THE BRIGHTEST CLUSTER GALAXY IN A85: THE LARGEST CORE KNOWN SO FAR

We have been aware of the enormous core of Holm 15A (Holmberg 1937), the brightest cluster galaxy (BCG) in A85, for over 30 yr. Hoessel (1980) fitted a modified Hubble law to its surface-brightness profile, and reported a core radius (i.e., the radius where the surface brightness reaches half of its central value) r_c = 572, which corresponds to a physical size r_c = 6.14 kpc at A85's restframe (z_clus = 0.05529 ± 0.00024; combined redshift data taken from the literature and our analysis of data from the Sloan Digital Sky Survey, SDSS DR9; Ahn et al. 2012). Extended H_α+[N ii] and [O iii] λ5007 emission was reported by Hu et al. (1985) and Fisher et al. (1995). Brown (1997), using LOCOS data (López-Cruz 1997), found that Holm 15A has a large core. Although Faber et al. (1997) had already suggested the presence of cores in galaxies brighter than M_V = −21 mag, establishing the core–cusp dichotomy, this was largely overlooked, in part, due to the lack of resolution that had hampered most BCG studies and the additional complication of using different parameterizations. As a result, Holm 15A has remained a curiosity.

Before the distribution of cores was established (e.g., Laine et al. 2003; Lauer et al. 2007), Faber et al. (1997) had advanced the idea that large cores were a manifestation of supermassive black hole (SMBH) binaries. Although, we are still unclear about details such as black hole (BH) merging times, hardening, and the so-called final parsec problem (e.g., Khan et al. 2013), the view that cores are expanded by the "scouring" action of SMBH over galactic cusps has gained wide acceptance (e.g., Merritt 2006; Kormendy & Ho 2013). Nevertheless, there are alternative scenarios, for example, scalar-field dark matter (e.g., Robles & Matos 2012), in which cores are formed without secondary mechanisms.

Añorve (2012) conducted a comprehensive study of the structure of galaxies using LOCOS clusters at z ⩽ 0.08, using the Driver for GALFIT on Cluster Galaxies (DGCG; Añorve 2012), a Perl script for GALFIT (Peng et al. 2010) that accounts for the effects of crowding and point-spread function (PSF) variations. DGCG allows the two-dimensional (2D) modeling (bulge+disk) of the surface-brightness distribution for cluster galaxies. Holm 15A was one of the few galaxies for which our DGCG analysis did not converge. This prompted us to perform the more detailed study that we report in this Letter.

We use the work of Postman et al. (2012, hereafter PL12) and Lauer et al. (2007, and references therein) to place Holm 15A into the context of the overall BCG population. To allow direct comparisons with PL12, we adopted Ω_M = 0.3, Ω_Λ = 0.7, and H₀ = 70 km s⁻¹ Mpc⁻¹.

2.1. Optical Observations

We work from a LOCOS image of Holm 15A. This image was taken under good weather conditions with an average 167 seeing (pixel scale =068 pixel⁻¹). To supplement our results, we looked for higher-resolution images. We failed to find Hubble Space Telescope (HST) images. However, A85 is part of the ongoing Multi-Epoch Nearby Cluster Survey (MENeaCS; Sand et al. 2011), from which we selected an image with 120 s integration in the SDSS r' band taken with the Canada–France–Hawaii Telescope (CFHT) 3.5 m telescope and MegaCam (see Table 1).

Table 1. Holm 15A: Nuker Law Fits

μb	rb	rb	α	β	γ	rγ	e	P. A.	Data
[mag./arcsec2]	[arcsec]	[kpc]				[kpc]		[deg.]
21.78 ± 0.01	17.21 ± 0.04	18.48 ± 0.04	1.24 ± 0.01	3.33 ± 0.02	0.0 ± 0.0	4.57 ± 0.06	0.26	−33.33	a
22.32	19.09	20.50	1.22	3.62	0.0	4.57	0.24	−34.07	b

Notes. Centroid: $\alpha _{2000} = 00^{\mathrm{h}} 41^{\mathrm{m}}50\buildrel{\mathrm{s}}\over{.}467$, δ₂₀₀₀ = −09°18'1157. Columns 1: surface brightness, 2: break radius in arcsec, 3: break radius in kiloparsecs, 4: power index at r_b, 5: outer power index, 6: inner power index, 7: cusp radius in kiloparsecs, 8: ellipticity, 9: position angle in degrees, 10: data source. ^aLOCOS (López-Cruz 1997), Telescope: KPNO 0.9 m, CCD: T2KA; pixel scale: 068/pixel. filter: R (Kron–Cousins), exposure time: 900 s, seeing: 16 FWHM, FOV: 232 × 232. ^bMENeaCS (Sand et al. 2011), Telescope: CFHT 3.5 m, CCD: MegaCam; pixel scale: 0187/pixel, filter: r' (SDSS), exposure time: 120 s, seeing: 074 FWHM, FOV: 096 × 094.

Download table as:  ASCIITypeset image

The LOCOS image was reduced with IRAF according to standard reduction procedures (e.g., López-Cruz et al. 2004). MENeaCS data was processed using the Elixir pipeline (Sand et al. 2011). Optical spectra from SDSS DR9 were used to determine cluster membership, galaxy and cluster velocity dispersions, and line ratios.

2.1.1. Photometry

The Nuker law, introduced by Lauer et al. (1995), is a "broken" power law given by

$$\begin{equation} I(r)= 2^{(\frac{\beta -\gamma }{\alpha })} I_{b} \left(\frac{r_{b}}{r}\right)^{\gamma } \left[ 1 + \left(\frac{r}{r_{b}}\right)^{\alpha }\right]^{(\frac{\gamma - \beta }{\alpha })}, \end{equation} \tag{ 1 }$$

where γ is the power index of the inner cusp; β is that of the envelope; α is that at the "break radius" r_b, the radius of maximum curvature in (log I, log r) space; and I(r_b) = I_b is the intensity at the break radius. r_b has been used as a scale indicator, but the "cusp radius" r_γ ≡ r_b((1/2) − γ)/(β − (1/2))^1/α, the radius where (dlog I/dlog r) = −(1/2), correlates better with other galaxy properties (e.g., Carollo et al. 1997; Lauer et al. 2007).

GALFIT was used to fit Holm 015A with an elliptical generalization of Equation 1 including centroid, axis ratio, and position angle (i.e., nine free parameters in total), simultaneously with single-Sérsic fits to eight neighboring galaxies. The input parameters for those eight galaxies were taken from a DGCG run over the whole LOCOS image. The star SDSS J004159.70-091937.3 (m_R = 15.85 mag) 27 from the center of Holm 15A was selected to generate the local PSF for Holm 15A. GALFIT proceeds by convolving a local PSF with the model component on a pixel center and minimizes χ² against the galaxy of interest data (see Ravindranath et al. 2001; Peng et al. 2010 for details). The sky background was modeled without including gradients. In simulated data, it has been found that fitting neighboring bright objects simultaneously with the galaxy of interest gives more accurate results than simple pixel masking (e.g., Añorve 2012; Häussler et al. 2007). Since the background is the most important source of uncertainty in surface-brightness modeling, we carefully investigated how our model parameters depend on the background level using the combination of one to five Sérsic models to remove the light of Holm 15A with other objects masked. For each (fixed) value of the background, a new fit to (1) was generated. This allowed us to make a realistic assessment of the errors for each fitted parameter (see Huang et al. 2013). The results are given in Table 1. The large core size of Holm 15A, r_γ = 426 ± 006, and its low redshift allow accurate parameter estimates even under the modest seeing conditions of the LOCOS frame: the value of r_b is more than 10 times larger than the seeing FWHM. The original LOCOS image, the model, and the residual are shown in Figure 1.

Zoom In Zoom Out Reset image size

We confirmed our result using the MENeaCS image. The fitting strategy was the same as for the LOCOS, except that we found a slight background gradient and did not perform such a detailed error analysis. Despite differences in telescope size, CCD format, pixel resolution, seeing, waveband, etc., the resulting parameter set (Table 1, second row) agrees well with the LOCOS result. The agreement of the fitted values of r_γ and γ is noticeable. The inner slope of Holm 15A is flat (γ = 0, see Table1) but less luminous galaxies tend to have γ > 0.3 (e.g., Ravindranath et al. 2001, who applied GALFIT to a large sample of early-type galaxies; ETGs). We note that the value of the outer slope β found from either fit is the largest ever reported—compare the mean and maximum of $\bar{\beta }=1.4\pm 0.2$ and max (β) = 2.63 in Laine et al. (2003).

We used the IRAF ellipse package on the LOCOS image to derive a one-dimensional (1D) surface brightness profile of Holm 15A (blue points with errors in Figure 2). We worked on a deconvolved image (the PSF was generated using neighboring stars) generated using STSDAS Maximum Entropy Method (MEM) inversion, with errors calculated by applying ellipse to 50 simulated images, where the background was randomly varied ±1 σ, pixel by pixel. We compared deconvolved images using MEM and the Richarson–Lucy approach (e.g., PL12), and we found negligible differences. From this 1D profile, we can determine r_b and r_γ directly using their definitions (given above). These non-parametric (^np) estimates are $r_{b}^{{\rm np}}=17{\buildrel{\prime\prime}\over{.}} 4\pm 0.2\; (18.7\pm 0.2\, \mathrm{kpc)}$ and $r_{\gamma }^{{\rm np}}=4{\buildrel{\prime\prime}\over{.}} 27\pm 0{\buildrel{\prime\prime}\over{.}} 05\;(4.59\pm 0.05\,{\rm kpc})$, which agree to better than one pixel with the results in Table 1. To compare the results from 1D and 2D surface-brightness modeling, we use the parameters resulting after the GALFITNuker Law fit Holm 15A, whose parameters are given in Table 1 (see first row). The resulting profile (red continuous line in Figure 2) closely matches the blue dots, indicating excellent agreement between the 1D and 2D models. We conclude that the core scale is reliably measured by either a parametric or non-parametric approach.

Zoom In Zoom Out Reset image size

The total stellar light in Holm 15A was modeled by the sum of two Sérsic profiles. We obtained a total absolute magnitude $M_{V}^{\mathrm{total}}=-23.81\pm 0.10\; (m_V^{\mathrm {total}}=13.28)$ (see Donzelli et al. 2011), where we have assumed (V − R) = 0.61, as is typical for a giant early ETG. We attempted to fit a de Vaucouleurs profile (dVP) to the LOCOS image (using GALFIT) and the 1D profile. This function rises too steeply at small radii and gives large residuals inside r_b (Figure 2). We estimate the missing light in this fit by comparing the dVP integrated magnitude $M_{V}^{\mathrm{dVP}}=-24.9^{+0.3}_{-0.4}$, where the errors are estimated by comparing the 1D and 2D dVP fits to Holm 15A, with the value of $M_{V}^{\mathrm{total}}$ (an approach similar to that advocated by Kormendy & Ho 2013). Hence, the approximate luminosity missing at r < r_b relative to the dVP is $L_{V,{\rm def}} \simeq 5^{+3}_{-2} \times 10^{11}\,L_{V\odot }\,(M_{V\odot }=4.83)$. Missing light has been found to correlate with SMBH mass (e.g., Kormendy & Bender 2009; Kormendy & Ho 2013).

2.1.2. Spectroscopy

By applying Fourier quotient and cross-correlation methods and the synthesis code STARLIGHT (Cid Fernandes et al. 2005) on the SDSS DR9 spectrum of Holm 15A, we obtain velocity dispersions of 226 ± 50 km s⁻¹ and 305 ± 15 km s⁻¹. While SDSS DR9 reports a velocity dispersion of 322 ± 13 km s⁻¹, Fisher et al. (1995) reported 289 ± 31 km s⁻¹ and a slight velocity gradient. Based on these values, our best estimate for the velocity dispersion is σ_gal = 310 ± 15 km s⁻¹, or about 80 km s⁻¹ less than the velocity dispersion of A2261-BCG (σ = 387 ± 16 km s⁻¹, PL12).

SDSS DR9 reports coordinates $\alpha _{2000}=00\buildrel{\mathrm{h}}\over{.}41\buildrel{\mathrm{m}}\over{.}50\buildrel{\mathrm{s}}\over{.}46$, δ₂₀₀₀ = −09181134 for the 3 arcsec diameter fiber on Holm 15A. We used this to search for lines associated with active galactic nucleus (AGN) activity by correcting for extinction, removing stellar continuum using STARLIGHT (as in Torres-Papaqui et al. 2012), and fitting Gaussian line profiles. We found line ratios

$$\begin{eqnarray*} \log \left(\frac{[{\rm O\,\scriptsize{III}}]\,\lambda 5007}{{\rm H}_{\beta }}\right) &=& 0.013 \log \left(\frac{[{\rm N\,\scriptsize{II}}]\,\lambda 6584}{{\rm H}_{\alpha }}\right) = 0.302 \\ \log \left(\frac{[{\rm O\,\scriptsize{I}}]\, \lambda 6300}{{\rm H}_{\alpha }}\right) &=& -0.567 \log \left(\frac{[{\rm S\,\scriptsize{II}}]\,\lambda \lambda 6717,31}{{\rm H}_{\alpha }}\right)\\ &=& -0.014 \end{eqnarray*}$$

with mean errors of about ±0.08 (Coziol et al. 2011).

2.2. Infrared Data

2.3. Radio and X-Ray Observations

We have used optical images taken with different telescope configurations using parametric and non-parametric approaches, and we consistently find a large cusp radius for Holm 15A (r_γ = 4.57 ± 0.06 kpc). This value of the cusp radius makes Holm 15 about 1 kpc larger than A2261-BCG (r_γ = 3.2 ± 0.1 kpc), hitherto the largest-cored BCG (PL12). Holm 15A's cusp radius is about 18 times the mean cusp radius for cored BCG (based on data from Laine et al. 2003). Figure 3 shows that although Holm 15A is the largest cusp BCG, it nevertheless lies less than 1σ above the correlation between r_γ and luminosity (as does A2261-BCG). This might suggest that a common mechanism is responsible for the formation of cores over a wide range of scales.

Zoom In Zoom Out Reset image size

The line ratios (Section 2.1.2) suggest that Holm 15A is a LINER (e.g., Ho 2008; Torres-Papaqui et al. 2012). By comparison, A2261-BCG is devoid of emission lines (PL12). Low emission from molecular gas (Salomé & Combes 2003) and dust (Quillen et al. 2008) suggests that there is little star formation in Holm 15A. Indeed, the radio power of Holm 15A (Section 2.2), being larger than P_1.4 GHz = 10^22.75 W Hz⁻¹, cannot be explained by star formation alone (Morrison et al. 2003). Hence, we conclude that optical line emission and radio continuum are dominated by AGN activity. A2261-BCG is a radio AGN slightly more luminous than Holm 15A (P_1.4 GHz = 5 × 10²³ W Hz⁻¹, PL12)

Using known scaling relations, we formed the five SMBH mass estimates given in Table 2. The mass inferred from the M_• − σ relation is the lowest. This might be a consequence of the breakdown of the M_• − σ scaling law for σ ≳ 270 km s⁻¹ (e.g., Lauer et al. 2007; Kormendy & Ho 2013). Three scaling relations suggest BH masses M_• ≳ 10¹¹ M_☉, but these relations have large scatter and were derived with limited samples. Therefore, we conservatively suggest that Holm 15A hosts a SMBH with M_• ∼ 10¹⁰ M_☉.

A SMBH has a strong stellar-dynamical influence within radius r_f = (GM_•/σ²). If we adopt M_• = 10¹⁰ M_☉ and velocity dispersion σ ≈ 310 km s⁻¹, r_f ∼ 450 pc (042). Thus, only with the largest mass estimate from Table 2 can we interpret the cusp radius as due to the gravitational influence of a SMBH causing r_f ∼ r_γ.

Let us suppose that Holm 15A contains an SMBH binary with total mass $\mathrm{{\it M}_{\bullet _B}}$. We would then expect the separation between the SMBH components to be $a_{B} \sim ({ \mathrm{{\it G M}_{\bullet _B}}}/{2\sigma ^2}) \sim ({r_f}/{2})$ (Dotti et al. 2012). For a SMBH binary total mass $\mathrm{{\it M}_{\bullet _B}}=10^{10}\; \mathrm{{\it M}_\odot }$, this implies a_B ∼ 225 pc (021), which is close to the average cusp scale of BCGs but much less than the scale of Holm 15A.

The morphology of the radio emission in Holm 15A is not clear in current data since most of the flux density is in rather diffuse structure, which could be much distorted. If this is the case, then by analogy with RBS 797 (Gitti et al. 2013), a high-sensitivity, high-resolution map of the center of Holm 15A could test whether Holm 15A hosts a SMBH binary.

J004150 appears to be at far higher redshift than Holm 15A. However, if the light of Holm 15A was not properly modeled, the photometric redshift estimated for J004150 could be in error. Moreover, if J004150 is a quasar at z = 0.9, then extended X-ray emission from the hot interstellar medium is unlikely to have been detected in the current Chandra exposure. Instead, the apparent X-ray extension is more consistent with a low-z AGN. It is possible that J004150 is a third SMBH component associated with Holm 15A. This can be tested by optical spectroscopy of J004150.

We follow Kormendy & Ho (2013, Section 6.10) to estimate the mass of the SMBH associated with the cluster dark matter halo. A85 has velocity dispersion σ_cl = 752 ± 34 km s⁻¹ (from our own analysis and the literature) and hence should host a central M_• ∼ 1.5 × 10¹¹ M_☉ SMBH. Holm 15A is the central galaxy of A85 and three of the mass estimates in Table 2 are consistent with the expected cluster central SMBH mass. Such agreement is unexpected unless dark matter halos are scale-free and the SMBH–dark matter coevolution is independent from the effects of baryons (see Kormendy & Ho 2013). Otherwise, we should accept that ultramassive (M_• ⩾ 10¹⁰ M_☉) BHs in BCGs follow special scalings as suggested by Hlavacek-Larrondo et al. (2012), and that the masses of SMBHs cannot grow indefinitely (Natarajan & Treister 2009).

We have found that Holm 15A has the largest core known so far, with r_γ = 4.57 ± 0.06 kpc. A central AGN supports the presence of a central BH, which could be ultramassive. Very large cores (r_γ ≳ 1 kpc) seem to be rare (see Figure 3) and may represent a relatively brief phase in the evolution of BCGs, as calculated merging times for SMBH binaries appear to be relatively short (Khan et al. 2014). Other mechanisms might be at work because timescales for cusp regeneration seem too long (e.g., Merritt 2006). If SMBH growth is regulated by galaxy mergers (e.g., Merritt 2006; Booth & Schaye 2011), their final masses were set, perhaps, by initial conditions (e.g., Treister et al. 2013).

The physical condition of Holm 15A may represent one of the best laboratories for testing the SMBH "scouring" scenario for the creation of BCG cores. Follow-up observations that might test for the presence of a second, or third, SMBH component include improved radio mapping, optical spectroscopy of J004150, and stellar- and gas-dynamical mapping using the HST or a large ground-based telescope (e.g., Dalla Bontà et al. 2009; McConnell et al. 2012) to investigate velocities on scale 04. Since only a few BCGs have dynamically determined BH masses (e.g., Table 2 in Kormendy & Ho 2013), further studies on Holm 15A (and A2261-BCG) could provide crucial tests of the applicability of SMBH mass scaling laws, the core "scouring" scenario, and hence on the coevolution of BHs.

Funding for SDSS-III has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the NSF, and the U.S. Department of Energy Office of Science. We thank Chien Peng for comments and advice.