SUPERMASSIVE BLACK HOLES WITH HIGH ACCRETION RATES IN ACTIVE GALACTIC NUCLEI. I. FIRST RESULTS FROM A NEW REVERBERATION MAPPING CAMPAIGN

The most important criterion for the selection of targets is based on their normalized accretion rate as estimated by the (less accurate) single-epoch method. In particular, we search for objects with $\dot{m}\ge 1$. As explained, the suggestion is that in such cases, the radiative efficiency η is no longer a simple function of the BH spin and advection becomes an important process that affects the source luminosity. We decided to avoid radio loud AGNs although, in a few cases, a radio loud source was discovered after the project had begun. Such cases will be discussed separately.

Selecting AGNs by their high accretion rates depends on their L_Bol which in itself depends on the SED. Unfortunately, in standard ADs, much of the radiation is emitted beyond the Lyman limit at 912 Å (Laor & Netzer 1989; Elvis et al. 1994, 2012; Marconi et al. 2004; Richards et al. 2006; Davis & Laor 2011). According to the standard SS model, the optical bolometric correction factor, κ_B, i.e., the factor converting λL_λ measured at optical wavelengths to L_Bol, depends on the accretion rate and BH mass such that $\kappa _{\rm B}\propto (\dot{m}/M_{\bullet })^{1/3}$. For ADs around 10⁸–10⁹ M_☉ BHs, this agrees with global empirical estimates that suggest κ_B ∼ 5–10 (e.g., Marconi et al. 2004). However, much larger and much smaller values of κ_B are predicted by thin AD models for smaller and larger BHs, respectively (see specific examples in Netzer & Trakhtenbrot 2013). Since most of the ∼50 measured BH masses in the RM campaign have BH masses that fall below 10⁸ M_☉, we suspect that $\dot{m}$ in at least several of these sources has been underestimated in the past. Thus, at least some of these sources can be SEAMBHs.

Alternatively, we can select targets based on the slope of their 2–10 keV X-ray continuum. As argued in Wang et al. (2013), some SEAMBH models predict that such objects would show steeper X-ray slopes compared with low-$\dot{m}$ sources. While a possible $\dot{m}\hbox{--}\Gamma _{2\hbox{--}10}$ relation has been suggested by Wang et al. (2013), this depends on the poorly understood disk–corona structure. The underlying physics is based on the link between the accretion rate in the cold gas phase, the increased emission from the UV and optical bands, and the enhancement of Comptonization cooling, leading to a suppressed hot corona emission. A correlation between $\dot{m}$ and Γ_2–10 has indeed been reported in both low and high luminosity AGNs (see, e.g., Wang et al. 2004; Risaliti et al. 2009; Zhou & Zhang 2010; Shemmer et al. 2006; Brightman et al. 2013). A sample of 60 SEAMBH candidates selected from the literature was presented by Wang et al. (2013). The $\dot{m}\hbox{--}\Gamma _{2\hbox{--}10}$ correlation in this sample is strong but with a large scatter. Much of the scatter can be the result of the uncertain BH masses and bolometric luminosities of these sources.

Our target selection follows several criteria that remove the dependence of the disk structure on the uncertain L_Bol.

• 1.  
  The sources are NLS1s suspected to be SEAMBHs. Their L_Bol and M_• are based on the single-epoch mass determination method and satisfy $\dot{m}>1$. Their Eddington rates are based on the SS thin disk model and hence⁷

  $$\begin{equation} \dot{m}_{_{\rm SS}} \simeq 20.1\left(\frac{L_{44}}{\cos i}\right)^{3/2} M_7^{-2}\eta _{_{\rm SS}}, \end{equation} \tag{ 3 }$$

  where L₄₄ = λL_λ/10⁴⁴ erg s⁻¹ at λ = 5100 Å, M₇ = M_•/10⁷ M_☉, and i is the inclination angle. We require that $\dot{m}_{_{\rm SS}}>1$ for cos i = 0.75 (an inclination typical of type-I AGNs that cannot be observed from a much larger inclination angle due to obscuration by the central torus). This expression contains the unknown value of η due to the unknown BH spin (a). However, a lower limit on $\dot{m}_{_{\rm SS}}$ can be obtained by assuming the lowest possible efficiency of thin ADs, $\eta _{_{\rm SS}}=0.038$, corresponding to a spin parameter of a = −1. We should point out, again, that $\dot{m}_{_{\rm SS}}$ given by Equation (3) is a lower limit for the Eddington ratios $\dot{m}$ if BH accretion proceeds via a slim AD.
• 2.  
  The targeted signal-to-noise ratio is high enough to enable two-dimensional (2D) velocity-resolved RM in the Hβ and Mg ii λ2798 lines.
• 3.  
  The 2–10 keV photon index is very steep, Γ_2–10 > 2.
• 4.  
  Targets can be followed from the ground for a period of at least 80–180 days for low and high redshift AGNs, respectively.
• 5.  
  Objects with proven continuum variations. We note that there is some evidence that the variability amplitudes of AGNs are anti-correlated with Eddington ratios (e.g., Ai et al. 2013; Zuo et al. 2012). In particular, some NLS1s show small optical continuum variations (Giannuzzo & Stirpe 1996; Giannuzzo et al. 1998), but some are very small (Shemmer & Netzer 2000; Klimek et al. 2004; Yip et al. 2009). However, significant variations are found in many SDSS NLS1s located in the Strip 82 region (Ai et al. 2013). We chose sources from the Catalina database⁸ and excluded candidates showing variations of less than 10% in the Catalina light curves.

2.2.1. The Shangri-La Telescope and Spectrograph

2.2.2. Spectrophotometry

2.2.3. Photometric Observations

2.3.1. Light Curve Measurements

2.3.2. Variability Characteristics

To characterize the continuum variability of the sources we use the variability characteristic F_var defined by Rodríguez-Pascual et al. (1997),

$$\begin{equation} F_{\rm var}=\frac{(\sigma ^2-\Delta ^2)^{1/2}}{\langle F\rangle } \,, \end{equation} \tag{ 4 }$$

where $\langle F\rangle =N^{-1}\sum _i^NF_i$ is the averaged flux, N is the number of observations, and

$$\begin{equation} \sigma ^2=\frac{1}{N-1}\sum _i^N\left(F_i-\langle F\rangle \right)^2;\quad \Delta ^2=\frac{1}{N}\sum _i^N\Delta _i^2. \end{equation} \tag{ 5 }$$

Here Δ_i represents the uncertainty on the flux F_i. Below we apply Equation (4) to the V-band, F_Hβ and F₅₁₀₀ light curves.

2.3.3. Cross Correlation Analysis

In order to obtain the Eddington ratios (Equation (3)), we have to subtract the contaminations of host galaxies. Following the scheme described in Bentz et al. (2009a), we estimated the host galaxy contribution to the observed AGN flux at 5100 Å using archival Hubble Space Telescope (HST) images that are available for all our targets. In case of more than one HST observation, we chose those images with the longest exposure times in the optical band.

The procedure used here is quite standard and its main features are summarized here for clarity. We retrieve from the archive the calibrated data that were reduced by the latest version of the pipeline, and the best reference files. All of the exposures for a single object are combined with AstroDrizzle version 1.0.2 task in the DrizzlePac package to clean cosmic rays and make one distortion-corrected image of the host galaxy.⁹ Only the pixels flagged as good in the data quality frames are included in the combined images. Some exposures include saturated pixels. These were removed prior to the final combination of all data. Thus, all pixels marked with zeros in weighting Astrodrizzle output (i.e., no good pixel from any exposure) are masked out in the following fitting processes. Examples of the final combined distortion-free images are shown in the left panels of Figure 1 with spectroscopic monitoring apertures overlaid. The diagram confirms that galaxy contribution to the observed fluxes at 5100 Å is non-negligible in all three cases.

Zoom In Zoom Out Reset image size

We use GALFIT version 3.0 (Peng et al. 2002, 2010) to model the surface brightness distribution of the three host galaxies. The GALFIT algorithm fits 2D analytic functions to galaxies and point sources. Each of the objects in this study is fitted with a point-spread function (PSF) to model the AGN, one to several Sérsic profiles to model the host galaxy and a constant to model the sky background. PSFs were generated for each AGN by creating a distorted PSF model at its detector position in each of the exposed frames using TinyTim package (Krist 1993) and combining these models by Astrodrizzle using the same configuration that was used to combine the AGN images. Sérsic profiles are employed to model the various host galaxy components such as bulges, disks and bars. Thus the surface brightness is expressed as

$$\begin{equation} \Sigma (R) = \Sigma _e \exp \left\lbrace -\kappa _0 \left[ \left(\frac{R}{R_e} \right)^{1/n} - 1 \right] \right\rbrace, \end{equation} \tag{ 6 }$$

where Σ_e is the surface brightness at the effective radius R_e, n is the Sérsic power-law index, and κ₀ is coupled to n to make sure that half of the flux is within R_e (see details in Peng et al. 2002, 2010; n = 1 represents a galactic disk and n = 4 a bulge). We set no constraint on n when fitting bulge, bar or elliptical host, but fix n = 1 to model the disk component.

At least one Sérsic component was required to model the galaxy in addition to the PSF and sky components, and more were added if needed. For some cases where the PSF modeled by TinyTim does not match the nuclear surface brightness distribution well (see Krist 2003; Kim et al. 2008), an additional Sérsic component with small effective radius was used to modify the mismatch. When the FOV of the HST instrument (e.g., High Resolution Channel (HRC) of the Advanced Camera for Surveys (ACS)) is small, the area in the edge of the image used to constrain the fitted sky value was limited. We carefully adjusted the initial parameters to make sure the residuals and χ² were minimized. Other bright objects in the FOV of our targets were masked out.

Having completed the host galaxy modeling, we subtracted the PSF and sky components from the images of each object and extracted from the PSF-sky-free images the total counts due to the host galaxy inside the slit of the spectrograph. These counts were transformed to flux density units. Because of the difference between the pivot wavelength of the HST filter and the rest-frame 5100 Å, we used a bulge template spectrum (Kinney et al. 1996) to determine the required color correction. The template is redshifted and reddened by Galactic extinction based on the dust extinction map of Schlafly & Finkbeiner (2011). The synphot package is then employed to convolve the HST passband with the template to simulate the photometry.

Normalizing the counts to the observed total counts in the aperture provides the host galaxy flux at 5100 Å. We adopted a nominal 10% uncertainty on the host galaxy contributions due to the modeling procedures (see also analysis in Bentz et al. 2013).

In this first paper of the series we report the observations of three objects: Mrk 142, Mrk 335, and IRAS F12397+3333. Information on the three targets and comparison stars is given in Table 1 as well as the variability parameters (F_var) for the continuum and the Hβ emission line. Table 2 gives the data for the continuum and Hβ light curves. Mrk 142 was previously mapped by the LAMP collaboration (Bentz et al. 2009b). Mrk 335 has been monitored in two earlier campaigns (Kassebaum et al. 1997; Grier et al. 2012) and IRAS F12397+3333 is reported here for the first time. Table 3 and Figure 1 present data and images related to the host galaxy treatment in the three sources (see captions for details). The HST images of Mrk 142 and Mrk 335 have been analyzed by Bentz et al. (2009a, 2013) and the one for IRAS F12397+3333 has been treated by Mathur et al. (2012) who followed the same scheme described in Bentz et al. (2009a). Our results are consistent with both these earlier studies.

Table 1. Basic Data and Variability Amplitude

Object	α2000	δ2000	Redshift	E(B−V)	Monitoring Period	Nspec	Variability Amplitude (%)a			Comparison Stars
							Fvar(5100 Å)	Fvar(V)	Fvar(Hβ)	R*	P.A.
Mrk 335	00 06 19.5	+20 12 10	0.0258	0.030	2012 Oct–2013 Feb	91	5.2 ± 0.5	3.9 ± 0.3	3.4 ± 0.3	807	1745
Mrk 142	10 25 31.3	+51 40 35	0.0449	0.015	2012 Nov–2013 Apr	119	8.1 ± 0.6	5.5 ± 0.4	7.8 ± 0.5	1131	1552
IRAS F12397+3333	12 42 10.6	+33 17 03	0.0435	0.017	2013 Jan–2013 May	51	5.6 ± 0.6	3.2 ± 0.4	4.5 ± 0.6	1890	1300

Notes. E(B − V) is the Galactic extinction using the maps in Schlafly & Finkbeiner (2011). N_spec is the number of spectroscopic observing epochs, R_* is the angular distance to the target, and P.A. is the position angle. ^a Amplitudes were calculated using Equation (4). The uncertainties of F_var is calculated according to Edelson et al. (2002).

Download table as:  ASCIITypeset image

Table 3. Host Galaxy Decomposition

Object	Data Set	Observational Setup	$m_{\rm st}^{\rm a}$	Re	n	b/a	P.A.	Note	$\chi _{\nu }^2$
				('')			(deg)
Mrk 335	J9MU010	ACS, HRC, F550M	14.73					PSF	1.700
			17.78	0.09	0.89	0.04	44.45	Add'l PSF
			14.81	5.02	3.58	0.93	95.81	Elliptical
			−0.007					Sky
Mrk 142	IB5F010	WFC3, UVIS1, F547M	16.13					PSF	1.424
			18.27	0.41	0.45	0.78	14.25	Bulge
			16.34	4.73	[1.0]	0.56	39.53	Disk
			0.014					Sky
IRAS F12397	J96I090	ACS, HRC, F625W	16.27					PSF	1.630
			18.36	0.16	1.16	0.72	42.65	Nuclear spiral?
			17.40	0.79	0.75	0.96	146.4	Bulge
			16.51	3.69	[1.0]	0.49	53.84	Disk
			0.007					Sky

Notes. The values in square brackets are fixed in the fitting procedure. ^a m_st is the ST magnitude, an f_λ-based magnitude system, m_ST = −2.5log₁₀(f_λ) − 21.10, for f_λ in erg s⁻¹ cm⁻² Å⁻¹ (see Sirianni et al. 2005). The units of sky are electrons s⁻¹.

Download table as:  ASCIITypeset image

Figures 2–4 show mean observed spectra, light curves and CCFs for the three sources. As described below, we found that moon phases and the angular separation between the Moon and the target are the main factors affecting the V-band light curve. We demonstrate this in Figure 2 for one of the sources (Mrk 335) and no longer use these light curves since they are inferior to the F₅₁₀₀ light curves in all three sources. The variability characteristics, F_var, calculated from the light curves are given in Table 1. The averaged variability is typically ∼5% over the monitoring campaign. The largest peaks or dips are significantly larger than this value. For Mrk 335, Mrk 142 and IRAS F12397+3333, they show F₅₁₀₀ and Hβ variations of (F₅₁₀₀, Hβ) ≈ (15%, 10%), (25%, 20%), (15%, 10%), respectively. All values are significantly larger than the uncertainties associated with our line and continuum measurements.

Zoom In Zoom Out Reset image size

The lags computed from the CCFs of the Hβ emission line versus the F₅₁₀₀ light curves are tabulated in Table 4 and discussed in Section 3.2. We prefer the use of F₅₁₀₀ over the V-band since the latter may be influenced by the strong emission lines of Hβ and Fe ii  that can slightly affect the lag determination. The monochromatic luminosities, L₅₁₀₀, calculated after allowing for Galactic extinction based on Schlafly & Finkbeiner (2011), are listed in Table 5. We used the centroid time lag (τ_cent) to indicate our best value of $R_{_{\rm BLR}}$ that enters the calculation of the BH mass.

Table 4. Results of CCF Analysis in Rest Frame of Sources

	F5100 vs. Hβ
Object	τcent	τpeak	rmax
	(days)	(days)
Mrk 335	$10.6_{-2.9}^{+1.7}$	$8.2_{-1.1}^{+4.4}$	0.67
Mrk 142	$6.4_{-2.2}^{+0.8}$	$5.3_{-2.2}^{+2.6}$	0.68
IRAS F12397+3333	$11.4_{-1.9}^{+2.9}$	$11.9_{-2.2}^{+1.4}$	0.71

Download table as:  ASCIITypeset image

Table 5. Luminosity, BH Mass, and Eddington Ratios

Object	FWHMa	$F_{\rm gal}^{\rm b}$	$F_{\rm AGN}^{\rm b}$	L5100	$M_{\bullet }^{\rm c}$	$\dot{m}_{_{\rm SS}}$
	(km s−1)	(10−15 erg s−1 cm−2 Å−1)		(1043 erg s−1)	(106 M☉)	$\eta _{_{\rm SS}}=0.1$	0.038
Mrk 335	1997 ± 265	0.99 ± 0.10	5.20 ± 0.37	4.9 ± 0.4	$8.3_{-3.2}^{+2.6}$	1.6	0.6
Mrk 142	1647 ± 69	0.37 ± 0.04	1.27 ± 0.15	3.7 ± 0.4	$3.4_{-1.2}^{+0.5}$	5.9	2.3
IRAS F12397(I)	1835 ± 473	0.65 ± 0.06	1.44 ± 0.14	16.9 ± 1.7	$7.5_{-4.1}^{+4.3}$	12.1	4.6
IRAS F12397(II)	1835 ± 473	0.65 ± 0.06	1.44 ± 0.14	3.9 ± 0.4	$7.5_{-4.1}^{+4.3}$	1.3	0.5

Notes. IRAS F12397(I)/(II) mean with/without intrinsic reddening, respectively. ^a FWHM stands for V_FWHM in the text of the paper and is measured from the mean spectra. ^b F_gal and F_AGN are host galaxy and mean AGN fluxes at (1 + z) 5100 Å. F_gal + F_AGN = F_obs, where F_obs is the observed flux. ^c M_• are calculated using the centroid time lag, V_FWHM and f_BLR = 1.0. Uncertainties on M_• are the results of the errors on the time lags and the FWHMs. L₅₁₀₀ is the mean AGN luminosity at rest-frame wavelength of 5100 Å after the subtraction of the host galaxy contribution and allowing for Galactic extinction.

Download table as:  ASCIITypeset image

Our BH mass measurements are based on the FWHM of the broad Hβ lines as measured from the mean spectra. The subtraction of the narrow Hβ component is rather uncertain because of the very smooth line profiles and the limited spectral resolution of our observations (about 500 km s⁻¹ based on a comparison with the SDSS spectra). Similar to the approach used in Hu et al. (2012), we fitted the entire Hβ profile with a narrow Gaussian component (F_N) and a broad Gaussian–Hermite component (F_B). For a typical AGN with luminosity similar to the ones measured in our sample, $F_{\rm N}/F_{\rm [{O}\,{\scriptsize{III}}]} \simeq 0.1$, where $F_{\rm [{O}\,{\scriptsize{III}}]}$ is the flux of the narrow [O iii] λ5007 line. We used this ratio, and the shift and FWHM of the [O iii] λ5007 line, to define F_N. This was subtracted from the total profile to obtain the measured FWHM(broad Hβ). The uncertainty is obtained by following the same procedure assuming (1) F_N = 0 and (2) $F_{\rm N}=0.2F_{\rm [{O}\,{\scriptsize{III}}]}$. Finally we obtained the intrinsic FWHM(broad Hβ) by allowing for the instrumental broadening. The FWHMs and their uncertainties are listed in Table 5.

Regarding the preferred velocity measure, there are two options: $V_{_{\rm FWHM}}$ and σ_line (see Section 1 for definitions and references). The mean values for both are obtained through a comparison with the M_•–σ_* relationship in non-active galaxies. The most recent results based on 25 AGNs with reliably measured σ_* are those of Woo et al. (2013) who obtained $f_{_{\rm BLR}}(\sigma _{\rm line})=5.1\hbox{--}5.9$, depending on the exact method used. The uncertainty on these numbers are about ±1.5. Netzer & Marziani (2010) used the $V_{_{\rm FWHM}}$-based method and the somewhat inferior samples of Onken et al. (2004) and of Woo et al. (2010) to obtain $f_{_{\rm BLR}}(V_{_{\rm FWHM}})=1.0$. This is confirmed by the more recent results of J.-H. Woo (2013, private communication) that provides $f_{_{\rm BLR}}(V_{_{\rm FWHM}})$ for the above sample of 25 AGNs. The result of this work is $f_{_{\rm BLR}}(V_{_{\rm FWHM}})=0.98^{+0.28}_{-0.22}$, i.e., similar to the scatter of $f_{_{\rm BLR}}(\sigma _{\rm line}$). Since for a Gaussian line profile $V_{_{\rm FWHM}}\simeq 2.35 \times \sigma _{\rm line}$, the two methods are basically identical given the uncertainties.

The rms spectra over the region of the Hβ line are shown in Figure 5 where they are compared with the same parts of the mean spectra. The quality of the rms spectra depends on the seeing and the position of the target in the slit. The spectra tend to be noisy in particular in cases of small variability amplitude. Strong residuals may be present depending on the exact positioning of the slit and the accuracy of the wavelength calibration. In the cases studied here, this is most noticeable for the [O iii] λ5007 line, the strongest narrow feature in the spectrum, where the residual noise, at a level of about 3% of the line intensity, is clearly visible. This phenomenon is well known from earlier RM campaigns (e.g., Figures 1–3 in Peterson et al. 1998; Kaspi et al. 2000; Park et al. 2012). We have also checked the measured fluxes of the [O iii] λ5007 lines in all spectra and found that they fluctuated in a random way, uncorrelated with the continuum variations, by 3%, 4% and 4% for Mrk 335, Mrk 142, and IRAS F12397, respectively.

We measured the FWHM (Hβ) from the rms spectra after smoothing the profiles with a nine pixels boxcar filter. The uncertainty on the width was estimated by repeating the procedure with a 3 pixel boxcar filter that resulted in a narrower profile. These measurements resulted in FWHMs of 1418 ± 118, 1623 ± 110, and 1510 ± 194 km s⁻¹, for Mrk 335, Mrk 142, and IRAS F12397+3333, respectively. The uncertainties derived in this way are only due to the noisy rms profiles. The comparison with the FWHMs measured from the mean spectra suggests a somewhat narrower rms profile for Mrk 335, and similar widths, within the uncertainties, for the other two objects.

In this work we adopt the combination of the $V_{_{\rm FWHM}}$ method and the mean spectrum for several reasons: first, the measurement of $V_{_{\rm FWHM}}$, given our spectral resolution, is less uncertain than σ_line. Second, NLS1s tend to have Lorentzian-shaped line profiles that can result in a considerable increase of σ_line which is very sensitive to the accuracy of the very extended wings measurement. Finally, in many cases the combination of the rms spectrum and σ_line gives similar BH mass to the combination of $V_{_{\rm FWHM}}$ and the mean spectrum (Kaspi et al. 2000). This is not surprising given that the scaling factor $f_{_{\rm BLR}}$ is an average over a large number of BLRs, with different radial gas distributions and inclinations to the lines of sight, some described better by $V_{_{\rm FWHM}}$ and some by σ_line. Regardless of the exact method, the only secure way to obtain the correct mass and its normalization is by calibrating the entire sample against the M_•–σ_* relationship.

The results of the three BH mass measurements are listed in Table 5 with their associated uncertainties. We also list lower limits on Eddington ratios that were computed by adopting the most conservative estimate on the radiation efficiency (see discussion below). All three sources are super-Eddington accretors.

3.2.1. Mrk 335

Mrk 335 is a well-known NLS1. It has been monitored for 6 yr by Kassebaum et al. (1997) and re-analyzed by Zu et al. (2011). The X-ray observations show large variations at soft and hard X-ray energies (Arévalo et al. 2008). The optical monitoring shows variability at a level of 10% (Peterson et al. 1998). Figure 2 shows the line and continuum light curves obtained by us with variability amplitudes of ∼15% in both. The measured lag, $10.6^{+1.7}_{-2.9}$ days, is similar to the previous measurements that range from 12.5 to 16.8 days (see Table 6 for details and references). This lag is still in reasonable agreement with the Bentz et al. (2013) expression for the $R_{_{\rm BLR}}\hbox{--}L$ relationship. The virial mass of the BH measured by us is 8.3 × 10⁶ M_☉ and the Eddington ratios $\dot{m}_{_{\rm SS}}\approx 1.6$, large enough to indicate a SEAMBH. The more conservative approach assuming $\eta _{_{\rm SS}}=0.038$, gives $\dot{m}_{_{\rm SS}}\approx 0.6$ which is also beyond the SS regime of thin ADs (Laor & Netzer 1989).

Comparison with previous RM experiments for this source, listed in Table 6, suggest that the present response of the BLR is shorter than in all previous campaigns although the deviation is still consistent with the estimated uncertainties. We suggest that the size of the BLR in this source changed significantly, over a period of about 10 yr since the earlier RM measurements. The corresponding change in the 5100 Å luminosity is smaller. This interesting possibility requires more data to instigate in detail.

Table 6. Mrk 335: BLR Size and Continuum Luminosity at Different Epochs

Epoch	τcent	τpeak	FAGN[5100 Å(1 + z)]	L5100	Reference
(JD2,400,000+)	(days)	(days)	(10−15 erg s−1 cm−2 Å−1)	(1043 erg s−1)
49156−49338	$16.8^{+4.8}_{-4.2}$	$18^{+5}_{-6}$	6.12 ± 0.19	5.8 ± 0.2	P04, B13
49889−50118	$12.5^{+6.6}_{-5.5}$	$13^{+9}_{-7}$	7.25 ± 0.21	6.8 ± 0.2	P04, B13
55440−55568	$14.3^{+0.7}_{-0.7}$	$14.0^{+0.9}_{-0.9}$	5.84 ± 0.29	5.5 ± 0.3	G12, B13
56222−56328	$10.6^{+1.7}_{-2.9}$	$8.2^{+4.4}_{-1.1}$	5.20 ± 0.37	4.9 ± 0.4	This paper

Notes. P04: Peterson et al. (2004); G12: Grier et al. (2012); B13: Bentz et al. (2013).

Download table as:  ASCIITypeset image

3.2.2. Mrk 142

This object was mapped by the LAMP project (Bentz et al. 2009b) who reported a time lag of τ = 2.7 ± 0.8 days measured from the peak correlation coefficient (r_max ≈ 0.5). The significance of this measurement was the lowest among all sources monitored in that project. The reason was the lack of a clear peak in the Hβ light curve.

Figure 3 shows the light curves of Mrk 142 obtained by the SLT. Our observations are superior to the measurements of Bentz et al. (2009b) for several reasons: the sampling is more homogeneous, the error bars are smaller, and the light curves exhibit two clear minima and maxima. The CCF shows a very sharp peak around τ_cent ≈ 6.4 days with a correlation coefficient of r_max ≈ 0.7. Our measured lag is larger than the previously measured value by more than a factor two which brings the source closer to the $R_{_{\rm BLR}}\hbox{--}L$ relationship of Bentz et al. (2013). However, the object is still a clear outlier of the relationship.

Zoom In Zoom Out Reset image size

According to Equation (3) and Table 4, the value of $\dot{m}_{_{\rm SS}}$ derived for Mrk 142 is at least 2.3 ($\eta _{_{\rm SS}}=0.038$) and more likely around 5.9 ($\eta _{_{\rm SS}}=0.1$). This is at the high end of the $\dot{m}_{_{\rm SS}}$ distribution of all AGNs with reliable RM-based BH mass measurements. As alluded to earlier, such accretion rates are inconsistent with the thin disk idea and we expect that the real accretion rate, $\dot{m}$, is even larger making Mrk 142 a clear case of a SEAMBH.

3.2.3. IRAS F12397+3333

IRAS F12397+3333 is an infrared luminous source identified as an AGN by Keel et al. (1988). The optical SED of the source is "red" compared with most AGNs, including those in the RM sample. This is typical of other IRAS sources. The hypothesis of significant dust attenuation is supported by polarization measurements (Grupe et al. 1998) and by the weak signal obtained by Galaxy Evolution Explorer (GALEX) at λ = 1528 Å (Atlee & Mathur 2009) which, when combined with our own measurements, gives L₅₁₀₀/L₁₅₂₈ ≃ 4–8, far above what is observed in unreddened AGNs, like those in the SDSS sample (note that these are not contemporaneous observations).

To further check this point, we fitted the SDSS spectrum of the source and decomposed the Hα and Hβ lines into narrow and broad components.¹⁰ We found FWHMs of 380 and 1680 km s⁻¹, for the narrow and broad components, respectively. This results in (Hα/Hβ)_broad ≈ 5.71 and (Hα/Hβ)_narrow ≈ 6.14. This steep Balmer decrement suggests that reddening significantly affects the narrow emission lines and possibly also the broad emission lines. Finally, the X-ray spectrum of the source shows strong and broad Kα line emission at around 6.4 keV (Bianchi et al. 2009; Zhou & Zhang 2010) but the hydrogen column density is low, N_H = (2.74 ± 1.38) × 10²⁰ cm⁻² (Grupe et al. 2001). Assuming galactic dust-to-gas ratio, this corresponds to A_v ≃ 0.15.

Given the likely continuum extinction, and the conflicting results obtained from the X-ray column density and the GALEX results, we considered two different possibilities regarding the reddening. The first is no attenuation but extreme value of Hα/Hβ in the BLR and the second is dust extinction with an intrinsic Balmer decrement of (Hα/Hβ)_broad = 3.5, a ratio typical of a large number of "blue" AGNs with no suspected continuum attenuation (e.g., Vanden Berk et al. 2001). Balmer decrements in this range are fairly typical and are consistent with the conditions expected in the BLR (Netzer 2013).

Assuming first no reddening, we get L₅₁₀₀ = 3.9 × 10⁴³ erg s⁻¹. Using Equation (3) and our measured BH mass (7.5 × 10⁶ M_☉) we get $\dot{m}_{_{\rm SS}}\simeq 1.3$. In the second case, L₅₁₀₀ = 1.69 × 10⁴⁴ erg s⁻¹ which results in $\dot{m}_{_{\rm SS}}\simeq 12.1$. The corresponding values for $\eta _{_{\rm SS}}=0.038$ are 0.5 and 4.6. All values are above or close the limiting case of $\dot{m}_{_{\rm SS}}=1$ and the ones corrected for reddening are the largest among all cases published so far in AGNs with directly measured BH mass. Comparing Eddington ratio estimations in other samples of NLS1s with estimates based on the virial method, e.g., the one shown in Figure 4 of Wang & Netzer (2003), we find that IRAS F12397+3333 and Mrk 142 have the highest Eddington ratios measured so far.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As argued in numerous publications (e.g., Kelly et al. 2011, and references therein), understanding AGN variability is key to the understanding of the accretion mechanism, including the AD properties (Liu et al. 2008). Thus, the results of our monitoring campaign shed new light on several aspects of NLS1s in general and SEAMBHs in particular. First, we detected significant line and continuum variations, ∼15%, in three fast accreting BHs. Such variations have not been detected in most other known cases of NLS1s. The variations can provide important constraints on the properties of slim ADs (Mineshige et al. 2000), a topic that we are currently investigating.

Second, all three objects are accreting close to or above the limit of $\dot{m}_{_{\rm SS}}\approx 1$ even under the most conservative estimate of $\eta _{_{\rm SS}}=0.038$. Two of the sources, Mrk 142 and IRAS F12397+3333, show the highest values of $\dot{m}_{_{\rm SS}}$ measured so far. They indicate physical conditions that are far outside the nominal conditions for thin SS type disks. Among reverberation-mapped AGNs, these are perhaps the best candidates for having slim ADs.

Third, the newly measured lag for Mrk 335 is smaller than the earlier measurements but still in reasonable agreement with the $R_{_{\rm BLR}}\hbox{--}L$ relation obtained by Bentz et al. (2013). This is not the case for Mrk 142, and for IRAS F12397+3333 assuming there is extinction along the line of sight to this source. Both these objects fall well below the above correlation. This leads to a suggestion that SEAMBHs may exhibit a somewhat different $R_{_{\rm BLR}}\hbox{--}L$ relationship. One possibility is that in such cases, the very different geometry of the slim disk results in a different radiation pattern. In particular, the face-on luminosity can be very large compared with the equatorial value leading to the possibility that equatorial BLR gas clouds are exposed to much weaker radiation than assumed in isotropic radiation fields, which affect their location. Such ideas have been around for some time (Netzer 1987, 2013; Korista et al. 1997) but have never been studied in relation to slim disks. In addition, the balance between gravity and radiation pressure force, which is proportional to $\dot{m}$, may affect BLR gas around slim disks in SEAMBHs more than in other AGNs. This can lead to marginally bound, or perhaps even unbound, cloud systems (see discussion of these ideas in Marconi et al. 2008 and Netzer & Marziani 2010). If correct, the BH mass determination in such sources may be less secure. On the other hand, 7 of the sources with measured σ_* studied by Woo et al. (2013) are NLS1s and their virial-based BH mass is in reasonable agreement with the mass obtained by the M_•–σ_* relationship. Unfortunately none of the three sources studied here is included in the Woo et al. (2013) sample.

Fourth, the normalized accretion rates obtained here, at least for two of the cases, are so large that they must be important to the general issue of BH growth and BH duty cycle. In particular, in the exponential growth scenario, the growth time of massive BHs is inversely proportional to $\dot{m}$. Typical continuous growth times for local active BHs, based on typical Eddington rates of such sources (∼0.1), are of order the Salpeter time, ∼4 × 10⁸ yr (e.g., Netzer & Trakhtenbrot 2007). For the objects considered here, the growth time could be an order of magnitude or much shorter (see Netzer & Trakhtenbrot 2013 for similar considerations regarding growth times of SDSS AGNs).

Finally, the direct measurements of SEAMBH masses will allow us to estimate the cosmological distance to such objects using the relationship between the BH mass and the saturated luminosity of such objects (Wang et al. 2013). This can provide a new way to measure cosmological distances up to very high redshifts. We are extending the monitoring project to larger numbers and to higher redshifts in order to provide more precise calibration of this method.