Astron. Astrophys. 321, 379-388 (1997)

2. Observations and data reductions

We obtained 12 high-resolution integrated-light spectra of 10 globular clusters belonging to M 31. The observations were made September 9-11, 1994 (see Table 1), with the Hamilton echelle spectrograph (Vogt 1987) operated at the coudé focus of the 3-m Shane telescope of the Lick Observatory. The detector was a thinned, backside-illuminated TI CCD, with 800   800 pixels of 15   15 µm each, and with a readout noise of about 6 electrons.

Table 1. Radial velocities and velocity dispersions of M 31 globular clusters

During the first night, a relatively wide entrance slit of 0   0 was used to match poor seeing conditions. A slit of 5   0 was used during the second and the third nights. As a consequence, the instrumental resolution, as estimated from the FWHM of the emission lines of thorium calibration spectra, was slightly lower during the first night (31 000 or 9.7 km s^-1) than during the second and the third ones (35 000 or 8.6 km s^-1). Both slits used have widths larger than the apparent half-light radii (see Table 2) of all our clusters.

Table 2. Mass-Luminosity Ratios for M 31 Globular Clusters

Table 1 gives the date of the observations, the exposure time (Texp) and the signal-to-noise (S/N) ratio of the cluster spectra. During each observing night, many radial velocity standard stars were also measured. Spectra of a thorium-argon lamp were taken 6-8 times through each of the three nights.

The spectra were reduced with INTER - TACOS, a new software package developed in Geneva by D. Queloz and L. Weber. The first-night spectra contain 50 useful orders ranging from 4800 to 8300 Å, and those obtain the second and third night contain 49 orders ranging from 4600 to 7500 Å. The different orders do not overlap, i.e., the echelle spectra exhibit holes in the wavelength coverage. The orders are never rebinned, nor merged together.

The shifts in velocity between the different thorium-argon lamp spectra taken during one particular night are always smaller than 1 km s^-1. However, a velocity shift as large as 5 km s^-1 is observed between the second and the third nights. As a first step, for each night, one wavelength solution is computed from a thorium-argon spectrum, and applied to all the spectra taken during that particular night. The cluster spectra also contain numerous strong night-sky OH and O2 emission lines, because of the long exposure time on relatively faint objects. The residual velocity zero-point shift of each globular cluster spectrum is then derived by measuring ¹ the mean shift (in velocity) of the emission lines compared to their accurate rest wavelengths (taken from Osterbrock et al. 1996). As a second step of the wavelength calibration, all the globular cluster spectra are corrected according to these velocity zero-point shifts.

The reduced spectra are cross-correlated with an optimized numerical mask, used as a template (see Dubath et al. 1990 for the details of the cross-correlation technique). In short, our template contains lower and upper wavelength limits for each line from a selected sample of narrow spectral lines. For a particular radial velocity, the value of the cross-correlation function (CCF) is given by the sum, over all spectral lines, of the integral of the considered spectrum within the lower and the upper line limits shifted according to the given velocity. The CCF is thus built step by step over the velocity range of interest. The CCF is not affected by the fact that the spectra are composed of a succession of independent orders, with holes in the wavelength coverage. The CCF is a kind of mean spectral line over the approximately 1400 useful lines spread over the spectral ranges of the different orders. The lower and upper limits of the lines are computed from a mix of observed and theoretical high-resolution spectra of K2 giants, in such a way as to optimize the CCF (see Dubath et al. 1996a).

When the number of considered lines is large enough ( several hundreds), this cross-correlation technique produces a CCF which is nearly a perfect Gaussian. A Gaussian function is fitted to each derived CCF in order to determine three physical quantities: (1) the location of the minimum equal to the radial velocity , (2) its depth D, and (3) its standard deviation , related to line broadening mechanisms.

Fig. 1 displays the CCFs of the integrated-light spectra obtained for the M 31 globular clusters in our sample.

Fig. 1. Cross-correlation functions (CCFs) of the M 31 globular cluster spectra. The dots represent the CCFs themselves, and the continuous lines represent the fitted Gaussians. Cluster designations are from Battistini et al. (1980).

We have a total of 32 measurements of 11 standard stars, mostly giant stars of spectral type G8 to K4, collected during the same observing run. The comparison of the standard-star radial velocities with reference values provided by CORAVEL measurements (Mayor, private communication) shows that the instrumental radial velocity accuracy is of order 0.5 km s^-1 and that the zero-point shift between the two datasets is not significant. In the case of the M 31 clusters, the accuracy of the radial velocity zero-point is very probably even better owing to our use of the night-sky emission lines to calibrate the zero point of each cluster spectrum.

The mean value of the widths () of the CCFs obtained for our sample of standard stars is 5.8 0.3 km s^-1 for the first night, and 5.5 0.2 for the two others. This difference is the direct consequence of the different slit widths used. We see no dependence of the CCF width on spectral type over the range considered here (G8 to K4). We also know that the CCF width does not depend on the star metallicity from previous measurements of a large number of various types of metal-deficient stars (see Fig. 6 of Dubath et al. 1996a).

© European Southern Observatory (ESO) 1997

Online publication: June 30, 1998
helpdesk.link@springer.de