THE HST/ACS COMA CLUSTER SURVEY. II. DATA DESCRIPTION AND SOURCE CATALOGS^*

The nominal observing sequence for each field consists of four dither positions ("DITHER-LINE" pattern type) each having an exposure time of 350 s and 640 s for the F814W and F475W filters, respectively. Only two dither positions are available across the ACS inter-CCD chip gap resulting in a degraded 3'' band that runs horizontally across the center of every image. Four fields were observed with less than four dither positions due to the failure of the ACS instrument while the sequence was partially complete. In Table 1, we list the details for each field including a unique identifier given by the HST visit number, coordinates, image orientation, number of dithers, and the integrated exposure time in each band.

The dithered exposures were individually dark- and bias-subtracted, and flat-fielded through the standard HST-CALACS pipeline software. The MultiDrizzle software (Koekemoer et al. 2002) was used to combine the dither positions and create a final image. The steps in the MultiDrizzle process include (1) aligning the dithered exposures, (2) identifying cosmic rays using a median filter and "cleaning," (3) weighting pixels from the individual dithers by the inverse variance of the background/instrumental noise, and (4) mapping the weighted exposures to an output grid with a pixel size of 005 (pixfrac = 0.8 and scale = 1.0).

The identification of cosmic rays by the MultiDrizzle software was inefficient in regions of the image not covered by the full dither pattern. The lacosmic routine (van Dokkum 2001) was therefore used to assist with cosmic ray identification within the ACS chip gap and also across the full image for fields with less than four dither positions. A side effect of the lacosmic routine is that pixels associated with extreme surface brightness gradients in the image may be erroneously assigned large (negative or positive) values. Although this affects only a small number of pixels, we have corrected these pixels in the DR2 images by replacing their values with the local median. Owing to erratic behavior, we do not apply lacosmic to regions at the image edge not covered by the full dither pattern, thus the majority of remaining cosmic rays found in DR2 images are located in these regions.

A slight spatial offset between F475W and F814W images (∼0.3 pixels) was discovered soon after the initial data release. This offset has a non-negligible effect on the aperture-matched color measurements of unresolved or marginally resolved objects. Images were realigned by measuring the residual shift between the F475W and F814W frames using compact sources as a reference, and then re-drizzling the F475W band to match the original F814W image. For three fields (visits 3, 10, and 59), it was necessary to re-drizzle the images in both filters to a common pixel position. The DR2 images are now aligned to within ∼0.05 pixels along each axis.

A two-step process was used to calculate the astrometry for each ACS image, starting with the solution calculated by MultiDrizzle and then performing a fine adjustment using the Scamp software (v1.4; Bertin 2006). The initial spatial registration of images was calculated with MultiDrizzle using reference stars in the HST Guide Star Catalog. Astrometry solutions were reliable to only ∼1'' (1σ rms offset, as compared to the Sloan Digital Sky Survey (SDSS) coordinate system) owing to the dearth of reference stars in the Coma fields, which are located at high Galactic latitude (b ∼ 88°).

Second pass solutions for the image astrometry were performed using Scamp with the goal of aligning ACS images to the SDSS coordinate system. Both stars and galaxies from SDSS DR6 were selected as reference objects, which increased the ACS and SDSS matched sample to 18–36 objects for each field. We excluded saturated stars, ACS objects that were blended with nearby sources, or objects located in regions of the ACS image with reduced exposure time. Astrometry solutions were calculated for each field using a range of values for the Scamp input parameters CROSSID_RADIUS, ASTRCLIP_NSIGMA, and FWHM_THRESHOLDS. We forced Scamp to solve the ACS astrometry without using higher order distortion terms (DISTORT_DEGREES = 1) as these values are well constrained by the HST-ACS pipeline (Meurer et al. 2003). We chose the Scamp configuration that resulted in both the "best" reduced chi-squared fit (between 0.9 ≲ χ² ≲ 1.2) and the smallest final offset between ACS and SDSS reference objects.

This procedure was performed in the F814W band and the astrometry solutions were applied to images in both filters. The ACS images are now aligned to the SDSS system with a 1σ rms spread of 01–02 for each coordinate, which is consistent with the precision of the SDSS astrometry for extended objects (Pier et al. 2003).

We created rms maps for every image that are used to establish the detection threshold for source identification and to derive photometric errors. The rms maps were constructed from the inverse-variance maps produced by MultiDrizzle, which are stored in the [WHT] extension of the fits images. The WHT maps estimate the background and instrumental uncertainties related to the flat-field, dark current, read noise, and the effective exposure time across the image.

We must, however, apply a correction to the rms maps in order to derive accurate photometric errors, i.e., the single-pixel rms values cannot be summed in quadrature because MultiDrizzle introduces a correlation of the noise in neighboring pixels (see Fruchter & Hook 2002). We therefore apply a scaling factor to the rms map to recover its uncorrelated value over length scales larger than a few pixels (e.g., Casertano et al. 2000). The corrected rms map is given by

where f is the scalar correction for correlated noise, commonly referred to as the "variance reduction factor."

We have measured the variance reduction factor for a subset of our images by selecting a relatively empty 512 × 512 pixel region inside each image, and then we perform the following steps: (1) normalization of the pixels by the effective exposure time, (2) masking of real objects identified using a median filter, (3) measurement and subtraction of the background, and (4) photometry on its autocorrelation image in order to measure the peak pixel flux relative to the total flux, e.g., the correlation factor is f = peak/total, where a value of 1 indicates that the image is uncorrelated at any pixel scale. We have measured the correction factor for a subset of images in both filters, recovering an average value of f = 0.77 ± 0.08 (1σ rms). The average value is applied to all WHT maps as the correction factor depends primarily on the choice of MultiDrizzle parameters pixfrac and scale (which are constants in our image pipeline).

Flag maps were created for every image in order to identify pixels with low effective exposure time. The effective exposure time varies abruptly across each image owing to regions that are only partially covered by the dither pattern. These regions include the ACS chip gap and the image border, which cover 5%–10% of the total image area. We have added additional flags to identify pixels that may be affected by very bright galaxies (described in Section 3.1) and for pixels in close proximity to the image edge. We have assigned a single binary-coded flag value to each pixel that indicate the following: [1] associated with a bright galaxy, [2] located within 32 pixels of the image edge, [4] an effective exposure time less than 2/3 the total integration time, and [8] zero effective exposure. The order was chosen such that higher flags correspond to regions of the image where object detection and photometry are more likely to have systematic effects.

The main purpose of the flag maps is to identify large regions of each image with similar completeness limits. We note that cosmic rays cleaned by MultiDrizzle often leave a few pixels with low effective exposure, but have little impact on the ability to detect underlying sources. To avoid flagging these pixels, we applied a 7 pixel median filter to the effective exposure map prior to imposing the low-exposure criteria. A few heavily cleaned cosmic ray events remain flagged in each map.

We have subtracted the light distribution from bright galaxies in DR2 images in order to improve object detection and photometry for faint underlying sources (e.g., dwarf galaxies, GCs). The candidates for light subtraction are the 31 Coma member galaxies that have magnitudes brighter than F814W = 15.0. We modeled their light distribution using the software package GALPHOT (Franx et al. 1989) within the Astro-WISE system (Valentijn et al. 2007). GALPHOT was chosen because it is a well-defined standard for isophote fitting (e.g., Jorgensen et al. 1995) and has provided reliable fits for galaxies in HST images (e.g., Sikkema et al. 2007; Balcells et al. 2007). GALPHOT uses harmonic fitting for elliptical and higher-order isophotal light models which has the advantage of not requiring an initial guess (e.g., as compared with the IRAF task ELLIPSE); note that we fit bright galaxies in visit-19 using both GALPHOT and ELLIPSE and visually confirmed that only negligible differences exist between their residual images.

Objects located inside the light distribution of the target galaxy were masked prior to performing our fitting procedure and fits were performed inside image cutouts of each galaxy. We selected the optimal light model by visual inspection of the residual images, and truncated the best-fitting model at the location where it is indistinguishable from the background noise. The best-fit light model was subtracted and the cutout was inserted back into the original image. We could not obtain a satisfactory solution for three bright galaxy candidates: one spiral galaxy was too irregular to be modeled with elliptical isophotes, and the other two galaxies were overlapping thus preventing a good solution. Basic properties of the remaining 28 galaxies are listed in Table 2.

In order to perform reliable source detection within regions of galaxy subtraction, we have corrected the rms maps to account for photon noise from the bright galaxy and fitting errors. Specifically, we added the following term σ²_GSub to the variance map:

where σ_poisson is the Poisson noise taken from the light distribution model, and we have added an additional 5% error to account for an imperfect fit (e.g., Del Burgo et al. 2008). We ran SExtractor on the galaxy-subtracted images (with the updated rms maps) using a configuration that is otherwise identical to that used for the original images.

New source detections were merged with the original source catalog using the following procedure. We defined a circular annulus for each subtracted galaxy with an inner and an outer radius (r_min, r_max) as listed in Table 2. We added to our final source catalog all sources from the galaxy-subtracted catalog that are located in the annulus. For duplicate detections in the annulus, we adopt photometry from the galaxy-subtracted analysis. Objects located within the inner radius were removed as large systematic residuals prevent the reliable detection of sources. At locations beyond the outer radius, we retain the original measurements as both methods include virtually the same real sources. Figure 4 shows a pair of overlapping galaxies both before and after performing galaxy subtraction, including the location of original and new source detections. As a result of subtracting bright galaxies, we have recovered 3291 additional sources that were detected out to 45'' from the centers of the bright galaxies.

Zoom In Zoom Out Reset image size

In the following section, we present an initial attempt to classify objects, real or otherwise, in the DR2.1 source catalogs. Object classifications are stored in the catalog FLAGS_OBJ parameter. A thorough description of these classifications is given below, and a summary of the flag values is also provided in the Appendix. We also discuss spurious sources in our catalogs owing to inaccurate galaxy separation and the chance alignment of random noise.

3.2.1. Extended Galaxies and Point Sources

The SExtractor CLASS_STAR parameter, or stellarity index, is a dimensionless value that classifies objects as extended or point-like based on a neural network analysis that compares the object scale and the image point-spread function (PSF). The parameter varies between 0 and 1, such that extended objects have values near 0 and point-like objects are closer to 1; less reliable classifications are assigned mid-range values. The galaxy/point source classification is improved by also considering the 50% light radius measured in the F814W band. Specifically, we classify as extended galaxies those objects with CLASS_STAR < 0.5 and FLUX_RADIUS_3 > 2.5 pixels (2.5 pix = 0125, or 61 pc at the distance of Coma). These objects are assigned FLAGS_OBJ = 0. Objects that do not meet this criteria are considered point sources and are assigned FLAGS_OBJ = 1. From visual inspection we estimate that this classification scheme is ∼95% reliable for objects brighter than F814W = 24.0 mag. The CLASS_STAR parameter is not reliable at fainter magnitudes, thus we use the 50% light radius criteria alone to classify these objects. According to this classification scheme, the slight majority of all detections are unresolved (60%).

In Figure 5, we have plotted catalog sources in a magnitude–size diagram separated as extended galaxies and point sources. We have visually inspected objects that lie outside their typical magnitude–size parameter space, and adjusted FLAGS_OBJ for sources that were obviously misclassified. Our classification scheme clearly separates extended galaxies and intrinsically compact sources at magnitudes brighter than F814W ≈25; at fainter magnitudes, the point source class includes an increasing population of unresolved galaxies. This can be seen more clearly from the magnitude histogram for extended galaxies and point sources shown in Figure 6. The histogram for the extended galaxy population flattens at F814W = 25 mag, or ∼1–1.5 mag brighter than the total source population, owing to a higher fraction of unresolved galaxies. The relatively sharp increase in the point source population at magnitudes fainter F814W = 22.5 may initially reflect the UCD population in the Coma cluster (K. Chiboucas et al. 2011, in preparation), and then the large number of GCs in the cluster with F814W ≳ 24 mag (Paper IV).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.2.2. Cosmic Rays

3.2.3. Image Artifacts

3.2.4. Galaxy Shredding

3.2.5. Random Noise

We have compared our catalog to previous optical surveys as a check on the accuracy of the SExtractor photometry. Several optical photometric surveys have been performed in the Coma cluster (e.g., Godwin et al. 1983; Terlevich et al. 2001; Komiyama et al. 2002), but we limit this comparison to the surveys with I-band observations that closely match the ACS F814W band, such as the Canada–France–Hawaii Telescope (CFHT) deep-field coverage of the central region in Coma (Adami et al. 2006a) and SDSS DR6 wide-field coverage of Coma (Adelman-McCarthy et al. 2008). We matched ACS-CFHT galaxies using a 3'' search radius and ACS-SDSS galaxies with a 06 radius, and limited the comparison to ACS regions with nominal exposure time and objects with no SExtractor flags.

In Figure 7, we show the magnitude offset between our photometry and both CFHT and SDSS as a function of F814W magnitude. The Adami et al. (2006a) CFHT observations are significantly deeper and have better spatial resolution than SDSS, resulting in less scatter at a given ACS magnitude. The ACS-CFHT magnitude comparison shows systematic offsets at both bright and faint magnitudes. From visual inspection, the faint magnitude offset results from both blended objects in the Adami et al. (2006a) catalog or shredded galaxies in our catalog. The offset at bright magnitudes likely results from saturated galaxies in the CFHT observations, which Adami et al. (2006a) had replaced with measurements from previous studies. In contrast, the ACS-SDSS photometry have good agreement at bright magnitudes (F814W⩽17), but the systematic offset at fainter magnitudes indicates that the SDSS flux may be slightly underestimated. For objects with unreliable ACS measurements, we therefore recommend adopting the SDSS photometry at bright magnitudes (F814W⩽17 mag; F814W_AB = SDSS_i − 0.01 ± 0.05) and CFHT photometry for fainter sources (17 < F814W < 22.75 mag; F814W_AB = CFHT_I + 0.47 ± 0.06–0.20 across this magnitude range).

Zoom In Zoom Out Reset image size

It is well known that the SExtractor detection efficiency is a function of object size and magnitude (e.g., Bershady et al. 1998; Cristóbal et al. 2003; Eliche-Moral et al. 2006). As such, we have measured the detection efficiency separately for models with similar size and magnitude, as well as Sérsic index and ellipticity, and discuss how the detection efficiency depends on these parameters.

Figure 8 shows the detection efficiency versus magnitude for our models. We have separated the models into four subsets based on Sérsic index and ellipticity, which throughout this section are referred to as models FC, FE, PC, and PE: flat (F) and peak (P) models have Sérsic indices that are smaller/larger than n_Ser = 2.25, respectively; circular (C) and elliptical (E) models have ellipticities smaller/larger than e = 0.4, respectively. The panels in Figure 8 show the detection efficiency for each model subset, which we further separate into six equally spaced bins in logarithmic effective radius that span the range R_eff = 0.5–60 pixels. The left-most curves in each panel trace the largest R_eff , which for a given magnitude represent galaxies with the lowest surface brightness. All efficiency curves show the same pattern: at the bright end efficiencies are near 100%, and then fall to zero with a similar slope. None of the curves reaches 100% efficiency (maximum rates are ∼96%–98%) owing to artificial sources that were injected near relatively brighter galaxies. We note that our SExtractor catalogs reach 100% efficiency for moderately bright objects upon subtracting the brightest 28 galaxies from our images prior to source detection (Section 3.1). For a given magnitude, the detection efficiency depends primarily on object size (small R_eff gives higher efficiency), followed by Sérsic index (models with large n_Ser have higher efficiencies), and to lesser extent on the ellipticity (high-ellipticity models have slightly higher detection rates), i.e., the detection efficiency is positively correlated with any model parameter that increases the central surface brightness.

Zoom In Zoom Out Reset image size

In Table 3, we list the expected 80% completeness limits for our SExtractor catalogs based on our simulations. The completeness limits are given for the four model subgroups (FC, FE, PC, and PE) separated into the six logarithmically spaced bins in R_eff (F814W results are shown in Columns 2–7, and F475W in Columns 8–13). The 80% completeness limits for point sources occur at F475W = 27.8 and F814W = 26.8 mag for images with the nominal four dither positions (corresponding to an average S/N = 5). The F814W 80% completeness limit is consistent with the peak in the magnitude histogram shown earlier in Figure 2.

Next we have tested the accuracy of the SExtractor photometry by comparing its flux measurements to the magnitudes of the artificial objects. The SExtractor Kron photometry (Kron 1980) is known to capture a varying fraction of the total light, which is a function of both the Sérsic profile and S/N, e.g., the SExtractor Kron apertures capture relatively less light for galaxies with high Sérsic index or low surface brightness (see Graham & Driver 2005). In this section, we derive corrections for the SExtractor Kron magnitudes that provide a better estimate of total magnitudes. The following analysis is performed for F814W measurements alone, as SExtractor apertures were not defined in the F475W band.

In Figure 9, we show the difference between the SExtractor Kron magnitude and the input magnitude plotted against the SExtractor mean effective surface brightness in the F814W band. The mean effective surface brightness (mag arcsec⁻¹) is defined as

where MAG_AUTO and FLUX_RADIUS_3 are the SExtractor catalog entries for Kron magnitude and the 50% light radius (in pixels), respectively, and PSCALE = 0.05 arcsec pixel⁻¹ is the plate scale of our ACS images. We have separated the comparison into four subsets based on Sérsic index and ellipticity (models FC, FE, PC, and PE as described in Section 4.1). We have trimmed the suite of models used in this analysis to include only those that overlap in R_eff–magnitude space with galaxies detected in other deep HST surveys, e.g., the Hubble Ultra Deep Field (Beckwith et al. 2006), the Groth Strip Survey (Simard et al. 2002; Vogt et al. 2005), and the Galaxy Evolution through Morphology and SEDs (GEMS; Rix et al. 2004) surveys. This was done so that corrections are estimated from models that are representative of the objects potentially detectable in our survey; the parameter space covered by these models includes the dwarf LSB galaxy population in the Coma cluster (e.g., Graham & Guzmán 2003).

Zoom In Zoom Out Reset image size

The four panels in Figure 9 show a similar pattern, viz. photometric errors increase toward fainter surface brightness, with a systematic bias of the recovered magnitudes toward fainter values. We also note the following.

• 1.  
  Despite that objects with large Sérsic indices ("P" models) have a higher detection efficiency, their SExtractor photometry have both large systematic offset in magnitude and large scatter as compared to other models.
• 2.  
  The systematic magnitude offset for models with low Sérsic index ("F" models) is relatively insensitive to surface brightness (i.e., does not exceed 0.1 mag) for models brighter than 〈μ〉^SE_e = 23.0.
• 3.  
  The magnitudes of high-ellipticity objects ("E" models) are better recovered than the magnitudes for objects with low ellipticity ("C" models). The typical magnitude residual of a high-ellipticity model is similar to the magnitude residual of a low-ellipticity model around 0.5 mag brighter in 〈μ〉^SE_e.
• 4.  
  The surface brightness detection limits are the same for all four models (〈μ〉^SE_e = 24.75).

In order to correct the SExtractor photometry for missing light we require the true effective radius (R_eff) and the Sérsic index of the galaxy (e.g., Graham & Driver 2005). The SExtractor half-light radius is a proxy for R_eff but is often underestimated at faint magnitudes, and also for LSB objects, owing to relatively fewer pixels that satisfy the detection threshold. Instead, we derive an empirical relation between the input effective radius of our models and the output SExtractor half-light radius (FLUX_RADIUS_3), which is a function of the Sérsic profile, ellipticity, and 〈μ〉^SE_e. This relationship is shown in Figure 10, which was fit with a fifth-order polynomial in 〈μ〉^SE_e for each of the four model groups. It is seen that the FLUX_RADIUS_3 parameter provides a better estimate of the true R_eff for flat (FE or FC) galaxy models than those with higher Sérsic indices. The best-fit conversion factor is also more or less constant for flat models brighter than 〈μ〉^SE_e = 23, while it is a steadily growing function of 〈μ〉^SE_e for high-Sérsic peak models (PC or PE). This is consistent with the systematic magnitude offsets seen in Figure 9. There are also secondary effects for the galaxy ellipticity, such that SExtractor recovers the true effective radius slightly better for low-ellipticity models. We conclude that we are able to estimate the true effective radius with good accuracy (∼20% rms error in the R_eff/FLUX_RADIUS_3 ratio) based on the SExtractor measurements alone, and thus perform a variable aperture correction to our catalog sources.

Zoom In Zoom Out Reset image size

We have tested the reliability of performing a variable aperture correction on our suite of galaxy models. The fraction of missing light is calculated based on the ratio of the SExtractor Kron radius and the derived effective radius, and taking the known Sérsic index of the models (e.g., see Figure 9 in Graham & Driver 2005). The results are shown in Figure 11, which shows the magnitude difference between the aperture-corrected SExtractor magnitudes and the input magnitude as a function of 〈μ〉^SE_e. The magnitude residuals now follow a well-defined "trumpet" diagram and are not nearly as dependent on galaxy structure/shape as shown in the raw magnitude comparison in Figure 9. There still exists a small systematic offset for all models at low surface brightness (especially for PE models), for which we will apply a fine correction based on the average residuals measured in bins of surface brightness. Our aperture correction is reliable for objects with surface brightness brighter than 〈μ〉^SE_e = 23.25 in the F814W band. This limit corresponds to our ability to recover the total magnitude with a 1σ rms error of 0.2 mag, but the uncertainties increase rapidly for fainter objects.

Zoom In Zoom Out Reset image size

We have applied this variable aperture correction to all sources in our SExtractor catalogs. We do not have prior knowledge of the Sérsic profile; thus we have performed two aperture corrections for every object using the following values for the Sérsic profile: (1) a flat profile (n_Ser = 1.525) that is representative of the majority of objects in our catalog (MAG_AUTO_CORRA), and (2) a variable Sérsic profile given by Graham & Guzmán (2003) that scales with absolute B-band magnitude and is applicable to early-type galaxies in the Coma cluster (MAG_AUTO_CORRB); specifically, we use the formula F475W = −9.4 log n_Ser + 20.6, assuming a color offset F475W = B − 0.1 for early-type galaxies in Coma and a distance modulus of 35.00 mag. The first correction is a statistical estimate of the total magnitude and is not meant for accurate measurements of individual galaxies. The second correction provides a reliable estimate of the total magnitude for individual early-type galaxies in the Coma cluster that are brighter than F814W = 20; fainter Coma member galaxies have surface brightness values that are lower than the reliable limit of our aperture correction (〈μ〉^SE_e = 23.25). The errors for the SExtractor corrected magnitudes are taken directly from the measured rms scatter shown in Figure 11. These errors account not only for the counting errors that are measured accurately by SExtractor, but now include uncertainties related to our ability to recover the true magnitude. We include an additional 0.05 mag error for both methods to account for the uncertainty of the Sérsic profile. The corrections for F475W Kron magnitudes may be estimated indirectly by adding the F475W−F814W color to the corrected F814W Kron magnitude (assuming the galaxy has a negligible color gradient across the SExtractor Kron aperture).

In Figure 12, we show the color–magnitude diagram (CMD) for all point sources and extended galaxies in our catalog, including members of the Coma cluster. Colors are measured inside the SExtractor isophotal apertures as defined in the F814W band. The GC population is identified as the concentration of point sources at magnitudes fainter than F814W = 24 mag with color values near F475W−F814W ≈ 1 (Paper IV). We suspect that unresolved background galaxies are responsible for the extension of point sources to bluer colors at magnitudes fainter than F814W = 25. Stars are scattered across other regions of the CMD.

Zoom In Zoom Out Reset image size

The cluster red sequence (RS) is easily identified as the dense horizontal band of cluster galaxies that stretch between 13 mag ≲ F814W ≲ 22.5 mag. It is well known that the cluster RS consists primarily of quiescent early-type galaxies, and that massive RS galaxies (as traced by luminosity) have redder colors owing to their deep potential wells that are able to retain metals for future generations of stars with cooler main-sequence turnoffs (e.g., Bower et al. 1992; Kodama & Arimoto 1997). In contrast, RS dwarf galaxies have bluer colors owing to both feedback mechanisms that remove processed elements (e.g., supernovae winds; Dekel & Silk 1986) and more extended star formation histories (e.g., Smith et al. 2008; Sánchez-Blázquez et al. 2009).

Figure 13 shows a separate CMD for objects in the Coma cluster. Magnitudes and colors are adjusted for Galactic reddening and K-corrections, and we use the Kron magnitudes corrected for missing light (Section 4.2). Colors were measured inside fixed circular apertures with sizes that were selected to match the typical R_eff of cluster RS galaxies: a 30 radius (1.4 kpc) for galaxies brighter than F814W = 16 mag, a 225 radius (1.0 kpc) for galaxies between 16 mag < F814W < 20 mag, and a 15 radius (0.7 kpc) for fainter galaxies (e.g., Graham et al. 2003; Ferrarese et al. 2006). Fixed apertures avoid the color bias owing to variable apertures that may not sample the same physical regions in galaxies (Scodeggio 2001). The slope of the RS appears constant across the full magnitude range, which confirms the results of previous studies (e.g., Secker et al. 1997; López-Cruz et al. 2004; Adami et al. 2006a), but extends the relation to a few magnitudes fainter. A robust bisector fit to the full RS gives the relation F475W−F814W = −0.060 × F814W + 2.12 mag, with a 1σ dispersion of 0.06 mag. This relation is not affected by the relatively large errors for the corrected Kron photometry at faint magnitudes, e.g., we obtain an identical fit to the RS for galaxies brighter than F814W = 20 mag.

Zoom In Zoom Out Reset image size

Interestingly, the RS dispersion remains roughly constant (0.06 mag) when measured in discrete magnitude intervals across the full RS. This result is in contrast to the relatively large 0.3 mag dispersion in B − R color reported for faint (R ∼ 22 mag) dwarf galaxies of both the Coma cluster (Adami et al. 2009b) and Perseus cluster (Conselice et al. 2003), which were attributed to either varied star formation histories or effects from dynamical interactions. Large uncertainties for their color measurements, or contamination from background galaxies, may have obscured an underlying tighter relation with the RS. Other possible explanations for the low dispersion reported here are (1) our redshift coverage and visual classification scheme may preferentially select faint Coma member galaxies that lie near the RS, and (2) our ACS fields do not cover the regions of the Coma cluster where many faint dwarf galaxies were reported to be offset from the RS (e.g., west of the cluster center, and in the immediate vicinity of NCG 4889; Adami et al. 2009a). Future studies will provide a more accurate analysis of the RS using magnitudes and colors obtained by fitting the light distribution of each individual Coma member galaxy.

In Figure 13, we have separated Coma members into broad morphology classes based on visual inspection and following the galaxy templates of Sandage & Binggeli (1984). This classification scheme is only meant to provide a cursory description of the Coma cluster morphologies and to identify large-scale trends. Future papers will provide less subjective classifications based on multi-component fits to the light distributions of Coma member galaxies. As expected, early-type galaxies (E/S0) are the majority of bright RS galaxies and most are located in the central ACS fields, while spiral galaxies (Sa-Sd) are located below the RS and were detected primarily in the outer ACS fields. Although we did not consider the SExtractor measurements as part of our classification, our E/S0 and dE galaxy classes can be reproduced with a separation in surface brightness at 〈μ〉^SE_e = 22 mag arcsec⁻¹ in the F814W band. Galaxies classified as dEs all have magnitudes fainter than F814W ≈ 18 (M_F814W = −17), which corresponds to a B-band magnitude of M_B ≈ −16. The dE galaxies were subdivided into two classes: (1) nucleated dE galaxies (dE,N) have unresolved central nuclei and are the dominant subclass, especially among bright dE galaxies, and (2) non-nucleated dE galaxies (dE,NN) lack a visible central nucleus but are otherwise similar in appearance to dE,N galaxies, and are an increasing fraction of total dE population at fainter magnitudes. Compact galaxies are located primarily above the RS and consist of the seven cE candidates discussed in Price et al. (2009) and another three UCD-like objects that have redshifts. Ten galaxies have irregular morphology but are conservatively classified as Im/dE as it is difficult to differentiate between these galaxy types at faint magnitudes.

Analysis of the subgroups of early-type galaxies requires more detailed measurements than presented here, but we note that our classes show no obvious trends in the CMD shown in Figure 13. The exception, however, relates to the RS outliers at intermediate magnitudes (17.5 ≲ F814W ≲ 21) as (1) several E/S0 and dE,N galaxies are located above the 1σ RS but are rare below the RS, and (2) there is a relative lack of dE,NN galaxies outside the 1σ RS. Interestingly, the majority of these outliers occupy the same region of the CMD as cEs and UCDS, which are thought to originate from the tidally stripped nuclei of E/S0 and dE,N galaxies (e.g., Bekki et al. 2001, 2003; Price et al. 2009). It is intriguing to speculate that the outliers above the RS may have experienced at least low-level tidal interactions, or represent a phase in cE/UCD creation, but such analysis is beyond the scope of this paper and will be explored in future works.

Nearly all galaxies classified as Im/dE (9/10) are located either above or below the 1σ RS. These galaxies may be analogs of the scattered population of faint dwarf LSB galaxies detected in the Coma and the Perseus clusters (Adami et al. 2009a; Conselice et al. 2003), albeit as a much smaller fraction of the total galaxy population than reported in these other studies. Adami et al. (2006b) suggested that tidal stripping may be responsible for similar faint red galaxies in their sample, while the galaxies blueward of the 1σ RS may be tidal dwarf galaxies formed from the stripped disk material. Our observations are somewhat consistent with this scenario such that, unlike the bright blue spirals that are found primarily in the outer ACS fields, all of the Im/dE galaxies (blue or red) were detected in the central ACS fields where interactions are more common. The nature of these galaxies will be studied in future papers of this series.

A graphical representation of the CMD is presented in Figure 14, which shows two-color postage-stamp images for a subset of our Coma member galaxies. The color axis is normalized to the RS and color intervals are in standard deviations of the measured dispersion (1σ = 0.06 mag). This diagram demonstrates the wide range of morphology, star formation activity, and evolution of cluster members detected in our survey, such as bright E/S0 galaxies and barred disk galaxies that follow the RS and extend to the dE galaxy population at fainter magnitudes (middle row); compact galaxies located above the 3σ RS (top row) that include cEs at intermediate magnitudes and a UCD at the extreme faint end; star-forming galaxies located below the 3σ RS (bottom row); and more rare galaxies in our sample such as possible interacting galaxies (Row 1, Column 2), an anemic spiral (Row 2, Column 2), and a faint dwarf irregular galaxy located well below the RS (bottom-right corner).

Zoom In Zoom Out Reset image size