6 Discussion

Figure 10: Comparison of photometric and spectroscopic metallicity estimates. In the left panel the [MgFe] index is plotted against (V-I) colours (including the pseudo (V-I) colours derived from the spectra). Overplotted as solid line is the prediction for a 12 Gyr stellar population model by Maraston (2002, in preparation) and Thomas et al. (2002a). In the middle panel we show a comparison between an empirical calibration of (V-I) colour against metallicity (dashed line; Kundu & Whitmore 1998) and predictions from stellar population models (solid line, 12 Gyr; Maraston 2002, in preparation). In the right panel, photometric metallicities are calculated according to Kundu & Whitmore (1998), while spectroscopic metallicities are derived from the [MgFe] index in comparison to models of Thomas et al. (2002a) assuming a constant age of 12 Gyr. The data point at [Fe/H] $_{\rm spec~[MgFe]} = -2.25$ represents a globular cluster with line-strength weaker than the currently available model predictions and is therefore only an upper limit. The lines at the right hand side of the plot show the peak positions of the bimodal distributions of GCs observed by Kundu & Whitmore (1998, solid lines) and RGB stars observed by Elson (1997, dotted lines).

Before we discuss the implications of our results it needs to be noted that our sample of GCs in NGC 3115 is dominated by the bright end of the luminosity function. Our faintest object (Slitlet ID: 17) which we could use for the line-strength analysis is 0.9 mag brighter than the peak of the GC luminosity function. Furthermore, since this project was only a pilot study we observed only a small number of clusters over a limited region across NGC 3115. Although our data can give clear insights into the globular cluster system of NGC 3115 it is not necessarily a representative sample. The comparison with the Lick/IDS observations of MW and M 31 GCs needs to be viewed with caution since these samples cover a larger range in absolute GC magnitude.

Using their HST V, I photometry Kundu & Whitmore (1998) develop the following formation scenario for NGC 3115. The blue, metal poor clusters are formed with the halo/bulge component of the galaxy very early on. Then about $4\pm3$ Gyr later an unequal mass, gas-rich merger event forms the disk component and the associated red, metal rich clusters. The authors point out that the spatial distribution of the blue and red clusters are consistent with the halo/bulge and disk components respectively. Furthermore, there is evidence from optical imaging (Silva et al. 1989), that the disk is bluer and therefore perhaps younger than the halo/bulge component, consistent with the above outlined scenario.

In our spectroscopic study of the GCs in NGC 3115 we find that the clusters are consistent with being coeval at about 11-12 Gyr. There is perhaps a weak hint of the red clusters being younger, but by no more than $\sim$2 Gyr. This on its own would not rule out the scenario by Kundu & Whitmore (1998), however our estimates of the abundance ratios show that at least the population of red, metal rich clusters is not homogeneous. For these objects we find a range in abundance ratios from solar to about ${\rm [Mg/Fe]} \simeq 0.5$. In our small sample of limited spatial extend we do not find any clear trends of the chemical composition of GCs with the kinematics (see Fig. 3). The relative velocities of most of the clusters in our sample are consistent with rotation. A larger sample of more complete spatial coverage is needed to establish possible trends between abundance ratios and e.g., age, metallicity, spatial position and kinematics.

What we can however say is that the metal-rich clusters with solar abundance ratios must have been made out of well mixed material which incorporates the products of both SN II (the main producer of alpha elements) and SN Ia (the main producer of Fe-peak elements). Since SN Ia are somewhat delayed compared to SN II the solar abundance ratio clusters must have formed after the initial star burst in NGC 3115. There are many possible scenarios to explain the observed abundance ratio distribution, but it is hard to fit them into a simple picture of only two distinct formation events which create the red and blue globular cluster sub-populations (see Beasley et al. 2002a).

One scenario which we would like to put forward for further discussion is the following. The metal poor (blue), non-solar abundance ratio clusters are associated with the halo formation, and the metal-rich (red), non-solar abundance ratio clusters are formed together with the bulge as was similarly proposed for the Milky Way by Carney et al. (1990) and Wyse & Gilmore (1992). The metal rich, solar ratio clusters are then formed with the disk of this lenticular galaxy $\sim$1-2 Gyr after the initial star-burst perhaps in connection with a merger. This scenario would then require the disk to have also close to solar abundance ratios, which can be tested by observations of the integrated light. Furthermore, if this scenario is correct the spatial distribution and kinematics of disk GCs will be distinct from the population of halo and bulge GCs in NGC 3115. Future, larger samples of NGC 3115 GCs will be very valuable to explore the connection between disk formation and metal rich globular cluster formation.

More spectroscopic observations of GCs in nearby galaxies of various types will be very valuable to improve our stellar population models and learn more about the early star-formation epochs in early-type galaxies. However, the mismatch of the models and some observed indices demonstrates that we are also in urgent need for a new, high-quality flux-calibrated spectral library in order to exclude simple observational offsets. Only then can we make good progress with improvements on the input physics of stellar population models and their application to extragalactic objects.