Astron. Astrophys. 351, 619-626 (1999)

2. Observations and results

2.1. Angular separation measurements

A variety of astrometric observations of different kinds and of different quality have been collected for Gliese 570BC over the years. The close pair was first resolved by Mariotti et al. (1990) who obtained infrared 1D speckle observations on three occasions and derived a first visual orbit. We refer to their paper for the description of the instrument and the observing and data reduction procedures. HMcC later measured Gl 570BC once with a 2D infrared speckle instrument. For readers' convenience, we list these published measurements in Table 1 together with our new 2D speckle and adaptive optics (hereafter AO) observations.

Table 1. Angular separation measurements.
Notes:
Observation dates are listed as offsets relative to Julian Day 2400000.
References:
1) Mariotti et al. 1990; 2) Henry & Mc Carthy 1993; 3) This paper, Kitt Peak, 2D speckle imager; 4) This paper, ESO 3.6m, COME-ON adaptive optics system; 5) CFHT, CIRCUS 2D speckle imager; 6) ESO 3.6m, ADONIS adaptive optics system; 7) This paper, CFHT, PUE'O adaptive optics system.

Two measurements were obtained in February 1991 and April 1991 using the speckle mode of the 2D infrared imagers then installed respectively at the KPNO 4.2m telescope and the CFHT 3.6m telescope. Each imager was designed to permit acquisition of exposures short enough to substantially freeze the seeing under standard atmospheric conditions ( 50 to 100 ms) in bands H and K , independently of the overhead due to read-out time and data transfer time. Several sequences of a few hundred such short exposures were obtained, alternating every few minutes between the source and a nearby unresolved star (usually Gl 570A) used as a point spread function (PSF) calibrator. The whole observation took one hour or less.

In principle, this observing procedure allows an almost simultaneous PSF calibration, and consequently estimates the visibility modulus with 1 to 5% accuracy. We used a software package written specifically for this type of data reduction by E. Tessier (Tessier et al. 1994). It produces an unbiased visibility estimator for the source and then extracts the binary parameters and their estimated variance from this visibility. The actual detector scale and position angle (P.A.) origin for those observations were calibrated from observations of the astrometric binary  Aqr (Heintz 1989).

Most of the new measurements were obtained at the 3.6m Canada-France-Hawaii Telescope (CFHT) on top Mauna Kea, using the CFHT Adaptive Optics Bonnette (Arsenault et al. 1994, Rigaut et al. 1998) and two different infrared cameras (Nadeau et al. 1994, Doyon et al. 1998). Delfosse et al. (1999b) provide a detailed description of the observing procedure, which we therefore don't repeat here. Two observations were also obtained with the ESO 3.6m telescope (La Silla, Chile) AO system COME-ON+ (Rousset & Beuzit 1999), then ADONIS (Beuzit et al. 1997), equipped with the SHARP-II infrared camera, using a similar observing procedure. This COME-ON+ observation has been published in Mariotti et al. (1991). For recent CFHT measurements the corrected point spread function obtained from the AO system was synthesized from simultaneous recordings of the wavefront sensor measurements and deformable mirror commands, as described by Véran et al. (1997). For pre-1997 measurements it was instead obtained from observations of a reference single star of similar R-band magnitude. Astrometric calibration fields such as the central region of the Trapezium Cluster in the Orion Nebula (McCaughrean & Stauffer 1994), were observed to accurately determine the actual detector plate scale and position angle (P.A.) origin. Uncertainties on these parameters do not apreciably contribute to the overall separation error for the small separations in the Gl 570BC system.

The separation, position angle and magnitude difference between the two stars were determined using uv plane model fitting in the GILDAS (Grenoble Image and Line Data Analysis System) software, as well as with the deconvolution algorithm described by Véran et al. (1999), coded within IDL. With approximate initial values of the positions of the two components along with the PSF reference image, the fitting procedures gave as output the flux and pixel coordinates of the primary and secondary. Application of the astrometric calibrations then yields the desired parameters.

2.2. Radial velocity measurements

All radial velocity measurements are listed in Table 2 (available in electronic form only). Most of them were obtained with the two CORAVEL radial velocity scanners (Baranne et al. 1979) on the Swiss 1m telescope at Observatoire de Haute Provence (France) and on the Danish 1.54m telescope at La Silla (Chile). The earlier data were previously published by Duquennoy & Mayor (1988) and the system has been regularly observed since then. Gl 570BC is relatively bright for the CORAVEL instruments (V=8.09, Leggett 1992), but both stars are poor matches to the fixed K0III correlation mask. They therefore only produce shallow correlation dips (of which examples were displayed in Duquennoy & Mayor (1988)) and their velocities are measured with typical precisions of respectively 0.7 and 4 km s^-1, instead of the usual CORAVEL precision of 0.3 km s^-1. The measurements of the M3V secondary in particular were at the limit of the CORAVEL capabilities. During the orbit adjustment discussed below they were found to have sizeable systematic errors at phases where the two profiles are even slightly blended. This didn't measurably affect the derived orbital elements since these noisier measurements carried essentially no weight anyway, but they were nonetheless ignored in the final solution. For clarity, they are also not plotted in Fig. 1.

Fig. 1. Radial velocity orbit of the Gl 570BC system. Filled triangles represent measurements of the primary star, and filled squares measurements of the secundary stars. The ELODIE mesurements can be distinguished from the CORAVEL ones by their much smaller error bars.

Table 2.

Over the last three years, we have obtained considerably more accurate measurements with the ELODIE spectrograph (Baranne et al. 1996) on the 1.93m telescope of Observatoire de Haute Provence and the CORALIE (Queloz et al. in preparation) spectrograph on the Swiss 1.2m Euler telescope at La Silla (Chile). These echelle spectra were analysed by numerical cross-correlation with an M4V one-bit (i.e. 0/1) mask, as described by Delfosse et al. (1999b). Radial velocities were initially determined by adjusting double gaussians to the correlation profiles. Those however had systematic phase-dependent residuals during the orbital adjustment, which were particularly large (400 m s^-1 rms) for the fainter secundary star. Upon analysis, they were found to originate from the low level wiggles in the correlation profile of the primary: while the core of this profile is well described by a gaussian function, its baseline doesn't drop to zero as a gaussian would, but instead keeps oscillating at the 0.2% level, about one tenth of the depth (2%) of the secondary star's dip. Depending on its position on this oscillating background, the measured velocity of the fainter star could therefore be incorrect by up to 1 km s^-1. Similar errors of course also affected the measured velocities of the primary star, but they were smaller by the square of the relative depths of the two correlation dips, or a factor of 16. Thanks in part to the excellent stability of the spectrograph, the shape of the correlation profile for a given star is very stable. We could therefore obtain an excellent estimate of the wings of the intrinsic correlation profile of each star, by averaging all profiles of the system, after aligning them at the measured velocity of the star and blanking all pixels within two profile widths of the velocity of the other star. The residuals of a gaussian adjustment to these average profiles are then subtracted from all correlation profiles, at the measured velocity of each star. This decreases the fluctuation level in the profile baseline to a level of 0.05%. The radial velocities measured by a double gaussian fit to these corrected profiles have typical accuracies of 40 m s^-1 for the primary and 100 m s^-1 for the fainter secondary. These residuals are twice larger than would be measured for single-lined spectra with equivalent S/N ratios and correlation dip parameters, but show no systematic phase dependency. They should thus cause no systematic errors on measured parameters.

© European Southern Observatory (ESO) 1999

Online publication: November 3, 1999
helpdesk.link@springer.de