The Astrophysical Journal, 806:22 (25pp), 2015 June 10 doi:10.

1088/0004-637X/806/1/22
© 2015. The American Astronomical Society. All rights reserved.

SUPERMASSIVE BLACK HOLES WITH HIGH ACCRETION RATES IN ACTIVE GALACTIC NUCLEI. IV.
Hβ TIME LAGS AND IMPLICATIONS FOR SUPER-EDDINGTON ACCRETION
Pu Du1, Chen Hu1, Kai-Xing Lu1,2, Ying-Ke Huang1, Cheng Cheng3, Jie Qiu1, Yan-Rong Li1, Yang-Wei Zhang4,
Xu-Liang Fan4, Jin-Ming Bai4, Wei-Hao Bian5, Ye-Fei Yuan6, Shai Kaspi7, Luis C. Ho8,9, Hagai Netzer7, and
Jian-Min Wang1,3,10
(SEAMBH Collaboration)
1
Key Laboratory for Particle Astrophysics, Institute of High Energy Physics, Chinese Academy of Sciences, 19B Yuquan Road, Beijing 100049, China
2
Astronomy Department, Beijing Normal University, Beijing 100875, China
3
National Astronomical Observatories of China, Chinese Academy of Sciences, 20A Datun Road, Beijing 100020, China
4
Yunnan Observatories, Chinese Academy of Sciences, Kunming 650011, China
5
Physics Department, Nanjing Normal University, Nanjing 210097, China
6
Department of Astronomy, University of Science and Technology of China, Hefei 230026, China
7
Wise
8
Observatory, School of Physics and Astronomy, Tel Aviv University, Tel Aviv 69978, Israel
Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing 100871, China
9
Department of Astronomy, School of Physics, Peking University, Beijing 100871, China
Received 2015 January 26; accepted 2015 April 7; published 2015 June 4

ABSTRACT
We have completed two years of photometric and spectroscopic monitoring of a large number of active galactic
nuclei (AGNs) with very high accretion rates. In this paper, we report on the result of the second phase of the
campaign, during 2013–2014, and the measurements of five new Hβ time lags out of eight monitored AGNs. All
five objects were identified as super-Eddington accreting massive black holes (SEAMBHs). The highest measured
accretion rates for the objects in this campaign are M˙  200 , where M˙ = M˙ · L Edd c-2 , M˙ · is the mass accretion
rates, L Edd is the Eddington luminosity and c is the speed of light. We find that the Hβ time lags in SEAMBHs are
significantly shorter than those measured in sub-Eddington AGNs, and the deviations increase with increasing
accretion rates. Thus, the relationship between broad-line region size (R Hb ) and optical luminosity at 5100 Å,
R Hb -L 5100 , requires accretion rate as an additional parameter. We propose that much of the effect may be due to
the strong anisotropy of the emitted slim-disk radiation. Scaling R Hb by the gravitational radius of the black hole
(BH), we define a new radius–mass parameter (Y ) and show that it saturates at a critical accretion rate of
M˙ c = 6 ~ 30 , indicating a transition from thin to slim accretion disk and a saturated luminosity of the slim disks.
The parameter Y is a very useful probe for understanding the various types of accretion onto massive BHs. We
briefly comment on implications to the general population of super-Eddington AGNs in the universe and
applications to cosmology.
Key words: accretion, accretion disks – black hole physics – galaxies: active – quasars: supermassive black holes
Supporting material: machine-readable tables

1. INTRODUCTION (Ferrarese et al. 2001; Onken et al. 2004; see Kormendy and
Ho 2013 for an extensive review). This method is an efficient
Reverberation mapping (RM) experiments, which measure
and routine way to estimate BH mass at basically all redshift, at
the delayed response of the broad emission line gas to the
distances that are well beyond the resolving power of all
ionizing continuum in active galactic nuclei (AGNs), are the
ground-based telescopes. The RM experiments have
best way to map the gas distribution and derive several
been extensively discussed and reviewed in the literature
fundamental properties such as its emissivity-weighted radius
(e.g., Peterson 1993; Kaspi et al. 2000 for earlier works and
and the black hole (BH) mass. The method was suggested by
Shen et al. 2015 and de Rosa et al. 2015 for more recent
Bahcall et al. (1972) and its theoretical foundation explained in results).
detailed in Blandford & McKee (1982). Numerous RM Most RM experiments, so far, have focused on the time lag
experiments, since the late 1980s (e.g., Calvel et al. 1991; of the broad Hβ line (t Hb in the rest frame) relative to the
Bentz 2011; Maoz et al. 1991; Peterson et al. 1991, 1993;
Dietrich et al. 1993; Wanders et al. 1993; Kaspi et al. 2000; AGN continuum luminosity (lL l ) at rest-frame wavelength of
Bentz et al. 2009a; Denney et al. 2010; Grier et al. 2012; Du 5100 Å (hereafter L5100). Results for 41 such measurements,
et al. 2014; Wang et al. 2014a; Barth et al. 2015) have based on AGN luminosity that is corrected for host
succeeded in mapping the emissivity distribution of Hβ and galaxy contamination, are summarized in Bentz et al. (2013).
several other lines in the broad-line region (BLR) in more than They lead to a simple, highly significant correlation of the form
40 AGNs. BH virial mass measurements based on the method b
R Hb » al44 ltd, (1 )
was shown to be consistent with the M·-s relationship, or
with measurements based on stellar dynamics* in local objects, where R Hb = ct Hb is the emissivity-weighted radius of the BLR
where s is the stellar velocity dispersion in the host galaxy
* and l 44 = L 5100 10 44 erg s-1. We refer to this type of
10
Author to whom any correspondence should be addressed. relationship as the R Hb -L 5100 relationship. The constants α

1
 The Astrophysical Journal, 806:22 (25pp), 2015 June 10
Table 1
The SEAMBH Project: Targets and Observations

Object a2000 d 2000 Redshift Monitoring Period Nspec S/N Comparison Stars Note on tHb
mr¢ mr¢
(Lj) (WO) Spec R* P.A.
First Phase: SEAMBH2012 Sample

Mrk 335 00 06 19.5 +20 12 10 0.0258 2012 Oct–2013 Feb 91 50 L 21 80.7 174◦. 5 Yes
Mrk 1044 02 30 05.5 −08 59 53 0.0165 2012 Oct–2013 Feb 77 91 L 21 207.0 −143◦. 0 Yes
IRAS 04416+1215 04 44 28.8 +12 21 12 0.0889 2012 Oct–2013 Mar 92 12 L 31 137.9 −55◦. 0 Yes
Mrk 382 07 55 25.3 +39 11 10 0.0337 2012 Oct–2013 May 123 25 L 3 198.4 −24◦. 6 Yes
Mrk 142 10 25 31.3 +51 40 35 0.0449 2012 Nov–2013 Apr 119 45 L 10 113.1 155◦. 2 Yes
MCG +06-26-012 11 39 13.9 +33 55 51 0.0328 2013 Jan–2013 Jun 34 24 L 37 204.3 46◦. 1 Yes
Mrk 42 11 53 41.8 +46 12 43 0.0246 2013 Jan–2013 Apr 53 24 L 15 234.4 33◦. 8 No
IRAS F12397+3333 12 42 10.6 +33 17 03 0.0435 2013 Jan–2013 May 51 53 L 32 189.0 130◦. 0 Yes
Mrk 486 15 36 38.3 +54 33 33 0.0389 2013 Mar–2013 Jul 45 100 L 44 193.8 −167◦. 0 Yes
Mrk 493 15 59 09.6 +35 01 47 0.0313 2013 Apr–2013 Jun 27 40 L 45 155.3 98◦. 5 Yes

Second Phase: SEAMBH2013 Sample

2

SDSS J075101.42+291419.1 07 51 01.4 +29 14 19 0.1208 2013 Nov–2014 May 38 50 50 40 133.3 −41◦. 3 Yes
SDSS J080101.41+184840.7 08 01 01.4 +18 48 40 0.1396 2013 Nov–2014 Apr 34 50 50 50 118.8 −98◦. 2 Yes
SDSS J080131.58+354436.4 08 01 31.6 +35 44 36 0.1786 2013 Nov–2014 Apr 31 50 50 30 100.0 145◦. 2 Uncertain
SDSS J081441.91+212918.5 08 14 41.9 +21 29 19 0.1628 2013 Nov–2014 May 34 30 30 30 79.0 73◦. 9 Yes
SDSS J081456.10+532533.5 08 14 56.1 +53 25 34 0.1197 2013 Nov–2014 Apr 27 40 ... 50 164.5 −172◦. 9 Yes
SDSS J093922.89+370943.9 09 39 22.9 +37 09 44 0.1859 2013 Nov–2014 Jun 26 40 30 30 175.1 −139◦. 0 Yes
SDSS J094422.13+103739.7 09 44 22.1 +10 37 40 0.2410 2014 Jan–2014 Apr 17 30 30 15 109.3 170◦. 1 No
SDSS J100055.71+314001.2 10 00 55.7 +31 40 01 0.1948 2013 Nov–2014 Jun 26 100 15 40 151.2 116◦. 8 No

Note. We denote the samples monitored during the 2012–2013 and 2013–2014 observing seasons as SEAMBH2012 and SEAMBH2013, respectively. The SEAMBH2012 group is described in Papers I, II and III. No
observational points were removed from the SEAMBH2012 sample due to S/N. Nspec is the numbers of spectroscopic epochs, R* is the angular distance between the object and the comparison star and P.A. the position
angle from the AGN to the comparison star. The last column contains notes on the Hβ time lags: “Yes” means significant lag, “No” lags that could not be measured and “uncertain” refers to the case of J080131 described
in the text. m r ¢ (Lj/WO) are referred to photometry at Lijiang Station of Yunnan Observatory and Wise Observatory. We calculated S/N of the SEAMBH2012 and SEAMBH2013 samples from light curves. We removed
those points of SEAMBH2013 sample in poor observing conditions in light of the lowest S/N listed here. Only a few points with extremely large error bars were removed in SEAMBH2012 sample (see details in Papers I
and II). Objects marked with “L” were not observed at the Wise observatory.

Du et al.
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

and β differ slightly from one study to the next, depending on 2. OBSERVATIONS AND DATA REDUCTION
the number of sources and their exact luminosity range. For the 2.1. Target Selection
Bentz et al. (2013) work, a = 33.65 and b = 0.533. Scaling
relationships based on other emission lines have been used to Unlike the first phase described in Papers I and II, in which
estimate BH mass in high-redshift AGNs. These are either most objects are NLS1s, for the second phase we selected
targets from the quasar sample of the Sloan Digital Sky Survey
based on the Mg II l 2798 line, which is scaled to the Hβ line
Data Release 7 (SDSS DR7) using the pipeline employed in
(e.g., McLure & Dunlop 2004; Vestergaard & Peterson 2006;
Hu et al. (2008a). The objects have similar spectroscopic
Vestergaard & Osmer 2009; Shen et al. 2011; Trakhtenbrot & characteristics to the NLS1s from Papers I and II, in particular:
Netzer 2012) or the C IV l1549 line, for which few direct lag (1) strong optical Fe II lines, (2) narrow (2000 km s-1) Hβ
measurements are available (e.g., Kaspi et al. 2007). So far, lines, and (3) weak [O III] lines (Osterbrock & Pogge 1987;
there is not enough information about the dependence of the Boroson & Green 1992). The objects selected for the first year
R Hb -L 5100 relationship on accretion rate or Eddington ratio study are NLS1s with extremely steep 2–10 keV continuum.
L Bol L Edd , where L Bol is the bolometric luminosity, For the second year sample reported here, we do not have X-
ray data and use, instead, the dimensionless accretion rate M˙ ,
L Edd = 1.5´1038 (M· M ) erg s-1 is the Eddington luminosity
which can be estimated through the physics of thin accretion
for a solar composition gas, and M· the BH mass. disks as formulated by Shakura & Sunyaev (1973,
We are conducting a large RM monitoring campaign hereafter SS73). In such systems, the accretion rate can be
targeting high-accretion rate AGNs. The aims are to understand directly calculated from the part of the spectrum where
better the physical mechanism powering these sources, the Ln µ n 1 3, regardless of the value of the BH spin (e.g., Collin
dependence of the R Hb -L 5100 relationship on accretion rate, et al. 2002; Davis & Laor 2011). The only significant
and the possibility to use such objects to infer cosmological uncertainty in this estimate is the disk inclination to the line
distances especially at high z (Wang et al. 2013). We coin these of sight, i (see more details in Paper II). We take cos i = 0.75
sources “super-Eddington accreting massive black holes” in this series of papers.11 The standard thin disk equations give
(SEAMBHs). Earlier attempts in this direction, based on (see Paper II),
monitoring narrow-line Seyfert 1 galaxies (NLS1s), failed
mostly because of the small variability amplitude of the æ l44 ö3 2 -2
M˙ = 20.1 çç ÷ m7 , (2 )
selected sources (Giannuzzo & Stirpe 1996; Giannuzzo et al. çè cos i ÷÷ø
1998; Shemmer & Netzer 2000; Klimek et al. 2004). Bentz
et al. (2011) summarized results for a few NLS1s, and found where m 7 = M· 10 7 M. For thin accretion disks, we have
relatively small L Bol L Edd for this class of objects and time lags L Bol L Edd = h M˙ , where L Bol is the disk bolometric luminos-
that are not significantly different from the ones given by ity, and η is the mass-to-radiation conversion efficiency which
Equation (1). Our monitoring campaign started in 2012 and its depends on the BH spin.
first two phases are already completed. The ∼20 targets When comparing thin to slim accretion disks it is important
observed so far are listed in Table 1. Results from the first to note that in both systems, the observed 5100 Å emission
phase have been published in Du et al. (2014; hereafter Paper comes from large disk radii and thus is less influenced by the
I) and Wang et al. (2014a; hereafter Paper II). We also studied radial motion of the accretion flow compared with the regions
the time-dependent variability of the strong Fe II lines in nine closer in that emit the shorter wavelength photons. At these
of the sources, and the results are reported in Hu et al. (2015, large radii, Keplerian rotation dominates, radiation cooling
hereafter Paper III). We characterize SEAMBHs by their locally balances the release of gravitational energy through
dimensionless accretion rate, M˙ = M˙ · L Edd c-2 , where M˙ · is viscosity, and all effects arising from radial advection and the
the accretion rate. Typical values of this parameter range from a BH spin can be neglected. As direct integration of the disk
few to ∼100 for the objects in the first phase of the project. spectral energy distribution (SED) is not practical in almost all
cases, due to the lack of far-UV observations, measuring
Such high accretion rates are characteristics of slim accretion
L Bol L Edd directly is not possible, and Equation (2) is the best
disks (Abramowicz et al. 1988) that are thought to power these
way to estimate M˙ , which is directly related to the SS73
objects (Szuszkiewicz et al. 1996; Wang et al. 1999; Wang &
accretion disk model.
Zhou 1999; Mineshige et al. 2000). Paper II shows that such
To estimate the BH mass, we followed the standard approach
objects may eventually become new standard candles for and assume that the BLR gas is moving in Keplerian orbits and
cosmology. the rest frame time lags (t Hb ) provide reliable estimates of R Hb .
This paper reports the results of the second year of our RM
This gives
campaign. Target selection, observation details, and data
reduction are described in Section 2. Hβ lags, BH mass, and 2
R Hb VFWHM
accretion rates are provided in Section 3. The new Hβ lags are M· = fBLR = 1.95 ´ 10 6 fBLR V32 t10 M, (3 )
discussed in Section 4, showing that for a given luminosity, the G
Hβ lag gets shorter with increasing M˙ . In Section 5 we briefly where G is the gravitational constant, V3 = VFWHM 103 km s-1
discuss the implications of the new findings to the under-
standing of BLR physics and geometry, and to accretion is the FWHM of the Hβ line profile in units of 103 km s-1 and
physics. Section 6 gives a brief summary of the paper. t10 = t Hb 10 days. The factor fBLR is calibrated from the known
Throughout this work we assume a standard cosmology with 11
cos i = 0.75 represents a mean disk inclination for a type 1 AGN with a
H0 = 67 km s-1 Mpc-1, WL = 0.68, and W M = 0.32 (Ade torus covering factor of about 0.5, assuming the torus axis is co-aligned with
et al. 2014). the disk axis.

3
 The Astrophysical Journal, 806:22 (25pp), 2015 June 10
Table 2
Light Curves of J075101 and J080101

J075101 J080101
Continuum Combined Continuum Line Continuum Combined Continuum Line
JD F5100 JD F5100 JD FHb JD F5100 JD F5100 JD FHb
3.466 6.219 ± 0.169 W 3.466 6.219 ± 0.169 10.277 4.312 ± 0.055 13.310 6.835 ± 0.089 L 13.310 6.835 ± 0.089 16.257 6.696 ± 0.076
8.588 5.850 ± 0.181 W 8.588 5.850 ± 0.181 13.247 4.413 ± 0.074 16.257 7.120 ± 0.051 S 16.257 7.120 ± 0.051 21.430 6.484 ± 0.055
10.277 6.325 ± 0.046 S 10.277 6.325 ± 0.046 15.248 4.508 ± 0.091 21.430 6.834 ± 0.046 S 21.440 6.793 ± 0.057 24.365 6.294 ± 0.058
13.247 6.559 ± 0.075 S 13.256 6.455 ± 0.146 22.222 4.019 ± 0.053 21.449 6.753 ± 0.085 L 24.377 6.803 ± 0.064 29.270 6.330 ± 0.051
6.352 ± 0.103 L 7.155 ± 0.074 3.699 ± 0.058 6.758 ± 0.076 S 6.866 ± 0.060 6.485 ± 0.062
4

13.265 15.248 25.255 24.365 29.283 33.432

-1
Note. JD: Julian dates from 2,456,600; F5100 and FHb are fluxes at (1 + z ) 5100 Å and Hβ emission lines in units of 10-16 erg s-1 cm-2 Å and 10-14 erg s-1 cm-2 . L: photometry from Lijiang. S: spectrum from Lijiang.
W: photometry from Wise.
(This table is available in its entirety in machine-readable form.)

Du et al.
 The Astrophysical Journal, 806:22 (25pp), 2015 June 10
Table 3
Light Curves of J080131 and J081441

J080131 J081441
Continuum Combined Continuum Line Continuum Combined Continuum Line
JD F5100 JD F5100 JD FHb JD F5100 JD F5100 JD FHb
3.478 2.146 ± 0.072 W 3.478 2.146 ± 0.072 6.296 1.301 ± 0.022 3.490 2.720 ± 0.042 W 3.490 2.720 ± 0.042 18.318 2.590 ± 0.054
6.296 2.230 ± 0.020 S 6.311 2.326 ± 0.136 7.336 1.318 ± 0.021 8.612 2.725 ± 0.036 W 8.612 2.725 ± 0.036 23.290 2.856 ± 0.035
6.325 2.422 ± 0.029 L 7.351 2.258 ± 0.040 9.343 1.218 ± 0.020 18.318 2.502 ± 0.064 S 18.318 2.502 ± 0.064 30.409 2.782 ± 0.022
7.336 2.287 ± 0.015 S 8.600 2.146 ± 0.068 12.401 1.221 ± 0.027 23.290 2.599 ± 0.028 S 23.309 2.626 ± 0.039 34.369 2.743 ± 0.028
5

7.367 2.230 ± 0.029 L 9.343 2.241 ± 0.012 21.278 1.175 ± 0.024 23.328 2.654 ± 0.097 L 30.409 2.772 ± 0.010 37.301 2.728 ± 0.022

(This table is available in its entirety in machine-readable form.)

Du et al.
 The Astrophysical Journal, 806:22 (25pp), 2015 June 10
Table 4
Light Curves of J081456 and J093922

J081456 J093922
Continuum Combined Continuum Line Continuum Combined Continuum Line
JD F5100 JD F5100 JD FHb JD F5100 JD F5100 JD FHb
6.344 6.026 ± 0.028 S 6.349 6.089 ± 0.090 6.344 3.507 ± 0.041 3.589 2.673 ± 0.086 W 3.589 2.673 ± 0.086 17.372 1.300 ± 0.033
6.354 6.152 ± 0.031 L 7.380 6.066 ± 0.030 7.380 3.441 ± 0.040 17.372 2.668 ± 0.014 S 17.372 2.668 ± 0.014 66.424 1.137 ± 0.029
7.380 6.066 ± 0.030 S 10.430 6.083 ± 0.040 10.422 3.297 ± 0.039 50.596 2.554 ± 0.071 W 50.596 2.554 ± 0.071 68.304 1.070 ± 0.032
10.422 6.054 ± 0.030 S 13.346 6.017 ± 0.046 19.319 3.184 ± 0.036 54.414 2.377 ± 0.078 W 54.414 2.377 ± 0.078 79.421 1.029 ± 0.038
6

10.438 6.112 ± 0.024 L 19.319 5.974 ± 0.051 22.382 3.026 ± 0.046 56.441 2.537 ± 0.076 W 56.441 2.537 ± 0.076 81.239 1.140 ± 0.022

(This table is available in its entirety in machine-readable form.)

Du et al.
 The Astrophysical Journal, 806:22 (25pp), 2015 June 10
Table 5
Light Curves of J094422 and J100055

J094422 J100055
Continuum Combined Continuum Line Continuum Combined Continuum Line
JD F5100 JD F5100 JD FHb JD F5100 JD F5100 JD FHb
3.614 1.500 ± 0.038 W 3.614 1.500 ± 0.038 62.387 1.051 ± 0.020 9.512 2.921 ± 0.248 W 9.512 2.921 ± 0.248 18.388 0.697 ± 0.040
9.476 1.405 ± 0.063 W 9.476 1.405 ± 0.063 67.447 1.043 ± 0.027 18.388 3.649 ± 0.042 S 18.388 3.649 ± 0.042 44.397 0.823 ± 0.046
53.437 1.157 ± 0.035 W 53.437 1.157 ± 0.035 81.351 1.030 ± 0.029 43.551 2.932 ± 0.237 W 43.551 2.932 ± 0.237 64.438 0.747 ± 0.030
56.469 1.216 ± 0.040 W 56.469 1.216 ± 0.040 84.394 0.964 ± 0.024 44.397 3.483 ± 0.038 S 44.397 3.483 ± 0.038 67.261 0.798 ± 0.029
7

61.528 1.222 ± 0.040 W 61.528 1.222 ± 0.040 87.234 0.961 ± 0.018 45.611 3.203 ± 0.265 W 45.611 3.203 ± 0.265 71.420 0.802 ± 0.033

(This table is available in its entirety in machine-readable form.)

Du et al.
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

M·-s relation (e.g., Onken et al. 2004; Ho & Kim 2014). This 2.4 m photometry, and r′-band Wise Observatory photometry.
is still a matter of some debate and the quoted uncertainties are We used the Bayesian algorithm described in Li et al. (2014),
large. The study of Ho & Kim (2014) shows that for AGNs in who assume that the AGN optical continuum variations follow
host galaxies with pseudo-bulges, fBLR is smaller than in AGNs a damped random walk model (Kelly et al. 2009; Li et al.
2013). We applied a multiplicative scale factor and an additive
hosted by classical bulges or ellipticals. However, fBLR gets flux adjustment to bring the different measurements to a
larger for AGNs with higher Eddington ratios. It is not very common flux scale. The algorithm performs inter-calibration on
clear what is the end result of these two opposite trends. For all the datasets simultaneously. This enables us to relax the
most SEAMBHs in Papers I and II, the host galaxies were requirements for the sampling rates and retain the highest
observed by the Hubble Space Telescope (HST) and show achievable temporal resolution. We selected the Lijiang
indications for pseudo-bulges. They also have very high spectroscopy as the common scale. The best parameters for
L Bol L Edd . For the present sample we do not have HST images the inter calibration are determined by a Markov chain Monte
but these sources also are of high L Bol L Edd . Given all this, we Carlo (MCMC) implementation. In that way we get the inter-
use fBLR = 1, as in Papers I and II, but note the large calibrated continuum light curves of photometry (Lijiang and
Wise data) and spectral continuum (Lijiang data). Finally, the
uncertainty on this number. combined light curves are obtained by averaging all the points
Employing Equations (1) and (3), we estimated M˙ from of the same day in the inter-calibrated light curves. All the
Equation (2) for quasars in the SDSS DR7 and selected sources continuum and Hβ light curves for the eight objects are listed in
with the highest M˙ . We discarded radio-loud objects based on Tables 2–5, and shown in Figure 1. We also calculated the
the reported FIRST observations. We also chose a redshift mean and rms spectra and present them in Appendix A where
range of z = 0.1–0.3 and L 5100  10 44.5 erg s-1 to make sure we briefly discuss these data.
that the lag, as estimated from Equation (1), can be measured
in 5–6 months of monitoring campaign and magnitude 2.3. Host Galaxies
r ¢ ⩽ 18.0 to ensure high enough signal-to-noise ratio (S/N).
Details of the sources are given in Table 1. Most of the objects described in Papers I and II had previous
HST images that were used to subtract the host galaxy
contribution at 5100 Å. Such information is not available for
2.2. Photometry and Spectroscopy the sources reported in this paper. Instead, we used the empirical
The second year observations described here are essentially relation suggested by Shen et al. (2011) based on SDSS spectra:
identical to those of the first year, and readers are referred to
host
L 5100 AGN
L 5100 = 0.8052 - 1.5502x + 0.912x 2 - 0.1577x 3, for
Papers I and II for detailed information about the telescope and x < 1.053, where x = log ( L 5100
tot
10 44 erg s-1) and L 5100
tot
is the
spectrograph. In short, the telescope is a 2.4 m alt-azimuth total emissions of AGNs and their host at 5100 Å. For
mounted Ritchey–Chrétien located in Lijiang, Yunnan Pro- x > 1.053 we have L 5100
host AGN
 L 5100 , and the host contamination
vince, and operated by the Yunnan Observatories. We use the can be neglected. These authors obtained the geometric mean
Yunnan Faint Object Spectrograph and Camera with a back- composite spectra for objects binned in log L 5100 normalized at
illuminated 2048 × 4608 pixel CCD covering a field of 3000 Å and found spectral flattening toward long wavelengths.
10¢ ´ 10¢. The differences between the first and second year The flatten part of the spectra depends on host contaminations.
observations are: (1) we adopted a 5-wide slit, compared with This empirical relation was obtained by the difference of the
the previously used 2″.5 slit, to minimize the influence of flatten part through comparisons among the spectra for quasars
atmospheric differential refraction, and used Grism 3 with in 5100 Å luminosity bins.
spectral resolution of 2.9 Å pixel−1 and wavelength coverage of The above procedure was developed for the 3 SDSS fiber
3800–9000 Å. (2) Instead of the Johnson V filter used in 2012, noted as x 3 . This is considerably smaller than the effective
we used SDSS r′-band filter to avoid the potential contamina-
tion by emission lines such as Hβ and Hα. (3) The aperture (5 ) used by us. Denote x 5 as the flux measured from
observations now include photometry from the Wise Observa- our mean spectra and Dx = ∣ x 3 - x 5 ∣. We find that on average
tory, 1 m telescope in Israel, where we used a Princeton Dx x 3 < 0.3. We also find that the differences in L 5100host AGN
L 5100
Instruments CCD camera, an SDSS r′-band filter and a typical between the 3 fiber and our 5 aperture is less than 10%. The
exposure time of 900 s. The reduction of the Wise and Yunnan fraction is estimated by comparing F5 and F3 fluxes at 5100 Å
Observatory photometry data was done in a standard way using in our sample, where F5 and F3 are measured from Lijiang and
IRAF routines. For the Lijiang data, the reduction, and the SDSS observations, respectively. In general, more than 90% of
procedures to measure 5100 Å and F(Hβ), are given in Paper I the light is contained within the 3″ fiber for z ~ 0.2 quasars.
while for Wise data, the flux measurements were done using Thus, the present subtraction of host contamination is quite
point-spread function photometry. The light curves were robust, and the added uncertainty due to this difference
produced by comparing the instrumental magnitudes to those is not larger than the intrinsic uncertainty in Shen et al.ʼs
of constant stars in the field (see, e.g., Netzer et al. 1996, for expression.
details). The uncertainties on the photometric measurements
include the fluctuations due to photon statistics and the scatter
3. Hβ LAGS, BH MASS, AND ACCRETION RATES
in the measurement of the stars used. Detailed information of
observations, such as, comparison stars etc. is provided in We used a standard cross-correlation analysis to determine
Table 1. We also list the critical S/N used as a low threshold to the lags of the Hβ line relative to the combined
reject poor weather condition observations. 5100 Å continuum. The procedure is quite standard and was
The 5100 Å light curves used in the analysis were obtained described, in detail, in Papers I and II, where we also list all the
by inter-calibration of the 5100 Å spectroscopy, r′-band Lijiang relevant references. The uncertainties on the lags are

8
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 1. Light curves and cross correlation results. Each object has six panels: ((a), (b), (c)) are light curves of the inter-calibrated continuum, combined 5100 Å
continuum, and Hβ emissions, respectively; ((d), (e), (f)) are auto correlation function (ACF) of the combined continuum, cross correlation function (CCF) of the
combined continuum and Hβ emission and the Monte-Carlo simulations of peaks (red) and centroid (blue) of lags, respectively. In panels (d) and (e), the solid lines
show the results of ICCF method and the points with error bars are from ZDCF (Z-transformed discrete correlation function). F5100 and FHb are in unit of
-1
10-16 erg s-1 cm-2 Å and 10-14 erg s-1 cm-2 for all quasars, respectively. In panel (a) of J075101, the red dots, green squares and blue diamonds are fluxes from
Lijiang spectroscopy, photometry and Wise photometry, respectively. The color points in other objects have the same meanings with that of J075101. Bars with
terminals as systematic errors are plotted in the corners of the panels (see Paper I for details).

determined through the “flux randomization/random subset response to the two shorter duration features. The CCF of the
sampling” method (RS/RSS) (Peterson et al. 1998, 2004), the entire campaign has a low peak, at rmax  0.5, which is
cross-correlation centroid distribution (CCCD) and cross- consistent with a zero lag. Carrying the analysis for the first 70
correlation peak distribution (CCPD) generated by RS/RSS days only, as shown in the second diagram of J080131 panel in
method (Maoz & Netzer 1989; Peterson et al. 1998, 2004; Figure 1, we found a very significant peak with a centroid lag
+8.4
Denney et al. 2006, 2010 and reference therein), which are of 11.5- 3.6 days (with a very high coefficient of rmax = 0.81) in
shown in Figure 1. For a successful detection of Hβ lag we rest-frame.
require: (1) non-zero lag from the CCF peak and (2) a Our search of the literature shows that the unusual
maximum correlation coefficient larger than 0.5. combination of line and continuum light curves observed in
Five of the eight sources show well-determined lags. All J080131 is rare (see similar, but not identical, behavior in NGC
these sources are listed in Table 6. The other three failed our 7469 (Peterson et al. 2014) and in the UV light curve of
above criteria. Figure 1 shows that the CCFs of two of them NGC 5548; de Rosa et al. 2015). We can think of various
(J094422 and J100055) show two comparable peaks and no scenarios that could cause such an event in a SEAMBH with
significant Hβ lags. The third source, J080131, has an unusual extremely high accretion rate. For example, the inner part of
combination of line and continuum light curves. The slim disks can be so thick that self-shadowing effects lead to
continuum light curve shows a well-determined, broad feature strong anisotropy of the emitted radiation. The signal received
followed by two major short “bursts” centered at around and measured by a remote observer can differ substantially
JD24566000+, 120, and 140. The Hβ light curve shows a clear from the ionizing radiation that reaches the Hβ-emitting clouds
response to the first continuum dip and burst feature, but no (see discussions by Wang et al. 2014b). Given the rarity of

9
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 1. (Continued.)

such objects, we have no way to test this idea and thus do not details of the procedure are given in Papers I and II and the
include this object in the remaining analysis. distributions of M˙ in the two groups of the previously and
We used Equations (2) and (3) to calculate accretion rates, currently monitored objects are shown in Figure 2. As clearly
M˙ and BH masses for the five sources listed in Table 6. The shown in the diagram, our Lijiang-observed sources occupy the

10
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Table 6
Hβ Reverberations of the SEAMBH Targets

tHb FWHM sline log ( M· M) log M˙ log L 5100 log L Hb EW(Hβ)
Objects
(days) (km s-1) (km s -1) -1 -1 (Å)
( erg s ) ( erg s )
SEAMBH2012
+1.6 +0.10 +0.37
Mrk 335 8.7- 1.9 2096 ± 170 1470 ± 50 6.87- 0.14 1.28- 0.30 43.69 ± 0.06 42.03 ± 0.06 110.5 ± 22.3
+3.3 +0.12 +0.40
Mrk 1044 10.5- 2.7 1178 ± 22 766 ± 8 6.45-0.13 1.22- 0.41 43.10 ± 0.10 41.39 ± 0.09 101.4 ± 31.9
+2.9 +0.19 +0.69
Mrk 382 7.5- 2.0 1462 ± 296 840 ± 37 6.50- 0.29 1.18- 0.53 43.12 ± 0.08 41.01 ± 0.05 39.6 ± 9.0
+1.2 +0.07 +0.23
Mrk 142 7.9- 1.1 1588 ± 58 948 ± 12 6.59- 0.07 1.65- 0.23 43.56 ± 0.06 41.60 ± 0.04 55.2 ± 9.5
+8.4 +0.14 +0.37
MCG +06-26-012a 24.0- 4.8 1334 ± 80 785 ± 21 6.92- 0.12 -0.34- 0.45 42.67 ± 0.11 41.03 ± 0.06 114.6 ± 32.5
+5.5 +0.27 +0.98
IRAS F12397b 9.7- 1.8 1802 ± 560 1150 ± 122 6.79-0.45 2.26- 0.62 44.23 ± 0.05 42.26 ± 0.04 54.2 ± 8.4
+7.5 +0.12 +0.20
Mrk 486 23.7- 2.7 1942 ± 67 1296 ± 23 7.24- 0.06 0.55- 0.32 43.69 ± 0.05 42.12 ± 0.04 135.9 ± 20.3
+1.2 +0.04 +0.33
Mrk 493 11.6-2.6 778 ± 12 513 ± 5 6.14-0.11 1.88-0.21 43.11 ± 0.08 41.35 ± 0.05 87.4 ± 18.1
+13.9 +0.31 +0.16
IRAS 04416c 13.3- 1.4 1522 ± 44 1056 ± 29 6.78- 0.06 2.63- 0.67 44.47 ± 0.03 42.51 ± 0.02 55.8 ± 4.7

SEAMBH2013
+15.6 +0.17 +0.25
SDSS J075101 33.4- 5.6 1495 ± 67 1055 ± 32 7.16- 0.09 1.34- 0.41 44.12 ± 0.05 42.25 ± 0.03 68.1 ± 8.6
+9.7 +0.34 +0.39
SDSS J080101 8.3-2.7 1930 ± 18 1119 ± 3 6.78-0.17 2.33- 0.72 44.27 ± 0.03 42.58 ± 0.02 105.5 ± 8.3
+12.7 +0.23 +0.63
SDSS J081441 18.4- 8.4 1615 ± 22 1122 ± 11 6.97- 0.27 1.56- 0.57 44.01 ± 0.07 42.42 ± 0.03 132.0 ± 23.7
+7.7 +0.12 +1.03
SDSS J081456 24.3- 16.4 2409 ± 61 1438 ± 32 7.44- 0.49 0.59- 0.30 43.99 ± 0.04 42.15 ± 0.03 74.4 ± 7.6
+2.1 +0.07 +0.71
SDSS J093922 11.9- 6.3 1209 ± 16 835 ± 11 6.53- 0.33 2.54- 0.20 44.07 ± 0.04 42.09 ± 0.04 53.0 ± 6.7

Note. (1): Hβ lags of J080131 are not listed in this table, but given in the second paragraph of Section 3 in the main text. (2): All SEAMBH2012 measurements are
taken from Paper III, but 5100 Å fluxes are from I and II.
a
MCG +06-26-012 was selected as a super-Eddington candidate but later was identified to be a sub-Eddington accretor (M˙ = 0.46 ).
b
For IRAS F12397, we use the fluxes of the case with local absorption correction (see the details in Paper I and III).
c
The time lag of IRAS 04416 cannot be obtained significantly using the integration method in Papers II and II, but has been detected by the fitting procedures in
Paper III.

accurate rates, in which BH mass is estimated by the R Hb -L 5100

relation (e.g., Netzer et al. 2004).
In Paper II we classified SEAMBHs as those objects with
m˙ = h M˙ ⩾ 0.1. This is based on the idea that beyond this
value, the accretion disk becomes slim and the radiation
efficiency is reduced due to photon trapping and other effects.
Since we cannot observe the entire SED, we have no direct way
to measure L Bol L Edd , and this criterion is used as an
approximate tool for identifying SEAMBH candidates. To be
on the conservative side, we chose the lowest possible
efficiency, h = 0.038 (retrograde disk with a = -1, see
Bardeen et al. 1972). Thus SEAMBHs are objects with
Figure 2. The distribution of dimensionless accretion rates of the mapped M˙ = 2.63. For simplicity, in this paper we use M˙ min = 3 as the
AGNs and quasars. The previous sample refers to the sample of all mapped required minimum. Later on, in Section 5, we introduce an
AGNs summarized by Bentz et al. (2013) (the repeated monitored AGNs are empirical way, based on our own mass measurements, to define
regarded as individual ones), and adds four recently mapped AGNs: NGC 7469
updated by Peterson et al. (2014), KA 1858+4850 (Pei et al. 2014), Mrk 1511
this group more accurately.
(Barth et al. 2013) and NGC 5273 (Bentz et al. 2014). The dashed line is for Table 6 lists all observables, M· and M˙ for the first
the previous sample and the solid is for the SEAMBH sample. Our campaign (SEAMBH2012) and second (SEAMBH2013) year observa-
selected those candidate sources with extremely high accretion rates (M˙  10 ) tions. All objects, except for MCG +06-26-012 (see note in the
. Though some sources monitored previously have M˙ ~ 10 , most of them
have lower accretion rates. Right panel shows the distribution of EW(Hβ),
table caption), show M˙ > 3, indicating they are SEAMBHs. In
showing that SEAMBH sample tends to have low EW(Hβ). Paper III, more accurate measurements of Hβ lags have been
derived by using a scheme of simultaneously fitting spectra of
host and AGNs. The newly measured lags are the ones listed
high tail of the M˙ distribution. This is not surprising given the Table 6. Comparing with the previous measurements in Papers
way we selected our targets. We also show in Figure 2 the EW I and II, the updated lags are consistent with the previous ones,
distribution of Hβ. On average, our high accretion rate sources but the error bars are much smaller. Including our new
have lower mean EW(Hβ). An anti-correlation between EW observations, the majority of SEAMBHs with RM-based mass
(Hβ) and L Bol L Edd has been found in other samples with less measurements come from our Lijiang RM campaign.

11
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

4. Hβ TIME LAGS IN SEAMBHS Denney et al. 2010; Grier et al. 2012) while others prefer the
use of FWHM(Hβ) (e.g., Kaspi et al. 2005). Our procedure is
To understand better the fundamental differences between
based on the FWHM method (see explanation in Papers I and
high and low M˙ sources, we must use the measured correlation
II). As shown recently in Woo et al. (2015), the scatter in the
between time lag and source luminosity for all RM AGNs. For
scaling parameter ( fBLR ) derived in this method is very similar
this, we have to take into account that some of the sources have
to the scatter in the method based on the rms spectrum.
been mapped more than once and hence there are various
We have therefore collected from the literature all the
measured values for their luminosity, time lag, and FWHM
required information about the Hβ line and carried out our own
(Hβ) for the same BH mass. There are two possible ways to
mass measurements for the sources where earlier mass
address this issue. The first is to average the BH mass from the
measurements were based on the rms method. In most of the
individual campaigns, and then obtain mean (or median) values
cases, the agreement between the older rms method and the
for L5100, L Hb , and the time lag. This scheme, which was used
value obtained here is within the uncertainty on the BH mass.
in Kaspi et al. (2005), Bentz et al. (2009a), and other papers, is In a few cases (five objects: NGC 4051, NGC 4151, PG 0804
referred here as the “average scheme.” The second is to treat +761, PG 1613+658, and PG 2130+099) the deviation is larger
the independent RM campaigns of a single object as different which we interpret as unrealistically small uncertainty on the
objects (e.g., Bentz et al. 2013). We call this the “direct mass. The deviation is generally smaller than ∼0.5 dex except
scheme.” for NGC 4051 where the deviation is ∼0.7 dex. Some more
The goal of the present paper is to find the basic properties of details for FWHM measurements of a few objects are given in
the central power house, presumably an accretion disk, as a the caption of Table 7. We list all the measured values and their
function of BH mass and normalized accretion rate. The averages in this table.12
measured Hβ time lag is the tool we use to study these All correlations shown in this paper were calculated with the
properties. This is changing with the ionizing continuum but it FITEXY method in the version adopted by Tremaine et al.
does not represent real changes in the gas distribution in the (2002), where scatter is allowed for by increasing the
BLR during one observing season. Since the BH mass does not uncertainties in small steps until c 2 is about unity (this is
change on a short time scale, and the time scale to change the typical for many of our correlations). We also used the BCES
global accretion rate through the disk is also very long, we method (Akritas & Bershady 1996) but prefer not to use its
must be seeing short time scale fluctuations that do not change results since it is known to give unreliable results in samples
these fundamental properties but, nevertheless, affect the containing a few outliers (there are a few objects with quite
ionization of the BLR. To understand the population properties, large uncertainties of M˙ ).
we must give equal weights to all objects, i.e., use the average Figure 3(a) shows Hβ lags versus L5100 for all the
scheme approach. This means that the mean luminosity is used observations of mapped AGNs (i.e., multiple measurements
to derive M˙ . This can be done during one campaign that lasts a for each object when available). Our objects are marked with
few months, or during several campaigns that last a few years. blue (first year, data reported in Papers I and II) and red
The only exception is, perhaps, when the total time exceeds the (second year, this paper). Using the data in Table 7 we get:
dynamical time scale of the BLR.
The direct scheme can be useful to derive other properties,
such as the ionization distribution across a “typical BLR.” ( )
log R Hb ltd =
While this is not the goal of the present work, we have ì
ï (1.45  0.03) + (0.50  0.03) log l44 (entire sample)
calculated all the fundamental correlations discussed below ï
ï
ï (1.54  0.03) + (0.53  0.03) log l44 ( M˙ < 3),
using this approach, too, and show them in Appendix B. We í
ï
ï
(1.30  0.05) + (0.54  0.06) log l44 ( M˙ ⩾ 3), (4)
note that despite the fundamentally different approaches, which ï
result in a different BH mass distribution (since several masses ï
î
are counted more than once), the results are not very different.
with intrinsic scatters sin = (0.21, 0.15, 0.24) for (4a, 4b, 4c),
respectively. The regression for the M˙ < 3 sample is almost
4.1. Continuum Luminosity Versus BLR Size identical to Bentz et al. (2013). Equations (4b) and (4c) show
Table 7 lists the objects summarized by Bentz et al. (2013), different intercepts for the M˙ < 3 and the M˙ ⩾ 3 samples. Most
plus a few newly mapped sources. Since some of the of the SEAMBHs from our campaign are located below the
procedures used in earlier campaigns differ in several aspects, R Hb -L 5100 relation, increasing the scatter in the relationships.
we decided to apply the same uniform procedure for all the
sources. For objects with multiple RM measurements, our For more extreme SEAMBHs with M˙ > 10, the regression
procedure for obtaining the average values is as follows: we shows log (R Hb ltd) = (1.18  0.07) + (0.48  0.10) log l 44 ,
first calculate M· for each of the individual campaigns and then indicating they deviate from the M˙ < 3 group more than the
average all these values, using weighted standard deviations, to M˙ ⩾ 3 group. It is clear that the SEAMBH sources increase the
get a mean M· and its error. This value is used to obtain scatter considerably, especially over the limited luminosity range
individual values of M˙ for the various campaigns, using L5100 occupied by the new sources.
from this campaign. The mean M˙ for the object is the average In the following analysis, we use the averaged lags and the
of the individual M˙ , and its errors include the standard averaged mass as listed in Table 7 (note that unlike in Paper II,
deviation. For the mean luminosity and t Hb , we used here we include in the analysis the few radio-loud AGNs with
logarithmic averages. measured time lags and BH mass). In Figure 3(c), we divided
Regarding BH mass calculations in individual campaigns,
this is more problematic since some works used the rms 12
Following Bentz et al. (2013), we included in the uncertainties of the
spectrum (e.g., Peterson et al. 2004; Bentz et al. 2009b; 5100 Å luminosities also the uncertainties due to distances.

12
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Table 7
Reverberation Mapping AGNs and Results

tHb FWHM log ( M· M) log M˙ log L 5100 log L Hb EW(Hβ) References
Objects
(days) (km s-1) ( erg s-1) ( erg s-1) (Å)
+1.6 +0.10 +0.31
Mrk 335 8.7- 1.9 2096 ± 170 6.87- 0.14 1.17- 0.30 43.69 ± 0.06 42.03 ± 0.06 110.5 ± 22.3 1, 2, 3
+4.8 +0.11 +0.30
16.8- 4.2 1792 ± 3 7.02- 0.12 1.28-0.29 43.76 ± 0.06 42.13 ± 0.06 119.7 ± 23.3 4, 5, 6, 7
+6.6 +0.18 +0.30
12.5- 5.5 1679 ± 2 6.84- 0.25 1.39- 0.29 43.84 ± 0.06 42.18 ± 0.06 111.2 ± 21.1 4, 5, 6, 7
+0.7 +0.11 +0.30
14.3- 0.7 1724 ± 236 6.92- 0.14 1.25-0.29 43.74 ± 0.06 41.99 ± 0.07 89.5 ± 19.5 4, 8a
+4.6 +0.10 +0.18
14.0- 3.4 1707  79 6.93- 0.11 1.27- 0.17 43.76 ± 0.07 42.09 ± 0.09 108.2 ± 16.8 L
+24.1 +0.09 +0.28
PG 0026+129 111.0- 28.3 2544 ± 56 8.15- 0.13 0.65- 0.20 44.97 ± 0.02 42.93 ± 0.04 46.2 ± 4.7 4, 5, 6, 9
+24.5 +0.11 +0.31
PG 0052+251 89.8- 24.1 5008 ± 73 8.64- 0.14 - 0.59- 0.25 44.81 ± 0.03 43.13 ± 0.05 107.4 ± 14.8 4, 5, 6, 9
+3.2 +0.07 +0.31
Fairall9 17.4- 4.3 5999 ± 66 8.09- 0.12 -0.71- 0.21 43.98 ± 0.04 42.67 ± 0.04 249.8 ± 32.0 4, 5, 6, 10
+3.5 +0.07 +0.24
Mrk 590 20.7- 2.7 2788 ± 29 7.50- 0.06 - 0.22- 0.25 43.59 ± 0.06 41.92 ± 0.06 107.0 ± 22.0 4, 5, 6, 7
+8.5 +0.22 +0.28
14.0- 8.8 3729 ± 426 7.58-0.48 - 0.91-0.30 43.14 ± 0.09 41.58 ± 0.16 142.8 ± 62.2 4, 5, 6, 7
+4.9 +0.07 +0.25
29.2- 5.0 2744 ± 79 7.63- 0.09 -0.54- 0.26 43.38 ± 0.07 41.75 ± 0.07 119.7 ± 28.1 4, 5, 6, 7
+3.6 +0.05 +0.24
28.8- 4.2 2500 ± 42 7.55- 0.07 - 0.13- 0.25 43.65 ± 0.06 41.92 ± 0.07 94.9 ± 20.7 4, 5, 6, 7
+6.5 +0.07 +0.36
25.6- 5.3 2716  202 7.55- 0.08 -0.41- 0.36 43.50 ± 0.21 41.85 ± 0.12 108.6 ± 20.2 L
+3.3 +0.12 +0.40
Mrk 1044 10.5- 2.7 1178 ± 22 6.45- 0.13 1.22- 0.41 43.10 ± 0.10 41.39 ± 0.09 101.4 ± 31.9 1, 2, 3
+21.3 +0.19 +0.37
3C 120 38.1- 15.3 2327 ± 48 7.61- 0.22 0.03- 0.37 44.07 ± 0.05 42.37 ± 0.06 100.9 ± 18.3 4, 5, 6, 7
+2.3 +0.05 +0.37
25.9- 2.3 3529 ± 176 7.80- 0.06 - 0.17- 0.37 43.94 ± 0.05 42.36 ± 0.05 135.9 ± 22.3 4, 8a
+8.7 +0.15 +0.30
26.2- 6.6 2472  729 7.79- 0.15 -0.07- 0.30 44.00 ± 0.10 42.36 ± 0.04 118.8 ± 28.9 L
+13.9 +0.31 +0.16
IRAS 04416 13.3- 1.4 1522 ± 44 6.78- 0.06 2.63- 0.67 44.47 ± 0.03 42.51 ± 0.02 55.8 ± 4.7 1, 2, 3
+8.3 +0.07 +0.24
Ark 120 47.1- 12.4 6042 ± 35 8.53- 0.13 -1.48- 0.23 43.98 ± 0.06 42.60 ± 0.05 211.5 ± 37.5 4, 5, 6, 7
+4.8 +0.05 +0.27
37.1- 5.4 6246 ± 78 8.45- 0.07 - 2.01- 0.27 43.63 ± 0.08 42.43 ± 0.07 321.1 ± 77.6 4, 5, 6, 7
+8.5 +0.07 +0.41
39.5- 7.8 6077  147 8.47- 0.08 -1.70- 0.41 43.87 ± 0.25 42.54 ± 0.13 244.8 ± 80.3 L
+8.3 +0.28 +0.41
Mrk 79 9.0- 7.8 5056 ± 85 7.65- 0.88 - 0.75- 0.34 43.63 ± 0.07 41.89 ± 0.07 92.4 ± 21.0 4, 5, 6, 7
+6.6 +0.15 +0.41
16.1- 6.6 4760 ± 31 7.85- 0.23 -0.59- 0.34 43.74 ± 0.07 41.92 ± 0.08 78.1 ± 18.3 4, 5, 6, 7
+6.4 +0.15 +0.41
16.0- 5.8 4766 ± 71 7.85-0.20 - 0.70- 0.34 43.66 ± 0.07 41.89 ± 0.07 86.0 ± 19.3 4, 5, 6, 7
+5.1 +0.12 +0.25
15.6- 4.9 4793  145 7.84- 0.16 -0.68- 0.21 43.68 ± 0.07 41.90 ± 0.05 85.4 ± 13.3 L
+15.6 +0.17 +0.25
SDSS J075101 33.4- 5.6 1495 ± 67 7.16- 0.09 1.34- 0.41 44.12 ± 0.05 42.25 ± 0.03 68.1 ± 8.6 11
+2.9 +0.19 +0.69
Mrk 382 7.5- 2.0 1462 ± 296 6.50- 0.29 1.18- 0.53 43.12 ± 0.08 41.01 ± 0.05 39.6 ± 9.0 1, 2, 3
+9.7 +0.34 +0.39
SDSS J080101 8.3- 2.7 1930 ± 18 6.78- 0.17 2.33- 0.72 44.27 ± 0.03 42.58 ± 0.02 105.5 ± 8.3 11
+18.8 +0.05 +0.15
PG 0804+761 146.9- 18.9 3053 ± 38 8.43- 0.06 0.00- 0.13 44.91 ± 0.02 43.29 ± 0.03 122.5 ± 10.3 4, 5, 6, 9
+12.7 +0.23 +0.63
SDSS J081441 18.4- 8.4 1615 ± 22 6.97- 0.27 1.56- 0.57 44.01 ± 0.07 42.42 ± 0.03 132.0 ± 23.7 11
+7.7 +0.12 +1.03
SDSS J081456 24.3- 16.4 2409 ± 61 7.44- 0.49 0.59- 0.30 43.99 ± 0.04 42.15 ± 0.03 74.4 ± 7.6 11
+13.7 +0.15 +0.57
PG 0844+349 32.3-13.4 2694 ± 58 7.66- 0.23 0.50- 0.42 44.22 ± 0.07 42.56 ± 0.05 111.2 ± 22.1 5, 6, 9, 12
+5.5 +0.09 +0.35
Mrk 110 24.3- 8.3 1543 ± 5 7.05- 0.18 0.81- 0.32 43.68 ± 0.04 42.12 ± 0.05 139.6 ± 20.4 4, 5, 6, 7
+10.5 +0.18 +0.34
20.4- 6.3 1658 ± 3 7.04- 0.16 0.92- 0.32 43.75 ± 0.04 42.02 ± 0.05 94.8 ± 14.7 4, 5, 6, 7
+14.9 +0.16 +0.35
33.3- 10.0 1600 ± 39 7.22-0.16 0.58-0.33 43.53 ± 0.05 41.97 ± 0.04 139.4 ± 20.5 4, 5, 6, 7
+8.9 +0.13 +0.26
25.6- 7.2 1634  83 7.10- 0.14 0.77- 0.25 43.66 ± 0.12 42.03 ± 0.08 123.8 ± 29.1 L
+2.1 +0.07 +0.71
SDSS J093922 11.9- 6.3 1209 ± 16 6.53- 0.33 2.54- 0.20 44.07 ± 0.04 42.09 ± 0.04 53.0 ± 6.7 11
+21.6 +0.06 +0.16
PG 0953+414 150.1- 22.6 3071 ± 27 8.44- 0.07 0.39- 0.14 45.19 ± 0.01 43.29 ± 0.04 64.7 ± 5.9 4, 5, 6, 9
+0.8 +0.09 +0.38
NGC 3227 3.8- 0.8 4112 ± 206 7.09- 0.12 -1.34- 0.36 42.24 ± 0.11 40.38 ± 0.10 71.0 ± 23.6 4,13a
+1.2 +0.07 +0.85
Mrk 142 7.9- 1.1 1588 ± 58 6.59- 0.07 1.90- 0.86 43.56 ± 0.06 41.60 ± 0.04 55.2 ± 9.5 1, 2, 3
+0.7 +0.10 +0.82
2.7-0.8 1462 ± 2 6.06- 0.16 1.96- 0.82 43.61 ± 0.04 41.66 ± 0.05 57.6 ± 8.6 4, 14
+7.3 +0.38 +0.59
6.4- 3.4 1462  86 6.47- 0.38 1.93- 0.59 43.59 ± 0.04 41.62 ± 0.06 56.6 ± 6.6 L
+1.0 +0.05 +0.41
NGC 3516 11.7- 1.5 5384 ± 269 7.82- 0.08 - 1.97- 0.52 42.79 ± 0.20 41.06 ± 0.18 94.7 ± 59.2 4,13a
+0.6 +0.11 +0.51
SBS 1116+583 A 2.3- 0.5 3668 ± 186 6.78- 0.12 - 0.87- 0.71 42.14 ± 0.23 40.70 ± 0.07 186.8 ± 104.1 4, 14
+0.5 +0.05 +0.30
Arp 151 4.0- 0.7 3098 ± 69 6.87- 0.08 - 0.44- 0.28 42.55 ± 0.10 40.95 ± 0.11 130.0 ± 44.4 4,14
+3.3 +0.12 +0.45
NGC 3783 10.2- 2.3 3770 ± 68 7.45- 0.11 - 1.58- 0.59 42.56 ± 0.18 41.01 ± 0.18 144.0 ± 83.7 4, 5, 6, 15
+8.4 +0.14 +0.37
MCG +06-26-012 24.0- 4.8 1334 ± 80 6.92- 0.12 - 0.34- 0.45 42.67 ± 0.11 41.03 ± 0.06 114.6 ± 32.5 1, 2, 3
+0.6 +0.07 +0.35
Mrk 1310 3.7- 0.6 2409 ± 24 6.62- 0.08 - 0.31- 0.39 42.29 ± 0.14 40.56 ± 0.10 94.3 ± 38.2 4, 14
+0.5 +0.23 +1.29
NGC 4051 1.9- 0.5 851 ± 277 5.42- 0.53 1.59- 0.84 41.96 ± 0.19 40.19 ± 0.18 86.8 ± 51.7 4,13a
+1.1 +0.07 +0.37
NGC 4151 6.6- 0.8 6371 ± 150 7.72- 0.06 - 2.81- 0.57 42.09 ± 0.21 40.56 ± 0.20 150.8 ± 100.6 4, 5, 6, 16
+25.6 +0.11 +0.63
PG 1211+143 93.8- 42.1 2012 ± 37 7.87- 0.26 0.84- 0.35 44.73 ± 0.08 43.02 ± 0.06 100.2 ± 22.9 5, 6, 9, 12
+1.7 +0.20 +0.59
Mrk 202 3.0- 1.1 1471 ± 18 6.11- 0.20 0.66- 0.65 42.26 ± 0.14 40.40 ± 0.09 70.6 ± 27.5 4, 14
+1.6 +0.10 +0.36
NGC 4253 6.2- 1.2 1609 ± 39 6.49- 0.10 0.36- 0.42 42.57 ± 0.12 40.77 ± 0.12 81.1 ± 31.6 4, 14
+68.5 +0.09 +0.33
PG 1226+023 306.8- 90.9 3509 ± 36 8.87- 0.15 0.70- 0.20 45.96 ± 0.02 44.13 ± 0.04 74.6 ± 6.9 4, 5, 6, 9
+27.6 +0.24 +0.52
PG 1229+204 37.8- 15.3 3828 ± 54 8.03- 0.23 - 1.03- 0.55 43.70 ± 0.05 42.31 ± 0.06 209.7 ± 38.3 4, 5, 6, 9
+0.8 +0.08 +0.41
NGC 4593 3.7- 0.8 5143 ± 16 7.28- 0.10 - 0.73- 0.52 42.87 ± 0.18 41.17 ± 0.18 101.6 ± 59.0 4, 6, 17
+1.3 +0.13 +0.41
4.3- 0.8 4395 ± 362 7.21- 0.12 - 1.47- 0.52 42.38 ± 0.18 40.73 ± 0.18 115.4 ± 67.6 18b
+0.8 +0.09 +0.60
4.0- 0.7 5142  572 7.26- 0.09 - 1.10- 0.64 42.62 ± 0.37 40.95 ± 0.33 108.3 ± 45.7 L
+5.5 +0.27 +0.98
IRAS F12397 9.7- 1.8 1802 ± 560 6.79- 0.45 2.26- 0.62 44.23 ± 0.05 42.26 ± 0.04 54.2 ± 8.4 1,2,3c
+1.6 +0.11 +0.61
NGC 4748 5.5- 2.2 1947 ± 66 6.61- 0.23 0.10- 0.44 42.56 ± 0.12 40.98 ± 0.10 136.8 ± 50.1 4, 14

13
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Table 7
(Continued)

tHb FWHM log ( M· M) log M˙ log L 5100 log L Hb EW(Hβ) References
Objects
(days) (km s-1) -1
( erg s ) -1
( erg s ) (Å)
+36.0 +0.13 +0.53
PG 1307+085 105.6- 46.6 5059 ± 133 8.72- 0.26 - 0.68- 0.28 44.85 ± 0.02 43.13 ± 0.06 98.4 ± 15.1 4, 5, 6, 9
+1.2 +0.19 +1.33
NGC 5273 2.2- 1.6 5688 ± 163 7.14- 0.56 -2.50- 0.67 41.54 ± 0.16 39.74 ± 0.11 82.2 ± 37.1 19
+3.9 +0.09 +0.33
Mrk 279 16.7- 3.9 5354 ± 32 7.97- 0.12 -0.89- 0.30 43.71 ± 0.07 42.12 ± 0.06 132.2 ± 28.7 4,5,6,20
+61.0 +0.17 +0.63
PG 1411+442 124.3- 61.7 2801 ± 43 8.28- 0.30 - 0.23- 0.38 44.56 ± 0.02 42.85 ± 0.03 99.7 ± 8.2 4,5,6,9
+1.5 +0.03 +0.46
NGC 5548 19.7- 1.5 4674 ± 63 7.92- 0.04 - 1.62- 0.49 43.39 ± 0.10 41.79 ± 0.10 128.1 ± 40.3 4, 5, 6, 21
+2.1 +0.05 +0.47
18.6- 2.3 5418 ± 107 8.03- 0.06 -1.99- 0.51 43.14 ± 0.11 41.61 ± 0.13 151.3 ± 57.9 4, 5,6,21
+2.9 +0.07 +0.45
15.9-2.5 5236 ± 87 7.93- 0.08 - 1.68- 0.49 43.35 ± 0.09 41.72 ± 0.10 119.7 ± 37.9 4, 5, 6, 21
+1.9 +0.07 +0.47
11.0- 2.0 5986 ± 95 7.89- 0.09 - 2.10- 0.52 43.07 ± 0.11 41.52 ± 0.17 144.5 ± 66.7 4, 5, 6, 21
+1.6 +0.05 +0.46
13.0- 1.4 5930 ± 42 7.95- 0.05 - 1.72- 0.49 43.32 ± 0.10 41.75 ± 0.09 135.9 ± 41.2 4, 5, 6, 21
+3.8 +0.11 +0.45
13.4- 4.3 7378 ± 39 8.15- 0.17 -1.64- 0.49 43.38 ± 0.09 41.73 ± 0.10 114.4 ± 37.0 4, 5, 6, 21
+2.6 +0.05 +0.45
21.7- 2.6 6946 ± 79 8.31- 0.06 - 1.43- 0.48 43.52 ± 0.09 41.82 ± 0.09 102.4 ± 30.2 4, 5, 6, 21
+1.2 +0.03 +0.45
16.4- 1.1 6623 ± 93 8.15- 0.03 - 1.56- 0.48 43.43 ± 0.09 41.75 ± 0.10 106.3 ± 33.3 4, 5, 6, 21
+2.0 +0.05 +0.46
17.5- 1.6 6298 ± 65 8.13- 0.04 - 1.85- 0.49 43.24 ± 0.10 41.72 ± 0.10 153.5 ± 50.7 4, 5, 6, 21
+4.3 +0.07 +0.45
26.5- 2.2 6177 ± 36 8.30- 0.04 - 1.33- 0.48 43.59 ± 0.09 41.87 ± 0.10 98.1 ± 30.0 4, 5, 6, 21
+3.2 +0.05 +0.45
24.8- 3.0 6247 ± 57 8.28-0.06 - 1.45- 0.48 43.51 ± 0.09 41.83 ± 0.09 106.6 ± 31.5 4, 5, 6, 21
+5.7 +0.27 +0.47
6.5- 3.7 6240 ± 77 7.69- 0.37 - 2.04- 0.51 43.11 ± 0.11 41.64 ± 0.13 172.8 ± 66.1 4, 5, 6, 21
+5.9 +0.15 +0.47
14.3- 7.3 6478 ± 108 8.07-0.31 - 2.03- 0.51 43.11 ± 0.11 41.55 ± 0.14 139.8 ± 55.7 4, 5, 6, 21
+2.6 +0.15 +0.49
6.3-2.3 6396 ± 167 7.70- 0.20 -2.27- 0.55 42.96 ± 0.13 41.12 ± 0.10 74.4 ± 27.3 4, 22
+0.9 +0.08 +0.47
4.2- 1.3 12771 ± 71 8.12- 0.16 -2.19- 0.51 43.01 ± 0.11 41.33 ± 0.10 105.3 ± 35.4 4, 14
+2.7 +0.09 +0.48
12.4- 3.9 11481 ± 574 8.50- 0.17 - 2.21- 0.53 42.99 ± 0.11 41.27 ± 0.10 96.6 ± 34.4 4, 13a
+6.4 +0.16 +0.33
17.6- 4.7 7241  2200 8.10- 0.16 -1.80- 0.33 43.29 ± 0.20 41.64 ± 0.23 117.2 ± 25.9 L
+29.9 +0.12 +0.47
PG 1426+015 95.0- 37.1 7113 ± 160 8.97- 0.22 - 1.51- 0.28 44.63 ± 0.02 42.83 ± 0.04 80.1 ± 9.2 4, 5, 6, 9
+3.9 +0.08 +0.35
Mrk 817 19.0- 3.7 4711 ± 49 7.92- 0.09 - 0.81- 0.35 43.79 ± 0.05 42.07 ± 0.05 98.0 ± 16.5 4, 5, 6, 7
+3.7 +0.09 +0.35
15.3- 3.5 5237 ± 67 7.91- 0.11 - 0.98- 0.35 43.67 ± 0.05 42.00 ± 0.06 108.5 ± 20.4 4, 5, 6, 7
+6.5 +0.08 +0.35
33.6- 7.6 4767 ± 72 8.17- 0.11 - 0.98- 0.35 43.67 ± 0.05 41.92 ± 0.05 91.1 ± 15.3 4, 5, 6, 7
+3.4 +0.09 +0.35
14.0-3.5 5627 ± 30 7.94- 0.12 - 0.73- 0.35 43.84 ± 0.05 41.77 ± 0.05 43.2 ± 7.1 4, 13
+9.9 +0.14 +0.22
19.9- 6.7 5348  536 7.99- 0.14 - 0.87- 0.22 43.74 ± 0.09 41.93 ± 0.14 78.5 ± 34.3 L
+0.9 +0.07 +0.24
Mrk 1511 5.7- 0.8 4171 ± 137 7.29- 0.07 - 0.34- 0.24 43.16 ± 0.06 41.52 ± 0.06 115.5 ± 23.1 18b
+1.2 +0.07 +0.23
Mrk 290 8.7- 1.0 4543 ± 227 7.55- 0.07 - 0.85- 0.23 43.17 ± 0.06 41.64 ± 0.06 153.0 ± 29.0 4,13a
+7.5 +0.12 +0.20
Mrk 486 23.7- 2.7 1942 ± 67 7.24- 0.06 0.55- 0.32 43.69 ± 0.05 42.12 ± 0.04 135.9 ± 20.3 1, 2, 3
+1.2 +0.04 +0.33
Mrk 493 11.6- 2.6 778 ± 12 6.14- 0.11 1.88- 0.21 43.11 ± 0.08 41.35 ± 0.05 87.4 ± 18.1 1, 2, 3
+15.0 +0.14 +0.45
PG 1613+658 40.1- 15.2 9074 ± 103 8.81- 0.21 - 0.97- 0.31 44.77 ± 0.02 43.00 ± 0.03 86.7 ± 7.6 4, 5, 6, 9
+29.6 +0.15 +0.58
PG 1617+175 71.5- 33.7 6641 ± 190 8.79- 0.28 - 1.50- 0.33 44.39 ± 0.02 42.74 ± 0.05 114.8 ± 15.1 4,5,6,9
+45.9 +0.08 +0.17
PG 1700+518 251.8- 38.8 2252 ± 85 8.40- 0.08 1.08- 0.17 45.59 ± 0.01 43.78 ± 0.02 78.9 ± 4.5 4, 5, 6, 9
+6.2 +0.10 +0.60
3C 390.3 23.6- 6.7 12694 ± 13 8.87- 0.15 - 3.35- 0.65 43.68 ± 0.10 42.29 ± 0.05 206.2 ± 50.7 4, 5, 6, 23
+3.6 +0.03 +0.51
46.4- 3.2 13211 ± 28 9.20- 0.03 - 2.12- 0.51 44.50 ± 0.03 42.78 ± 0.04 97.1 ± 10.0 4, 24
+27.6 +0.23 +0.95
44.5-17.0 12796  361 9.18- 0.23 -2.62- 0.96 44.43 ± 0.58 42.60 ± 0.35 108.8 ± 58.8 L
+2.0 +0.07 +0.25
KA 1858+4850 13.5- 2.3 1820 ± 79 6.94- 0.09 0.75- 0.21 43.43 ± 0.05 41.89 ± 0.04 146.9 ± 21.1 25d
+0.9 +0.05 +0.46
NGC 6814 6.6- 0.9 3323 ± 7 7.16- 0.06 - 1.64- 0.80 42.12 ± 0.28 40.50 ± 0.28 121.6 ± 112.2 4,14
+6.1 +0.03 +0.13
Mrk 509 79.6- 5.4 3015 ± 2 8.15-0.03 - 0.52- 0.14 44.19 ± 0.05 42.61 ± 0.04 132.7 ± 19.1 4, 5, 6, 7
+1.2 +0.08 +0.23
PG 2130+099 9.6- 1.2 2450 ± 188 7.05- 0.10 1.69- 0.20 44.20 ± 0.03 42.65 ± 0.03 142.1 ± 13.6 4,8a
+3.4 +0.12 +1.83
NGC 7469e 10.8- 1.3 4369 ± 6 7.60- 0.06 0.90- 1.87 43.51 ± 0.11 41.60 ± 0.10 63.0 ± 21.0 4,26
+0.7 +0.06 +1.85
4.5-0.8 1722 ± 30 6.42- 0.09 0.63-1.90 43.32 ± 0.12 41.71 ± 0.09 124.9 ± 44.3 27, 5
+5.8 +0.84 +1.31
6.5- 3.0 4343  2859 6.92- 0.84 0.76- 1.35 43.43 ± 0.15 41.66 ± 0.10 86.9 ± 47.1 L

Notes.
a
Means that the literature does not provide FWHMs from mean spectra or Hβ fluxes with narrow-line components subtracted. For these, we scanned and digitized the
mean spectra from the papers, then calculated those numbers ourselves.
b
Means that the AGN continuum is obtained through spectral fitting decomposition (Reference 18).
c
For IRAS F12397, we use fluxes for the case with local absorption correction (see the details in Papers I and III).
d
Means that the contribution of host galaxy is provided in Reference 25.
e
The virial products of the two independent campaigns are quite different, yielding a large error bars on the mean M· and M˙. In Appendix B, NGC 7469 is shown as
an SEAMBH from Collier et al. (1998) and Peterson et al. (2014). Both campaigns have high quality data. Numbers in boldface are the weighted averages of all the
measurements of this object. The details of the average scheme are given in the main text.
References. (1) Paper I, (2) Paper II, (3) Paper III, (4) Bentz et al. (2013), (5) Collin et al. (2006), (6) Kaspi et al. (2005), (7) Peterson et al. (1998), (8) Grier et al.
(2012), (9) Kaspi et al. (2000), (10) Santos-Lleó et al. (1997), (11) This paper, (12) Bentz et al. (2009a), (13) Denney et al. (2010), (14) Bentz et al. (2009b), (15)
Stirpe et al. (1994), (16) Bentz et al. (2006), (17) Denney et al. (2006), (18) Barth et al. (2013), (19) Bentz et al. (2014), (20) Santos-Lleó et al. (2001), (21)
Peterson et al. (2002) and references therein, (22) Bentz et al. (2007), (23) Dietrich et al. (1998), (24) Dietrich et al. (2012), (25) Pei et al. (2014), (26) Peterson
et al. (2014), (27) Collier et al. (1998).

14
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 3. The R Hb -L 5100 and R Hb -L Hb relationships (panels (a), (c), (e)) and the deviations of R Hb from the regression relationships (panels (b), (d), (f)). The
dotted is the fit for the M˙ < 3 sample whereas the dashed is the fit for the M˙ > 3 sample in all panels. Panels (a) and (b) show the comparison of the previous sample
and the SEAMBH campaigns. Clearly SEAMBHs are below the R Hb -L 5100 relation, and the scatter of the relation increases with the inclusion of the SEAMBH2012/
( )
2013 samples (see the intrinsic scatter given by the numbers below Equations (4) and (5)). Panel (b) shows DRHb = log R Hb R Hb , R - L , where R Hb , R - L is given by
Equation (4b). Panels (c) and (d) show the two samples of M˙ ⩾ 3 and M˙ < 3 AGNs with averaged RM values for the R Hb -L 5100 relation and the residual of DR Hb ,
where R Hb , R - L is given by Equation (5b). Panels (e) and (f) show the R Hb -L Hb and DR Hb -M˙ relations.

the population into two sub-groups, those with M˙ ⩾ 3 (11 to wait for future observations of luminous, high accretion rate
from earlier studies and 13 from our study) and those with sources.
M˙ < 3. The diagram emphasizes that much of the intrinsic
scatter is caused by a systematic deviation of the large M˙ 4.2. Hβ Luminosity Versus BLR Size
sources toward shorter lags. The regression calculations give
Figure 3(c) shows the new R Hb -L Hb relationship. Such
(
log R Hb ltd = ) relationships have been used in the past to supplement the
continuum-based relationships and to correlate the BLR size
ìï (1.45  0.04) + (0.50  0.03) log l44 (entire sample) more closely with the ionizing continuum radiation (see e.g.,
ïï Kaspi et al. 2005). The regression analysis gives
ï (1.55  0.04) + (0.53  0.04) log l44 ( M˙ < 3),
í
ïï
ïï (1.32  0.06) + (0.53  0.06) log l44 ( M˙ ⩾ 3), (5)
î
(
log R Hb ltd = )
ì
ï (1.30  0.03)
with intrinsic scatters of sin = (0.21, 0.16, 0.22) for (5a, 5b, ï
ï (entire sample) ,
ï
ï + (0.51  0.03) log L Hb ,42
5c), respectively. The slope of the correlation for the SEAMBH ï
sample, those with M˙ ⩾ 3, is comparable to that of sub- ï
ï
ï (1.36  0.04)
Eddington AGNs, but the normalization is significantly í
ï + ( 0.52  0.04 ) log L
( M˙ < 3), (6 )
ï H b ,42
different. Clearly, the Hβ region of SEAMBHs is smaller than ï
ï
sub-Eddington AGNs (see the different intercepts in ï (1.21  0.06)
ï
ï ( M˙ ⩾ 3),
Equations (5b) and (5c)). ï
ï + ( 0.53  0.06 ) log L H b ,42
A more complete approach is to analyze the dependence î
of t Hb on L5100 and M˙ together, however, several reasons where L Hb,42 = L Hb 10 42 erg s-1, and sin = (0.20, 0.18, 0.22)
prevent us from performing such an analysis. First, the current
sample is still small; and second the luminosity range of for (6a, 6b, 6c), respectively. As in the continuum luminosity
SEAMBHs is very narrow. There are only 4 luminous case, the inferred BLR size for SEAMBHs is smaller than the
SEAMBHs (PG 0026+129, PG 1211+143, PG 1226+023, size of the sources with M˙ < 3 but the differences are smaller
PG 1700+518) identified from the previous campaigns, but in this case.
their M=˙ (4.5, 6.9, 5.0, 12.0), respectively, are smaller than The R Hb -L Hb relationship has been examined by Wu et al.
the average of this group. Thus, the search for a new (2004) and Kaspi et al. (2005). A comparison with the results
relationship of the type of R Hb = R Hb (L 5100, M˙) will have of these papers shows similar intercepts but different slopes,

15
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 4. Panels (a), (b), (c), and (d) show the DR Hb -M˙ relation and DR Hb distributions for R Hb -L 5100 and R Hb -L Hb relations, respectively. The dotted lines
indicate M˙ = 3 in panels (a) and (c). Both panels show that DR Hb ~ 0 up to M˙ ~ 3, beyond which ∣ DR Hb ∣ increases with M˙. This implies that, relative to normal
AGNs of the same 5100 Å luminosity, SEAMBHs have Hβ lags that become systematically shorter with increasing accretion rate.

with the present slopes being significantly flatter than AGNs is a poor approximation for a more complex situation
0.685 ± 0.106 found in Wu et al. (2004) and 0.694 ± 0.064 where both the luminosity and the accretion rate determine this
found by Kaspi et al. (2005). It is not at all clear that these are relationship. From the regression, we get the dependence of the
significant variations since the present sample is much larger deviations of R Hb from the R Hb -L 5100 relation in Figure 4(a)
and the distribution of sources along the L Hb axis is quite
different. The correlations involving L5100 and L Hb (Figures 3 DR Hb = (0.36  0.13) - (0.46  0.10)
and 9 in Appendix B) show a very similar scatter but, for a
´ log M˙ ( for M˙ ⩾ 3), (7)
given Hβ luminosity, SEAMBHs have shorter lags compared
with low-accretion rate AGNs. Such differences may be related
to the properties of slim accretion disks (Wang et al. 2014b), or with sin = 0.07. From the R Hb -L Hb relation in Figure 4(c), we
other types of anisotropies (e.g., Goad & Korista 2015; get
Netzer 1987; O’Brien et al. 1994; Goad & Wanders 1996;
Ferland et al. 2009). DR Hb = (0.33  0.15) - (0.38  0.10)
´ log M˙ ( for M˙ ⩾ 3), (8)
4.3. M˙-dependent BLR Size
To test the dependence of the BLR size on accretion rates, with sin = 0.13. We have tested the above correlations also for
(
we define a new parameter, DR Hb = log R Hb R Hb, R - L that ) M˙ < 3. The FITEXY regressions give slopes of around zero
specifies the deviations of individual objects from the with very large uncertainties, DR Hb µ M˙ -0.085  0.051 and
R Hb -L 5100 L Hb relationship of the sub-sample of M˙ < 3.0 -0.059  0.054
DR Hb µ M˙ for Figures 4(a) and (c), respectively,
sources (i.e., R Hb, R - L as given by Equations (4b), (5b), and
implying that DR Hb does not correlate with M˙ for M˙ < 3
(6b). The scatter of DR Hb is calculated by s=
group. All this confirms that M˙ is an additional parameter
é ù1 2
ê å i ( DRHb, i - áDRHb ñ ) N ú , where N is the number of controlling the R Hb -L 5100 relation in AGNs with high accretion
2

ë û rates.
objects and áDRHb ñ is the averaged value. Figure 3 provides To summarize, the new SEAMBHs observed in 2012 and
the value ofσ for comparison. 2013 significantly increase the scatter of the R Hb -L 5100
Figure 4 shows plots of DR Hb versus M˙ , and DR Hb
relation. We find that SEAMBHs have significantly shorter
distributions for the M˙ ⩾ 3 and M˙ < 3 samples. A Kolmo- lags than those of sub-Eddington AGNs with similar 5100 Å
gorov–Smirnov (KS) test comparing the two shows that the luminosity, and the shortened lags increase with the dimension-
probability of the same parent distributions is pKS = 0.0054 for less accretion rate. Given this, we recommend to use Equations
the R Hb -L 5100 case and pKS = 0.014 for R Hb -L Hb case. This (4b) or (5b) when trying to estimate the BLR size in sub-
provides a strong indication that the main cause of deviation Eddington AGNs. We suggest that the dependence of R Hb on
from the old R Hb -L 5100 relationship (or the R Hb -L 5100 the dimensionless accretion rate could be a consequence of the
relationship for the sub-Eddington AGNs) is the extreme anisotropic radiation of slim disks. We come back to this issue
accretion rate. Thus, a single R Hb -L 5100 relationship for all in the following section.

16
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

5. DISCUSSIONS 5.2. Accretion Disks and the Maximum Size of the BLR
The new results presented here illustrate that the small scatter 5.2.1. Geometrically Thin Disks
in the earlier R Hb -L 5100 relationships (Kaspi et al. 2005; Bentz
Here we explore the possibility that the observed changes in
et al. 2013 and references therein), only applies to low BLR as a function of M˙ are due to the transition from thin
accretion rate AGNs, which, according to our definition, are (M˙ < 3) to slim (M˙ ⩾ 3) disks. To start, we express the
sources with M˙ < 3. The addition of high accretion rate measured Hβ lags in units of the gravitational radius,
sources, those referred to as SEAMBHs, changes this picture. (Equation
Rg = 1.5 ´ 10 5 (M· M ) cm. Since R Hb µ l 44 b
The scatter increases and there is a clear, statistically significant
2 3
tendency for AGNs with higher accretion rates to show smaller (1)), and l 44 µ M˙ M· (Equation (2), recast), we expect
4 3

R Hb . Moreover, the deviation increases with M˙ . This holds for

R Hb
m 7-0.29,
0.35
both L5100 and L Hb although the deviations are larger when r Hb = = 1.9 ´ 10 4 M˙ (9 )
Rg
using the 5100 Å luminosity. The deviations are most notice-
able over the luminosity range where most SEAMBHs are where we use Equation (5b). Equation (9) concisely mergers
concentrated, between 1043 and few ´10 44 erg s-1. Below we the empirical R Hb -L 5100 relation with the basic expectations of
explore the physical implications of the new results focusing on
the earlier suggestion (Paper II) that SEAMBHs are powered geometrically thin disks.
by slim accretion disks with properties that are considerably We now defined a new parameter, Y = m 70.29 r Hb , which we
different from those of thin disks. coin “the radius–mass parameter,” i.e.,
0.35
Y = 1.9 ´ 10 4M˙ . (10)
5.1. M˙ and the Origin of the Shorter Time Lag
We calculated this combined parameter for all sources in order
The deviations from the earlier R Hb -L 5100 relationship raise to test whether it depends on accretion rate. This test would
several interesting possibilities that can explain the new results. provide clues about the various modes of accretion and their
The first possibility is that the ionizing flux reaching the BLR,
relationship with the size of the BLR.
which determines its RM-measured size, is different from the
Similar to the definition of DR Hb in Section 4.3, we define
flux inferred by a remote observer due to an unusual SED and/
or the anisotropy of the radiation emitted from the disk. This DX = log (X X reg), where X is the measured value for
possibility has been mentioned in the past with regards to the individual sources and X reg is the value obtained for the group
covering factor and the level of ionization of the BLR. Netzer from the regression analysis. In this section we consider two
(1987) presented a detailed model where the combination of an cases, X = r Hb and X = Y . The various tests performed on the
angle-dependent SED of geometrically thin accretion disks (see accretion rate groups are shown in Figure 5 and discussed
also Czerny & Elvis 1987; Li et al. 2010 for geometrically thick below.
parts of the disks), with an isotropic X-ray source, results in a Our first test for r Hb is performed on the two accretion rate
weaker ionizing continuum along the plane of the disk groups, M˙ < 3 and M˙ ⩾ 3. Figure 5(a) shows the r Hb-M˙
compared with the pole-on direction. This will result in a
complicated angle-dependent level of ionization. The recent relation. For M˙ < 3 objects, we find r Hb µ M˙ 0.50  0.08 with
work by Wang et al. (2014b) suggests that large SED sin = 0.24, which is steeper than expected (Equation (9)). By
variations, especially in the ionizing luminosity as a function contrast, r Hb µ M˙ 0.35  0.12 for the M˙ ⩾ 3 group, consistent
of the polar angle, will be most noticeable in slim accretion with Equation (9). It is clear that the r Hb-M˙ relation is
disks where the anisotropy of the disk is more extreme because markedly different for the two groups, suggesting that
of self-shadowing effects. Such a geometry would lead to a SEAMBHs are physically distinct from sub-Eddington AGNs.
smaller emissivity-weighted BLR radius at large angles relative Dr Hb and its distributions are shown in the lower panels of
to the polar axis. The second possibility is that much of the
Figure 5(a). A KS test gives the probability of
effect is due to the change of the bolometric luminosity with
polar angle due to the reasons discussed above. This will pKS = 1.3 ´ 10-3 and demonstrates the differences again for
reduce the dust sublimation radius that has a large effect on the the M˙ ⩾ 3 and the M˙ < 3 groups. We suspect that the
0.35
measured RM size (see below). Obviously, the covering factor deviations from the expected dependences on M˙ are due to
of the BLR, which determines L Hb , and the anisotropy of the the dependence on m7 in our sample that contain a large range
Balmer line emission, could also affect these relationships. of BH mass.
The search for a direct ionizing luminosity indicator initiated Figure 5(b) shows the Y-M˙ relation, which has much
several attempts to replace L5100 by L Hb in the R Hb -L 5100 smaller scatter than the r Hb-M˙ relation, for both groups. The
relationship because Hβ responds more directly to the ionizing regression gives Y µ M˙ 0.41  0.07 for the M˙ < 3 group with
luminosity (e.g., Peterson et al. 1993). Kaspi et al. (2005) sin = 0.15, which, within the errors, agrees well with
show that this substitution had no affect on the deduced BLR Equation (10) confirming the idea of thin accretion disks in
size. Similar ideas have been proposed by Wang & Zhang objects with M˙ < 3. The diagram clearly shows a change of
(2003), Wu et al. (2004), and others. Our new observations slope at a transition accretion rate of M˙ c  3 from a rising to
hint to the possibility that the changes in the L Hb relationship an almost horizontal constant (saturated) value. This transition
relative to the relationship for the M˙ < 3 group are smaller corresponds to a saturated-Y of Ysat ~ 3 ´ 10 4 . This is
than in the L5100 case. Unfortunately, the relatively small size consistent with the idea that beyond this critical rate, which
of the SEAMBH sample prevents us from quantifying these we name M˙ c , the central power-house is no longer a thin
ideas. accretion disk. Obviously, such a test could not be performed

17
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 5. Black hole mass-scaled BLR and the radius–mass parameter vs. dimensionless accretion rate. In each panel, there are three plots to show their relation and
deviations. Correlations and probabilities of KS tests are marked in the plots. Panel (a) shows that the M˙ < 3 sample has a steeper correlation than that given by
Equation (9). Panel (b) shows that SEAMBHs deviate from this relation and that Y is independent of M˙. Clearly Y behaves in a different way in the two accretion rate
groups.

without the new observational data from SEAMBH2012 and

SEAMBH2013. This is not entirely unexpected since previous
works specifies in the introduction to this paper, and explain in
more detail below, show that the SS73 disk model becomes
invalid above a certain accretion rate (e.g., Laor &
Netzer 1989).

5.2.2. The Transition Accretion Rate

Figure 6 shows the determination of the transition accretion
rate and the saturated-Y. To determine M˙ c from the data, we
first test the dependence of the index b of the R Hb -L 5100
relation on M˙ c , and find b = (0.527  0.038, 0.528
0.030, 0.533  0.028) for samples of objects with
M˙ ⩽ M˙ c = (3, 10, 20). These changes are so small that we
decided to adopt b = 0.53 for all the groups. In order to obtain
M˙ c and Ysat , we use the functional form13

ïìï æ M˙ ök 0
ï Ysat çç ÷÷ ( M˙ < M˙ c ),
Y = ïí çèç M˙ c ÷÷ (11)
ø
ïï Figure 6. Determination of the saturated-Y and the transition accretion rate. We
ïï Ysat
ïî ( M˙ ⩾ M˙ c ), find Y µ M˙
0.27  0.04
, supporting the SS73 disk model below M˙ c and
saturation beyond the numbers listed inside the panel supporting the idea of
slim accretion disks in SEAMBHs. The outlier with large error bars is NGC
where Ysat , k0, and M˙ c are to be determined by the data. We 7469 (see Table 7).
make use of the Levenberg–Marquardt method to fit the entire
sample with the inclusion of error on M˙ through Monte-Carlo the uncertainties of Y until c 2 = 1. This gives
simulations (e.g., Press et al. 1992). In the fitting, we increase
+ 19.6 + 0.6
M˙ c = 13.8- 8.1 , Ysat = 3.3- 0.5 ´ 10 , and
4
13
We also tried to fit the data with a broken power law, k 0 = 0.27  0.04. (12)
Y = Ysat M˙ 1 [1 + (M˙ M˙ c )k2 ]-1, where k1, k2, Ysat , and M˙ c can be obtained
k

from the fit. This did not result in a satisfactory fit. The reasons are: (1.) Many The value of k0 is consistent, within 2s , with the expected
more objects with low M˙ regardless of the exact value of M˙ c . (2.) The overall
range in M˙ for low accretion rate sources (more than 3.5 dex) is much larger value of 0.35, providing empirical support, from the basis of
than for high accretion rate sources (less than 1.5 dex). RM experiments, for the geometrically thin disk model.

18
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

The analysis shows high uncertainty on the value of M˙ c . significantly increases the output power of radiation from slim
This is due to the limited numbers of objects with M˙ > 13.8 disks. However, such a process could be greatly suppressed by
(13 objects). The precision we can determine the values of the fast radial motion of the accretion flows so that the radiation
M˙ c and Ysat would most likely improve when large samples of remains at a level of low efficiency as the classical model
high accretion rates become available. (Abramowicz et al. 2013). The detailed discussion of these
theoretical issues is beyond the scope of this paper. Here we
5.2.3. Slim Accretion Disks focus on the attempt to use our new sample in order to
qualitatively compare with the saturated luminosity and the
To understand the dependence of the radius–mass parameter transition accretion rates. Quantitative comparison with models
on the accretion rates, we need to consider the properties of of slim disks is given in a forthcoming paper.
slim accretion disks. Such disks are the result of the increasing The best measured value of Ysat (Equation (12)) results in
mass accretion flow through a thin accretion disk which leads two expressions for the critical BLR size,
to a build-up of radiation pressure close to the BH, the
thickening of the disk, and the trapping of the photons that are + 0.6
trying to escape from these regions. Such systems have been r Hb = 3.3- 0.5 ´ 10 m 7
4 0.29
; or
discussed in numerous papers (e.g., Abramowicz et al. 1988; + 3.4
R crit (Hb ) = 18.8- 1.29
(14)
2.8 m 7 ltd,
Wang & Zhou 1999; Mineshige et al. 2000; Watarai
et al. 2001; Sadowski 2009). The general result obtained from
these studies is a simple dependence of the bolometric which are independent of the accretion rates for AGNs with
luminosity of the system linearly on the BH mass and only a M˙  M˙ c . In the context of photoionization, the critical BLR
weak logarithmic dependence on the mass accretion rate (so size corresponds to a saturation of the ionizing luminosity, and/
called “saturated luminosity”). The characteristic saturated or the bolometric luminosity. In particular, the small scatter of
bolometric luminosity can be expressed as (Mineshige Ysat indicates that the scatter of the saturated luminosity is small
et al. 2000; Watarai et al. 2001) in the current SEAMBH sample. Obviously, the Y -saturation
qualitatively supports the saturated luminosity shown by
L Bol = ℓ0 éê 1 + ln ( M˙ M˙ c ) ùú,L Edd ( for M˙ ⩾ M˙ c ), (13) Equation (13), but Equation (14) provides a quantitative way
ë û
to test theoretical models and numerical simulations of slim
where ℓ0 = 2 ~ 3 and M˙ c ~ 20 . The uncertainties on ℓ0 and disks. The ultimate test for this finding will be performed once
M˙ c depend on the details of the vertically averaged equations there are enough objects with the same BH mass with directly
of slim disks. Calculations of emergent spectra of slim measured R Hb .
23
disks show that L 5100 µ M˙ M·4 3 very similar to the SS73 While the uncertainty on the preferred value of the transition
disk, whereas the hydrogen ionizing luminosity accretion rate is large, because of the small number of objects,
¥
(L ion = ò L d , where ϵ is photon energy) increases with it is clear that the value obtained from fitting our sample is
13.6eV 
M much slower than L5100 (Wang et al. 1999, 2014b;
˙ considerably lower than M˙ c = 50 preferred by Mineshige
Mineshige et al. 2000; Watarai et al. 2001) and, for small to et al. (2000), but covers M˙ c = 20 within the error bar given
by Watarai et al. (2001). It has been suggested that the classical
intermediate size BHs, L ion » L Bol for large M˙ , showing a
slim disk models may have underestimated the effects of
characteristic saturation of L ion with M˙ . A canonic spectral photon trapping. In fact, M˙ c  10 is predicted by model A of
energy distribution, L µ  -1 exp (-  0 ), is expected from Ohsuga et al. 2002; their Figure 1). In this model, it is assumed
the photon trapping part of slim disks, where  0 is the cut-off that most of dissipated gravitational energy is released in the
energy determined by the maximum temperature of the disk. mid-plane and the vertical structure of gas density is presumed;
Such a SED follows from a temperature distribution of however, this assumption is not self-consistent. We also note
Teff µ R-1 2 , where R is the radius. This is significantly flatter that the minimum accretion rates of M˙ min = 3 used by us to
than that in the SS73 disk (Wang & Zhou 1999; Mineshige define SEAMBHs is outside the range found here. However,
there must be a smooth transition from thin to slim disks and a
et al. 2000). 0.35
The calculations of the structure and emitted spectrum of relatively large range in M˙ from linear dependence on M˙
slim accretion disks are very challenging because of the to full saturation.
complicated coupling between structure and radiation transfer. In summary, the Y -saturation measured from RM experi-
The classical model (Abramowicz et al. 1988) uses the ments provides a new tool to test the physics of accretion onto
vertically averaged equations (their validity regimes were BHs. We find that the current data support both sides of disk
discussed by Gu & Lu 2007; Cao & Gu 2015), which probably paradigm, from thin disks with the expected dependence on
underestimate photon trapping effects (Ohsuga et al. 2002). mass and accretion rate, to slim disks with their expected
The model neglects the time delay between energy generation saturation.
around mid-plane and photon escaping from the surface of the
disks. Therefore, both the saturated luminosity and the
transition accretion rates are not well determined by theory. 5.2.4. The Maximum Size of the BLR
Numerical simulations of slim disks show different, perhaps We can also calculate the maximum expected dimensionless
more realistic properties, but the agreement between 2D fully radius, r Hb , using pure observational considerations. For this we
general relativistic (Sadowski et al. 2014) and 3D magnetohy- use Equation (3) and the definition of r Hb from Equation (9).
drodynamic calculations (Jiang et al. 2014) is not very good. In
-1
particular, in 3D simulations, the inclusion of vertical This gives r Hb = R Hb Rg = f BLR ( c VFWHM )2 . This allows us to
transportation of radiation fluxes driven by magnetic buoyancy use the smallest observed Hβ line width to derive a mass-

19
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

independent maximum dimensionless size for all RM sources, There are several interesting ideas about the location of the
innermost walls of the torus around slim accretion disks (e.g.,
æ
-1 ç c ÷
ö2 -1 -2
Kawakatu & Ohsuga 2011; Kawaguchi 2013). In particular, the
r Hmax = f BLR çç ÷ = 9 ´ 10 4 f BLR V min,3, (15) large attenuation of the slim disk radiation at high polar angles
b
çè Vmin ÷÷ø (Wang et al. 2014b) would predict a small Rdust at these
locations, perhaps as small as the self-gravity radius of the disk.
where Vmin,3 = Vmin 103 km s-1 is the minimum observed This is not evident in the present data which suggests that the
FWHM(Hβ). geometry of the inner torus walls is more complicated than the
Our sample includes one source, Mrk 493, with simplistic model suggested in various papers.
FWHM(Hb) ~ 780 km s-1. This width is, arguably, the
smallest FWHM observed so far (see e.g., Zhou et al. 2006; 5.4. SEAMBHs and Cosmology
Hu et al. 2008b). This translates to r Hmax b
 1.5 ´ 10 5. As The new results presented here suggest that the R Hb -L 5100
explained, this maximum gravitationally scaled size is relationship for AGNs requires an additional parameter. In
independent of the BH mass. For systems powered by thin particular, the use of the earlier correlations that are based on
accretion disks we use the definition of low accretion rate sources, tend to over-estimate the BLR size
rcrit = 2.7 ´ 10 4 m 7-0.29 (by inserting M˙ = 3 into Equation of high-M˙ sources and hence their BH mass. The shorter Hβ
(9)) to find the BH mass requires to get such a scaled radius, lags found here indicate that these objects have smaller BLR
m 7  0.003. The data in Table 6 shows that for Mrk 493 size than the ones obtained by applying the old R Hb -L 5100
m 7  0.14 and M˙  76, which indicates that this AGN is not relationships. This implies smaller BH mass and, therefore,
powered by a thin disk. It is therefore not surprising that the higher Eddington ratios. Recent work by Nobuta et al. (2012),
two mass estimates are not in agreement. Kelly & Shen (2013), and Netzer & Trakhtenbrot (2014) all
show the large fraction of high L Bol L Edd sources at high
5.3. Slim Disks and the Size of the Torus redshifts. This number is likely to increase given the necessary
correction to the BH mass found here.
The emissivity-weighted radius of the BLR measured by RM There are several ways of using AGNs for cosmology. These
experiments is directly related to dust in the outskirts of the include using the measured diameters of various BLRs (Elvis
BLR, probably in a toroidal-shaped structure (the “central & Karovska 2002), the empirical R Hb -L 5100 relation (Horne
torus”). As argued in Barvainis (1987) and Netzer & Laor
(1993), the innermost boundary of the torus is set by the dust et al. 2003; Watson et al. 2011; Czerny et al. 2013), torus inner
sublimation temperature which introduces a simple edge reverberation (Hönig et al. 2014; Yoshii et al. 2014), and
Rdust µ L Bol
12
relationship. Recent dust RM experiments (Suga- rms X-ray spectra (La Franca et al. 2014). The use of
numa et al. 2006; Koshida et al. 2014) provided Rdust in 17 SEAMBHs as cosmological distance indicators has been
sources by following the response of the dust emission in the suggested by Wang et al. (2013) and explained in great detail
K-band to the variable V-band continuum. The results are in in Paper II where we examined the consequences of increasing
very close agreement with the estimated graphite-grain M˙ that leads to an almost constant L Bol M·. Marziani &
sublimation radius assuming a mean grain size of about Sulentic (2014a, 2014b) used a similar approach based on a
0.05 μm. Mor & Netzer (2012) estimated this radius to be sample of highly accreting AGNs identified by their 4D
eigenvector 1 (see also King et al. 2014). While we used RM-
Rdust = 0.5L 46 T1800 pc, where L 46 = L Bol 10 46 erg s-1 and
1 2 2.8
based BH mass measurements (Paper II), they used the
T1800 = Tsub 1800 K is the sublimation temperature.
standard single-epoch method, combined with photoionization
A comparison of the Koshida et al. (2014) observations and
modeling, to estimate R Hb .
the ones presented here for M˙ < 3 sources shows that, on
average, Rdust R Hb  4. For M˙ ⩾ 3 objects, we find The usefulness of the present method for cosmology depends
on the number of SEAMBHs with accurately measured M· at
Rdust R Hb  7 (based only on four objects). Since the numbers high redshifts. As shown above, the reduction in the BLR size
are too small to derive a meaningful correlations for the two in SEAMBHs compared to the size in low accretion rate
accretion rate sub-groups, we focus on the comparison of the sources (used so far to estimate BH mass) depends on M˙ and
measured BLR and torus sizes. can reach a factor of 3 ~ 4 for M˙ = 100 (Figure 4(a) and
Figure 7 compares the dust and Hβ radii as a function of Equation (7)). It is reasonable to expect that higher values of
M˙ in two different ways. The left panel (a) shows the M˙ are associated with an even greater reduction relative to
dimensionless radii of the BLR and the torus inner walls for 14 expressions currently used in the literature. At redshifts
objects in the Koshida et al. (2014) sample for which we have z = 1 ~ 2, the reduction is more significant than the increase
measured Hβ lags. The right panel (b) shows the ratio of the in the measured lag due to the cosmological dilution factor
two radii for all these objects. The surprising result is the (1 + z ) suggesting that many SEAMBHs at high redshifts,
unexpectedly large scatter in this ratio. While Hβ and the hot even those with higher luminosities, can be monitored,
dust follow the continuum luminosity in basically an identical successfully, and their R Hb measured, during one observing
way (size proportional to L1 2 ), and differ only in their season, given adequate observing facilities.
normalization, this does not result in similar ratios between the
two distances for all sources. In fact, the ratios we measure
6. CONCLUSIONS
range from about 1.4 for Mrk 590 to about 10 for Mrk 335 and
NGC 4593. The median ratio, shown in the diagram, is about 4. We have completed two years of spectroscopic monitoring
The large scatter and the very small number of sources with of 18 AGNs that are high-accretion candidates. The second
measured dust lags prevent us from obtaining a meaningful year observations resulted in significant Hβ lag measurements
relationship between Rdust and M˙ . of five AGNs, all of which are identified as SEAMBHs. The

20
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 7. Black hole mass-scaled BLR (purple and gray points) and dust reverberation radii (black points). Panel (a) shows Rdust Rg and R Hb Rg vs. M˙. Panel (b)
shows ratios of Rdust R Hb as a function of M˙. Note the large scatter in both low and high M˙ groups.

highest accretion rate reaches M˙  200 in a few objects. The ingredients and the observations presented here will be
main results of our study can be summarized as follows. compared with more detailed calculations once these become
available.
1. For a given L5100, the SEAMBHs monitored by us
generally have shorter Hβ lags by a factor of up to ∼3–4 We are grateful to the anonymous referee for many useful
compared with lower accretion rate AGNs. The reduction and thoughtful comments and suggestions that helped us to
gets larger with increasing dimensionless accretion rates. improve the paper. We acknowledge the support of the staff of
2. We considered several possible explanations for the new the Lijiang 2.4 m telescope. Funding for the telescope has been
results and suggest that they may be related to the strong provided by CAS and the People’s Government of Yunnan
anisotropy and self-shadowing of the central slim disk Province. This research is supported by the Strategic Priority
radiation. We defined a new radius–mass parameter, Research Program—The Emergence of Cosmological Struc-
Y = m 70.29 r Hb and demonstrated that it reaches saturation tures of the Chinese Academy of Sciences, grant No.
when the accretion rate exceeds a critical value of XDB09000000, by NSFC grants NSFC-11173023,
+19.6 -11133006, -11373024, -11233003, and -11473002, and by
M˙ c = 13.8- 8.1 . This transition accretion rate is sig-
nificantly smaller than the one predicted by classical slim Israel-China ISF-NSFC grant 83/13.
disks model, but is in better agreement with modified slim
disk models. APPENDIX A
3. The shorter Hβ lags imply that the number of SEAMBHs MEAN AND RMS SPECTRA
in the universe probably has been underestimated in To supplement the material presented in the body of the
earlier works that did not take into account the smaller paper, we provide in Figure 8 the mean and rms spectra of the
BLR in such sources. It also helps to obtain lag objects in the SEAMBH2013 sample calculated by the standard
measurements in a shorter time, thus making the method (e.g., Peterson et al. 2004). Prior to the calculation of
application of SEAMBHs to cosmology more feasible. these spectra, we corrected the data for the slight wavelength
shift caused by the relatively wide aperture used in 2013 (5″)
Our observed sample, combined with SEAMBHs obtained using the [O III]l 5007 line as our wavelength reference. We
from the literature, is too small to put stronger constraints on define
the radius–mass saturation parameter. Larger SEAMBH
samples with a large dynamic range in M˙ , will provide better 1 N
understanding of the nature of transition from thin to slim disk F ( l) = åFi (l) ,
N i=1
(16)
systems. Better and more accurate BH mass measurements,
based on improved numerical tools such as the MCMC
technique as applied to velocity-resolved RM observations and
(Pancoast et al. 2011; Li et al. 2013), will help too. Finally, we
ì
ï1 N 2ü
ï
12
S ( l) = ï ï
are aware of the limitations of the current theory and numerical
í åë i é F ( l ) - F ( l ) ù
ûý , (17)
simulations of slim accretion disks. Present super-Eddington ï
î N i=1
ï ï
ï
þ
accretion models are still missing important physical

21
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 8. Mean and rms spectra (observed flux vs. rest-frame wavelength) of the objects in the SEAMBH2013 group.

where Fi (l) is the ith spectrum and N the total number of with scatters of sin = (0.20, 0.15, 0.24) for (19a, 19b, 19c),
spectra obtained during the campaign. The line dispersion is respectively. For DR Hb -M˙ relation, we have

sline
2
(l) = l2 - l 02 , (18) DR Hb = (0.49  0.14) - (0.58  0.10)
´ log M˙ ( for M˙ ⩾ 3), (20)
where l 0 = ò lP (l) dl ò P (l) dl ,
ál2ñ = ò l2P (l) dl ò P (l) dl , and P (l) is the line profile. and
The calculated values of sline are listed in Table 6.
DR Hb = (0.44  0.15) - (0.49  0.11)
APPENDIX B
R Hb CORRELATIONS USING THE DIRECT METHOD ´ log M˙ ( for M˙ ⩾ 3), (21)

The correlations listed in the main body of the paper were with sin = (0.01, 0.11) for deviations from R Hb -L 5100 and
calculated using the average scheme where each object is R Hb -L Hb relations, respectively. These correlations are shown
represented by one point. The results shown in this section are
in Figures 9 and 10.
for the direct scheme, where each observing campaign is
For the Y-M˙ relation as shown by Figure 11, we obtained
represented by one point.
The following equations correspond to Equation (6) = 10.8-+ 12.0 + 0.4
M˙c 5.7 , Ysat = 3.0- 0.4 ´ 10 , and
4

( )
log R Hb ltd = k 0 = 0.28  0.03, (22)

ïìï (1.29  0.02) from Figure 12, yielding the critical radius of
ïï (entire sample) ,
ïï + (0.52  0.03) log L Hb ,42 + 0.4
r Hb = 3.0- 0.4 ´ 10 m 7
4 0.29
; or
ïï
ï (1.35  0.03) + 2.3
R crit (Hb ) = 17.1- 1.29
(23)
í
ïï + (0.53  0.03) log L Hb ,42
( M˙ < 3), (19) 2.3 m 7 ltd.
ïï The results are very similar, and entirely consistent with the
ïï (1.17  0.05)
ïï + (0.55  0.06) log L ( M˙ ⩾ 3), results of the average scheme presented in the main body of the
ïî Hb ,42 paper.

22
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 9. Same as Figure 3 but for the direct scheme. The left panel and Figure 3(a) are identical, the right panel corresponds to Figure 3(e).

Figure 10. Same as Figure 4 but for the direct scheme. The object with the largest error bar is NGC 7469.

23
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Figure 11. Same as Figure 5 but for the direct scheme.

Bentz, M. C., Denney, K. D., Cackett, E. M., et al. 2006, ApJ, 651, 775
Bentz, M. C., Denney, K. D., Cackett, E. M., et al. 2007, ApJ, 662, 205
Bentz, M. C., Denney, K. D., Grier, C. J., et al. 2013, ApJ, 767, 149
Bentz, M. C., Horenstein, D., Bazhaw, C., et al. 2014, ApJ, 796, 8
Bentz, M. C., Peterson, B. M., Netzer, H., et al. 2009a, ApJ, 697, 160
Bentz, M. C., Walsh, J. L., Barth, A. J., et al. 2008, ApJL, 689, L21
Bentz, M. C., Walsh, J. L., Barth, A. J., et al. 2009b, ApJ, 705, 199
Blandford, R. D., & McKee, C. F. 1982, ApJ, 255, 419
Boroson, T. A., & Green, R. F. 1992, ApJS, 80, 109
Calvel, J., Reichert, G. A., Alloin, D., et al. 1991, ApJ, 366, 64
Cao, X., & Gu, W.-M. 2015, MNRAS, 448, 3514
Collier, S. J., Horne, K., Kapsi, S., et al. 1998, ApJ, 500, 162
Collin, S., Boisson, C., Mouchet, M., et al. 2002, A&A, 388, 771
Collin, S., Kawaguchi, T., Peterson, B. M., & Vestergaard, M. 2006, A&A,
456, 75
Czerny, B., & Elvis, M. 1987, ApJ, 321, 305
Czerny, B., Hryniewicz, K., Maity, I., et al. 2013, A&A, 556, 97
Davis, S. W., & Laor, A. 2011, ApJ, 728, 98
de Rosa, G., Peterson, B. M., Ely, J., et al. 2015, ApJ, submitted
(arXiv:1501.05954)
Denney, K. D., Bentz, M. C., Peterson, B. M., et al. 2006, ApJ, 653, 152
Denney, K. D., Peterson, M. C., Pogge, R. W., et al. 2010, ApJ, 721, 715
Dietrich, M., Kollatschny, W., Peterson, B. M., et al. 1993, ApJ, 408, 416
Dietrich, M., Peterson, B. M., Albrecht, P., et al. 1998, ApJS, 115, 185
Dietrich, M., Peterson, B. M., Grier, C. J., et al. 2012, ApJ, 757, 53
Du, P., Hu, C., Lu, K.-X., et al. (SEAMBH collaborations) 2014, ApJ, 782, 45
Elvis, M., & Karovska, M. 2002, ApJ, 581, 67
Ferrarese, L., Pogge, R. W., Peterson, B. M., et al. 2001, ApJL, 555, L79
Figure 12. Same as Figure 6 for the direct scheme. Ferland, G. J., Hu, C., Wang, J.-M., et al. 2009, ApJ, 707, 82
Giannuzzo, M. Z., Mignoli, M., Stirpe, G. M., & Comastri, A. 1998, A&A,
330, 894
REFERENCES Giannuzzo, M. Z., & Stirpe, G. M. 1996, A&A, 314, 419
Goad, M. R., & Korista, K. T. 2014, MNRAS, 444, 43
Abramowicz, M. A., Czerny, B., Lasota, J.-P., & Szuszkiewicz, E. 1988, ApJ, Goad, M., & Wanders, I. 1996, ApJ, 469, 113
332, 646 Grier, C. J., Peterson, B. M., Pogge, R. W., et al. 2012, ApJ, 755, 60
Abramowicz, M. A., Horák, J., & Kluźniak, W. 2013, AcA, 63, 267 Gu, W.-M., & Lu, J.-F. 2007, ApJ, 660, 541
Ade, P. A. R., Aghanim, N., Armitage-Caplan, C., et al. 2014, A&A, 571, 16 Ho, L. C., & Kim, M. 2014, ApJ, 789, 17
Akritas, M. G., & Bershady, M. A. 1996, ApJ, 470, 706 Hönig, S. F., Watson, D., Kishimoto, M., & Hjorth, J. 2014, Natur, 515, 528
Bahcall, J., Kozlovsky, B.-Z., & Salpeter, E. E. 1972, ApJ, 171, 467 Horne, K., Korista, K. T., & Goad, M. G. 2003, MNRAS, 339, 367
Bardeen, J. M., Press, W. H., & Teukolsky, S. A. 1972, ApJ, 178, 347 Hu, C., Du, P., Lu, K.-X., et al. (SEAMBH collaboration) 2015, ApJ, in press
Barth, A. J., Bennert, V. N., Canalizo, G., et al. 2015, ApJS, 217, 26 (arXiv:1503.03611)
Barth, A. J., Pancoast, A., Bennert, V. N., et al. 2013, ApJ, 769, 128 Hu, C., Wang, J.-M., Ho, L. C., et al. 2008a, ApJ, 687, 78
Barvainis, R. 1987, ApJ, 320, 537 Hu, C., Wang, J.-M., Ho, L. C., et al. 2008b, ApJL, 683, L115
Bentz, M.C. 2011, in Narrow-Line Seyfert 1 Galaxies and Their Place in the Jiang, Y.-F., Stone, J. M., & Davis, S. W. 2014, ApJ, 796, 106
Universe, ed. L. Foschini, M. Colpi, L. Gallo, et al. (Milan: Civic Kaspi, S., Brandt, W.N., Maoz, D., et al. 2007, ApJ, 659, 997
Acquarium Auditorium), 33, http://nls1.brera.inaf.it/ Kaspi, S., Maoz, D., Netzer, H., et al. 2005, ApJ, 629, 61

24
The Astrophysical Journal, 806:22 (25pp), 2015 June 10 Du et al.

Kaspi, S., Smith, P. S., Netzer, H., et al. 2000, ApJ, 533, 631 Peterson, B. M., Wanders, I., Bertram, R., et al. 1998, ApJ, 501, 82
Kawaguchi, T. 2013, arXiv:1306.0188 Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 1992,
Kawakatu, N., & Ohsuga, K. 2011, MNRAS, 417, 2562 Numerical Recipes in FORTRAN (2nd ed.; Cambridge: Cambridge Univ.
Kelly, B. C., Bechtold, J., & Siemiginowska, A. 2009, ApJ, 698, 895 Press)
Kelly, B. C., & Shen, Y. 2013, ApJ, 764, 45 Sadowski, A. 2009, ApJS, 183, 171
Klimek, E. S., Gaskell, C. M., & Hedrick, C. H. 2004, ApJ, 609, 69 Sadowski, A., Narayan, R., McKinney, J. C., & Tchekhovskoy, A. 2014,
King, A. L., Davis, T., Denney, K. D., & Vestergaard, M. 2014, MNRAS, MNRAS, 439, 503
441, 3454 Santos-Lleó, M., Chatzichristou, E., de Oliveira, C. M., et al. 1997, ApJS,
Koshida, S., Minezaki, T., Yoshii, Y., et al. 2014, ApJ, 788, 159 112, 271
La Franca, F., Bianchi, S., Ponti, G., et al. 2014, ApJL, 787, L12 Santos-Lleó, M., Clavel, J., Schulz, B., et al. 2001, A&A, 369, 57
Laor, A., & Netzer, H. 1989, MNRAS, 238, 897 Shakura, N. I., & Sunyaev, R. A. 1973, A&A, 24, 337
Li, Y.-R., Wang, J.-M., Ho, L. C., Du, P., & Bai, J.-M. 2013, ApJ, 779, 110 Shemmer, O., & Netzer, H. 2000, arXiv:astro-ph/0005163
Li, Y.-R., Wang, J.-M., Hu, C., Du, P., & Bai, J.-M. 2014, ApJL, 786, L6 Shen, Y., Brandt, W. N., Dawson, K. S., et al. 2014, ApJS, 216, 4
Li, G.-X., Yuan, Y.-F., & Cao, X. 2010, ApJ, 715, 623 Shen, Y., Brandt, W. N., Dawson, K. S., et al. 2015, ApJS, 216, 4
Maoz, D., & Netzer, H. 1989, MNRAS, 236, 21 Shen, Y., Richards, G. T., Strauss, M. A., et al. 2011, ApJS, 194, 45
Maoz, D., Netzer, H., Mazeh, T., et al. 1991, ApJ, 367, 493 Stirpe, G. M., Winge, C., Altieri, B., et al. 1994, ApJ, 425, 609
Marziani, P., & Sulentic, J. 2014a, AdSpR, 54, 1331 Suganuma, M., Yoshii, Y., Kobayashi, Y., et al. 2006, ApJ, 639, 46
Marziani, P., & Sulentic, J. 2014b, MNRAS, 442, 1211 Szuszkiewicz, E., Malkan, M. A., & Abramowicz, M. A. 1996, ApJ, 458, 474
McLure, R. J., & Dunlop, J. S. 2004, MNRAS, 352, 1390 Trakhtenbrot, B., & Netzer, H. 2012, MNRAS, 427, 3081
Mineshige, S., Kawaguchi, T., Takeuchim, M., & Hayashida, K. 2000, PASJ, Vestergaard, M., & Osmer, P. S. 2009, ApJ, 699, 800
52, 499 Vestergaard, M., & Peterson, B. M. 2006, ApJ, 641, 689
Mor, R., & Netzer, H. 2012, MNRAS, 420, 526 Wanders, I., van Groningen, E., Alloin, D., et al. 1993, A&A, 269, 39
Netzer, H. 1987, MNRAS, 225, 55 Tremaine, S., Gebhardt, K., Bender, R., et al. 2002, ApJ, 574, 740
Netzer, H., Heller, A., Loinger, F., et al. 1996, MNRAS, 279, 429 Wang, J.-M., Du, P., Hu, C., et al. (SEAMBH collaboration) 2014a, ApJ,
Netzer, H., & Laor, A. 1993, ApJL, 404, L51 793, 108
Netzer, H., Shemmer, O., Maiolino, R., et al. 2004, ApJ, 614, 558 Wang, J.-M., Du, P., Valls-Gabaud, D., Hu, C., & Netzer, H. 2013, PhRvL,
Netzer, H., & Trakhtenbrot, B. 2014, MNRAS, 438, 672 110, 081301
Nobuta, K., Akiyama, M., Ueda, Y., et al. 2012, ApJ, 761, 143 Wang, J.-M., Qiu, J., Du, P., & Ho, L. C. 2014b, ApJ, 797, 65
O’Brien, P. T., Goad, M. R., & Gondhalekar, P. M. 1994, MNRAS, 268, 845 Wang, J.-M., Szuszkiewicz, E., Lu, F.-J., & Zhou, Y.-Y. 1999, ApJ, 522, 839
Ohsuga, K., Mineshige, S., Mori, M., & Umemura, M. 2002, ApJ, 574, 315 Wang, J.-M., & Zhou, Y.-Y. 1999, ApJ, 516, 420
Onken, C. A., Ferrarese, L., Merritt, D., et al. 2004, ApJ, 615, 645 Wang, T.-G., & Zhang, X.-G. 2003, MNRAS, 340, 793
Osterbrock, D. E., & Pogge, R. W. 1987, ApJ, 323, 108 Watarai, K.-Y., Mizuno, T., & Mineshige, S. 2001, ApJL, 549, L77
Pancoast, A., Brewer, B., & Treu, T. 2011, ApJ, 730, 139 Watson, D., Denney, K. D., Vestergaard, M., & Davis, T. M. 2011, ApJ,
Pei, L., Barth, A. J., Aldering, G. S., et al. 2014, ApJ, 795, 38 740, 49
Peterson, B. M. 1993, PASP, 105, 247 Woo, J.-H., Yoon, Y., Park, S., Park, D., & Kim, S. C. 2015, ApJ, 801, 38
Peterson, B. M., Ali, B., Horne, K., et al. 1993, ApJ, 402, 469 Wu, X.-B., Wang, R., Kong, M.-Z., Liu, F.-K., & Han, J.-L. 2004, A&A,
Peterson, B. M., Balonek, T. J., Barker, E. S., et al. 1991, ApJ, 368, 119 424, 793
Peterson, B. M., Berlind, P., Bertram, R., et al. 2002, ApJ, 581, 197 Yoshii, Y., Kobayashi, Y., Minezaki, T., Koshida, S., & Peterson, B. A. 2014,
Peterson, B. M., Ferrarese, L., Gilbert, K. M., et al. 2004, ApJ, 613, 682 ApJL, 784, L11
Peterson, B. M., Grier, C. J., Horne, K., et al. 2014, ApJ, 795, 149 Zhou, H.-Y., Wang, T.-G., Yuan, W., et al. 2006, ApJS, 166, 128

25