Remarkably high mass and high velocity dispersion of molecular gas associated with a regular, absorption-selected type-I quasar

We present 3-mm observations of the quasar J0015+1842 at z=2.63 with the NOrthern Extended Millimeter Array (NOEMA). Our data reveals molecular gas, traced via a Gaussian CO(3-2) line, with a remarkably large velocity dispersion (FWHM=1010+/-120 km/s) and corresponding to a total molecular mass MH2~(3.4-17)x10^10 Msun, depending on the adopted CO-to-H2 conversion factor alphaCO=(0.8-4.0) Msun (km/s pc^2)^-1. Assuming the 3-mm continuum emission is thermal, we derive a dust mass of the order of Mdust ~5x10^8 Msun. J0015+1842 is located in the molecular gas-rich region in the IR vs CO line luminosity diagram, in-between the main locus of main-sequence and sub-millimetre galaxies and that of most other AGNs targeted so far for CO measurements. While the large velocity dispersion of the CO line suggests a merging system, J0015+1842 is observed to be a regular, only very moderately dust-reddened (Av~0.3-0.4) type-I quasar from its UV-optical spectrum, from which we infer a mass of the super-massive black hole be around MBH~6x10^8 Msun. We suggest that J0015+1842 is observed at a galaxy evolutionary stage where a massive merger has brought significant amounts of gas towards an actively accreting super-massive black hole (quasar). While the host still contains a large amount of dust and molecular gas with high velocity dispersion, the quasar has already cleared the way towards the observer, likely through powerful outflows as recently revealed by optical observations of the same object. High angular resolution observations of this and similar systems} should help determining better the respective importance of evolution and orientation in the appearance of quasars and their host galaxies and have the potential to investigate early feedback and star-formation processes in galaxies in their quasar phases.


Introduction
There is accumulating evidence for a strong link between the evolution of massive galaxies and the super-massive black hole (SMBH) that they generally host in their centre (e.g. Heckman et al. 2004). When matter is accreted onto the disc surrounding the SMBH, enormous amounts of energy can be released through radiation (or relativistic jets), triggering the so-called galactic nuclear activity. Major mergers have been proposed as an efficient mechanism to make the matter lose most of its angular momentum and fall down to the inner regions of the galaxy in a relatively short time scale (e.g. Silk & Rees 1998;Volonteri et al. 2003;Springel et al. 2005, but see Miki et al. 2021). At the same time, merging systems are known to induce intense starformation activity through compression of the gas. In fact, starburst galaxies and luminous unobscured active nuclei (quasars) are possibly the same systems observed at different stages of the galaxy-SMBH co-evolution (Hopkins et al. 2008). In the rapid SMBH growth phase, huge amounts of dust and molecules are brought to the galaxy centre, and the SMBH activity remains heavily obscured. Bright, unobscured quasars would then correspond to a later phase, when the SMBH have almost fully assembled and are radiating close to their Eddington limit. The energy released could then be sufficient to significantly clear dust and gas from the entire galaxy through powerful winds. Such a feedback mechanism could be responsible for quenching star formation in the host (e.g. Zubovas & King 2012;Pontzen et al. 2017;Terrazas et al. 2020), although the observed large-scale outflows may not necessarily arise from propagation of energy from the accretion disc to the interstellar medium (Fabian 2012;Veilleux et al. 2017), i.e. powered by AGN activity, but could also result from the intense star-formation activity or tidal ejection during the merging phase (e.g. Puglisi et al. 2021).
Since cold molecular gas is expected to support both the star-formation and the growth of the SMBHs, large efforts have been devoted to study the properties of this phase (e.g. Omont et al. 1996;Barvainis et al. 1997;Lewis et al. 2002;Bertoldi et al. 2003;Riechers et al. 2006;Weiß et al. 2007;Wang et al. 2010;Salomé et al. 2012, among many other works). In particular, many observing campaigns aimed at detecting the CO emission lines and constrain the molecular reservoirs of galaxies to investigate the evolutionary sequence between starburst (SB) galaxies and quasars, as well as AGN feedback (either positive or negative) on star-formation in the host (e.g. Weiß et al. 2012;Nesvadba et al. 2020). The studies have mostly focused on the Article number, page 1 of 5 arXiv:2103.09542v2 [astro-ph.GA] 5 May 2021 A&A proofs: manuscript no. main bright end of the infra-red luminosity distribution and showed that luminous quasars generally have low ratio of molecular gas masses to star-formation rates (SFR) (e.g. Bischetti et al. 2021), while starburst galaxies have much larger molecular masses to SFR ratios, supporting the evolutionary paradigm. Several studies have also searched for CO emission in obscured quasars, i.e. possibly at the short-lived intermediate stage between the SB and optically-bright quasar phase (e.g. Polletta et al. 2011, Brusa et al. 2015), but differences with luminous quasars seem to appear only for Compton-thick cases, more likely associated to the initial steps of the blow-out phase (Perna et al. 2018).
On the other hand, according to the AGN unification scheme (Antonucci 1993), a given AGN appears obscured or not depending on its orientation with respect to the observer, i.e. whether the line of sight crosses large amounts of dust in the circumnuclear region or not. It is therefore not absolutely clear how much the appearance of a system depends on its evolutionary stage and its orientation. It is also possible that the disc of the host galaxy also contributes significantly to obscuring the nuclear region (e.g. Gkini et al. 2021).
Recently, Noterdaeme et al. (2019) uncovered a population of unobscured quasars in the Sloan Digital Sky Survey (SDSS York et al. 2000) featuring strong H 2 absorption lines at the quasar redshift in its optical spectrum. Detailed spectroscopic investigations of one of them, SDSS J001514.82+184212.34 (hereafter J0015+1842) with X-shooter on the Very Large Telescope, suggested that the absorption system belongs to a galactic-scale multi-phase outflow, which is also revealed by spatially resolved [O iii] and Ly-α emission (Noterdaeme et al. 2021). In this Letter, we present the detection of strong and remarkably broad CO(3-2) emission with the NOrthern Extended Millimeter Array (NOEMA) in this quasar. We discuss these exceptional characteristics for an otherwise apparently regular quasar and suggest that the orientation of this system allows us to observe early AGN feedback during or after a merger phase.

Observations
Observations were carried out with the NOrthern Extended Millimeter Array (NOEMA) and the PolyFIX correlator in the 3mm band in May and June, 2020. The CO(3-2) line at z = 2.631 is redshifted at the 95.234 GHz frequency. We observed in Dconfiguration, with 10 antennas on 23 and 29 May, and 9 antennas on June 3. On May 23 rd , the weather conditions were good and stable and we obtained 2.4 h of integration. On May 29 th , most of the data was flagged because of poor weather, leaving 0.8 h of integration. On June 3 rd , the weather was fine, providing us 4.9 h of integration time. Calibration was done using 5 sources, 3C345, 3C454.3, 0007+171, 2010+723 and MWC349. The absolute flux calibration is accurate at the 10% level. The data were calibrated with the CLIC package and mapped with the MAPPING package in the GILDAS software 1 . Using CLARK cleaning in natural weighting, the compact D configuration provided a beam of 3 . 9×3 . 7, with a PA of 69 • for the CO(3-2) line and the upper sideband continuum and 5 . 1×4 . 5, with a PA of 125 • for the lower sideband continuum. The continuum was also computed with the wider 7.7 GHz upper side band, with a beam of 3 . 9×3 . 4, with a PA of −53 • .
The quasar was observed in dual polarisation mode in 4 basebands, with 3.9 GHz total bandwidth per baseband, distributed in lower and upper sidebands distant by 15.5 GHz. The CO(3-2) line was observed in the upper side-band, and no other line was

Results
Fig. 1 presents the CO(3-2) integrated map, the velocity dispersion map, and the continuum maps in the two sidebands. The moments of the line cube have been taken above a threshold of 3σ in 120 km/s channels, corresponding to an integrated level of 0.063 Jy/beam km/s.

CO(3-2) emission
The CO(3-2) emission is clearly detected in the integrated map, but the source remains unresolved, given the large beam size of 4 , i.e. corresponding to 32 kpc at the quasar's redshift. The one moment map of Fig. 1 presents a possible East-West velocity gradient, when the outflowing gas has likely a North-South direction projected on the sky (Noterdaeme et al. 2021).
In Fig. 2, we present the 3mm spectrum extracted at the position of the quasar. We fitted the CO(3-2) line with a Gaussian function, yielding a total integrated flux F CO(3−2) = 1.1 ± 0.2 Jy km s −1 , FWHM = 1010 ± 120 km s −1 and a central velocity v = −210 ± 50 km s −1 with respect to the systemic redshift 2 . Such shift, if real, may be due to blending of CO lines in a complex system. We converted the flux to luminosity following Solomon & Vanden Bout (2005) and obtained L CO(3−2) ≈ 4.1 × 10 10 K km s −1 pc 2 . To estimate the mass of molecular gas from the CO line luminosity, we assumed a typical CO spectral line energy distribution for quasars and took the CO(3-2)/CO(1-0) intensity ratio r 31 = 0.97 from Carilli & Walter (2013), consistent with observations of AGN host galaxies at z > 2 ( Sharon et al. 2016). The CO-to-H 2 conversion factor depends on the average conditions in the molecular gas which we do not know. We hence obtain a conservative range M H 2 = α CO L CO(1−0) ≈ (3.4 − 17) × 10 10 M assuming α CO = 0.8 M (km s −1 pc 2 ) −1 (as generally adopted for quasars Walter et al. e.g. 2003;Wang et al. e.g. 2010;Bolatto et al. e.g. 2013) and α CO = 4 M (km s −1 pc 2 ) −1 (standard value).
Finally, we also note a possible structure about 8 (∼60 kpc at z = 2.631) southwards of the quasar in both the zero and one moment maps. This could be due to a gas-rich companion galaxy not seen in the optical. The corresponding extracted spectrum has F CO(3−2) = 0.18 ± 0.06 Jy km s −1 , FWHM = 240 ± 60 km s −1 and v = −115 ± 30 km s −1 . This corresponds to a molecular mass of 2.8×10 10 M , assuming the same r 31 as above and the standard α CO . Deeper observations with higher spatial resolution are necessary to better unveil the quasar and its environment.

Continuum 3 mm emission and FIR luminosity
We detect the 3 mm continuum emission at the position of the quasar on the map with a flux density, determined by fitting uv visibilities, of 65±14 µJy and 104±14 µJy for the lower (81.76 GHz or 3.67 mm) and upper (97.24 GHz or 3.08 mm) sidebands, respectively. The flux density is varying with frequency as a power-law with slope of ν α with α = 2.7, i.e. consistent with dust emission, with a dust opacity varying as ν β , with β = 0.7 albeit with a large (±1.2) statistical uncertainty 3 . Following Carniani et al. (2017, their Eq. 2), but assuming a dust temperature in the range 40-80 K and β within the range obtained above, we infer a dust mass of M dust ≈ 5 × 10 8 M , within a factor of two, assuming thermal emission only.
We constrained the AGN spectral energy distribution (SED) from the fluxes measured in SDSS and Wide-field Infrared Sur-vey Explorer (WISE, Wright et al. 2010) filters 4 , using the template by Polletta et al. (2007), only moderately reddened by dust (A V ∼ 0.3, consistent with A V = 0.4 ± 0.1 derived from the X-shooter spectrum and template, Noterdaeme et al. 2021), see Fig. 3. Based on this template, the intrinsic bolometric luminosity is found to be log L bol /L 13.4. We note that the relative contribution of the host galaxy is probably different than predicted by this template which over-predicts the 3 mm continuum emission. We fitted this emission using a modified blackbody emission law (see e.g. Eq. 2 from Rangwala et al. 2011), as expected for reprocessed cold dust emission. Using standard values for power law slope (β = 1.6, consistent with our constraint) and normalisation point of the frequency dependence of the effective optical depth (ν 0 = 1.5 THz), and assuming a dust temperature T d = 60 K, we get a total IR luminosity associated with the host log L IR /L ∼ 12.7. We caution however that this estimate is very uncertain in the absence of measurements at 100-1000 µm to constrain T d and β. Assuming T d in the range 40-80 K results in about one order of magnitude uncertainty of log L IR /L = 12.2 − 13.1. We also note that if a merging/companion galaxy contributes to this IR luminosity (Bischetti et al. 2021), then the value could be considered as an upper-limit to the IR luminosity of the quasar host alone.

Discussion and conclusions
In order to maximise the chances of detection, the vast majority of z > 1 AGNs observed so far in CO lines have been selected for being among the brightest objects, in particular at millimetre or infra-red wavelengths (e.g. Coppin et al. 2008;Wang et al. 2010;Simpson et al. 2012;Wang et al. 2013;Feruglio et al. 2014;Wang et al. 2016;Fan et al. 2019;Banerji et al. 2021). For example, WISE has played an important role in selecting these hyperluminous objects, in combination (e.g. Bischetti et al. 2017) or not (Fan et al. 2018) with photometry from the SDSS. Other works focused on possibly less luminous objects, but still preferentially dust-obscured (e.g. Polletta et al. 2011;Banerji et al. 2017;Kakkad et al. 2017;Perna et al. 2018), radio-loud (e.g. Willott et al. 2007) or on the highest redshifts detected in [C ii]λ158µm (e.g. Venemans et al. 2017). In turn, J0015+1842 is not particularly bright, with e.g. i-band magnitude in only the  Fig. 3. Fit to the spectral energy distribution of the quasar J0015+1842 using photometric points from SDSS and WISE (black squares). The solid (resp. dotted) blue line shows the reddened (resp. unreddened) quasar type I template from Polletta et al. (2007). The NOEMA continuum measurements at 3.08 and 3.67 mm are represented by the red circles with error bars, and the associated constraint on the cold dust emission component assuming T d = 60 ± 20 K is shown in green.
second brightest quartile of SDSS quasars at the same redshift, and only moderately reddened (A V ∼ 0.3 − 0.4). It has also ∼3 mag fainter WISE magnitudes than those of WISSH quasars (Bischetti et al. 2017).
In Fig. 4, we compare the width and luminosity of the detected CO line with those seen in other z > 1 AGNs from the literature. We use the compilation of unlensed AGNs by Bischetti et al. (2021) and retrieved the line widths from the original detection publications. In a few cases where several spectrally resolved components were reported, we considered conservatively the widest one. To avoid systematics related to the assumption of CO-to-H 2 conversion factor, we compare directly the CO line luminosity instead of the H 2 mass. This compilation also includes the sample by Circosta et al. (2021) who recently performed a X-ray selection of both type I and type II AGNs with a range of bolometric luminosities, independently on their mm or IR emission. This more homogeneous sample also presents the advantage of being observed all in the CO(3-2) line, so that direct comparison can be made with J0015+1842 regardless of the assumption of the excitation correction. We therefore distinguish this sample from other AGNs in Figs. 4 and 5.
Both the luminosity and the line width of J0015+1842 are remarkably high, and the width measured here, FWHM = 1010 ± 120 km s −1 is well above the vast majority of other measurements in the compiled sample, only equated by the powerful obscured quasar SWIRE J022513-043419 that has FWHM ≈ 1020 ± 110 km s −1 . Like J0015+1842, the latter object also presents extended [O iii] emission that may trace outflowing gas .
We compare the CO content of J0015+1842 with other AGNs as a function of the IR luminosity in Fig. 5, using the same compilation as previously, but restricted to systems for which Bischetti et al. (2021) consider the IR measurement to be reasonably reliable (their section 6.2). We also add sub-millimetre galaxies (SMGs) to the comparison. The IR luminosity, integrated over the range 8-1000 µm, is generally considered as a 9.0 9.5 10.0 10.5 11.0 11.5 log(L CO(1 0) /(K kms 1 pc 2 )) 0 200 400 600 800 1000 1200 FWHM(CO) (km s 1 ) AGNs (compil. by Bischetti et al. 2021) X-ray AGNs -BL X-ray AGNs -NL J0015+1842 Fig. 4. Full-width at half maximum versus luminosity of the CO emission line in J0015+1842 (red star) compared to that in z > 1 active galactic nuclei (compiled by Bischetti et al. (2021), with X-ray selected broad-line (BL) and narrow-line (NL) AGNs from Circosta et al. (2021) shown by orange triangles). Measurements using CO(3-2) are shown with darker symbols than those from other CO lines.
proxy of dust emission related to star formation in the host, once the AGN contribution has been removed. The IR-to-CO luminosity ratio is then a widely used proxy for star-formation efficiency. Remarkably, the high CO-line luminosity of J0015+1842 is similar to that of the most luminous AGNs in IR. However, with its likely lower cold dust emission, J0015+1842 is located closer to the region populated by MS galaxies and SMGs, and with molecular content well above that of most Xray selected AGNs. The only AGN in the sample of Circosta et al. (2021) with similar CO luminosity is cid_1253 (COSMOS J100130.56+021842.6) with log(L CO(3−2) /(K km s −1 pc −2 )) = 10.80±0.04 and FWHM(CO) = 810±93 km s −1 . This object is a merger, hosting a narrow-line (type II) AGN, while J0015+1842 is a regular quasar with broad emission lines (type I).
According to the evolutionary scheme, major mergers of gasrich galaxies trigger starburst activity and bring a significant amount of matter towards the nuclear regions, initiating the build up of the supermassive black hole. These objects would mostly appear first as obscured AGNs, with star-formation efficiencies much higher than in main sequence galaxies, then transition to optically bright systems with low gas fractions. Indeed, the presence of powerful multiphase galactic-scale outflows show that AGNs are capable of removing large amounts of gas from their host galaxies. As recently discussed by Perna et al. (2018), neither the evolutionary sequence nor orientation effects with respect to obscuring medium alone are able to explain the observed SFEs in obscured and unobscured AGNs.
J0015+1842 is bringing an important piece to this puzzle. The exceptionally broad Gaussian profile of the CO line in J0015+1842 suggests a recent or ongoing merger, with a very significant amount of molecular gas and dust. However, the line of sight to the active nucleus is only very moderately reddened and the presence of broad emission lines corresponds to a reg- AGNs (compil. by Bischetti et al. 2021) SMGs X-ray AGNs -BL X-ray AGNs -NL J0015+1842 Fig. 5. CO line luminosity against the infra-red luminosity of the host integrated in the 8-1000 µm range. Unfilled symbols corresponding to CO upper limits in case of non-detection and green crosses to submillimetre galaxies. Other symbols are as per Fig. 4. The dashed line corresponds to the relation for main-sequence galaxies (Sargent et al. 2014). ular type-I quasar powered by a highly accreting super-massive black hole with Eddington ratio 5 λ Edd = L bol /L Edd 1. From the analysis of ionised emission lines together with absorption from H 2 , Noterdaeme et al. (2021) suggest that a multiphase outflow is observed oriented almost towards the observer. This could provide a natural explanation to the low extinction along the line of sight. J0015+1842 supports then a picture in which feedback processes can start early in the evolutionary sequence, with outflows clearing the view towards the nuclear region at least in some directions, while a large amount of molecular gas is still available in the host. In addition, although requiring confirmation, the star-formation efficiency may not have yet reached a level as high as seen in other, possibly more evolved, quasars. Measuring the flux density at ∼1 mm would help measuring better the IR luminosity and host SFR of J0015+1842 and similar sources.
If, as suggested by Noterdaeme et al. (2021), the presence of proximate H 2 absorbers and leaking Ly-α emission provide an efficient way to identify multi-phase outflows in regular quasars, then observations of CO emission in a sample of them could bring further clues to constrain the relative importance of orientation and evolutionary sequence in the appearance of quasars. Deep observations at high angular resolution should allow one to confirm or not the derived configurations and enable detailed investigation of early feedback mechanisms in more regular quasars than those usually targeted for millimetre studies.