SMM J04135+10277: A distant QSO-starburst system caught by ALMA

The gas content of galaxies is a key factor for their growth, starting from star formation and black hole accretion to galaxy mergers. Thus, characterising its properties via observations of tracers like the CO emission line is of big importance in order to understand the bigger picture of galaxy evolution. We present Atacama Large Millimeter/submillimeter Array (ALMA) observations of dust continuum, CO(5-4) and CO(8-7) line emission in the quasar--star-forming companion system SMM J04135+10277 (z=2.84). Earlier low-$J$ CO studies of this system found a huge molecular gas reservoir associated to the companion galaxy, while the quasar appeared gas-poor. Our CO observations revealed that the host galaxy of the quasar is also gas-rich, with an estimated molecular gas mass of $\sim(0.7-2.3)\times10^{10}$ M$_{\odot}$. The CO line profiles of the companion galaxy are broad ($\sim1000$ km s$^{-1}$), and show signs of rotation of a compact, massive system. In contrast to previous far-infrared observations, we resolve the continuum emission and detect both sources, with the companion galaxy dominating the dust continuum and the quasar having a $\sim25\%$ contribution to the total dust emission. By fitting the infrared spectral energy distribution of the sources with \textsc{MR-MOOSE} and empirical templates, the infrared luminosities of the quasar and the companion are in the range of $L_{\rm IR, QSO}\sim(2.1-9.6)\times10^{12}$ L$_{\odot}$ and $L_{\rm IR, Comp.}\sim(2.4-24)\times10^{12}$ L$_{\odot}$, while the estimated star formation rates are $\sim210-960$ M$_{\odot}$ yr$^{-1}$ and $\sim240-2400$ M$_{\odot}$ yr$^{-1}$, respectively. Our results demonstrate that non-detection of low-$J$ CO transition lines in similar sources does not necessarily imply the absence of massive molecular gas reservoir but that the excitation conditions favour the excitation of high-$J$ transitions.


INTRODUCTION
An intense phase of star formation and supermassive black hole (SMBH) growth at high-redshift is necessary to explain the observed correlation found in local elliptical galaxies between the mass of the SMBH and some properties of the host galaxy, such as the velocity dispersion or bulge mass (e.g. Magorrian et al. 1998;Gebhardt et al. 2000;Merritt & Ferrarese 2001;Tremaine et al. 2002;Marconi & Hunt 2003;Häring & Rix 2004;McConnell & Ma 2013). The found correlations suggest that the evolution of the SMBHs and E-mail: judit.fogasy@chalmers.se their host galaxies is tied together. Theoretical studies of galaxy evolution provide a possible framework for this coevolution in form of interactions and major merger events between massive and gas-rich galaxies (e.g. Sanders et al. 1988;Di Matteo et al. 2005, 2008Hopkins et al. 2008Hopkins et al. , 2006Narayanan et al. 2010). As the star formation and black hole activity peaks at around z ∼ 2 − 3 (e.g. Madau et al. 1998;Hopkins 2004;Richards et al. 2006;Zheng et al. 2009;Aird et al. 2010Aird et al. , 2015, by probing the far-infrared (FIR) properties and molecular gas content of high-z active galactic nuclei (AGNs) and their close environment, we can probe this scenario. Moreover, there is an extensive literature with ambigous conclusions on whether there is a strong correlation between the star formation rate (SFR) and AGN luminosity of high-redshift AGNs (e.g. Schweitzer et al. 2006;Lutz et al. 2010;Bonfield et al. 2011;Mullaney et al. 2012;Santini et al. 2012;Stanley et al. 2015;Bernhard et al. 2016;Dong & Wu 2016;Netzer et al. 2016;Pitchford et al. 2016;Duras et al. 2017;Ramasawmy et al. 2019;Schulze et al. 2019). Based on these studies the growth of AGNs can follow two different paths: low or moderate luminosity AGNs can evolve through secular processes, thus the SMBH growth is not directly linked to the star formation of the host galaxy; while the most luminous AGNs grow through major mergers which enhance the star formation of the hosts.
Indeed, in the last decade several quasar studies at submm wavelength found signs of intense star formation in their hosts by tracing dust continnum emission and [C ii] cooling line emission (e.g. Carilli et al. 2001;Omont et al. 2001;Maiolino et al. 2005;Wang et al. 2013;Venemans et al. 2016). However, most of these studies are limited in resolution and/or sensitivity, thus they could miss the detection of close companion galaxies, which also have a contribution to the submm emission.
Thanks to the high angular resolution and sensitivity of the Atacama Large Millimeter/submillimeter Array (ALMA), many recent studies of high-z quasars found companion galaxies in their vicinity, in some cases more than one (e.g. Wagg et al. 2012;Silva et al. 2015;Trakhtenbrot et al. 2017;Banerji et al. 2017Banerji et al. , 2018Carniani et al. 2017;Decarli et al. 2017;Fan et al. 2018). These results are in agreement with our earlier study, where we used galform Lagos et al. 2014), a galaxy formation and evolution model with the aim of investigating the expected frequency of finding quasarcompanion galaxy systems (Fogasy et al. 2017).
In this paper we focus on the quasar-star-forming companion galaxy system SMM J04135+10277 (z = 2.84). The AGN component of the system is a type-1 quasar, originaly discovered at submm wavelength (Knudsen et al. 2003) using the Submillimetre Common-User Bolometer Array (SCUBA). Subsequent single-dish observations of the quasar tracing low-J CO transitions revealed a massive molecular gas reservoir associated with the AGN (M H 2 ∼ 10 11 M ; Hainline et al. 2004;Riechers et al. 2011b). However, followup interferometric observations showed that the gas reservoir was offset from the position of the quasar by ∼ 5 (41.5 kpc) and associated with a close companion galaxy (Riechers 2013). Both the quasar and its companion are slightly gravitationally lensed by the foreground galaxy cluster Abell 478, by a gravitational magnification factor of µ QSO = 1.3 and µ Comp. = 1.6 ± 0.5, respectively (Knudsen et al. 2003;Riechers 2013). Given the molecular gas distribution and the separation between the sources, the system has been proposed to be in an early stage of a wet-dry merger event, the companion being gas-rich and the quasar being gas-poor, making this case unique (Riechers 2013). Based on the spectral energy distribution (SED) of the companion galaxy we can infer that it is a heavily dust-obscured star-forming galaxy (A V ∼ 2.8 mag), with a star formation rate of ∼ 700 M yr −1 and dust mass of ∼ 5 × 10 9 M (Fogasy et al. 2017). Considering the high SFR, molecular gas and dust mass of the companion galaxy, it is possible that the companion has a significant contribution to the SCUBA detected submm emission and the majority of star formation happens in the companion rather than in the host galaxy of the quasar. In order to further investigate this system we obtained dust continuum, CO(5-4) and CO(8-7) observations with ALMA. Our goal is to constrain the excitation properties of SMM J04135+10277 through modelling of the spectral line energy distribution (SLED) of both sources and to deepen our understanding of the processes shaping the evolution of this system. This is important because prior to ALMA, the majority of studies finding AGN-companion systems were focusing on the large scale environment of AGNs and did not study the systems in details.
In section 2 we present the ALMA observations, in section 3 we describe the results of the observations and the data analysis, including the SLED modelling of the sources. In section 4 we discuss our findings and compare them with other studies found in the literature. Finally, we summarise our results in section 5. Throughout this paper we adopt WMAP7 cosmology with H 0 = 70.4 km s −1 Mpc −1 , Ω m = 0.272 and Ω Λ = 0.728 (Komatsu et al. 2011).

OBSERVATIONS
SMM J04135+10277 was observed with the 12-m ALMA array during cycle 3 (project code 2015.1.00661.S; P.I.: Fogasy). One of the spectral windows with a bandwidth of 1.875 GHz was tuned to the central frequency of 149.843 GHz to cover the redshifted CO(5-4) line in band 4, and to 239.648 GHz to cover the redshifted CO(8-7) line in band 6. The other three spectral windows of each band, with a bandwidth of 2 GHz each, were used to observe the continuum. Other details of the observations are given in Table  1.
The calibration of the data was done using the ALMA Science Pipeline. The calibration process includes standard calibration and reduction steps, such as flagging, bandpass calibration, flux and gain calibration. The quasar J0423-0120 was used as a bandpass calibrator source. Due to the strong variability of the flux calibrator source at the time of the observations (∼ 10% in both bands), the flux calibration was redone with updated flux values. After this correction we assume a conservative 10% absolute flux calibration error, which is not included in the flux densities reported in this paper.
To image the data we used the Common Astronomy Software Applications (CASA; McMullin et al. 2007) with the ARC Node provided scripts. We applied the CLEAN algorithm to create the continuum images and line data cubes using natural weighting. In case of the spectral line data the continuum was fitted using the line free channels of every spectral window and was subtracted in the uv-plane using the uvcontsub task in CASA. For each band the continuum subtracted line data was imaged with 77 km s −1 channel width.

Dust continuum emission, FIR luminosities and star formation rates
The ALMA observations enabled us to resolve the far-IR emission of SMM J04135+10277 and detect both sources for the first time ( Fig.1), in contrast to previous singledish submm observations, where the quasar and its companion remained unresolved within a single beam (e.g. SCUBA beam is 15 at 850 µm, the separation between the quasar and the companion is ∼ 5 ). Thanks to the high-sensitivity of ALMA we achieved a robust detection, yielding a high signal-to-noise ratio of 14 and 46 in the case of the quasar and 44 and 163 in the case of the companion galaxy, in band 4 and band 6, respectively. The dust emission of the quasar is much weaker compared to the companion galaxy, only about 22 − 25% of the total emission in both bands (Table 2). In band 4 the cD galaxy of A478 is detected northwest from the phase centre (R.A. = 04 h 13 m 25 s .27, Dec. = +10 • 27 54 .69), at ∼ 33 (∼ 264 kpc) distance from the quasar but it is not detected in the band 6 data.
The UV-to-IR SED of the companion galaxy was constructed in Fogasy et al. (2017), here we focus on the FIR SED of the quasar and the companion. To complement our ALMA observations we used previous SCUBA and ArTeMiS data to obtain the SED (Knudsen et al. 2003;Fogasy et al. 2017). As the resolution of the single-dish data is not sufficient to separate emission associated to the quasar and the companion galaxy, we use the multi-wavelength SED fitting code MR-MOOSE (Drouart & Falkendal 2018). The code is designed to treat upper limits consistently as well as large variations of resolution and to fit blended sources using a Bayesian framework, by allowing the user to use the spatial information. MR-MOOSE simultaneously fits the SEDs of the quasar and the companion, and the total emission when it is impossible to resolve sources individually, e.g. in the SCUBA bands. The best fit of models and the possible range of solutions are shown on Figure 2. In the case of the quasar the dust emissivity and dust temperature of the best-  fit models are β dust,QSO = 2.5 ± 0.3 and T dust,QSO = 24 ± 6 K. In the case of the companion the best-fit models predicts β dust,Comp. = 3.3 ± 0.4 and T dust,Comp. = 16 ± 3 K. Considering the small number of available data points, it is difficult to properly constrain the absolute values of β and T dust , but the MR-MOOSE result still highlights that the quasar has a consistently higher temperature compared to the companion.
The FIR luminosities of the quasar and its companion based on the best-fit MR-MOOSE SED models are L FIR,QSO = (2.1 +1.7 −1.2 ) × 10 12 L and L FIR,Comp. = (2.4 ± 1) × 10 12 L , corrected for gravitational lensing magnification. From the obtained infrared luminosities, the SFR of the sources can be estimated according to the relation SFR ∼ δ MF × 10 −10 L IR , assuming that the dust is heated by star formation. As MR-MOOSE only fits the FIR part of the SED, the total IR luminosity of the sources could be higher. Thus we treat the SFR calculated from the L FIR as a lower limit. By adopting a Chabrier initial mass function (δ M F ∼ 1), we get a SFR of ∼ 210 M yr −1 . On all panels the symbols represent the same observations. The black diamonds represent the ALMA observations; the black unfilled diamonds show the total ALMA emission of the quasar and the companion. The black circles show the total SCUBA 450 µm and 850 µm emission; the black unfilled circles represent the SCUBA observations, scaled by the assumed relative contribution of the quasar (25% ) and companion (75% ) to the total emission. The upper limit shows the 3σ ArTeMiS measurement of the system. and ∼ 240 M yr −1 for the quasar and the companion, respectively.
In addition, we used the empirical SED model grid of Chary & Elbaz (2001) as an alternative approach to fit the IR SED of quasar and the companion. The grid includes 105 templates with a wide range of infrared luminosities. As the SED templates have no free parameters, we chose the templates that fit our ALMA detections the best. For the SCUBA bands we assumed a 25% contribution from the quasar, based on the ALMA detections. The best-fit templates are shown on Figure 2 together with the MR-MOOSE models. Based on the best-fit Chary & Elbaz (2001) templates the IR luminosities of the quasar and the companion are L IR,QSO ∼ 9.6 × 10 12 L and L IR,Comp. ∼ 2.4 × 10 13 L (the luminosities are corrected for the gravitational lensing magnification). The SFR of the quasar and its companion according the same relation as above is ∼ 960 M yr −1 and ∼ 2400 M yr −1 , respectively.
By comparing the results of the two approaches, the Chary & Elbaz (2001) templates cannot fit well the data unless the quasar's contribution at the 450 µm SCUBA wavelength is higher ( ∼ 35%). Furthemore, the Chary & Elbaz (2001) templates are constructed for low-redshift starforming galaxies and might not be suitable for high-z AGNs. As MR-MOOSE only fits a modified-blackbody SED to the FIR part of the SED and the Chary & Elbaz (2001) templates cover the whole luminosity, the results of these two approaches can be treated as lower and upper limits for the total IR luminosity and the SFR of the sources.

CO(5-4) and CO(8-7) line emission
We detect CO(5-4) and CO(8-7) line emission from both the quasar and the companion galaxy (Fig. 3, 4). As it was expected from earlier CO(1-0) and CO(3-2) observations, both the CO(5-4) and CO(8-7) emission is dominated by the companion galaxy. However, while the quasar remained undetected at low-J CO transitions, it appears bright in our high-J transition observations. This indicates that the host galaxy of the quasar has a significant amount of warm, highly-excited molecular gas.
The line emitting regions of both sources are compact and in the case of the companion, we do not see the elongated shape previously found by Riechers (2013). For further inspection, the spectra of the quasar and companion were extracted from the line data cube. We fitted the spectra using Gaussian line profiles in order to determine the line widths and velocity integrated fluxes of the sources.
The CO line profiles of the quasar are narrow and shifted by ∼ 700 km s −1 with respect to the redshift of the quasar derived from rest-frame ultraviolet emission lines (z UV = 2.837 ± 0.003; Knudsen et al. 2003). This is in agreement with previous studies of high-redshift quasars, which found that the broad UV emission lines of quasars tend to be blueshifted by several hundreds of km s −1 compared to emission lines probing the interstellar medium of the host galaxies, such as the [C ii] and CO lines (e.g. Riechers 2011;Willott et al. 2015;Venemans et al. 2016;Trakhtenbrot et al. 2017;Decarli et al. 2018).
In contrast, the CO(5-4) and CO(8-7) line profiles of the companion are very broad, 1127 ± 49 km s −1 and 1038 ± 110 km s −1 respectively, when fitting a single Gaussian profile. There is a descrepancy between the line widths of the different CO transitions: the CO(1-0) and CO(3-2) lines are narrower (679 ± 120 km s −1 and 765 ± 222 km s −1 ; Riechers 2013; Sharon et al. 2016) compared to the higher-J transition lines reported here. The fitted CO redshifts of the companion shown in Table 3 are consistent within the un-   Knudsen et al. 2003). The top-axis shows the relative velocity offset with respect to the fitted redshifts.
certainties of the CO(1-0) redshift reported in Sharon et al. (2016) and slightly lower compared to the CO(3-2) redshift presented in Riechers (2013), which we attribute to the much higher signal-to-noise ratio of our measurements.

CO kinematic properties of the companion
For the companion galaxy, the ALMA observations reveal that the emission of the CO(5-4) and CO(8-7) line is spatially extended over a few beams, though less than suggested in Riechers (2013). We analysed the intensity weighted velocity field map (moment-1 map), the velocity dispersion map (moment-2 ) and the position-velocity (PV) diagram of the source to characterise the kinematic properties and understand the discrepancy between the linewidth of the low-J and high-J data (Fig. 5). For the PV diagram we chose the angle towards the highest velocity gradient (−30 • ) and used the same angle for both transitions. The position and angle from which the PV diagram was derived is indicated as a blue line on Fig. 5. The velocity map and the PV diagram of the CO(8-7) emission line are less well-defined compared to that of the CO(5-4) line, which might be the result of the lower signal-to-noise ratio achieved in band 6. Both the velocity field maps and the PV diagrams of the high-J transition lines show that the emission is consistent with rotation of a massive and compact system with maximal velocity values of |V max sin i| 300 km s −1 and show no clear signs of outflows, despite the very broad linewidth. The velocity dispersion maps of the companion show very high dispersion across the source. A common characteristics of rotating systems is an increase of the velocity dispersion towards the centre. The resolution of the natural weighted image is insufficient to show this, and beam smearing could cancel out this feature. Thus a proper analysis of the kinematics is not possible given that several resolution elements would be needed on either side of a potential rotation curve (e.g. Epinat et al. 2010;Di Teodoro & Fraternali 2015). However, in Appendix A we present a simple kinematic model with rotation using GalPak 3D (Bouché et al. 2015).
In contrast to the companion galaxy, the quasar host  galaxy does not display a complex line profile and signs of a velocity gradient, hence we do not carry out a similar analysis for this source. In Appendix B we present the moment-1 and moment-2 maps of the quasar.

CO line luminosities and gas masses
The CO(5-4) and CO(8-7) line luminosities are derived using the following equation from Solomon et al. (1997): where S CO dV is the velocity integrated flux, ν obs is the observed frequency of the line in GHz and D L is the luminosity distance in Mpc. The derived luminosities of the sources are corrected for the gravitational lensing and are summarised in Table 3.
To estimate the molecular gas mass of the system, we use the CO(5-4) line observations together with the CO(1-0) line observations of Sharon et al. (2016). We take this approach as the molecular gas mass is best estimated from the ground transition of CO and for the companion interferometric CO(1-0) detection is available, while the quasar is only detected in our higher-J ALMA observations. In the further analysis we adopt a conversion factor of α CO = 0.8 M (K km s −1 pc 2 ) −1 (Downes & Solomon 1998), which is typically used for starburst galaxies.
In the case of the companion Sharon et al. (2016) reported a line luminosity of L CO(1−0) = (8.6 ± 1.7) × 10 10 K km s −1 pc 2 , which converts to a molecular gas mass of (6.8 ± 1.4) × 10 10 M . This is about a factor of two lower than the molecular gas mass determined by Riechers (2013), and can be a result of the different resolution of the observations. Regardless of this difference, it is clear that the companion has a massive gas reservoir.
The quasar was not detected by Sharon et al. (2016) but a 3σ upper limit is obtained, yielding a CO(1-0) line luminosity upper limit of 1.9 × 10 10 K km s −1 pc 2 , which converts to a molecular gas mass of 7.4 × 10 9 M . We also calculated the molecular gas mass from the CO(5-4) emission of the quasar, as this is the lowest-J detection available for this source. We assume that the CO(5-4) transition is thermally excited, based on CO(4-3) observations of high-z type  Table 3. Velocity integrated fluxes, line widths and line luminosities of the system of SMM J04135+10277.

RADEX modelling of the CO spectra
To put constraints on the excitation properties of SMM J04135+10277, we modelled the SLED of both sources using our ALMA detections combined with CO(1-0) and CO(3-2) line observations from the literature (Riechers 2013;Sharon et al. 2016). We used the non-LTE molecular radiative trans-fer code RADEX (van der Tak et al. 2007) to calculate the line intensities of CO from transitions J up = 1 to J up = 8, within a wide range of physical parameters: kinetic temperature (T kin ), number density of molecular hydrogen (n H 2 ), and column density of CO (N CO ). During each run we set the background radiation temperature to the value of the cosmic microwave background at z = 2.84, adopted spherical geometry and assumed that the only collisional partner of CO is molecular hydrogen. As the linewidths of the lowand high-J CO lines are different, we use an average value during the RADEX modelling. The RADEX code only considers photodissociation regions and does not include mechanical and X-ray heating. We compared our observations with the models using χ 2 -analysis. In the first run we varied all three physical parameters in case of each source but as most of the models converged to the same CO column density, we fixed N CO and continued with fitting only T kin and n H 2 . During the RADEX modelling of the quasar we set the line width of 300 km s −1 , fixed N CO to 10 16 cm −2 and run RADEX on a grid of T kin = [10 − 400] K and log(n H 2 ) = [2 − 7] cm −3 . It is important to note, that as we only have two detections for the quasar and two upper limits, we do not expect to arrive to a straightforward conclusion about the excitation properties of this source. However, by using the available data we can narrow down the possible temperature and density ranges. Figure 6 shows how the χ 2 value of the models vary as a function of temperature and density. The upper-J transitions of the quasar can be fitted by models with a continous range of kinetic temperatures of 60 − 400 K and molecular hydrogen density between 10 4 − 10 6.2 cm −3 . Models with the lowest χ 2 values are outside of the typical temperature range for quasars (T kin = 40 − 60 K; Riechers et al. 2006Riechers et al. , 2009 and the corresponding densities are much higher compared to the typical quasar value (10 3.6 − 10 4.3 cm −3 ). This might be due to the lack of data points to produce a reasonable fit but it could also suggest the presence of an extremly compact region of highly excited gas close to the central AGN. In contrast, there are models with low χ 2 value and densities within the typical range for quasars but their temperature is not well-constrained. In order to illustrate how the RADEX models fit the observations, on Figure 7 we plot three model SLEDs with low χ 2 value but with different temperatures and densities. In addition, we calculated the molecular gas mass of the quasar using the CO(1-0) line intensity of the the best-fit models, which gives a mass range of (1.2 − 1.8) × 10 9 M . The RADEX estimated gas mass is lower compared to the gas mass derived from the CO(5-4) transition and the CO(1-0) upper limit.
For the modelling of the companion galaxy we used a line width of 850 km s −1 , fixed N CO to 10 17 cm −2 and run RADEX on a grid of T kin = [10 − 400] K and log(n H 2 ) = [2 − 7] cm −3 . From the χ 2 -analysis we found that models with low χ 2 value give a consistent estimate for the molecular hydrogen density, between 10 3 − 10 4 cm −3 , but can have temperatures ranging from 110 K to 400 K (Fig.6). While the temperatures of the best-fit models are not well constrained, they could narrow down the possible molecular hydrogen density range of the companion, and the yielded range is within the typical values found for SMGs (10 2.7 − 10 3.5 cm −3 ; Carilli & Walter 2013). Using the CO(1-0) line intensity of the best-fit models, the molecular gas mass of the companion is (2.9 − 4.5) × 10 10 M , which is fairly close to the gas mass derived from the CO(1-0) transition. On Figure 7 we show three models SLEDs which have low χ 2 value, but have very different kinetic temperatures. Based on Figure 7 it is clear that even the best-fit RADEX models cannot fit well the data and the temperature cannot be simply constrained.

FIR emission of SMM J04135+10277
Thanks to the high-resolution of ALMA we were able the resolve the system of SMM J04135 and detect FIR emission from both the quasar and its companion. However, fitting the FIR SED of the sources is still challenging, as the ALMA observations only trace emission at the Rayleigh-Jeans side of the FIR SED and adequate observations around the peak of the FIR SED are lacking. It has been shown that the wavelength of the rest-frame SED peak can be used as a proxy for dust temperature and it decreases with increasing total infrared luminosity but it is dependent on the emissivity and opacity of the used model (Lee et al. 2013). Furthermore, the dust temperature increases with infrared luminosity (Casey et al. 2012). Thus, the infrared luminosities and SFRs of the quasar and the companion estimated from the MR-MOOSE models can be considered as lower limits, and a better sampling around the peak of the SEDs is required to get the whole picture.
Another important note about the FIR SED of the quasar is the possible contribution of the AGN to the total FIR emission. Several studies addressed this question with contradicting results and this topic is still highly debated(e.g. Schweitzer et al. 2006;Leipski et al. 2014;Schneider et al. 2015;Lutz et al. 2016;Symeonidis 2017). What can be assumed based on these studies is that the most luminous AGNs can have a significant contribution to IR emission, mostly at near-and mid-IR wavelength originating from the dusty torus component of AGNs. The main obstacle, which needs to be eliminated is the resolution factor, as many studies use large-beam Herschel observations, thus potentially detect dust emission from blended sources. Since in our case the dust emission of SMM J04135 is only resolved in the ALMA bands, it is very difficult to estimate the AGN contribution at lower wavelength, in the SCUBA bands in particular.

Rotation of a massive system
The observed CO(5-4) and CO(8-7) lines of the companion are very broad, with about a 1000 km s −1 FWHM in each band. These values are much higher compared to lower-J transitions reported in the literature (Riechers 2013;Sharon et al. 2016). Looking at other high-z sources, only a few submm galaxies and quasars have been reported to have such broad CO lines (Genzel et al. 2003;Neri et al. 2003;Greve et al. 2005;Coppin et al. 2008;Polletta et al. 2011;Frayer et al. 2018), making the companion galaxy special.
In order to find the reason of observing such a broad line, we obtained the kinematics of the companion galaxy through velocity maps, velocity dispersion maps and PV diagrams. Based on our analysis we found no features characteristic to molecular outflows and the velocity map of the companion suggests rotation of a massive and compact galaxy with maximal velocity values of |V max sin i| 300 km s −1 . The velocity dispersion map shows an almost uniform dispersion all over the companion galaxy, which could be an indication of spatially unresolved rotation (Gnerucci et al. 2011).
As we did not find signs of outflows, we consider two possible explanations for the origin of the broad CO lines, which can happen separately or at the same time. The broad line could be the result of not having sufficient resolution to map the companion galaxy and observe the velocity gradients both at large and small scales. Beam smearing has been known as one of the main effects causing broad lines and increasing the observed dispersion of galaxies (e.g. Teuben 2002;Newman et al. 2013;De Breuck et al. 2014;Harborne et al. 2019). At high-z this effect can be even more prominent. As the ALMA beam sizes of our observations are comparable of the CO size of the companion galaxy, this effect is a plausible explanation for observing such broad lines.
Even if the broad line is consistent with rotation, another reason why the observed CO lines are broad could be turbulence, as it has been found in case of recent studies of high-z galaxies (e.g. Gnerucci et al. 2011;Williams et al. 2014). Turbulence could arise from several components, such as having a hidden AGN or from a late stage merger event in the companion galaxy, where the merging galaxies are in coalescence. Given the high SFR of the companion galaxy, the supernova rate might be also high, thus the kinematic energy released by supernovae could also account for part of the turbulence. To investigate these scenarios, higher resolution observation are required, preferably tracing even higher-J transitions as well.

CO excitation properties of SMM J04135+10277
The detection of CO(5-4) and CO(8-7) line emission in both sources complemented with low-J CO observations provides the opportunity to infer the excitation properties of this system. However, there are still some limitations to get the whole picture about the excitation conditions, such as having too few data points to fit the SLED. In case of the quasar we only have detections in the ALMA bands and upper limits for the CO(1-0) and CO(3-2) emission. This of course affects the χ 2 -analysis and thus the RADEX models with low χ 2 value yield very different solutions.
In addition to lacking enough data points, another reason we cannot fit well the observations might be the model itself. By using a model with a more complex radiation field and geometry other than spherical could change the outcome and better fit the data. It has been shown that high-J transitions starting from J = 7 are better fitted with models including X-ray dominated regions (XDR; Meijerink et al. 2007). Such models could be especially relevant for the quasar, where the central compact region maybe better modelled by an XDR, while the more diffuse gas associated to star formation is better modelled by a PDR (Meijerink & Spaans 2005). However, models including XDRs have more fitted parameters than the RADEX model used in this paper, and thus could be done provided the SLED was better sampled including more data for J > 8 transitions.
The detection of highly excited molecular gas in the host galaxy of the quasar shows that one has to be careful when interpreting non-detections of low-J CO transitions in such sources. For example, Kakkad et al. (2017) presented CO(2-1) observations of z ∼ 1.5 quasars, with a detection rate of only 30%. Based on our results we note that the low detection rate could be related to the compactness of the CO emitting regions and also because the excitation conditions favour the excitation of high-J transitions, rather than the absence of a massive molecular gas reservoir.
The situation is not less complicated when we look at the SLED of the companion galaxy. While we have four CO detections, it is clear that the companion cannot be fitted with a single phase component. In case of SMGs, it has been already demonstrated that their SLED is often best modelled with a combination of an extended, diffuse component and a compact component with higher excitation (Riechers et al. 2011a;Harris et al. 2010;Hodge et al. 2013). However, by fitting two separate components to the low-and high-J CO transitions we encounter the same problem, such as having two observations and two fitted parameters. Therefore, we cannot infer both the temperature and density of the sources.

Quasar-star-forming companion systems
In recent years, many high-redshift quasars have been observed with ALMA taking advantage of its sensitivity and high-resolution.While some of these studies, tracing dust and molecular gas emission, found vigorous star formation and huge gas reservoirs in the host galaxies of the quasars (e.g. Wang et al. 2013;Banerji et al. 2017), the number of detected quasar-star-forming companion galaxy systems has also been growing (e.g. Banerji et al. 2017;Carniani et al. 2017;Decarli et al. 2017;Trakhtenbrot et al. 2017). This is in agreement with our simulation results reported in Fogasy et al. (2017), where we used GALFORM Lagos et al. 2014), a galaxy formation and evolution model to investigate the expected frequency of finding systems similar to SMM J04135+10277. According to the simulations, at a distance of < 350 kpc, 33% of the simulated quasar sample have a companion galaxy (M > 10 8 M ) and 2.4% have bright companions with a SFR > 100 M yr −1 . However, the system of SMM J04135+10277 seems to differ from the other detected quasar-companion pairs, as the dust emission and the SFR is dominated by the companion 1 . This could be due to the different redshifts of the studied sources, but it might indicate that the companion galaxy of SMM J04135+10277 harbours a hidden AGN, which also has a contribution to the FIR emission. SMGs harbouring hidden AGNs is not a new concept, and in some sources, Xray observations have already revealed obscured AGNs (e.g. Alexander et al. 2003;Borys et al. 2005). Although both Chandra and XMM-Newton have observed the field, the observations targeted the foreground X-ray bright galaxy cluster making it difficult to resolve and disentangle emission from the QSO and companion.
Another possible explanation why we do not see many more systems reminiscent of SMM J04135+10277 is the time scale issue: it could be that we captured the system of SMM J04135+10277 at a special time of its evolution, with the companion galaxy having an intense star-forming phase and the quasar being active. Given the short duration of the starburst phase (< 10 8 yr) and quasar lifetime (< 10 8.5 yr, Di Matteo et al. 2008;Hopkins et al. 2008), it might be that we simply missed this window of time in case of other known systems and observed them in a later stage of their evolution.

CONCLUSIONS
We have presented ALMA observations of the dust continuum, CO(5-4) and CO(8-7) line emission of the quasarcompanion galaxy system SMM J04135+10277 (z = 2.84). Compared to previous studies of the system using large beam observations, we resolve the continuum emission and detect dust emission associated with both sources. Based on the ALMA continuum data, the dust emission is dominated by the star-forming companion galaxy but the quasar has a non-negligible contribution of 25% to the total emission. We fitted the SED of the sources using the SED fitting code MR-MOOSE, which is designed to treat upper limits and blended sources using a Bayesian approach. Based on the SED fitting the dust temperature of the quasar is higher compared to that of the companion, resulting in similar FIR luminosities for each source.
As the companion galaxy shows a very broad line profile in both CO transitions (∼ 1000 km s −1 ), we studied the kinematics of the galaxy. The ALMA results show signs of rotation, however, in the absence of high resolution observations a proper analysis of the kinematics is not possible and the observed line width of the companion is possibly affected by beam smearing.
While previous observations of the CO(1-0) and CO(3-2) transitions only detected molecular line emission associated with the companion galaxy, we detect a significant molecular gas reservoir in each source. The estimated molecular gas mass of the companion galaxy and the quasar host is ∼ 7×10 10 M and ∼ 0.7−2.3×10 10 M , assuming a conversion factor of α CO = 0.8 M (K km s −1 pc 2 ) −1 . In the light of our observations, it is clear that the quasar is not gas-poor as it was suggested by previous studies and has a significant molecular gas mass, only visible at higher frequencies due to excitation.
Using the results of low-J CO transitions observations found in the literature and our ALMA CO(5-4) and CO(8-7) detections, we model the SLED of each source using the radiative transfer code RADEX. In the case of the quasar we only have limited amount of detections, thus the model cannot constrain well the excitation properties of the source. However, models with low χ 2 value could narrow down the possible temperature and density range. For the companion the RADEX models cannot put constraints on the temperature but give a consistent estimate for the molecular hydrogen density. The reason of this might be the limited number of observations and the simplicity of the model. Finally, we compared the case of SMM J04135+10277 to other quasar-companion systems observed by ALMA. In comparison, SMM J04135+10277 stands out from similar systems as the quasar does not dominate the dust emission, while still having a significant amount of molecular gas.