MNRAS 000, 1–18 (2023) Preprint 22 December 2023 Compiled using MNRAS LATEX style file v3.

Probing supermassive black hole growth and its dependence on stellar

mass and star-formation rate in low-redshift galaxies
O. Torbaniuk1,2 ∗, M. Paolillo3,4,5, R. D’Abrusco6 , C. Vignali1,2, A. Georgakakis7 , F. J. Carrera8 , F. Civano9
1 Department of Physics and Astronomy ‘Augusto Righi’, University of Bologna, via Piero Gobetti 93/2, I-40129 Bologna, Italy
2 INAF – Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, Via Gobetti 101, I-40129 Bologna, Italy
arXiv:2312.13869v1 [astro-ph.GA] 21 Dec 2023

3 Department of Physics, University of Napoli Federico II, via Cinthia 9, 80126, Napoli, Italy
4 INAF – Osservatorio Astronomico di Capodimonte, via Moiariello 16, 80131, Napoli, Italy
5 INFN – Sezione di Napoli, via Cinthia 9, 80126, Napoli, Italy
6 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA, 02138, USA
7 Institute for Astronomy & Astrophysics, National Observatory of Athens, V. Paulou & I. Metaxa, Athens, 11532, Greece
8 Instituto de Física de Cantabria (CSIC-Universidad de Cantabria), Avenida de los Castros, 39005 Santander, Spain
9 Astrophysics Science Division, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA

Accepted XXX. Received YYY; in original form ZZZ

ABSTRACT

We present an improved study of the relation between supermassive black hole growth and their host galaxy properties in the
local Universe (𝑧 < 0.33). To this end, we build an extensive sample combining spectroscopic measurements of star-formation
rate (SFR) and stellar mass from Sloan Digital Sky Survey, with specific Black Hole accretion rate (sBHAR, 𝜆sBHAR ∝ 𝐿 X /M∗ )
derived from the XMM-Newton Serendipitous Source Catalogue (3XMM-DR8) and the Chandra Source Catalogue (CSC 2.0).
We find that the sBHAR probability distribution for both star-forming and quiescent galaxies has a power-law shape peaking at
log 𝜆sBHAR ∼ −3.5 and declining toward lower sBHAR in all stellar mass ranges. This finding confirms the decrease of AGN
activity in the local Universe compared to higher redshifts. We observe a significant correlation between log 𝜆sBHAR and log SFR
in almost all stellar mass ranges, but the relation is shallower compared to higher redshifts, indicating a reduced availability of
accreting material in the local Universe. At the same time, the BHAR-to-SFR ratio for star-forming galaxies strongly correlates
with stellar mass, supporting the scenario where both AGN activity and stellar formation primarily depend on the stellar mass
via fuelling by a common gas reservoir. Conversely, this ratio remains constant for quiescent galaxies, possibly indicating the
existence of the different physical mechanisms responsible for AGN fuelling or different accretion mode in quiescent galaxies.
Key words: galaxies: active – galaxies: elliptical and lenticular, cD – galaxies: spiral – galaxies: star formation – X-ray: galaxies
– accretion, accretion discs

1 INTRODUCTION 2023; Poitevineau et al. 2023; Sahu et al. 2023). However, the phys-
ical origin of such correlations is still poorly understood.
The growth of galaxies (via stellar formation processes) and of the
supermassive black hole (SMBH) at their centers (via mass-accretion Since the rate of star formation on galactic scales is directly re-
potentially triggering an Active Galactic Nucleus or AGN) appear to lated to the availability of cold gas (and its efficiency in forming stars,
proceed coherently over cosmic times. As was suggested by several Bigiel et al. 2008; Peng & Maiolino 2014; Catinella et al. 2018), it
studies, the global cosmological star-formation rate and AGN accre- is reasonable to speculate that the nuclear activity can be fed by
tion rate show similar evolution with redshift, reaching a peak at red- the same gas reservoir being accreted onto the central SMBH. This
shift 𝑧 ∼ 1−3 and declining rapidly towards more recent cosmic times scenario also agrees with studies showing that moderate-to-high lu-
(Delvecchio et al. 2014; Madau & Dickinson 2014; Aird et al. 2015; minosity AGN predominately reside in galaxies with higher star-
Malefahlo et al. 2022; D’Silva et al. 2023). At the same time, the formation rates (Merloni et al. 2010; Rosario et al. 2013; Heinis et al.
mass of central SMBHs seem to be tightly correlated with the proper- 2016; Aird et al. 2018; Stemo et al. 2020; Torbaniuk et al. 2021).
ties of their host galaxies (e.g. stellar velocity dispersion, bulge mass, Still, as AGN accretion operates typically on smaller spatial scales
total stellar mass; see Ferrarese & Merritt 2000; Gültekin et al. 2009; than star formation, the confirmation of the presence of the com-
McConnell & Ma 2013; Kormendy & Ho 2013; Reines & Volonteri mon gas reservoir for SMBH accretion and star formation re-
2015; Shankar et al. 2016; González-Lópezlira et al. 2022; Li et al. quires a deeper understanding of the mechanism responsible for
gas transportation from the galaxy outskirts all the way to their
centers. Individual observations of the closest galaxies and hydro-
dynamical simulations suggest that such mechanism can be pro-
∗E-mail: [email protected] vided by large-scale gravitational torques formed by disk insta-

© 2023 The Authors

2 O. Torbaniuk et al.
bilities or by major mergers and minor interactions among galax- the lack of deep, uniform, wide-areas X-ray surveys most stud-
ies (Fathi et al. 2006; Hopkins & Quataert 2010; Fischer et al. 2015; ies on the co-evolution between galaxies and their central SMBH
Storchi-Bergmann & Schnorr-Müller 2019; Quai et al. 2023). In ad- have focused on intermediate/high redshift from 𝑧 ∼ 0.25 up to
dition, the feedback provided by stellar evolution and AGN activity 𝑧 ≈ 4.0 (Chen et al. 2013; Rosario et al. 2013; Delvecchio et al.
may be pivotal in controlling the amount and distribution of cold gas 2015; Rodighiero et al. 2015; Aird et al. 2018, 2019; Stemo et al.
and consequently alter AGN growth. For instance, the stellar feed- 2020; Spinoglio et al. 2022; Pouliasis et al. 2022) and therefore, the
back can significantly affect nuclear accretion reducing the gas supply link between BH accretion rate and SFR (or stellar mass) have been
by star formation or, on the contrary, enhancing it through the turbu- investigated mainly for the moderate- and high-luminosity AGNs.
lence injection produced by supernova explosions and/or strong wind Understanding the BHAR–SFR relation also in local galaxies, in
from massive stars (Schartmann et al. 2009; Hopkins et al. 2016; addition to probing the bulk of the accreting BH population, i.e.
Byrne et al. 2023). At the same time, accretion onto the SMBH can low-to-moderate luminosity AGN, also provides information on the
generate energetic outputs in the form of electromagnetic radiation physics of low efficiency SMBH accretion, which is difficult to trace
(i.e. radiative or quasar mode) or powerful jets (i.e. radio or jet mode) at higher redshifts. Moreover, the population of galaxies with fad-
depending on the accretion efficiency (> 1 and ≪ 1 per cent Ed- ing star formation (i.e. quiescent galaxies) in the local Universe is
dington, respectively; Fabian 2012; Heckman & Best 2014). Hence, a crucial laboratory to evaluate the role of AGN activity in star-
interacting with the gas in the host galaxy through radiation pres- formation suppression and to explore the alternative mechanisms of
sure that produces a powerful wind (quasar mode) or an outflow of AGN fuelling in environments with a low reservoir of cold gas.
relativistic particles (jet mode) AGN can suppress the formation of
stars by heating and/or blowing the cold gas away from the galaxy In Torbaniuk et al. 2021 (hereinafter Paper I) we presented a first
or, vice versa, trigger it by compressing dense clouds in the in- study of the correlation between star formation and AGN activ-
terstellar medium, Schawinski et al. 2009; Ishibashi & Fabian 2012; ity in the local Universe (𝑧 < 0.33) using a homogeneous Sloan
Zubovas et al. 2013; Leslie et al. 2016; Combes 2017; Fiore et al. Digital Sky Survey (SDSS DR8) optical galaxy sample with ro-
2017; Park et al. 2023; Ferrara et al. 2023). In addition to being able bust SFR (in the range 10−3 to 102 M ⊙ year−1 ) and M ∗ estimates
to quench star formation in the host galaxy, AGN feedback can also (from 106 to 1012 M ⊙ ) in combination with X-ray data from XMM-
significantly reduce its own accretion, thus resulting in self-regulation Netwon Serendipitous Source Catalogue (3XMM-DR8). This al-
of the nuclear activity (Fabian 2012; Paspaliaris et al. 2023). lowed us to estimate the specific BH accretion rate 𝜆 sBHAR , trac-
However, all mechanisms of gas transportation discussed above ing the level of AGN activity per unit stellar mass of the host
are not able to produce the continuous and regular gas flow required galaxy. We found that the local Universe contains a low fraction
to feed the SMBH and therefore, they trigger AGN activity in a of efficiently accreting SMBH, while the majority of local SMBH
much more stochastic manner compared to the smooth and coherent accrete at very low rates. We observed a significant correlation be-
process of star formation. Such stochasticity is usually observed as tween sBHAR–SFR for almost all stellar masses, but the population
the variability of the AGN activity, which may lead to the change of SMBH hosted by quiescent galaxies is accreting at systemat-
of AGN luminosity by orders of magnitude on intermediate-long ically lower levels than star-forming systems. This work however
timescales of ∼ 105−7 yr (see Mullaney et al. 2012a; Aird et al. 2013; suffered from the limited size of the sample and from the low reso-
King & Nixon 2015; Sartori et al. 2018). Thus, observations of any lution of XMM data, which could affect the estimate of the intrin-
individual AGN are probing only a fraction of its activity cycle and sic nuclear activity and thus the estimate of the sBHAR, especially
therefore do not allow to study directly the connection between the for low-luminosity AGN. Thus, in order to confirm the results ob-
growth of the central SMBH and the overall galaxy properties; this tained in Paper I, in the present work we improve the analysis us-
agrees with recent observations swing that galaxies with similar stel- ing also X-ray data from the Chandra Source Catalogue (CSC2.0).
lar masses and SFR contain AGN with a very broad range of accre- Since the Chandra telescope has the highest resolution among all
tion rates (Bongiorno et al. 2012; Aird et al. 2012; Georgakakis et al. X-ray telescopes available nowadays, it allows us to better discrim-
2014; Aird et al. 2018; Torbaniuk et al. 2021). Therefore, the proper inate the nuclear source from the host-galaxy contribution, and test
investigation of the relation between galaxy properties and AGN ac- our previous results. Furthermore, the combined 3XMM+CSC2.0
tivity requires the study of representative samples of the whole AGN dataset represents the largest serendipitous X-ray survey of the lo-
population. The availability of large statistical samples of galaxies al- cal Universe and will represent a reference sample both for high-
lows to probe the AGN activity over cosmological scales through the z studies as well as for next-generation surveys such as eROSITA
BHAR probability density function, i.e. the probability of a galaxy (Comparat et al. 2022; Liu et al. 2022; Mountrichas et al. 2022b;
with certain properties (e.g. stellar mass, SFR, morphological type) Aspegren et al. 2023; Comparat et al. 2023; Mountrichas & Shankar
to host a SMBH accreting at a certain rate. According to recent 2023).
works, the BHAR probability function seems to have a power-law
shape with an exponential cut-off at high BHAR and with flatten- This paper is organised as follows. In Section 2 of this paper we
ing or even decreasing toward low accretion rates (Aird et al. 2012; present a short description of the primary catalogue of host galaxy
Bongiorno et al. 2012; Aird et al. 2018). properties from SDSS DR8. The extraction and correction of new X-
Studies of such correlation between AGN activity and host galaxy ray data from CSC2.0, their cross-calibration with 3XMM-DR8 data,
properties require careful separation of the nuclear and stellar emis- and the completeness of the obtained X-ray sample are discussed in
sion. The detection of X-ray radiation produced by the innermost Section 3. In Section 4 we present the intrinsic sBHAR probability
regions of active nuclei is an efficient method for probing SMBH distribution for star-forming and quiescent galaxies with different
accretion over a wide range of redshifts. Moreover, it allows us to stellar masses. In addition, we study X-ray luminosity and 𝜆 sBHAR
access even relatively low-luminosity AGN in the local Universe, distributions as a function of stellar mass and galaxy properties as
where the identification in the optical and infrared bands is not triv- well as the correlation between SFR and 𝜆 sBHAR . We summarise
ial due to the contamination of the host galaxy (Brandt & Hasinger our findings in Section 5. Throughout this paper, we adopt a flat
2005; Alexander & Hickox 2012; Merloni 2016). However, due to cosmology with ΩΛ = 0.7 and H0 = 70 km s−1 Mpc−1 .

MNRAS 000, 1–18 (2023)

 SMBH growth in low-z galaxies 3
2 THE SDSS GALAXY SAMPLE optical and IR bands is challenging due to the dominance of the host
galaxy emission in these bands. At the same time, the host galaxy’s
The results in this paper are based on the same initial optical galaxy
contribution to the total X-ray emission is generally smaller allowing
sample used in Paper I, described fully therein and briefly sum-
to use of the X-ray as a robust technique for the AGN identifica-
marised here. This sample is based on the galSpec catalogue of
tion. In Paper I we used 1953 sources with reliable photometry3 and
galaxy properties1 produced by the MPA–JHU group as the sub-
point-like X-ray morphology from the XMM-Newton Serendipitous
sample from the main galaxy catalogue of the 8th Data Release
Source Catalogue (3XMM-DR8, Rosen et al. 2016); their distribu-
of the Sloan Digital Sky Survey (SDSS DR8). The stellar masses
tion on SFR–M ∗ plane is shown at the top panel of Figure 1. In this
(M ∗ ) were obtained through Bayesian fitting of the SDSS ugriz pho-
work, we present a similar analysis using the X-ray data from Chan-
tometry to a grid of models (see details in Kauffmann et al. 2003a;
dra X-ray Observatory. It has the highest resolution among all X-ray
Tremonti et al. 2004). The estimates of the SFR were done in two
telescopes available nowadays and therefore, allows us to better dis-
different ways depending on the object classification according to
criminate the nuclear source from the host-galaxy contribution. To
the BPT criteria (Baldwin et al. 1981). The values of SFRs for star-
increase the number of studied objects in our sample and improve the
forming galaxies were determined using the H𝛼 emission line lumi-
statistics of the studied relations we will later combine the data from
nosity (Brinchmann et al. 2004), while for all other spectral classes
CSC2.0 and 3XMM-DR8 in the final X-ray sample (see Section 3.2).
(e.g. AGN, composite and unclassified objects) the empirical rela-
tion between SFR and the Balmer decrement, D4000, was used (see
details in Kauffmann et al. 2003b).
The entire galSpec catalogue provides information for about 1.5 3.1 The Chandra data set selection
million galaxies with redshift 𝑧 < 0.33, but in our work we se- We extracted our data from the Chandra Source Catalogue4
lected only objects with reliable spectroscopic parameters (i.e. with (CSC 2.0, Evans et al. 2010). Comparing our SDSS galaxy sam-
RELIABLE != 0) and redshift estimate (i.e. with zWarning = 0). Fur- ple (presented in Section 2) and the Chandra footprint we found that
thermore, we excluded duplicates and objects with low-quality SDSS 22 836 objects from the SDSS sample are falling in the area of the
photometry using the basic photometric processing flags2 . sky observed by the Chandra X-ray Observatory.
To establish the level of AGN activity for different galaxy pop- In order reduce the X-ray sample size, we used the CSCview tool
ulations, all galaxies in our sample have been classified as ‘star- to select all potential X-ray counterparts within a conservative ra-
forming’ (SFGs) or ‘quiescent’ according to their position on the dius of 10 arcsec, to be sure to include also poorly resolved sources
SFR–M ∗ plain (Fig. 1). Since SFG are found to follow a rela- far from the Chandra optical axis; CSCview provided us with co-
tively tight correlation between the current SFR and M ∗ , the so- ordinates, as well as positional uncertainties (i.e. the position error
called main sequence (MS) of star-formation (Noeske et al. 2007; ellipses5 ), for 2 271 sources. Finally, we refined the cross-match re-
Blanton & Moustakas 2009; Speagle et al. 2014; Renzini & Peng quiring the (optical vs X-ray) source position distance to be less
2015; Tomczak et al. 2016; Santini et al. 2017; Pearson et al. 2018; than the sum of the Chandra positional errors (err_ellipse_r0,
Huang et al. 2023), these two classes can be separated using the po- err_ellipse_r1 and err_ellipse_ang) and the SDSS positional
sition of each individual galaxy relative to the evolving MS of SFG. uncertainty of 0.18 arcsec (i.e. the SDSS fiber position uncertainty,
Similarly to the approach used in Paper I, we set the threshold be- Pier et al. 2003)6 , reducing the number of objects to 1 512. The false
tween the two classes 1.3 dex below the main sequence defined by match rate for such crossmatch, estimated by randomly shifting our
Aird et al. (2017), as follows: sources by 20 arcsec and repeating the match, is near 0.4 per cent (i.e.
6 sources for our sample).
log SFRcut (𝑧) [M ⊙ year −1 ] =
To avoid including spatially extended objects (e.g. hot gas regions
= −7.6 + 0.76 log M ∗ /M ⊙ + 2.95 log(1 + 𝑧). (1) or galaxy clusters) we only selected objects with zero extension pa-
Galaxies that fall above the above threshold (shown by a black band rameter (extend_flag == 0)7 . Finally, we choose only the objects
in Fig. 1) were classified as star-forming while those below as quies- with available detection in the hard band (2.0–7.0 keV) where the
cent. Note that the relation in Equation (1) is redshift-dependent, so AGN emission is likely dominant. As a result, our final CSC-SDSS
the galaxy classification was done considering the redshift of each sample contains 912 objects (454 star-forming and 458 quiescent
individual object, and the range of the thresholds corresponding to galaxies); their distribution on SFR–M ∗ plane is shown at the bot-
galaxies in our redshift range. tom panel of Figure 1.
Our final sample consists of 703 422 galaxies, of which 376 938
are classified as star-forming (53.6 per cent) and 326 484 as quiescent
galaxies (46.4 per cent), respectively. 3
We selected detections from the most sensitive PN camera, however, when
the data from PN camera were missing, we used those from MOS1 or MOS2
cameras.
4 https://cxc.cfa.harvard.edu/csc/
3 X-RAY AGN SAMPLE 5 https://cxc.cfa.harvard.edu/csc/columns/positions.html
6 We used the TOPCAT sky ellipse matching algorithm which creates a
In order to quantify the AGN activity we need to distinguish the
position ellipse for each source based on the given central position, ma-
nuclear emission produced by the accretion of the material onto the
jor and minor radii, and position angle of the ellipse for both databases
central SMBH from the stellar emission of the host galaxy, espe- and compares these elliptical regions on the sky for overlap. See details at
cially in the circumnuclear region. Since the local Universe contains http://www.starlink.ac.uk/topcat/ and Taylor (2005).
mainly low-to-moderate luminosity AGN, their identification in the 7 However, we did an additional crossmatch of spatially extended sources

in the CSC2.0 (extend_flag != 0) with our 3XMM-SDSS sample from

Paper I to check how many of them were previously included to our study as
1 https://www.sdss.org/dr12/spectro/galaxy_mpajhu/ ‘point-like’ due to smaller resolution of XMM-Newton compared to Chandra
2 https://www.sdss.org/dr12/algorithms/photo_flags_recommend/ observatory (see more details in Section 3.2).

MNRAS 000, 1–18 (2023)

 4 O. Torbaniuk et al.

Number1 of objects in2SDSS galaxy sample AGN spectrum with the absorption column density log 𝑁H /cm−2 ≃
100 10 10 103 104 22.5 (also see Tozzi et al. 2006; Liu et al. 2017).

2 3.2 Cross-calibration of Chandra and XMM-Newton data

To increase the number of objects in our sample we decided to com-
1 bine the sample compiled from CSC 2.0 catalogue (see the previous
section) together with the 3XMM-SDSS sample from our Paper I
(see sample #4 defined in table 1) compiled based on 3XMM-Newton
log SFR [M /year]

Serendipitous Source Catalogue (3XMM-DR8). However, before us-

0 ing the data of these samples together we need to cross-calibrate their
fluxes. In order to do so, we used 170 objects from our SDSS sam-
ple, which were detected in the hard band both by Chandra and
-1 XMM-Newton observatories. However, a fraction of nearby galaxies
detected as point-like sources in the 3XMM sample may be resolved
as spatially extended sources by Chandra because of its higher res-
olution compared to XMM-Newton. We found that an additional 42
-2 point-like sources from our 3XMM-SDSS sample have extended
counterparts in the CSC2.0 catalogue. At this stage, we decided to
XMM (N = 1781) include these 42 sources in our cross-calibration analysis, but they
XMM & CSC (N = 170)
-3 will be excluded from further studies (i.e. not included in our final
CSC-SDSS sample) because of their X-ray host galaxy contamina-
2 tion; as a result, we have 212 objects for the cross-calibration analysis.
Since the hard band is represented in slightly different energy
ranges in CSC2.0 (2–7 keV) and in 3XMM (2–12 keV) we first

1 rescaled all 3XMM fluxes to the energy band used in the CSC2.0.
For this, using WebSpec tool9 we simulated two spectra within en-
ergy ranges of 2–12 keV and 2–7 keV assuming a simple power-
log SFR [M /year]

law model with spectral index Γ ∼ 1.4. As a result, we ob-

0 tained that the total fluxes obtained for two spectra correlate as
𝑓 [2−12 keV] / 𝑓 [2−7 keV] = 1.72. This value was used as a scaling
coefficient to transform all 3XMM fluxes available in our sample to

-1 the 2–7 keV energy range.

Even so, significant discrepancies may still exist between Chandra
and XMM-Newton due to calibration uncertainties. Therefore, for a
proper combination of the data, we used the flux relations obtained
-2 by Tsujimoto et al. (2011) in the cross-calibration analysis of X-
ray detections of the pulsar wind nebulae G21.5–0.9. In our work
CSC (N = 742) we calibrate fluxes from the three XMM-Newton pn, MOS1 and
XMM & CSC (N = 170)
-3 8 9 10 11 12
MOS2 cameras relative to the flux from Chandra ACIS camera as
𝑓ACIS / 𝑓PN = 1.194, 𝑓ACIS / 𝑓MOS1 = 1.108 and 𝑓ACIS / 𝑓MOS2 =

log[M⊙ /M ] 1.128. These calibration coefficients are also in agreement with the
results of another cross-calibration analysis based on the sample of
Figure 1. The distribution of star-formation rate vs. stellar mass for our final galaxy clusters (Nevalainen et al. 2010).
SDSS galaxy sample. The grey shaded band shows the main sequence (MS) The comparison between Chandra fluxes (not-calibrated) and
of star-forming galaxies defined by Eq. (1). The black shaded band represents XMM-Newton fluxes calibrated with the coefficients mentioned above
a cut 1.3 dex below the MS of SFG used for the division of the studied sample for 212 individual sources is shown in Fig. 2. The difference between
into star-forming and quiescent galaxies. Both areas correspond to the redshift Chandra and XMM-Newton fluxes is 0.4 dex on average, exceeding
interval of our SDSS sample 𝑧 = 0.00 − 0.33. The individual objects with 1 dex only for a few extreme cases. The observed scatter is due to the
X-ray detection in the hard band in 3XMM-DR8 (top) and CSC2.0 (bottom) internal uncertainties of each catalogue (the average errors for high,
catalogues are shown by blue crosses and green pluses, respectively. Red
medium, and low flux values are presented by grey markers in Fig. 2),
circles show individual objects that have X-ray detection in both catalogues.
combined with the uncertainties in the flux conversion factors and
the scatter introduced by variability of individual nuclear sources. In
To compute the rest-frame X-ray luminosities in the hard band (2– conclusion, we assume that the two calibration steps described above
7 keV) we use the aperture-corrected net energy flux in the ACIS hard are sufficient for our study, as the residual systematics will not affect
(2–7 keV) energy band (flux_aper_h) available in the CSC Master our final results (see Section 4).
Source Table8 . Following the same step as in Paper I we also applied Based on the obtained results, we applied the same calibration
a K-correction for each value of X-ray luminosities: we assume a corrections to the 3XMM fluxes of 1 741 from 1 953 objects detected
photon index Γ = 1.4 which corresponds to a moderately obscured only in the 3XMM-SDSS sample. The rest of the objects have been

8 https://cxc.cfa.harvard.edu/csc/organization.html 9 https://heasarc.gsfc.nasa.gov/webspec/webspec.html

MNRAS 000, 1–18 (2023)

 SMBH growth in low-z galaxies 5
(2006); Boroson et al. (2011); Kim & Fabbiano (2013); Civano et al.
Star-formation galaxies (2014); Jones et al. (2014). To eliminate the contribution of LMXB
Npoint=96, N =23
ext
and AB+CV we used a relation of 𝐿 X and galaxy luminosity in
Quiescent galaxies
log Flux (XMM, calibrated), erg cm−2 s−1

-12 Npoint=74, N =19

ext
the 𝐾𝑆 -band from Boroson et al. (2011), while for the hot gas, we
chose to use 𝐿 X − 𝐿 KS relation defined in Civano et al. (2014). The
galaxy luminosity in the 𝐾𝑆 -band was calculated based on 𝐾𝑆 mag-
nitudes from the 2MASS10 Point and Extended Source Catalogues
-13 (Skrutskie et al. 2006). The detailed description of 𝐾𝑆 -band lumi-
nosity calculation and all applied relations are presented in Paper I.
In this work we label an object as an AGN if its observed ‘total’
X-ray luminosity exceeds the value of the predicted for its host galaxy
-14 derived as explained above. Thus to isolate the AGN contribution, we
subtracted the host-galaxy contribution from the total X-ray luminos-
ity of the source. As a result, we identify 2 223 AGN with positive
residual X-ray luminosity after the correction 1 449 of which are
-15 1:1 3XMM-only sources and 774 are CSC2.0 sources (the latter includ-
point-like in CSC & XMM ing 144 sources in common with 3XMM sources). The total X-ray
point-like in XMM, extended in CSC
luminosity (and the value of correction for the host galaxy contribu-
-15 -14 -13 -12 tion) vs redshift distribution in the hard band, for star-forming and
log Flux (CSC), erg cm−2 s−1 quiescent galaxies, are shown in Fig. 3. Since the same distribution
Figure 2. The relation between the hard-band (2–7 keV) fluxes in CSC2.0 for 3XMM sources has been already shown in Paper I, in this pa-
and 3XMM-DR8 catalogue for 212 galaxies from our SDSS sample (119 per we present distributions only for objects with available detection
star-forming and 93 quiescent galaxies). 170 sources detected as point-like in the CSC2.0 catalogue. It is worth mentioning that both the ob-
both in the 3XMM and the CSC2.0 are presented by circles, while squares served X-ray luminosity and the scaling relations used to evaluate the
show 42 sources detected as point-like in the 3XMM and as an extended in the host galaxy contribution have their own uncertainties, and therefore
CSC2.0. The 3XMM flux is calibrated in accordance with the CSC2.0 flux sources with relatively low luminosities can be falsely excluded (or
considering the systematic difference between the two X-ray observatories
included) from our sample using the criterion of the positive residual
(see details in the text).
X-ray luminosity. We found that changing the threshold by the un-
certainty in the scaling relations affects only 1.5 per cent of SFG and
detected also by Chandra, so for 170 ‘true’ point-like objects (both in 5.3 per cent of quiescent galaxies in our sample, while 6.7 per cent
the 3XMM-DR8 and CSC2.0) we further use the hard band flux from of all sources have luminosities consistent with the threshold within
the CSC2.0. As a result, the combined sample contains 2 653 sources. the photometric errors. Thus the fraction of objects potentially af-
Hereinafter, we refer to the sources with only 3XMM detections as fected by these uncertainties is relatively small (11.4 per cent in total);
3XMM sources and those with CSC2.0 detections as CSC2.0 sources, furthermore, the effect due to the scaling relation uncertainty is sys-
even though some of them also have 3XMM detections. tematic in nature, while the photometric uncertainties are statistical,
so the results presented in Section 4 will be scarcely affected.
3.3 Contribution of the host galaxy to the total X-ray emission
To assess the level of AGN activity from the X-ray emission we 3.4 The stellar mass completeness of CSC+XMM sample
need first to determine the contribution of the host galaxy to the total Our SDSS galaxy sample is magnitude-limited to a Galac-
X-ray luminosity. The Chandra telescope has a higher resolution tic extinction-corrected Petrosian magnitude of 𝑟 = 17.5 (see
than XMM-Newton and is able to more easily separate the nuclear SDSS DR7 Target selection page11 ). As a result, star-forming and
emission from the host-galaxy contribution, the faintest objects or the quiescent galaxies have different M ∗ limits because of the differ-
galaxies at greater distances (i.e. smaller angular size of the object). ent mass-to-light ratios of their stellar populations. This introduces
In principle, we should estimate the fractional contribution of the host incompleteness as quiescent galaxies of a given stellar mass drop
galaxy within the region where the X-ray flux is measured; however, out of the sample at lower redshifts compared to SFG of the same
since the scaling relations we use are computed for the entire galaxy stellar mass. This source of bias is well visible plotting stellar mass
and the X-ray flux extraction radii are different for each source based as a function of redshift separately for star-forming and quiescent
on the position within the telescope field of view, we followed a galaxies (see Figure 4). To minimise this effect of bias we used
simpler approach already used for 3XMM data in Paper I estimating the same approach as proposed in Georgakakis et al. (2014) apply-
an upper limit to the corrections assuming that the contribution is ing a redshift-dependent mass limit which corresponds to a max-
due to the entire galaxy. imally old (i.e. maximal mass-to-light ratio) galaxy. We used the
Since our sample contains both star-forming and quiescent galax- Bruzual & Charlot 2003 model of the mass-to-light ratio evolution
ies (see definition in Section 2) we need to consider the contribu- considering Kroupa IMF (Kroupa 2001) and for each redshift, we
tions of different galaxy components to its total X-ray emission. For calculated the value of stellar mass that corresponds to an observed
star-forming galaxies we calculated the expected X-ray luminosi- magnitude of 𝑟 = 17.5 mag (see solid grey line in Fig. 4). Above this
ties (𝐿 X,host ) due to low and high-mass X-ray binaries (LMXB and limit, the galaxy sample should not be affected by the incompleteness
HMXB) based on the scaling relation between 𝐿 X and SFR, stellar of the survey because all galaxies have a mass-to-light ratio smaller
masses M ∗ and redshift 𝑧 of galaxies from Lehmer et al. (2016).
On the contrary, quiescent galaxies are dominated by hot gas, low-
mass X-ray binaries (LMXB), and some emission from coronally 10 https://old.ipac.caltech.edu/2mass/
active binaries (ABs) and cataclysmic variables (CVs); see Fabbiano 11 https://classic.sdss.org/dr7/products/general/target_quality.html

MNRAS 000, 1–18 (2023)

 6 O. Torbaniuk et al.

44
log L X [2−7 keV] (erg s −1 )

43

42

41

40

39

38 SFG: before corrections (N = 429) Quiescent: before corrections (N = 345)

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30
Redshift, z Redshift, z
Figure 3. The X-ray luminosity vs redshift distribution of 774 objects with positive residual X-ray luminosity from CSC-SDSS sample. The observed (uncorrected)
LX values for SFGs and quiescent galaxies are presented as blue and red circles, respectively. The change in LX due to the subtraction of the predicted host
galaxy contribution (i.e. LX,host , see description in the text) for each object is shown by a solid line.

12.0 4 THE SPECIFIC BLACK HOLE ACCRETION RATE

Following the approach presented in Paper I we calculate the spe-
11.5 cific Black Hole accretion rate (𝜆 sBHAR ), the rate of accretion onto
the central SMBH scaled relative to the stellar mass of the host
11.0
log[M∗ /M⊙ ]

galaxy. We followed the definition from Bongiorno et al. (2012,

2016); Aird et al. (2018):
10.5
𝑘 bol 𝐿 X,hard [erg s−1 ]
10.0 𝜆 sBHAR =
1.3 · 1038 [erg s−1 ] × 0.002 M ∗ /M ⊙
, (2)

9.5 Star-forming galaxies

where 𝑘 bol is a bolometric correction factor for the hard band, 𝐿 X,hard
Quiescent galaxies XMM is the 2–7 keV X-ray luminosity. Although the bolometric correc-
9.0 CSC tion factor is dependent on the luminosity (Marconi et al. 2004;
0.05 0.10 0.15 0.20 0.25 0.30 Lusso et al. 2010, 2012; Bongiorno et al. 2016; Duras et al. 2020),
Redshift, z here we adopted an average bolometric correction of 𝑘 bol = 25
since our sample is not probing the range of high bolometric lumi-
Figure 4. The host galaxy stellar mass vs redshift distribution of combined
CSC+XMM sample, where sources with X-ray detection in CSC2.0 and
nosities, where 𝑘 bol increases significantly with respect to 25, and
3XMM presented by crosses and circles, respectively. Red and blue colors thus the other systematics discussed below dominate the final uncer-
show quiescent and star-forming galaxy populations, respectively. The grey tainty. We assumed that the Black Hole mass scales with the host
solid curve shows the redshift-dependent mass limit for a galaxy with a galaxy stellar mass as M BH = 0.002 M ∗ /M ⊙ as in Häring & Rix
maximally old stellar population at limiting SDSS magnitude 𝑟 = 17.5 mag. (2004). The additional scale factors are defined from the request that
The horizontal dashed and solid black lines show the limit of 1010 M ⊙ and 𝜆 sBHAR ≈ 𝜆 Edd , where the Eddington ratio 𝜆 Edd ∝ LX /M BH .
0.2 dex above the derived SDSS mass limit, respectively, used to remove the
part of the sample with significantly lower fraction of X-ray detected quiescent
galaxies with respect to SFG (see the text for details). 4.1 The sBHAR distribution as a function of stellar mass
To study the distribution of 𝜆 sBHAR in the local Universe we measure
𝑝(log 𝜆 sBHAR |M ∗ ), which represents the probability density func-
tion (PDF) that a galaxy with a certain stellar mass hosts a SMBH
accreting with a given 𝜆 sBHAR . In what follows, 𝑝(log 𝜆 sBHAR |M ∗ )
than that of the maximally old stellar population model. However, was calculated for the full galaxy sample as well as separately for star-
this limit refers to the entire SDSS galaxy sample and does not rep- forming and quiescent galaxies, allowing to evaluate the difference
resent the ‘real’ limit for our subsample of X-ray detected sources. in AGN activity for galaxy populations characterised by different
In order to allow a fair comparison, we chose to exclude regions in morphology, star-formation history, gas content and transportation
Fig. 4 where the fraction of X-ray detected quiescent galaxies is sig- processes.
nificantly lower than the fraction of SFG. Thus, we limit our study to
higher mass values > 1010 M ⊙ and 0.2 dex above the derived SDSS
4.1.1 Methodology
limit (see dashed and solid black lines in Fig. 4).
As a result, the final XMM+CSC sample contains 1 938 objects As we mentioned before, while our X-ray sample contains sources
(967 star-forming and 971 quiescent galaxies). above a certain detection likelihood in the full band, to compute the

MNRAS 000, 1–18 (2023)

 SMBH growth in low-z galaxies 7
1.0
intrinsic nuclear accretion rate distribution we use sources reliably 10 0 log[M∗ M ] 10 5
. < / < . 10 5 log[M∗ M ] 11 0
. < / < .

detected in the hard band, so that we can compute a proper complete- 0.8

source fraction
ness correction. In order to consider the fraction of missed sources
0.6
as a function of flux, we need to account for the sensitivity varia-
tions of the X-ray observations covering our SDSS galaxy sample 0.4

X-ray
across the sky (due to different detector efficiency, exposure time, 0.2 XMM: PN
XMM: MOS1
off-axis angle, etc). Since our CSC+XMM sample is a mixture of XMM: MOS2
0.0
two X-ray catalogues, the sensitivity corrections are estimated sepa-
1.0
rately for each X-ray sample. For sources with X-ray detection only M∗ M ]
11.0 < log [ / < 11.5 log [ M∗ M ]
/ > 11.5

in the 3XMM-DR8 catalogue, only 405 of 1 257 sources have a re- 0.8

X-ray source fraction

liable likelihood detection DET_ML > 6 (3𝜎) in the hard band. We 0.6
collected the values of the survey flux sensitivity limit (from XMM
0.4
FLIX, Carrera et al. 2007) at the position of each source in our orig-
inal optical sample falling in the 3XMM footprint. The cumulative 0.2 CSC: marginal
curves in four stellar mass ranges are shown in Figure 5. In essence, CSC: true

the three curves describe the likelihood of detecting the X-ray coun-
0.0
-15 -14 -13 -12 -15 -14 -13 -12
log (Flux Upper/Sensitivity Limit) erg cm⊙ s⊙ 2 1
log (Flux Upper/Sensitivity Limit) erg cm ⊙2 s⊙1
terpart of our galaxies at each flux level in each of the three XMM
cameras (pn, MOS1, and MOS2). These cumulative values were ap- Figure 5. The cumulative histogram of flux upper-limits in the hard band
(2–12 keV) for three XMM cameras (pn, MOS1 and MOS2 cameras by red,
plied as statistical weights to the number of sources used to compute
blue and green color, respectively) from the XMM FLIX service and of flux
𝑝(log 𝜆 sBHAR |M ∗ ); see later Figure 6.
sensitivity-limits in the hard band (2–7 keV) for MARGINAL (black solid) or
On the other hand, the source detection process in the CSC2.0 TRUE (black dashed) detections in the CSC2.0 for four stellar mass ranges.
catalogue does not provide the source detection likelihood directly,
but the classification as FALSE, MARGINAL or TRUE from the
analysis of stacked images12 . Therefore, we used the 427 sources PDF was normalized using the statistical weight from the cumula-
(out of 681 with X-ray detection in the CSC2.0), with hard fluxes tive curves in Figure 5. As a result, both uncorrected and corrected
higher than the corresponding MARGINAL or TRUE flux sensitiv- 𝑝(log 𝜆 sBHAR |M ∗ ) are presented Figure 6 by dashed and solid lines,
ity limits provided by CSCview service. The two additional CSC respectively.
cumulative curves in Figure 5 represent the likelihood of detecting Finally, to quantify the detection limit of our data, we estimated
the MARGINAL or TRUE X-ray counterpart of our galaxies with a the minimum sBHAR log 𝜆 sBHAR,min , that can either be detected
given flux in the hard band in the CSC2.0. These cumulative curves with XMM-Newton or corresponding to a ‘marginal’ detection in
were applied as statistical weights to the CSC2.0 detected sources, CSC2.0. In order to do this, for each individual galaxy falling inside
as done for XMM sources above. the 3XMM and CSC2.0 footprint (𝑁gal ) we use the minimum flux
The binned corrected distribution of sBHAR in our hard X-ray sensitivity from the cumulative curves in Figure 5 and converted it
galaxy sample, in the −6 < log 𝜆 sBHAR < 0 range, is presented in to the lowest detectable sBHAR through Eq. (2), using the proper 𝑧
Figure 6 for the entire galaxy population (left panel), and separately and M ∗ . From the cumulative distribution of log 𝜆 sBHAR,min in each
for star-forming and quiescent galaxies (central and right panels), in stellar mass range we define our detection limit as the sBHAR value
four stellar mass ranges. The errors for each probability point were for which there is a probability of detecting at least one AGN in our
calculated using the confidence limits equation from Gehrels (1986). sample. The obtained sBHAR detection limits for marginal detection
We note that, in comparison with the more advanced analysis in the CSC2.0 and the most sensitive detection by PN camera in the
presented in other works using deep surveys (e.g. Bongiorno et al. 3XMM are shown in Figure 6 by shaded grey and dotted color areas,
2016; Georgakakis et al. 2017; Aird et al. 2018), we are using only respectively.
X-ray fluxes and upper limits. This prevents us from using more
complex approaches like, e.g., Bayesian modelling, which requires
the availability of individual photons (e.g. source and background 4.1.2 The analysis of the local sBHAR distribution and its
counts) instead of the archival data products. As an alternative, we comparison with the literature
estimated a continuous probability distribution function assuming The completeness-corrected 𝑝(log 𝜆 sBHAR |M ∗ ) distribution in
that the likelihood of observing the certain value of log 𝜆 sBHAR in a Fig. 6 has an approximately power-law shape with flattening (or even
given galaxy can be described by a normal distribution centered on turnover) toward low accretion rates for all stellar mass ranges indicat-
the best log 𝜆 sBHAR estimate and with width derived from flux error ing the prevalence of low-efficiency accretion in the local Universe.
of each AGN using the same host galaxy parameters (𝑧 and M ∗ ). This trend is broadly consistent with the studies of the 𝜆 sBHAR prob-
The total probability distribution was then estimated as the sum of ability functions presented in Birchall et al. (2022, 2023), derived
the individual PDFs for all objects in the corresponding stellar mass via non-parametric models as in Aird et al. (2012); Bongiorno et al.
range (𝑁AGN ) normalised by the total number of galaxies (within (2012); Georgakakis et al. (2017), by adopting analytic models for
the same mass range) falling inside the X-ray footprint (𝑁gal ). To the sBHAR distribution convolved with the galaxy mass function as
correct the effect of the variable sensitivity across, each individual in Bongiorno et al. (2016), or using the Bayesian mixture modelling
approach as in Aird et al. (2018).
12 We find that the 𝑝(log 𝜆 sBHAR |M ∗ ) distributions for all galax-
The MARGINAL and TRUE source detection likelihood thresholds are
detected from simulations and correspond to false source rates of ∼1 and ies and stellar mass ranges cover a wide range of the BH accretion
∼0.1 false sources per stack, respectively. More detailed description of this rates pointing that the variability of the AGN activity happens in
process can be found in the section ‘Limiting sensitivity and Sky cover- shorter timescales compared to the long-term host galaxy processes.
age’ on the webpage of the statistical properties of the CSC2.0 catalogue: This AGN variability can be the result of the stochastic nature of
https://cxc.harvard.edu/csc/char.html the processes responsible for the gas transportation to the nuclear

MNRAS 000, 1–18 (2023)

8 O. Torbaniuk et al.

Figure 6. The probability distribution of specific Black Hole accretion rates, log 𝜆sBHAR , as a function of the stellar mass of the host galaxy (increasing from top
to bottom) for all (both star-forming and quiescent, left column), star-forming (center) and quiescent (right) galaxies. The black and grey points showed the same
probability distribution obtained from the observed distributions of log 𝜆sBHAR , while the solid and dashed lines represent the probability distribution obtained
as a sum of Gaussian distributions for each individual AGN in the studied sample (see detailed description in the text). The total number of galaxies falling inside
X-ray footprints and the total number of AGN with hard X-ray detection in each stellar mass range are given in the legend of each panel. The sBHAR detection
limits for marginal detection in CSC2.0 catalogue and the most sensitive detection in the PN camera in the 3XMM catalogue are shown by shaded grey (CSC)
and dotted color (XMM) areas, respectively. The power-law fits of sBHAR distributions estimated by Aird et al. 2012 (𝑧 = 0.6, 9.5 < log [ M ∗ /M ⊙ ] < 12.0)
and Bongiorno et al. 2012, 2016 (0.3 < 𝑧 < 0.8, 8.0 < log [ M ∗ /M ⊙ ] < 12.0) are shown with violet solid, yellow dashed and orange solid lines, respectively
(lines are identical for all stellar mass panels). The sBHAR probability distributions for three stellar mass ranges (log [ M ∗ /M ⊙ ] = 10.0–10.5, 10.5–11.0 and
11.0–11.5) and 𝑧 = 0.0–0.5 obtained by Georgakakis et al. (2017) are presented by pink shaded areas which corresponds to 90 per cent confidence intervals.
The light blue solid line and corresponding shaded area (90 per cent confidence interval) show the sBHAR distributions obtained by Aird et al. (2018) within
0.1 < 𝑧 < 0.5 for three stellar mass ranges for star-forming, quiescent and all galaxies. The power-law fits (and their 1𝜎 uncertainty) of sBHAR distributions
of Birchall et al. (2022, 2023) are shown by the dash-dotted lines (and corresponding dotted areas) for all galaxies (brown, Birchall et al. 2022) and separately
for SFG and quiescent (cyan, Birchall et al. 2023) within four stellar mass ranges and 𝑧 < 0.3.

MNRAS 000, 1–18 (2023)

 SMBH growth in low-z galaxies 9

Table 1. The values of AGN fraction, 𝑓 (log 𝜆sBHAR > −5.0) and 𝑓 (log 𝜆sBHAR > −2.0), and the average specific accretion rate, log h𝜆sBHAR i, for four stellar
mass ranges for all, star-forming and quiescent galaxies. The estimates are derived by integrating 𝑝 (log 𝜆sBHAR | M ∗ ) distributions presented in Figure 8 (see
also the description in the text).

𝑓 (log 𝜆sBHAR > −5.0) [%] 𝑓 (log 𝜆sBHAR > −2.0) [%] log h𝜆sBHAR i
# Stellar mass range
All Star-forming Quiescent All Star-forming Quiescent All Star-forming Quiescent

1 10.0 < log [ M ∗ /M ⊙ ] < 10.5 11.3 15.1 5.0 0.2 0.3 0.1 −3.82 −3.69 −4.21
2 10.5 < log [ M ∗ /M ⊙ ] < 11.0 24.8 47.3 7.3 0.4 0.7 0.07 −3.62 −3.32 −4.24
3 11.0 < log [ M ∗ /M ⊙ ] < 11.5 20.7 15.9 22.3 0.3 0.9 0.2 −3.68 −3.40 −3.83
4 log [ M ∗ /M ⊙ ] > 11.5 6.9 27.2 4.5 0.2 1.1 0.09 −4.06 −3.34 −4.37

region of galaxies, as well as of AGN and stellar feedback pro- for all, star-forming and quiescent galaxy populations in four stellar
cesses that are able to heat and/or remove the gas from the nu- mass ranges are presented in Table 1. As we see, the fraction of AGN
clear region, thus preventing its accretion onto the SMBH. The peak with log 𝜆 sBHAR > −5.0 lies in the range from 7 to 24 per cent,
of the derived sBHAR distribution in the local Universe occurs at while the fraction of so-called ‘classical’ AGN with moderate-to-
low BH accretion rates −4 ≤ log 𝜆 sBHAR ≤ −3, which tends to high accretion rate reaches only 0.4 per cent, which supports the fact
be offset for quiescent galaxies log 𝜆 sBHAR ≈ −4 relative to SFG of the dominance of low-efficiency accretion in the local Universe.
(log 𝜆 sBHAR ≈ −3). This finding agrees with results obtained by Moreover, we found that star-forming galaxies show higher AGN
Georgakakis et al. (2014); Aird et al. (2018); Birchall et al. (2022, fractions in almost all stellar mass ranges (15 – 47 per cent) relative
2023) for low redshift samples. The lower normalisation of the sB- to quiescent galaxies (ranging from 5 to 22 per cent), which may
HAR distributions of Birchall et al. (2022, 2023) with respect to ours indicate the difference in accretion modes and/or mechanisms re-
can be possibly due to differences in the completeness correction for sponsible for AGN fuelling for different galaxy populations. At the
the X-ray sensitivity variations (i.e. difference in the X-ray correc- same time, AGN fractions do not show a strong tendency to increase
tion energy band) used in these works with respect to us. At the with stellar mass both for SFG and quiescent galaxies. These find-
same time, Figure 6 shows that the probability of a galaxy to host the ings are also in agreement with low redshift results in Aird et al.
SMBH accreting at relatively high sBHAR (i.e. log 𝜆 sBHAR > −2) is (2018); Birchall et al. (2022, 2023). To quantify the average accre-
smaller in the local Universe compared to the studies at high redshifts tion rate in the local Universe we calculate the value of h𝜆 sBHAR i
(Bongiorno et al. 2012, 2016; Georgakakis et al. 2017; Aird et al. within a given galaxy sample and stellar mass range by integrating
2018) likely due to a smaller amount of the gas available for AGN 𝜆 sBHAR × 𝑝(log 𝜆 sBHAR |M ∗ ) distribution over the entire studied
feeding or the rarity of large-scale events (e.g. galaxy interaction and range of sBHAR. The analysis of the derived values of logh𝜆 sBHAR i
mergers) able to trigger intensive gas supply to the central regions in Table 1 showed that SFG with different stellar mass seem to main-
necessary to fuel high accretion rate AGN. However, it needs to be tain accretion at the same rate, while quiescent galaxies tend to have
noted that the lack of sources with log 𝜆 sBHAR ≥ −1 in our sBHAR lower values of log h𝜆 sBHAR i with a weak tendency to increase with
distributions is also due to the fact that our optical sample by defini- stellar mass.
tion excludes bright AGNs (Seyfert 1 and quasars), where the AGN
continuum dominates the host galaxy emission. On the contrary, at
low BH accretion rates log 𝜆 sBHAR ≤ −2 the shape of the distri- 4.2 The X-ray luminosity and sBHAR correlation with stellar
bution begins to flatten and possibly turnover for log 𝜆 sBHAR ≤ −4 mass and star-formation rate
similarly to those distributions obtained by Georgakakis et al. (2017);
Aird et al. (2018). Such behavior of the sBHAR distribution at lower 4.2.1 Stellar mass
BH accretion rates most likely reflects the natural lower limit of AGN
In the previous section, we see that AGN in the local Universe show a
activity, i.e. the minimal fuelling level necessary to trigger radiatively
broad range of accretion rates. Therefore, to analyse the dependence
efficient AGN observed in X-ray. However, probing the turnover of
between AGN activity and the properties of the host galaxy (i.e. the
the sBHAR distribution is quite challenging firstly because of its
total stellar mass and SFR), following the same steps as in Paper I,
proximity to the sensitivity limits of current X-ray telescopes, but
we divided our CSC+XMM sample in bins of SFR and M ∗ (with
also due to the difficulty of separating the nuclear emission from the
binwidth of 0.25 dex) and calculated the median 𝜆 sBHAR and 𝐿 X
host galaxy, which becomes dominant in X-ray at such luminosities.
in each bin. The resulting distribution of 𝜆 sBHAR (and 𝐿 X ) on the
Finally, based on the 𝑝(log 𝜆 sBHAR |M ∗ ) distributions we can SFR–M ∗ diagram is shown in Figure 7.
derive the AGN fraction 𝑓 (log 𝜆 sBHAR ), which is fully accounted The figure shows that the median value of 𝐿 X increases with M ∗
for the varying sensitivity of the X-ray observations across the sky, both for star-forming and quiescent galaxies (left panel of Fig. 7).
representing the fraction of galaxies in the local Universe that con- This trend is consistent with our previous results in Paper I and
tain a central black hole that is accreting above a certain limit also those found in Mullaney et al. (2012b); Delvecchio et al. (2015);
in log 𝜆 sBHAR . In order to do this, we integrate our estimates of Heinis et al. (2016); Carraro et al. (2020); Stemo et al. (2020) show-
𝑝(log 𝜆 sBHAR |M ∗ ) down to two limits of log 𝜆 sBHAR = −5.0 ing that more massive galaxies have the tendency to host AGN with
and −2.0. These two sBHAR limits were chosen so as to evalu- higher X-ray luminosity than galaxies with smaller stellar masses.
ate the fraction of the ‘entire’ X-ray selected AGN population (with On the contrary, the relation between median log 𝜆 sBHAR and stellar
log 𝜆 sBHAR > −5.0, i.e. down to the CSC sensitivity limit) and the mass (see the right panel of Fig. 7) is not so straightforward. For
fraction of galaxies hosting AGN with moderate-to-high accretion instance, star-forming galaxies show similar values of the median
rates, i.e. black holes are growing above ∼ 1 per cent of their Ed- log 𝜆 sBHAR for all stellar mass ranges, while the median sBHAR
dington limit (log 𝜆 sBHAR > −2.0). The estimated AGN fractions for quiescent galaxies seems to decrease with M ∗ . Similar weak

MNRAS 000, 1–18 (2023)

10 O. Torbaniuk et al.
X-ray luminosity specific BH accretion rate
44.0 -1.5
2.0 43.3
2.0 -2.0

41.1 42.7 -3.8 -2.4

1.5 -2.0
40.9 41.9 40.9 42.6
1.5 -3.7 -2.9 -4.2 -2.8
43.0
40.6 40.9 41.7 42.0 42.4 43.5 43.2 -3.9 -3.7 -3.1 -3.2 -3.0 -2.2 -2.6

⊙1 (median)
1.0 1.0
40.9 40.8 42.2 42.0 42.1 43.2 -3.5 -3.8 -2.6 -3.1 -3.3 -2.3 -2.5
M/year]

M⊙/year]
40.8 40.6 41.5 42.1 42.2 42.2 42.1 -3.4 -3.7 -3.2 -2.8 -3.0 -3.2 -3.5
0.5 0.5

log λsBHAR (median)

40.6 41.0 41.5 41.8 42.3 42.3 42.3 -3.4 -3.4 -3.2 -3.0 -2.9 -3.0 -3.3
42.0
-3.0

log L X [2 7 keV], erg s

40.9 41.1 41.4 41.8 41.8 42.2 41.4 41.8 -3.2 -3.2 -3.2 -3.0 -3.4 -3.2 -4.2 -4.0
0.0 40.9 41.0 41.9 41.7 41.9 41.8 42.1
0.0 -3.3 -3.4 -2.8 -3.2 -3.3 -3.5 -3.5
log SFR [

log SFR [
40.9 41.7 41.2 41.9 41.8 41.4 42.1 -3.1 -2.6 -3.5 -2.9 -3.4 -3.9 -3.4
-0.5 41.1 40.8 41.1 41.4 41.4 41.5 41.6
-0.5 -3.1 -3.7 -3.5 -3.5 -3.7 -3.9 -4.1 -3.5
41.0

⊙
40.8 40.4 41.2 41.0 41.4 41.3 40.8 40.7 -3.3 -4.0 -3.6 -3.9 -3.8 -4.1 -4.7 -5.2
-1.0 41.6 41.2 40.6 40.7 41.0 40.9 40.8
-1.0 -2.5 -3.2 -4.0 -4.3 -4.2 -4.5 -4.8

41.1 40.7 40.5 40.6 40.5 41.1 40.8 -3.2 -3.7 -4.2 -4.4 -4.6 -4.2 -4.9
-4.0
-1.5 39.8 40.6 40.7 40.1 39.7 40.9
-1.5 -4.4 -3.8 -4.0 -4.8 -5.3 -4.5
40.0
40.2 40.1 40.3 39.6 41.7 -4.0 -4.2 -4.3 -5.2 -3.3
-2.0 39.1 40.2 40.8
-2.0 -5.1 -4.1 -3.7
-4.5

-2.5 -2.5
39.0 -5.0
10.0 10.5
M∗/M]
log [
11.0 11.5 12.0 10.0 10.5
M ∗ /M ⊙ ]
log [
11.0 11.5 12.0

Figure 7. The distribution of X-ray luminosity (left) and the specific BH accretion rate 𝜆sBHAR (right) on SFR–M ∗ plane for Chandra AGN sample. The actual
median value of 𝜆sBHAR (X-ray luminosity) for each bin of SFR and M ∗ is written inside the square. The black and grey shaded areas are the same as in Figure 1.
The number of points in both diagrams ranges from 104 in the central part to 2-3 in the edges.

correlation between sBHAR and M ∗ was also found in Table 1 and −2.5 < log SFR < 2.0, where the uncertainty of hlog 𝜆 sBHAR i was
Paper I. At the same time, a number of studies showed that BH accre- computed using jackknife resampling. The hlog 𝜆 sBHAR i–log SFR
tion rate correlates positively with stellar mass at different redshift correlation is presented in four stellar mass ranges in Figure 8. To
(Rosario et al. 2013; Xu et al. 2015; Yang et al. 2018; Carraro et al. all derived values of h𝜆 sBHAR i we applied the regression analysis
2020); however, this relation seems to become weaker toward the and fitted the sBHAR-SFR correlation using the linear approxima-
local Universe (Yang et al. 2018) due to increasing number of mas- tion and least-squares regression model. The best fit parameters are
sive quiescent galaxies with less abundant cold gas fuelling the less listed in Table 2. Figure 8 confirms a trend of increasing sBHAR
luminous AGN (Rosario et al. 2013; Aird et al. 2017). In addition, with SFR from quiescent to star-forming galaxies for all stellar mass
the weakening of the sBHAR-M ∗ relation toward lower redshifts ranges. However, the linear regression analysis confirms that the
may be a product of strong AGN evolution with redshift, whereby hlog 𝜆 sBHAR i correlates with SFR at > 95 per cent confidence (𝑃-
AGN feedback in the form of wind produced by high-luminous AGN value < 0.05) only for two intermediate stellar mass intervals (the
may expel the cold gas from the host galaxy and thus reduce BH intervals #2 and 3 on Table 2). At the same time, at the two lowest
accretion (i.e. the self-regulation of the SMBH growth in massive M ∗ ranges (#1 and #2) sBHAR seems to increase with SFR only for
galaxies). It should be also mentioned, that previous sBHAR-M ∗ quiescent galaxies, while for star-forming galaxies h𝜆 sBHAR i-SFR
relations studied in the literature usually focus on highly accreting relation flattens and possibly shows a drop for log SFR & 1. The
AGN (e.g. quasars; especially in high redshift studies), which are comparison of the sBHAR-SFR relation with those available in the
missing in our sample. literature for low and high redshift AGN samples is presented in the
following Section 4.3.

4.2.2 Star-formation rate

4.2.3 The influence of selection effects on the sBHAR-SFR relation
Analysing two different galaxy populations in Figure 7 we found
that quiescent galaxies possess a smaller median sBHAR (and X-ray In interpreting the flattening of hlog 𝜆 sBHAR i–log SFR relation for
luminosity) with respect to star-forming galaxies at fixed M ∗ (iden- star-forming galaxies with relatively low stellar masses (see 10.0 <
tically to the trend obtained from 𝑝(log 𝜆 sBHAR |M ∗ ) in the previous log [M ∗ /M ⊙ ] < 10.5 and 10.5 < log [M ∗ /M ⊙ ] < 11.0 ranges
Section; see Table 1). A similar difference of log 𝜆 sBHAR for the two in Fig. 8) we need to be aware of the selection effects affecting our
different galaxy populations was also presented in Delvecchio et al. sample. As it can be seen in Fig. 4 the lowest stellar mass range
(2015); Rodighiero et al. (2015); Aird et al. (2018) and Paper I, and (i.e. 10.0 < log [M ∗ /M ⊙ ] < 10.5) contains only nearby objects
can be explained by a scenario where both star-formation and AGN (i.e. redshift 𝑧 < 0.1) and misses objects at higher redshifts. At the
activity are triggered by fuelling from a common cold gas reservoir same time, the highest stellar mass range (i.e. log [M ∗ /M ⊙ ] > 11.5)
(Alexander & Hickox 2012). We need to note that the separation into shows a dearth of massive objects at 𝑧 < 0.15. Since both sBHAR
star-forming and quiescent galaxies in this work has been done as- and SFR present a rapid decline toward 𝑧 ∼ 0 (Delvecchio et al. 2014;
suming a linear cut 1.3 dex below the MS of SFG (see Section 2), Madau & Dickinson 2014; Aird et al. 2015; D’Silva et al. 2023), the
but in reality galaxy populations in the local Universe usually show observed difference could reflect differences in the sample average
mixed properties (e.g. late-type spiral galaxies with quenched SF evolutionary stage.
vs star-forming galaxies with elliptical early-type morphology in To check the influence of the selection effect mentioned above
Paspaliaris et al. 2023). on the shape of hlog 𝜆 sBHAR i–log SFR relation we use two redshift-
To probe the relation between SFR and sBHAR in more detail limited subsamples as shown in Fig. 9: the first subsample is limited to
we calculated the mean h𝜆 sBHAR i in 10 bins of SFR in the range 𝑧 < 0.1 with stellar mass log [M ∗ /M ⊙ ] > 10, while the second one

MNRAS 000, 1–18 (2023)

 SMBH growth in low-z galaxies 11

Figure 8. The jackknife mean value of sBHAR versus SFR for star-forming (diamond) and quiescent galaxies (circles) for four stellar masses ranges. The
individual objects from our X-ray AGN sample are represented by grey crosses (SFGs) and pluses (quiescent). The errorbars were calculated as a variance of the
jackknife mean. The dash-dotted line shows the least-square linear best fit with 95 per cent confidence interval. The best fitting and goodness-of-fit parameters
are presented in Table 3.

Table 2. The best fit parameters obtained from a linear relation between hlog 𝜆sBHAR i and log SFR for four stellar mass ranges for the CSC+XMM sample
(see Figure 8). The slope, intercept with their standard errors, and all statistics parameters (𝐹-statistic, 𝑃 value, and 𝑅 2 ) were found from the least-square linear
regression. In this work, we consider the confidence level as 𝑃-value < 0.05 (stellar mass ranges satisfying this criterion are marked in blue). 𝑁 is the number
of points in each stellar mass bin.

# Stellar mass range slope intercept F-statistic 𝑃 value (F-stat) 𝑅2 𝑁

1 10.0 < log [ M ∗ /M ⊙ ] < 10.5 0.05 ± 0.13 −3.29 ± 0.14 0.14 0.7231 0.02 10
2 10.5 < log [ M ∗ /M ⊙ ] < 11.0 0.34 ± 0.07 −3.31 ± 0.07 22.44 0.0015 0.74 10
3 11.0 < log [ M ∗ /M ⊙ ] < 11.5 0.43 ± 0.05 −3.38 ± 0.05 72.58 6.1 · 10−5 0.91 9
4 log [ M ∗ /M ⊙ ] > 11.5 0.28 ± 0.18 −3.65 ± 0.13 2.61 0.1670 0.34 7

contains objects within 0.1 < 𝑧 < 0.2 and log [M ∗ /M ⊙ ] > 10.5. 4: 11.0 < log [M ∗ /M ⊙ ] < 11.5 and log [M ∗ /M ⊙ ] > 11.5, re-
As a result, we obtained 1 001 objects in the 𝑧 < 0.1 subsample spectively). However, it should be noted that the stellar mass range
(528 SFGs and 473 quiescent galaxies) and 744 objects in the second #4 for 0.1 < 𝑧 < 0.2 contains only 5 star-forming galaxies with
0.1 < 𝑧 < 0.2 subsample (373 SFGs and 371 quiescent). log SFR > 0 and therefore, the best-fit may not reflect the intrinsic
slope of the sBHAR-SFR relation in this stellar mass range. Com-
Following a similar approach as in the previous section we com- pared to the h𝜆 sBHAR i–log SFR relation for the entire sample (see
puted h𝜆 sBHAR i in SFR bins separately for 𝑧 < 0.1 and 0.1 < 𝑧 < 0.2 Fig. 8) the results for two redshift-limited subsamples show the pres-
redshift subsamples. The h𝜆 sBHAR i–log SFR relation and the best fit- ence of the sBHAR drop at higher SFR for almost all stellar mass
ting parameters of its linear approximation are shown in Fig. 10 and ranges (see in Fig. 10). This suggests that the lack of the drop ob-
Table 3. A statistically significant hlog 𝜆 sBHAR i–log SFR correlation served in Figure 8 is related to the fact that at high masses we are
at > 95 per cent confidence level (𝑃 value < 0.05) was confirmed only probing higher redshifts where the h𝜆 sBHAR i–log SFR relation can
for one stellar mass range in 𝑧 < 0.1 (#2: 10.5 < log [M ∗ /M ⊙ ] < be different (see Sec. 4.3). Moreover, repeating the regression anal-
11.0) and for two stellar mass ranges in 0.1 < 𝑧 < 0.2 (#3 and

MNRAS 000, 1–18 (2023)

 12 O. Torbaniuk et al.

NSFG = 528 NSFG = 373 2016; Shankar et al. 2017, 2020; González-Lópezlira et al. 2022;
12.0 NQuiescent = 473 NQuiescent = 371 Graham & Sahu 2023; Sahu et al. 2023) showed that the BH-to-
stellar mass relation for local galaxies varies depending on mor-
11.5 phology type. For instance, early-type galaxies (i.e. spheroids or
classical bulges) have a tendency to follow the canonical BH-to-
log[M∗ /M⊙ ]

bulge mass relation (Häring & Rix 2004; Kormendy & Ho 2013;
11.0 McConnell & Ma 2013; Li et al. 2023), which is considered to cause
also the observed BH-to-stellar mass relation. At the same time, late-
type galaxies with pseudobulges usually show a weaker correlation
10.5 between their BH mass and the host galaxy properties (like the mass
of the pseudobulge or the disk component), and direct estimations of
Star-forming galaxies XMM
10.0 Quiescent galaxies CSC their BH masses suggest smaller values than those obtained from BH-
to-stellar mass relation (Shankar et al. 2016; Li et al. 2023). This may
0.05 0.10 0.15 0.20 0.25 0.30 result in an underestimation of the sBHAR (i.e. 𝜆 sBHAR ∝ 𝐿 X /M ∗ )
Redshift, z and may be responsible for the flattening of sBHAR-SFR relation at
Figure 9. The same host galaxy stellar mass vs redshift distribution as in higher SFR.
Figure 4. The green solid lines show two selected subsamples in two redshift
intervals 𝑧 < 0.1 and 0.1 < 𝑧 < 0.2. The horizontal lines show the mass
limits 1010 and 1010.5 M ⊙ for these two subsamples. 4.3.1 Methodology
To test the hypothesis that the flattening of sBHAR-SFR relation
at higher SFR is caused by uncertainties in the BH-to-stellar mass
ysis for the stellar mass ranges #2 and #3 considering points only
scaling relation we study the absolute BH accretion rate (𝑚¤ BH ), which
with log SFR < 1 we found that hlog 𝜆 sBHAR i positively correlates
represents the mass growth rate (in M ⊙ /year units) of the central
with log SFR at > 95 per cent confidence level for both redshift in-
BH, using the definition from Alexander & Hickox 2012; Chen et al.
tervals and both stellar mass ranges (see best-fit parameters in square
2013; Delvecchio et al. 2015; Rodighiero et al. 2015; Stemo et al.
brackets in Table 3 and the grey area in Fig. 10).
2020:
Several works suggest a change in AGN X-ray luminosity (and
BH accretion rate) depending on the position of the host galaxy on 𝜖 𝑘 𝐿 [erg s−1 ]
𝑚¤ BH [M ⊙ year −1 ] = 0.15 · bol X 45 , (3)
the SFR-M ∗ diagram with respect to the MS of SFG. The enhance- 0.1 10
ment or suppression of X-ray luminosity (and BH accretion rate) where 𝜖 is the mass-energy efficiency conversion (typically estimated
for galaxies above the MS of SFG (i.e. starbursts) compared to the to be 𝜖 ≈ 0.1, Marconi et al. 2004) and 𝑘 bol 𝐿 X = 𝐿 bol is the bolo-
‘normal’ star-formation population of galaxies is rather controver- metric luminosity defined similarly to Section 4. This also allows
sial: according to some works, starbursts are less efficient in SMBH us to compare our results with the BHAR-SFR relations derived in
feeding than galaxies inside and below MS (Masoura et al. 2018; the literature, for different AGN samples within the wide range of
Carraro et al. 2020), while others support a simultaneous increase of redshifts, without depending on the specifics of BH mass derivation.
the BHAR and SFR even for galaxies above the MS (Pouliasis et al. Following the same approach as presented in Section 4.2 we calcu-
2022; Mountrichas et al. 2022a) or the absence of correlation at all lated the mean 𝑚¤ BH in 10 bins of SFR in the range −2.5 < log SFR <
(Rovilos et al. 2012; Shimizu et al. 2015). We highlight the posi- 2.0 (with similar step in 0.5 dex) for the entire CSC+XMM sample.
tion of the MS in Fig. 10 calculated using Eq. (1). The position of In this case, we did not divide the sample into stellar mass or redshift
the drop and the MS weakly correlate; however, it is hard to tell ranges in order to facilitate comparison with results available in the
whether this is revealing an intrinsic physical dependence, especially literature. The uncertainties of each 𝑚¤ BH were calculated similarly
considering that according to most studies, the MS flattens toward as in the previous section using jackknife resampling. The result-
high stellar masses due to an increased fraction of bulge-dominated ing relation and the best fitting parameters are presented in Fig. 11
galaxies at higher M ∗ (Erfanianfar et al. 2016; Tomczak et al. 2016; together with the hlog 𝑚¤ BH i–log SFR relations from literature.
Schreiber et al. 2016; Popesso et al. 2019; Dimauro et al. 2022).
Thus the MS position in Fig. 10 may actually need to be shifted
toward lower SFR relative to the observed sBHAR drop. Further- 4.3.2 Consistency with results in the literature
more, the deficiency of high accreting AGN (log 𝜆 sBHAR > −2)
Figure 11 shows that hlog 𝑚¤ BH i correlates positively with log SFR
in our sample can also flatten our sBHAR-SFR relation because,
with a best-fit slope of 0.35 ± 0.07 (𝑃-value = 0.0013), confirming
according to low- and high-redshifts studies, quasars preferentially
the correlation between hlog 𝜆 sBHAR i and log SFR found in the pre-
reside inside and above the MS (i.e. high SFR; Pouliasis et al. 2022;
vious Section. Furthermore, even using hlog 𝑚¤ BH i, we still observe
Zhuang & Ho 2022). Finally, the flattening of the sBHAR-SFR re-
that the relation between BH accretion rate and SFR flattens toward
lation toward high values of SFR may be also caused by deviations
larger SFRs (log SFR > 0) compared to ‘quiescent’ galaxies with
in the stellar-to-BH mass scaling relation, as discussed in the next
log SFR < 0.
Section.
The best-fit slope of our hlog 𝑚¤ BH i–log SFR relation is com-
patible with the one found in Delvecchio et al. (2015) at low-redshift
(0.01 < 𝑧 < 0.25), while the higher normalization is partly explained
4.3 The relation between BH growth and SFR: comparison
by the fact that their low redshift subsample contains galaxies with
with the literature
lower stellar masses (log [M ∗ /M ⊙ ] . 10.8). At the same time,
The specific BH accretion rate defined in Eq. (2) is affected by un- we observe that the slope of the low-redshift hlog 𝑚¤ BH i–log SFR
certainties in the underlying BH-to-stellar mass scaling relation. relations (both in Delvecchio et al. 2015 and this work) are systemat-
A number of studies (Reines & Volonteri 2015; Savorgnan et al. ically flatter compared to high redshift studies (Chen et al. 2013;

MNRAS 000, 1–18 (2023)

 SMBH growth in low-z galaxies 13

Figure 10. The same jackknife mean value of sBHAR vs SFR as in Fig. 8 for star-forming (diamond) and quiescent galaxies (circles) in the redshift intervals
𝑧 < 0.1 (left) and 0.1 < 𝑧 < 0.2 (right). The black dash-dotted line shows the least-square linear best fit with 95 per cent confidence interval (grey) considering
only points with log SFR < 1.0. The grey shaded area represents the position of the MS of SFG, defined 1.3 dex above the cut used for the star-forming and
quiescent galaxies separation (see definition in Eq. (1) and the text of Section 2). To plot the MS area for each separate panel we used the extreme values of
log [ M ∗ /M ⊙ ] for each stellar mass range and the maximum value of 𝑧 for the corresponding redshift subsample.

MNRAS 000, 1–18 (2023)

14 O. Torbaniuk et al.

Table 3. The same as in Table 2, but for two redshift intervals 𝑧 < 0.1 and 0.1 < 𝑧 < 0.2 presented in Fig. 10. The values in square brackets correspond to the
best fit parameters obtained from a linear hlog 𝜆sBHAR i–log SFR relation considering only points with log SFR < 1.0.

𝑧 interval # Stellar mass range slope intercept F-statistic 𝑃 value (F-stat) 𝑅2 𝑁

1 10.0 < log [ M ∗ /M ⊙ ] < 10.5 0.04 ± 0.14 −3.32 ± 0.15 0.08 0.7901 0.01 9
2 10.5 < log [ M ∗ /M ⊙ ] < 11.0 0.32 ± 0.10 −3.53 ± 0.10 10.71 0.0170 0.64 8
𝑧 < 0.1 [0.36 ± 0.10] [−3.49 ± 0.09] [14.30] [0.0129] [0.74] [7]
3 11.0 < log [ M ∗ /M ⊙ ] < 11.5 0.16 ± 0.12 −3.97 ± 0.12 1.87 0.2211 0.24 8
[0.42 ± 0.08] [−3.72 ± 0.08] [30.15] [0.0027] [0.86] [7]
4 log [ M ∗ /M ⊙ ] > 11.5 – – – – – –

1 10.0 < log [ M ∗ /M ⊙ ] < 10.5 – – – – – –

2 10.5 < log [ M ∗ /M ⊙ ] < 11.0 0.16 ± 0.08 −2.98 ± 0.07 3.56 0.0961 0.31 10
0.1 < 𝑧 < 0.2 [0.25 ± 0.05] [−2.95 ± 0.04] [23.39] [0.0029] [0.80] [8]
3 11.0 < log [ M ∗ /M ⊙ ] < 11.5 0.35 ± 0.05 −3.27 ± 0.05 41.53 0.0007 0.87 8
[0.39 ± 0.05] [−3.24 ± 0.04] [61.63] [0.0005] [0.93] [7]
4 log [ M ∗ /M ⊙ ] > 11.5 0.77 ± 0.08 −3.32 ± 0.07 95.52 0.0103 0.98 4

Figure 11. The hlog 𝑚¤ BH i–log SFR relation for CSC+XMM sample estimated in the same way as sBHAR-SFR relation presented in Fig. 8 together with results
obtained by Chen et al. (2013); Delvecchio et al. (2015); Rodighiero et al. (2015); Yang et al. (2019); Stemo et al. (2020); Zhuang & Ho (2020); Spinoglio et al.
(2022) for AGN samples at different redshift intervals (see details in the text).

Delvecchio et al. 2015; Rodighiero et al. 2015; Yang et al. 2019; 0.2 < 𝑧 < 0.5 range (see blue circles in Fig. 11) Stemo et al. 2020 is
Stemo et al. 2020). This suggests that the correlation evolves with in good agreement with our hlog 𝑚¤ BH i–log SFR relation.
time, steepening at higher redshifts. However, we should also point
out that those works do not sample well the low-SFR regime (i.e.
they typically miss the quiescent galaxy population). In fact, the As mentioned before, our sample is missing highly accreting AGN
hlog 𝑚¤ BH i–log SFR relations found in Yang et al. (2019) for bulge- (e.g. quasars) by construction, and the absence of such systems can
dominated systems (which are mainly located below the MS of SFG) be responsible (at least partially) for the flattening of the hlog 𝑚¤ BH i–
for 0.5 < 𝑧 < 1.5 and 1.5 < 𝑧 < 2.5 reveal flatter slopes compared to log SFR relation. This is supported by the findings of Pouliasis et al.
the trends observed for the same intermediate-to-high redshift inter- (2022); Zhuang & Ho (2022) showing that quasars at low redshift are
vals in Chen et al. (2013); Delvecchio et al. (2015); Rodighiero et al. mainly located in the galaxies with high SFR. In fact, the 𝑚¤ BH -SFR
(2015), and it agrees well with our local relation at log SFR < 0. relation derived by Zhuang & Ho (2020) for the sample of type 1
In the case of Stemo et al. (2020), they fit a large redshift range AGN (i.e. log 𝑚¤ BH > −2) at relatively low redshifts 𝑧 = 0.3 shows
(0.2 < 𝑧 < 2.5) and this makes it difficult to compare with re- a steeper slope compared to our local relation. A similar 𝑚¤ BH -SFR
sults derived on narrower redshift ranges; however at low redshift relation was presented by Spinoglio et al. (2022) for the combined
sample of type 1 and type 2 AGN at 𝑧 < 0.9, obtaining a best-fit
slope steeper than other high redshift samples.

MNRAS 000, 1–18 (2023)

 SMBH growth in low-z galaxies 15

Star-forming galaxies h𝑚¤ BH /SFRi are in good agreement with those present in the literature
-2.5 for different redshift ranges (Mullaney et al. 2012b; Rodighiero et al.
2015; Aird et al. 2019; Delvecchio et al. 2019). In this way, our re-
sult supports the scenario where the ratio between average black
hole growth and average galaxy growth remains constant over cos-
-3.0 mic times, despite the significant evolution of both the typical
SMBH growth rates and star-formation rates over cosmic times
(Madau & Dickinson 2014; Aird et al. 2015; Malefahlo et al. 2022).
log ṁBH /SFR

-3.5 An increase of h𝑚¤ BH /SFRi with stellar mass was also found in
Yang et al. (2018); Carraro et al. (2020) but their absolute values are
systematically lower than ours, likely due to a different 𝑘 bol assump-
tions; in fact if we adopt a luminosity-dependent 𝑘 bol as they did, the
-4.0

two results are in close agreement. However, Yang et al. (2018) also
suggest that h𝑚¤ BH /SFRi evolves with redshift up to 𝑧 < 3.0.
Mullaney+ 2012 (z ⊙ 1,z ⊙ 2) In contrast, we find that quiescent galaxies show no change
Rodighiero+ 2015 (1.5 <z< 2.5)

-4.5 Yang+ 2018 (0.5 <z< 3.0) of h𝑚¤ BH /SFRi with stellar mass, which agrees with findings in
Aird+ 2019 (0.1 <z< 2.5)
Carraro et al. 2020. Actually, quiescent galaxies have a higher
Delvecchio+ 2019 (0.5 <z< 3.0)
Carraro+ 2020 (0.1 <z< 3.5) level of log h𝑚¤ BH /SFRi (near −2.08 on average13 ) indicating that
This work (z< 0.33) they are comparatively more efficient in feeding the SMBH than
9 10 11 12
log[M∗ /M ]
in forming stars. Moreover, this is pointing toward the exis-
tence of the different physical mechanisms responsible for AGN
fuelling in quiescent galaxies via stellar mass-loss or cold ac-
Figure 12. The h 𝑚 ¤ BH /SFRi as a function of stellar mass M ∗ for star- cretion flows (Rafferty et al. 2006; Kauffmann & Heckman 2009;
forming galaxies in our sample (𝑧 < 0.33) as well as results obtained in Woodrum et al. 2022; Bambic et al. 2023; Guo et al. 2023) and/or
Mullaney et al. 2012b; Rodighiero et al. 2015; Yang et al. 2018; Aird et al. different accretion mode (e.g. the convection/advection-dominated
2019; Delvecchio et al. 2019; Carraro et al. 2020 (see description in the text). accretion flows, Bondi accretion of hot gas; Narayan et al. 1997;
Quataert & Gruzinov 2000; Allen et al. 2006; Hardcastle et al. 2007;
Russell et al. 2013).
4.3.3 The role of stellar mass in triggering the BH growth and
stellar formation
As we saw above, although the average BH accretion rate correlates
5 SUMMARY AND CONCLUSIONS
with SFR over a wide redshift range, the exact form of the correlation
depends on the studied galaxy sample properties and the investigated This paper presents a study of the correlation between AGN ac-
redshift. In any case, this correlation does not clarify if there is a direct tivity and stellar formation in the nearby Universe, improving and
physical link (i.e. feedback) between these two processes or rather it extending the analysis performed in Paper I. We started from the
arises from a common dependence on a more fundamental quantity, same parent galaxy sample extracted from the SDSS, contained in
e.g. the amount of cold molecular gas in the host galaxy (Aalto et al. the galSpec catalogue, which provides spectroscopical estimates of
2012; Alexander & Hickox 2012; Davies et al. 2012; Combes et al. SFR and M ∗ for each galaxy. In order to quantify the nuclear activity,
2014; Sharon et al. 2016; Kakkad et al. 2017; Shimizu et al. 2019; we combined X-ray data from the Chandra Source Catalog 2.0, with
Woo et al. 2020; Yesuf & Ho 2020; Circosta et al. 2021; Koss et al. the 3XMM-DR8 data previously used in Paper I. This allowed us to:
2021; Zhuang et al. 2021; Salvestrini et al. 2022). Taking into ac- i) increase the AGN sample size, deriving more robust constraints
count the presence of the SFR–M ∗ correlation for the star-forming on the specific BH accretion rate distribution in the local Universe,
galaxies, it is uncertain whether the SFR or the stellar mass of 𝑝(log 𝜆 sBHAR |M ∗ ), as well as on the correlation between SFR and
the host galaxy plays a dominant role in triggering/regulating the AGN activity, ii) adopt more stringent selection criteria to avoid
SMBH growth (Yang et al. 2018; Aird et al. 2019; Yang et al. 2019; mass-related biases and iii) demonstrate that the resolution limit of
Carraro et al. 2020). To explore this point, in Fig. 12 we derived the XMM-Newton was not significantly affecting our previous results.
unitless quantity h𝑚¤ BH /SFRi for star-forming galaxies, which repre- We found that 𝑝(log 𝜆 sBHAR |M ∗ ) has an approximately
sents the ratio of the mass accreted onto the SMBH relative to the power-law shape, flattening or declining toward lower sBHAR
mass accumulated into stars, in four stellar mass ranges. (log 𝜆 sBHAR . −3.5) for a wide range of stellar masses, sup-
Figure 12 shows that the ratios h𝑚¤ BH /SFRi in SFG tend to in- porting a picture where the local Universe contains predominately
crease with stellar mass indicating that, in comparison to low mass SMBHs accreting at low efficiency with respect to earlier epochs.
galaxies, more massive systems are more effective at feeding their Furthermore, the fraction of ‘classical’ AGN with high-efficient
central SMBH (and fuel faster SMBH growth) rather than forming accretion (log 𝜆 sBHAR > −2.0) reaches only 0.4 per cent rela-
stars. Such behaviour may be caused by more efficient transporta- tive to 7–24 per cent of the ‘entire’ local AGN population with
tion of cold gas toward the galaxy center, for instance, aided by log 𝜆 sBHAR > −5.0. At the same time, star-forming galaxies show
the presence of a denser core in more massive galaxies (i.e. bulge, generally higher AGN fraction (up to 47 per cent) compared to qui-
Fang et al. 2013; Ni et al. 2021; Aird et al. 2022; Di et al. 2023). escent galaxies (up to 22 per cent).
The reduced accretion efficiency in lower mass galaxies may be also We investigated the correlation between AGN activity and host
caused by the increased influence of the stellar feedback (e.g. su-
pernova explosions), which reduces or interrupt the gas inflow to
the central SMBH (Fabian 2012; Dubois et al. 2015; Hopkins et al. 13 Note that we do not show h 𝑚 ¤ BH /SFRi points for quiescent galaxies in
2016; Emerick et al. 2018; Byrne et al. 2023). Our measurements of Fig. 12 to avoid confusion with the results for SFG.

MNRAS 000, 1–18 (2023)

16 O. Torbaniuk et al.
galaxy properties, such as stellar mass and SFR, confirming that Source Catalog, provided by the Chandra X-ray Center (CXC) as
the nuclear X-ray luminosity depends on the stellar mass of the host part of the Chandra Data Archive and from the 3XMM-Newton
galaxy for both star-forming and quiescent systems, while the median serendipitous source catalogue compiled by the 10 institutes of the
log 𝜆 sBHAR (i.e. 𝐿 𝑋 normalised to the host galaxy/SMBH mass) XMM-Newton Survey Science Centre selected by ESA.
seems independent from the stellar mass for SFG and is possibly This publication makes use of data products from the Two Mi-
anticorrelated for quiescent galaxies. Additionally, quiescent galaxies cron All Sky Survey, which is a joint project of the University
show systematically lower X-ray luminosity and 𝜆 sBHAR at a given of Massachusetts and the Infrared Processing and Analysis Cen-
stellar mass with respect to the star-forming galaxies, implying a ter/California Institute of Technology, funded by the National Aero-
significant correlation between log 𝜆 sBHAR and log SFR for almost nautics and Space Administration and the National Science Founda-
all stellar masses. We discuss the difficulties in comparing studies at tion.
different redshifts, due to the different 𝐿 𝑋 , mass, and SFR ranges that
they probe. In general, however, we observe a weaker dependence of
the absolute SMBH accretion rate on SFR, in the sense that our best-
DATA AVAILABILITY
fit relation appears flatter, suggesting a smaller amount of accreting
material in low-redshift galaxies. All the data presented/used in this work are publicly available.
At the same time, we found that the SMBH accretion rate rel- The SDSS data (galSpec catalogue) are accessible through
ative to the mass formed into stars (for star-forming galaxies), i.e. the online service CasJobs SDSS SkyServer or the web-page
h𝑚¤ BH /SFRi, increases with stellar mass, which is fully consistent https://www.sdss.org/dr12/spectro/galaxy_mpajhu/.
with results at high redshifts. This may indicate that the rate at which The XMM-Newton data are available in
both AGN and star-formation are triggered in star-forming galax- the XMM-Newton Survey Science Centre
ies primarily depends on the total stellar mass. On the other end, (http://xmmssc.irap.omp.eu/Catalogue/3XMM-DR8/3XMM_DR8.html).
the absence of h𝑚¤ BH /SFRi–M ∗ correlation for quiescent galaxies The count/flux upper limits for XMM data are obtained from the
suggests that a different physical mechanism is responsible for the XMM FLIX at https://www.ledas.ac.uk/flix/flix.html.
triggering and fuelling AGNs in quiescent galaxies. The Chandra data is available to access
Our results can be further validated through the studies of the cold through the official CSCview application (see
molecular gas content in the host galaxies and its correlation with the https://cxc.harvard.edu/csc/gui/intro.html). The
properties of AGN and SF processes. However, currently available 2MASS data can be found through the Infrared Processing &
studies are limited to the small samples of mainly local galaxies and Analysis Center (https://old.ipac.caltech.edu/2mass/).
present rather contradictory results showing an enhanced/depleted
(or no difference) fraction of molecular gas in AGN relative to non-
AGN host galaxies. Thus, also points to the need for a more systematic
REFERENCES
approach to searching for cold molecular gas for larger samples of
galaxies and AGN. Aalto S., Garcia-Burillo S., Muller S., Winters J. M., van der Werf P., Henkel
C., Costagliola F., Neri R., 2012, A&A, 537, A44
Aird J., et al., 2012, ApJ, 746, 90
Aird J., et al., 2013, ApJ, 775, 41
ACKNOWLEDGEMENTS Aird J., Coil A. L., Georgakakis A., Nandra K., Barro G., Pérez-González
P. G., 2015, MNRAS, 451, 1892
Authors are grateful to Ivan Delvecchio for helpful discussions and Aird J., Coil A. L., Georgakakis A., 2017, MNRAS, 465, 3390
for providing his relevant data for proper comparison with our results. Aird J., Coil A. L., Georgakakis A., 2018, MNRAS, 474, 1225
Also, we thank Jarle Brinchmann for his help with understanding the Aird J., Coil A. L., Georgakakis A., 2019, MNRAS, 484, 4360
data presented in galSpec catalogue. Aird J., Coil A. L., Kocevski D. D., 2022, MNRAS, 515, 4860
OT and MP acknowledge the financial contribution from the agree- Alexander D. M., Hickox R. C., 2012, New Astron. Rev., 56, 93
ment ASI-INAF n.2017-14-H.O. and from PRIN-MIUR 2022. MP Allen S. W., Dunn R. J. H., Fabian A. C., Taylor G. B., Reynolds C. S., 2006,
also acknowledges support from the Timedomes grant within the MNRAS, 372, 21
Aspegren O., Urry C., Balokovic M., LaMassa S., Peca A., Reske C., 2023,
“INAF 2023 Finanziamento della Ricerca Fondamentale”. AG ac-
in American Astronomical Society Meeting Abstracts. p. 301.11
knowledges support from the EU H2020-MSCA-ITN-2019 Project
Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP, 93, 5
860744 “BiD4BESt: Big Data applications for Black hole Evo- Bambic C. J., Russell H. R., Reynolds C. S., Fabian A. C., McNamara B. R.,
lution Studies”14 and the Hellenic Foundation for Research and Nulsen P. E. J., 2023, MNRAS, 522, 4374
Innovation (HFRI) project “4MOVE-U” grant agreement 2688, Bigiel F., Leroy A., Walter F., Brinks E., de Blok W. J. G., Madore B.,
which is part of the programme “2nd Call for HFRI Research Thornley M. D., 2008, AJ, 136, 2846
Projects to support Faculty Members and Researchers”. FJC acknowl- Birchall K. L., Watson M. G., Aird J., Starling R. L. C., 2022, MNRAS,
edges support from the grant PID2021-122955OB-C41 funded by 510, 4556
MCIN/AEI/10.13039/501100011033 and by ERDF A way of mak- Birchall K. L., Watson M. G., Aird J., Starling R. L. C., 2023, MNRAS,
ing Europe. 523, 4756
Funding for SDSS-III has been provided by the Alfred P. Sloan Blanton M. R., Moustakas J., 2009, ARA&A, 47, 159
Bongiorno A., et al., 2012, MNRAS, 427, 3103
Foundation, the Participating Institutions, the National Science Foun-
Bongiorno A., et al., 2016, A&A, 588, A78
dation, and the U.S. Department of Energy Office of Science. The Boroson B., Kim D.-W., Fabbiano G., 2011, ApJ, 729, 12
SDSS-III web site is http://www.sdss3.org/. Brandt W., Hasinger G., 2005, Annual Review of Astronomy and Astrophysics,
This research has made use of data obtained from the Chandra 43, 827
Brinchmann J., Charlot S., White S. D. M., Tremonti C., Kauffmann G.,
Heckman T., Brinkmann J., 2004, MNRAS, 351, 1151
14 www.bid4best.org Bruzual G., Charlot S., 2003, MNRAS, 344, 1000

MNRAS 000, 1–18 (2023)

 SMBH growth in low-z galaxies 17
Byrne L., Faucher-Giguère C.-A., Stern J., Anglés-Alcázar D., Wellons S., Liu T., et al., 2017, ApJS, 232, 8
Gurvich A. B., Hopkins P. F., 2023, MNRAS, 520, 722 Liu T., et al., 2022, A&A, 661, A5
Carraro R., et al., 2020, A&A, 642, A65 Lusso E., et al., 2010, A&A, 512, A34
Carrera F. J., et al., 2007, A&A, 469, 27 Lusso E., et al., 2012, MNRAS, 425, 623
Catinella B., et al., 2018, MNRAS, 476, 875 Madau P., Dickinson M., 2014, ARA&A, 52, 415
Chen C.-T. J., et al., 2013, ApJ, 773, 3 Malefahlo E. D., Jarvis M. J., Santos M. G., White S. V., Adams N. J., Bowler
Circosta C., et al., 2021, A&A, 646, A96 R. A. A., 2022, MNRAS, 509, 4291
Civano F., Fabbiano G., Pellegrini S., Kim D. W., Paggi A., Feder R., Elvis Marconi A., Risaliti G., Gilli R., Hunt L. K., Maiolino R., Salvati M., 2004,
M., 2014, ApJ, 790, 16 MNRAS, 351, 169
Combes F., 2017, Frontiers in Astronomy and Space Sciences, 4, 10 Masoura V. A., Mountrichas G., Georgantopoulos I., Ruiz A., Magdis G.,
Combes F., et al., 2014, A&A, 565, A97 Plionis M., 2018, A&A, 618, A31
Comparat J., et al., 2022, A&A, 666, A156 McConnell N. J., Ma C.-P., 2013, ApJ, 764, 184
Comparat J., et al., 2023, A&A, 673, A122 Merloni A., 2016, in Haardt F., Gorini V., Moschella U., Treves A., Colpi M.,
D’Silva J. C. J., et al., 2023, MNRAS, 524, 1448 eds, , Vol. 905, Lecture Notes in Physics, Berlin Springer Verlag. p. 101,
Davies R., Mark D., Sternberg A., 2012, A&A, 537, A133 doi:10.1007/978-3-319-19416-5_4
Delvecchio I., et al., 2014, MNRAS, 439, 2736 Merloni A., et al., 2010, ApJ, 708, 137
Delvecchio I., et al., 2015, MNRAS, 449, 373 Mountrichas G., Shankar F., 2023, MNRAS, 518, 2088
Delvecchio I., et al., 2019, ApJ, 885, L36 Mountrichas G., Masoura V. A., Xilouris E. M., Georgantopoulos I., Buat V.,
Di Y., Li Y., Yuan F., Shi F., Caradonna M., 2023, MNRAS, 523, 1641 Paspaliaris E. D., 2022a, A&A, 661, A108
Dimauro P., et al., 2022, MNRAS, 513, 256 Mountrichas G., Buat V., Yang G., Boquien M., Burgarella D., Ciesla L.,
Dubois Y., Volonteri M., Silk J., Devriendt J., Slyz A., Teyssier R., 2015, Malek K., Shirley R., 2022b, A&A, 663, A130
MNRAS, 452, 1502 Mullaney J. R., et al., 2012a, MNRAS, 419, 95
Duras F., et al., 2020, A&A, 636, A73 Mullaney J. R., et al., 2012b, ApJ, 753, L30
Emerick A., Bryan G. L., Mac Low M.-M., 2018, ApJ, 865, L22 Narayan R., Garcia M. R., McClintock J. E., 1997, ApJ, 478, L79
Erfanianfar G., et al., 2016, MNRAS, 455, 2839 Nevalainen J., David L., Guainazzi M., 2010, A&A, 523, A22
Evans I. N., et al., 2010, ApJS, 189, 37 Ni Q., et al., 2021, MNRAS, 500, 4989
Fabbiano G., 2006, ARA&A, 44, 323 Noeske K. G., et al., 2007, ApJ, 660, L43
Fabian A. C., 2012, ARA&A, 50, 455 Park M., et al., 2023, ApJ, 953, 119
Fang J. J., Faber S. M., Koo D. C., Dekel A., 2013, ApJ, 776, 63 Paspaliaris E. D., Xilouris E. M., Nersesian A., Bianchi S., Georgantopoulos
Fathi K., Storchi-Bergmann T., Riffel R. A., Winge C., Axon D. J., Robinson I., Masoura V. A., Magdis G. E., Plionis M., 2023, A&A, 669, A11
A., Capetti A., Marconi A., 2006, ApJ, 641, L25 Pearson W. J., et al., 2018, A&A, 615, A146
Ferrara A., Zana T., Gallerani S., Sommovigo L., 2023, MNRAS, 520, 3089 Peng Y.-j., Maiolino R., 2014, MNRAS, 443, 3643
Ferrarese L., Merritt D., 2000, ApJ, 539, L9 Pier J. R., Munn J. A., Hindsley R. B., Hennessy G. S., Kent S. M., Lupton
Fiore F., et al., 2017, A&A, 601, A143 R. H., Ivezić Ž., 2003, AJ, 125, 1559
Fischer T. C., Crenshaw D. M., Kraemer S. B., Schmitt H. R., Storchi- Poitevineau R., Castignani G., Combes F., 2023, A&A, 672, A164
Bergmann T., Riffel R. A., 2015, ApJ, 799, 234 Popesso P., et al., 2019, MNRAS, 483, 3213
Gehrels N., 1986, ApJ, 303, 336 Pouliasis E., et al., 2022, A&A, 667, A56
Georgakakis A., et al., 2014, MNRAS, 440, 339 Quai S., Byrne-Mamahit S., Ellison S. L., Patton D. R., Hani M. H., 2023,
Georgakakis A., Aird J., Schulze A., Dwelly T., Salvato M., Nandra K., MNRAS, 519, 2119
Merloni A., Schneider D. P., 2017, MNRAS, 471, 1976 Quataert E., Gruzinov A., 2000, ApJ, 539, 809
González-Lópezlira R. A., et al., 2022, ApJ, 941, 53 Rafferty D. A., McNamara B. R., Nulsen P. E. J., Wise M. W., 2006, ApJ,
Graham A. W., Sahu N., 2023, MNRAS, 518, 2177 652, 216
Gültekin K., et al., 2009, ApJ, 698, 198 Reines A. E., Volonteri M., 2015, ApJ, 813, 82
Guo M., Stone J. M., Kim C.-G., Quataert E., 2023, ApJ, 946, 26 Renzini A., Peng Y.-j., 2015, ApJ, 801, L29
Hardcastle M. J., Evans D. A., Croston J. H., 2007, MNRAS, 376, 1849 Rodighiero G., et al., 2015, ApJ, 800, L10
Häring N., Rix H.-W., 2004, ApJ, 604, L89 Rosario D. J., et al., 2013, ApJ, 771, 63
Heckman T. M., Best P. N., 2014, ARA&A, 52, 589 Rosen S. R., et al., 2016, A&A, 590, A1
Heinis S., et al., 2016, ApJ, 826, 62 Rovilos E., et al., 2012, A&A, 546, A58
Hopkins P. F., Quataert E., 2010, MNRAS, 407, 1529 Russell H. R., McNamara B. R., Edge A. C., Hogan M. T., Main R. A.,
Hopkins P. F., Torrey P., Faucher-Giguère C.-A., Quataert E., Murray N., Vantyghem A. N., 2013, MNRAS, 432, 530
2016, MNRAS, 458, 816 Sahu N., Graham A. W., Hon D. S. H., 2023, MNRAS, 518, 1352
Huang R., Battisti A. J., Grasha K., da Cunha E., Lagos C. d. P., Leslie S. K., Salvestrini F., et al., 2022, A&A, 663, A28
Wisnioski E., 2023, MNRAS, 520, 446 Santini P., et al., 2017, ApJ, 847, 76
Ishibashi W., Fabian A. C., 2012, MNRAS, 427, 2998 Sartori L. F., Schawinski K., Trakhtenbrot B., Caplar N., Treister E., Koss
Jones T. M., Kriek M., van Dokkum P. G., Brammer G., Franx M., Greene M. J., Urry C. M., Zhang C. E., 2018, MNRAS, 476, L34
J. E., Labbé I., Whitaker K. E., 2014, ApJ, 783, 25 Savorgnan G. A. D., Graham A. W., Marconi A., Sani E., 2016, ApJ, 817, 21
Kakkad D., et al., 2017, MNRAS, 468, 4205 Schartmann M., Meisenheimer K., Klahr H., Camenzind M., Wolf S., Hen-
Kauffmann G., Heckman T. M., 2009, MNRAS, 397, 135 ning T., 2009, MNRAS, 393, 759
Kauffmann G., et al., 2003a, MNRAS, 341, 33 Schawinski K., et al., 2009, ApJ, 690, 1672
Kauffmann G., et al., 2003b, MNRAS, 346, 1055 Schreiber C., Elbaz D., Pannella M., Ciesla L., Wang T., Koekemoer A.,
Kim D.-W., Fabbiano G., 2013, ApJ, 776, 116 Rafelski M., Daddi E., 2016, A&A, 589, A35
King A., Nixon C., 2015, MNRAS, 453, L46 Shankar F., et al., 2016, MNRAS, 460, 3119
Kormendy J., Ho L. C., 2013, ARA&A, 51, 511 Shankar F., Bernardi M., Sheth R. K., 2017, MNRAS, 466, 4029
Koss M. J., et al., 2021, ApJS, 252, 29 Shankar F., et al., 2020, Nature Astronomy, 4, 282
Kroupa P., 2001, MNRAS, 322, 231 Sharon C. E., Riechers D. A., Hodge J., Carilli C. L., Walter F., Weiß A.,
Lehmer B. D., et al., 2016, ApJ, 825, 7 Knudsen K. K., Wagg J., 2016, ApJ, 827, 18
Leslie S. K., Kewley L. J., Sanders D. B., Lee N., 2016, MNRAS, 455, L82 Shimizu T. T., Mushotzky R. F., Meléndez M., Koss M., Rosario D. J., 2015,
Li J. I. H., et al., 2023, ApJ, 954, 173 MNRAS, 452, 1841

MNRAS 000, 1–18 (2023)

18 O. Torbaniuk et al.
Shimizu T. T., et al., 2019, MNRAS, 490, 5860
Skrutskie M. F., et al., 2006, AJ, 131, 1163
Speagle J. S., Steinhardt C. L., Capak P. L., Silverman J. D., 2014, ApJS,
214, 15
Spinoglio L., Fernández-Ontiveros J. A., Malkan M. A., 2022, ApJ, 941, 46
Stemo A., Comerford J. M., Barrows R. S., Stern D., Assef R. J., Griffith
R. L., 2020, ApJ, 888, 78
Storchi-Bergmann T., Schnorr-Müller A., 2019, Nature Astronomy, 3, 48
Taylor M. B., 2005, in Shopbell P., Britton M., Ebert R., eds, Astronomical
Society of the Pacific Conference Series Vol. 347, Astronomical Data
Analysis Software and Systems XIV. p. 29
Tomczak A. R., et al., 2016, ApJ, 817, 118
Torbaniuk O., Paolillo M., Carrera F., Cavuoti S., Vignali C., Longo G., Aird
J., 2021, MNRAS, 506, 2619
Tozzi P., et al., 2006, A&A, 451, 457
Tremonti C. A., et al., 2004, ApJ, 613, 898
Tsujimoto M., et al., 2011, A&A, 525, A25
Woo J.-H., Son D., Rakshit S., 2020, ApJ, 901, 66
Woodrum C., et al., 2022, ApJ, 940, 39
Xu L., Rieke G. H., Egami E., Haines C. P., Pereira M. J., Smith G. P., 2015,
ApJ, 808, 159
Yang G., et al., 2018, MNRAS, 475, 1887
Yang G., Brandt W. N., Alexander D. M., Chen C. T. J., Ni Q., Vito F., Zhu
F. F., 2019, MNRAS, 485, 3721
Yesuf H. M., Ho L. C., 2020, ApJ, 901, 42
Zhuang M.-Y., Ho L. C., 2020, ApJ, 896, 108
Zhuang M.-Y., Ho L. C., 2022, ApJ, 934, 130
Zhuang M.-Y., Ho L. C., Shangguan J., 2021, ApJ, 906, 38
Zubovas K., Nayakshin S., King A., Wilkinson M., 2013, MNRAS, 433, 3079

This paper has been typeset from a TEX/LATEX file prepared by the author.

MNRAS 000, 1–18 (2023)