The (U-V)₀ color histogram for 71 globular clusters in NGC 5128 indicates a bimodal distribution (Fig. 10). Note that the (U-V)₀ color distribution of the MW globular clusters does not show bimodality, while the one of M 31 does. The (U-V)₀color distribution of the MW globular clusters is in fact biased due to lack of U-magnitude and higher reddening values (only clusters with E(B-V)<0.5 are plotted) of a bulge (more metal-rich and redder) cluster. The blue peaks of the MW, M 31 and NGC 5128 globular cluster color distributions coincide. The blue cutoff in the three galaxies is quite sharp and is driven principally by the colors of blue horizontal branch stars. The red side of the color distribution is much broader. In M 31 it can be partially driven by differential reddening inside the host spiral galaxy, because the colors for M 31 clusters presented here were corrected only for the mean foreground reddening of E(B-V)=0.08. Examination of the stellar (U-V) vs. (V-K) diagram in the two fields in NGC 5128 indicates that the amount of reddening does not vary much in the halo of this giant elliptical (Rejkuba et al. 2001b). Harris et al. (1992) and Holland et al. (1999) also found more red (metal-rich) clusters in their NGC 5128 data.

The color distribution of the MW globular clusters in Fig. 10, being dominated by halo clusters, is representative of the GCS of our Galaxy as if observed from outside. The NGC 5128 clusters in Harris et al. (1992) and in this work belong as well to the halo population. The differences in the color (metallicity) distribution of clusters and stars in the halo of a spiral, like MW and a gE like NGC 5128, are reflected in Fig. 10 with a much larger metallicity spread in the gE galaxy.

The line overplotted on the color distribution of NGC 5128 is the best nonparametric kernel estimate (Silverman 1986) of the color distribution used to measure accurately the blue and the red peaks of the distribution. The blue and red peaks are measured at ( $U-V)_0=0.73\pm0.01$ and $1.28\pm0.01$ , respectively.

$\begin{figure} { \resizebox{\hsize}{!}{\includegraphics[angle=270]{10447f11.eps}} } \end{figure}$

Figure 11: The (U-V)₀ vs. [Fe/H] color-metallicity linear relation (full line) derived from the MW globular clusters with E(B-V)<0.5 (stars). The M 31 clusters with spectroscopic metallicities and accurate (U-V) colors plotted for comparison (squares) were not used for the fit. Typical errors in measured metallicities are plotted on in the box. Short dashed lines are the SSP models from Kurth et al. (1999) for ages of 5, 10 and 15 Gyrs, from blue to left, respectively. Each model spans metallicities from Z=0.0001 to 0.05

From the color-color diagram (Fig. 7) it is obvious that the globular clusters in NGC 5128 are older than 1 Gyr. Unfortunately, the models do not allow one to better constrain the ages (Fig. 7; see also Barmby & Huchra 2000). Assuming that the globular clusters in NGC 5128 have similar ages as the MW ones, one can use the Galactic globular clusters to calibrate the metallicity scale. It is known that for the very low and very high metallicities the color-metallicity relation cannot be expressed by a simple linear function, but extrapolation from lower metallicities towards the higher ones by using the higher polynomial fit might lead to unphysical results. Thus I prefered to make a linear fit with Gaussian errors in $\sigma{\rm (Fe/H)}=0.1$ in order to calibrate the metallicity as a function of intrinsic (U-V)₀ color. Only MW globular clusters from the W. Harris (1996) web list with E(B-V)<0.5 (equivalent to E(U-V)<0.82; Rieke & Lebofsky 1985) were used in the fit. The following color-metallicity relation is based on the Zinn & West (1984) scale and is valid only for the range of metallicities observed in the MW globular clusters (Fig. 11):

Although it is not obvious from the (U-V)₀ color distribution, but rather from the spectroscopically measured [Fe/H] distribution, the metallicity distribution of MW globular clusters is bimodal with the metal-poor and metal-rich peaks at -1.6 and -0.6 dex respectively (Harris 2000). These values are practically the same as the ones derived here for globular clusters in NGC 5128 from the (U-V)₀ colors. On the other hand, on the basis of Washington photometry of 62 globular clusters in NGC 5128, Harris et al. (1992) conclude that "in both the range and distribution of abundance, the NGC 5128 GCS is clearly different from the Galactic GCS and quite similar to that of the large elliptical NGC 1399''. They measured a mean metallicity of $\langle{\rm [Fe/H]}_{C-T_1}\rangle=-0.8\pm0.2$ dex, substantially more metal-rich than the mean value for the MW GCS. They, however, mention that the values exceeding ${\rm [Fe/H]}\sim -0.25$ are extrapolated, a fact that is also true in the metallicity calibration from (U-V)₀ color distribution. The extrapolation leads to large errors and a possible overestimation of abundances of the metal-rich clusters. Because of this I do not report here the metallicities of individual clusters.

Note also that I restricted the (U-V)₀ color to be within the range observed in MW and bulk of the M 31 globular clusters. In this way the sample was selected to be as clean as possible from the very compact and red galaxies. It is not surprising that the clusters with supra-solar abundances are not present, reducing the mean abundance of the sample. However, relaxing the red color cut and adding the 6 "good'' globular cluster candidates redder than ( U-V)₀=2.5 (filled symbols in Fig. 5) the mean of the color distribution shifts from $\langle(U-V)_0\rangle=1.05\pm0.45$ to $\langle(U-V)_0\rangle=1.22\pm0.75$ , corresponding to mean metallicities of $\langle{\rm [Fe/H]}\rangle=-1.1\pm0.1$ dex and $\langle{\rm [Fe/H]}\rangle=-0.7\pm0.2$ dex, respectively. The latter value is very similar to the mean metallicity derived from Washington photometry by Harris et al. (1992). In order to assess the real nature of the objects redder than ( U-V)₀=2.5spectroscopic observations are necessary.

5 Color and metallicity distributions