A Collection of New Dwarf Galaxies in NGC 5128's Western Halo

We detect dwarf candidates using a full-color RGB image constructed from our u'g'z' imaging with custom Python image processing scripts. This filter combination samples the full optical spectral energy distribution (SED) and is sensitive to old, metal-poor stellar populations expected of primordial dwarf galaxies, while the inclusion of the u'-band also serves to sample flux from younger stellar components arising from more recent star formation. Figure 1 shows the RGB image with orange dashed curves indicating NGC 5128-centric iso-radial contours. We visually searched this image for faint, extended sources displaying smooth morphologies with shallow surface-brightness (μ) profiles typical of low-luminosity dwarfs. Using this method, we unambiguously find 16 dwarf galaxy candidates, of which a single previously known dwarf is recovered (KK98a 189; Karachentseva & Karachentsev 1998) and the remaining 15 are reported here for the first time, noting that one dwarf (dw1317-4255) has recently been confirmed to be a true member of the Centaurus A group Crnojević et al. 2018). The set of zoom-in images in Figure 1 show the dwarf candidates and measure 2' on a side, corresponding to physical ∼2.2 × 2.2 kpc² regions at the distance of NGC 5128. We point out that by employing a "by eye" detection technique, we are unable to make a robust estimate of the completeness limit to our imaging in a straightforward manner and thus cannot make broad statements on the true total population of dwarfs in this region. We instead defer such an analysis to a future work detailing the overall properties of dwarfs detected throughout our survey region, noting that a lack of completeness estimate does not affect our conclusions.

Zoom In Zoom Out Reset image size

We use galfit (v3.0.5; Peng et al. 2010) to measure the structural and photometric properties of the dwarfs. We first create 365 × 365 (4 × 4 kpc²) image cutouts in all filters centered on each dwarf. We model the dwarfs using 1D Sérsic (Sérsic 1963) profiles defined as $I(r)={I}_{\mathrm{eff}}\exp \{-{b}_{n}[{(r/{r}_{\mathrm{eff}})}^{1/n}-1]\}$, described by effective radius r_eff, the intensity I_eff at r_eff, and parameters defining the model shape n and b_n that are linked such that half of the total model light is contained within r_eff.

For each band we use an iterative surface-brightness (μ) model-fitting approach (Muñoz et al. 2015; Eigenthaler et al. 2018; Ordenes-Briceño et al. 2018a) by starting with reasonable initial guesses for magnitudes and sizes, and run SE on each cutout with a low detection threshold (detect_thresh = 0.5) to construct a segmentation map to mask non-dwarf sources. We then run galfit on the masked image and allow all of the parameters—including the minor-to-major iso-photal axis ratio (b/a) and position angles (PAs)—to vary simultaneously. In most cases, models quickly converge to a solution, and in the cases that fail we fix the input magnitude and r_eff to obtain reasonable estimates on b/a and n, and then fix those parameters while freeing the magnitudes and sizes for refinement. Once an initial model solution has been derived, we then subtract it from the original image, use SE to derive an improved mask and generate a more refined dwarf model, and iterate this process until dwarf properties negligibly vary.

The results of this technique are shown in Figure 2 for a selection of dwarfs, where each block is titled by the dwarf identifiers. In each block, the original monochromatic imaging is shown in the upper rows with the corresponding models in the middle rows above the residual images. Close inspection of the residuals shows that the models estimate the μ profiles very well, with the iterative masking strategy effectively decoupling the dwarf signal even in the presence of foreground star contamination.

Zoom In Zoom Out Reset image size

We list the full suite of photometric measurements in Table 1 alongside absolute V-band magnitudes (M_V), r_eff, n, and stellar mass estimates (${{ \mathcal M }}_{\star }$; see Section 3.2). In all cases, the recovered photometric centroids for the dwarf candidates agree to <1'' across filters, providing consistent anchors upon which the remaining model parameters are based. Total apparent magnitudes fall in the range 22.31 ≳ m_u'/mag ≳ 18.82 and 19.97 ≳ m_z'/mag ≳ 16.40 in the bluest and reddest filters, corresponding to foreground extinction-corrected (Schlafly & Finkbeiner 2011) −7.42 ≳ M_V/mag ≳ −10.82, using the conversion of Jester et al. (2005). Overall we find filter-dependent shape parameters in the range 0.15 ≤ n ≤ 2.73, and 0.27 ≤ b/a ≤ 1.00. Averaging across filters for each dwarf gives shape parameter ranges of 0.37 ≤ n ≤ 2.14, and 0.47 ≤ b/a ≤ 0.89 with a sample average $\langle n\rangle \approx 0.94$ and $\langle b/a\rangle \approx 0.72$, consistent with typical structural parameters of other dwarf galaxies in the nearby universe (e.g., Eigenthaler et al. 2018).

Table 1.  Photometric Properties of the Dwarf Galaxy Candidates

ID	αJ2000	δJ2000	u'	g'	r'	i'	z'	MV	reff	n	$\mathrm{log}{{ \mathcal M }}_{\star }$
	(hh: mm: ss)	(dd: mm: ss)	(mag)	(mag)	(mag)	(mag)	(mag)	(mag)	('')		(M⊙)
dw1312-4246	13:12:10.18	−42:46:48.53	20.27 ± 0.11	18.76 ± 0.12	18.81 ± 0.17	17.76 ± 0.19	17.86 ± 0.13	−9.37	6.23 ± 0.70	1.32 ± 0.35	${5.89}_{-0.36}^{+0.34}$
dw1312-4244	13:12:10.93	−42:44:43.66	21.90 ± 0.26	20.26 ± 0.19	19.88 ± 0.12	19.18 ± 0.33	18.75 ± 0.47	−8.12	7.51 ± 1.32	0.68 ± 0.24	${5.35}_{-0.38}^{+0.35}$
dw1312-4218	13:12:22.48	−42:18:41.58	21.98 ± 0.18	21.23 ± 0.24	20.16 ± 0.13	19.58 ± 0.12	19.97 ± 0.15	−7.54	6.39 ± 0.66	0.68 ± 0.07	${5.17}_{-0.35}^{+0.34}$
dw1313-4246	13:12:42.87	−42:46:50.57	21.76 ± 0.11	20.32 ± 0.12	20.11 ± 0.13	19.02 ± 0.13	18.92 ± 0.11	−7.95	5.39 ± 0.65	1.32 ± 0.19	${5.43}_{-0.49}^{+0.35}$
dw1313-4211	13:13:34.28	−42:11:08.38	19.74 ± 0.09	18.49 ± 0.08	18.29 ± 0.07	17.13 ± 0.08	17.14 ± 0.09	−9.79	12.00 ± 0.96	1.07 ± 0.11	${6.15}_{-0.44}^{+0.35}$
dw1313-4214	13:13:36.40	−42:14:08.11	20.11 ± 0.12	18.78 ± 0.16	18.43 ± 0.15	17.31 ± 0.16	17.61 ± 0.17	−9.58	9.69 ± 1.19	1.06 ± 0.20	${6.03}_{-0.40}^{+0.35}$
dw1314-4204	13:14:08.17	−42:04:08.51	20.78 ± 0.05	19.18 ± 0.05	18.97 ± 0.05	18.06 ± 0.05	18.22 ± 0.04	−9.11	4.47 ± 0.30	0.91 ± 0.14	${5.78}_{-0.35}^{+0.35}$
dw1314-4230	13:14:21.93	−42:30:41.87	20.96 ± 0.20	19.35 ± 0.12	19.05 ± 0.16	18.07 ± 0.09	17.86 ± 0.11	−8.96	5.92 ± 0.72	0.89 ± 0.28	${5.89}_{-0.56}^{+0.35}$
dw1314-4142	13:14:44.82	−41:42:28.27	21.62 ± 0.09	20.30 ± 0.07	20.55 ± 0.04	19.44 ± 0.04	19.79 ± 0.08	−7.74	3.89 ± 0.25	1.01 ± 0.40	${5.23}_{-0.30}^{+0.35}$
dw1315-4232	13:15:02.98	−42:32:17.78	20.88 ± 0.21	19.58 ± 0.23	19.02 ± 0.14	18.02 ± 0.18	18.32 ± 0.21	−8.91	8.44 ± 1.78	1.50 ± 0.37	${5.76}_{-0.42}^{+0.35}$
dw1315-4309	13:15:33.97	−43:09:27.18	22.31 ± 0.12	21.17 ± 0.18	20.63 ± 0.18	19.51 ± 0.15	19.31 ± 0.17	−7.42	6.05 ± 0.76	0.46 ± 0.14	${5.27}_{-0.51}^{+0.36}$
dw1316-4224	13:16:42.27	−42:24:05.32	19.34 ± 0.22	17.77 ± 0.15	17.94 ± 0.25	16.86 ± 0.20	16.84 ± 0.14	−10.32	14.34 ± 2.01	2.14 ± 0.58	${6.30}_{-0.35}^{+0.34}$
dw1317-4255	13:17:48.49	−42:55:40.45	21.54 ± 0.22	19.70 ± 0.21	19.89 ± 0.17	18.61 ± 0.25	18.73 ± 0.30	−8.46	12.93 ± 1.11	0.37 ± 0.09	${5.55}_{-0.31}^{+0.35}$
dw1318-4233	13:18:05.59	−42:33:37.10	20.67 ± 0.57	19.47 ± 0.42	18.84 ± 0.46	18.41 ± 0.38	18.82 ± 0.34	−9.14	16.78 ± 3.95	0.37 ± 0.13	${5.61}_{-0.42}^{+0.32}$
dw1319-4203	13:19:21.26	−42:03:38.74	20.47 ± 0.05	19.12 ± 0.06	19.05 ± 0.08	18.11 ± 0.04	18.11 ± 0.07	−9.15	4.91 ± 0.40	0.52 ± 0.09	${5.82}_{-0.34}^{+0.34}$
KK98a 189	13:12:45.23	−41:49:55.23	18.82 ± 0.08	17.40 ± 0.04	17.35 ± 0.04	16.42 ± 0.03	16.40 ± 0.04	−10.82	14.94 ± 0.68	0.73 ± 0.05	${6.48}_{-0.33}^{+0.35}$

Note. Column (1) lists dwarf identifiers based on the on-sky coordinates listed in columns (2)–(3). Columns (4)–(8) show the total apparent magnitudes in each filter, followed by the foreground extinction-corrected absolute V-band luminosities, effective radii, average Sérsic index, and stellar mass estimates in columns (9), (10), (11), and (12), respectively.

Download table as:  ASCIITypeset image

While galfit is excellent for model fitting, it is deficient in error estimation. For this, we use the ensemble of all five filters to place uncertainty estimates on the overall sizes and magnitudes. Table 1 lists the mean of each set of filter-dependent model r_eff along with corresponding 1σ error budgets. To estimate errors on the total dwarf magnitudes we use the variance in the sets of modeled r_eff, n, and b/a to generate new models by fixing each at their maximized/minimized values, and leaving the magnitudes unconstrained. We use the resulting sets of six models to determine the set of (r_eff, n, b/a) that produces the brightest/faintest magnitudes and adopt the difference as the uncertainties listed in Table 1.

We compute ${{ \mathcal M }}_{\star }$ by fitting the extinction-corrected total dwarf magnitudes to SEDs predicted by simple stellar population (SSP) synthesis models (Bruzual & Charlot 2003). We allow SSP ages and metallicities in the ranges 1–15 Gyr and 0.0001–0.05 [Z/H]. We calculate the χ² goodness of fit for masses predicted between our observed photometric SEDs, and all SSP models. We assume an average ${{ \mathcal M }}_{\star }$, weighted by $\exp (-{\chi }^{2})$, which together with the allowed ranges in age/metallicity help to mitigate the age–metallicity degeneracy (Zhang et al. 2017). The resulting ${{ \mathcal M }}_{\star }$ estimates are listed in the last column of Table 1, and we find that they occupy a relatively narrow range of $5.17\lesssim \mathrm{log}{{ \mathcal M }}_{\star }/{M}_{\odot }\lesssim 6.48$, with a mean $\mathrm{log}\langle {{ \mathcal M }}_{\star }/{M}_{\odot }\rangle =5.73$. With sizes of 70 ≲ r_eff/pc ≲ 300, these objects are broadly consistent with properties of LG dSphs, in particular those that show the highest stellar surface densities, and have many analogs in the Local Volume (LV; Chiboucas et al. 2009; McConnachie 2012; Müller et al. 2015, 2017a, see also Figure 3).

Zoom In Zoom Out Reset image size

Figure 3 compares the modeled r_eff and total luminosities to other low-mass stellar systems in the NGC 5128-M83 complex (Crnojević et al. 2014, 2016; Müller et al. 2015, 2017a), and for additional context to various other dwarfs in the LV, including the M81, M101, M96, and Leo systems (Chiboucas et al. 2009; Merritt et al. 2014; Javanmardi et al. 2016; Bennet et al. 2017; Henkel et al. 2017; Müller et al. 2017b, 2018a; Park et al. 2017), and the Fornax galaxy cluster core (Muñoz et al. 2015; Eigenthaler et al. 2018). We also show low-mass satellites identified in the LG (McConnachie 2012; Bechtol et al. 2015; Smercina et al. 2017; Homma et al. 2018; Muñoz et al. 2018b). For consistency with the LG dwarf data, where possible we show absolute V-band magnitudes converted from g'- and r'-band magnitudes (Jester et al. 2005); however, for some of the dwarfs (Chiboucas et al. 2009; Crnojević et al. 2016; Eigenthaler et al. 2018) neither g' nor r' imaging is available, so we defer to those that are available.

The size–luminosity relation of the new candidates shows a similar slope as other low-luminosity dwarfs in the nearby universe. Our selection technique biases us toward higher μ_V such that the new candidates fall almost parallel to lines of iso-μ_V (gray dashed lines), a trend that is not replicated by LG systems. The faintest of the new dwarfs populate a parameter space mostly devoid of LG analogs, but with several representative systems found throughout the LV. We note that four candidates—namely T18-dw1312-4247, 1314-4204, 1314-4231, and 1319-4203—fall below the main r_eff–M_V relation with r_eff ≃ 100 kpc and M_V ≃ −9 mag. These objects have no known analogs and we thus consider the possibility of them being associated with background galaxies at a distance of ∼50 Mpc. Should this be the case, we indicate the size–luminosity parameter locations for the whole sample by dashed star symbols. We discuss this notion further in Section 4, where we find them more likely to be associated with NGC 5128.

Many of the new candidates coincide with the faintest dwarf galaxies known in the Fornax galaxy cluster at M_V ≃ −8 mag (Eigenthaler et al. 2018), and overlap with several dwarf galaxies elsewhere in the LV; however, a sharp truncation in the size–luminosity relation is seen beyond the faintest dwarf candidates, with a noticeable gap near −7.5 ≲ M_V/mag ≲ −5.5. We refrain from speculating on the dearth of LG dwarfs in this region, but note that such compact dwarfs beyond the LG can exhibit similar morphologies to background elliptical galaxies in monochromatic imaging, and only reveal themselves via RGB imaging with sufficiently wide SED coverage. Given this and the overall difficulty identifying such objects beyond the Local Group, it is likely that this region will continue to be filled out as future imaging campaigns probe ever deeper into various giant galaxy environments.

Figure 4 shows a comparison of the new dwarf candidates to those in the Fornax galaxy cluster with u', g', and i' imaging (Muñoz et al. 2015; Eigenthaler et al. 2018), with the same symbol/color definitions as in Figure 3. We also plot nuclear star clusters (NSCs) corresponding to the nucleated dwarf galaxies present in the Fornax sample (orange; Ordenes-Briceño et al. 2018b), and compact stellar systems (CSSs) including GCs and ultra-compact dwarf galaxies (UCDs) also confirmed to be members of the Fornax cluster (brown; Wittmann et al. 2016). Finally, we also show for comparison a large sample of NGC 5128 GCs (red points; Taylor et al. 2017)

Zoom In Zoom Out Reset image size

The left column shows (u' − i')₀, (u' − g')₀, and (g' − i')₀ versus M_g' color–magnitude diagrams (CMDs) for the various samples. The trend toward bluer colors at fainter luminosities is apparent for the Fornax dwarfs in all three panels, most prominently for the (u' − i')₀ CMD, and is typically considered to represent a mass–metallicity relation. The new NGC 5128 dwarf candidates are offset from the Fornax dwarf sample toward redder average colors by Δ(u' − i')₀ ≈ 0.6 and Δ(g' − i')₀ ≈ 0.4 mag, and appear more consistent—in terms of stellar population properties—with the nuclei and CSS populations, despite having much more diffuse morphologies.

The right-hand column in Figure 4 shows a comparison of the sample color distributions, indicated by the shaded histograms aligned to the y-axes, and corresponding Epanechnikov-kernel density estimates (thicker solid lines). We find that while there is some overlap with the generally older and metal-poor Fornax dwarf galaxies, the bulk of the NGC 5128 dwarfs have colors consistent with the secondary red peaks shown by the CSS samples, which shows up most prominently in the (u' − i')₀ color. If the redder colors exhibited by the NGC 5128 dwarfs are due to a metallicity effect, then this might imply early formation within the halos of the giant galaxy progenitors of NGC 5128 itself, where they could incorporate material rapidly enriched by the giants at early times. Alternatively, higher metallicities could be due to either a prolonged primordial star formation history enabling self-enrichment, or a non-primordial burst of star formation may have occurred, possibly by previously enriched gas shocked upon infall into NGC 5128's halo. The former scenario might be expected from the density-morphology relation (Dressler 1980), where early star formation failed to be suppressed owing to the relatively low-density environment of the Centaurus A group precluding a high number of harassing encounters. In either case, it would require that these dwarfs must have at one point been embedded in dark matter halos of sufficient mass to retain enriched material expelled by early supernovae, which was likely subsequently stripped during infall, thus preventing very recent star formation that would give rise to bluer colors.