Brought to Light. III. Colors of Disk and Clump Substructures in Dwarf Early-type Galaxies of the Fornax Cluster

The FDS (Iodice et al. 2016; Peletier et al. 2020) is a joint effort of the guaranteed-time observation surveys FOCUS (P.I. R. F. Peletier) and VEGAS (P.I. M. Capaccioli; Capaccioli et al. 2015), carried out with the Very Large Telescope Survey Telescope (Arnaboldi et al. 1998) located at the ESO, Paranal. Using the OmegaCAM instrument (Kuijken et al. 2002; Kuijken 2011), the FDS targets the Fornax galaxy cluster out to its virial radius (0.7 Mpc; Drinkwater et al. 2001), and also the Fornax A infalling subgroup. The FDS data consist in deep, multiband imaging in the u, g, r, and i bands, although the Fornax A subgroup lacks u-band imaging. With OmegaCAM having a field of view of 1 × 1 deg², the FDS consists of a total of 32 fields of this size. In practice, we only work with 25 fields, as these are the ones covered by the FDS dwarf galaxy catalog (see following Section 2.2).

After assessing the available imaging data and associated metadata, we decide to discard the u- and i-band data, since, when compared to the g- and r-band data, the former (a) are shallower, and (b) lack an analytical parameterization of the point-spread function (PSF) core and wings. The ideal situation would be to be able to use all available bands, as more bands means more colors, and consequently better constraints on the properties, such as age and metallicity, of the stellar populations of the galaxies. However, the aforementioned restrictions of the u and i-band data (plus the limited spatial coverage of the Fornax cluster in the u band) cements our choice to focus our analysis solely on the g- and r-band reduced, background-subtracted, and flux-calibrated FDS images. For a detailed description of the data reduction, we refer to Venhola et al. (2018).

The median depth and median PSF FWHM of the FDS fields in the g and r bands are provided in Table A.1 of Venhola et al. (2018). The depth is defined as the surface brightness of a source with a signal-to-noise ratio (S/N) = 1 at the pixel scale of the FDS data, which corresponds to 0''.2 pixel⁻¹. After correcting for Galactic foreground extinction (for details, see Section 3.6), the g-band depth lies in the range of 25.9–26.8 mag arcsec⁻², with a median of μ_g = 26.6 mag arcsec⁻², while the r-band depth lies in the range of 25.5–26.2 mag arcsec⁻², with a median of μ_r = 26.0 mag arcsec⁻². Therefore, the g-band imaging is on average about half a magnitude deeper. With regard to the FWHM of the PSF, the g band has a median FWHM_g = 1''.11, while the r band has slightly better seeing with a median FWHM_r = 0''.95.

The FDSDC (Venhola et al. 2018) contains a total of 564 dwarf galaxies spread throughout 26 FDS fields. Their membership to the cluster has been established based on the distribution and correlation of their properties (such as brightness, color, size, and concentration), which allows one to distinguish them from background galaxies. The catalog considers all galaxies with an r-band absolute magnitude fainter than M_r = −18.5 mag to be dwarfs, and reaches a 50% completeness limit at M_r = −10.5 mag and at a limiting r-band surface brightness of μ_r = 26.0 mag arcsec⁻². For more details on the construction and overall characteristics of the FDSDC, we refer to Venhola et al. (2018).

For each dwarf galaxy, the catalog provides its ICRS coordinates, several structural and photometric parameters, and a morphological classification. The structural parameters (such as Sérsic index, effective radius, axis ratio, and position angle) were derived from the r band with GALFIT (Peng et al. 2002), using either a single Sérsic function (Sérsic 1968) or a combination of a Sérsic function and a PSF function if a nucleus is present. The photometric parameters were then derived by using the structural information provided by the r-band fit, and correspond to the apparent magnitude of the galaxy in the g and r bands, the apparent magnitude of its nucleus (if present) in the g and r bands, and the apparent magnitude within the effective radius in the u, g, r, and i bands. These photometric measurements are given in the SDSS u, g, r, and i filters calibrated to the AB magnitude system, and are not corrected for Galactic foreground extinction. Additionally, based on the appearance and overall color of the galaxies, they are classified into early and late morphological types.

In practice, we work with the FDSDC information of 560 dwarf galaxies located in 25 FDS fields. Our analysis excludes the four dwarf galaxies contained in field 33, as the inner and outer regions of the PSF were not modeled for this particular field. Nonetheless, a quick check reveals that no substructure features are embedded in these four galaxies, so no loss is incurred by excluding them.

The first step is to construct cutouts of the g- and r-band FDS images of each dwarf galaxy present in the catalog using its coordinates and the World Coordinate System provided in the FDS image headers. We choose to make the size of each cutout a variable parameter that depends on the overall extension of the galaxy. Consequently, for each of the 560 FDSDC galaxies, we construct cutouts of their g- and r-band FDS images with dimensions 20 R_e × 20 R_e , where R_e is the effective radius ⁵ in the r band as listed in Table 2. These dimensions ensure that the galaxy is well contained within the cutout, and that the cutout samples the local background residuals.

To be able to perform reliable measurements in the g- and r-band galaxy cutouts simultaneously, it is necessary to take into account their different PSFs. For this purpose, we use the analytic functions describing the core and the wings of the PSF separately as provided by Venhola et al. (2018). The PSF core, extending over the central 10'' of the PSF, is modeled by the sum of a Gaussian function with a Moffat function. The PSF core parameters of each FDS field in the g and r bands are provided in Table A.2 of Venhola et al. (2018). In some cases, a Gaussian function is not required in order to describe properly the PSF core, so only a Moffat function is used. The PSF wings, defined as the radial region beyond 40'', are characterized by an exponential function, whose parameters are the same for all the FDS fields, and are provided in Section 5.1.2 of Venhola et al. (2018).

In order to match the g- and r-band PSFs of each galaxy image cutout, we make use of the Python utilities provided by the astropy package and its affiliated package photutils (Bradley et al. 2021). We first create a 2D model of the composite PSF for each FDS field and band. Then, for each galaxy, we translate their g- and r-band composite PSF models into kernel images, where the kernel size is a variable parameter that depends on the observable galaxy size. As the process of convolution redistributes the galaxy light across the image, it is important that the size of the convolution kernel matches the overall observable extension of the galaxy, so that no light becomes unnecessarily dispersed and lost beyond the galaxy's boundaries. For this reason, we choose to create g- and r-band PSF kernel images of dimensions 6 R_e × 6 R_e (i.e., of 3 R_e in radial length) for each galaxy. As a consequence, all measurements and analyses in the following sections are performed only out to a radial extension of 3 R_e .

The next step consists of identifying, for each galaxy, which band has the narrowest (i.e., best) PSF, and which has the broadest (i.e., worst) PSF. This is necessary as the quality of the FDS imaging varies from field to field, so while in some fields the g-band imaging has the narrowest PSF and the r band has the broadest PSF, in other fields it can be the other way around. We evaluate the PSF quality by comparing the g- and r-band PSF FWHMs provided in Table A.1 of Venhola et al. (2018) for each FDS field. Once we have identified the narrow and the broad PSF, we create a matching kernel image from the previously constructed g- and r-band PSF kernel images. Finally, to match the PSFs of the two bands, we convolve the galaxy image cutout of the band that has the narrowest PSF with the matching PSF kernel that we just constructed, thus effectively degrading its PSF to match the band that has the broadest PSF. As a result, we obtain PSF-matched g- and r-band image cutouts for each of the FDSDC galaxies.

The PSF-matched g- and r-band galaxy cutouts we created allow us to make reliable photometric measurements in both bands simultaneously. However, we are also interested in stacking both bands in order to create a single g + r galaxy image with increased S/N and depth. As we aim to analyze substructure features that are intrinsically faint, the increase in depth allows us to better detect and identify them. Moreover, the stacking ensures that both bands are taken into account during the substructure detection process. At this stage, we still have no knowledge about the brightness and color of the substructure features. This means that we do not know in which band they are brighter, or if there is a band in which they are barely visible. The impact of these unknowns is minimized when we combine the contribution of both bands into one image.

We thus construct a g + r image for each galaxy by coadding their PSF-matched g- and r-band image cutouts. We adopt a simple approach when stacking: the g and r bands are assigned the same weights, so their contribution to the total flux of the final coadded image is the same. On average, the g + r coadded images are able to reach 0.5 mag deeper when compared to the r-band images.

Unsharp masking constitutes an efficient way to assess if faint substructure features are embedded in the bright diffuse body of a galaxy. The creation of an unsharp mask image requires first to smooth the galaxy image, and then to divide the original galaxy image by the smoothed-out version of itself. This way, the majority of the smooth light is removed, thus revealing any nonsmooth features that would normally lie hidden to the naked eye.

Using utilities from the astropy package in Python, we create a series of unsharp mask images from the g + r coadded image of the FDSDC galaxies. As the size of the smoothing kernel affects the width of the substructures that are revealed, we adopt a range of smoothing kernel sizes. For each FDS field and band, we transform their PSF FWHM into a Gaussian standard deviation σ_PSF. Then, we create five different Gaussian smoothing kernels, with standard deviations corresponding to 1, 3, 5, 7, and 9 × σ_PSF of the PSF, respectively. With regard to the shape and orientation of the kernels, we adopt the axis ratio and position angle given by the FDSDC, in order to match the overall geometry of the galaxy when smoothing. As a result, we obtain five unsharp mask images for each FDSDC galaxy, where each unsharp mask is tuned to reveal the potential substructure features of a particular width and extent.

For each of the 560 FDSDC galaxies, we visually inspect their g + r coadded image and their associated unsharp mask images. To become part of our sample of Fornax dwarf ETGs with substructures, a dwarf galaxy must comply with the following three criteria:

• 1.  
  Red color ( g − r ≥ 0.5 mag). The galaxy must have a red global color, which we define as g − r ≥ 0.5 mag. The total integrated g − r color of the galaxy is derived from their g- and r-band apparent magnitudes reported by the FDSDC.
• 2.  
  Early-type morphology. In the g + r coadded image, the galaxy must exhibit an overall early-type morphology, such as that of a dE, dS0, or dSph. The morphological appearance is assessed based on a visual inspection of the image.
• 3.  
  Disk or clump substructure features. In the unsharp mask images, the galaxy must show either disk or clump substructure features. Disk substructures encompass bars, spiral arms, rings, and dumbbells, as per the definition of Lisker et al. (2006b). Clump substructures encompass irregular light overdensities such as star-forming regions, dust lanes, and off-center nuclei.

According to the morphological classification of the FDSDC, there are a total of 194 dwarf galaxies brighter than M_r = − 13 mag that simultaneously fulfill criteria (1) and (2), i.e., that are red dwarf ETGs. Out of these, 23 dwarf ETGs also show substructures as described in criterion (3), which amounts to 12% of them. These 23 dwarf ETGs constitute our working sample, where 9 have disk substructures (40% of the sample) and 14 have clump substructures (60% of the sample). For a more detailed look into the magnitude distribution of these galaxies, refer to Figure 7 at the end of Section 4.2. Finally, for purposes of this work, we shall from now assume that the Fornax dwarf ETG population is constituted by (a) all of the dwarf galaxies morphologically classified as early types in the FDSDC, plus (b) all of the galaxies in our sample.

In order to run the "residual method" (Michea et al. 2021) on our sample, we first need to mask out the foreground and background sources in the projected vicinity of our objects, so that the subsequent measurements and analyses of the galaxy images will not be contaminated by their light. We do so by visually identifying interloping sources in the g + r coadded images of our sample, whose positions and extents are manually registered and then translated into a bad pixel mask (BPM) image.

We use this BPM image in the IRAF STSDAS ellipse task (based on Jedrzejewski 1987) with which we independently fit the PSF-matched g- and r-band images of each sample galaxy, while allowing their center coordinates, isophote ellipticity, and isophote position angle to change freely with galactocentric radius. We then use the IRAF bmodel task to build the g- and r-band galaxy model images from the fitted isophotes. Finally, we replace the values of the bad pixels in the PSF-matched g- and r-band galaxy images with their corresponding values in the g- and r-band galaxy models, respectively. By masking the interloping sources, we obtain PSF-matched g- and r-band galaxy images that have a much cleaner and regular appearance. Nonetheless, we note that the regions marked on the BPM images are still omitted during any kind of measurement or analysis.

We also need to correct for Galactic foreground extinction, as both the FDS images and the photometric measurements provided by the FDSDC do not take dust reddening into account. For this purpose, we employ the NASA/IPAC Extragalactic Database Extinction Calculator, ⁶ based on the Galactic extinction maps of Schlafly & Finkbeiner (2011) and the Galactic extinction law of Fitzpatrick (1999) with R_V = 3.1, and compute the A_g and A_r extinction coefficients for each galaxy in our sample. In the g band, we find that the extinction coefficient varies between 0.015 and 0.079 mag across our sample, with a median value of A_g = 0.040 mag. In the r band, it lies in the range of 0.010–0.055 mag, with a median value of A_r = 0.028 mag. In the sections that follow, all photometric measurements are corrected for Galactic foreground extinction and are given in the AB magnitude system.

The photometric properties of our dwarf ETG sample are provided in Table 1. From the g- and r-band integrated apparent magnitudes given by the FDSDC, we derive the g- and r-band integrated absolute magnitudes by assuming a Fornax cluster distance modulus of 31.51 ± 0.15 mag (equivalent to a distance of 20.0 ± 1.4 Mpc; Blakeslee et al. 2009). We assume this Fornax cluster distance from here onwards. We also derive the integrated g − r color of the galaxies. By construction, the galaxies in our sample were selected based on their integrated red colors, and thus span the range 0.51 ≤ g − r ≤ 0.76 mag, with a median of g − r = 0.61 mag. Being dwarf systems, they are also characteristically faint. Their absolute magnitudes in the g band lie in the range between −14.3 ≥ M_g ≥ −18.0 mag, with a median of M_g = −16.8 mag; while in the r band, they cover the range between −14.9 ≥ M_r ≥ −18.7 mag, with a median of M_r = −17.5 mag.

Table 1. Central Coordinates and Photometric Properties of Our Fornax Dwarf ETG Sample

Dwarf Galaxy	Central Coordinates			μe			m			M		g − r
	(J2000.0)			(mag arcsec−2)			(mag)			(mag)		(mag)
	R.A.	Decl.		g Band	r Band		g Band	r Band		g Band	r Band
(1)	(2)	(3)		(4)	(5)		(6)	(7)		(8)	(9)	(10)
F01D145	03h46m3338	$-34^\circ 41^{\prime} 10\buildrel{\prime\prime}\over{.} 32$		24.52	23.90		17.23 ± 0.24	16.66 ± 0.21		−14.28 ± 0.28	−14.85 ± 0.25	0.57 ± 0.31
F02D000	03h50m3672	$-35^\circ 54^{\prime} 33\buildrel{\prime\prime}\over{.} 84$		22.23	21.53		14.17 ± 0.11	13.51 ± 0.09		−17.34 ± 0.19	−18.00 ± 0.18	0.66 ± 0.15
F04D000	03h42m4555	$-33^\circ 55^{\prime} 12\buildrel{\prime\prime}\over{.} 36$		22.49	21.83		14.49 ± 0.11	13.84 ± 0.09		−17.02 ± 0.19	−17.67 ± 0.18	0.65 ± 0.14
F04D001	03h41m360	$-33^\circ 46^{\prime} 44\buildrel{\prime\prime}\over{.} 76$		21.66	20.99		13.48 ± 0.10	12.83 ± 0.09		−18.03 ± 0.18	−18.68 ± 0.17	0.65 ± 0.13
F04D002	03h43m2266	$-33^\circ 56^{\prime} 19\buildrel{\prime\prime}\over{.} 68$		22.46	21.77		15.22 ± 0.15	14.55 ± 0.12		−16.29 ± 0.21	−16.96 ± 0.19	0.67 ± 0.19
F04D053	03h41m4541	$-33^\circ 47^{\prime} 29\buildrel{\prime\prime}\over{.} 40$		24.24	23.67		15.90 ± 0.16	15.35 ± 0.14		−15.61 ± 0.22	−16.16 ± 0.21	0.55 ± 0.21
F04D061	03h41m2119	$-33^\circ 46^{\prime} 9\buildrel{\prime\prime}\over{.} 84$		24.29	23.71		15.61 ± 0.17	15.03 ± 0.15		−15.90 ± 0.23	−16.48 ± 0.21	0.57 ± 0.22
F05D000	03h41m3254	$-34^\circ 53^{\prime} 19\buildrel{\prime\prime}\over{.} 68$		21.99	21.41		13.74 ± 0.11	13.20 ± 0.09		−17.77 ± 0.18	−18.31 ± 0.18	0.54 ± 0.14
F07D000	03h45m358	$-35^\circ 58^{\prime} 21\buildrel{\prime\prime}\over{.} 72$		20.79	20.08		13.76 ± 0.07	13.15 ± 0.06		−17.75 ± 0.17	−18.36 ± 0.16	0.62 ± 0.09
F09D255	03h36m4973	$-33^\circ 27^{\prime} 39\buildrel{\prime\prime}\over{.} 24$		23.44	22.81		16.34 ± 0.16	15.74 ± 0.14		−15.17 ± 0.22	−15.77 ± 0.21	0.60 ± 0.22
F09D492	03h39m5503	$-33^\circ 03^{\prime} 11\buildrel{\prime\prime}\over{.} 88$		23.26	22.69		16.81 ± 0.14	16.26 ± 0.12		−14.70 ± 0.21	−15.25 ± 0.19	0.54 ± 0.19
F10D189	03h38m917	$-34^\circ 31^{\prime} 7\buildrel{\prime\prime}\over{.} 68$		23.52	22.84		15.20 ± 0.15	14.58 ± 0.13		−16.31 ± 0.21	−16.93 ± 0.20	0.62 ± 0.20
F11D279	03h36m5431	$-35^\circ 22^{\prime} 28\buildrel{\prime\prime}\over{.} 92$		22.41	21.64		14.35 ± 0.09	13.59 ± 0.08		−17.16 ± 0.18	−17.92 ± 0.17	0.76 ± 0.12
F14D144	03h33m3406	$-33^\circ 34^{\prime} 17\buildrel{\prime\prime}\over{.} 04$		23.33	22.62		14.76 ± 0.13	14.10 ± 0.11		−16.75 ± 0.20	−17.41 ± 0.18	0.66 ± 0.17
F15D384	03h34m3082	$-34^\circ 17^{\prime} 50\buildrel{\prime\prime}\over{.} 64$		23.26	22.63		15.26 ± 0.15	14.66 ± 0.13		−16.25 ± 0.21	−16.85 ± 0.20	0.61 ± 0.19
F15D417	03h32m4769	$-34^\circ 14^{\prime} 19\buildrel{\prime\prime}\over{.} 32$		22.20	21.55		14.70 ± 0.11	14.06 ± 0.09		−16.81 ± 0.18	−17.45 ± 0.18	0.64 ± 0.14
F17D227	03h31m828	$-36^\circ 17^{\prime} 24\buildrel{\prime\prime}\over{.} 36$		21.91	21.35		14.51 ± 0.09	13.97 ± 0.08		−17.00 ± 0.17	−17.54 ± 0.17	0.54 ± 0.12
F19D001	03h27m1802	$-34^\circ 31^{\prime} 35\buildrel{\prime\prime}\over{.} 40$		21.28	20.49		13.77 ± 0.10	13.02 ± 0.08		−17.74 ± 0.18	−18.49 ± 0.17	0.75 ± 0.13
F22D244	03h26m2503	$-37^\circ 07^{\prime} 40\buildrel{\prime\prime}\over{.} 08$		22.48	21.94		15.66 ± 0.12	15.15 ± 0.11		−15.85 ± 0.19	−16.36 ± 0.18	0.51 ± 0.16
F26D000	03h23m5441	$-37^\circ 30^{\prime} 36\buildrel{\prime\prime}\over{.} 00$		22.13	21.57		13.72 ± 0.11	13.16 ± 0.10		−17.79 ± 0.19	−18.35 ± 0.18	0.56 ± 0.15
F26D003	03h24m5837	$-37^\circ 00^{\prime} 34\buildrel{\prime\prime}\over{.} 56$		21.83	21.16		14.03 ± 0.12	13.39 ± 0.10		−17.48 ± 0.19	−18.12 ± 0.18	0.64 ± 0.15
F26D141	03h22m2270	$-37^\circ 23^{\prime} 51\buildrel{\prime\prime}\over{.} 36$		23.04	22.41		15.14 ± 0.15	14.53 ± 0.13		−16.37 ± 0.21	−16.98 ± 0.20	0.61 ± 0.20
F31D196	03h29m4325	$-33^\circ 33^{\prime} 25\buildrel{\prime\prime}\over{.} 20$		23.70	23.06		14.46 ± 0.16	13.86 ± 0.14		−17.05 ± 0.22	−17.65 ± 0.20	0.60 ± 0.21

Note. Column (1): name of the dwarf galaxy. Columns (2) and (3): R.A. and decl. of the central coordinates in the International Celestial Reference System (ICRS), according to the FDSDC (Venhola et al. 2018). Columns (4) and (5): surface brightness at one effective radius in the g and r bands, respectively. Columns (6) and (7): total apparent magnitude in the g and r bands, respectively, according to the FDSDC. Columns (8) and (9): total absolute magnitude in the g and r bands, respectively, assuming a Fornax cluster distance modulus of 31.51 ± 0.15 mag (equivalent to a distance of 20.0 ± 1.4 Mpc; Blakeslee et al. 2009). Column (10): total g − r color. All photometric measurements are in the AB magnitude system and have been corrected for Galactic foreground extinction.

Download table as:  ASCIITypeset image

We also report in Table 1 the g- and r-band surface brightness derived from the isophote whose semimajor axis corresponds to the one effective semimajor axis in the r band according to the FDSDC, which is listed in Table 2. Specifically, we computed the effective surface brightness μ_e,g and μ_e,r from the ellipse isophotes fitted to the PSF-matched r-band image by keeping their center coordinates, ellipticity, and position angle free to vary with galactocentric distance. We then ran ellipse on the PSF-matched g-band image by imposing the geometry of the r-band isophotes in a no-fit photometry-only mode. This way, the galaxy geometry stays the same in both bands, making all photometric measurements in the g and r bands comparable on an isophote-by-isophote basis. From Table 1, we can see that the effective surface brightness of our sample has a median value of μ_e,g = 22.5 mag arcsec⁻² in the g band, and μ_e,r = 21.8 mag arcsec⁻² in the r band. Therefore, both the absolute magnitude and surface brightness distributions of our sample indicate that the galaxies are brighter in the r band than in the g band.

The structural properties of our dwarf ETG sample in the r band are provided in Table 2, where the ellipticity and position angle belong to the isophote whose semimajor axis corresponds to either the one or two effective semimajor axes. For simplicity, we will simply refer to the effective semimajor axis as the effective radius (R_e ) throughout this work. Our galaxy sample features an ample variety of sizes and shapes. The effective radius varies between 6''.7 and 24''.3 (0.6–2.4 kpc), with a median of R_e = 14''.6 (1.4 kpc). The axis ratio at one effective radius lies in the range between 0.29 and 0.94, with a median of b/a = 0.57. The galaxy sample shows a similar distribution at two effective radii. Furthermore, while for some galaxies the ellipticity and position angle stay approximately the same at one and two effective radii, there are also cases in which the galaxy geometry changes drastically with galactocentric radius. Consequently, the structural properties of the sample are quite heterogeneous, in contrast to its more homogeneous photometric properties.

In order to visualize our Fornax dwarf ETG sample, we present the PSF-matched r-band galaxy images and their unsharp mask images in Figure 1. A clean version of the galaxy images is shown, in which the interloping sources have been masked out. In some unsharp masks, wavy patterns appear on extended masked regions, but we note that these are only artifacts and not real features. The unsharp mask images are created by following the approach described in Section 3.4. For the size of the Gaussian smoothing kernel, we adopt a Gaussian standard deviation of size 4 × σ_PSF along its semimajor axis, derived from the FWHM of the r-band PSF. For its geometry, we adopt the ellipticity and position angle of the isophote at two effective radii in the galaxy, which are given in Table 2. We observe that the galaxy images have a predominantly smooth appearance, although in some cases it is already possible to discern some of their embedded substructure features. By removing the bright diffuse light of the galaxies, a rich variety of features is clearly revealed in their unsharp mask images. These include disk substructures, such as edge-on disks and rings; and clump substructures, such as irregular light overdensities and off-center nuclei. We note the possibility that the observed off-center nuclei could instead be nuclear star clusters, as those found in the central regions of bright dwarf galaxies (Johnston et al. 2020; Paudel & Yoon 2020; Fahrion et al. 2021).

Zoom In Zoom Out Reset image size

Finally, to understand the context of our selected dwarf ETGs, we provide a view into their color–magnitude and projected spatial distributions compared to the other FDSDC galaxies in Figure 2. Our sample is represented by red and blue points, based on the classification of their substructures into the disky and clumpy categories, respectively (according to criterion (2) in Section 3.5). Our sample is consistent with being part of the red sequence defined by the dwarf galaxies of the Fornax cluster, with the disky subsample having overall redder colors than the clumpy subsample. Additionally, the majority of the disky subsample lies inside the projected virial radius of the Fornax cluster, while the clumpy subsample tends to lie outside of it instead. Only the galaxies belonging to the clumpy subsample are found in the projected vicinity of the Fornax A subgroup. When considered together, these differences could be an indication that dwarf ETGs with disk and clump substructures may constitute two different galaxy subpopulations. Or, alternatively, it may imply that the disk and clump subsamples are currently in different stages of their evolution. We will continue to address how the properties of the disk and clump subsamples differ from one another in the upcoming Sections 4 and 5.

Zoom In Zoom Out Reset image size

In order to handle images in two bands, we first run the residual method on the PSF-matched r-band images of the sample closely following the steps described in Michea et al. (2021). This means that, for each galaxy, we obtain a series of r-band residual images, where each one corresponds to a particular configuration of the method's parameters. Next, to run the residual method on the g band, we make use of the geometry information of the galaxy obtained during the run in the r band. Specifically, for each parameter configuration set, we extract the radial shape and orientation of the diffuse component of the galaxy at the r-band stopping iteration. By using the inellip parameter of the IRAF ellipse task, this geometry is then imposed and kept fixed when running the method on the PSF-matched g-band images. The method is then allowed to iterate as usual, until a g-band stopping iteration is reached.

In summary, the residual method is first run freely on the r band, then the geometry of the r-band diffuse component is extracted, and finally the method is run with this constrained geometry on the g band. This way, the geometry of the diffuse component is kept fixed between the g and r bands, making any photometric measurements directly comparable. We note that a given parameter configuration set still has its own stopping iteration that is independent for each galaxy in each band. In the g band, the iterative aspect of the method requires between one to three iterations, with a median of two iterations. In contrast, in the r band, it requires between one to two iterations, with a median of one iteration.

With respect to the parameter configuration of the method, the same values are implemented in both the g- and r-band runs. The specific parameters that must be tuned to the properties of the data set correspond to the smoothing kernel size and the sampling step size.

On the one hand, the smoothing kernel size dictates the amount of smoothing that the galaxy image is subjected to. It should be tuned to match the average half-width of the substructure features of the galaxies. By inspecting the unsharp mask images of our dwarf ETG sample (see Figure 1), we identify the substructure features and measure that their average half-widths lie in the range of 1''.2–4''.7 (0.1–0.5 kpc), which expressed in terms of the PSF FWHM of the data transforms into a Gaussian standard deviation corresponding to 3–8 × σ_PSF. Consequently, we adopt a Gaussian smoothing kernel with a standard deviation equal to 3, 4, 5, 6, 7, and 8 × σ_PSF.

On the other hand, the sampling step size dictates the frequency at which the radial light profile of the galaxy is sampled. It should be tuned based on the image resolution and provide the ability to sample the lower S/N regions at the galaxy outskirts. Taking into consideration that the resolution of the FDS data corresponds to 0''.2 pixel⁻¹, we carry out some test runs, and determine that the optimal separation between the successive isophotes is given by a growth rate of the step size length in the range of 18%–23%. Therefore, we adopt sampling step sizes with a growth rate equal to 18%, 19%, 20%, 21%, 22%, and 23%.

Together, these parameter ranges create a 6 × 6 configuration grid. Consequently, the residual method is independently run a total of 36 times on the g- and r-band images of a galaxy, with each run adopting a particular smoothing-sampling parameter pair. As a result, small variations will be present throughout the multiple residual images being obtained for a single galaxy. When quantifying the contribution of the residual light to the total galaxy light, these variations then translate into the uncertainty range of the median measurement.

By adopting the aforementioned parameter configuration, we proceed to apply the residual method to our sample of Fornax dwarf ETGs to measure the light in their substructure features. In Figure 3, we show r-band residual images of the sample. For illustration purposes, we do not show g-band residual images, as the appearance of the substructure features is extremely similar (if not completely the same) in both bands. We observe that the rich variety of substructures that were originally revealed in the unsharp mask images (see Figure 1) have also been faithfully captured in the residual images. The appearance of the residuals supports our substructure classification into disk substructures and clump substructures described in criterion (2) of Section 3.5. Out of the 23 dwarf ETGs in our sample, we classify 9 (40%) as disky, and 14 (60%) as clumpy. The substructure classification of each galaxy is provided in column (2) of Table 3. We highlight that 10 dwarf ETGs of our sample are also in the work of Hamraz et al. (2019), in which Hubble Space Telescope Advanced Camera for Surveys data is used to study Fornax galaxies. These images have higher spatial resolution compared to the FDS images. Notably, their unsharp mask images (see Figure 8 of Hamraz et al. 2019) reveal the very similar substructure features as the ones we observe in our unsharp mask and residual images that are based on the shallower FDS data, thus serving as further confirmation of our performed classification.

Zoom In Zoom Out Reset image size

Table 3. Substructure Classifications and Residual Light Fractions of Our Fornax Dwarf ETG Sample

Dwarf Galaxy	Classification	Residual Light Fraction
		Within 1 Re			Within 2 Re
		g Band	r Band		g Band	r Band
(1)	(2)	(3)	(4)		(5)	(6)
F01D145	Clumpy	${0.065}_{-0.006}^{+0.005}$	${0.038}_{-0.005}^{+0.006}$		${0.110}_{-0.008}^{+0.007}$	${0.060}_{-0.006}^{+0.012}$
F02D000	Clumpy	${0.067}_{-0.005}^{+0.006}$	${0.039}_{-0.004}^{+0.005}$		${0.085}_{-0.007}^{+0.008}$	${0.046}_{-0.006}^{+0.006}$
F04D000	Clumpy	${0.086}_{-0.019}^{+0.004}$	${0.050}_{-0.003}^{+0.003}$		${0.087}_{-0.016}^{+0.005}$	${0.053}_{-0.005}^{+0.005}$
F04D001	Disky	${0.051}_{-0.009}^{+0.009}$	${0.050}_{-0.009}^{+0.009}$		${0.050}_{-0.007}^{+0.006}$	${0.051}_{-0.007}^{+0.006}$
F04D002	Disky	${0.043}_{-0.006}^{+0.005}$	${0.032}_{-0.006}^{+0.004}$		${0.063}_{-0.006}^{+0.011}$	${0.050}_{-0.006}^{+0.005}$
F04D053	Clumpy	${0.101}_{-0.006}^{+0.007}$	${0.077}_{-0.006}^{+0.006}$		${0.129}_{-0.009}^{+0.013}$	${0.095}_{-0.007}^{+0.009}$
F04D061	Clumpy	${0.078}_{-0.007}^{+0.007}$	${0.065}_{-0.006}^{+0.005}$		${0.109}_{-0.008}^{+0.008}$	${0.084}_{-0.007}^{+0.006}$
F05D000	Clumpy	${0.107}_{-0.015}^{+0.009}$	${0.087}_{-0.009}^{+0.005}$		${0.103}_{-0.015}^{+0.009}$	${0.092}_{-0.010}^{+0.008}$
F07D000	Disky	${0.055}_{-0.009}^{+0.009}$	${0.057}_{-0.010}^{+0.009}$		${0.052}_{-0.006}^{+0.007}$	${0.053}_{-0.007}^{+0.007}$
F09D255	Disky	${0.050}_{-0.011}^{+0.010}$	${0.033}_{-0.007}^{+0.008}$		${0.091}_{-0.010}^{+0.010}$	${0.063}_{-0.008}^{+0.010}$
F09D492	Clumpy	${0.084}_{-0.005}^{+0.007}$	${0.046}_{-0.004}^{+0.004}$		${0.099}_{-0.006}^{+0.008}$	${0.054}_{-0.004}^{+0.007}$
F10D189	Disky	${0.062}_{-0.006}^{+0.004}$	${0.037}_{-0.004}^{+0.005}$		${0.079}_{-0.004}^{+0.012}$	${0.045}_{-0.004}^{+0.011}$
F11D279	Disky	${0.051}_{-0.006}^{+0.005}$	${0.051}_{-0.006}^{+0.007}$		${0.053}_{-0.006}^{+0.006}$	${0.054}_{-0.006}^{+0.006}$
F14D144	Clumpy	${0.051}_{-0.007}^{+0.007}$	${0.037}_{-0.004}^{+0.004}$		${0.067}_{-0.010}^{+0.010}$	${0.045}_{-0.005}^{+0.006}$
F15D384	Disky	${0.057}_{-0.005}^{+0.006}$	${0.037}_{-0.003}^{+0.005}$		${0.083}_{-0.007}^{+0.007}$	${0.049}_{-0.004}^{+0.005}$
F15D417	Disky	${0.027}_{-0.009}^{+0.010}$	${0.029}_{-0.010}^{+0.011}$		${0.055}_{-0.007}^{+0.009}$	${0.052}_{-0.007}^{+0.008}$
F17D227	Clumpy	${0.086}_{-0.018}^{+0.022}$	${0.064}_{-0.008}^{+0.008}$		${0.085}_{-0.015}^{+0.018}$	${0.061}_{-0.007}^{+0.007}$
F19D001	Disky	${0.032}_{-0.003}^{+0.004}$	${0.034}_{-0.003}^{+0.004}$		${0.047}_{-0.005}^{+0.007}$	${0.048}_{-0.005}^{+0.006}$
F22D244	Clumpy	${0.100}_{-0.023}^{+0.004}$	${0.068}_{-0.003}^{+0.021}$		${0.101}_{-0.021}^{+0.006}$	${0.074}_{-0.004}^{+0.019}$
F26D000	Clumpy	${0.223}_{-0.010}^{+0.009}$	${0.208}_{-0.059}^{+0.005}$		${0.201}_{-0.008}^{+0.005}$	${0.169}_{-0.039}^{+0.008}$
F26D003	Clumpy	${0.106}_{-0.013}^{+0.017}$	${0.068}_{-0.004}^{+0.005}$		${0.102}_{-0.011}^{+0.015}$	${0.068}_{-0.006}^{+0.007}$
F26D141	Clumpy	${0.070}_{-0.006}^{+0.006}$	${0.040}_{-0.005}^{+0.006}$		${0.123}_{-0.008}^{+0.010}$	${0.053}_{-0.006}^{+0.008}$
F31D196	Clumpy	${0.072}_{-0.005}^{+0.005}$	${0.045}_{-0.004}^{+0.003}$		${0.091}_{-0.006}^{+0.006}$	${0.052}_{-0.006}^{+0.004}$

Note. Column (1): name of the dwarf galaxy. Column (2): classification of the substructures into the disky or clumpy categories. Columns (3) and (4): residual light fraction measurements within the one effective radius isophote in the g and r bands, respectively. Columns (5) and (6): residual light fraction measurements within the two effective radii isophote in the g and r bands, respectively.

Download table as:  ASCIITypeset image

We note that some of our galaxies with embedded edge-on disks present artifacts in the central region of their residual images. These artifacts manifest as an hourglass-shaped light overdensity that is perpendicular to the major axis of the disk, and can be either very prominent (e.g., F04D002, F15D417) or quite small (e.g., F04D001, F07D000). These fake features arise as a direct consequence of smoothing something that is circular in shape with a highly elliptical kernel, such as a prominent spherical bulge or nucleus. In the residual method, the smoothing kernel adopts the shape and orientation of the galaxy isophote at two effective radii, as the objective is to match the overall geometry of the diffuse light and not the geometry of the most central region. Consequently, this makes the presence of such an artifact unavoidable in some cases. To address it, we carefully identify the affected regions in the residual images, and flag the pixels as bad in the corresponding BPM images. This way, all measurements and analyses that we carry out consistently exclude these artificial features, whenever present.

To quantify the contribution of the substructure features to the total galaxy light, we compute the residual light fraction (RLF) within the isophotes at one and two effective radii of the galaxies. The RLF is defined as the ratio of the residual light (contained in the galaxy residual image) to the total light (contained in the galaxy original image), and it is always quantified in its respective band. The results obtained are presented in Table 3. In the g band, we find that the RLF ranges between 2.7% and 22.3% within one effective radius, with a median value of 6.7%, and between 4.7% and 20.0% within two effective radii, with a median value of 8.7%. Similarly, in the r band, it ranges between 2.9% and 20.8% within one effective radius, with a median value of 4.6%, and between 4.5% and 16.9% within two effective radii, with a median value of 5.3%. Overall, the RLF tends to be larger at two effective radii than at one effective radius, and larger in the g band than in the r band.

To visualize these results, and to establish if there are any differences between the disky and clumpy subsamples, we plot the g-band versus the r-band RLFs in Figure 4. It is evident that there is a strong correlation between the measurements: if the substructures have a high relative brightness in the g band (i.e., a large RLF), then they will also appear bright in the r band, and vice versa. To quantify this correlation, we compute the Pearson correlation coefficient r_P of the distributions and its associated p-value. We find that the distribution within one effective radius (left plot) has a Pearson correlation coefficient r_P = 0.94 and a p-value = 1 × 10⁻¹¹. Similarly, the distribution within two effective radii (right plot) has an r = 0.84 and a p-value = 5 × 10⁻⁷. Thus, the g- and r-band RLFs are strongly correlated within both one and two effective radii of the galaxies. However, we also note that, for a given r-band RLF, the g-band RLF tends to be comparatively larger (i.e., above the one-to-one relation, the gray line in Figure 4). This difference is accentuated when dividing the dwarf ETGs into the disky and clumpy subsamples. Overall, the disky subsample presents smaller RLFs in both bands, and also smaller differences between its g- and r-band measurements. In contrast, the clumpy subsample presents larger RLFs, with its g-band measurements being larger than the r-band measurements. This is a possible indication that the substructure features are comparatively bluer in the clumpy subsample than in the disky subsample.

Zoom In Zoom Out Reset image size

We would like to note that dwarf galaxy F26D000, located in the upper right regions of Figure 4, appears to be an outlier, as it presents a RLF that is significantly larger compared to the rest of the sample. However, our choice of including this galaxy in our sample is well justified: it fulfills the three selection criteria of Section 3.5, and it is not even the brightest nor the bluest among the sample galaxies. Based on Figures 1 and 3, F26D000 appears to have quite bright, numerous clumpy substructures, which logically translate into bright residuals, and thus into a large RLF.

In Figure 5, we plot the g- and r-band RLFs versus the g- and r-band total absolute magnitude of the galaxies, respectively. Two main things are noticeable. First, there appear to be no significant trends, which we corroborate by computing the Pearson correlation coefficient r_P of the distributions and its associated p-value. However, if we exclude the outlier galaxy F26D000, we find that there is one weak correlation: the g-band RLF measured within two effective radii slightly correlates with the total g-band absolute magnitude, with fainter galaxies having larger RLFs. This distribution has a Pearson correlation coefficient r_P = 0.55 with a p-value = 8 × 10⁻³. We note that, similarly, there was no significant trend between the RLF and the galaxy luminosity in the Virgo dwarf ETGs analyzed in "Brought to Light I" (Michea et al. 2021). The second noticeable thing is that the distributions have a larger scatter in the g band than in the r band. Overall, we recognize the same behavior that we observed in Figure 4: the clumpy subsample tends to have larger RLFs than the disky subsample in both the g and r bands, albeit the difference is more substantial in the g band.

Zoom In Zoom Out Reset image size

In Figure 6, we plot the g- and r-band RLFs versus the total g − r color of the galaxies. We can clearly appreciate how the RLF in both bands increases with bluer galaxy colors, although the correlation is stronger in the g band than in the r band. In the g band, the distributions have a Pearson correlation coefficient r_P = −0.51 (r = −0.60) and a p-value = 1 × 10⁻² (p-value = 2 × 10⁻³) for the measurements at one and two effective radii, respectively. The correlations are less significant in the r band, with a Pearson correlation coefficient r_P = −0.37 (r = −0.45) and a p-value = 8 × 10⁻² (p-value = 3 × 10⁻²) for the measurements at one and two effective radii, respectively. The main result, however, is how the disky and clumpy subsamples become separated by the g − r color of the galaxies. The disky subsample is comparatively redder (g − r ≥ 0.6 mag), being relegated to the right half of the plots; while the clumpy subsample is comparatively bluer (g − r ≤ 0.65 mag), and is thus relegated to the left half of the plots. In the upcoming Section 5, we provide a more in-depth view into these color differences, as we compute and analyze the g − r colors of the diffuse and substructure components of the galaxies separately.

Zoom In Zoom Out Reset image size

As a way to compare our sample of dwarf ETGs with substructures to the whole dwarf ETG population of the Fornax cluster, in Figure 7, we plot the brightness distribution of both populations. As a reminder, we have defined the Fornax dwarf ETG population as all of the galaxies classified as early types in the FDSDC plus all of the galaxies in our sample (see Section 3.5). As the main result, we find that the brighter a dwarf ETG is, the higher the probability of it having embedded substructure features, either disky or clumpy. Thus, as shown in the top panel, the overall fraction of dwarf ETGs with substructures can be as high as 80% on the bright end (M_r ≈ −19 mag), dropping to 5% on the faint end (M_r ≈ −15 mag). If we only consider disk substructures, as shown in the middle panel, the fraction of dwarf ETGs with disk-like features can reach up to 40% on the bright end, dropping to 5% already by M_r ≈ −16 mag. We provide a comparison with the disk fraction of dwarf ETGs in the Virgo cluster in the discussion in Section 6. If now we only consider clump substructures, as shown in the bottom panel of Figure 7, the fraction of dwarf ETGs with clump-like features apparently follows a similar behavior. However, we caution against a physical interpretation of this result, as it is likely that our selection criteria (which require galaxies to have an early-type morphology) has removed a lot of the lower-luminosity dwarf galaxies where clumpy structures are likely even more common (e.g., normal dwarf irregulars), but where the clumps became too bright for the galaxies to be considered early types. Nonetheless, we highlight the fact that substructures, in either the form of disk features or clump features, are highly probable to be present in a dwarf ETG if the galaxy is bright enough (M_r ≤ −18 mag).

Zoom In Zoom Out Reset image size

As explained in Section 4.1, the residual method provides us with the g- and r-band images of the diffuse and substructure components for each galaxy in our sample. When running the method, we imposed the r-band galaxy geometry on the g band, thus making any photometric measurements directly comparable in both bands on a pixel-by-pixel (and isophote-by-isophote) basis. Consequently, for each galaxy, we are able to robustly compute the g − r color of its diffuse component out to three galactocentric radii by comparing the g- and r-band model images, and likewise the color of its substructure component by comparing the g- and r-band residual images.

We now proceed with a strong note of caution. While our method succeeds in isolating the excess light coming from substructure features in a residual image, the light contained in each pixel still consists of a mix of substructure and diffuse light. Therefore, there is no way to completely separate the photons from substructures and the photons from diffuse components in order to measure their true, uncontaminated color, without some kind of additional information that is not available in photometry alone, or without making artificial assumptions about the light profile shape where the substructure and diffuse light are combined. Indeed, in some of the radial light profiles (see Appendix Figure 12), the substructure color becomes slightly more blue at radii where the residual light becomes strongest. We cannot be certain if this is due to contamination from substructure features, or because of the complex color profiles that dwarf ETGs are known to have, even in the absence of substructures (Urich et al. 2017). Nevertheless, some color contamination could arise, especially if the substructures have a much bluer color than the diffuse component. In such a case, we predict that the true color of the substructure may be even bluer than the values we measure, and the diffuse component's color may be slightly redder. However, as the residual fraction is generally only a few percent, the diffuse component's color is unlikely to be significantly altered by the presence of the substructure. Also, in cases where the substructure and residual are observed to have a similar color, obviously we do not have to worry about color contamination.

In Table 4, we provide the integrated g − r colors of the diffuse and substructure components of our dwarf ETG sample. These color measurements are integrated within the isophotes at one and two effective radii of the galaxies, for which we derive a median color value and its associated uncertainties from the 16th and 84th percentiles of the distributions. We find that the color of the diffuse component lies in the range between 0.47 and 0.82 mag within one effective radius, with a median g − r = 0.67 mag; and in the range between 0.53 and 0.80 mag within two effective radii, with a median g − r = 0.68 mag. In contrast, the substructure component covers a wider color range and is comparatively much bluer on average: its color lies in the range between −0.16 and 0.84 mag within one effective radius, with a median g − r = 0.23 mag; and in the range between −0.27 and 0.83 mag within two effective radii, with a median g − r = 0.26 mag. It would appear that, when considering the color of an individual component, the measurements within one and two effective radii are quite similar and thus comparable.

Table 4.  g − r Color of the Diffuse and Substructure Components of Our Fornax Dwarf ETG Sample

Dwarf Galaxy	g − r Color
	Diffuse Component			Substructure Component
	Within 1 Re	Within 2 Re		Within 1 Re	Within 2 Re
	(mag)	(mag)		(mag)	(mag)
(1)	(2)	(3)		(4)	(5)
F01D145	${0.639}_{-0.020}^{+0.002}$	${0.670}_{-0.031}^{+0.003}$		$-{0.105}_{-0.076}^{+0.362}$	$-{0.163}_{-0.038}^{+0.370}$
F02D000	${0.732}_{-0.002}^{+0.002}$	${0.734}_{-0.002}^{+0.003}$		${0.115}_{-0.029}^{+0.038}$	${0.032}_{-0.044}^{+0.041}$
F04D000	${0.701}_{-0.030}^{+0.001}$	${0.701}_{-0.031}^{+0.002}$		${0.029}_{-0.012}^{+0.405}$	${0.094}_{-0.030}^{+0.405}$
F04D001	${0.685}_{-0.002}^{+0.002}$	${0.675}_{-0.001}^{+0.001}$		${0.695}_{-0.040}^{+0.031}$	${0.688}_{-0.030}^{+0.018}$
F04D002	${0.682}_{-0.001}^{+0.018}$	${0.689}_{-0.001}^{+0.027}$		${0.491}_{-0.456}^{+0.025}$	${0.539}_{-0.428}^{+0.020}$
F04D053	${0.500}_{-0.002}^{+0.022}$	${0.552}_{-0.002}^{+0.034}$		${0.250}_{-0.236}^{+0.016}$	${0.263}_{-0.263}^{+0.013}$
F04D061	${0.595}_{-0.002}^{+0.002}$	${0.599}_{-0.001}^{+0.004}$		${0.400}_{-0.017}^{+0.020}$	${0.294}_{-0.013}^{+0.008}$
F05D000	${0.543}_{-0.028}^{+0.006}$	${0.560}_{-0.032}^{+0.004}$		${0.226}_{-0.032}^{+0.275}$	${0.373}_{-0.016}^{+0.314}$
F07D000	${0.734}_{-0.001}^{+0.001}$	${0.714}_{-0.001}^{+0.001}$		${0.760}_{-0.015}^{+0.013}$	${0.744}_{-0.012}^{+0.008}$
F09D255	${0.661}_{-0.022}^{+0.005}$	${0.672}_{-0.030}^{+0.005}$		${0.090}_{-0.205}^{+0.441}$	${0.090}_{-0.056}^{+0.382}$
F09D492	${0.559}_{-0.006}^{+0.007}$	${0.598}_{-0.004}^{+0.005}$		$-{0.164}_{-0.046}^{+0.079}$	$-{0.123}_{-0.046}^{+0.055}$
F10D189	${0.721}_{-0.002}^{+0.001}$	${0.720}_{-0.002}^{+0.002}$		${0.125}_{-0.029}^{+0.052}$	${0.031}_{-0.055}^{+0.084}$
F11D279	${0.815}_{-0.001}^{+0.001}$	${0.787}_{-0.001}^{+0.001}$		${0.832}_{-0.022}^{+0.017}$	${0.828}_{-0.019}^{+0.012}$
F14D144	${0.724}_{-0.010}^{+0.011}$	${0.728}_{-0.015}^{+0.015}$		${0.358}_{-0.222}^{+0.229}$	${0.288}_{-0.242}^{+0.252}$
F15D384	${0.666}_{-0.003}^{+0.003}$	${0.677}_{-0.003}^{+0.003}$		${0.142}_{-0.035}^{+0.017}$	${0.021}_{-0.017}^{+0.029}$
F15D417	${0.660}_{-0.001}^{+0.002}$	${0.646}_{-0.001}^{+0.001}$		${0.816}_{-0.052}^{+0.082}$	${0.642}_{-0.020}^{+0.025}$
F17D227	${0.473}_{-0.026}^{+0.012}$	${0.540}_{-0.029}^{+0.009}$		$-{0.028}_{-0.013}^{+0.361}$	$-{0.021}_{-0.015}^{+0.392}$
F19D001	${0.784}_{-0.001}^{+0.001}$	${0.796}_{-0.001}^{+0.001}$		${0.845}_{-0.033}^{+0.025}$	${0.806}_{-0.013}^{+0.013}$
F22D244	${0.486}_{-0.005}^{+0.029}$	${0.525}_{-0.005}^{+0.031}$		${0.311}_{-0.338}^{+0.016}$	${0.374}_{-0.345}^{+0.024}$
F26D000	${0.546}_{-0.007}^{+0.067}$	${0.588}_{-0.004}^{+0.054}$		${0.427}_{-0.327}^{+0.027}$	${0.369}_{-0.306}^{+0.029}$
F26D003	${0.709}_{-0.018}^{+0.021}$	${0.705}_{-0.019}^{+0.021}$		${0.179}_{-0.185}^{+0.191}$	${0.217}_{-0.210}^{+0.212}$
F26D141	${0.661}_{-0.003}^{+0.002}$	${0.712}_{-0.004}^{+0.003}$		${0.032}_{-0.067}^{+0.070}$	$-{0.273}_{-0.063}^{+0.059}$
F31D196	${0.667}_{-0.002}^{+0.002}$	${0.675}_{-0.002}^{+0.002}$		${0.118}_{-0.019}^{+0.014}$	${0.025}_{-0.045}^{+0.020}$

Note. Column (1): name of the dwarf galaxy. Columns (2) and (3): g − r color of the diffuse component measured within the isophotes at one and two effective radii, respectively. Columns (4) and (5): g − r color of the substructure component measured within the isophotes at one and two effective radii, respectively.

Download table as:  ASCIITypeset image

As a way to visualize these results, in Figure 8, we plot the integrated g − r color of the diffuse and substructure components versus the total r-band absolute magnitude of the galaxies. We can clearly appreciate that while the diffuse component tends to have more consistently red colors, the substructure component displays a richer diversity, with the majority adopting bluer colors than their diffuse counterparts. Overall, there are weak correlations between the component color and the galaxy brightness: the brighter galaxies tend to have redder components. For the diffuse component, we compute a Pearson correlation coefficient of r_P = −0.37 (r = −0.32) and an associated p-value = 8 × 10⁻² (p-value = 1 × 10⁻¹) for the measurements at one and two effective radii, respectively. In comparison, the correlations are slightly stronger for the substructure component, having an r = −0.57 and a p-value = 5 × 10⁻³ for both measurements at one and two effective radii. Additionally, we observe that both components in the clumpy subsample tend to have comparatively bluer colors than in the disky subsample. This color difference, however, is significantly more pronounced in the substructure component than in the diffuse component. In other words, clump substructures are on average bluer than disk substructures. If the color of the diffuse component were heavily altered by contamination from the substructures, we might expect that those galaxies with high residual fractions and blue substructures would deviate farther downwards from the sequence seen in the left-hand panel. We tested for this possibility but saw little evidence for a dependency on residual fraction and/or color difference, which supports the notion that the color of the diffuse component is not strongly altered by the presence of substructures.

Zoom In Zoom Out Reset image size

To provide a different perspective, in Figure 9, we plot the integrated g − r colors of the substructure component versus the diffuse component. The colors of the components appear to be weakly correlated. We compute a Pearson correlation coefficient r_P = 0.42 with an associated p-value = 4 × 10⁻² for the measurements within one effective radius, while no correlation of any significance is obtained for the measurements within two effective radii. In the upper panel, we can appreciate the striking difference between our two subsamples. On the one hand, the clumpy subsample has a substructure component that is always bluer than the diffuse component (i.e., below the one-to-one relation shown in gray). The colors of both components also display a large scatter. On the other hand, the disky subsample is divided into two: five galaxies present a disk substructure as red as the diffuse component, while for four galaxies the disk substructure is notably bluer. We tested but saw no evidence for a dependence between the substructure color and clustercentric distance. When considering our dwarf ETG sample as a whole, it is clear that the substructure features are never redder than the diffuse light of a galaxy, a point that will be discussed further in Section 6. Next, in the middle panel, we color-code the measurements based on the total r-band absolute magnitude of the galaxies. Corroborating our findings of Figure 8, we can indeed observe that the brighter galaxies tend to present redder colors in both of their components, albeit with a large scatter. Finally, the results shown in the lower panel are addressed in the upcoming Section 5.3.

Zoom In Zoom Out Reset image size

To further characterize the diffuse and substructure components, we construct their g- and r-band surface brightness profiles and g − r color profiles. For the radial profiles to be comparable on both bands, we sample the model and residual images of a galaxy using a fixed set of isophotes. These isophotes follow the geometry of the diffuse component, in order to minimize any effect that the substructure component may have in driving the galaxy geometry. Specifically, we construct concentric elliptical annuli of a constant 5 pixels (1'') in width that follow the ellipticity and position angle of the isophotes of the r-band galaxy model image. These annuli are then used to sample the galaxy model and galaxy residual images in both the g and r bands, extract their light profiles, and construct their surface brightness and color profiles. As an example, we present these radial profiles for the case of the dwarf galaxy F02D000 in Figure 10, while corresponding figures for the rest of our dwarf ETG sample are provided in Appendix.

Zoom In Zoom Out Reset image size

In regard to the surface brightness profiles of our sample, we observe that the diffuse component is brighter than the substructure component at all galactocentric radii. From the galaxy center out to two effective radii, the diffuse component in the g band is on average between 2.0 and 3.3 mag brighter than the substructure component, with a median of 2.6 mag. Similarly, in the r band, it is on average between 2.2 and 3.5 mag brighter, with a median of 3.0 mag. The diffuse component tends to have a smooth profile that becomes steadily fainter with increasing radius, while in contrast the substructure component presents a more irregular profile with fluctuations, which declines less steeply with increasing radius.

To obtain a more quantitative assessment of these observations, we attempt to model the surface brightness profile of the diffuse component, but not of the substructure component. Due to the fluctuating nature of the latter, it would be very difficult to model it properly. Consequently, we choose to only run the 2D fitting algorithm GALFIT (Peng et al. 2002) on the galaxy model images in the r band, using the r-band PSFs constructed in Section 3.2. As a result, we find that the r-band diffuse components of our dwarf ETG sample appear to be described well by a single Sérsic profile (Sérsic 1968) of index n ranging between 0.60 and 2.28, with a median value of n = 1.33. Clearly, the diffuse components of our sample are more consistent with being described by an exponential disk profile (n = 1) than by a de Vaucouleurs profile (n = 4; de Vaucouleurs 1948). Furthermore, when separated into subsamples, we find that the Sérsic indices of the disky subsample range between 1.04 and 2.20, with a median of n = 1.61; while the Sérsic indices of the clumpy subsample range between 0.60 and 2.28, with a median of n = 1.16. Therefore, the diffuse component of the disky subsample presents on average larger Sérsic indices than the clumpy subsample. In other words, the disky subsample has a comparatively steeper inner profile with more extended wings on the outer profile. These results are consistent with the work of Su et al. (2021), who perform multicomponent decompositions of the FDS galaxies with GALFIT. Interestingly, for our dwarf ETG sample, the Sérsic indices they obtain by modeling the r-band galaxy images are very similar to the ones we obtain by modeling our r-band model images. This is likely because the residual fractions are low enough not to significantly change the overall light distribution.

Finally, similarly to the surface brightness profile, the color profile of the diffuse component tends to be smooth, while the color profile of the substructure component usually presents strong fluctuations. For the majority of the galaxies in our sample, their substructure component is bluer than their diffuse component at all galactocentric radii, although there are also cases in which both components have approximately the same radial color. In conclusion, for our sample, their substructure component is always either bluer than or the same color as their diffuse component at all radii within the two effective radii isophote.

To go one step further, we use the g − r color information we have extracted to shed some light on the properties of the stellar populations of the diffuse and substructure components of our dwarf ETG sample. However, we proceed with caution, as a single color is usually not enough to break the age–metallicity degeneracy (Worthey 1994).

While taking this degeneracy into account, we make use of the stellar population synthesis models based on the Medium-Resolution Isaac Newton Telescope Library of Empirical Spectra ⁷ (MILES; Vazdekis et al. 2010), an empirical stellar spectra library. In particular, we use the Extended MILES models (E-MILES; Vazdekis et al. 2012, 2016), which cover stellar spectral energy distributions (SEDs) in a wider spectral range (168–5000 nm). From the E-MILES SEDs, photometric predictions are derived for a variety of photometric filters, encompassing a large range of ages and metallicities. As the only caveat, these photometric predictions assume simple stellar populations (SSPs), meaning that all of the stars are assumed to have formed instantaneously in an initial burst of star formation. This clearly is a simplified approach to a possibly more complex star formation history of a real galaxy, but nonetheless, we use SSPs as a first-order approximation to the problem.

Photometric predictions in the form of AB magnitudes are available in the SDSS g and r bands, allowing us to construct g − r color models that we can directly compare with our observational data. Following Venhola et al. (2019), we select SSPs with the following properties. For the initial mass function (IMF), we choose a Kroupa universal (Kroupa 2001) IMF, which is a multipart power-law function with a logarithmic slope of 1.30. For the construction of the evolutionary tracks of the SSPs, we use the Padova 2000 isochrones (Girardi et al. 2000). Then, we consider all available ages within the approximate age of the universe, from 0.063 to 14 Gyr, increasing by ∼12% between successive steps. Finally, low-mass galaxies (M/M_⊙ ≲ 10⁹) are expected to have subsolar metallicities (${\mathrm{log}}_{10}(Z/{Z}_{\odot })\approx -0.6;$ Gallazzi et al. 2005; Toloba et al. 2015; Sybilska et al. 2017). However, to provide a wider range as a reference, we consider subsolar to solar metallicities: ${\mathrm{log}}_{10}(Z/{Z}_{\odot })=-1.3$, −0.7, −0.4, and 0.0. We note that the observed colors are corrected for Galactic foreground extinction but not for the dust extinction intrinsic to the galaxies. Hence, the age estimates derived here are to be considered upper limits. This is particularly true for the age estimates of the substructures whose colors are possibly still contaminated by the dominant, redder diffuse components.

The above setup allows us to construct four g − r color models, each one with a specific metallicity, evolving from 0 to 14 Gyr. In Figure 11, we plot these evolutionary tracks and superimpose the median integrated g − r colors of the diffuse and substructure components (dashed and solid lines in red and blue, respectively) of the disky (left-hand panel) and clumpy (right-hand panel) galaxies in our sample, as well as the 16th and 84th percentiles of their color distributions (shown as shaded red and blue regions, respectively). If we compare the behavior of the evolutionary tracks, we observe that there are no significant differences in color in the first 1 Gyr. After that, the stellar populations that are metal-richer transition into redder colors at a faster rate, and are able to reach redder colors overall. The age–metallicity degeneracy is evident: a young, metal-rich stellar population can have the same g − r color as a comparatively older, metal-poorer stellar population. Despite this effect, we find that the g − r color of the galaxy components can still give us some insight on their stellar populations.

Zoom In Zoom Out Reset image size

On the one hand, the diffuse components of the disky and clumpy galaxies display a small color scatter, and share a similar median g − r color (0.69 and 0.63 mag, respectively) within 1σ. Such a color is consistent with either intermediate-age (2–4 Gyr) stellar populations at solar metallicity or older (3–14 Gyr) stellar populations at subsolar metallicity. On the other hand, the disk substructure components exhibit a large color scatter, since 40% are as blue as the clump substructures (refer to Figure 9). According to the evolutionary tracks in Figure 11, the red (g − r ≥ 0.6 mag) disks have similar stellar populations as the diffuse components, whereas the blue disks (median g − r ≈ 0.2 mag) are consistent with stellar populations younger than ∼1–2 Gyr. The latter also applies to the clump substructure components, whose median color is g − r = 0.16 mag.

Another way to acquire some knowledge on the stellar populations of galaxies is by considering their gas content. The presence of gas serves as an indicator of possible recent or ongoing star formation, and thus of the existence of a young stellar population. Therefore, we match our sample with the ALMA Fornax Cluster Survey (Zabel et al. 2019) targeting the CO(1–0) emission line, the ATCA H i Survey (Loni et al. 2021), and the SAMI-Fornax Dwarfs Survey (SAMI-FDS; Scott et al. 2020) for the detection of Hα emission. We note that, while Hα is a strong indicator of recent star formation, H i and CO may just sit there passively without leading to star formation. Conversely, in galaxies with only H i or CO, low-level star formation may be going on leading to the formation of a low number of massive stars, which may not be enough to significantly ionize the interstellar medium (ISM).

As summarized in Table 5, we find that some kind of ISM has been detected in 7 of our dwarf ETGs: there are 5 CO, 2 H i, and 3 Hα detections. Concerning the 5 CO detections, the morphology and kinematics of the molecular gas of the galaxies is classified as disturbed, as opposed to regular. In regard to the 2 H i detections, only one galaxy shows a disturbed H i morphology. Interestingly, all of the galaxies in our sample with either CO, H i, or Hα detections have been classified in this work as having clump substructures. This means that half of the the galaxies in the clumpy subsample (7/14 galaxies) have been confirmed to contain gas, while in contrast gas was detected in none of the galaxies in the disky subsample (0/9 galaxies).

These results can be visualized in the lower panel of Figure 9. We observe that the galaxies with gas detections have a substructure component that is always bluer than their diffuse component. Therefore, we can hypothesize that their diffuse component is mostly quenched; or, alternatively, their star formation rate is too low to make a difference in the observed g − r color. Given the comparatively bluer color of the substructure component, we would instead expect that the galaxies are forming stars in their clumpy substructure regions, although the resolution of the surveys is too low to confirm a possible spatial correlation. This would imply that the bluer colors of their substructures are caused by the presence of young stellar populations.

As part of the analysis of our Fornax dwarf ETG sample (see Section 4.2), we showed how the fraction of galaxies that display disk-like features is dependent on the galaxy brightness, steeply decreasing as the dwarf ETGs become fainter. As illustrated in the middle panel of Figure 7, we find that the disk fraction of Fornax dwarf ETGs can be as high as 40% on the bright galaxy end, dropping to a low 5% on the faint galaxy end, where we define the location of the faint end as three magnitudes below from the bright end. In comparison, while Lisker et al. (2006b) observe that the same overall behavior is followed by the disky dwarf ETGs of the Virgo cluster, they find a comparatively higher disk fraction. As shown in the top panel of Figure 12 of Lisker et al. (2006b), the disk fraction of the Virgo dwarf ETGs corresponds to about 60% on the bright end, dropping to 5% once reaching the faint end (again defined to be three magnitudes below). This indicates that the dwarf ETGs with disk substructures are in relative terms more numerous in Virgo compared to Fornax.

When comparing the results of Michea et al. (2021) for nine dwarf ETGs in Virgo with those obtained here for 23 dwarf ETGs in Fornax, we notice an interesting difference. In the Virgo sample, we detected numerous examples of spectacular spiral-like features, such as tightly wrapped grand-design spiral arms. Meanwhile, in the Fornax sample, we generally do not detect these dramatic spiral patterns. We do detect some edge-on disks that form the basis of our disky sample, and there are two or three cases that appear to contain bars. However, the lack of tightly wrapped spiral arm structures is unexpected, and might indicate some fundamental difference, either in the dwarf internal properties or in their external environment, between the Fornax and Virgo samples.

Another less physical reason for this discrepancy could be a selection bias, as the Virgo galaxies analyzed by Michea et al. (2021) were hand-picked based on the appearance of spiral features in their SDSS images. Nevertheless, several more galaxies like these ones have been reported to exist in the Virgo cluster (Lisker et al. 2006b). Another explanation could be an observational bias. Although not detailed here, we ran a series of tests on the spatial resolution and depth of the data, and found that the better resolution and shallower depth of the Fornax (FDS) images compared to the Virgo (SDSS) ones is not fundamentally hindering our ability to detect substructure features of any kind, including spiral-like features. We should thus be capable of detecting substructures of similar brightness and appearance in Fornax as the ones observed in our Virgo sample. We also believe it is rather unlikely that all of the disky Fornax dwarf ETGs with spiral arms are systematically projected edge-on with respect to our line of sight.

The differences between the disky dwarf ETG populations in Virgo and Fornax could therefore be physical, and possibly related to the different properties of these two clusters. As a matter of fact, Virgo is more massive (4 × 10¹⁴ M_⊙) than Fornax (6 × 10¹³ M_⊙, McLaughlin 1999; Drinkwater et al. 2001; Weinmann et al. 2011), has a denser and hotter intracluster medium (Jones et al. 1997; Schindler et al. 1999; Paolillo et al. 2002; Janz et al. 2021), and is dynamically younger with a larger fraction of star-forming galaxies (Ferguson 1989). This would make environmental effects such as galaxy-cluster harassment (Moore et al. 1996), ram pressure stripping (Gunn & Gott 1972), and thermal evaporation (Cowie & McKee 1977; Cowie & Songaila 1977) comparatively more efficient in triggering the disk substructure features in Virgo dwarf ETGs.

In the second paper of this series ("Brought to Light II," Smith et al. 2021), we simulated dwarf ETGs along different orbits in a Virgo-like cluster potential. We found that tidal triggering can reproduce the disk substructures observed in Virgo dwarf ETGs only when a galaxy contains a cold and thin disk embedded in a more massive and hot disk, and experiences very plunging pericenter passages (r_peri < 0.25 r_vir). Therefore, perhaps the dwarf ETGs in Virgo contain more cold stellar disk substructures, and/or their orbits could be more plunging than in Fornax. In a forthcoming paper of this series (R. Smith et al. 2022, in preparation), we will carry out a more in-depth analysis of substructures in Virgo dwarf ETGs using deep multiwavelength survey data (rather than the small sample of pointed observations that were used in Michea et al. 2021) for comparison with the Fornax dwarf ETG population.

While the disky dwarf ETG populations in Virgo and Fornax seem to differ in the morphology of their substructure features, they appear instead to be comparable in their RLF in the r band at fixed M_r range (compare our Figure 5 with Figure 5 of Michea et al. 2021). This would exclude a dependence of the stellar light and mass of the disk substructures on the properties of the host environment.

For what concerns our Fornax sample alone, we note that the disky dwarf ETGs tend to comprise a brighter magnitude range, and have redder integrated g − r colors. In comparison, the clumpy dwarf ETGs tend to comprise a wider magnitude range that can reach fainter magnitudes, and have bluer (less red) global g − r colors. Moreover, the disk substructures tend to contribute a smaller fraction of the total galaxy light (i.e., have smaller RLFs in both g and r bands) independently of the integrated g − r color of their host galaxy. In comparison, the clump substructures tend to contribute a larger fraction of the total galaxy light (i.e., have larger RLFs), which increases with bluer g − r colors of their host galaxy. We also note that some of these clumpy dwarf ETGs have been detected either in Hα, H i, or CO; hence they are relatively more gas-rich than their disky counterparts.

In all cases, we measure that substructures are either similar or bluer in color than their diffuse components. If we assume their color to indicate the ages of the stellar populations (considering all caveats discussed throughout Section 5), we might conclude that the substructures were either formed at the same time as the main disk (in the case of matching color) or subsequently added to an existing galaxy at some later stage (in the case of bluer substructures). The color of the clump substructures is consistent with them being regions of on-going or recent star formation in host galaxies that are likely on the way to quench their star formation activity. This picture is also supported by the detected presence of gas, either cold or ionized. We note that, as addressed in Section 5.1, all color measurements that we provide are not free from contamination, as the residual method does not allow us to fully disentangle the true underlying color of each component based on imaging alone. As such, we believe that our analysis would deeply benefit if supplemented with spectroscopic data, which would provide additional insight on the stellar populations (and also the dynamics) of the galaxies. Several studies have already shown how Integral Field Unit spectroscopy can be used to thoroughly dissect and extract the properties of galaxies of the Fornax cluster (Iodice et al. 2019; Johnston et al. 2020; Lara-López et al. 2022), including dwarf galaxies. This is something we would like to incorporate in our future work.

Interestingly, the clumpy dwarf ETGs in our sample are spatially located, in projection, at the Fornax virial radius or outside it (see Figure 2). Their spatial distribution could indicate that the clumpy dwarf ETGs were originally star-forming galaxies, such as dwarf irregulars. One possibility is that these dwarf irregulars are having their star formation suppressed, which might explain why clumps make up only a small fraction of their total light compared to the more dominant star-forming regions of normal dwarf irregulars. If so, the quenching process appears to have occurred beyond the cluster virial radius, perhaps because their low masses make them more susceptible to environmental effects. These may be examples of backsplash galaxies (Gill et al. 2004), objects that previously entered the cluster but are now located beyond the virial radius. Alternatively, their quenching process may have begun in substructures beyond the cluster (so-called "pre-processing"; Mihos 2004). Indeed, four of them appear to be associated with the infalling Fornax A subgroup.

Conversely, all of the disky dwarf ETGs except one are found at or within the cluster virial radius, and none of them are at the position of the Fornax A subgroup. As they are deeper inside the Fornax cluster, they could have been subjected to stronger environmental effects such as strangulation and ram pressure stripping, which could explain why a large fraction of them show red colors, as if their star formation has been quenched. However, we saw no evidence that the blue disky sample is spatially distributed in a different way from those dwarf ETGs with red colors. The progenitors of these galaxies could have been simple dwarf ETGs that later gathered a new gas disk, perhaps via a merger of cosmic web gas accretion. Alternatively, their progenitors may have been late-type dwarf spiral galaxies that already contained a thick and thin disk component since formation.

In summary, we hypothesize that the dwarf ETGs with clumpy substructures are likely a transition population, caught in the transformation from normal star-forming dwarf irregulars to nonstar-forming dwarf ETGs. The origin of the disky substructures is less clear, and they may have simply formed that way or later formed the substructures via gas accretion or a merger. Follow-up spectroscopy is needed in order to use their stellar properties, such as age and metallicity, and their star formation histories to better understand their evolutionary path(s). Nevertheless, although we do see the presence of thin disk substructures, and in rare cases bars, we notice an absence of the tightly wrapped spiral features that were seen in our previous study of Virgo dwarf ETGs (Michea et al. 2021). We will investigate this apparent difference in an upcoming study.