A Massive, Clumpy Molecular Gas Distribution and Displaced AGN in Zw 3146
A. N. Vantyghem, B. R. McNamara, C. P. O'Dea, S. A. Baum, F. Combes, A. C. Edge, A. C. Fabian, M. McDonald, P. E. J. Nulsen, H. R. Russell, P. Salome
FF EBRUARY
2, 2021
Preprint typeset using L A TEX style emulateapj v. 12/16/11
A MASSIVE, CLUMPY MOLECULAR GAS DISTRIBUTION AND DISPLACED AGN IN ZW 3146
A. N. V
ANTYGHEM , B. R. M C N AMARA , , , C. P. O’D EA , S. A. B AUM , F. C OMBES , , A. C. E DGE , A. C. F ABIAN , M.M C D ONALD , P. E. J. N ULSEN , , H. R. R USSELL , AND
P. S
ALOMÉ University of Manitoba, Department of Physics and Astronomy, Winnipeg, MB R3T 2N2, Canada; [email protected] Department of Physics and Astronomy, University of Waterloo, Waterloo, ON N2L 3G1, Canada Waterloo Centre for Astrophysics, Waterloo, ON N2L 3G1, Canada Perimeter Institute for Theoretical Physics, Waterloo, Canada LERMA, Observatoire de Paris, PSL University, CNRS, Sorbonne University Paris, France Collège de France, 11 place Marcelin Berthelot, 75005 Paris Department of Physics, Durham University, Durham DH1 3LE Institute of Astronomy, Madingley Road, Cambridge CB3 0HA Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA ICRAR, University of Western Australia, 35 Stirling Hwy, Crawley, WA 6009, Australia
February 2, 2021
ABSTRACTWe present a recent ALMA observation of the CO(1-0) line emission in the central galaxy of the Zw 3146galaxy cluster ( z = 0 . Chandra ob-servations. The 5 × M (cid:12) supply of molecular gas, which is confined to the central 4 kpc, is marginallyresolved into three extensions that are reminiscent of the filaments observed in similar systems. No velocitystructure that would be indicative of ordered motion is observed. The three molecular extensions all trail X-raycavities, and are potentially formed from the condensation of intracluster gas lifted in the wakes of the risingbubbles. Many cycles of feedback would be require to account for the entire molecular gas reservoir. Themolecular gas and continuum source are mutually offset by 2.6 kpc, with no detected line emission coincidentwith the continuum source. It is the molecular gas, not the continuum source, that lies at the gravitational centerof the brightest cluster galaxy. As the brightest cluster galaxy contains possible tidal features, the displacedcontinuum source may correspond to the nucleus of a merging galaxy. We also discuss the possibility that agravitational wave recoil following a black hole merger may account for the displacement. Keywords: galaxies: active — galaxies: clusters: individual (ZwCl 3146) — galaxies: ISM — galaxies:kinematics and dynamics INTRODUCTION
Galaxy clusters are permeated by a hot (10 − K), dif-fuse intracluster medium (ICM) that shines brightly in X-rays. In cool core galaxy clusters the central density is sharplypeaked at the position of the brightest cluster galaxy (BCG).The correspondingly high X-ray luminosity in the cluster coreshould result in the ICM condensing into clouds of cold,molecular gas and forming stars at rates of hundreds to thou-sands of solar masses per year (Fabian 1994; Peterson &Fabian 2006). Indeed, the BCGs in cool core clusters arethe largest and most luminous elliptical galaxies in the Uni-verse. They are burgeoning with star formation proceeding atrates of ∼ −
500 M (cid:12) yr − (e.g. McNamara 2004; O’Dea et al.2008; McDonald et al. 2011; Donahue et al. 2015; Tremblayet al. 2015; McDonald et al. 2018), host luminous emission-line nebulae (e.g. Lynds 1970; Heckman 1981; Cowie et al.1983; Hu et al. 1985; Crawford et al. 1999), and harbourmolecular gas reservoirs more massive than even gas-rich spi-ral galaxies, sometimes exceeding 10 M (cid:12) (Edge 2001; Edgeet al. 2002; Edge & Frayer 2003; Salomé & Combes 2003).These signatures of gas condensation are prevalent in systemswhose central cooling times fall below a sharp threshold of1 Gyr (Cavagnolo et al. 2008; Rafferty et al. 2008; Pulido et al.2018). Nevertheless, they account for less than 10% of whatwould be expected from an unimpeded cooling flow.Active galactic nuclei (AGN) feedback regulates the rate ofICM cooling (for reviews, see McNamara & Nulsen 2007, 2012; Fabian 2012). The “radio-mode” of AGN feedbackis nearly ubiquitous in giant ellipticals and galaxy clusterswith short central cooling times (Burns 1990; Dunn & Fabian2006; Best et al. 2007; Bîrzan et al. 2012). Radio jetslaunched by the central AGN inflate bubbles, drive shockfronts, and generate sound waves in the hot atmosphere (e.g.McNamara et al. 2000; Churazov et al. 2000, 2001; Blantonet al. 2001; Fabian et al. 2006). The sizes and surroundingpressures of radio bubbles, which manifest as cavities in theX-ray surface brightness, are used in a direct estimate of theAGN power output. In a large sample of objects, the AGNpower output closely matches the radiative losses from thesurrounding ICM (Bîrzan et al. 2004; Dunn & Fabian 2006;Rafferty et al. 2006; Nulsen et al. 2009). This, along with thepreponderance of systems with central cooling times < a r X i v : . [ a s t r o - ph . GA ] F e b V ANTYGHEM ET AL .& Soker 2005, 2010). This results in “chaotic cold accre-tion” (Gaspari et al. 2013), where the stochastic production ofcold clouds feeds black hole growth and regulates the feed-back loop between ICM cooling and AGN heating.Though molecular gas is the fuel for AGN feedback, its dis-tribution is also shaped by the AGN. Radio jets drive fast out-flows of ionized, neutral, and molecular gas on pc to kpcscales (e.g. Morganti et al. 2005, 2013; Nesvadba et al. 2006;Alatalo et al. 2011; Dasyra & Combes 2011; Sturm et al.2011; Tadhunter et al. 2014; Morganti et al. 2015). A gen-tler coupling between AGN feedback and molecular gas isobserved in BCGs (Olivares et al. 2019; Russell et al. 2019).The rotationally-supported disks that would be expected fromthe long-lived accumulation of molecular clouds in the galac-tic center are rare (Hamer et al. 2014; Russell et al. 2019).Instead, billions of solar masses of cold gas are situated inkpc-scale filaments that trail X-ray cavities (e.g. Salomé et al.2006, 2008; Lim et al. 2008, 2012; McDonald et al. 2012a;McNamara et al. 2014; Russell et al. 2014, 2016, 2017a,b;Vantyghem et al. 2016, 2018, 2019). The detection of red-shifted absorption lines indicates that some of this gas rainsback onto the central galaxy in a circulation flow (David et al.2014; Tremblay et al. 2016, 2018; Rose et al. 2019b,a, 2020).Motivated by these observations, McNamara et al. (2016)proposed the “stimulated feedback” paradigm (see also Re-vaz et al. 2008). As radio bubbles rise buoyantly, they upliftlow-entropy ICM from the cluster core to an altitude where itbecomes thermally unstable. The cold clouds decouple fromthe radio bubbles and fall back onto the central galaxy as inchaotic cold accretion. In this way, AGN feedback stimulatesthe production of its own fuel.Mapping the spatial distribution of cold gas in BCGs inall but the closest and most gas-rich systems has only be-come possible with the arrival of the Atacama Large Mil-limeter Array (ALMA). Here we present a new ALMA ob-servation targeting the CO(1-0) emission line in the BCG ofZw 3146, which is one of the most extreme cool core clus-ters and brightest CO emitters known. If left unimpeded, itssoft (0 . − . × erg s − (Ebel-ing et al. 1998; Böhringer et al. 2000) would lead to a massdeposition rate of 1250 M (cid:12) yr − (Edge et al. 1994). Indeed,the BCG in Zw 3146 is exceptionally active, showing a bluecontinuum and luminous nebular emission lines (Allen et al.1992; Hicks & Mushotzky 2005), an infrared luminosity of4 × L (cid:12) (Egami et al. 2006b), an 8 × M (cid:12) reservoirof cold molecular gas (Edge 2001; Edge & Frayer 2003), andstrong molecular H emission lines from ∼ M (cid:12) of warmH (Egami et al. 2006a). The infrared luminosity implies astar formation rate of ∼
70 M (cid:12) yr − (Egami et al. 2006b; Edgeet al. 2010; Hicks et al. 2010; Hoffer et al. 2012).Spectroscopic measurements indicate that the true mass de-position rate is much lower than implied by the X-ray lumi-nosity (e.g. Egami et al. 2006b). Any continuous radiativecooling flow from the bulk temperature at 4 keV to 0 .
01 keVis less than 50 M (cid:12) yr − (Liu et al. in preparation, see alsoLiu et al. 2019). The rate of gas cooling to 0 . (cid:12) yr − , with less than 78 M (cid:12) yr − cooling to lower tem-peratures. AGN feedback is likely responsible for the heav-ily suppressed cooling rate. The X-ray atmosphere hostsseveral cavities that replenish energetic losses at a rate of ∼ × erg s − (Rafferty et al. 2006). The combination ofa massive reservoir of molecular gas and powerful AGN feed-back make Zw 3146 a prime candidate to contain molecular filaments trailing X-ray cavities, as observed in other BCGs.Throughout this work we assume a standard Λ -CDM cos-mology with H = 70 km s − Mpc − , Ω m , = 0 .
3, and Ω Λ , =0 .
7. At the redshift of Zw 3146 ( z = 0 . (cid:48)(cid:48) = 4 .
35 kpc and the luminosity distance is1500 Mpc. OBSERVATIONS AND DATA REDUCTION
The BCG of the ZwCl 3146 galaxy cluster (2MASXJ10233960+0411116 – R.A.: 10:23:39.6340, decl.:+4:11:10.660) was observed with ALMA in Band 3 (Cycle 4,ID 2018.1.01056.S, PI Vantyghem), centered on the CO(1-0)line at 89 .
316 GHz. The observations were conducted inthree blocks from 15-19 August 2019, with a total on-sourceintegration time of 106 minutes. Each observing block wassplit into ∼ (cid:48)(cid:48) . A single spectral window in the frequency divisioncorrelator mode was used to study the CO spectral line with1.875 GHz bandwidth and 488 kHz (1 . − ) frequencyresolution, though the data were later smoothed to a coarservelocity grid. An additional three basebands with the timedivision correlator mode, each with a 2 GHz bandwidth andfrequency resolution of 15 .
625 MHz, were employed in orderto measure the continuum source.The observations were calibrated in
CASA version 5.4.0(McMullin et al. 2007) using the pipeline reduction scripts.Continuum-subtracted data cubes were created using
UVCON - TSUB and
CLEAN . Additional phase self-calibration was at-tempted, but the continuum source was too faint to provide animprovement in signal-to-noise. Images of the line emissionwere reconstructed using Briggs weighting with a robust pa-rameter of 0.5. The final CO(1-0) data cube had a synthesizedbeam of 0 . × .
26 arcsec (P.A. − . ◦ ) and was binned to avelocity resolution of 20 km s − . The RMS noise in line-freechannels was 2 . − . Systemic Velocity
The gas velocities presented in this work are measured inthe molecular gas rest frame. This was determined by extract-ing a spectrum from a 3 (cid:48)(cid:48) × (cid:48)(cid:48) region that encompasses allof the observed line emission, as shown in Fig. 1. A single-gaussian fitted to the spectrum yields a molecular gas redshiftof 0 . ± . . ± . ±
30 km s − from this nebular redshift.The discrepancy may originate from differences in the distri-bution of gas traced by each observation. The molecular gasis confined to the central few kpc of the BCG, while nebularemission is present on larger scales. RESULTS
Molecular Gas Mass
The integrated flux ( S CO ∆ v ) of the CO(1-0) line can be con-verted to molecular gas mass through (Solomon et al. 1987; OLECULAR G AS IN Z W Table 1
Parameters of Molecular FeaturesRegion χ / dof Velocity Center FWHM Integrated Flux † Gas Mass † ( km s − ) ( km s − ) ( Jy km s − ) (10 M (cid:12) )All Emission 221.2/197 0 . ±
10 353 ±
24 2 . ± .
17 504 ± − . ± . ±
19 0 . ± .
013 36 . ± . . ± . ±
14 0 . ± .
02 52 . ± . ±
12 288 ±
28 0 . ± .
036 78 . ± . − ±
17 415 ±
41 0 . ± .
04 83 . ± .
800 600 400 200 0 200 400 600 800
Velocity [km/s] F l u x D e n s i t y [ m J y ] Centre: -0.011 ± 10 km/sFWHM: 353 ± 24 km/s
Figure 1.
CO(1-0) spectrum extracted from a 13 ×
13 kpc (3 (cid:48)(cid:48) × (cid:48)(cid:48) ) box thatcontains all of the observed emission. Solomon & Vanden Bout 2005; Bolatto et al. 2013) M mol = 1 . × X CO X CO , gal (cid:18) S CO ∆ v D L + z (cid:19) M (cid:12) , (1)where z is the redshift of the source, D L is the luminosity dis-tance in Mpc, and S CO ∆ v is in Jy km s − . The CO-to-H con-version factor, X CO , has been studied extensively in the localUniverse. Its nominal value is the mean empirically-derivedGalactic value, X CO , gal = 2 × cm − ( K km s − ) − (Bolattoet al. 2013).Lacking an independent calibration of X CO in BCGs, wefollow standard practice by adopting the Galactic valuethroughout this work. The only estimate of X CO in a BCG wasobtained recently through the detection of CO(3-2) emis-sion in RXJ0821.0+0752 (Vantyghem et al. 2017). As COis an optically thin emission line, it provided a direct mea-surement of the CO column density. In combination witha measurement of CO(1-0) and CO(3-2) line intensities anda number of assumptions, X CO was estimated to be half ofthe Galactic value. This is reassuringly close to the Galacticvalue given the significant morphological differences betweenBCGs and the Milky Way disk. Until a direct calibration inBCGs is available, we continue to follow the standard practiceof adopting the Galactic X CO .The primary reasons for the CO-to-H conversion factor tovary between galaxies are the gas metal abundance and ex-citation conditions. In metal-poor dwarf galaxies X CO is ele-vated by more than an order of magnitude because CO tracesa smaller fraction of the overall H distribution (Bolatto et al. 2013). The molecular gas in BCGs likely forms by condensa-tion of the hot atmosphere, which has typical metallicities of ∼ . − . (cid:12) (e.g. De Grandi & Molendi 2001; De Grandiet al. 2004). The metal abundance of molecular clouds inBCGs is therefore expected to be similar to that of the MilkyWay. X CO is driven downward by factors of ∼ (cid:12) yr − in the most extreme cases (e.g. the PhoenixCluster; McDonald et al. 2012b). However, the molecular gasin BCGs is often distributed over much larger spatial scalesthan in ULIRGs, where the gas is confined to the central kpc.Despite their elevated SFRs, BCGs still lie on the Kennicutt-Schmidt relation (Kennicutt 1998; Kennicutt & Evans 2012)alongside normal galaxies (see e.g. McNamara et al. 2014;Russell et al. 2014; Vantyghem et al. 2018).The SFR in Zw 3146, 70 M (cid:12) yr − (Egami et al. 2006b), iscomparable to other BCGs for which the Galactic X CO hasbeen adopted. However, Zw 3146 is also the most luminousand distant H emitter observed, with a line luminosity of L [H − S (3)] = 6 . × erg s − (Egami et al. 2006b).A cosmic-ray dominated region or excitation through shocksdriven by the AGN may be required to explain such strong ro-tational H lines (Ferland et al. 2009; Bayet et al. 2010; Guil-lard et al. 2012). This could impact the value of X CO . StrongH line emission is observed in many cool core clusters (Don-ahue et al. 2011) and radio galaxies (Ogle et al. 2010).A spatially-integrated CO(1-0) spectrum was extractedfrom a 13 ×
13 kpc (3 (cid:48)(cid:48) × (cid:48)(cid:48) ) box containing all of the detectedline emission. The spectrum, shown in Fig. 1, is best-fit bya single Gaussian with total flux 2 . ± .
17 Jy km s − . Thiscorresponds to 5 . ± . × M (cid:12) of molecular gas. Alist of all fitted parameters is provided in Table 1. No otherCO(1-0) line emission was detected throughout the 70 (cid:48)(cid:48) pri-mary beam of these observations.Previous single-dish and interferometric radio observationsof Zw 3146 with the IRAM-30m telescope and Owens ValleyRadio Telescope (OVRO) yielded CO(1-0) integrated fluxesof 5 . ± . − and 5 . ± . − , respectively(Edge 2001; Edge & Frayer 2003). Our ALMA observationrecovers 50% of this flux; a similar fraction to other BCGs(e.g. David et al. 2014; Russell et al. 2014; McNamara et al.2014; Vantyghem et al. 2018). The line emission in the OVROobservations, which recovers all of the single-dish flux, is un-resolved by the 6 . (cid:48)(cid:48) × . (cid:48)(cid:48) beam. This suggests that ourALMA observations are missing large-scale flux, predomi-nantly on scales of ∼ (cid:48)(cid:48) (22 kpc). V ANTYGHEM ET AL . Figure 2.
A multiwavelength view of the Zw 3146 cluster. The top-left panel shows the lightly-smoothed
Chandra
X-ray 0 . − HST
WFPC2 F606W image, Ly- α , and far-UV (Tremblay et al. 2015).The most prominent substructures in the HST F606W image have been identified. Note that the F606W filter includes a contribution from H β and [O III ] lineemission. The rectangle in each of the panels on the right indicate the field-of-view of the ALMA CO(1-0) image, shown on the bottom-left. The solid blackellipse is the synthesized beam. The blue contours and + indicate the CO(1-0) emission and its centroid, respectively. The red contours correspond to 4 σ steps inthe mm continuum, beginning at 3 σ and with σ = 7 . µ Jy beam − . OLECULAR G AS IN Z W Figure 3.
Maps of velocity centroid and line FWHM determined by thepixel-by-pixel fitting scheme discussed in Section 3.2. These images havethe same field-of-view as the CO(1-0) flux map in Fig. 2. The solid blackellipse is the synthesized beam.
Molecular Gas Distribution and Kinematics
Maps of integrated flux, velocity, and FWHM of the CO(1-0) line were created by fitting the spectra extracted from indi-vidual pixels within the data cube. Each spectrum was aver-aged over a box the size of the synthesized beam. We allowedfor the possibility of multiple coincident velocity structures byfitting each spectrum with up to two Gaussian components.The significance of each velocity component was tested us-ing a Monte Carlo analysis employing 10000 iterations, witha detection requiring 5 σ significance. The presence of onecomponent was required before attempting to fit a second.We found that a single Gaussian component was sufficient toaccurately model the spectrum for each pixel. Instrumentalbroadening has been incorporated into the model.The integrated flux map is presented in the lower left panelof Fig. 2, and the velocity centroid and full-width at half max-imum (FWHM) maps are presented in Fig. 3. The molecu-lar gas distribution is compact, with all of the detected gasconfined to a 1 (cid:48)(cid:48) (4 .
35 kpc) radius. It is marginally resolvedinto three extensions to the NE, SE, and W. These extensionsdo not exhibit any clear velocity gradients. The emission isat its most blueshifted ( −
150 km s − ) at the central peak andin the farthest extent of the SE extension. The NE and W Figure 4.
Maps of the integrated CO(1-0) flux after subtracting a pointsource centered on the peak of the gas distribution. The contours correspondto the original distribution of integrated flux, as shown in Fig. 3. extensions have more moderate velocities (0 to 50 km s − ),but the velocity transition to the central value occurs overthe span of about a beam, so it is likely the result of theresolution element smearing two distinct velocities together.The linewidths, which range from 100 to 400 km s − FWHM,are broader than those seen along filaments in other systems( <
100 km s − ) (e.g. Russell et al. 2016; Vantyghem et al.2016, 2018), consistent with a superposition of gas structuresat a range of velocities. Chandra
X-ray,
HST
WFPC2 F606W, Ly- α ( HST
ACS/SBC F140LP), and FUV (
HST
ACS/SBC F165LP) im-ages (O’Dea et al. 2010; Tremblay et al. 2015) are shownalongside the CO(1-0) total flux map in Fig. 2. The boxin the X-ray image corresponds to the fields-of-view for the
HST images, while the ALMA field-of-view is shown in the
HST images. The central galaxy contains bright knots of neb-ular emission and FUV flux from recent star formation, oneof which is coincident with the molecular gas. A brighter Ly- α and FUV peak is located 3 .
25 kpc ESE of the moleculargas peak. No molecular gas is present at this position; theCO emission truncates where the Ly- α and FUV fluxes be-gin to increase. Dust attenuation may account for this anti-correlation between the molecular gas and FUV emission.Features that may be indicative of a merger remnant are seenin the HST
F606W image. The most prominent clumps andfilament are identified in Fig. 2. However, this filter also in-cludes H β and [O III ] line emission, which may account forsome of the observed features.The molecular gas is offset by 10 kpc to the south of thepeak in X-ray emission, which itself does not coincide withstructure at any other wavelength. The bright X-ray emissionin the cluster core arcs around the position of the CO(1-0)peak. In addition, the X-ray atmosphere contains a series ofdepressions that likely correspond to cavities inflated by thecentral AGN (see Section 3.5).The molecular extensions can be better visualized by sub-tracting the contribution from an unresolved point source co-incident with the CO(1-0) peak. This is shown in Fig. 4. Firstwe created an image of the beam centered on the pixel withthe largest integrated flux. This image was further convolvedwith a beam-shaped tophat kernel, as the pixel-based spectralfitting used to create the maps extracts the emission from aregion of this shape. The image was then scaled to match the V
ANTYGHEM ET AL .
800 600 400 200 0 200 400 600 800
Velocity [km/s] F l u x D e n s i t y [ m J y ] Centre: 17.2 ± 6.1 km/sFWHM: 177 ± 14 km/s
800 600 400 200 0 200 400 600 800
Velocity [km/s] F l u x D e n s i t y [ m J y ] Centre: 28 ± 12 km/sFWHM: 288 ± 28 km/s
800 600 400 200 0 200 400 600 800
Velocity [km/s] F l u x D e n s i t y [ m J y ] Centre: -69 ± 17 km/sFWHM: 415 ± 41 km/s
Figure 5.
CO(1-0) spectral to the three clumps identified in Fig. 4. flux of the brightest pixel and subtracted from the integratedflux map. The unresolved emission subtracted from the inte-grated flux map amounted to 22% of the total flux.The resulting map shows that the gas morphology is sep-arated into the three extensions, although none of them areresolved. They have been identified in Fig. 4. The corre-sponding spectra are shown in Fig. 5, and their spectral fitsare listed in Table 1.
Figure 6.
Radio spectral energy distribution, with the ALMA continuumsource measurement shown in blue. The best-fitting power law, indicated bythe dashed line, has a slope of − . ± . Millimetre Continuum
An image of the continuum at 96 .
193 GHz (3 . (cid:48)(cid:48) on each side. A single, unresolved sourcewas identified (R.A.: 10:23:39.5948, decl.: +4:11:11.032),which had a total integrated flux density of 133 ± µ Jy. Thecontours for this source are shown in red Fig. 2. These con-tours begin at 3 σ , where σ = 7 . µ Jy beam − , and increasein 4 σ steps. A two-dimensional gaussian fit to the continuumflux yielded an accuracy for the centroid of 0 .
006 arcseconds.We searched for, but did not detect, evidence for absorptionagainst the continuum source using the native velocity resolu-tion of the observations.In Fig. 6 we plot the ALMA continuum alongside flux mea-surements at lower frequencies, which were obtained fromeither the VLA or BIMA (Egami et al. 2006b; Coble et al.2007; Giacintucci et al. 2014). Error bars are included, butare smaller than the markers for all measurements exceptthose provided in this work. The spectral energy distribu-tion (SED) is best fit by a power law with spectral index α = 0 . ± . The Molecular Gas-AGN Offset
As is evident in Fig. 2, the molecular gas and continuumsource are mutually offset. The two peaks are 2 . . (cid:48)(cid:48) )apart, and only a minimal overlap is present in the fainterreaches of the two distributions. This offset cannot originatefrom a phase calibration issue, as the continuum and line mea-surements are obtained from the same observation.Independent observations of the radio continuum at lowerresolution confirm the position of the mm continuum sourceidentified here. In a VLA 4.9 GHz observation the continuumcentroid is 0 . (cid:48)(cid:48) SSW of the ALMA continuum source (O’Deaet al. 2010). Though the coarse resolution (1 . (cid:48)(cid:48) × . (cid:48)(cid:48) ) pre-vents the CO(1-0) centroid and continuum source from beingdefinitively distinguished, the VLA centroid is closer to theALMA continuum position. Note that the VLA coordinateswere reported in the B1950 equinox, so have been convertedto J2000. Giacintucci et al. (2014) also report a continuumcentroid corresponding to VLA 4.9 and 8.5 GH observations Following the convention S ν ∝ ν − α OLECULAR G AS IN Z W Figure 7.
DECaLS r-band image of Zw 3146 in the same field-of-view as theHST F606W image in Fig. 2. Contours of the ALMA CO(1-0) emission andcontinuum source are shown in blue and red, respectively. Te black contourscorrespond to the DECaLS r-band flux. The black rectangle indicates theALMA CO(1-0) field-of-view as shown in Fig. 2. in the C configuration. The position is consistent with thelocation of the ALMA continuum source, although no uncer-tainties are provided and the beam sizes are larger. Never-theless, these independent radio observations indicate that theALMA continuum position reported here is likely accurate.The stellar continuum is difficult to disentangle from nebu-lar emission in the HST images. In order to determine the po-sition of the galactic center, we obtained a Dark Energy Cam-era Legacy Survey (
DECaLS ) r -band image using the legacyskyviewer , which is shown in Fig. 7. This image is uncon-taminated by nebular line emission and stellar features from amerger remnant, providing a smooth tracer of the stellar con-tinuum. Contours of molecular line emission and continuumsource are overlaid in blue and red, respectively, in Fig 7,along with black contours tracing the stellar emission. It isthe molecular gas that is coincident with the galactic center,not the continuum emission from the AGN.All taken together, these results imply that the moleculargas resides at the galactic center while the continuum sourceis offset by 2 . X-ray Cavities
In order to investigate any possible connection between themolecular gas and X-ray cavities, we present a new search forcavities in Zw 3146 using the
Chandra data reduced by Hoganet al. (2017) and Pulido et al. (2018). The 0 . − Figure 8.
X-ray image processed using the GGM algorithm to enhance sur-face brightness edges, shown with the same field-of-view as the X-ray imagein Fig. 2. Six candidate X-ray cavities have been identified. The locationof the cluster center is indicated by the black "Y". The blue + indicates themolecular gas centroid. have processed this image using the gaussian gradient magni-tude algorithm (GGM; Sanders et al. 2016a,b) with a radially-dependent filtering kernel. This algorithm enhances edges insurface brightness. The resulting image is shown in Fig. 8,sharing a field-of-view with the X-ray image in Fig. 2. TheX-ray atmosphere contains a series of surface brightness de-pressions and edges corresponding to X-ray cavities and coldfronts, respectively (Forman et al. 2002; Rafferty et al. 2006).We manually identified six candidate X-ray cavities usinga combination of the GGM filtered image and original X-rayimage. The sizes were determined qualitatively by estimat-ing the size of the surface brightness depressions, using theedges identified in the GGM image as a guide. The cavitiesare shown and labelled in Fig. 8. Two cavities were previ-ously reported by Rafferty et al. (2006), though their workused a shallower X-ray image. Judging from the radial dis-tances and sizes of these cavities, one corresponds directly toCavity "A". The other cavity reported by Rafferty et al. isidentified as two in our work – "B" and "C". Cavity "D" islocated close to the cluster center and may correspond to mo-tions unrelated to AGN activity, but we tabulate its derivedproperties nonetheless. Cavities "E" and "F" are located out-side of a cold front and are less well-defined than their coun-terparts.A list of all cavity parameters can be found in Table 2. Aqualitative figure of merit (FOM), which describes a cavity’scontrast with its surroundings, was provided for each cav-ity (Rafferty et al. 2006). High contrast cavities (FOM = 1)are surrounded by bright rims, medium contrast (FOM = 2)are partially surrounded by bright rims, and low contrast(FOM = 3) have no or faint rims. All necessary thermody-namic quantities were obtained by interpolating the depro-jected radial profiles from Pulido et al. (2018). Note that theproximity of cavity D to the cluster center has a strong sys-tematic effect on the cavity age, and therefore its power. The V ANTYGHEM ET AL .cluster centroid was determined from an aperture with a 20 (cid:48)(cid:48) (87 kpc) radius. It is offset from the X-ray peak by 6 kpc.The enthalpy required to inflate a cavity filled with relativis-tic gas is given by E cav = 4 pV , where p is the cavity pressureand V is its volume. The cavity was assumed to be in pressurebalance with the surrounding gas. The confining pressure wasinterpolated from the deprojected pressure profile to a radiusequal to the distance to the cavity center ( R ).The cavity volume was estimated as the geometric meanof the volumes of oblate and prolate geometries, with semi-axes a and b matching those of the fitted ellipse, so that V = π ( ab ) / . This is equivalent to assuming that the lengthof the semiaxis along the line-of-sight is r = √ ab . The un-certainty for each cavity volume was determined in two ways.First, a 15% uncertainty was assumed in both a and b andpropagated. Second, we used the prolate and oblate geome-tries as measures of the minimum and maximum volumes,respectively. The volume uncertainty was then taken to be thedifference between maximum (minimum) volume and cavityvolume, in absolute value. This evaluates to a fractional un-certainty of (cid:112) a / b − − (cid:112) b / a forprolate, which are respectively the positive and negative er-rors. This is the dominant uncertainty when a / b ≥ .
74 whencompared to a 15% uncertainty in both a and b .The cavity age was estimated using both the sound crossingtime and buoyant rise time (Bîrzan et al. 2004). The soundcrossing time, which assumes the cavity rises at the speed ofsound ( c s ), is simply a function of ICM temperature, and isgiven by t c s = R / c s = R (cid:112) µ m H /γ kT , (2)where γ = 5 / µ = 0 . v t ), t buoy = R / v t (cid:39) R (cid:112) SC / gV . (3)Here C = 0 .
75 is the drag coefficient (Churazov et al. 2001)and S is the bubble cross-section, which is assumed to be itsprojected area. The acceleration under gravity at radius R wasdetermined using the cluster mass profile from Pulido et al.(2018). The two timescales agree to within 20% for all cav-ities. The buoyancy time was used to determine the meancavity power, P cav = E cav / t buoy .These quantities were then used to place mass constraintson the AGN outbursts. The first, M acc = E cav /(cid:15) c , is the massthat must be accreted onto the black hole to fuel the out-burst at an efficiency of (cid:15) , which we assume to be 0 .
1. Thetotal accreted gas mass of 4 . × M (cid:12) can easily be sup-plied by the observed 5 × M (cid:12) of molecular gas. Thecorresponding accretion rate, ˙ M acc = P cav /(cid:15) c , ranges from ∼ . − . (cid:12) yr − for each cavity. The other interestingmass is that of hot gas displaced by the cavities, as this massdictates the maximum mass that can be lifted by the cavities.The mass of displaced gas is given by M disp = n µ m H V , where n = n e + n H is the density of the surrounding gas. Each of thethree larger cavities can lift a few 10 M (cid:12) , so the observedmolecular gas supply could have been lifted by the cavities. DISCUSSION
Origin of the Molecular Gas-AGN Offset
As discussed in Section 3.4, the molecular gas and contin-uum source are mutually offset. The molecular gas residesat the galactic center, while the continuum source is located2 . ∼ −
100 pc displace-ments, which are consistent with an oscillating SMBH follow-ing a gravitational wave recoil (e.g. Lena et al. 2014). M87 isan example of such a system, as its SMBH is offset from thephoto-center by 6 . − − (Chiaberge et al.2017). Emission from a lobe
Instead of originating from the supermassive black hole’saccretion disk, the detected continuum emission could insteadoriginate from the lobe of a radio jet. This would naturallyexplain the offset, although it requires that the emission fromboth the core and a second lobe are undetected. Given theRMS in the continuum image of 7 . µ Jy beam − , any sourcebrighter than 22 . µ Jy would be detected in our observations.The continuum emission, with a spectral index of 0 .
7, isconsistent with originating from an AGN. Spectral indicessteeper than 0 . The SMBH of a secondary galaxy
The spatially offset continuum emission could also origi-nate from the SMBH in a secondary galaxy that is projectednear the BCG core. This SMBH could either be in an inde-pendent background galaxy or the center of a galaxy mergingwith the BCG.The probability that the continuum source is a chance align-ment with an unrelated background radio source is low, basedon our estimates of the source density of radio continuumsources at 100 GHz. Lacking a 100 GHz survey sensitiveenough to detect the source density down to 0 .
13 mJy, wehave instead used a 31 GHz survey of known extragalacticsources (Mason et al. 2009). A source density of 16 . − isappropriate for sources with flux densities ≥ . . − . This exercise results in a probability of 1 . × − that two sources are separated by < . (cid:48)(cid:48) . While our assumed OLECULAR G AS IN Z W Table 2
Cavity MeasurementsCavity FOM a Major Minor R pV t cs t buoy P cav M acc ˙ M acc M disp kpc kpc kpc 10 erg Myr Myr 10 erg s − M (cid:12) M (cid:12) yr − M (cid:12) A 1 21 . . . + − + . − . . ± . . + . − . . + . − . . + . − . . ± .
85B 2 22 . . . + − + . − . . ± . . + . − . + . − . . ± .
12 3 . ± .
1C 2 30 15 . + − . + . − . . + . − . . + . − . . + . − . . + . − . . + . − . D 1 9 .
01 6 .
72 7 .
65 30 . ±
11 8 . + . − . . + . − . . + . − . . + . − . . + . − . . ± .
26E 3 20 . . . . + − . + . − . . ± . . + . − . . + . − . . + . − . . + . − . F 3 22 . . + − . + . − . . + . − . . ± . . ± . . ± .
029 1 . ± . a The cavity figure of merit, which is a qualitative measure of the contrast between a cavity and its surroundings: (1) high contrast, bright rim surroundscavity; (2) medium contrast, bright rim partially surrounds cavity; and (3) low contrast, no rim, or faint rim surrounds cavity. density is based on a flux limit of 1 mJy, which is higher thanthe Zw 3146 continuum source flux (0 .
133 mJy), the actualsource density would need to be several orders of magnitudelarger to exceed a probability of close but random coincidenceof even a few percent.The likelihood that the continuum source originates fromthe core of a merger remnant is much higher. Clumpsand trails possibly corresponding to stellar emission from amerger remnant are detected in the
HST
F606W image (Fig.2). H β and [O III ] line emission may also contribute to thesefeatures. Unlike in systems with dual nuclei, we do not de-tect excess optical emission at the position of the continuumsource. A similar conclusion was drawn in Zw 8193, wherethe unresolved radio emission is offset from the center ofthe BCG by 3 arcsec and may be associated with a merginggalaxy (O’Dea et al. 2010).
Gravitational wave recoil
As two galaxies merge, their black holes migrate towardthe center of the gravitational potential through dynamicalfriction (Begelman et al. 1980). The resulting binary sys-tem tightens through interactions with nearby stars or grav-itational drag from gas until the SMBHs eventually coalesce(Merritt & Milosavljevi´c 2005; Mayer et al. 2007). During thefinal stage of coalescence, the anisotropic emission of gravita-tional waves imparts a recoil velocity to the coalesced object(González et al. 2007; Campanelli et al. 2007). This “kick”can in rare circumstances be as large as 4000 km s − , so canresult in large displacements between the resulting SMBH andthe galactic center.When a SMBH experiences a large recoil ( v kick ∼ − r c ∼ ∼ yr. Then, when the amplitude becomes comparable tothe galaxy’s core radius, the oscillations of the SMBH persistfor ∼ . yr. In a merger between twogalaxies with ∼ M (cid:12) SMBHs, coalescence occurs within1 − HST
F606W image (Fig. 2) may correspond to tidal featuresinduced by a merger. Thus, it is plausible that the SMBHs have recently coalesced and that gravitational wave recoil ac-counts for the displacement of the continuum source.
Acceleration by an asymmetric radio jet
If the radio jets are intrinsically asymmetric in their poweroutput, then the SMBH will experience a net thrust that willdisplace it from its equilibrium position (Shklovsky 1982).When the black hole is accreting near the Eddington limit, theresulting thrust can cause it to oscillate within the host galaxy(Wang et al. 1992; Tsygan 2007; Kornreich & Lovelace 2008;Lena et al. 2014). Clusters are generally not expected to con-tain asymmetric jets, as even in the most powerful systems theSMBH resides at the center of the BCG and, on parsec scales,the jets are two-sided (Liuzzo et al. 2010). However, whilemost cavities in Zw 3146 are balanced by another on the op-posite side of the BCG, the innermost cavity (cavity D) has nodiscernible counterpart. Since this cavity is positioned oppo-site the displacement direction, it can conceivably account forthe displacement, provided it was inflated by an asymmetricjet.The momentum flux for a one-sided relativistic jet is P jet / c .This is the reaction force applied to the black hole from whichthe jet originates. The resulting acceleration for a SMBH ofmass M • is a • = P jet M • c ≈ . × − (cid:18) P jet erg s − (cid:19) (cid:18) M • M (cid:12) (cid:19) − cm s − . (4)The black hole mass in the Zw 3146 BCG, estimated usingthe 2MASS K-band magnitude, is M • = 1 . × M (cid:12) (Mainet al. 2017). The black hole acceleration for cavity D, using P cav as an estimate for jet power (see Table 2), evaluates to1 . × − cm s − .A displaced black hole will experience a restoring accel-eration from the gravitational pull of the BCG. Assumingthe BCG is represented by a singular isothermal sphere withvelocity dispersion σ ∗ = 300 km s − , the gravitational accel-eration (2 σ / R ) at a displacement of R = 1 kpc is 5 . × − cm s − . This is 40 times larger than the accelerationthat would be imparted by cavity D assuming that it has nocounter-jet. An asymmetric jet can therefore not account forthe 2 . Comparison to other BCGs
The majority of BCGs contain molecular gas distributionsthat do not resemble those in gas-rich spiral galaxies. Insteadof rotationally-supported disks or rings, the cold gas in BCGsis generally filamentary (e.g. Olivares et al. 2019; Russell0 V
ANTYGHEM ET AL ..
ANTYGHEM ET AL .. Figure 9.
Comparison between total integrated flux maps for Zw3146 CO(1-0) with PKS0745-191 CO(1-0) and CO(3-2). et al. 2019). In a sample of 12 BCGs, only two have ≤ − <
5% of the emission is located atthe vertex of its three filaments. Without additional higherresolution observations of Zw 3146 we cannot conclusivelydetermine its true molecular gas morphology. However, thesimilarities with PKS0745-191 suggest that it could share themorphology of three radial filaments with little central emis-sion, observed with intermediate resolution.
Origin of the Molecular Gas
The massive (5 × M (cid:12) ) reservoir of cold gas residingin the central 4 kpc of the Zw 3146 BCG places stringent re- quirements on its origin. Several tens of gas-rich spiral galax-ies would need to be stripped to account for this gas supply.In cluster cores these are rare, as most central galaxies are red,gas-poor ellipticals (Best et al. 2007). Sustained stellar massloss will contribute to the molecular gas supply, but cannotaccount for the entire reservoir (Sparks et al. 1989; Voit &Donahue 2011). Instead, like most BCGs, the primary mech-anism for forming the molecular gas is the condensation ofthe hot atmosphere.Although the hot atmosphere contains ample fuel to pro-duce the molecular gas reservoir, the region covered bymolecular emission is dominated by molecular gas. Only2 . ± . × M (cid:12) of hot gas is present within the central10 kpc, while the 5 × M (cid:12) of cold gas is located in the cen-tral 4 kpc. Zw 3146 is not unique in this regard. In other sys-tems observed by ALMA, the molecular gas mass is approx-imately equal to the hot gas mass when considering only theregion that contains molecular emission (Russell et al. 2019).Forming all of the molecular gas in this region would result ina significant inflow as the central hot atmosphere is depleted.Alternatively, the cold gas could be formed on larger scalesin filaments that are too faint to be detected. The Ly- α emis-sion, which could be coincident with this fainter molecularline emission, extends well beyond the inner 4 kpc in whichthe molecular gas is confined. Gas condensing out of the cen-tral 20 kpc would have 2 × M (cid:12) of hot gas to draw from.In the stimulated feedback paradigm, molecular gas con-denses from intracluster gas that has been entrained and up-lifted by a radio bubble (McNamara et al. 2016). Zw 3146harbours multiple X-ray cavities that have likely arisen from2 − M (cid:12) of hot gas (see Table 2). This isseveral times larger than the mass of the molecular gas in theextensions (Table 1). Combined, the cavities displace a totalof 1 . × M (cid:12) of hot gas. Thus these cavities could haveinitiated the formation of the entire 5 × M (cid:12) reservoir of OLECULAR G AS IN Z W Chandra suggest that the hot atmosphere is condens-ing at a rate of 300 M (cid:12) yr − (Egami et al. 2006b). However,a recent spectral analysis conducted using archival data fromthe XMM-Newton
Reflection Grating Spectrometer, followingthe prescription of Liu et al. (2019), measured <
50 M (cid:12) yr − of continuous radiative cooling from the bulk temperature(4 keV) down to 0 .
01 keV (Liu et al. in preparation). Thisincludes 130 ±
60 M (cid:12) yr − of gas cooling to 0 . <
78 M (cid:12) yr − cooling below that. This is closely matchedto the star formation rate of 70 ±
14 M (cid:12) yr − . It would take1 Gyr to produce the observed reservoir with condensationproceeding at 50 M (cid:12) yr − . Condensation at this rate over theage of a single cavity would contribute a few 10 M (cid:12) of coldgas. The age of the AGN two most energetic AGN outbursts,determined by the buoyancy time of their respective cavities,are ∼ × yr and ∼ × yr. This translates to the for-mation of 2 − . × M (cid:12) of molecular gas, 4 −
7% of thetotal supply.Stimulated feedback can therefore contribute significantlyto the reservoir of molecular gas, but many cycles of AGNfeedback would be required to account for the entire reservoir.Each cycle of AGN feedback can stimulate the formation of afew 10 M (cid:12) of molecular gas. This process must be sustainedfor 1 Gyr or longer in order to accumulate the cold gas massthat is observed. The consumption of cold gas in star forma-tion complicates this picture, however, as the star formationrate is comparable to the mass deposition rate. CONCLUSIONS
Our ALMA CO(1-0) observations of the Zw 3146 BCGhave revealed a massive (5 × M (cid:12) ) reservoir of molecu-lar gas located within 4 kpc of the galactic center. The gasdistribution is marginally resolved into three filaments. Wehave also identified six cavities in the Chandra
X-ray image.Each molecular filament trails one of the X-ray cavities.The cavities have displaced a mass of intracluster gas sev-eral times larger than the molecular gas mass, suggesting thatstimulated cooling – where low entropy gas from the clustercore is uplifted by rising radio bubbles until it becomes ther-mally unstable and condenses – may be responsible for theformation of the molecular gas. However, stimulated coolingwould have difficulty accounting for the entire reservoir, asthe cavity ages are more than ten times shorter than the 1 Gyrit would take to form 5 × M (cid:12) of cold gas at the observedcondensation rate. The total gas supply is likely accumulatedover many cycles of AGN feedback, with each cycle shapingits distribution.While the molecular gas is centered within the BCG, thecontinuum source originating from the AGN is not. Thecontinuum source is located 2 . < yr) blackhole merger may account for this displacement. Alternativemechanisms – such as a chance alignment with a backgroundgalaxy, the continuum corresponding to the lobe of a jet, oran asymmetric jet accelerating the SMBH of the BCG – are unlikely.We thank the anonymous referee for their helpful com-ments that improved this paper. ANV thanks Yjan Gordonand Cameron Lawlor-Forsyth for their support. ANV, BRM,CPO, and SAB are supported by the Natural Sciences and En-gineering Research Council of Canada. This paper makes useof the ALMA data ADS/JAO.ALMA 2018.1.01056.S. ALMAis a partnership of the ESO (representing its member states),NSF (USA) and NINS (Japan), together with NRC (Canada),NSC and ASIAA (Taiwan), and KASI (Republic of Korea), incooperation with the Republic of Chile. The Joint ALMA Ob-servatory is operated by ESO, AUI/NRAO, and NAOJ. Thisresearch made use of Astropy, a community-developed corePython package for Astronomy. This research made use ofAPLpy, an open-source plotting package for Python hosted athttp://aplpy.github.com.REFERENCES Alatalo, K., Blitz, L., Young, L. M., et al. 2011, ApJ, 735, 88Allen, S. W., Edge, A. C., Fabian, A. C., et al. 1992, MNRAS, 259, 67Bartlett, D. J., Desmond, H., Devriendt, J., Ferreira, P. G., & Slyz, A. 2020,arXiv e-prints, arXiv:2007.01353Batcheldor, D., Robinson, A., Axon, D. J., Perlman, E. S., & Merritt, D.2010, ApJ, 717, L6Bayet, E., Hartquist, T. W., Viti, S., Williams, D. A., & Bell, T. A. 2010,A&A, 521, A16Begelman, M. C., Blandford, R. D., & Rees, M. J. 1980, Nature, 287, 307Best, P. N., von der Linden, A., Kauffmann, G., Heckman, T. M., & Kaiser,C. R. 2007, MNRAS, 379, 894Bîrzan, L., Rafferty, D. A., McNamara, B. R., Wise, M. W., & Nulsen,P. E. J. 2004, ApJ, 607, 800Bîrzan, L., Rafferty, D. A., Nulsen, P. E. J., et al. 2012, MNRAS, 427, 3468Blanton, E. L., Sarazin, C. L., McNamara, B. R., & Wise, M. W. 2001, ApJ,558, L15Böhringer, H., Voges, W., Huchra, J. P., et al. 2000, ApJS, 129, 435Bolatto, A. D., Wolfire, M., & Leroy, A. K. 2013, ARA&A, 51, 207Burns, J. O. 1990, AJ, 99, 14Campanelli, M., Lousto, C. O., Zlochower, Y., & Merritt, D. 2007,Phys. Rev. Lett., 98, 231102Cavagnolo, K. W., Donahue, M., Voit, G. M., & Sun, M. 2008, ApJ, 683,L107Chiaberge, M., Ely, J. C., Meyer, E. T., et al. 2017, A&A, 600, A57Churazov, E., Brüggen, M., Kaiser, C. R., Böhringer, H., & Forman, W.2001, ApJ, 554, 261Churazov, E., Forman, W., Jones, C., & Böhringer, H. 2000, A&A, 356, 788Coble, K., Bonamente, M., Carlstrom, J. E., et al. 2007, AJ, 134, 897Cowie, L. L., Hu, E. M., Jenkins, E. B., & York, D. G. 1983, ApJ, 272, 29Crawford, C. S., Allen, S. W., Ebeling, H., Edge, A. C., & Fabian, A. C.1999, MNRAS, 306, 857Dasyra, K. M., & Combes, F. 2011, A&A, 533, L10David, L. P., Lim, J., Forman, W., et al. 2014, ApJ, 792, 94De Grandi, S., Ettori, S., Longhetti, M., & Molendi, S. 2004, A&A, 419, 7De Grandi, S., & Molendi, S. 2001, ApJ, 551, 153Donahue, M., de Messières, G. E., O’Connell, R. W., et al. 2011, ApJ, 732,40Donahue, M., Connor, T., Fogarty, K., et al. 2015, ApJ, 805, 177Dunn, R. J. H., & Fabian, A. C. 2006, MNRAS, 373, 959Ebeling, H., Edge, A. C., Bohringer, H., et al. 1998, MNRAS, 301, 881Edge, A. C. 2001, MNRAS, 328, 762Edge, A. C., Fabian, A. C., Allen, S. W., et al. 1994, MNRAS, 270, L1Edge, A. C., & Frayer, D. T. 2003, ApJ, 594, L13Edge, A. C., Wilman, R. J., Johnstone, R. M., et al. 2002, MNRAS, 337, 49Edge, A. C., Oonk, J. B. R., Mittal, R., et al. 2010, A&A, 518, L47Egami, E., Rieke, G. H., Fadda, D., & Hines, D. C. 2006a, ApJ, 652, L21Egami, E., Misselt, K. A., Rieke, G. H., et al. 2006b, ApJ, 647, 922Fabian, A. C. 1994, ARA&A, 32, 277—. 2012, ARA&A, 50, 455Fabian, A. C., Sanders, J. S., Taylor, G. B., et al. 2006, MNRAS, 366, 417Ferland, G. J., Fabian, A. C., Hatch, N. A., et al. 2009, MNRAS, 392, 1475Forman, W., Jones, C., Markevitch, M., Vikhlinin, A., & Churazov, E. 2002,arXiv e-prints, astro
ANTYGHEM ET AL ..