Astron. Astrophys. 324, 505-522 (1997)

10. Discussion and conclusion

We present, in this paper, the radial velocity and velocity dispersion measurements deduced from integrated-light spectra obtained with CASPEC at the ESO 3.6-m telescope at La Silla, Chile, for a large sample of Galactic and Magellanic globular clusters. These measurements can be affected by large statistical uncertainties if the integrated light through the spectrograph slit is dominated by a small number of bright stars. We present a very extensive set of numerical simulations, which indicates that statistical uncertainties on the velocity dispersion values are small, i.e., comparable or smaller than the measurement uncertainty, when the absolute magnitude of the integrated light within the measurement sampling area is brighter than an absolute V magnitude of about -4.5 (see Fig. 12). This result suggests that statistical uncertainties on our (and any other) velocity dispersion estimates are important for most Galactic clusters, and small in the case of Magellanic clusters.

However, the three following facts suggest that our simulation results may be somewhat pessimistic. First, there is a reasonably good agreement between our integrated-light measurements and the measurements based on radial velocities of individual stars found in the literature. Second, our results agree well with previous independent integrated-light up measurements, achieved over different sampling areas. Third, on average, our values are consistent with those expected from King models, given the cluster total absolute magnitude and structural parameters, as is the case for the values derived from the radial velocities of individual stars. This last point rules out the possible presence of a large bias in our measurements of Galactic clusters resulting from the light in our sampling area coming predominantly from a few bright giants.

In short, our simulations provide upper limits for the uncertainties due to the domination of too small a number of bright stars on the values. Consequently, when using our measurements one should keep in mind that the quoted statistical uncertainties may be overestimated. Because of the larger distances to the Magellanic globular clusters, all but one of the measurements obtained for the Magellanic clusters are fully reliable, with relatively small uncertainties, which nevertheless could also be overestimated.

We shall present, in a following paper (Dubath et al. 1996), a determination of the structural parameters for the Magellanic clusters, together with their estimates based on the present measurements.

No significant deviation from the predictions of the King models, or from the Virial theorem, are observed in the globular cluster fundamental plane, as already discussed by several authors (e.g., Pryor & Meylan 1993) in the case of Galactic clusters. This shows that the globular cluster is, within the accuracy of the data, constant from cluster to cluster, and that Magellanic clusters do not have a dramatically different from those of Galactic clusters. As a consequence, the fundamental plane correlations can also be used to test our results. For example, the remarkably small scatter of the Magellanic cluster data around the relation expected from King models observed in Fig. 16 is another indication of the good quality of our measurements for these clusters.

The results of the simulations can also be used to predict the accuracy of any measurement from integrated-light observation. In order to get a reliable measurement, the absolute V magnitude of the integrated light should be brighter than -4.5. For a cluster of known brightness profile, this can be translated into a minimum aperture (i.e., a minimum number of stars) needed to get a meaningful measurement.

© European Southern Observatory (ESO) 1997

Online publication: May 26, 1998

helpdesk.link@springer.de