WIDE-FIELD MULTIBAND PHOTOMETRY OF GLOBULAR CLUSTER SYSTEMS IN THE FORNAX GALAXY CLUSTER

The observations were carried out on the nights of 2006 November 24 and 28 (UT) with the Mosaic II CCD imager mounted on the prime focus of the 4 m Blanco telescope at Cerro Tololo Inter-American Observatory (CTIO). The observing conditions were clear and photometric with an average seeing of around 1''. The Mosaic II consists of eight 2048 × 4096 pixel CCDs with a pixel scale of 027 providing a total field of view of 36' × 36'.

Our target field was chosen such that the brightest elliptical galaxy in the Fornax cluster, NGC 1399, and its three neighboring galaxies, NGC 1404, NGC 1387, and NGC 1389, were all imaged in a single telescope pointing. Figure 1 shows the median combined image of our target field as imaged in the B-filter, and Table 1 lists the basic properties of the four members of the Fornax cluster. We observed the target field with u, B, V, and I filters. We note that the standard Johnson U filter was not available at the time of our observations, so we used the Sloan Digital Sky Survey (SDSS) u filter as an alternative and later calibrated the photometry to the Johnson U system. The resulting effect of this calibration procedure is discussed below. The total exposure times in u, B, V, and I were 27000, 5400, 1800, and 1800 s, respectively. We split our observations in each band into two sets of three dithered exposures (five sets for u) to fill in the gaps between the individual CCD chips in the mosaic, and to minimize the impact on our photometry from CCD blemishes (e.g., hot pixels and bad columns). A summary of observations is provided in Table 2.

Zoom In Zoom Out Reset image size

All images were pre-processed using the MSCRED package (Valdes 1998) in IRAF.⁵ We used the CCDPROC task to correct for cross-talk between the CCD chips, apply and trim the overscan, correct for the bias level, and apply the flat-field correction. We used MSCPIXAREA to correct for the variations in pixel scales across the frame produced by geometric distortion in the Mosaic II imager. Each mosaic image was then split into eight individual images using the MSCSPLIT task.

To produce a clean high signal-to-noise image in each filter, we combined the images using the MONTAGE2 routine in the ALLFRAME package (Stetson 1994). We then applied a ring median filter (Secker 1995) having an inner radius of 2.5 times the mean FWHM and a width of 5 pixel on the combined images and subtracted off these images from the original images to remove the rapidly varying background light of the galaxies while preserving the point sources. Objects were detected from the background-subtracted images using the DAOPHOT II/ALLSTAR routines through the following steps. First, we derived a list of objects and created new images with these objects subtracted. From these images, we detected fainter objects that were missed in the previous step and created another set of object-subtracted images. This procedure was iterated five times, and all detected objects were merged into a master catalog. The master catalog was then input to the ALLFRAME program (Stetson 1994) along with all of the images and their point-spread function (PSF) models for final PSF photometry. Aperture corrections were computed through the DAOGROW routine (Stetson 1990) using bright and isolated point sources, and we applied them to our photometry above. Finally, an error-weighted mean instrumental magnitude for each object was obtained using the DAOMATCH/DAOMASTER routines (Stetson 1993).

The photometric calibration was achieved using standard stars in the fields of Landolt (1992) and Stetson (2000). During the course of our observing runs, we imaged eight standard fields placing the stars of interest on chip 6. Each image contains 15–25 standard stars, which covers an appropriately large range in color. We also observed the standard fields in many different airmasses. In the calibration of the U band, however, we used all the standard stars that are observed in subsequent observing runs during 2006, 2007, and 2008, because the photometric uncertainty in the U band is quite large and the number of standard stars is too small to derive reliable solutions. To calibrate the instrumental magnitudes to the Johnson-Cousins standard magnitudes, we used the transformation equation of the form

$$\begin{equation} M_{{\rm std}}=m_{{\rm inst}}\,{+}\,{\rm c1\,{+}\,c2}\,{\times}\, {\rm airmass}\,{+}\,{\rm c3} \,{\times}\, {\rm color}\,{+}\,{\rm c4}\,{\times}\, {\rm color}^2, \end{equation} \tag{ 1 }$$

where M_std is the calibrated magnitude, m_inst is the instrumental magnitude that is normalized to 1 s, c1 is the photometric zero-point offset, c2 is the extinction coefficient for airmass corrections, and both c3 and c4 are the color coefficients. The colors used in the calibration are U − B, B − V, B − V, and V − I for U, B, V, and I, respectively. The photometric coefficients are determined by an iterative process of error-weighted fitting of the transformation equation to the standard magnitudes. The quadratic color term is included only in the equation of U band, because a significant deviation from a linear fit occurs in the transformation of U band. Figure 2 shows the U magnitude difference between the standard and instrumental magnitudes corrected for both of the photometric zero point and the atmospheric extinction, as a function of U − B. The dashed lines represent two best-fit functions for blue and red standard stars, respectively. However, if we divide the observed data into two groups by U − B and use the two functions in calibration separately, the discontinuity between two functions may lead to an artificial clump in color distributions of the data. To avoid this problem, we connected the two functions (red lines) at the internal division points of them in the range 0.29 < U − B < 1.35 (two vertical lines). Finally, the instrumental magnitudes were converted to the standard Johnson-Cousins UBVI system using the set of photometric coefficients listed in Table 3.

Zoom In Zoom Out Reset image size

Calibrating the SDSS u photometry to the Johnson U system may cause systematic shifts in the colors based on this band since the system throughputs are different between the two filters. To estimate these shifts, we simulated our observations using the system throughputs of the two filters, synthetic spectra of GCs from the YEPS model (Chung et al. 2012), and stars from the Kurucz ATLAS9 model. We calculated the magnitude through each filter for each model GC, and converted it to the standardized Johnson U magnitude using the synthetic spectrum of a star having the same color of the GC. We then calculated the differences between these standardized magnitudes from the respective filters. The differences range from 0.002 mag to 0.022 mag, depending on the colors of the model GCs. We conclude that these are small enough to have a negligible impact on our photometric analyses throughout the paper.

As the individual CCD chips of the Mosaic may have different color terms, we attempted to correct for this "chip-to-chip variation" as follows. On the night of 2008 May 5, we took multiple uBVI observations of the Landolt standard fields SA98 and SA107 with the Mosaic II, such that the same set of standard stars was positioned on each chip. Using these data, we derived the color terms that correct for the chip-to-chip variation. Although these color terms were very small, we applied them to our data in hopes of making the multiple measurements of the same source more consistent. The equation used in correction is of the form

$$\begin{equation} M_{{\rm cor}}=m+{\rm c1+c2} \times {\rm color}, \end{equation} \tag{ 2 }$$

where M_cor is the magnitude converted to that on the reference chip 6, m is the magnitude measured on each chip, c1 is the zero-point offset, and c2 is the color coefficient. The colors used in the calibration are U − B, B − V, B − V, and V − I for U, B, V, and I, respectively. Table 4 lists the zero-point offsets and color coefficients for each chip.

We corrected for the foreground Galactic extinction using the reddening maps provided by Schlegel et al. (1998). To account for the atmospheric dispersion corrector (ADC) of the 4 m Blanco telescope that significantly modifies the system throughputs, we computed the effective wavelengths of our four bands from the total system throughputs including the effect of the ADC, and we derived A_λ/E(B − V) values based on the extinction law provided by Cardelli et al. (1989). The mean extinction values in U, B, V, and I are 0.06, 0.05, 0.04, and 0.02 mag, respectively.

Our images were calibrated for astrometry using the USNO-B 1.0 catalog stars (Monet et al. 2003) and the MSCTPEAK task. The typical rms of our astrometric solutions is at the level of 02. The R.A./decl. coordinates of our subsequent source catalogs use the positions derived from these images.

We performed artificial star tests to estimate the completeness of our photometry. We used the ADDSTAR routine of DAOPHOT II to generate 4000 artificial stars in each chip with magnitudes and color ranges consistent with those of the point sources detected and measured above. The luminosity function of our artificial stars was chosen to be flat to preserve statistical significance in all magnitude ranges. Artificial stars were added to the PSF-subtracted images using the empirical PSFs derived during our photometry. We used the same photometry procedure as above to detect and measure artificial stars in these images. This process was repeated seven times with the positions of artificial stars altering randomly each time, resulting in a total of 28,000 artificial stars per chip. In each 0.3 mag bin, we calculated the detection efficiency, i.e., the ratio of the number of recovered artificial stars to the number of added ones. Central regions of galaxies and the gaps between CCD chips were masked for the calculation. To obtain a limiting magnitude at which the completeness drops to 50%, the detection efficiencies were fitted to the analytic function by Fleming et al. (1995),

$$\begin{equation} f(m)=\frac{1}{2}\left (1-\frac{\alpha (m-m_{{\rm lim}})}{\sqrt{1+\alpha ^2(m-m_{{\rm lim}})^2}}\right), \end{equation} \tag{ 3 }$$

where m_lim is the limiting magnitude, and α is the parameter that controls the steepness of the completeness function near the limiting magnitude. Figure 3 shows the measured completeness (dots) and the fitted functions (solid lines) in U, B, V, and I bands depicted by purple, blue, green, and red colors, respectively. The limiting magnitudes of the four bands are 24.4, 24.4, 23.7, and 22.8 mag in U, B, V, and I bands, respectively. We note that the limiting magnitudes slightly vary in the <0.2 mag level among individual CCDs due to variations in detector characteristics.

Zoom In Zoom Out Reset image size

We also investigated the radial variations of completeness around NGC 1399, NGC 1404, and NGC 1387. In the central region of a galaxy, the detection effectiveness at a certain magnitude decreases as the surface brightness of the galaxy increases, and it finally becomes zero at a saturated region. To quantify this effect in our observations, we performed separate artificial star tests near the galaxy centers. We added 8000 artificial stars on the individual CCD chips that contain the galaxies (NGC 1399, NGC 1404, and NGC 1387), this time with a centrally concentrated radial distribution centered on each galaxy. This experiment was repeated five times with the positions of artificial stars altering each time. The completeness was then calculated as a function of magnitude in concentric circles with radial intervals of 02. Figure 4 shows the V band completeness functions at various radial bins from the center of each galaxy. The thick black line represents the completeness function for the overall region, as derived above. As expected from the strong background lights near the centers of our target galaxies, we find that the completeness is very low near the center. The completeness increases as one moves out from the center, and then flattens out beyond 13 for NGC 1399, 07 for NGC 1404, and 05 for NGC 1387.

Zoom In Zoom Out Reset image size

At the distance of the Fornax cluster (∼20.0 Mpc; Blakeslee et al. 2009), 1 pc is equal to 001, so a typical GC appears as a point source in our Mosaic images. We therefore can not select GCs based on their apparent sizes but must rely on photometric colors. Details on the selection process of GC candidates are given below.

We start by using the ALLFRAME parameters CHI and SHARP to select point sources from our photometric catalog. Figure 5 shows the SHARP, CHI, and the ALLFRAME photometric error as a function of U₀ magnitude with solid lines showing the selection criteria. While the figure only shows the U-band selection, we applied the same selection criteria to all bands simultaneously, and the resulting selection includes 5500 point-source objects. We then employ color cuts using our multiband photometry as demonstrated in Figure 6 to select GC candidates. The blue selection boxes shown in the three color–color diagrams were determined on the basis of the findings below. First, we matched our point-source objects with the spectroscopic catalog of Schuberth et al. (2010) and found that the NGC 1399 GCs, as identified by their measured line-of-sight velocities, form a tight sequence (red dots in Figure 6) in the color–color diagrams, whereas foreground stars are located along an extended sequence (cyan dots in Figure 6). Second, we downloaded Hubble Space Telescope (HST)/ACS images of NGC 1399 from the archive and identified genuine GCs based on their sizes. The sizes were measured using ISHAPE (Larsen 1999) and following the same procedure outlined in Strader et al. (2006). We then cross-identified the GCs with our point-source objects, and found that most of them lie inside the tight sequence (orange dots in Figure 6). To select GC candidates, we further applied a magnitude cut of V₀ ⩽ 20.6 (corresponding to M_V ≲ −11) to the color–color-selected sample. This magnitude limit was determined based on the brightest GC in the spectroscopic catalog of Schuberth et al. (2010), and we note that it is consistent with the limit used by Mieske et al. (2006) to separate GCs from their ultra-compact dwarf galaxies sample. The color–color and magnitude cuts result in a final sample of 2037 GC candidates.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 6 demonstrates the advantage of using the U band for separating star-forming background galaxies from the extragalactic GCs. The background galaxies have (B − V)₀ colors similar to the bluer GCs, but due to their star-forming nature, galaxies exhibit significantly bluer colors in (U − V)₀ and (U − B)₀ than the GCs. This allows us to effectively reduce contamination by background galaxies. The same effect was found by Dirsch et al. (2003) using the C − T1 Washington color index. Nonetheless, our selection box still contains several foreground stars, and it is important to quantify the contamination level of our color–color-selected objects. We provide the estimation of the contamination level below.

To estimate the number of contaminating sources (foreground stars and background galaxies) in our GC candidate list, it would be best to use a control field observed near the science field with the same observational conditions. However, this approach is impractical for us because the U-band observations require very long exposure times. As an alternative, we used two independent methods to estimate the contamination level of our GC candidate list.

On average, chip 1 is located farthest away from the target galaxies NGC 1399, NGC 1387, and NGC 1404, so it has the least number of GCs expected among the eight CCD chips.⁶ In Figure 7, we compare the (U − B)₀ versus (B − I)₀ color–color diagrams of the 5500 point-source objects for each chip. In chip 1, there are 41 objects that are selected as GC candidates using the color–color selection box as in Figure 6. If we assume that all 41 objects are contaminants, this implies that our catalog of GC candidates has a contamination level of 16%. We note that 16% is in fact an upper limit because chip 1, which is at an average distance of 24' (∼140 kpc) from the center of NGC 1399, is likely to contain several GCs given that the NGC 1399 GCs are known to extend out to a projected distance of at least 45'  (∼260 kpc; Bassino et al. 2006a).

Zoom In Zoom Out Reset image size

We further estimated the number of foreground stars using the Besançon model of the Milky Way population (Robin et al. 2003). We generated a synthetic catalog of stars in the direction of the sky consistent with our target field and then selected stars using the same color–color cut as used in Figure 6. The estimated number of foreground stars in our GC candidate list turned out to be 180 and the corresponding contamination level is 9%. As expected, this is lower than the upper limit set by using chip 1 as the control field above. Based on the two estimates, the contamination level of our GC candidate list is likely to be in the 9%–16% range.

Although the contamination level itself may not be that high, it is important to test whether the contaminating sources will affect our GC color distribution analyses. Figure 8 shows the comparison of (U − B)₀ color distributions for all point sources (shaded histogram), our GC candidate list (blue), and synthetic stars from the Besançon model (red) between chips 7 and 1. Based on the similarities found in the red and blue color distributions in the upper panel of Figure 8, chip 1 seems to be a good control field. From the same figure, we find that the foreground MW stars are expected to have a (U − B)₀ color distributions that is peaked toward the bluer side similar to the (U − B)₀ color distribution of GC candidates. However, as demonstrated in the lower panel of Figure 8, the number of these foreground stars, when compared to that of the GC candidates, is too small to have any impact on our color distribution analyses presented in Sections 4.3 and 4.4.

Zoom In Zoom Out Reset image size

Table 5 presents the photometric catalog of total 2037 GC candidates. The columns give the following information.

Column 1. Identification number

Columns 2 and 3. Right ascension and declination (J2000)

Columns 4 and 5. U₀ magnitude and its rms uncertainty

Columns 6 and 7. B₀ magnitude and its rms uncertainty

Columns 8 and 9. V₀ magnitude and its rms uncertainty

Columns 10 and 11. I₀ magnitude and its rms uncertainty.

We note that all magnitudes in the catalog are corrected for Galactic extinction.

Figure 9 maps the spatial distributions of GC candidates (left panel) and foreground/background objects (right). The color-filled contours show the distribution of surface number density. The black crosses mark the positions of the optical centers of galaxies (from top, NGC 1399, NGC 1387, NGC 1404, and NGC 1389). In the left panel, three GC systems are clearly visible around NGC 1399, NGC 1404, and NGC 1387. The number of GCs around NGC 1389 is too small, so we omit the galaxy from the following analysis. In the right panel, the distribution of foreground/background objects appears fairly uniform. The slight overdensity in the region between NGC 1399 and NGC 1387 is likely due to the presence of a distant galaxy cluster, given that many objects in this region are bluer than −0.3 in (U − B)₀. Another overdense region is right on the center of NGC 1399, which indicates that GCs in highly dense regions can be misclassified as foreground/background objects in our GC selection procedure, as mentioned in Section 3. Note that the residuals of the mosaic CCD gaps are discernible as a vertical feature in the middle of the panel.

Zoom In Zoom Out Reset image size

It is noteworthy that the GC distribution around NGC 1399 shows interesting features. On the one hand, there is an overabundance of GCs in the outer region of NGC 1399 toward NGC 1404 and NGC 1387. This was also pointed out by Bassino et al. (2006a), and it may indicate interactions of NGC 1399 with NGC 1404 and NGC 1387 in the recent past (Forbes et al. 1997; Bekki et al. 2003). On the other hand, a ∼05 displacement is found between NGC 1399's optical center and the center of the GC distribution in the inner region of the galaxy as denoted by a red contour in the density map. Paolillo et al. (2002) reported an asymmetric X-ray halo for NGC 1399, consisting of three components with different centers; a central component, a galactic component centered 1' southwest of NGC 1399, and a cluster component centered ∼56 northeast of the galaxy. While the center of the GC distribution does not coincide with the X-ray centers, the asymmetric distribution of the GCs may suggest that the NGC 1399 is not yet dynamically relaxed and may be undergoing merger events.

Figure 10 presents the surface number densities against the galactocentric radius for the three galaxies. In order to examine the radial extent of each GC system, we calculated the surface GC number density as a function of the distance from the galaxy center. We first set up a series of radial bins (annulus), each containing approximately equal numbers of GCs (∼100 GCs for NGC 1399; ∼40 GCs for NGC 1404; ∼25 GCs for NGC 1387). The number of GCs in each bin was then corrected using the corresponding completeness function of the bin. We finally divided the corrected numbers of GCs by the area of annuli to get surface densities. The surface density profile of each galaxy is well described by de Vaucouleurs' R^1/4 fits (dashed lines). We determine the limiting radii of GC systems of NGC 1404 (34) and NGC 1387 (37) as the radial points where the density profiles begin to depart from de Vaucouleurs' law. Since the GC system of NGC 1399 is extended (∼45'; Bassino et al. 2006a) well beyond the field of view (36' × 36') of our Mosaic observations, we regard the entire GCs in the images as NGC 1399's GC system, except for the regions inside the limiting radii of NGC 1404 and NGC 1387.

Zoom In Zoom Out Reset image size

Figure 11 presents the CMDs for all point sources. Magnitudes and colors were corrected for the Galactic extinction, as mentioned in Section 2.3. The black and gray dots represent the GC candidates and foreground/background objects, respectively. The dashed lines indicate the limiting magnitude of 50% completeness. The tilt of the limiting magnitude lines results from lower detection efficiency of red GCs than blue GCs in the U band. The (U − B)₀ colors corresponding to GCs, background galaxies, and foreground stars can be determined both by the (U − B)₀ versus (B − I)₀ diagram (the bottom panel of Figure 6) and by the V₀–(U − B)₀ CMD (the left panel of Figure 11). GCs are placed at −0.2 < (U − B)₀ < 0.8, while most background galaxies are bluer at (U − B)₀ ≃ −0.8 and foreground stars are at (U − B)₀ = −0.3 and 1.1 close to both ends of the GC range. The V₀–(B − V)₀ CMD (right panel) shows that many of the background galaxies have similar colors to the GCs, making it difficult to discriminate GCs from galaxies in this color. The foreground star sequences, particularly on the red side, are more clearly separated from GCs than in (U − B)₀. This comparison corroborates that the use of various color combinations including U-band colors provides a powerful tool for discriminating GCs from other objects.

Zoom In Zoom Out Reset image size

It is interesting to note that the CMDs do not show bimodality for GCs brighter than V₀ ∼ 21.5. The disappearance was previously reported for NGC 1399 (Ostrov et al. 1998; Dirsch et al. 2003; Blakeslee et al. 2012), a few galaxies in the Virgo Cluster (Mieske et al. 2006; Strader et al. 2006), and other giant galaxies (Harris et al. 2006; Faifer et al. 2011). The unimodal distribution is often explained by merging of blue and red sequences at the highest luminosities as a result of the blue tilt phenomenon (Harris et al. 2006), which is a trend for the blue GCs to have redder colors at higher luminosities. However, Ostrov et al. (1998) and Forte et al. (2007) have reported that there is no blue tilt in the GC system of NGC 1399. Interestingly, our data reveal that the number of blue GCs rapidly drops at the brightest magnitudes (V₀ < 21.5). This suggests that the unimodal distribution of the brightest GCs is not the consequence of the blue tilt, but is instead owing to the relative scarcity of blue, brightest GCs.

Figure 12 shows three representative color histograms of the entire GC system within the limiting radius in NGC 1399 (top), NGC 1404 (middle), and NGC 1387 (bottom). The GCs fainter than the limiting magnitudes and/or the ones located in the innermost region of each galaxy, where the completeness test is unreliable (see Section 2.3), are taken out from the subsequent analysis. The thick solid lines are smoothed histograms with Gaussian kernels. The filled histograms in the top row are for the brightest GCs, as mentioned in the previous section. For NGC 1399 (top row), GCs show bimodal distributions for all colors considered in agreement with previous findings in C − T1 (Dirsch et al. 2003; Bassino et al. 2006a; Forte et al. 2007). For NGC 1404 (middle) and NGC 1387 (bottom), the dotted histograms are for all GCs within the limiting radii of the galaxies. The solid histograms represent "corrected" distributions, for which the contribution of NGC 1399 GCs is statistically subtracted from the original distributions and will be used in the further analysis. NGC 1404 GCs exhibit pronounced bimodality with an evident dip between two groups in all colors, consistent with the previous studies for NGC 1404's inner GCs by Grillmair et al. (1999) and Larsen et al. (2001). The clear separation between blue and red GCs for NGC 1387 GCs is also consistent with what was found in C − T1 by Bassino et al. (2006b). Blue GCs are much less abundant in NGC 1404 and NGC 1387 than in NGC 1399.

Zoom In Zoom Out Reset image size

A closer scrutiny of Figure 12 reveals that the exact morphologies of the GC color distributions change depending on colors considered. The most distinctive case is given by the (U − B)₀ distribution of NGC 1399. Compared to the (B − V)₀ and (V − I)₀ histograms, the (U − B)₀ distribution exhibits a more prominent blue GC peak and have red GCs with a weaker peak and larger dispersion in color. The (U − B)₀ distribution shows spread in colors of red GCs that is nearly a factor of two larger than that of blue GCs, while the (B − V)₀ and (V − I)₀ histograms have similar variances for the red and blue GCs. The histograms for NGC 1404 and NGC 1387 also show changes in their shape depending on colors.

To be more quantitative in the bimodality analysis, we used the Gaussian Mixture Modeling (GMM) code by Muratov & Gnedin (2010). Table 6 presents the results of the GMM analysis for six color combinations ((U − B)₀, (U − V)₀, (U − I)₀, (B − V)₀, (B − I)₀, and (V − I)₀) for the three galaxies. It gives the color index, the mean (μ), and the standard deviation (σ) of blue and red GC colors, the red GC fraction (f_r = N_red/N_total), and the ratio of the standard deviations between blue and red GCs (σ_r/σ_b). The last three columns summarize the probabilities of preferring a unimodal distribution over a bimodal distribution (p-values) derived based on the likelihood ratio test (LRT; χ²), on the separation of the means relative to their variances (DD) and on the kurtosis of a distribution (kurt).⁷

Figure 13 presents the results from the GMM analysis for all six color combinations. To quantify the shape of histograms, we use the red GC fractions (left) and the ratios of standard deviations between blue and red GC colors (right). NGC 1399 GC sample (solid black lines) does not include GCs around NGC 1404 and NGC 1387, and the GC sample of NGC 1404 (dashed red lines) and NGC 1387 (dotted blue) is corrected for NGC 1399 GC contribution. For NGC 1399, we find that both the numbers and dispersions of blue and red GCs change significantly depending on colors that are used. The trend is more evident when comparing the U-band colors with the rest. For NGC 1404 and NGC 1387, the interesting features emerge after the correction for the contribution of NGC 1399 GCs. First, the red GC fractions in all colors remain constant regardless of the colors, as expected in Figure 12 showing distinct separation between blue and red GCs. It is plausible that the early interactions of the galaxies with NGC 1399 have preferentially left the central, red GCs and later blue GCs have been accreted from outer region of NGC 1399. Second, the color dispersions vary significantly depending on colors, following a similar pattern to the case of NGC 1399 for V − I, B − V, and B − I. For U-band colors, however, the patterns are different in the sense that they tend to be constant within the errors. The variation of dispersion as a function of colors is not expected if the metallicity-to-color converting relations are straight. Instead, the slopes seem different between the blue and red parts of the metallicity–color relations and such an effect varies systematically from color to color (Yoon et al. 2006, 2011a, 2011b, 2012; Cantiello & Blakeslee 2007; Chies-Santos et al. 2012; Blakeslee et al. 2012).

Zoom In Zoom Out Reset image size

Figure 14 displays the color distributions as a function of galactocentric radius for GCs in NGC 1399. The subsequent analysis is only valid for NGC 1399 because small number statistics prohibit us making any firm statement on the radial variations of GC colors for the two fainter galaxies after the correction for the NGC 1399 GC contamination. In order to investigate the radial variation of color bimodality, we set up a series of radial bins containing approximately equal number (∼300) of GCs and perform the GMM analysis for each bin. We focus particularly on three main parameters of the GMM analysis that characterize the bimodal distributions; the number ratio between blue and red GCs, the mean colors of the groups, and their color dispersions.

Zoom In Zoom Out Reset image size

Figure 15 shows the variations of color bimodality properties as a function of galactocentric radius. There is an obvious radial trend for the mean colors of both blue and red GCs getting bluer with increasing radius in all colors. The radial color gradient of red GCs is steeper than that of blue GCs. The red GC fraction in each color rapidly decreases with increasing galactocentric radius out to r ≃ 4', and it remains fairly constant beyond the point. There is a weak radial trend in GC color dispersions, in that the ratio between dispersions of blue and red GCs increases with radius.

Zoom In Zoom Out Reset image size