AGN vs. host galaxy properties in the COSMOS field
G. Lanzuisi, I. Delvecchio, S. Berta, M. Brusa, A. Comastri, R. Gilli, C. Gruppioni, S. Marchesi, M. Perna, F. Pozzi, M. Salvato, M. Symeonidis, C. Vignali, F. Vito, M. Volonteri, G. Zamorani
aa r X i v : . [ a s t r o - ph . GA ] F e b Astronomy&Astrophysicsmanuscript no. multi˙cosmos˙v9˙rev2 c (cid:13)
ESO 2017February 27, 2017
AGN vs. host galaxy properties in the COSMOS field
G. Lanzuisi , ⋆ , I. Delvecchio , S. Berta ⋆⋆ , M. Brusa , , A. Comastri , R. Gilli , C. Gruppioni , S. Marchesi , , M.Perna , , F. Pozzi , , M. Salvato , M. Symeonidis , C. Vignali , , F. Vito , M. Volonteri , and G. Zamorani . Dipartimento di Fisica e Astronomia, Universit`a di Bologna, Viale Berti Pichat 6 /
2, I-40127 Bologna, Italy INAF - Osservatorio Astronomico di Bologna, Via Ranzani 1, I–40127 Bologna, Italy Department of Physics, University of Zagreb, Bijeniˇcka cesta 32, HR-10002 Zagreb, Croatia University of Zagreb, Physics Department, Bijeniˇcka cesta 32, 10002 Zagreb, Croatia Department of Physics & Astronomy, Clemson University, Clemson, SC 29634, USA Max-Planck-Institut f¨ur extraterrestrische Physik, Giessenbachstrasse, 85748 Garching, Germany Mullard Space Science Laboratory, University College London, Holmbury St. Mary, Dorking, Surrey RH5 6NT, UK Department of Astronomy and Astrophysics, 525 Davey Laboratory, The Pennsylvania State University, U. Park, PA 16802, USA Institut d’Astrophysique de Paris, UPMC et CNRS, UMR 7095, 98 bis bd Arago, F-75014 Paris, FranceReceived 25 October 2016; Accepted 22 February 2017
ABSTRACT
Context.
The coeval active galactic nuclei (AGN) and galaxy evolution and the observed local relations between super massive blackholes (SMBHs) and galaxy properties suggest some sort of connection or feedback between the SMBH growth (i.e., the AGN activity)and the galaxy build-up (i.e., the star formation history).
Aims.
We have looked for correlations between average properties of X-ray detected AGN and their FIR detected, star forming hostgalaxies, in order to find quantitative evidences for this connection, that has been highly debated in the latest years.
Methods.
We exploit the rich multi-wavelength data set (from X-ray to FIR) available in the COSMOS field for a large sample (692sources) of AGN and their hosts, in the redshift range 0 . < z <
4. We use X-ray data to select AGN and determine their properties,such as X-ray intrinsic luminosity and nuclear obscuration, and broad-band (from UV to FIR) SED fitting results to derive host galaxyproperties, such as stellar mass ( M ∗ ) and star formation rate (SFR). Results.
We find that the AGN 2-10 keV luminosity ( L X ) and the host 8 − µ m star formation luminosity ( L SFIR ) are significantlycorrelated, even after removing the dependency of both quantities with redshift. However, the average host L SFIR has a flat distributionin bins of AGN L X , while the average AGN L X increases in bins of host L SFIR with logarithmic slope of ∼ .
7, in the redshiftsrange 0 . < z < .
2. We also discuss the comparison between the full distribution of these two quantities and the predictions fromhydrodynamical simulations. No other significant correlations between AGN L X and host properties is found. On the other hand, wefind that the average column density ( N H ) shows a clear positive correlation with the host M ∗ , at all redshifts, but not with the SFR(or L SFIR ). This translates into a negative correlation with specific SFR, at all redshifts. The same is true if the obscured fraction iscomputed.
Conclusions.
Our results are in agreement with the idea, introduced in recent galaxy evolutionary models, that BH accretion and SFrates are correlated, but occur with di ff erent variability time scales. Finally, the presence of a positive correlation between N H and host M ∗ suggests that the column density that we observe in the X-rays is not entirely due to the circum-nuclear obscuring torus, but mayalso include a significant contribution from the host galaxy. Key words.
Galaxies: active – Galaxies: nuclei – Galaxies: evolution – Infrared: galaxies – X-ray: galaxies
1. Introduction
Super massive black hole (SMBH) growth and galaxy build upfollow a similar evolution through cosmic history, with a peak at z ∼ − z =
0, SMBHand their hosts sit on tight relations that link the SMBH massand the bulge properties of the host, such as luminosity, stel-lar mass and velocity dispersion (Kormendy & Richstone 1995,Magorrian et al. 1998, Kormendy & Ho 2013). Therefore theSMBH growth and the star formation history are likely relatedin some way during the cosmic time.The key parameter that regulates both processes seems to becold / molecular gas supply (Lagos et al. 2011, Vito et al. 2014,Delvecchio et al. 2015, Saintonge et al. 2016). Star formation- ⋆ e-mail: [email protected] ⋆⋆ Visiting scientist related processes (e.g. supernova and stellar wind) are knownto produce galaxy-wide outflows that can regulate the in-fallof gas and therefore the star formations itself (e.g. Genzel etal. 2011). But more powerful mechanisms, globally identifiedas ‘AGN feedback’, have been invoked in numerical and semi-analytic models of galaxy evolution (e.g. Granato et al. 2004; DiMatteo, Springel & Hernquist 2005, Menci et al. 2008, Sijackiet al. 2015, Dubois et al. 2016, Pontzen et al. 2016) in order toreproduce the observed galaxy population, and particularly thehigh mass end of the galaxy mass function.Observationally, the role of AGN activity in influencing theevolution of the global galaxy population is not clear yet. Thisissue has been investigated, in the past, looking for correlationsbetween average AGN and host properties, such as BH accretionrates (BHAR) or AGN luminosity (typically in the 2 −
10 keVband, L X hereafter) on one side, and SF rates (SFR) or IR lumi-nosity (in the 8 − µ m , L IR hereafter) on the other side, taking
1. Lanzuisi et al.: AGN vs. host properties in COSMOS advantage of the wealth of multi-wavelength information col-lected in deep extragalactic surveys. However, di ff erent, some-what contradictory results have been reported in the past years.Several studies have found a flat distribution computing av-erage L IR in bins of L X of X-ray selected sources (or SFR andBHAR, respectively, e.g. Shao et al. 2010, Rovilos et al. 2012,Rosario et al. 2012) at low redshift and luminosities. A sig-nificant positive correlation instead appears for luminous AGN( L X > − erg s − ) and high redshifts ( z > − ff erent triggering mechanisms at high and low luminosi-ties, via merger and secular evolution, respectively.Other groups have found a linear correlation at all z and L X ,when computing the average L X in bins of L IR (in log-log space)of IR selected sources (Chen et al. 2013, Delvecchio et al. 2015,Dai et al. 2015), with the ratio Log(SFR / BHAR) ∼
3, roughlyconsistent with the local M bulge / SMBH value (Magorrian et al.1998, Marconi & Hunt 2003). Finally, other authors have foundno correlation at all, regardless of the L X and z range (e.g.Mullaney et al. 2012, Stanley et al. 2015).Looking at AGN obscuration, Rovilos et al. (2012) exploredfor the first time the possible relation between AGN column den-sity ( N H ), as measured from the X-ray spectra, and host prop-erties, finding no correlation on a sample of 65 sources in theXMM-CDFS survey (Comastri et al. 2011). Rosario et al. (2012)found similar result from hardness ratios (HR) on a larger sam-ple in COSMOS, while Rodighiero el al. (2015) found a positivecorrelation between N H and M ∗ , again based on HR, on a sampleof z ∼ ff erences may partlyarise from di ff erent biases and analysis methods. For example,given that only a small fraction of the X-ray detected sources areFIR detected, and vice-versa, most of these studies rely on X-rayor FIR staking in order to recover the average properties of largesamples of AGN / host systems, or are limited to small subsam-ples. Mullaney et al. (2015) pointed out that modeling the SFRdistribution of X-ray selected hosts as a log-normal distribution,and including upper-limits, gives di ff erent results than comput-ing the linear mean of the distribution (i.e. via staking), that isinstead driven upwards by the bright outliers.Another issue was raised by Symeonidis et al. (2016), show-ing that the intrinsic AGN SED in the FIR is cooler than usuallyassumed. Therefore in some cases there is no ‘safe’ photometricband which can be used in calculating the SFR, without sub-tracting the AGN contribution. On the other hand, several of theworks cited above, take directly the FIR photometry (typicallyat 60 µ m ) in order to estimate SFR, thus potentially introducinga spurious correlation at high AGN luminosities.Recently, from theoretical studies, a physical mechanism hasbeen proposed to explain part of these contradictory results:Volonteri et al. (2015a,b) explain these di ff erent observationswith the way we analyze the data: the bivariate distribution ofAGN and SF luminosities gives two very di ff erent results, de-pending on the binning axis. Hickox et al. (2014) reached sim-ilar conclusions starting from the simple assumptions that longterm AGN activity and SFR are perfectly correlated, that the ob-served SFR is the average over ∼
100 Myr, while the AGN ac-tivity, traced by X-ray emission, varies on a much shorter timescales. In these models the di ff erent time scales involved in AGNand SF variability wash out the linear dependency between thetwo quantities, if the rapidly variable AGN luminosity is usedto build the subsamples to be studied ‘on average’. This resultwas also confirmed observationally by Dai et al. (2015) usingshallow data from XMM-LSS. Furthermore, Volonteri et al. (2015a) suggest that spatialscales are important: the BH accretion rate should be corre-lated with the nuclear ( <
100 pc) SFR, while it is less corre-lated with the total ( < ff ect the whole host galaxy.Of course, the SFR that can be inferred from the FIR luminosityis the global, galaxy-scale SFR (with the exception of the localUniverse, see e.g. Diamond-Stanic & Rieke 2012), and this in-troduce another source of uncertainty in the observational com-parison between BHAR and SFR.Here we explore the possible correlations between AGN andhost properties for a large sample of X-ray and FIR detectedsources thanks to the extensive Chandra , XMM–Newton and
Herschel coverage on the COSMOS field (Scoville et al. 2007,Hasinger et al. 2007, Elvis et al. 2009, Lutz et al. 2011, Oliveret al. 2012). This approach avoids the uncertainties related tothe staking, and allow for a proper SED deconvolution, sourceby source. This of course limits the significance of our findingsto the brightest, most accreting and most star forming systems.These are, however, the most interesting ones: the ones for whichthere is less agreement in the literature on the presence of a cor-relation between AGN and SF, and also the ones for which the-oretical models predict that the correlation should be stronger.The paper is organized as follows: section 2 describes thesample and source properties; section 3 presents the analysis of L X and L IR distributions; in section 4 we compare our resultswith a set of hydrodynamical simulations; in section 5 we dis-cuss correlations between nuclear obscuration and host proper-ties and in section 6 we discuss our results. Throughout the pa-per, we assume a standard Λ CDM cosmology with H =
70 kms − Mpc − , Ω Λ = .
73 and Ω M = .
27 (Bennett et al 2013).
2. The sample
We performed X-ray spectral fitting for all the
Chandra and
XMM–Newton detected sources (from the catalogs of Brusaet al. 2010 and Civano et al. 2015 respectively) with morethan 30 counts, in Marchesi et al. (2016) and Lanzuisi et al.(2013, 2015), respectively. This sample consist of 2333 indi-vidual sources (1949
Chandra and 1187
XMM–Newton sources,with 803 source in common ).For all the Herschel detected sources in the COSMOS field(Lutz et al. 2011, ∼ > σ inone of the FIR bands, from 100 to 500 µ m ), an SED deconvolu-tion with 3 components — stellar emission, AGN torus emissionand SF-heated dust emission — performed using photometricpoints from the UV to sub-mm, is available from Delvecchio etal. (2014, 2015, D15 hereafter), following the recipe describedin Berta et al. (2013).We then selected all the XMM–Newton and
Chandra de-tected sources, having at least one FIR detection (and thereforeSED deconvolution). The final sample comprises 692 sourcesX-ray and FIR detected (the ’X-FIR sample’ hereafter), all ofthem with an available redshift, 459 spectroscopic and 233 pho-tometric (Civano et al. 2012, Brusa et al. 2010, Salvato et al.2009, Marchesi et al. 2015). This is to date the largest sample ofAGN / host systems for which X-ray spectral parameters, such ascolumn density and absorption-corrected 2-10 keV luminosity,are known in combination with host properties such as M ∗ andSFR. For sources in common the
Chandra data from Marchesi et al.(2016) are used, given the deeper coverage.2. Lanzuisi et al.: AGN vs. host properties in COSMOS
Fig. 1:
Left : Distribution of rest frame, absorption corrected 2-10 keV luminosity vs. redshift for the
XMM–Newton (blue squares)and
Chandra (red crosses) detected sources that are also
Herschel detected (X-FIR sample). The dotted (dashed) line marks thesensitivity limit of the
XMM–Newton ( Chandra ) surveys. The redshift bins adopted in the text are marked by vertical gray dashedlines.
Right : Distribution of intrinsic column density vs. redshift for the sample. Symbols as in left panel. Arrows show upper-limitsfor unobscured sources. The average 1 σ error-bars on L X and N H are shown in the top left of both panels. Figure 1 (left) shows the distribution of L X vs redshift for theX-FIR sample. The average 1 σ error bar on L X is shown in theupper left corner. The absorption-corrected L X is a ff ected by un-certainties related to both the number of net counts (observedflux uncertainties) and the spectral shape of each source (un-certainties on N H and spectral slope). Therefore, the errors havebeen derived, for each source, using the equivalent in Sherpa (Fruscione et al. 2006) of the cflux model component in
Xspec (Arnaud 1996), applied to the best-fit unabsorbed powerlaw. Theflux and errors are then computed in the observed band corre-sponding to 2-10 keV rest-frame, and converted into luminosity.The redshift bins that will be used in the following analysisare shown with vertical dashed lines. The intervals have beenchosen with the aim of having a fairly large number of sourcesin each bin ( ∼ − L X bins that will be used in the following (1 bin per dex) areshown as horizontal dashed lines.Figure 1 (right) shows the column density distribution for theX-FIR sample. Arrows show sources for which the obscurationis constrained only by an upper-limit. The average 1 σ error baron N H is shown in the upper left corner. The distribution of N H from X–ray spectral analysis has a clear upper boundary aroundCT column densities due to the strong flux decrement associ-ated with CT obscuration in the 2-10 keV band. Also, the min-imum measurable N H increases with redshift, as the low energycut-o ff due to obscuration move outside the observing band.The global fraction of X-ray obscured sources (those with N H > cm − ) in the X-FIR sample is ∼ Lanzuisi et al. (2015a,b) present the CT sources detected by
XMM–Newton , while Lanzuisi et al. 2017 in prep. will present the ones de-tected by
Chandra . Fig. 2: SFR vs. M ∗ distribution for the entire sample of Herschel detected sources ( ∼ σ error-bars are shown in the top leftas a black cross. Chandra - and
XMM–Newton -COSMOS (Lanzuisi et al. 2015,Marchesi et al. 2016). Indeed, the FIR luminosity (and therefore
3. Lanzuisi et al.: AGN vs. host properties in COSMOS
Fig. 3: Fractional distribution of the host properties, SFR, M ∗ , sSFR, and MS o ff set, from top left to bottom right, for the X-FIRsample (black open histogram), and for the whole Herschel sample (gray filled histogram), in redshift bins.
Herschel detection rate) of type-2 AGN seems to be higher thanfor type-1 QSO (Chen et al. 2015).
The host properties (SFR vs. M ∗ ) of the 692 sources in the X-FIR sample are shown in figure 2 (red circles) divided in fiveredshift bins as described above. The values are taken from D15:the SFR has been derived by converting the IR luminosity (rest8 − µ m ) of the best-fitting galaxy SED (i.e. subtractingthe AGN emission when present) with the SF law of Kennicutt(1998), scaled to a Chabrier (2003) initial mass function (IMF).The M ∗ is derived from the SED decomposition itself, and basedon Bruzual & Charlot (2003) models, with a consistent IMF.Table 1 (full version available on-line) summarizes the multi-wavelength properties of the sources in the X-FIR sample.The host properties of the sample of Herschel detectedsources (from D15, ∼ σ error-barsresulting from the SED fit are shown in the top left corner.The errors on M ∗ follow a log-normal distribution, with aver-age h err ( M ∗ ) i = .
14 dex and standard deviation σ = . Systematic errors like, for example, uncertainties related to theadopted IMF or SF law, are not included in the error budget.
The mean error on SFR is h err (SFR) i = .
10 dex, and stan-dard deviation of σ = .
07 as for L SFIR (see sec. 3.1), since theSFR is derived from L SFIR adopting a Kennicutt (1998) law. Theredshift-dependent MS of star forming galaxies as described inWhitaker et al. (2012) is also shown in each panel. The FIRselected sources broadly follow the MS relation. However, the
Herschel -based selection is sensitive to the most star formingsystems, introducing a cut in SFR that moves towards higher val-ues with increasing redshift. (e.g. Rodighiero et al. 2011, D15).X-ray detected AGN are preferentially found at the highest M ∗ , i.e. the fraction of X-ray detected sources increases as afunction of M ∗ , in the first three redshift bins at least. This isa well known e ff ect (Kau ff mann et al. 2003, Bundy et al. 2008,Brusa et al. 2009, Silverman et al. 2009, Mainieri et al. 2011,Santini et al. 2012, Delvecchio et al. 2014). Aird et al. (2012)suggested that it is the result of an observational bias, such thatmore massive galaxies (i.e. more massive BHs), can be detected,at a given X-ray flux limit, with a variety of accretion rates, whilelower mass systems can be detected only if they have a high ac-cretion rate. This, combined with a steep Eddington ratio dis-tribution (i.e. sources with low Eddington ratio are much morecommon than sources with high Eddington ratio) can explain theobserved M ∗ distribution (see also Bongiorno et al. 2012).
4. Lanzuisi et al.: AGN vs. host properties in COSMOS
Table 1: Multi-wavelength properties of the 692 sources in the X-FIR sample.
ID RA DEC z Log( L SFIR ) Log( M ∗ ) SFR Log( N H ) Log( L X ) Log( L Bol ) XID CIDdeg deg erg / s M ⊙ M ⊙ / yr cm − erg / s erg / s(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)1846545 150.500 2.862 0.102s 44 . ± .
07 10 . ± .
13 15.8 < .
77 41 . ± .
27 41.96 60095 lid21001883498 150.065 2.929 0.102s 43 . ± .
16 10 . ± .
09 1.3 < .
42 42 . ± .
06 43.8 5617 lid38589570 150.372 1.609 0.104s 44 . ± .
08 10 . ± .
09 2.8 22 . + . − . . ± .
04 44.29 2021 cid16781612003 150.550 2.628 0.113s 43 . ± .
28 10 . ± .
09 0.4 < .
36 41 . ± .
36 42.6 — lid31891197519 150.335 2.304 0.123s 44 . ± .
06 10 . ± .
09 4.7 21 . + . − . . ± .
29 42.11 1533 cid967...
Notes.
Catalog entries are as follows: (1) Source ID from Capak et al. (2007); (2) and (3) right ascension and declination of the optical / IRcounterpart; (4) redshift ( s for spectroscopic or p photometric); (5) Log( L SFIR ) with 1 σ errors; (6) Log( M ∗ ) with 1 σ errors; (7) SFR derived from L SFIR ; (8) Log( N H ) with 1 σ errors or upper-limits; (9) Log( L X ) with 1 σ errors; (10) Log( L Bol ) computed from L X using Marconi et al. (2004);(11) and (12) XMM-COSMOS and Chandra-COSMOS IDs (from Brusa et al. 2010 and Marchesi et al. 2016 respectively). The full table will beavailable in electronic form at the CDS via http://cdsweb.u-strasbg.fr/cgi-bin/qcat?J/A+A/ . In our case there is a threshold at around log M ∗ ∼ M ⊙ in the first 3 redshift bins. A simple calculation shows that thisvalue can be roughly derived from the X-ray flux limit of the Chandra and
XMM–Newton surveys, using standard values forbolometric corrections ( k Bol = − λ Edd ∼ .
05) and BH-host mass ratios ( M ∗ / M BH = − M ∗ and L X distributions, will be presentedin Suh et al. (2017 submitted).Several studies in the local Universe suggest that the fractionof galaxies hosting an AGN increases also with IR luminosity(e.g. Lutz et al. 1998, Imanishi et al. 2010, Alsonso-Herrero etal. 2012, Pozzi et al. 2012). We also tested that the observedthreshold in mass is not driven by our requirement of Herschel detection: also using the M ∗ -SFR distribution of Bongiorno etal. 2012, computed for the full XMM-COSMOS catalog, a dropin the number of X-ray detected AGN below log M ∗ = M ⊙ is visible up to z = . ff ect is that the X-FIRsample has a M ∗ distribution shifted toward higher M ∗ with re-spect to the global Herschel sample (Fig. 3 top right). The dis-tribution of SFR for the X-FIR sample, instead, is roughly con-sistent with that of the global
Herschel sample (Fig. 3 top left).This have important implications when measuring, e.g., sSFRand MS o ff set (Fig. 3 bottom left and right): due to this selectione ff ect the X-FIR sample has lower sSFR with respect to the MSof star forming galaxies (or to the Herschel sample), if the twosamples are not properly mass-matched (Silverman et al. 2009,Xue et al. 2010). L X vs. L IR distributions The two quantities that have been more often used in order tolook for BHAR-SFR correlations are the AGN luminosity, of-ten represented by the L X , and the SF luminosity in the form of L IR (or L µ m , Santini et al. 2012, Rosario et al. 2012, Chen etal. 2013). It is generally assumed that the total FIR luminosityis not significantly a ff ected by any contamination from the AGNemission. However, recent studies have shown that the AGN maycontribute significantly to the IR emission and in some case evenin the FIR band (Symeonidis et al. 2016). Therefore, the SFR de-rived directly from FIR photometry can be overestimated, espe- Fig. 4: L X vs. L SFIR for the X-FIR sample. Di ff erent colors rep-resent di ff erent redshift bins: blue for 0 . < z < .
4, cyan for0 . < z < .
8, green for 0 . < z < .
2, red for 1 . < z < < z <
4. The average 1 σ errors on L X and L SFIR areshown in the upper left corner.cially in high luminosity AGN hosts. Thanks to the SED decom-position available, we will use in the following the L IR computedfor the SF component only ( L SFIR hereafter), after subtracting theAGN contribution, modeled with the SED templates of Fritz etal. (2006, see also Feltre et al. 2012). This will allow us to avoidintroducing a spurious correlation between AGN and SF lumi-nosity, especially at the highest luminosities.Clearly two luminosities are always correlated in any samplethat is flux limited in both directions, due to the combination ofthe luminosity-distance e ff ect and of the fact that the sourcestend to cluster at the flux limit (Malmquist bias, e.g. Feigelson& Berg 1983). Figure 4 shows the distribution of L X vs. L IR forthe X-FIR sample.
5. Lanzuisi et al.: AGN vs. host properties in COSMOS
The 1 σ errors on L SFIR follow a log-normal distribution withaverage value h err ( L SFIR ) i = .
10 dex, and standard deviation of σ = .
07. As mentioned in sec. 2.1, the errors on the absorp-tion corrected luminosity follow a much broader distribution,depending both on the number of counts available and on thespectral shape. They range from < ∼ . − . ∼ . − . σ error is h err ( L X ) i ∼ . σ = .
18. We show the averageerrors with a black cross in the left panel of Fig. 4, while thespecific value for each source is used in the following analysis.In order to look for intrinsic correlations between these twoquantities, one possibility is to compute the partial
Spearmanrank correlation between two variables in presence of a third,and to assess the statistical significance of such correlation (e.g.Macklin 1982). To derive the correlation coe ffi cient between L X and L SFIR , conditioned by the distance, ρ ( L X , L SFIR , ˙ z ), we evaluatethe Spearman coe ffi cient ρ related to each couple of parametersand then combined them according to the expression: ρ ( a , b , ˙ c ) = ρ ab − ρ ca ρ bc q (1 − ρ ca )(1 − ρ bc ) (1)(Conover 1980) which returns the partial correlation between a and b , corrected for the dependency on c . The resulting ρ is 0.15,and the associated confidence level, in terms of standard devia-tions, that the first two variables are correlated, independently ofthe influence of the third, is ∼ . σ , following eq. 6 of Macklin(1982). Therefore, the two quantities appear to be significantlycorrelated, after the e ff ect of redshift on both of them is takeninto account. The second approach, often used in the literature, is to define asnarrow as possible redshift bins, to minimize the distance e ff ect,and look for correlations between the two quantities. Thanks tothe large sample collected in this work, we can divide the samplein five redshift bins. For every redshift bin, a large distributionin both luminosities can be observed, with the typical luminosityincreasing with redshift (Fig. 4).Most of the observational works mentioned in Sec. 1 lookedfor the distribution of average L SFIR in L X bins, or average L X in L SFIR bins (but see Gruppioni et al. 2016). Both hydrodynami-cal simulations (e.g. Volonteri et al. 2015a,b) and semi-analyticmodel (e.g. Neistein & Netzer 2014) show that, in the L SFIR - L X plane, there may be the superimposition of a weak correlationfor the bulk of the population, and a strong correlation only forthe most extreme merger phases, corresponding to the highest L X and L SFIR . If the underlying distribution shows such a complexshape, the results of the two approaches (average L SFIR in L X binsor average L X in L SFIR bins) may be very di ff erent.In Fig. 5 left, we show the result of plotting average L SFIR inbin of L X (both in log scale), in five redshift bins. As can be seen,there is no correlation at all L SFIR and, as expected, there are nosources below the relation computed for a pure AGN template inMullaney et al. (2011), similar to the one of Netzer et al. (2009).Following this approach, we are therefore able to reproduce theresults of Shao et al. (2010), Rosario et al. (2012) and others, thatclaim no correlation between AGN activity and SF over severalorders of magnitudes in luminosity.On the other hand, computing average L X in L SFIR bins (inlog scale), from the same bivariate distribution, gives di ff erent results. Fig. 5, right, shows that, at all redshifts, the average L X correlates with the L SFIR and the binned points are close to theSFR / BHAR ∼
500 ratio found in Chen et al. (2013, C13 here-after).In both panels, we computed the error on the average L X and L SFIR through a bootstrap re-sampling procedure, as done in sev-eral previous works. For each bin with N sources, we randomlyextract N sources, allowing repetitions, and computed the meanvalue. The process is iterated 10 times, and the standard devia-tion of the mean is taken as error on the average SFR.The two approaches described above are the equivalent ofcomputing the forward and inverse linear regression of one vari-able over the other. Table 2 reports the slopes α , intercept β andassociated error, for each redshift bin, in the log-log space, of theleast square (LS) fit of L SFIR as a function of L X (hereafter L SFIR — L X ), and L X as a function of L SFIR (hereafter L X — L SFIR ), respec-tively . Indeed the slopes in the left panel are all consistent with0 within ∼ σ c.l.. On the other hand, LS fits of ( L X — L SFIR )give steeper correlations at all z bins, and slopes not consistentwith 0 at ∼ σ c.l.The SFR / BHAR ∼
500 ratio plotted in Fig. 5 is the onefound in C13, for a sample of 121 FIR selected AGN-hosts at0 . < z < .
8. To compare with their results we should lookat our first two z-bins: While the z-bin 0 . ≤ z < . L X — L SFIR ) slope, possibly due to the small volume sam-pled, the 0 . ≤ z < . ∼ σ , therefore in broad agreement with theC13 findings. Interestingly, we can extend up to 0 . ≤ z < . / BHAR ∼
500 can be found. Above this redshift interval, theslopes become flatter. Therefore, we found a strong (almost lin-ear) correlation between log L X and log L SFIR , for ( L X — L SFIR ) atredshifts lower than the peak of the SF and AGN activity, i.e.between 4 and 8 Gyr ago, while at higher redshift the correlationis still present but weaker.The exact value of the ratio SFR / BHAR in terms of L X and L SFIR depends strongly on the assumptions made to scale be-tween these quantities, i.e. the accretion e ffi ciency and bolo-metric correction in the first case, and the SF law and initialmass function (IMF) in the second. C13 derived the SFR from L SFIR using the Kennicutt (1998) relation, modified for a ChabrierIMF (Chabrier 2003) and the BHAR from L X using a constant k bol = . ffi ciency of 0.1. They use as referencethe value of SFR / BHAR ∼
500 derived from the M
Bulge / M BH ra-tio observed in Marconi et al. 2004. The authors suggest that thefact that the detected sources sit on the SFR / BHAR ∼
500 ratiois a coincidence, due to the ratio between the X-ray and FIR fluxlimits in the Bo¨otes field.In the X-FIR sample in COSMOS we have a factor of ∼ Herschel data are only a factor 2-3 deeper ( ∼ µ m and 8 mJyat 250 µ m ) than in Bo¨otes. Nonetheless, our X-FIR detected sam-ple sit close to the C13 relation. We underline that, in both cases,the X-ray and FIR detected sources are a small minority of boththe original X-ray and FIR samples (few % up to ∼ The LS fit is performed with the BCES code (Akritas & Bershady1996), adopting 10 bootstrap re-samplings. Similar results are obtainedusing the LINMIX code (Kelly et al. 2007). In the first case slopes and intercepts refer to a relation in the form
Log L
SFIR = + α × ( Log L X − + β , while in the second in the form Log L X = + α × ( Log L
SFIR − + β .6. Lanzuisi et al.: AGN vs. host properties in COSMOS Fig. 5:
Left : Average Log( L SFIR ) in bins of Log( L X ) in five redshift bins. The short dashed line is the correlation derived in Mullaney etal. (2011) for a pure AGN SED. Right : Average Log( L X ) in bins of Log( L SFIR ). The long dashed line represents a constant SFR / BHARof 500, from C13. In both panels the vertical error-bars are computed through a bootstrap re-sampling procedure, while the horizontalerror-bars show the 1 σ dispersion of that bin.Table 2: Slopes α and intercept β of the linear LS fit of ( L SFIR — L X ), ( L X — L SFIR ), and of the bisector estimator, in each redshift bin.The first set of slopes and intercepts refers to a relation in the form
Log L
SFIR = + α × ( Log L X − + β , while the second and thirdin the form Log L X = + α × ( Log L
SFIR − + β . z-bin LS ( L SFIR — L X ) LS ( L X — L SFIR ) bisector( L X , L SFIR ) α β α β α β . ≤ z < . . ± . − . ± .
09 0 . ± . − . ± .
21 1 . ± . − . ± . . ≤ z < . . ± .
10 0 . ± .
04 0 . ± . − . ± .
05 1 . ± . − . ± . . ≤ z < . . ± .
09 0 . ± .
03 0 . ± . − . ± .
04 1 . ± . − . ± . . ≤ z < . . ± .
09 0 . ± .
04 0 . ± . − . ± .
08 1 . ± . − . ± . . ≤ z < . . ± .
08 1 . ± .
04 0 . ± .
18 0 . ± .
20 1 . ± . − . ± . as discussed also in C13, the flux limit has an important role inthe observed properties of the detected sources alone.As discussed in Hickox et al. (2014) and Volonteri et al.(2015), a possible physical explanation for this behavior is that,when looking at left panel of fig. 5, we are averaging a slowlychanging quantity, such as the host SFR, of galaxies grouped onthe basis of the rapidly changing AGN L X . In the right panel,instead, the average L X of a large sample of sources groupedon the basis of the slowly changing SFR, allows us to recoverthe underlying, long term correlation between AGN activity andSFR. In the same way, from a statistical point of view, it may bereasonable to interpret the L X as the dependent variable, in thiscontext, as it has larger uncertainties with respect to L SFIR (Hogget al. 2010), both in terms of measurement errors (see sec. 3.1)and noise (i.e. variability).If we instead assume that in this case there is no “depen-dent” and “independent” variables (see e.g. Tremaine et al. 2002,Novak et al. 2006), the two variables may need to be treated sym-metrically. We used again the BCES code, to derive slope andintercept, and their standard deviation, using a symmetric esti- mator such as the bisector regression (Isobe et al. 1990). Theresults are shown in figure 6, while the slopes and intercepts arereported in Table 2. At all redshift bins, the slopes of the linearregression, although always larger than 1, are consistent with 1within 1 σ c.l. Since the
Herschel (Pilbratt et al. 2010) PACS and SPIRE pointspread functions are much larger than the one in the optical andNIR bands, going from ∼ ′′ to ∼ ′′ FWHM (Poglitsch et al.2010; Gri ffi n et al. 2010), there is the possibility that the FIR fluxof our sources is contaminated by unresolved neighbors (see e.g.Scudder et al. 2016).We verified the e ff ect of contamination by excluding fromthe X-FIR sample all sources with a second HST catalog entry, We recall that the BCES estimators, both the LS and the symmetricones, are not immune from biases that arise from data truncation, whichis the case for flux-limited samples (see Akritas & Bershady 1996). 7. Lanzuisi et al.: AGN vs. host properties in COSMOS
Fig. 6: Linear regression for L X and L SFIR computed for each red-shift bin with the bisector estimator in BCES. The color code isthe same as Fig. 4.from the ACS F814W (I-band) catalog (28.6 AB limiting mag-nitude, Scoville et al. 2007, Koekemoer et al. 2007). We choosea circular area of diameter 8 ′′ around the optical position. Whilethis distance is not enough to ensure negligible contamination, ithas been chosen in order to retain a su ffi cient number of sourcesto allow an analysis in all the five redshift bins. The 146 “iso-lated” sources obtained in this way show the same behavior de-scribed above, with a flat distribution of average L SFIR computedin bin of L X , and an almost linear correlation of average L X com-puted in bin of L SFIR .We also verified that sources with a single PACS or SPIREdetections (more subject to contamination) do not a ff ect our re-sults. Indeed, excluding the 154 (out of 692) sources with onlyone detection (at 3 σ ) either in PACS or SPIRE photometry, doesnot change the results presented in sec. 3.2 and in the followingparagraphs. L Bol , M ∗ , sSFR and MSoffset Several authors have used the AGN bolometric luminosity( L Bol ), instead of the L X , to look for correlation with the L IR or SFR. The L Bol is generally derived from the L X through aluminosity dependent bolometric correction (e.g. Marconi et al.2004, Lusso et al. 2012). The net e ff ect of this procedure is tostretch the horizontal axis of Fig. 5, left (the high L X sourceshave a higher X-ray bolometric correction than the low L X ones),while keeping the L SFIR fixed. In Fig. 7 we show the result ofthis approach (here we used the Marconi et al. 2004 luminos-ity dependent bolometric correction, but the Lusso et al. 2012relation would have the same e ff ect): in each redshift bin, thesources populating the highest L X bin are now spread in two L Bol bins and the last L Bol bin at each redshift is now populated by asmaller number of more extreme sources. The relation found lo-cally for AGN-dominated systems in Netzer et al. (2009) is alsoshown. Once again, we are able to reproduce results obtained inother works (Shao et al. 2010, Rosario et al. 2012). However, Fig. 7: Average Log( L SFIR ) in bin of L Bol , for the X-FIR sample.The dashed line is the relation found in Netzer et al. (2009) forAGN dominated systems. Error-bars computed as in Fig. 5.we are now confident that this result is not in disagreement withwhat shown in Fig. 5 (right), and the apparent contradiction isonly dependent on the way the data are analyzed and grouped,as shown in e.g. Volonteri et al. (2015) and Dai et al. (2015).Finally, we found a flat distribution when computing aver-age L X in bins of M ∗ , sSFR and MS o ff set, and average M ∗ ,sSFR and MS o ff set, in bins of L X , in all the five redshift bins.Indeed, no significant partial correlation is found, between anypair of these quantities, following the approach described in sec.3.1 to take into account the redshift e ff ect, that a ff ects also M ∗ ,SFR and sSFR ( σ << M ∗ covered by our sample is limited to the very highmass end, 10 < Log ( M ∗ ) <
12 (to be compared with the under-lining galaxies M ∗ distribution shown e.g. in Laigle et al. 2016,7 < Log( M ∗ ) <
12 in the same redshift interval). Deeper X-raysurveys are needed to investigate the dependency of L X with thiscrucial quantity.
4. Comparison with simulations
Here we compare our results with predictions from the simu-lations of galaxy mergers presented in Volonteri et al. (2015a).They are based on very high spatial and temporal resolution sim-ulations, covering a large range of initial mass ratios (1:1 to1:10), several orbital configurations, and gas fraction (definedas M gas / M ∗ ) in the range f gas = . − .
6. The very high res-olution imposes a limit on the mass of the simulated galaxies,that typically have M ∗ ∼ (2 − × M ⊙ , i.e. much smaller thanthe typical mass of our observed galaxies (see Fig. 3). The pro-cess is divided into three phases: the stochastic phase, in whichthe galaxies behave as they do in isolation, that lasts until thesecond pericenter; the merger phase characterized by strong dy-namical torques and angular momentum loss; the remnant phase,that starts when the angular momentum returns to be constant intime. While the stochastic and remnant phases have the same du-ration (by construction), the merger phase is much shorter (typi-cally 1 /
10 of the total).
8. Lanzuisi et al.: AGN vs. host properties in COSMOS
Fig. 8:
Left : BHAR / M ∗ rate vs. sSFR contours obtained from the simulations presented in Volonteri et al. (2015a). In red thestochastic phase, in yellow the merger phase and in black the remnant phase. Right : BHAR / M ∗ vs. sSFR contours observed inCOSMOS in five redshift bins. Sources that are in a major merger state in the first three redshift bins are marked with black stars.To compare our data with this set of simulated galaxies, weconverted the AGN bolometric luminosity into a BH mass ac-cretion rate (BHAR), by assuming an e ffi ciency of η = . M ∗ (from SED fitting) is much less uncertain (see sec. 2.2 for theerror budget in our sample) than that of M BH , and is availablefor both type-1 and type-2 AGN. This value is then comparedwith the sSFR for each source. The contours of global (within5 kpc) sSFR vs. sBHAR, obtained from the simulations for thethree di ff erent phases (stochastic, merger and remnant), are colorcoded in Fig. 8 (left) with red, yellow and black, respectively.The results from the X-FIR sample are shown in Fig. 8(right) for the five redshift bins. As can be seen the observedcontours in the low redshift bins span a similar range of physi-cal properties, with respect to simulations, with the bulk of thepopulation concentrated between 5 × − and 5 × − yr − insSFR, and between 10 − and 10 − yr − in sBHAR, and with atail at higher sSFR and sBHAR, possibly produced by sourcesin the merger phase as in the simulations (yellow contours).Interestingly, the importance of this tail grows with increasingredshift, even if the selection e ff ect in both directions must betaken into account.We also exploited the deep HST ACS coverage in theCOSMOS field to identify sources in the merger phase. We se-lected only sources that appear to be in a clear major mergerphase, and over-plotted them in Fig. 8 (right) as black stars, inthe first three redshift bins (above z ∼ ffi cult toassess the AGN host morphology). This selection is not meant tobe complete: not all the sources are covered by ACS, and not forall of them it is possible to recognize the host morphology, due tobright point-like AGN contribution, for example. However, it isinteresting that AGN hosts clearly in merger state tend to coverthe highest sSFR and sBHAR range, as predicted by simulations. One caveat to be considered here is the fact that the simulationsare performed at high-z, starting at z = − M ∗ for these galaxies is in therange Log( M ∗ ) = − . M ⊙ ), i.e. in the low mass tail of themass distribution even for the lowest redshift bin of the observedsample. Since the e ffi ciency of SFR and BHAR is most prob-ably mass-dependent, the comparison between di ff erent massranges may not be straightforward. Volonteri et al. (2015a) ar-gue, however, that SFR and BHAR are self-similar, on the basisof the mass sequence of star forming galaxies and of the possiblepower-law dependence of the specific BHAR (Aird et al. 2012;Bongiorno et al. 2013, but see Kau ff mann & Heckman 2009,Lusso et al. 2012 and Schulze et al. 2015).Finally, the simulations are not cosmological, in the sensethat the gas mass is not replenished by cosmic inflows and gasaccretion, as it is the case for real galaxies. This leads to a pos-sible underestimate of SFR and BHAR towards the end of thesimulation, when galaxies have converted a large fraction of theirgas in stellar and BH mass (see also Vito et al. 2014). N H and host properties Here we discuss the possible correlations between the columndensity through the AGN line of sight, as measured by the X-ray N H , and the host galaxy properties, such as M ∗ , SFR and sSFRand MS o ff set. The partial correlation analysis described in sec.
9. Lanzuisi et al.: AGN vs. host properties in COSMOS
Fig. 9: Linear regression of N H vs. M ∗ (left) and sSFR (right) in five redshift bins. The regression is performed using the linmix code, that also takes into account the N H upper-limits. The color code is the same of Fig. 4. The gray squares in the left panelshow results from Rodighiero et al. (2015) at z ∼
2, obtained from the HR of X-ray stacked images of FIR detected galaxies in theCOSMOS field. The orange dashed line is the relation found in Buchner et al. (2017) for a sample of GRB hosts in a wide range ofredshifts (see text).Fig. 10: log M ∗ vs. log M gas as derived from the eq. 1 of Scovilleet al. (2016). The sources are color coded on the basis of theirgas fraction.3.1, gives a significant positive correlation (at > σ c.l.) between N H and M ∗ , in the entire sample, once the distance e ff ect (both N H and M ∗ tend to increase with redshift in two di ff erent ways,due to two di ff erent selection e ff ects) is removed. We also finda significant negative correlation (at > σ c.l.) between N H andsSFR, while we do not find any significant correlation of N H withSFR and MS o ff set. As in the case of L X vs. L IR , the binning direction (or the vari-able chosen as independent) is relevant for the final distributionof N H as a function of host properties and vice-versa: computingaverage SFR, M ∗ , sSFR and MS o ff set in bin of N H we founda remarkably flat distribution of all these quantities, in agree-ment with results from Shao et al. (2010), Rovilos et al. (2012),Rosario et al. (2012), where the authors do not find any evolutionof the average host properties in bins of N H .On the other hand, computing average N H values in binsof M ∗ gives a positive trend in each redshift bin, while com-puting the average N H in sSFR bins gives a negative trend, inagreement with partial correlation analysis. The situation in thiscase is however complicated by the presence of upper-limits in N H , that makes the problem inherently asymmetric. We there-fore performed the linear regression of (Y—X) with a Bayesianapproach using the linmix code (Kelly et al. 2007) that is ableto properly take into account the upper-limits on N H .The result is shown in Fig. 9: the linear regression gives aclear positive correlation of N H with the host stellar mass, in-creasing by one-two dex from low to high masses, at all redshifts(slopes in the range α = .
42 – 0 . N H decreases typically by one order ofmagnitude or more, going from low to high sSFR (slopes in therange α = − .
35 – − . N H with SFR, and that the sSFR is defined as SFR / M ∗ , the two rela-tions are clearly connected.A similar result between N H and M ∗ was found in Rodighieroet al. (2015) for a sample of z ∼ N H is globally ∼ N H fromhardness ratios of the X-ray stacking, which includes also highlyobscured and CT, undetected AGN.Interestingly, a recent study on the distribution of the obscu-ration observed in X-ray spectra of GRB, as a function of thehost galaxy mass, found a similar trend, in the redshift range
10. Lanzuisi et al.: AGN vs. host properties in COSMOS
Fig. 11: Fraction of obscured sources as a function of M ∗ (left) and sSFR (right), for the entire sample (black points). The blue (red)dashed points show the results for the first (forth) redshift bin, respectively.1 < ∼ z < ∼ N H from the GRB spectra is probing onlythe host obscuration, the authors conclude that a large fraction ofthe obscuration observed in AGN, at least in the Compton thinregime, is not due to the nuclear torus, but to the galaxy-scalegas in the host.These dependencies imply that at increasing galaxy massthere are more chances to have an additional component to theamount of gas and dust along the line of sight through the AGN.It is well established that the gas fraction is a strong decreas-ing function of the galaxy mass (e.g. Santini et al. 2014; Peng,Maiolino & Cochrane 2015). However, it is possible to showthat the total amount of gas is driven mainly by the total galaxymass, and not by the gas fraction. To this end, we computed gasmass for all our galaxies, following the empirical relation foundin Scoville et al. 2016 (their eq. 1), that links M ∗ , sSFR o ff setfrom the MS, and molecular gas mass. This is shown in Fig. 10,where the sources are color-coded on the basis of their gas frac-tion. Even if at increasing M ∗ the gas fraction is smaller, the totalamount of gas still increases with M ∗ .The well-known mass-metallicity relation (e.g. Tremonti etal. 2004, Mannucci et al. 2010) goes in the direction of havingmore metals (responsible for X-ray absorption) with increasing M ∗ . In particular, going from Log( M ∗ ) = ∼ z ∼ N H observed here: Measuring the N H with fixed metal-licity (as done here) for sources with such a range in metallicity,translates into a factor ∼ ff erence in measured N H , for a giveninput obscuration. To compare our results with the literature, we also looked at thefraction of obscured sources as a function of host properties. InFig. 11 we show the fraction of obscured sources, defined asN
Obs / N Tot where N
Obs is the number of sources with a detection of N H and N H > × cm − . As expected from what shown inthe previous section, the fraction of obscured sources increaseswith increasing M ∗ , and decreases with sSFR (for sSFR > − ). The decrease in sSFR is partly washed out by the factthat we are considering the full redshift interval ( z = . − ff er-ent subsamples shifts toward higher sSFR with redshift. For thisreason we also show in Fig. 11 the results for the first and forthbins (blue and red dashed points, respectively) as an example.Merloni et al. (2014) found a flat relation between the frac-tion of obscured sources and M ∗ in a sample of X-ray detectedAGN from the XMM-COSMOS catalog. However, they limitedtheir analysis to a narrow range in L X (in order to cover a widerange redshift), while the obscured fraction is known to evolvestrongly with L X (e.g. Ueda et al. 2015).Another group, instead, have found an increasing fractionof obscured sources as a function of sSFR and MS o ff set, in asample of 70 µ m selected galaxies at 0 . < z <
1, interpretedas an indication of increasing gas fraction or density in the host,that in turn would sustain the increased sSFR.(e.g. Juneau et al.2013, J13 hereafter).We note that the definition of obscured AGN adopted hereand in J13 are di ff erent, and in the latter, mostly based on thelack of X-ray detection: there are 64 sources (out of 99 AGN)classified as obscured AGN on the basis of the Mass-Excitationdiagram selection (MEX, Juneau et al. 2011), and the X-ray nondetection. If these objects are indeed highly obscured, Compton-thick AGN, this population is mostly missed in our X-ray basedsample.Another possibility is that a fraction of the MEX-selectedAGN are not actively / strongly accreting SMBHs. Indeed, a siz-able fraction ( ∼ M ∗ below Log ( M ∗ ) = .
5. As shownin Sec. 2.2, however, X-ray detected AGN are rare at low M ∗ .Therefore all the sources that are X-ray undetected for reasonsdi ff erent from obscuration (variability, intrinsic weakness, con-taminant non-AGN etc.) would appear as obscured, low M ∗ host(hence high sSFR) AGN, possibly a ff ecting the observed trends.
11. Lanzuisi et al.: AGN vs. host properties in COSMOS
6. Discussion
We collected a large sample of X-ray and FIR detected AGNand host systems in the COSMOS field, spanning ∼ L X , N H , L SFIR , M ∗ , and covering the redshift range0 . < z <
4. We applied X-ray spectral analysis down to verylow counts, ( >
30 net counts) and adopted the SED decomposi-tion results derived in D15, to recover both AGN and SF proper-ties of each source. With this data-set in hand, we demonstratedthat it is possible to reproduce both the flat distribution of aver-age L
SFIR in bins of L X and the steeper correlation of average L X in bins of L SFIR reported in the literature in the latest years (e.g.Shao et al. 2010, Rosario et al. 2012, Mullaney et al. 2012, C13,Stanley et al. 2015).The apparently contradictory results found in the literature,and reproduced in Sec. 3.2, are due to the di ff erent results that areobtained when binning along one axis or the other, the equivalentof a forward or inverse linear regression (i.e. L SFIR — L X vs. L X — L SFIR ), as proposed in Hickox et al. (2014) and Volonteri et al.(2015), and found in Dai et al. (2015) on shallow XMM-LSSdata.Both from a physical and a statistical point of view, it seemsmore appropriate to consider the results from L X — L SFIR , giventhe larger measurement uncertainties on L X , and the shorter timescale variability of L X , with respect to L SFIR , that adds a furtherterm of intrinsic scatter. Doing so, we found a linear correlationbetween L X and L SFIR with slope consistent with 1, at least in theredshift range 0.4-1.2, i.e. below the peak of the SF and BH ac-cretion history. Beyond that and up to z =
4, the slope becomessignificantly flatter, α = . − . ∼
1, at all redshifts. This wouldpoint toward an average one-to-one correlation between SF andBH accretion, in the last 12 Gyr of cosmic history.Even more interesting is the full distribution of BH and hostproperties, such as L X and L SFIR or sBHAR and sSFR, that can beonly qualitatively compared, for the moment, with predictionsfrom galaxy merger simulations, resulting in interesting similar-ities between observations and models.We stress again that these results apply to the small subsam-ple of AGN / host systems detected in both X-ray and FIR, thatrepresents only ∼
20% of the full X-ray sample and ∼
10% ofthe AGN FIR sample. Indeed, one of the main reasons why it isso di ffi cult for present observations to probe the AGN-SF con-nection, is the fact that (X-ray and / or FIR) detected systems spana limited range in AGN and SF activity, sampling only the high L X / SFR tail of the possible correlation, (e.g. Sijacki et al. 2015).It is, however, interesting that we are able to reproduce theresults obtained via stacking of samples where the vast majorityof the sources are not detected (e.g. 20% of FIR detected AGNselected in X-ray in Shao et al. 2010). As suggested in Mullaneyet al. (2015), the stacking analysis, being the equivalent of a lin-ear mean, may be dominated by the brightest sources.A crucial next step in the comparison between theory andobservations will be to select the observed systems in di ff erentevolutionary stage, to reach a similar level of detail as in the cur-rent simulations. This will be feasible for large samples only atlow redshift, while detailed and complete morphological stud-ies in COSMOS (and other deep fields) data are very di ffi cultalready at z > ∼
1. From the theoretical point of view, more de-manding galaxy merger simulations will be required, in order tocover, with the same high resolution, a mass range comparable to the one of observed systems, and to possibly move toward ahigh redshift environment.Finally, a positive correlation between N H and M ∗ , and a sim-ilar negative correlation with sSFR, have been found at all red-shift bins. A similar result was found by Rodighiero et al. (2015)in a large sample of high redshift galaxies, computing HR ofstacked X-ray images. A recent study on GRB hosts has found asimilar behavior (Buchner et al. 2017), implying that an impor-tant fraction (up to 40%) of the Compton thin obscuration foundin AGN can be ascribed to galaxy scale gas (Buchner & Bauer2017).Several studies have found no correlation between columndensity and host properties (Rovilos et al. 2012, Rosario et al.2012), while others (e.g. J13) have found a positive correlation ofthe fraction of obscured sources with sSFR. Further investigationin this direction will help to shed light on the role of the host incontributing to the obscuration through the AGN line of sight. Acknowledgements.
The authors thank the anonymous referee for valuablecomments. GL, MB, and MP acknowledge financial support from the CIGgrant “eEASY” n. 321913. GL acknowledges financial support from ASI-INAF2014-045-R.0. ID acknowledges the European Unions Seventh Framework pro-gramme under grant agreement 337595 (ERC Starting Grant, “CoSMas”). Weacknowledge the contributions of the entire COSMOS collaboration consist-ing of more than 100 scientists. More information on the COSMOS survey isavailable at http://cosmos.astro.caltech.edu/ . Based on observationsobtained with XMM-Newton, an ESA science mission with instruments and con-tributions directly funded by ESA Member States and NASA’, and data obtainedfrom the
Chandra
Data Archive.
References
Aird, J., Coil, A. L., Moustakas, J., et al. 2012, ApJ, 746, 90Akritas, M. G., & Bershady, M. A. 1996, ApJ, 470, 706Alonso-Herrero, A., Pereira-Santaella, M., Rieke, G. H., & Rigopoulou, D. 2012,ApJ, 744, 2Arnaud, K. A. 1996, Astronomical Data Analysis Software and Systems V, 101,17Berta, S., Lutz, D., Santini, P., et al. 2013, A&A, 551, AA100Bennett, C. L., Larson, D., Weiland, J. L., et al. 2013, ApJS, 208, 20Bongiorno, A., Merloni, A., Brusa, M., et al. 2012, MNRAS, 427, 3103Brusa, M., Fiore, F., Santini, P., et al. 2009, A&A, 507, 1277Brusa, M., Civano, F. Comastri, A., et al. 2010, ApJ, 716, 348Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000Buchner, J., Schulze, S., & Bauer, F. E. 2017, MNRAS, 464, 4545Buchner, J., & Bauer, F. E. 2017, MNRAS, 465, 4348Bundy, K., Georgakakis, A., Nandra, K., et al. 2008, ApJ, 681, 931-943Cai, Z.-Y., Lapi, A., Xia, J.-Q., et al. 2013, ApJ, 768, 21Capak, P., Aussel, H., Ajiki, M., et al. 2007, ApJS, 172, 99Capelo, P. R., Volonteri, M., Dotti, M., et al. 2015, MNRAS, 447, 2123Chabrier, G. 2003, ApJ, 586, L133Chen, C.-T. J., Hickox, R. C., Alberts, S., et al. 2013, C13, ApJ, 773, 3Chen, C.-T. J., Hickox, R. C., Alberts, S., et al. 2015, ApJ, 802, 50Civano, F., Elvis, M., Brusa, M., et al. 2012, ApJS, 201, 30Civano, F., Marchesi, S., Elvis, M., et al. 2015, submitted to ApJComastri, A., Ranalli, P., Iwasawa, K., et al. 2011, A&A, 526, L9Cresci, G., Hicks, E. K. S., Genzel, R., et al. 2009, ApJ, 697, 115Dai, Y. S., Wilkes, B. J., Bergeron, J., et al. 2015, arXiv:1511.06761Delvecchio, I., Gruppioni, C., Pozzi, F., et al. 2014, MNRAS, 439, 2736Delvecchio, I., Lutz, D., Berta, S., et al. 2015, arXiv:1501.07602Diamond-Stanic, A. M., & Rieke, G. H. 2012, ApJ, 746, 168Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604Dubois, Y., Peirani, S., Pichon, C., et al. 2016, arXiv:1606.03086Elvis, M., Civano, F., Vignali, C., et al. 2009, ApJS, 184, 158Erb, D. K., Shapley, A. E., Pettini, M., et al. 2006, ApJ, 644, 813Fabian, A. C., & Iwasawa, K. 1999, MNRAS, 303, L34Feigelson, E. D., & Berg, C. J. 1983, ApJ, 269, 400Feltre, A., Hatziminaoglou, E., Fritz, J., & Franceschini, A. 2012, MNRAS, 426,120Fritz, J., Franceschini, A., & Hatziminaoglou, E. 2006, MNRAS, 366, 767Fruscione, A., et al. 2006, Proc. SPIE, 6270Genzel, R., Newman, S., Jones, T., et al. 2011, ApJ, 733, 101Granato, G. L., De Zotti, G., Silva, L., Bressan, A., & Danese, L. 2004, ApJ,600, 580
12. Lanzuisi et al.: AGN vs. host properties in COSMOS
Gri ffi n, M. J., Abergel, A., Abreu, A., et al. 2010, A&A, 518, L3Gruppioni, C., Berta, S., Spinoglio, L., et al. 2016, MNRAS, 458, 4297Hasinger, G., Cappelluti, N., Brunner, H., et al. 2007, ApJS, 172, 29Hickox, R. C., Mullaney, J. R., Alexander, D. M., et al. 2014, ApJ, 782, 9Hogg, D. W., Bovy, J., & Lang, D. 2010, arXiv:1008.4686Imanishi, M., Maiolino, R., & Nakagawa, T. 2010, ApJ, 709, 801Isobe, T., Feigelson, E. D., Akritas, M. G., & Babu, G. J. 1990, ApJ, 364, 104Juneau, S., Dickinson, M., Alexander, D. M., & Salim, S. 2011, ApJ, 736, 104Juneau, S., Dickinson, M., Bournaud, F., et al. 2013, J13, ApJ, 764, 176Kau ff mann, G., Heckman, T. M., Tremonti, C., et al. 2003, MNRAS, 346, 1055Kau ff mann, G., & Heckman, T. M. 2009, MNRAS, 397, 135Kelly, B. C. 2007, ApJ, 665, 1489Kennicutt, R. C., Jr. 1998, ApJ, 498, 541Koekemoer, A. M., Aussel, H., Calzetti, D., et al. 2007, ApJS, 172, 196Kormendy, J., & Richstone, D. 1995, ARA&A, 33, 581Kormendy, J., & Ho, L. C. 2013, ARA&A, 51, 511Lagos, C. D. P., Baugh, C. M., Lacey, C. G., et al. 2011, MNRAS, 418, 1649Lanzuisi, G., Civano, F., Elvis, M., et al. 2013, MNRAS, 431, 978Lanzuisi, G., Ranalli, P., Georgantopoulos, I., et al. 2015a, A&A, 573, AA137(L15)Lanzuisi, G., Perna, M., Delvecchio, I., et al. 2015b, A&A, 578, A120Lusso, E., Comastri, A., Simmons, B. D., et al. 2012, MNRAS, 425, 623 (L12)Lutz, D., Spoon, H. W. W., Rigopoulou, D., Moorwood, A. F. M., & Genzel, R.1998, ApJ, 505, L103Lutz, D., Poglitsch, A., Altieri, B., et al. 2011, A&A, 532, A90Macklin, J. T. 1982, MNRAS, 199, 1119Madau, P., & Dickinson, M. 2014, ARA&A, 52, 415Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115, 2285Mainieri, V., Bongiorno, A., Merloni, A., et al. 2011, A&A, 535, A80Maiolino, R., Shemmer, O., Imanishi, M., et al. 2007, A&A, 468, 979Mannucci, F., Cresci, G., Maiolino, R., Marconi, A., & Gnerucci, A. 2010,MNRAS, 408, 2115Marchesi, S., Civano, F., Elvis, M., et al. 2016, ApJ, 817, 34Marchesi, S., Lanzuisi, G., Civano, F., et al. 2016, ApJ, 830, 100Marconi, A., & Hunt, L. K. 2003, ApJ, 589, L21Marconi, A., Risaliti, G., Gilli, R., et al. 2004, MNRAS, 351, 169Menci, N., Fiore, F., Puccetti, S., & Cavaliere, A. 2008, ApJ, 686, 219Merloni, A., Bongiorno, A., Brusa, M., et al. 2014, MNRAS, 437, 3550Mullaney, J. R., Alexander, D. M., Goulding, A. D., & Hickox, R. C. 2011,MNRAS, 414, 1082Mullaney, J. R., Pannella, M., Daddi, E., et al. 2012, MNRAS, 419, 95Mullaney, J. R., Alexander, D. M., Aird, J., et al. 2015, MNRAS, 453, L83Neistein, E., & Netzer, H. 2014, MNRAS, 437, 3373Novak, G. S., Faber, S. M., & Dekel, A. 2006, ApJ, 637, 96Oliver, S. J., Bock, J., Altieri, B., et al. 2012, MNRAS, 424, 1614Peng, Y., Maiolino, R., & Cochrane, R. 2015, Nature, 521, 192Pilbratt, G. L., Riedinger, J. R., Passvogel, T., et al. 2010, A&A, 518, L1Poglitsch, A., Waelkens, C., Geis, N., et al. 2010, A&A, 518, L2Pontzen, A., Tremmel, M., Roth, N., et al. 2016, arXiv:1607.02507Pozzi, F., Vignali, C., Gruppioni, C., et al. 2012, MNRAS, 423, 1909Rodighiero, G., Daddi, E., Baronchelli, I., et al. 2011, ApJ, 739, L40Rodighiero, G., Brusa, M., Daddi, E., et al. 2015, ApJ, 800, L10Rosario, D. J., Santini, P., Lutz, D., et al. 2012, A&A, 545, A45Rovilos, E., Comastri, A., Gilli, R., et al. 2012, A&A, 546, A58Saintonge, A., Catinella, B., Cortese, L., et al. 2016, arXiv:1607.05289Salvato, M., Hasinger, G., Ilbert, O., et al. 2009, ApJ, 690, 1250Santini, P., Rosario, D. J., Shao, L., et al. 2012, A&A, 540, A109Santini, P., Maiolino, R., Magnelli, B., et al. 2014, A&A, 562, A30Schulze, A., Bongiorno, A., Gavignaud, I., et al. 2015, MNRAS, 447, 2085Scoville, N., Aussel, H., Brusa, M., et al. 2007, ApJS, 172, 1Scoville, N., Sheth, K., Aussel, H., et al. 2016, ApJ, 820, 83Scudder, J. M., Oliver, S., Hurley, P. D., et al. 2016, MNRAS, 460, 1119Shao, L., Lutz, D., Nordon, R., et al. 2010, A&A, 518, L26Sijacki, D., Vogelsberger, M., Genel, S., et al. 2015, MNRAS, 452, 575Silverman, J. D., Lamareille, F., Maier, C., et al. 2009, ApJ, 696, 396Suh, H. Civano, F., Hasinger, G., et al. 2017, submittedSymeonidis, M., Giblin, B. M., Page, M. J., et al. 2016, MNRAS, 459, 257Tremaine, S., Gebhardt, K., Bender, R., et al. 2002, ApJ, 574, 740Tremonti, C. A., Heckman, T. M., Kau ff mann, G., et al. 2004, ApJ, 613, 898Ueda, Y., Akiyama, M., Hasinger, G., Miyaji, T., & Watson, M. G. 2014, ApJ,786, 104Vito, F., Maiolino, R., Santini, P., et al. 2014, MNRAS, 441, 1059Volonteri, M., Capelo, P. R., Netzer, H., et al. 2015a, MNRAS, 449, 1470Volonteri, M., Capelo, P. R., Netzer, H., et al. 2015b, MNRAS, 452, L6Whitaker, K. E., van Dokkum, P. G., Brammer, G., & Franx, M. 2012, ApJ, 754,LL29Xue, Y. Q., Brandt, W. N., Luo, B., et al. 2010, ApJ, 720, 368mann, G., et al. 2004, ApJ, 613, 898Ueda, Y., Akiyama, M., Hasinger, G., Miyaji, T., & Watson, M. G. 2014, ApJ,786, 104Vito, F., Maiolino, R., Santini, P., et al. 2014, MNRAS, 441, 1059Volonteri, M., Capelo, P. R., Netzer, H., et al. 2015a, MNRAS, 449, 1470Volonteri, M., Capelo, P. R., Netzer, H., et al. 2015b, MNRAS, 452, L6Whitaker, K. E., van Dokkum, P. G., Brammer, G., & Franx, M. 2012, ApJ, 754,LL29Xue, Y. Q., Brandt, W. N., Luo, B., et al. 2010, ApJ, 720, 368