BAT AGN Spectroscopic Survey. XX. Molecular Gas in Nearby Hard-X-Ray-selected AGN Galaxies

Our AGN parent sample consists of 836 ultrahard-X-ray-selected (14–195 keV) AGN included in the 70-month Swift-BAT all-sky catalog (Baumgartner et al. 2013). Of these, we focused on 348 un-beamed AGN with low Galactic extinction [E(B − V) < 0.5] and in a redshift range of 0.01 < z < 0.05. Sources outside of the Galactic plane (b > 10°) were given higher priority, but in seven cases, sources with low Galactic extinction within the Galactic plane were observed. The lower redshift bound was chosen so that a single pointing would recover most of the CO flux from the host galaxy (see relevant angular scales below), while the higher redshift bound was chosen to guarantee a high detection rate within a reasonable integration time. The results for the complementary sample of z < 0.01 BAT AGN galaxies are presented in Rosario et al. (2018). In total, 200 AGN galaxies were newly observed—165 with the Atacama Pathfinder Experiment (APEX) and 35 with the James Clerk Maxwell Telescope (JCMT)—while 13 were obtained from the literature. Figure 1 gives an overview of the sky distribution of the 213/348 (61%) observed and the 135/348 (39%) unobserved AGN galaxies between 0.01 < z < 0.05. A further study of BAT AGN galaxies in the northern hemisphere (δ > 0°), based on the IRAM 30 m telescope, will appear in a future publication (T. Shimizu et al. 2021, in preparation). An ALMA campaign on 32 BAT AGN galaxies in CO (2–1) has also recently been completed (T. Izumi et al. 2021, in preparation).

Zoom In Zoom Out Reset image size

2.1.1. APEX Observations

2.1.2. JCMT Observations

2.1.3. Literature Measurements

We measure the velocity-integrated line flux S_CO (in units of Jy km s⁻¹) from the reduced spectra by summing the signal within a spectral range defined by hand, following the technique in xCOLD GASS (Saintonge et al. 2017). If the CO(2–1) line is undetected or very weak, the window is set to a width of 300 km s⁻¹ to estimate uniform upper limits. The CO line luminosities are then calculated following the relation

$$\begin{eqnarray}&&L{{\prime} }_{\mathrm{CO}(2-1)}=3.25\times {10}^{7}{S}_{\mathrm{CO}(1-0)}{\nu }_{\mathrm{obs}}^{-2}{D}_{L}^{2}{\left(1+z\right)}^{-3},\end{eqnarray} \tag{ 1 }$$

as given in Solomon et al. (1997), where the line luminosity $L{{\prime} }_{\mathrm{CO}}$ (in K km s⁻¹) is derived from the velocity-integrated line flux, S_COΔv (in Jy km s⁻¹) and the luminosity distance, D_L (in Mpc). The error on the line flux is defined as the rms noise achieved around the CO(2–1) line, in spectral channels with width ${\rm{\Delta }}{w}_{\mathrm{ch}}^{-1}=50$ km s⁻¹, and W_CO is the width of the CO(2–1) line in km s⁻¹. The detection threshold is set to S/N = 3. We provide three examples of CO line detections and nondetections in Figures 2 and 3, respectively. For 15 sources characterized by particularly narrow CO emission, we produced a higher resolution spectrum to optimize the profile fit using either 5, 10, or 20 km/s binning. The complete sample of CO line measurements is found in the online figure set associated with Figure 2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (200 panels)

Tarball of figure set images

Zoom In Zoom Out Reset image size

We have additionally measured the line widths (W50_CO21) and systemic velocities using Gaussian profile fitting for the detected sources, following the same approach as used in the xCOLD GASS studies (Tiley et al. 2016). Each reduced spectrum is first fit with a single Gaussian through Monte Carlo minimization, using the SHERPAPython package in CIAO4.11. As discussed in Tiley et al. (2016), a symmetric Gaussian double-peak function may be more appropriate for higher S/N spectra, as it minimizes potential biases as a function of S/N, inclination (apparent width), and rotation velocity (intrinsic width). We therefore also fit each spectrum with a Gaussian double-peak function (a parabolic function surrounded by two equidistant and identical half-Gaussians forming the low- and high-velocity edges of the profile), which reduces to a single Gaussian in the limit where the half-width of the central parabola is zero. We report the results obtained from the fitting method that provides the lowest $\left|{\chi }_{\nu }^{2}\right|$ (where ${\chi }_{\nu }^{2}$ is the χ² of the fit divided by the degrees of freedom). We show three examples of line fits in Figure 4. We find good agreement between profile fitting and integrated measures (see Appendix D).

Zoom In Zoom Out Reset image size

We complement the new CO measurements with various galaxy and AGN physical properties, drawn from BASS (Koss et al. 2017)—a large effort to collect optical and X-ray spectra (Ricci et al. 2017) and various other measurements for Swift-BAT AGN, in order to assess black hole masses, accretion rates, and bolometric luminosity estimates for an unbiased local AGN sample. Optical spectra can be accessed through the BASS website.³⁴ We incorporate products from both Data Releases 1 (DR1) and 2 (DR2); DR1 is currently publicly available, while DR2 will be published in the near future (M. Koss et al. 2021, in preparation). For all the BAT AGN galaxies in our sample, we use the following key parameters (see Appendix B for a further discussion of the uncertainties in parameter estimation), listed in Table 1 or in associated BASS publications:

• 1.  
  BAT ID: Catalog ID in the BAT survey.³⁵
• 2.  
  α_J2000 and δ_J2000: Right ascension and decl. of the optical/IR counterpart of the BAT AGN, in decimal degrees, based on WISE positions.
• 3.  
  AGN type and redshift: Measurements from BASS/DR1 (Koss et al. 2017) and DR2 (K. Oh et al. 2021, in preparation), when available. The redshift is based on optical emission-line fitting (mainly of $\left[{\rm{O}}\,{\rm\small{III}}\right]\,\lambda 5007$). While the emission-line redshifts may have an offset from the systemic host velocities (Rojas et al. 2020), they provide a rough estimate of the likely CO centroid.
• 4.  
  rk20fe: The isophotal radius at a surface brightness of 20 mag arcsec⁻² in the K band, taken from the 2MASS extended source catalog (Jarrett et al. 2000).
• 5.  
  Inclination angle: Galaxy inclination angle based on r-band optical imaging from the SDSS, Kitt Peak (Koss et al. 2011a), or from the 2MASS extended source catalog (Jarrett et al. 2000). In order to determine the inclination from the observed ratio of the semiminor to semimajor axes of the galaxy, we use the prescription of Davis et al. (2011) for both early- and late-type galaxies.
• 6.  
  Morphology: Morphological classification from Galaxy Zoo (Lintott et al. 2008), when available, followed by the Third Reference Catalog of Bright Galaxies (RC3, de Vaucouleurs et al. 1995), and finally visual inspection following the approach of Koss et al. (2011a), which is based on optical imaging. Values can be "elliptical," "spiral," or "uncertain." See Appendix A for a further discussion of the uncertainty in morphology measures.
• 7.  
  Stellar mass: Stellar masses of the BAT AGN host galaxies. We combined near-IR data from 2MASS, which is more sensitive to stellar emission, with mid-IR data from the AllWISE catalog (Wright et al. 2010), which is more sensitive to AGN emission (i.e., reprocessed by the circumnuclear dusty gas). See Powell et al. (2018) for a further description of these measurements.
• 8.  
  Star formation rate: The majority (80%, 171/213) of SFR estimates were derived from the Herschel-based study of BAT AGN galaxies by Shimizu et al. (2017), which measured the total IR (8–1000 μm) luminosities by decomposing the IR SEDs into AGN- and galaxy-related components using WISE (3.4, 4.6, 12, and 22 μm), Herschel PACS (70 and 160 μm) and SPIRE (250, 350, and 500 μm) imaging. A subset of 19% (40/213) AGN galaxies were not part of this Herschel imaging publication. For these we adopt the results of Ichikawa et al. (2019), who performed a similar IR SED decomposition but included any available WISE, Akari, IRAS, and Herschel measurements. While both studies have similar values, we preferentially use the Herschel-based measurements of Shimizu et al. (2017) to avoid the much larger PSF of the IRAS data compared to the CO. Ultimately, 15/213 (7%) of the SED estimates are only upper limits because no FIR detection (>60 μm) was found. Additionally, 3/213 (1%) sources lie within 10° of the Galactic plane and were not part of either study. See Appendix B for a further discussion about the uncertainties and offsets in SFR measurements.
• 9.  
  Offset from the main sequence of star formation: A correlation between stellar mass and SFR has been found for galaxies in the local and distant universe (see review in, e.g., Tacconi et al. 2020). The offset, Δ(MS), is defined as the difference on a logarithmic scale between the observed SFR of a galaxy and its expected SFR in relation to the main sequence (MS) from Renzini & Peng (2015). A value of Δ(MS) = 1 would indicate that a galaxy is elevated above the expected value of SFR on the MS by a factor of 10, while Δ(MS) = −1 indicates a factor of 10 lower than expected.

Beam corrections are needed because the lowest redshift galaxies (z ∼ 0.01–0.02) with the largest angular sizes may have more flux extending outside of the beam that goes unmeasured compared to more distant sources (z ∼ 0.04–0.05) with smaller angular sizes, where most or all of the CO(2–1) flux is likely recovered.

For each galaxy we estimate an aperture correction separately using a model of the gas distribution and either Herschel/PACS imaging (Lamperti et al. 2019) or the galaxy diameter from the K-band effective radius (see Saintonge et al. 2011b, 2017, for more details on this technique). For the majority of the sample (81%, 179/213), we use Herschel/PACS imaging. We assume that the CO luminosity is traced by the PACS 160 μm Herschel emission (FWHM = 12'') from Meléndez et al. (2014). We use the PACS 160 μm images because the longer wavelength is less likely to be contaminated by AGN emission, which can still contribute to a significant fraction of the 70 μm emission (e.g., Shimizu et al. 2017; Ichikawa et al. 2019).

We first measure the total infrared 160 μm emission of the galaxy within a radius large enough to include the entire galaxy. Then we take the ratio between the flux from the map multiplied by the CO(2–1) Gaussian beam sensitivity function and the total IR flux, and we use this value to extrapolate the total CO(2–1) flux.

The images that we used to trace the distribution of the FIR emission are not maps of the "true" distribution, but instead are maps of the "true" distribution convolved with the PSF of PACS (12''). Therefore an additional correction is needed.

To correct for the effect of the PACS PSF smearing, we use a simulated galaxy gas profile, following the procedure described in Saintonge et al. (2011b). For each galaxy, we create a model galaxy simulating a molecular gas disk following an exponential profile, with a scale length equivalent to its half-light radius. Then the profile is tilted according to the inclination of the galaxy, and we measure the amount of flux that would be observed from this model galaxy, using an aperture corresponding to the size of the beam. We multiply the IR image by a 2D Gaussian centered on the galaxy center and with FWHM equal to the beam size to mimic the effect of the beam sensitivity of the telescope that took the CO observations. By taking the ratio of these two measurements, we estimate how much the flux changes due to the effect of the FIR PSF smearing. We perform this 160 μm based correction on 69% (148/213) of images. For an additional 7% (15/213), we use the 70 μm surface brightness because the 160 μm was a nondetection.

In cases where the PACS imaging resulted in a nondetection, or where no PACS imaging exist (51/213, or 24%), we directly use the simulated galaxy gas profile discussed above as described in the COLD GASS and xCOLD GASS papers (Saintonge et al. 2011b, 2017) and solely use the K-band galaxy diameter (D = 2 × r_k20fe) and inclination to estimate the correction factor. Further discussion of the beam corrections is presented in Appendix C.

As the properties of (molecular) gas in galaxies are known to be related to stellar mass and morphology (e.g., Saintonge et al. 2011b; Young et al. 2011), it is critical that our analysis of the BAT AGN galaxies is done in comparison to a large, unbiased control sample of inactive galaxies spanning a similar range in stellar mass.

The IRAM 30 m xCOLD GASS sample (Saintonge et al. 2017) comprises 532 galaxies from two large IRAM programs, with measurements and analyses of molecular gas across a large range of stellar masses [i.e., $9\lt \mathrm{log}({M}_{* }/{M}_{\odot })\lt 11.5$], and over the same redshift range as our sample of BAT AGN galaxies (i.e., 0.01 < z < 0.05).

The xCOLD GASS sample is primarily composed of IRAM measurements, meaning that CO(1–0), with a HPBW of ≈22'', is a good match to our APEX (HPBW = 27'') and JCMT (HPBW = 20'') CO(2–1) data, even though a different line is used to probe the molecular gas. The xCOLD GASS sample also contains 28 galaxies with APEX(2–1) measurements, and the authors found that the APEX CO(2–1) to IRAM CO(1–0) luminosity ratio for velocity-integrated measurements is r₂₁ = 0.79 ± 0.03, based on robust measurements of both lines, with no systematic variations across the sample properties (e.g., SFR, sSFR, or ${f}_{{{\rm{H}}}_{2}}$). The range typically varies between ∼0.6 and 1.0 in nearby galaxies (Leroy et al. 2009). We adopt this r₂₁ conversion factor to convert our AGN galaxy measurements from CO(2–1) into CO(1–0) before converting into molecular gas. Further analysis of the line ratio will be discussed in T. Shimizu et al. (2021, in preparation), who reobserved 56 BAT AGN galaxies from our APEX and JCMT sample using IRAM in both CO(1–0) and CO(2–1).

For the purposes of our study, we excluded 164 galaxies with low stellar masses $[\mathrm{log}({M}_{* }/{M}_{\odot })\lt 10$], which are only rarely present in our BAT AGN galaxy sample. We further ensured that the xCOLD GASS comparison sample does not include any AGN. First, no BAT AGN galaxies were present in xCOLD GASS, likely due to their low space density. Second, we used the SDSS nebular emission-line measurements from the OSSY database (Oh et al. 2011) to test for AGN activity based on strong line diagnostics for the entire xCOLD GASS sample. We identified 14 optically selected AGN based on the $\left[{\rm{O}}\,{\rm\small{III}}\right]/{\rm{H}}\beta$ versus $\left[{\rm{N}}\,{\rm\small{II}}\right]/{\rm{H}}\alpha$ diagnostics of Kewley et al. (2006). We then also checked against the SDSS-based catalog of broad Hα line emitters compiled by Oh et al. (2015), but no galaxies were found to be part of this catalog. We finally checked the Veron-Cetty Catalog of Quasars and AGN (Veron-Cetty & Veron 2010) and found that three xCOLD GASS sources were reported as hosting AGN. We also removed four radio-loud AGN galaxies (Best & Heckman 2012). In the process of reviewing the xCOLD GASS sample, we also realized that for three galaxies, the SDSS spectroscopic pipeline inaccurately targeted a spiral arm or other feature and that the nucleus of the galaxy is more than 5'' away. We have excluded these galaxies also because their photometry and measurements are unreliable (ID = 20790, 36169, and 31592). The final xCOLD GASS inactive galaxy comparison sample thus holds 344 inactive galaxies.

Throughout this analysis and for both samples, we adopt a Milky Way-like conversion factor from CO luminosity to H₂ mass of α_CO = 4.3 M_*(K km s⁻¹ pc²)⁻¹ (see e.g., Bolatto et al. 2013 for a review). Studies of nearby galaxies with kiloparsec-scale resolutions have found that α_CO is generally flat with galactocentric radius, with average values of 3.1 ± 0.3 for nearby galaxies. While ULIRG-type galaxies are thought to have lower α_CO values (e.g., α_CO ∼ 1 M_⊙/(K km s⁻¹ pc²); Narayanan et al. 2012), none of these sources are part of either sample. Given that the stellar mass–metallicity relation is rather flat over the average stellar mass that we study (log M_*/M_⊙ > 10.0, Yates et al. 2012) and the difficulty of measuring metallicity in AGN hosts using emission lines because of contamination, we prefer this simple approach rather than invoking metallicity-dependent gradients and conversion factors (e.g., Accurso et al. 2017).

The median beam correction for all the BAT AGN galaxies is 1.33, which is higher by 16% than the xCOLD GASS survey (1.15), mainly driven by the lower redshift (〈z〉 = 0.026 versus 〈z〉 = 0.036), and higher stellar masses of the galaxies in the AGN galaxy sample. This is somewhat by design as the redshift range (z = 0.026–0.05) was selected to ensure that the whole galaxy was in the IRAM beam for the most massive galaxies (log M_*/M_⊙ > 10.0) in xCOLD GASS. We verified that the results presented throughout the rest of the present study are not significantly affected by inadequate aperture corrections by confirming that the distributions of key quantities and/or relations between them do not depend on galaxy redshift (or distance).

We provide the full catalog of CO(2–1) measurements and key derived quantities for all 213 BAT AGN galaxies in our sample. A summary of possible trends between the CO detection fraction and various AGN galaxy properties is provided in Figure 5.

Zoom In Zoom Out Reset image size

The catalog includes the following quantities, included in Table 2:

• 1.  
  BAT ID: Unique ID in the Swift -BAT 70-month catalog (see footnote 35).
• 2.  
  Telescopeand HPBW: The telescope used for the CO observation—either JCMT or APEX for our own observations, or for the archival data, along with the associated beam size.
• 3.  
  FlagCO21: A detection flag for the CO(2–1) line, with "1" indicating a detection, and "0" indicating a nondetection. When the line is not detected, the tabulated line luminosities and molecular gas masses (see below) are 3σ upper limits.
• 4.  
  σ_CO21: rms noise achieved around the CO(1-0) line, in spectral channels with width Δw_ch = 50 km s⁻¹. The quoted rms is at 50 km s⁻¹ even for spectra with finer binning.
• 5.  
  Beam-correction factor: Beam aperture correction, derived either through simulations following the approach developed for the COLD GASS sample (Saintonge et al. 2011b), or by assuming the CO luminosity is traced by the 160 μm emission observed by Herschel/PACS (FWHM = 12''), as reported by Meléndez et al. (2014).
• 6.  
  ${{\mathtt{L}}}_{\mathrm{CO}21,\mathrm{corr}}^{{\prime} }$: Total CO(2–1) line luminosity, calculated from ${L}_{\mathrm{CO}21,\mathrm{obs}}^{{\prime} }$ and the aperture correction. The error includes the measurement uncertainty, a 10% flux calibration error, and the 15% uncertainty on the aperture correction (Saintonge et al. 2011b)—all combined in quadrature. If the source is undetected in CO (FlagCO21 = 0 in Col. 3), then the tabulated value corresponds to a 3σ upper limit.
• 7.  
  $\mathrm{log}{M}_{{\rm{H}}2}$: Total molecular gas mass, including the helium contribution, calculated from the integrated CO(2–1) line luminosity assuming a conversion of 0.79 from CO(2–1) to CO(1–0), and assuming a Milky Way-like conversion factor of α_CO = 4.3 M_* (K km s⁻¹ pc²)⁻¹ (Bolatto et al. 2013). If the source is undetected in CO (FlagCO21 = 0 in Col. 3), then the tabulated value corresponds to a 3σ upper limit.
• 8.  
  t_dep(H₂): The gas-depletion timescale (t_dep(H₂) ≡ M_H2/SFR). A small number (8%, 16/207) of AGN galaxies have no measurements (denoted as "...") because they lack robust SFR measurements. Most of these (14/16) were not observed within the Herschel program, and thus have less sensitive SFR upper limits, or they lie close to the Galactic plane where no measurements were performed (12%, 2/16). Most are also undetected in molecular gas (75%, 12/16). As they have upper limits on SFR and molecular gas, a gas-depletion timescale, which is the ratio of the two, cannot be estimated.
• 9.  
  Pro_type: CO line measured with either a single- or double-peak Gaussian profile (as indicated by W50_type; Tiley et al. 2016).
• 10.  
  ${L}_{\mathrm{pro},\mathrm{corr}}^{{\prime} }$:Total CO(2–1) line luminosity, calculated from the profile fit ${L}_{\mathrm{pro},\mathrm{corr}}^{{\prime} }$ with an aperture correction.
• 11.  
  W50, z_CO: FWHM of the line from the profile fit in km s⁻¹, the associated error, and redshift from the central velocity. The tabulated errors correspond to 1σ.

Table 2.  Catalog of CO(2–1) Measurements

ID	Tele.	HPBW	Flag	σ	Caper	S/Ninte	${L}_{\mathrm{inte},\mathrm{corr}}^{{\prime} }$	$\mathrm{log}{M}_{{\rm{H}}2}$	tdep(H2)	Protype	${L}_{\mathrm{pro},\mathrm{corr}}^{{\prime} }$	S/Npro	W50	zCO
		('')		(mK)	Factor		(108 K km s−1 pc2)	($\mathrm{log}{M}_{\odot }$)	(log yr)		(108 K km s−1 pc2)		( km s−1 )	( km s−1 )
3	APEX	27	1	1.8	1.05	5.5	3	9.11	8.44	1	3	3.2	74 ± 21	7659 ± 10
17	APEX	27	1	0.9	1.04	11	7.7	9.52	8.66	2	8.1	10.4	561 ± 158	8614 ± 55
28	APEX	27	1	1.7	1.31	14.4	11.7	9.7	8.79	2	11.4	12.5	461 ± 94	6604 ± 42
44	APEX	27	1	1.3	1.21	7.4	25.2	10.03	9.05	2	23	9.2	469 ± 123	14363 ± 56
50	APEX	27	1	1.1	1.16	7	2.9	9.1	8.75	2	4.8	11.9	559 ± 160	5737 ± 51
58	APEX	27	1	0.8	1.51	6	0.7	8.5	7.83	1	0.6	6.2	499 ± 237	3408 ± 79
62	APEX	27	1	2	1.78	16.3	5.7	9.39	8.9	2	4.7	7.8	304 ± 59	3572 ± 26
63	APEX	27	1	0.8	1.17	7	0.9	8.61	8.41	1	0.5	5.7	158 ± 135	3634 ± 18
72	APEX	27	1	0.7	1.19	4.3	1	8.63	8.28	1	1.7	8	601 ± 252	5748 ± 83
77	APEX	27	1	2.1	1.3	8.4	4	9.24	8.73	1	4	5.8	86 ± 23	5075 ± 7
79	APEX	27	1	1.1	1.02	10.7	14.6	9.79	8.73	2	14.9	10.2	522 ± 121	11416 ± 52
81	APEX	27	1	1.3	1.35	13.2	9.6	9.61	8.82	2	10.2	9.9	381 ± 120	7707 ± 42
83	APEX	27	1	4.4	1.11	12.8	10.3	9.64	8.86	1	11.1	9.9	268 ± 56	4934 ± 23
95	APEX	27	1	1	1.06	6.8	5.9	9.41	8.87	2	6.2	8.4	409 ± 166	10033 ± 54
96	APEX	27	1	2.9	2.19	8.4	12.8	9.74	8.67	1	10.5	7.9	281 ± 83	5006 ± 32
101	APEX	27	1	2.1	1.21	9.2	4.8	9.32	8.81	1	5.3	10.2	323 ± 62	4903 ± 31
102	APEX	27	1	0.5	2.88	5.9	1	8.66	8.41	1	0.4	2.2	42 ± 35	4170 ± 9
114	APEX	27	1	0.9	1.42	12.4	26.4	10.05	8.86	2	24.5	8.8	533 ± 115	14295 ± 60
116	APEX	27	1	1.3	1.52	11.2	8.4	9.56	8.77	2	10.1	8.4	252 ± 44	7926 ± 19
128	APEX	27	1	1.2	1.04	4.7	3.4	9.16	...	1	3	3	48 ± 54	10069 ± 7

Note. A detailed description of this table's contents is given in Section 3.6.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

The full BAT AGN galaxy CO catalog, including all CO spectra, SDSS and PS1 images, and galaxy and ancillary AGN observations from BASS will be available at the BASS website (http://www.bass-survey.com).

As shown in Figure 6, our sample of BAT AGN galaxies strongly skews toward higher stellar masses compared to the xCOLD GASS sample of inactive galaxies $[\mathrm{log}({M}_{* }/{M}_{\odot })\,=10.75\pm 0.02$ versus $\mathrm{log}({M}_{* }/{M}_{\odot })=10.58\pm 0.02$, respectively]. Importantly, considering the regime where most BAT AGN galaxies reside ($\mathrm{log}({M}_{* }/{M}_{\odot })\gt 10.5$), the median molecular gas mass and molecular gas fraction are systematically higher in BAT AGN galaxies than in the inactive galaxies at fixed stellar mass (p < 0.001;  right panels of Figure 6). Additionally, the BAT AGN galaxies typically lie on the MS of star-forming galaxies compared to inactive galaxies in xCOLDGASS, which lie below it for the majority of the sample (p < 0.001, $\mathrm{log}({M}_{* }/{M}_{\odot })\gt 10.2$).

Zoom In Zoom Out Reset image size

In Figure 7 we present the molecular gas content of the BAT AGN galaxies and the xCOLD GASS inactive galaxies in the context of other star formation properties. The AGN galaxies clearly occupy a region of elevated star formation (both in terms of SFR and sSFR). In particular, passive galaxies—i.e., those with $\mathrm{log}(\mathrm{sSFR}/{\mathrm{yr}}^{-1})\lt -11$—are a distinct minority among our BAT AGN galaxies (2%–7%, due to upper limits), whereas they comprise roughly half of the inactive galaxies (49%). Over the limited range of SFRs where the two samples overlap ($0.2\lt \mathrm{log}(\ \mathrm{SFR}/{M}_{\odot }\,{\mathrm{yr}}^{-1})\lt 0.7$), the amount of molecular gas is not significantly different between the inactive galaxies and AGN galaxies. The gas fractions at a given sSFR are lower for AGN galaxies than for inactive galaxies (p = 0.004), however. The gas-depletion timescales [t_dep(H₂) ≡ M_H2/SFR] for AGN galaxies are not significantly different.

Zoom In Zoom Out Reset image size

In Figure 8 we show the molecular gas and gas fraction timescales, and the depletion timescale with offset from the MS as defined by Renzini & Peng (2015). The BAT AGN tend to largely lie on this MS. The xCOLDGASS inactive galaxies have a much larger fraction of galaxies significantly below the MS consistent with their lower average SFR. Over the limited range in which the two samples overlap in offset from the MS (−0.5 < Δ(MS) < 0.5), the AGN galaxies have lower gas fractions (p < 0.01), but not significantly different molecular gas masses or depletion timescales.

Zoom In Zoom Out Reset image size

4.1.1. Galaxy Molecular Gas Luminosity Function

Another way to study how important molecular gas is to AGN activity is to compare the number of AGN galaxies with a given molecular gas mass for a given volume to the number of inactive galaxies. We do this by comparing the CO luminosity function from xCOLD GASS (see Figure 2, Fletcher et al. 2020), which was constructed to measure the molecular gas of all galaxies of a given stellar mass ($\mathrm{log}({M}_{* }/{M}_{\odot })\gt 9.0$. The mass function was then created by normalizing to match the stellar mass function of local galaxies, and thus be representative of local galaxies. The BAT AGN galaxy sample is then a subset of this local galaxy sample for all galaxies hosting an AGN above a certain X-ray luminosity across the sky. As the BAT sample is flux limited, we use the ${V}_{\max }$ correction method following Ananna et al. (2020) to estimate the correction factor for each BAT AGN galaxy based on the observability of its intrinsic X-ray flux within the volume studied (0.01 < z < 0.05) and the sensitivity of BAT across the sky (Baumgartner et al. 2013). We focus on the 144 more luminous AGN galaxies (${L}_{2-10\ \mathrm{keV}}^{\mathrm{in}}$ > 10^42.8 erg s⁻¹ or L_bol > 10⁴⁴ erg s⁻¹ ) in the sample where the survey sensitivity can detect sources in the majority of the volume surveyed and the ${V}_{\max }$ corrections are less uncertain. We also correct for the unobserved BAT AGN galaxies (39%) and the volume near the Galactic plane that was excluded (b < 10°).

The likelihood of a given galaxy hosting a luminous AGN clearly increases with gas mass (Figure 9), but the absolute fraction maxes out at 1%–10% even among the most gas-rich sources. Presumably, key factors such as the gas distribution, stochasticity, and ability to lose angular momentum play essential roles.

Zoom In Zoom Out Reset image size

4.1.2. Morphological Comparison

We next consider the molecular gas properties of the BAT AGN hosts and the xCOLD GASS inactive galaxies as a function of their morphologies (spiral, elliptical, and "uncertain"), as shown in Figure 10. For reference, we provide example images of elliptical, uncertain, and spiral host morphologies in Appendix A.

Zoom In Zoom Out Reset image size

Among spiral galaxies, the molecular gas masses and gas fractions are similar for both AGN galaxies and inactive galaxies, except at the highest stellar mass bins ($\mathrm{log}({M}_{* }/{M}_{\odot })\gt 10.8$), where AGN hosts have significantly more molecular gas (p = 0.001). For galaxies of uncertain morphologies, the difference between AGN and inactive galaxies extends across all stellar masses (p < 0.001), although the offset is not as large as the ellipticals. On the other hand, the difference in molecular gas content of BAT AGN ellipticals and xCOLD GASS ellipticals is profound across all stellar masses studied (p < 0.001). Roughly half (51%, 21/41) of the BAT AGN galaxies with elliptical morphology hosts contain significant molecular gas mass reservoirs $[\mathrm{log}({M}_{{{\rm{H}}}_{2}}/{M}_{\odot })\gt 9$], compared to almost no gas-rich ellipticals in the xCOLD GASS inactive sample (4%, 3/81).

Both the median molecular gas mass $[\mathrm{log}({M}_{{{\rm{H}}}_{2}}/{M}_{\odot })\,={8.91}_{0.17}^{+0.44}$ versus $\mathrm{log}({M}_{{{\rm{H}}}_{2}}/{M}_{\odot })\lt 8.4]$ and the gas fraction $[\mathrm{log}{f}_{{{\rm{H}}}_{2}}=-{1.71}_{0.38}^{+0.24}$ versus $\mathrm{log}{f}_{{{\rm{H}}}_{2}}=\lt 2.71]$ are significantly higher for elliptical hosts of AGN than for inactive ellipticals (p < 0.001), while the gas-depletion timescales $[\mathrm{log}({t}_{\mathrm{dep}}({{\rm{H}}}_{2})/\mathrm{yr})={8.88}_{0.12}^{+0.15}$ versus $\mathrm{log}({t}_{\mathrm{dep}}({{\rm{H}}}_{2})/\mathrm{yr})\lt 8.5]$ appear to be lower for inactive galaxies (p = 0.001), although we find this result is not significant when the upper limits are increased by 30%.

Finally, we note that if stricter criteria were used for "uncertain," the general trends would remain with small differences. The offset in molecular gas properties would be more significant for AGN galaxies in spirals compared to inactive galaxies. Conversely, the size of the offset would be somewhat smaller for AGN galaxies in ellipticals.

Here we examine the molecular gas properties of our BAT AGN galaxy sample as a function of various AGN- and SMBH-related properties, such as optical AGN classification, intrinsic X-ray luminosity, black hole mass, Eddington ratio, and line-of-sight column density. A summary of the median parameters and confidence limits from the survival analysis for these parameters can be found in Table 3.

Table 3.  Summary of Median Parameters for Molecular Gas and Host Galaxy Parameters of AGN Galaxies

	$\mathrm{log}{M}_{{\rm{H}}2}$	$\mathrm{log}{M}_{{\rm{H}}2}/{M}_{\odot }$	log tdep(H2)	log SFR	log sSFR	Δ(MS)	log M⋆
	log (M⊙)		log (yr)	log (M⊙ yr−1)	log(yr−1)		log(M⊙)
AGN Type
Sy 1	${9.35}_{0.08}^{+0.18}$	$-{1.47}_{0.07}^{+0.04}$	${9.02}_{0.05}^{+0.06}$	${0.33}_{0.11}^{+0.14}$	$-{10.45}_{0.07}^{+0.10}$	$-{0.19}_{0.12}^{+0.09}$	${10.80}_{0.03}^{+0.07}$
Sy 1.9	${9.37}_{0.07}^{+0.43}$	$-{1.26}_{0.22}^{+0.06}$	${8.98}_{0.15}^{+0.04}$	${0.54}_{0.15}^{+0.25}$	$-{10.25}_{0.10}^{+0.22}$	${0.00}_{0.18}^{+0.16}$	${10.79}_{0.08}^{+0.09}$
Sy 2	${9.36}_{0.12}^{+0.13}$	$-{1.42}_{0.23}^{+0.08}$	${8.88}_{0.07}^{+0.05}$	${0.40}_{0.16}^{+0.11}$	$-{10.32}_{0.13}^{+0.12}$	$-{0.08}_{0.19}^{+0.10}$	${10.75}_{0.07}^{+0.03}$
p < 0.05			p = 0.002
log(NH/ cm−2 )
< 20.3	${9.31}_{0.13}^{+0.21}$	$-{1.53}_{0.11}^{+0.07}$	${9.05}_{0.09}^{+0.06}$	${0.30}_{0.11}^{+0.16}$	$-{10.47}_{0.11}^{+0.09}$	$-{0.24}_{0.10}^{+0.12}$	${10.80}_{0.04}^{+0.11}$
20.3 − 22.5	${9.50}_{0.15}^{+0.17}$	$-{1.43}_{0.03}^{+0.11}$	${9.02}_{0.04}^{+0.02}$	${0.39}_{0.14}^{+0.10}$	$-{10.39}_{0.06}^{+0.15}$	$-{0.15}_{0.13}^{+0.15}$	${10.79}_{0.04}^{+0.08}$
22.5 − 23.4	${9.35}_{0.10}^{+0.16}$	$-{1.37}_{0.07}^{+0.11}$	${8.94}_{0.09}^{+0.05}$	${0.34}_{0.21}^{+0.21}$	$-{10.21}_{0.12}^{+0.09}$	$-{0.07}_{0.11}^{+0.17}$	${10.64}_{0.05}^{+0.07}$
>23.4	${9.32}_{0.16}^{+0.28}$	$-{1.59}_{0.12}^{+0.25}$	${8.73}_{0.07}^{+0.11}$	${0.53}_{0.18}^{+0.16}$	$-{10.27}_{0.25}^{+0.15}$	$-{0.06}_{0.21}^{+0.20}$	${10.81}_{0.04}^{+0.07}$
p < 0.05			p < 0.001		p = 0.034
log(${L}_{2-10\ \mathrm{keV}}^{\mathrm{in}}$/erg s−1)
<42.8	${9.28}_{0.07}^{+0.15}$	$-{1.44}_{0.08}^{+0.13}$	${8.99}_{0.05}^{+0.05}$	${0.28}_{0.13}^{+0.19}$	$-{10.32}_{0.14}^{+0.12}$	$-{0.18}_{0.10}^{+0.20}$	${10.60}_{0.06}^{+0.08}$
42.8 − 43.2	${9.34}_{0.20}^{+0.19}$	$-{1.49}_{0.08}^{+0.11}$	${8.93}_{0.09}^{+0.05}$	${0.34}_{0.21}^{+0.13}$	$-{10.40}_{0.16}^{+0.12}$	$-{0.19}_{0.17}^{+0.10}$	${10.75}_{0.06}^{+0.07}$
43.2 − 43.4	${9.50}_{0.15}^{+0.20}$	$-{1.34}_{0.23}^{+0.13}$	${8.96}_{0.12}^{+0.19}$	${0.55}_{0.19}^{+0.15}$	$-{10.18}_{0.17}^{+0.07}$	$-{0.03}_{0.10}^{+0.16}$	${10.78}_{0.02}^{+0.11}$
>43.4	${9.48}_{0.21}^{+0.18}$	$-{1.47}_{0.12}^{+0.04}$	${8.88}_{0.22}^{+0.14}$	${0.43}_{0.12}^{+0.20}$	$-{10.44}_{0.08}^{+0.15}$	$-{0.18}_{0.12}^{+0.17}$	${10.88}_{0.05}^{+0.11}$
p < 0.05							p = 0.001
log(L/LEdd)
< −2	${9.29}_{0.05}^{+0.13}$	$-{1.52}_{0.11}^{+0.19}$	${8.99}_{0.05}^{+0.01}$	${0.40}_{0.27}^{+0.11}$	$-{10.35}_{0.22}^{+0.16}$	$-{0.14}_{0.19}^{+0.20}$	${10.78}_{0.08}^{+0.09}$
−2 to −1.6	${9.35}_{0.10}^{+0.13}$	$-{1.48}_{0.20}^{+0.04}$	${8.94}_{0.03}^{+0.10}$	${0.37}_{0.17}^{+0.12}$	$-{10.42}_{0.09}^{+0.13}$	$-{0.20}_{0.11}^{+0.11}$	${10.80}_{0.05}^{+0.08}$
−1.6 to −1.1	${9.50}_{0.23}^{+0.20}$	$-{1.40}_{0.13}^{+0.15}$	${9.02}_{0.14}^{+0.06}$	${0.35}_{0.11}^{+0.18}$	$-{10.35}_{0.17}^{+0.13}$	$-{0.12}_{0.18}^{+0.07}$	${10.81}_{0.07}^{+0.07}$
> −1.1	${9.56}_{0.22}^{+0.10}$	$-{1.43}_{0.02}^{+0.11}$	${8.88}_{0.08}^{+0.07}$	${0.50}_{0.16}^{+0.18}$	$-{10.17}_{0.16}^{+0.12}$	${0.07}_{0.22}^{+0.09}$	${10.74}_{0.14}^{+0.04}$
p < 0.05	p = 0.037	p = 0.042
log (MBH/L⊙)
<7.4	${9.29}_{0.18}^{+0.24}$	$-{1.43}_{0.04}^{+0.14}$	${8.91}_{0.11}^{+0.05}$	${0.33}_{0.09}^{+0.16}$	$-{10.23}_{0.15}^{+0.12}$	$-{0.09}_{0.10}^{+0.24}$	${10.55}_{0.10}^{+0.12}$
7.4 − 7.8	${9.45}_{0.20}^{+0.18}$	$-{1.37}_{0.10}^{+0.14}$	${9.00}_{0.03}^{+0.08}$	${0.24}_{0.09}^{+0.27}$	$-{10.30}_{0.10}^{+0.15}$	$-{0.09}_{0.15}^{+0.14}$	${10.68}_{0.03}^{+0.08}$
7.8 − 8.1	${9.35}_{0.05}^{+0.16}$	$-{1.53}_{0.30}^{+0.08}$	${8.98}_{0.14}^{+0.05}$	${0.34}_{0.12}^{+0.25}$	$-{10.52}_{0.15}^{+0.12}$	$-{0.28}_{0.12}^{+0.13}$	${10.88}_{0.08}^{+0.02}$
>8.1	${9.42}_{0.15}^{+0.25}$	$-{1.47}_{0.21}^{+0.13}$	${8.93}_{0.12}^{+0.06}$	${0.55}_{0.15}^{+0.08}$	$-{10.35}_{0.22}^{+0.15}$	$-{0.06}_{0.21}^{+0.11}$	${10.91}_{0.03}^{+0.04}$
p < 0.05							p < 0.001

Note. Summary of median values for AGN host galaxies and the 1σ confidence limits based on a survival analysis using the KM estimator and the lifelines survival analysis package. A two sample survival analysis in time test is run between the two populations split in half between the highest and lowest values.

Download table as:  ASCIITypeset image

4.2.1. Optical AGN Classification

We use the public BASS/DR1 and proprietary DR2 optical spectroscopy available for our AGN to determine their optical, spectroscopic AGN classification. We classify sources that have both broad Hα and Hβ emission lines as Seyfert 1 s. This includes sources that are often referred to as Type 1.0, 1.5, 1.8 etc. Sources with broad Hα but only narrow Hβ are classified as Seyfert 1.9 s, and finally, Seyfert 2 s have only narrow Balmer lines. Our BAT AGN sample breaks down as 40% (83/207) Seyfert 1 s, 16% (34/207) Seyfert 1.9 s, and 43% (90/207) Seyfert 2 s.

Figure 11 shows the distributions of molecular gas mass, molecular gas mass fraction, the associated depletion timescales, SFRs, sSFRs, and offset from the MS. Overall, the distributions are similar. However, we do find the depletion timescales of Sy 2 galaxies are on average lower than Sy 1 galaxies (p = 0.002).

Zoom In Zoom Out Reset image size

4.2.2. Line-of-sight Column Density and Molecular Gas

We next investigate the possible links between two completely independent probes of the (cold) gas content in our BAT AGN galaxies, which in principle originate from very different scales: the line-of-sight column density, N_H—as determined from detailed spectral modeling of the X-ray SEDs (Ricci et al. 2017), and the new molecular gas measurements.

In Figure 12 we plot N_H versus molecular gas mass, gas fraction, and depletion timescales both for individual (left panels) and binned (right panels) sources.

Zoom In Zoom Out Reset image size

There are no significant trends between N_H and molecular gas mass. However, similar to the Seyfert 2 sample, a shorter depletion timescale is found for the absorbed systems (p = 0.001), which is most pronounced in the most absorbed systems (log(N_H/ cm⁻² ) > 23.4).

4.2.3. Intrinsic X-Ray Luminosity, Black Hole Mass, and Eddington Ratio

We next examine possible links between the molecular gas content of the BAT AGN hosts and key properties of the SMBHs that power their central engines. For this, we use measurements of intrinsic X-ray luminosities (${L}_{2-10\ \mathrm{keV}}^{\mathrm{in}}$; see Ricci et al. 2017), black hole mass (M_BH), and Eddington ratio (L/L_Edd; both from Koss et al. 2017), available through the BASS project.

In Figure 13 we plot ${L}_{2-10\ \mathrm{keV}}^{\mathrm{in}}$ versus the molecular gas mass, gas fraction, and depletion timescale. While there appears to be a slight increase in total molecular gas mass with X-ray luminosity, it is not statistically significant (p = 0.10) when comparing the higher luminosity half of our AGN to the lower luminosity half (above and below log(${L}_{2-10\ \mathrm{keV}}^{\mathrm{in}}$/erg s⁻¹) = 43.2, respectively). One area with a significant correlation is between the stellar mass and increased ${L}_{2-10\ \mathrm{keV}}^{\mathrm{in}}$ (p = 0.001, see Table 3).

Zoom In Zoom Out Reset image size

Figure 14 shows the molecular gas properties compared to M_BH. No significant trends are found in any of the properties (molecular gas mass, gas fraction, or depletion timescale, SFR, sSFR, Δ(MS)). There is a significant trend between stellar mass and increased black hole mass, as expected (p < 0.001).

Zoom In Zoom Out Reset image size

Finally, Figure 15 shows the molecular gas properties compared to L/L_Edd. The hosts of the highest Eddington ratio AGN (log L/L_Edd > −0.16) have molecular gas masses that are significantly higher (p = 0.037) than hosts with the lowest Eddington ratio AGN, based on survival analysis. Similarly, the highest Eddington ratio AGN have significantly higher (p = 0.042) gas fractions than the lowest L/L_Edd ones. The gas depletion timescales, SFR, sSFR, Δ(MS), and stellar mass show no statistically significant trends with Eddington ratio.

Zoom In Zoom Out Reset image size

The global Eddington ratio and gas fraction correlation, as well as the increased likelihood of a galaxy with high molecular gas to host a luminous AGN, may be tracing an important part of SMBH and host-galaxy coevolution. High gas content may lead naturally to general correlations found in the SMBH and host (e.g., the correlations between SMBH mass and bulge properties; Kormendy & Ho 2013) and the similarity between the redshift evolution of star formation and SMBH growth (see the review of Heckman & Best 2014).

The correlations found between the Eddington ratio and both the total molecular gas mass and gas fraction (over the entire host galaxy) suggest that substantial cold gas may be concentrated in the nucleus of the galaxy. There is some evidence for this from Herschel FIR imaging of AGN hosts (Mushotzky et al. 2014; Lutz et al. 2018), which showed that a significant fraction of Herschel/PACS emission originates from compact sources, with typical scales of <2 kpc. However, high-sensitivity interferometric observations of the central cores of AGN host galaxies are critical to confirm this scenario. Several ALMA campaigns are currently trying to pursue this challenging goal, including our own BASS-focused effort (T. Izumi et al. 2021, in preparation), but the samples are still rather small (<40 AGN hosts). These high-resolution studies, which can overcome some of the huge difference in scales between the entire AGN galaxy studied here and the small scales of ultimate accretion are critical.

While statistically significant global trends and/or correlations (p = 0.001 − 0.04) are found in the large sample studied here (N = 207), it is crucial to stress that there is significant variation between individual AGN hosts, most likely driven by the stochasticity of both SMBH fuelling and the emergent AGN emission, and the galaxy-scale star formation processes (e.g.,Hickox et al. 2014; Schawinski et al. 2015; Caplar & Tacchella 2019; Wang & Lilly 2019). This is similar to the result that when averaging over the full star-forming galaxy population, the average SMBH growth is correlated with SFR (e.g., Mullaney et al. 2012; Chen et al. 2013; Jones et al. 2019), as it is in the host galaxies of BAT AGN (Shimizu et al. 2017), although this correlation may be mainly linked to the host-galaxy stellar mass relation rather than star formation (Yang et al. 2017, 2018).

The ability to assess differences in star formation properties and offset from the MS complicated by systematic biases in measuring the SFR using different techniques in high stellar mass passive galaxies and AGN (and see Appendix B) and how "passive galaxies" are defined. However, the BAT AGN do have on average a higher fraction of massive spirals, bluer colors, more mergers, and more LIRGs (e.g., Koss et al. 2011a, 2013) than inactive galaxies, consistent with their higher molecular gas masses than inactive galaxies and typically higher SFRs than xCOLD GASS. We note that a previous study found that the BAT AGN predominantly lie below the MS of star-forming galaxies (but still well above passive galaxies; Shimizu et al. 2015) based on the Herschel observations, although we show that the BAT AGN galaxies are typically within the scatter of the MS. This is because there is significant difficulty establishing the low-reshift MS at high stellar masses ($\mathrm{log}({M}_{* }/{M}_{\odot })\gt 10.3$), such as studied here, in that there is potentially a flattening as well as increased scatter at high stellar masses (e.g., Popesso et al. 2019a, 2019b). Furthermore, the rarity of high-mass galaxies ($\mathrm{log}({M}_{* }/{M}_{\odot })\gt 11$) also contributes to this difficulty. This issue will be further explored in a subsequent publication (T. Shimizu et al. 2021, in preparation).

Given the significant connection between gas-rich AGN host galaxies and black hole growth, the next question is to what extent AGN are able to suppress star formation activity throughout the host galaxy, or the infall of more (cold) gas onto the host. There is no significant evidence that the most luminous X-ray AGN galaxies have depletion timescales that are significantly shorter than lower luminosity AGN galaxies. This suggests that these AGN are not the primary drivers of quenching or evolution in the M_*–SFR plane. This could be because the measured "instantaneous" luminosities are weakly related due to the much longer timescales on which average stochastic accretion occurs. Large samples with hundreds of sources, such as the one studied here, combined with studies of the AGN history on longer timescales (e.g., Sartori et al. 2018), are crucial to overcome AGN variability and understand the role of AGN outbursts and feedback. Finally, high-sensitivity interferometric observations of the central cores of AGN host galaxies would also test whether signs of AGN feedback are present.

The large amounts of molecular gas we find in BAT AGN that are hosted in elliptical galaxies is quite surprising and significant. Examples of some of the most gas-rich ellipticals in the BAT AGN galaxy sample are shown in Figure 16. We caution that some of these galaxies may be better classified as early types, as it is difficult to distinguish between true ellipticals and face-on lenticular hosts without deeper, higher quality imaging (e.g., from HST; see Appendix A)—as has been noted with Galaxy Zoo classifications (e.g., S0-SAs, Lintott et al. 2008). Importantly, ellipticals with significant molecular gas mass ($\mathrm{log}({M}_{{{\rm{H}}}_{2}}/{M}_{\odot })\gt 9$) are quite rare among inactive galaxies. Of the 81 inactive massive ($\mathrm{log}({M}_{* }/{M}_{\odot })\,\gt 10.2$) ellipticals from xCOLD GASS, only 4% (3/81) have a significant amount of molecular gas. Similarly, the fraction of early-type galaxies with ($\mathrm{log}({M}_{{{\rm{H}}}_{2}}/{M}_{\odot }\gt 9)$) within the ATLAS3D sample is even lower, about 1% (3/260; Young et al. 2011).

Zoom In Zoom Out Reset image size

In contrast, the majority (51%, or 21/41) of the elliptical hosts of BAT AGN are comparatively gas-rich (i.e., meeting the $\mathrm{log}({M}_{{{\rm{H}}}_{2}}/{M}_{\odot })\gt 9$ criterion). Interestingly, the ATLAS3D survey found a significant trend between the molecular gas content and stellar specific angular momentum. Combined with our finding of a high fraction of gas-rich ellipticals among BAT AGN hosts, this may suggest that the amount of galaxy angular momentum may be coupled to SMBH growth via stochastic accretion.

Given the lack of gas-rich early types among inactive galaxies, our results suggest that the many gas-rich early-type galaxies host luminous AGN. This is further supported by the most gas-rich early-type galaxy in xCOLD GASS (ID 3819) being excluded from the study due to hosting an AGN. Additionally, there is strong evidence that at least 2/3 early-type gas-rich galaxies in ATLAS3D host AGN based on radio data (Nyland et al. 2016). Further study is required to test this, however, because of the difference in volumes surveyed, and the rarity of both gas-rich early-type galaxies and of luminous AGN.³⁷

One obvious source of the additional gas in the massive galaxies is from mergers, which have already been shown to be significantly more common in BAT AGN hosts (e.g., Koss et al. 2010, 2011b, 2018). One might also expect an increase in the molecular-to-atomic gas ratio as measured from H i during the merger. The relatively high gas fractions of BAT AGN galaxies may explain the enhanced occurrence rate of dual AGN (Koss et al. 2012, 2018), which exceeds predictions from (low-redshift) simulations (Rosas-Guevara et al. 2016). As a specific example, looking at eight gas-rich BAT AGN galaxies with uncertain morphologies (Figure 17), 38% (3/8) are close mergers with separations between galaxy nuclei smaller than 3.5 kpc (Koss et al. 2018), whereas the rate among inactive galaxies for similarly close mergers is lower than 1%.

Zoom In Zoom Out Reset image size

These observations may be able to measure the importance of externally acquired gas (e.g., from mergers and cold accretion). This was already clearly demonstrated in some inactive early-type galaxies (e.g., in the ATLAS3D sample; Davis et al. 2011), as the specific angular momentum of the gas is dramatically different from that of the stars. Early efforts studying dozens of the nearest BAT AGN galaxies are currently underway with ALMA (T. Izumi et al. 2021, in preparation), in H i with the Australia Telescope Compact Array (I. Wong et al. 2021, in preparation), and for hundreds of BAT AGN galaxies with the Jansky Very Large Array (Smith et al. 2020).

The finding that higher column density AGN galaxies (log(N_H/ cm⁻² ) > 23.4) and Seyfert 2 AGN galaxies are associated with lower depletion timescales suggests a connection to the gaseous environment of the nucleus. The higher column density AGN are also associated with higher sSFR. Obscured AGN may prefer hosts with more molecular gas centrally concentrated in the bulge associated with the torus that may be more prone to quenching than galaxy-wide molecular gas. The relationship between X-ray line-of-sight column density, the amount of molecular gas, stellar mass, and the inclination angle of the galaxies will likely allow one to further understand the galaxy-scale distribution of dusty, molecular gas, star formation, and the associated galaxy-scale obscuration (B. Strittmatter et al. 2021, in preparation). When also combined with higher resolution data, the closer (lower redshift) subset of the BAT AGN galaxies sample may provide a unique ability to measure the amount of extended (0.1 − 1 kpc) molecular gas. This in turn will provide a template for interpreting the typical contribution of galaxy-scale contribution to the X-ray obscuration in distant AGN that is observed in deep surveys. At high redshift, the gas fractions are generally higher (Vito et al. 2014), and galaxy-scale obscuration is likely much more important for these surveys.