ABSTRACT
We present catalogs of globular cluster candidates for the 100 galaxies of the Advanced Camera for Surveys Virgo Cluster Survey, a large program to carry out imaging of early-type members of the Virgo Cluster using the Advanced Camera for Surveys (ACS) on the Hubble Space Telescope. We describe the procedure used to select bona fide globular cluster candidates out of the full list of detections based on model-based clustering methods with the use of expected contamination catalogs constructed using blank field observations and which are customized for each galaxy. We also present the catalogs of expected contaminants for each of our target galaxies. For each detected source we measure its position, magnitudes in the F475W (≈ Sloan g) and F850LP (≈ Sloan z) bandpasses, and half-light radii by fitting point-spread function convolved King models to the observed light distribution. These measurements are presented for 20,375 sources, of which 12,763 are likely to be globular clusters. Finally, we detail the calculation of the aperture corrections adopted for the globular cluster photometry.
Export citation and abstract BibTeX RIS
1. INTRODUCTION
In the eleventh Hubble Space Telescope (HST) observing cycle, we initiated the Advanced Camera for Surveys (ACS) Virgo Cluster Survey (ACSVCS; Côté et al. 2004; hereafter Paper I), a program to acquire F475W (≈ SDSS g) and F850LP (≈ SDSS z) images for 100 early-type members of Virgo using the ACS (Ford et al. 1998). Paper I describes the survey itself, including a brief overview of the scientific goals, the selection of the program galaxies and their ensemble properties, the choice of filters, and the field placement and orientation.
One of the primary scientific objectives of the survey is a homogeneous study of the thousands of globular clusters (GCs) belonging to the sample galaxies. A crucial first step in this analysis is the selection of bona fide GCs from all the detected sources in a frame. This method has to be general enough to be applicable to GC systems belonging to galaxies with a wide range in properties, implying a corresponding variety in the properties of their GC systems. In particular, the size of the GC systems in the field of view varies from a few tens of GCs to thousands of them, presenting clear differences in the importance of dealing with contaminating sources such as background galaxies and foreground stars, which will, modulo cosmic scatter, be rather uniform across our sample. Beyond keeping the contamination of the GC samples to a minimum, it is important to be able to assess, for a particular application, the expected amount of the contamination in the sample, and its expected distribution with respect to the variables being probed (e.g., size, magnitude, and color).
The data reduction procedures for the survey have been detailed in Jordán et al. (2004a; hereafter Paper II). Paper II describes the combination of the science images, the modeling and removal of the galaxy surface brightness distribution and subsequent object detection performed with SExtractor (Bertin & Arnouts 1996). An initial selection is done on the object catalogs in order to get a first set of GC candidates. These selection criteria, detailed in Section 2.6 of Paper II, are very conservative. The main aim is to remove the glaring contaminants. All the GC candidates from this first selection are subsequently run through a code (KINGPHOT) that fits PSF-convolved King (1966) models to their surface brightness profiles. This code is described in Jordán et al. (2005; Paper X). The parameters measured for each GC candidate are its magnitude, both in an aperture of 02 and the total model intensity, its celestial coordinates α and δ, its half-light radius rh and concentration c ≡ log(rc/rt), where rc and rt are the core and tidal radii respectively. The latter is an uncertain quantity even for the brighter objects in our catalogs.
After the initial selection of GC candidates mentioned above there are still residual contaminants, mainly foreground stars and background galaxies. For the brightest galaxies, this contamination is small enough that it is not an issue, but for the faintest members of the sample it can be a significant fraction of the overall signal; thus isolating the bona fide GC candidates becomes crucial. Traditionally, selection of GCs is accomplished by restricting in color such as to include only the expected colors of old (τ ≈ 13 Gyr) stellar populations with −2.5 ≲ [Fe/H] ≲ 0, and in magnitude to exclude faint, poorly measured objects. While these cuts will certainly restrict the amount of contaminants present in the samples, this procedure has some obvious drawbacks. First, unless the boundaries of the selection box are adjusted in some way or additional constraints in galactocentric distance are used, the amount of contaminants present will, in general, increase rapidly as we go from the brightest to fainter galaxies. Second, in this scheme there is no way to assess the likelihood of a given source to be a contaminant or a bona fide GC, nor in general is it easy to quantify how contamination will affect the inferred distribution functions of GC parameters.
In this work, we describe a GC selection method, which addresses the points discussed above. The method uses, in addition to a broad color cut, the measured half-light radius rh, which proves very useful in discriminating between GCs and contaminants given the distance to our targets and the characteristics of the telescope/detector combination used. We also discuss the aperture corrections adopted for the photometry and present catalogs of GC candidates for the 100 galaxies in the ACSVCS in machine readable tables available for download from the electronic edition of the Astrophysical Journal. Previous papers in this series have discussed the connection between GCs and low-mass X-ray binaries (Jordán et al. 2004b; Sivakoff et al. 2007), the measurement and calibration of surface brightness fluctuations magnitudes and distances (Mei et al. 2005a, b, 2007), the morphology, isophotal parameters and surface brightness profiles for early-type galaxies (Ferrarese et al. 2006a), the connection between GCs and ultracompact dwarf galaxies (Haşegan et al. 2005), the nuclei of early-type galaxies (Côté et al. 2006), the color distribution of GCs (Peng et al. 2006a), the half-light radii of GCs and their use as a distance indicator (Jordán et al. 2005), diffuse star clusters in early-type galaxies (Peng et al. 2006b), the connection between supermassive black holes and central stellar nuclei in early-type galaxies (Ferrarese et al. 2006b), and the luminosity function, color–magnitude relations, and formation efficiencies of GCs in early-type galaxies (Jordán et al. 2006, 2007; Mieske et al. 2006a; Peng et al. 2008). The GC selection method and the procedure to determine the aperture corrections described in this work have also been applied to the ACS Fornax Cluster Survey (Jordán et al. 2007b; Côté et al. 2007).
2. GC SELECTION METHOD
For each GC candidate we measure its celestial coordinates α (right ascension) and δ (declination), model magnitudes in the F475W and F850LP filters, and King model parameter estimates, half-light radii rh and concentrations c, in each of those bands. Note that henceforth, we will use g475 as shorthand to refer to the F475W filter, and z850 denotes F850LP. Additionally, rh will be taken to be the straight average of the g475- and z850-band measurements, i.e. .
In Figure 1 we show a plot of z850 versus rh for all GC candidates from the 100 galaxies in the survey after the first rough selection of Paper II. The points on this figure are culled in color by requiring 0.5 < (g475 − z850) < 1.9, a generous color cut that includes metallicities in the range −2.25 < [Fe/H] < +0.56 for all simple stellar populations with ages between 2 and 13 Gyr (Bruzual & Charlot 2003), assuming either a Chabrier (2003) or Salpeter (1955) initial mass function. We also cull in rh by requiring rh < 10 pc, a cut that is also very inclusive for typical GCs (see Jordán et al. 2005). Note that although we will quote values of rh in pc throughout this work for convenience, we are really always working with angular measurements. For the purposes of this work, angular measurements are transformed to pc adopting a distance of 16 Mpc.12 The conversion factor from pc to arcsec is then 78 pc arcsec−1. While in principle we could use the measured surface brightness fluctuations distances from our our survey (Mei et al. 2007) to convert angular to physical distances for each galaxy individually (as we have done in many of the survey papers), we have chosen not to do so for GC classification. The reason is that one of our scientific objectives is to test some characteristics of the GC systems (e.g., the GC luminosity function) as distance indicators and thus we decided it was best not to use any distance information in the selection of GC candidates to avoid possible biases in our results.
Three distinct groupings of data points can clearly be distinguished in Figure 1. At rh ∼ 0 there is a vertical clustering of points which corresponds to unresolved sources, i.e., sources that are most likely foreground stars. At rh ∼ 3 pc and z850∼ 20–25 mag there is a second group of points which corresponds to the GC population of the Virgo galaxies. Lastly, there is a diagonal swath of points which corresponds to faint, background galaxies. The diagonal shape of the faint end envelope of this group is due to a completeness effect in surface brightness, as at a given magnitude more extended sources are less likely to be detected. Figure 1 clearly shows the usefulness of having size information in order to separate GCs from background galaxies and foreground stars, as the different types of objects separate into three distinct groupings. We note that, in this context, the use of z850-band magnitudes is preferred to that of g475 band due to the fact that background galaxies are in general bluer than typical GCs, and thus they are fainter with respect to GCs in z850 than in g475.
It is evident from Figure 1 that the unresolved sources can easily be eliminated by requiring rh > 0.75 pc, which we adopt as the minimum angular size to be considered a bona fide GC. This is close to the angular limit down to which we can reliably measure rh for GCs (Jordán et al. 2005). This cut leaves the task of separating the GCs from the background galaxies. For this, we use the clustering method described below.
2.1. Clustering Method
After removing the data cluster corresponding to unresolved sources, the data are a mixture of points drawn mainly from two populations, namely the GCs and a population comprised mostly of background galaxies, which we will henceforth term "contaminants." Thus, their joint distribution can be modeled using a mixture model (see below) with two components, in which the total observed distribution in the rh–z850 plane is the result of summing these two components weighted by their respective sizes. In what follows we briefly outline some formalism regarding mixture models in general and the specialization to our problem in particular. The discussion that follows draws on Fraley & Raftery (2002), to which the reader is directed for more details.
In general, we can model a random variable as a mixture of N components
where (z1, ..., zN) ∈ {0, 1} indicate the component membership of (zi = 1, for only one i), and dk and θk are the probability densities and parameters of the kth component, respectively. The zi are assumed to be distributed as a multinomial of one draw from the N components with Pr(z1 = 1) = f1, ..., Pr(zN = 1) = fN, where fk is the probability that a given observation of belongs to the kth component (fk > 0 and ∑kfk = 1).
In order to estimate the parameters of a mixture model (θ1, ..., θN, f1, ..., fN) via maximum likelihood, the expectation-maximization (EM) algorithm is employed (Dempster et al. 1977; McLachlan & Krishnan 1997). This algorithm, which alternates between two steps, an "E" (expectation) and an "M" (maximization) step, is a general approach to maximum likelihood maximization in which the data consist of n observations that arise from , in which the are observed and the are not observed. In our case, none of our observables tell us directly to which of the groups a given data point belongs, but in our statistical description there will be two groups, and a given data point will belong to one of them. Thus, we have that zik = 1 if belongs to group k, and 0 otherwise.
Having an indicator variable that gives information on group membership is necessary in order to estimate the parameters θk of a given component. Take for instance the simple example of estimating the mean for a multivariate normal: it is clear that in calculating the mean for group k, only data points belonging to that group have to be considered, i.e. , where nk ≡ ∑izik.
Under quite general conditions, which are satisfied in our case, the observed data likelihood can be obtained from the complete data likelihood as ; the maximum likelihood estimate of θ based on the observed data maximizes . For a sample of n independent multivariate observations the observed likelihood of a mixture model with N components is
The corresponding log likelihood of the complete data is
This log likelihood is the one maximized by the EM algorithm, which can be shown to converge to local maximum of the observed data likelihood under mild regularity conditions. The E, or expectation step, is given by the assignment
where a hat indicates as usual in statistical work an estimate of a parameter. is the conditional expectation value of zik, i.e. an estimate given all other parameters of the group membership of . The M, or maximization, step corresponds to maximizing the expression given in Equation (2) over fk and θk with the zik fixed at the values obtained in the E step. The E and M steps are then iterated until the parameters have converged. The value of at a maximum of Equation (1) is the estimated probability that observation i belongs to group k. This quantity can then be used to classify a given observation into its most likely group.
It is a common practice to choose the dk to be multivariate normals ϕk with mean and covariance matrix Σk. Explicitly,
That a description using two multivariate normal components would be rather reasonable in our case can be seen in Figure 2 where a two-dimensional kernel density estimate of the sample with the unresolved sources already removed is shown.
Download figure:
Standard image High-resolution image2.1.1. Specialization to GC Classification in the ACSVCS
For our problem we have two components, i.e. k = 1, 2, with the first component being the GCs and the second the contaminants. Given models for the distributions d1 and d2 for each component, we can characterize them by the set of parameters , where are location (mean) parameters characterizing the distribution in the rh–z850 plane and are any set of shape parameters characterizing the distributions.
We will assume that the shape parameters of the GC distribution are universal, i.e. we assume the rough shape of the clusters of GC points in the rh–z850 plane is independent of the galaxy. This is a reasonable assumption, and a necessary one as well, as the faint galaxies do not have in general enough signal in the data cluster corresponding to GCs in order to reliably determine the form of its joint rh and z850 distribution.
While the shape of the distributions will be kept constant, it might be useful to let some components of depend on the particular galaxy under study. First, the mean magnitude of GCs will obviously depend on the distance and on the typical mass and metallicity of the GCs. Also, we cannot assume that the mean GC rh will be the same in all galaxies (indeed they are not; see Jordán et al. 2005). In the case of the contamination group, the mean magnitude and size may vary because of the varying levels of the galaxy surface brightness: the contamination group in a dwarf galaxy should have a fainter mean z850 due to the decreased mean surface brightness in the frame (Ferrarese et al. 2006a). In practice, we found that letting to be a free parameter leads to unsatisfactory solutions in some cases, and we therefore chose to fix it at mag for all galaxies, but left to be a free parameter. When letting be free, the dispersion around the assumed value of 22.7 mag was significantly smaller than the assumed dispersion σ = 1.3 mag of the magnitude distribution of GCs (see below), and therefore the net effect on the final GC classification is small.13
While modeling the two clusters with multivariate normals might provide a good first approximation as suggested by Figure 2, there is the concern that the steep decline of a multivariate Gaussian in the rh direction might result in the incorrect classification of points at large rh as contamination. Indeed, the distribution of rh for the Milky Way shows an extended tail toward large rh (see e.g., Jordán et al. 2005).
Instead of relying on multivariate normals to describe the GC distribution, we can instead make use of important prior information. It is well known that the luminosity function of GCs is rather universal, being reasonably well approximated by a Gaussian with σ ∼ 1.3 mag (Harris 2001).14 Additionally, it is known that the half-mass radii of GCs are uncorrelated with their luminosity for M ≲ 2 × 106 M☉, a mass limit which includes the great majority of GCs (McLaughlin 2000; Jordán et al. 2005).15 Therefore, magnitude and rh can be taken as independent, and the joint distribution can be taken as the product of a magnitude distribution and an rh distribution. Given these points, we have adopted the following model for the distribution function of the GCs, dGC,
where ggc is the distribution function of rh and σ = 1.3 mag. In order to model ggc we determine it empirically from our data by taking all GC candidates in M87 (VCC 1316) and M49 (VCC 1226) satisfying z850 < 23 mag. This sample is composed almost exclusively of GCs, contamination being almost negligible for these two galaxies. Using this sample, we determined a nonparametric form for ggc using a normal kernel density estimate (Silverman 1986). Even though the combined GC candidate sample of these two giant galaxies has negligible contamination as a whole, for rh ≳ 6 pc contamination is potentially an issue. Thus, we used the nonparametric density estimate for rh < 6 pc only. For larger rh, we extended the distribution with a power-law of the form r−ph, the parameter p determined by fitting to the observed distribution of half-light radii with 4 < rh < 6 pc. The final grh we used is shown in Figure 3. The form of this distribution and the power law behavior for large GCs (rh > 4 pc) is consistent with the parametric form presented by Jordán et al. (2005) for the size distribution of GCs. The tail of this distribution is important to classify correctly extended sources in our procedure.
Download figure:
Standard image High-resolution imageFor the contaminants, we model their distributions using customized control fields. The pipeline described in Paper II was run on a series of control fields, which are listed in Table 1 of Peng et al. (2006a; Paper IX). The source-detection algorithm was run in these 17 fields using the weight maps W'ij constructed for each of the survey galaxies in turn. The weight maps, whose construction is detailed in Paper II, encode the detection thresholds of each galaxy and therefore we obtain for each of the galaxies a catalog of contaminants that we would have observed had the galaxy been present in the control fields. In this sense, we are able to customize the control fields for each of our galaxies. Using the catalog of expected contaminants for each galaxy, we build a two-dimensional kernel density estimate dcont to represent the joint distribution function of the contaminants. This distribution has no free parameters and is thus kept fixed during the source classification process for each galaxy.
With the form of the density functions for each galaxy in hand, we need to estimate the parameters of the mixture model. The procedure we adopted is the following. First, the whole GC candidate sample of all galaxies (minus the "stars") is used to determine a mixture model of the form of Equation (1) using multivariate normals and two groups (GCs and contaminants). This model is constructed in order to then provide a two-component, zero-order, mixture model that will be used to provide (given the z850–rh data points of any given galaxy) initial values for the unobserved group indicator variables zik. In order to estimate the means and covariance matrices Σk, k = 1, 2 we separated the data points into two groups using a hierarchical model-based clustering method (Fraley 1998). The means of the multivariate mixture model thus obtained are and , where 1 denotes the GC group and 2 the contaminants. The covariance matrices are given by
The resulting separation into two groups predicted by the multivariate mixture model applied to the whole sample is shown in Figure 4.
Download figure:
Standard image High-resolution imageAt this point we have a multivariate mixture model of the form given in Equation (1) in hand that given a set of points in the size–magnitude plane will classify them into either group according to the estimates of zik given by Equation (3). For each galaxy we use this model to provide estimates for the zik that in turn provide the initial conditions for the M step in the EM algorithm that is used to estimate the parameters of the final mixture model for that galaxy (using the distributions dGC and dcont). In other words, the multivariate Gaussian model estimated using the whole sample is used to provide the initial conditions that are needed in order to maximize the likelihood given by (see Equation (2))
and thus obtain estimates for using the EM algorithm. Note again that dcont has no free parameters and that . The estimates we obtain for these parameters are listed in Table 1 for each of our program galaxies.
Table 1. Maximum Likelihood Parameters of Mixture Model
VCC (1) | (pc) (2) | fGC (3) | VCC (1) | (pc) (2) | fGC (3) |
---|---|---|---|---|---|
1025 | 2.24 | 0.676 | 1743 | 2.75 | 0.244 |
1030 | 2.68 | 0.595 | 1779 | 3.45 | 0.165 |
1049 | 3.69 | 0.267 | 1826 | 3.46 | 0.231 |
1062 | 3.09 | 0.622 | 1828 | 3.46 | 0.307 |
1075 | 3.48 | 0.346 | 1833 | 3.44 | 0.281 |
1087 | 3.47 | 0.452 | 1857 | 3.48 | 0.212 |
1125 | 3.31 | 0.479 | 1861 | 3.48 | 0.471 |
1146 | 2.85 | 0.572 | 1871 | 2.77 | 0.258 |
1154 | 2.77 | 0.680 | 1883 | 3.23 | 0.459 |
1178 | 3.03 | 0.524 | 1886 | 3.51 | 0.223 |
1185 | 3.45 | 0.317 | 1895 | 3.32 | 0.208 |
1192 | 3.23 | 0.670 | 1903 | 2.79 | 0.770 |
1199 | 3.08 | 0.666 | 1910 | 3.33 | 0.435 |
1226 | 2.72 | 0.941 | 1913 | 3.44 | 0.497 |
1231 | 2.85 | 0.763 | 1938 | 3.12 | 0.552 |
1242 | 3.08 | 0.588 | 1948 | 3.48 | 0.215 |
1250 | 3.30 | 0.447 | 1978 | 2.56 | 0.949 |
1261 | 2.98 | 0.394 | 1993 | 2.97 | 0.116 |
1279 | 2.82 | 0.664 | 200 | 3.41 | 0.306 |
1283 | 2.76 | 0.494 | 2000 | 3.01 | 0.707 |
1297 | 3.41 | 0.699 | 2019 | 3.48 | 0.386 |
1303 | 2.84 | 0.437 | 2048 | 3.44 | 0.295 |
1316 | 2.73 | 0.971 | 2050 | 3.53 | 0.284 |
1321 | 3.09 | 0.421 | 2092 | 2.91 | 0.497 |
1327 | 3.42 | 0.704 | 2095 | 3.09 | 0.509 |
1355 | 3.59 | 0.314 | 21 | 3.50 | 0.274 |
140 | 3.46 | 0.305 | 230 | 3.34 | 0.386 |
1407 | 3.28 | 0.479 | 33 | 3.45 | 0.186 |
1422 | 3.68 | 0.335 | 355 | 2.77 | 0.401 |
1431 | 3.27 | 0.495 | 369 | 3.10 | 0.639 |
1440 | 3.30 | 0.357 | 437 | 3.54 | 0.478 |
1475 | 2.82 | 0.525 | 538 | 3.15 | 0.178 |
1488 | 3.51 | 0.249 | 543 | 3.47 | 0.285 |
1489 | 3.54 | 0.276 | 571 | 2.98 | 0.208 |
1499 | 3.60 | 0.324 | 575 | 2.54 | 0.344 |
1512 | 2.37 | 0.211 | 654 | 3.08 | 0.390 |
1528 | 3.43 | 0.426 | 685 | 2.94 | 0.701 |
1535 | 2.98 | 0.585 | 698 | 3.25 | 0.600 |
1537 | 2.66 | 0.480 | 731 | 2.23 | 0.922 |
1539 | 3.48 | 0.494 | 751 | 3.03 | 0.232 |
1545 | 3.39 | 0.511 | 759 | 2.86 | 0.574 |
1619 | 2.85 | 0.464 | 763 | 2.58 | 0.900 |
1627 | 2.91 | 0.226 | 778 | 2.62 | 0.533 |
1630 | 2.46 | 0.487 | 784 | 2.87 | 0.529 |
1632 | 2.62 | 0.851 | 798 | 3.04 | 0.605 |
1661 | 3.46 | 0.264 | 828 | 2.99 | 0.542 |
1664 | 2.90 | 0.666 | 856 | 3.50 | 0.421 |
1692 | 2.90 | 0.657 | 881 | 2.77 | 0.792 |
1695 | 3.44 | 0.235 | 9 | 3.44 | 0.248 |
1720 | 3.27 | 0.427 | 944 | 2.77 | 0.568 |
Notes.Key to columns: (1) Galaxy VCC number; (2) Mean half-light radius of GC component (assuming D = 16 Mpc); (3) Estimated fraction of the total sample of the GC component.The corresponding quantity for the contaminants component, fcont is given by fcont ≡ 1 − fGC.
Download table as: ASCIITypeset image
After the parameters , μcont, fGC, and fcont have been estimated we assign for each point in that galaxy's sample the probability of it being a GC given by (see Equation (3))
and, given that there are just two components, the corresponding probability pcont of it being a contaminant is given simply by pcont = 1 − pGC.
A final step in the classification is that pGC ≡ 1 is assigned to all sources satisfying z < 23 mag and 1.5 pc <rh < 4 pc, as we want to consider these sources to be bona fide GC candidates regardless of the exact value of pGC returned by the algorithm (we note that of the 6475 sources satisfying these conditions only five originally have pGC < 0.5). Due to the high level of contamination of faint extended objects we also set pGC = 0 for z850 > 25.15 mag, g475 > 26.35, and rh > 10 pc. Sources fainter than that magnitude limit are almost certainly contamination, and in any case contain little useful information due to the large errors in their measured quantities, while sources more extended than the size limit are hard to select against the majority of background galaxies with those angular sizes (see Figure 5).
Download figure:
Standard image High-resolution imageWhat are the advantages of this method over just defining fixed boundaries in the rh–z850 plane? By estimating fGC and fcont for each galaxy, we naturally include the fact of varying ratios of contamination to GC candidates. In addition, by letting the be determined for each galaxy we naturally account for variations in the mean rh as well. Also, we have a quantitative measure of how likely a certain data point is to be a GC (under the assumed model) and this knowledge can be included in statistical estimators applied to the GC sample.
In order to illustrate the performance of the method we show the results for four galaxies that span the magnitude range of our sample: VCC 1226 (= M49), VCC 1422, VCC 2048 and VCC 1661. In terms of apparent blue luminosity, these are the 1st, 50th, 51th and 100th ranked galaxies in the ACSVCS sample, respectively. In Figure 5 we show the resulting classification in the rh–z850 plane for these galaxies along with the same classification applied to our custom control fields (scaled to a single field). In Figure 6, we show the resulting GC luminosity functions and color distributions when restricting the objects to those having pGC ⩾ 0.5. In the samples thus selected it will still be necessary to take into account the residual contamination that is classified as bona fide GCs (i.e., the false positives), but it should be clear from these figures that this contamination has been greatly reduced by our selection using a process that naturally takes into account the richness of the GC system of each galaxy.
Download figure:
Standard image High-resolution imageWhile the selection method effectively isolates the desired GC data cluster, studies that aim to study the shape of the GC distributions in either size or magnitude need to test their conclusions against any subtle biases that the selection of GCs might have imposed on them via the choice of the form of dGC. This can be easily done by considering the robustness of results when selecting alternate GC samples of GCs that do not rely on pGC, as we have done when studying the luminosity function of GCs (Jordán et al. 2006, 2007a).
3. GC PHOTOMETRY: APERTURE CORRECTIONS
The photometric zeropoints and foreground reddening corrections that we adopted are detailed in Sections 2.6 and 2.7 of Jordán et al. (2004a). A challenging aspect of obtaining accurate photometry of marginally resolved objects, such as GCs at the distance of the Virgo cluster, is that their aperture corrections depend on their intrinsic size and luminosity profile, which are not known a priori for each source. The usual practice in previous photometric work on GCs with HST has been to adopt an average correction obtained from bright, high signal-to-noise ratio GC candidates or, alternatively, from simulated King models with rh ≈ 3 pc and assumed fixed concentration. In the ACSVCS, we provide two measurements of the magnitude per filter for each GC candidate. The first measurement is of the total magnitude using the best-fit King (1966) model, a method which takes into account the proper aperture correction for each individual object. We refer to these magnitudes as "model magnitudes." The second measurement is of the best-fit aperture magnitude within a 4 pixel (=02) radius, with an aperture correction appropriate for an average GC (rh = 3 pc, c = 1.5). We refer to these as "average correction aperture magnitudes." The former method is best for obtaining an unbiased total GC flux. The latter method is useful because it relies only minimally on the King model fitting process, and thus may be better for studies concerned with the colors of faint GCs. We outline both methods below, and discuss the differences in their use.
3.1. Model Magnitudes: Total Magnitudes from King Model Fits
We obtain model magnitudes using the best-fit King model for each GC candidate in conjunction with the appropriate, size-dependent aperture corrections. As described in the Appendix of Jordán et al. (2005), the fitted King models are convolved with the appropriate point-spread function (PSF). These PSFs are constructed in the manner outlined by Jordán et al. (2004a, Section 2.6), and extend to a radius r = 05. We fit the observed light distributions of GC candidates within a fitting radius rfit (see the Appendix of Jordán et al. 2005), which means that in a strict sense, the PSF-convolved model matches the amount of light within that radius, with the rest (out to r = 05) added as prescribed by the PSF-convolved model. These magnitudes already take into account the size (rh) of each object. However, aperture corrections are still needed for the following reasons. First, the magnitudes require the aperture correction of the PSF from r = 05 to infinity. Second, the PSF we have used in the fitting will have some differences from the one used to define the aperture corrections in Sirianni et al. (2005), which we assume to be a representation of the "true" mean PSF.
To obtain the aperture corrections arising from these effects, we adopted the following procedure. We constructed a set of King models of various rh and convolved them with the PSF that is used for fitting the GC candidates in each of the z850 and g475 bands. Then, we convolved the same set of models with a PSF that was constructed up to a radius of 3'' using stars in the outskirts of 47 Tucanae (as used in Mei et al. 2005). Aperture corrections derived from these PSFs are consistent with those derived by the PSFs used in Sirianni et al. (2005). Given a fitting radius rfit, the aperture correction appropriate for a GC candidate of half-light radius rh and concentration c described by a King model is given by
where ⊗ denotes convolution. All King models k and PSFs are normalized to have a total flux of unity. Thus, the numerator within the logarithm in Equation (9) is the fraction of the -convolved model flux that is within the fitting radius. Correspondingly, the denominator is the fraction of the -convolved model flux that is within the fitting radius. The ratio of the two represents the aperture correction necessary to transform the fitted magnitude to a total model magnitude. Remember that the fitted magnitude that we are correcting assumes the PSF, , and already incorporates the numerator term. We define aperture corrections as values to be subtracted from the fitted magnitude.
In order to apply this correction for our GC candidates we assumed a fixed concentration c = 1.5 for all GCs, and then computed for rh = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 50, 100 pc and rfit = 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 pixel and interpolated from these values to apply an aperture correction for any given cluster with best-fit rh fitted using a radius rfit.
The resulting aperture corrections for both bands are illustrated in Figure 7. There is a very mild dependence on the aperture correction on the measured rh for GCs (which are all required to have rh < 10 pc), but there is some change of the order of a few hundredths of a magnitude for very extended objects. Thus, our aperture corrections are essentially equivalent to having applied an average correction for all clusters. Note that the latter is true only because the fitted magnitudes include the amount of light outside the fitting radius (within r < 05) as given by the best fit PSF-convolved King model. It would not be true if an average aperture correction is applied to a simple aperture magnitude, as has usually been done in previous photometric measurements of GC systems at the distance of Virgo. In that case, the aperture correction would get systematically larger as rh increases. We note that, as expected, at rh ∼ 0 the aperture corrections are similar to those expected for point sources from 05 to infinity (Sirianni et al. 2005). We note also that assuming a fixed c when deriving our aperture corrections is not a strong assumption: for rh < 10 pc, aperture corrections vary by less than 0.005 mag when varying c between 1 and 2.
Download figure:
Standard image High-resolution image3.2. Average Correction Aperture Magnitudes
In addition to measuring total magnitudes based on the best-fit PSF convolved King models, we also measured the best-fit amount of light in a 4 pixel (=02) aperture radius, and then applied an aperture correction appropriate for an average object. We note that the light measured within the aperture is measured on the best-fit King model rather than the data. This method has the advantage of dealing naturally with bad pixels and subpixel shifts that can affect aperture magnitudes measured directly from the data. The normalization of the model (i.e., the total flux within the fitting radius) is the most robustly fitted quantity even for our faintest objects, and because the 4 pixel aperture is almost always equal in size to the fitting radius (sometimes smaller), the aperture flux measured on the model is very reliable. These magnitudes are best used for studying the colors of faint GCs, as they provide a color measurement that relies only weakly on the fitting process, and thus can have smaller photometric errors compared to colors derived from model magnitudes.
To estimate the total magnitudes from these aperture magnitudes, we calculate the aperture correction for a King model with rh = 3 pc and c = 1.5, which is the quantity given by
The values obtained for are then 0.237 mag for the g475 band and 0.347 mag for the z850 band. Assuming c = 1 or c = 2 would only make a minimal difference (<0.005 mag). We note that model magnitudes and average corrected aperture magnitudes are consistent in the average as expected.
3.3. Further Notes on Aperture Corrections
As we have discussed, photometry of marginally resolved sources whose individual sizes are not known a priori is a challenge for HST studies of GCs at the distance of the Virgo Cluster. In principle, standard aperture photometry, where one measures the flux in a given aperture and applies the same aperture correction for all objects, is subject to biases such that the total fluxes of extended objects are underestimated. We emphasize that our model magnitudes fully account for this effect, by fitting for the size of each individual object, and applying the proper aperture corrections out to a radius of 3'' using a suite of PSF-convolved King models. Studies concerned with the magnitudes or colors of sources that are large (rh ≫ 3 pc), and where size-dependent biases may be important, should use model magnitudes. For example, Mieske et al. (2006a), the ACSVCS study of the color–magnitude relations in GC systems, solely used model magnitudes to avoid any size-dependent biases on the magnitudes or colors.
Our average corrected aperture magnitudes, however, are more similar to previous work, where the flux is measured in a single aperture and an average aperture correction is applied. This presents the possibility that if GCs have sizes significantly different from rh = 3 pc, the size at which we calculate our fiducial correction, then their magnitudes and colors could be systematically in error. The left panel in Figure 8 shows the aperture corrections in both filters to a 4 pixel aperture for King models with a range of sizes. This plot illustrates how GCs with sizes substantially larger than 3 pc can have significantly underestimated fluxes as measured by average corrected aperture magnitudes (∼1 mag for rh = 20 pc). This is why in all our ACSVCS studies, we always use model magnitudes for measures of total flux. The bias in color is the difference of the two curves in the left panel of Figure 8, and we show this in the right panel of Figure 8. Unlike the bias in the total magnitude, however, the bias in using average corrected aperture colors is quite small. Over the range of sizes, 1 < rh < 30 pc, the g–z aperture correction deviates by only +0.004−0.014 from the rh = 3 pc fiducial. This shows that average corrected aperture colors can be considered essentially unbiased for the typical range of GC sizes, and they have the added advantage of smaller photometric errors for faint sources as compared to colors from model magnitudes.
Download figure:
Standard image High-resolution imageWhen we further examine the effects of GC size on the g–z aperture correction, we can see that at small sizes, the aperture correction to the color gets redder with increasing GC size due to the different sizes of the PSFs in F475W and F850LP. However, once GCs are larger than rh ∼ 10 pc, the color correction becomes bluer. This can be understood because at that point, we enter the regime where the size difference between the PSFs in the two filters is small compared to the size of the GC (10 pc ∼0125 at Virgo). Therefore, GCs with increasingly larger sizes will not produce increasingly redder corrections. However, for studies where the objects of interest have sizes rh > 10 pc, we recommend the use of model magnitudes and colors only.
Recently, there has been a claim that aperture magnitude biases as described above can masquerade as astronomical effects (Kundu et al. 2008). In particular, the claim is that trends in the mean colors of metal-poor GCs in the color–magnitude diagram can be explained by a correlation between GC size and luminosity. This cannot be true because our study of the color–magnitude relation of metal-poor GCs (Mieske et al. 2006a) uses model magnitudes and colors, which fully account for size-dependent aperture corrections. Moreover, as we have just shown, even if using average corrected magnitudes, the color biases for the average correction aperture colors would be much too small to account for the color-magnitude relation seen in GC systems.
4. COMPLETENESS
For many studies, it is important to know the level of completeness for a GC with a given set of properties. Unlike for point sources in blank fields where object magnitude is the primary parameter driving the detection probability, the ACSVCS data present more parameters than can affect completeness, namely the GC size (i.e., surface brightness), and the brightness of the background from the integrated light of the host galaxy. For a given magnitude, a GC with a larger size will be more difficult to detect, as will a GC projected onto the bright central regions of the host galaxy. We have run an extensive set of simulations to quantify the completeness of GCs as a function of their size (rh), the background surface brightness (μb), and their total magnitude (m).
We created a suite of simulated GCs based on King models with c = 1.5 and sizes of rh = 1, 3, 6, and10 pc. These simulated GCs were scaled to a random magnitude with 21 < z < 27 mag and color g–z = 1.1. They were then inserted at semirandom positions in actual ACSVCS images for the galaxies VCC 1226 and 1833, the former being the brightest galaxy in the sample, and the latter being a dwarf galaxy with one of the lowest sky backgrounds. This allowed us to sample the full range of background surface brightnesses in the survey. Pixel fluxes for simulated GCs were given the appropriate random Poisson noise, and GC centering was randomly shifted by fractions of a pixel (as small as 1/60 of a pixel). The positions at which they were placed in both the F475W and F850LP images were random, but they were limited to regions that were at least 35 from a real object, and at least 2'' from the edge of the image. This was because our goal was to measure detection efficiency as a function of (m, rh, μb) and not due to crowding or deblending, which does not present a problem for real objects.
The images with simulated GCs were then analyzed using the exact same pipeline as was used to produce the GC catalog. In total, we simulated 4,993,501 fake GCs across the full range magnitude, size, and background surface brightness. For each of these GCs, we know whether or not they are detected, and we build completeness curves from this database. For each of the four sizes, and for each filter, we present the detection probability of a GC in 10 bins of background surface brightness roughly equally spaced in log(flux), and in 0.1 mag steps in magnitude for 21 < z < 27 mag. These completeness curves are presented in Table 2 for the g band and Table 3 for the z band. For each of the rh values simulated (indicated in the first column) we tabulate the completeness curves as a function of magnitude (indicated in the second column) and the 10 different background values (Columns 3–12, background values are specified in the header of these columns).
Table 2. g-Band Completeness Curves
rh | mg | bg | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
('') | AB mag | 0.0518 | 0.1046 | 0.1491 | 0.2359 | 0.4023 | 0.7175 | 1.2535 | 2.1876 | 3.6955 | 6.0713 |
0.0128 | 22.10 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.11 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.12 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.13 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.14 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.15 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.16 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.17 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.18 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.19 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.20 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.21 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.22 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.23 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.24 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.25 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.26 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.27 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.28 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 22.29 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.
Download table as: DataTypeset image
Table 3. z-Band Completeness Curves
bz | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
rh ('') | mz AB mag | 0.0326 | 0.0917 | 0.1403 | 0.2354 | 0.4044 | 0.7457 | 1.3262 | 2.2897 | 4.3344 | 7.6519 |
0.0128 | 21.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.01 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.02 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.03 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.04 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.05 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.06 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.07 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.08 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.09 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.10 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.11 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.12 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.13 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.14 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.15 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.16 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.17 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.18 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
0.0128 | 21.19 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.
Download table as: DataTypeset image
5. CATALOGS
5.1. Full Source Catalogs
In Table 4 we present our full catalog of sources satisfying the rough selection criteria presented in Section 2.6 of Paper II for all galaxies in the ACSVCS. The first column is the galaxy ID in the Virgo Cluster Catalogue (VCC; Binggeli et al. 1985; see Table 1 in Côté et al. 2004 for NGC and Messier equivalents). Columns 2 and 3 give the right ascension α (J2000) and declination δ (J2000) of each source and Column 4 gives the projected distance to the center of the host galaxy in arcseconds. Columns 5 and 6 give the total King model magnitude and the total magnitude inferred from a 02 aperture for the z850 band. These magnitudes have been dereddened as described in Section 2.7 in Paper II and have had aperture corrections applied as described in Section 3. Columns 7 and 8 give the corresponding quantities for the g475 band. Columns 9 and 10 give the best-fit half-light radii of the PSF-convolved King (1966) models in arcseconds for the z850 and g475 bands, respectively. The uncertainties do not include systematic uncertainties arising from the PSF modeling which can be estimated to be of order ≈0005 (see Jordán et al. 2005). In order to convert the half-light radii to physical units, the SBF distances to our galaxies presented in Mei et al. (2007) can be used. Column 11 gives the value of pGC for each source. Column 12 gives the adopted value of E(B − V) which is taken from the DIRBE maps of Schlegel et al. (1998). Finally, Columns 13 and 14 give the galaxy plus "sky" background in counts/s present under each source in the g475 and z850 bands, respectively. These quantities are necessary to estimate the expected completeness using the data presented in Tables 2 and 3. We note that while formally we obtain the best-fit concentrations, we do not provide them here given that they are rather uncertain even for the most luminous GC candidates. In any case, these quantities were used only for one galaxy (M87) in the work presented in Jordán et al. (2004) regarding the connection between low-mass X-ray binaries and GCs, and, as discussed in Sivakoff et al. (2007), the results of that work do not depend on the use of concentrations.16
Table 4. Photometric and Structural Catalog of Sourcesa
VCC (1) | α (J2000) (2) | δ (J2000) (3) | dgal('') (4) | mz (5) | mz,ap (6) | mg (7) | mg,ap (8) | rh,z (9) | rh,g (10) | pGC (11) | E(B − V) (12) | bz (13) | bg (14) |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1226 | 187.4439383 | 8.0000332 | 3.219 | 21.325 ± 0.050 | 21.271 ± 0.042 | 22.477 ± 0.057 | 22.477 ± 0.037 | 0.0301 ± 0.0041 | 0.0370 ± 0.0050 | 1.00 | 0.022 | 11.0150 | 7.5360 |
1226 | 187.4455059 | 8.0010107 | 3.437 | 22.700 ± 0.426 | 22.665 ± 0.115 | 23.797 ± 0.093 | 23.763 ± 0.084 | 0.0319 ± 0.0181 | 0.0283 ± 0.0084 | 1.00 | 0.022 | 10.4730 | 7.4200 |
1226 | 187.4436038 | 8.0011660 | 4.942 | 21.669 ± 0.043 | 21.608 ± 0.036 | 23.038 ± 0.046 | 22.988 ± 0.046 | 0.0229 ± 0.0042 | 0.0224 ± 0.0052 | 1.00 | 0.022 | 7.8560 | 5.5080 |
1226 | 187.4448279 | 7.9986757 | 6.244 | 20.916 ± 0.046 | 20.874 ± 0.028 | 22.297 ± 0.032 | 22.255 ± 0.028 | 0.0238 ± 0.0039 | 0.0286 ± 0.0027 | 1.00 | 0.022 | 6.3150 | 4.4670 |
1226 | 187.4469413 | 7.9997839 | 8.081 | 21.217 ± 0.031 | 21.164 ± 0.027 | 22.557 ± 0.029 | 22.511 ± 0.029 | 0.0225 ± 0.0033 | 0.0289 ± 0.0029 | 1.00 | 0.022 | 4.4850 | 3.2670 |
1226 | 187.4451763 | 8.0026245 | 8.123 | 22.770 ± 0.203 | 22.882 ± 0.097 | 23.611 ± 0.095 | 23.629 ± 0.061 | 0.0574 ± 0.0233 | 0.0411 ± 0.0088 | 1.00 | 0.022 | 4.6930 | 3.4040 |
1226 | 187.4454712 | 8.0026666 | 8.523 | 22.125 ± 0.087 | 22.114 ± 0.080 | 23.073 ± 0.038 | 23.032 ± 0.036 | 0.0348 ± 0.0086 | 0.0343 ± 0.0030 | 1.00 | 0.022 | 4.4050 | 3.2010 |
1226 | 187.4474911 | 8.0010406 | 9.987 | 22.309 ± 0.102 | 22.340 ± 0.050 | 23.501 ± 0.084 | 23.537 ± 0.054 | 0.0638 ± 0.0100 | 0.0419 ± 0.0075 | 1.00 | 0.022 | 3.2730 | 2.4030 |
1226 | 187.4423850 | 7.9987706 | 10.310 | 21.171 ± 0.036 | 21.147 ± 0.029 | 22.102 ± 0.020 | 22.076 ± 0.015 | 0.0325 ± 0.0035 | 0.0299 ± 0.0014 | 1.00 | 0.022 | 3.1990 | 2.3020 |
1226 | 187.4448252 | 8.0033496 | 10.594 | 21.498 ± 0.049 | 21.439 ± 0.037 | 22.893 ± 0.028 | 22.853 ± 0.025 | 0.0184 ± 0.0059 | 0.0148 ± 0.0026 | 0.94 | 0.022 | 3.5490 | 2.5860 |
1226 | 187.4464555 | 7.9979555 | 10.688 | 23.410 ± 0.578 | 23.437 ± 0.167 | 23.974 ± 0.512 | 23.926 ± 0.101 | 0.0077 ± 0.0898 | 0.0358 ± 0.0105 | 0.99 | 0.022 | 3.5530 | 2.5730 |
1226 | 187.4468670 | 7.9982008 | 10.933 | 19.889 ± 0.011 | 19.891 ± 0.009 | 21.288 ± 0.028 | 21.290 ± 0.021 | 0.0375 ± 0.0017 | 0.0349 ± 0.0018 | 1.00 | 0.022 | 3.4110 | 2.4890 |
1226 | 187.4440948 | 8.0035453 | 11.557 | 23.374 ± 0.101 | 23.318 ± 0.099 | 24.089 ± 0.123 | 24.095 ± 0.085 | 0.0326 ± 0.0079 | 0.0354 ± 0.0097 | 0.99 | 0.022 | 3.2620 | 2.3900 |
1226 | 187.4471733 | 8.0027498 | 12.050 | 22.594 ± 0.080 | 22.555 ± 0.066 | 23.934 ± 0.055 | 23.897 ± 0.052 | 0.0270 ± 0.0075 | 0.0141 ± 0.0056 | 1.00 | 0.022 | 2.6650 | 1.9680 |
1226 | 187.4433294 | 8.0034903 | 12.210 | 22.090 ± 0.045 | 22.024 ± 0.046 | 23.492 ± 0.052 | 23.445 ± 0.049 | 0.0219 ± 0.0052 | 0.0223 ± 0.0041 | 1.00 | 0.022 | 3.0790 | 2.2590 |
1226 | 187.4450661 | 7.9969838 | 12.365 | 22.292 ± 0.048 | 22.235 ± 0.045 | 23.785 ± 0.050 | 23.737 ± 0.053 | 0.0144 ± 0.0040 | 0.0270 ± 0.0058 | 1.00 | 0.022 | 2.9960 | 2.1770 |
1226 | 187.4479377 | 7.9985460 | 13.171 | 22.025 ± 0.027 | 21.956 ± 0.026 | 23.261 ± 0.109 | 23.220 ± 0.032 | 0.0249 ± 0.0035 | 0.0175 ± 0.0046 | 1.00 | 0.022 | 2.6190 | 1.9270 |
1226 | 187.4444844 | 8.0040618 | 13.196 | 21.760 ± 0.085 | 21.751 ± 0.036 | 23.246 ± 0.025 | 23.204 ± 0.024 | 0.0310 ± 0.0057 | 0.0260 ± 0.0033 | 1.00 | 0.022 | 2.7710 | 2.0380 |
1226 | 187.4416614 | 7.9980605 | 13.917 | 21.132 ± 0.022 | 21.109 ± 0.013 | 21.966 ± 0.044 | 21.938 ± 0.044 | 0.0227 ± 0.0020 | 0.0272 ± 0.0058 | 1.00 | 0.022 | 2.2020 | 1.5970 |
Notes.Key to columns: (1) Galaxy VCC number; (2) and (3) J2000 right ascension (α) and Declination (δ) in decimal degrees; (4) Galactocentric distance in arcseconds; (5) z850-band model magnitude obtained from the best-fits PSF convolved King model and an aperture correction as per Equation (9); (6) z850-band average correction aperture magnitude inferred from a 02 aperture and an aperture correction as per Equation (10); (7) Same as (5) but for the g475 band; (8) Same as (6) but for the g475 band; (9) and (10) Best-fit half-light radii measured in the z850 and g475 bands, respectively; (11) Probability that the source is a GC according to the maximum likelihood estimate of our assumed mixture model (see Section 7); (12) Foreground E(B − V) assumed for this source. The corrections for foreground reddening were taken to be Ag = 3.634E(B − V) and Az = 1.485E(B − V) in the g and z bands, respectively (see Jordán et al. 2004); (13) Background in the z850 band (counts/s); (14) Background in the g475 band (counts/s). aTable 4 present the structural and photometrical catalog of all ACSVCS sourcs that satisfy the selection criteria presented in Section 2.6 in Jordán et al. 2004a (Paper II). To select a sample of bona fide GCs the sources should be restricted to those having pGC ⩾ 0.5.
Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.
Download table as: DataTypeset image
We stress that to select a catalog of bona fide GC candidates from the full source list presented in Table 4, the sample needs to be restricted to sources satisfying pGC ⩾ 0.5. We include sources with pGC < 0.5 in Table 4 in order to allow the construction of GC samples using different criteria than those we have adopted in the ACSVCS. We caution though that parameters in Table 4 that are obtained by fitting PSF-convolved King (1966) models should be viewed only as rough indications for sources with pGC ≲ 0.5, as those sources are not expected to be well represented by King models in general given that they are most likely background galaxies.
5.2. Control Field Catalogs
In Table 5, we present our full catalog of contaminants satisfying the same selection criteria as the sources in Table 4 for all galaxies in the ACSVCS. These catalogs are obtained from 17 control fields that have been customized for each galaxy as described in Section 2.1.1. These catalogs therefore give 17 times the amount of contamination expected in the field of each target galaxy. These catalogs are useful in assessing the effects of the residual contamination for any study performed using the GC catalogs constructed from Table 4. We stress that background galaxies are not expected to be well-described by King models, and therefore the best-fit parameters presented in Table 4 are useful only for the purposes just mentioned. Additionally, note that the uncertainties in the photometric and structural parameters are calculated under the assumption that the objects can be described by King models and thus are not presented for objects in the contaminants catalogs where this assumption has no grounds.
Table 5. Catalog of Expected Contaminantsa
VCC (1) | dgal('') (2) | mz (3) | mz,ap (4) | mg (5) | mg,ap (6) | rh,z (7) | rh,g (8) | pGC (9) |
---|---|---|---|---|---|---|---|---|
1226 | 8.127 | 21.563 | 21.797 | 22.413 | 22.396 | 0.0023 | 0.0192 | 0.65 |
1226 | 8.683 | 22.921 | 22.872 | 23.525 | 23.524 | 0.0150 | 0.0128 | 0.88 |
1226 | 9.106 | 20.545 | 20.483 | 21.252 | 21.258 | 0.0081 | 0.0145 | 0.54 |
1226 | 12.556 | 21.977 | 21.923 | 23.414 | 23.432 | 0.0124 | 0.0152 | 0.88 |
1226 | 15.859 | 23.186 | 23.116 | 24.067 | 24.066 | 0.0158 | 0.0128 | 0.89 |
1226 | 18.963 | 19.329 | 19.502 | 20.312 | 20.787 | 0.0051 | 0.0169 | 0.39 |
1226 | 18.966 | 22.586 | 22.818 | 23.941 | 24.055 | 0.0807 | 0.0525 | 0.94 |
1226 | 23.310 | 23.593 | 23.515 | 25.487 | 25.542 | 0.0154 | 0.0247 | 0.97 |
1226 | 23.385 | 23.312 | 23.550 | 24.597 | 24.811 | 0.0930 | 0.0775 | 0.44 |
1226 | 24.904 | 21.590 | 21.530 | 22.824 | 22.836 | 0.0119 | 0.0138 | 0.80 |
1226 | 25.750 | 19.101 | 19.182 | 20.782 | 21.209 | 0.0168 | 0.0226 | 1.00 |
1226 | 25.870 | 22.497 | 22.512 | 23.360 | 23.419 | 0.0331 | 0.0311 | 1.00 |
1226 | 26.637 | 22.120 | 22.197 | 23.049 | 23.497 | 0.0180 | 0.0230 | 1.00 |
1226 | 26.855 | 23.170 | 23.568 | 24.588 | 24.918 | 0.1098 | 0.0887 | 0.27 |
1226 | 27.742 | 20.256 | 20.205 | 20.757 | 20.760 | 0.0076 | 0.0159 | 0.56 |
1226 | 28.061 | 22.977 | 23.297 | 23.885 | 24.124 | 0.1301 | 0.0742 | 0.36 |
1226 | 28.708 | 24.031 | 23.970 | 24.610 | 24.576 | 0.0144 | 0.0206 | 0.92 |
1226 | 29.768 | 20.871 | 20.793 | 21.492 | 21.535 | 0.0134 | 0.0070 | 0.50 |
1226 | 30.516 | 22.317 | 22.247 | 23.521 | 23.576 | 0.0121 | 0.0269 | 1.00 |
Notes.Key to columns: (1) Galaxy VCC number; (2) Galactocentric distance in arcseconds; (3) z850-band model magnitude obtained from the best-fits PSF-convolved King model and an aperture correction as per Equation (9); (4) z850-band average correction aperture magnitude inferred from a 02 aperture and an aperture correction as per Equation (10); (5) Same as (3) but for the g475 band; (6) Same as (4) but for the g475 band; (7) and (8) Best-fit half-light radii measured in the z850 and g475 bands, respectively; (9) Probability that the source is a GC according to the maximum likelihood estimate of our assumed mixture model (Section 7). aThis table presents the expected contaminants in 17 control fields customized to each galaxy. It can be used to infer 17 times the expected contamination in any given GC sample.
Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.
Download table as: DataTypeset image
The first column is the galaxy ID in the VCC (Binggeli et al. 1985; see Table 1 in Côté et al. 2004 for NGC and Messier equivalents). Columns 3 and 4 give the total King model magnitude and the total magnitude inferred from a 02 aperture for the z850 band. Columns 5 and 6 give the corresponding quantities for the g475 band. Columns 7 and 8 give the best-fit half-light radii of the PSF-convolved King (1966) models in arcseconds for the z850 and g475 bands, respectively. In order to convert the half-light radii to a linear distance, the SBF distances to our galaxies presented in Mei et al. (2007) can be used. Finally, Column 9 gives the value of pGC for each source.
6. SUMMARY
We have presented the selection procedure for GC candidates in the ACS Virgo Cluster Survey, a survey of 100 galaxies in the Virgo cluster of galaxies. This procedure is based on model clustering methods which we briefly describe in the context of our survey.
We have additionally presented the determination of the aperture corrections for our GC candidates. Finally, we present the results of our photometric and structural parameter measurement for 20,375 objects which satisfy the rough selection criteria presented in Paper II in these series. This full source catalog contains 12,763 bona fide GC candidates which have a probability pGC > 0.5 of being a GC according to our selection procedure. Additionally, we present catalogs of the contaminants expected to remain in such samples as deduced from observations of 17 control fields. These catalogs are presented as machine readable tables available for download from the electronic edition of the Astrophysical Journal. They are also available for download at the project's Web site http://www1.cadc.hia.nrc.gc.ca/community/ACSVCS/.
Support for program GO-9401 was provided through a grant from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. E.W.P. acknowledges the support of the National Research Council of Canada's Plaskett Research Fellowship at the Herzberg Institute of Astrophysics. P.C. acknowledges additional support provided by NASA LTSA grant NAG5-11714.
Facilities: HST(ACS)
Footnotes
- *
Based on observations with the NASA/ESA Hubble Space Telescope obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555.
- 12
In Mei et al. (2005a, 2007) we adopt a mean distance of 16.5 Mpc to Virgo. The work presented here was performed before we adopted this value, but it is well within its systematic uncertainties. The effect in the classification of sources of adopting 16.5 Mpc instead of 16 would be in any case very small.
- 13
≲0.5% of the sources would switch classification when performing the classification by letting free instead of being fixed.
- 14
- 15
- 16
Regarding this point, it is worth noting that Jordán et al. (2007c) have shown, using HST/ACS and Chandra observations of Centaurus A (=NGC 5128), that the King model concentrations are not a fundamental variable in determining the presence of low-mass X-ray binaries in GCs.