted echelle spectrograph described in Casey et al. (1992). The observations were made in June 26, 1994 and were reduced by Dr. N.B. Suntzeff using IRAF, which is gratefully acknowledged. The widths (FWHM) of the same lines in V 3903 Sgr were interpolated between those in the rotational standard stars (Slettebak et al. 1975) HR 6175 (O9.5V, =400 km/s, see below), HR 6165 (B0V, 10 km/s), HR 5953 (B0.3IV, =150 km/s) and HR 6462 (B1Ib, =230 km/s). The resulting mean rotational velocities for the two components of V 3903 Sgr are =230 23 km/s and =170 17 km/s, confirming measurements based on the same rotational standards made with the PDO's equipment. As the dispersion of CTIO echelle spectra is significantly higher than those for the traditional Coudé ones taken at PDO, we adopt the result above for the rotational projected velocities.
It should be mentioned that the CTIO echelle spectra showed asymmetries in most of the detected lines of both HR 5953 ( Sco) and HR 6175 ( Oph), especially conspicuous in H and H lines. Oph is known to have line-profile variability caused by non-radial pulsations (Walker et al. 1979, Janot-Pacheco et al. 1991). Its rotation rate was determined by Reid et al. (1993) to be =400 20 km/s. Oph is a prototype giving its name to the subclass of stars with similar line-profile variability behaviour (Balona 1990). For Sco the asymmetries observed suggest the occurrence of double lines. If this duplicity is real, it indicates velocity differences as large as 110 20 km/s. Naturally, the status of standard for rotation velocities of this star should be revised if this duplicity is confirmed.
The ratio of the equivalent widths of the relatively unblended He I lines of Table 2 for both components was measured yielding . The temperatures of both stars in V 3903 Sgr are significantly different and, therefore, it is not straightforward to associate the ratio of equivalent widths directly with the luminosity ratio. However, it is expected that the intensity of the He I lines decreases with increasing temperature and that the ratio of equivalent widths above is in fact an upper limit for . Using He II at 4686 Å and the CTIO spectra, we obtain , which should be a lower limit for the actual value of .
© European Southern Observatory (ESO) 1997
Online publication: April 6, 1998