A Low-Mass Black Hole in the Nearby Seyfert Galaxy Ugc 06728

The Astrophysical Journal, 831:2 (10pp), 2016 November 1 doi:10.3847/0004-637X/831/1/2 © 2016. The American Astronomical Society. All rights reserved. A LOW-MASS BLACK HOLE IN THE NEARBY SEYFERT GALAXY UGC 06728 Misty C. Bentz, Merida Batiste, James Seals, Karen Garcia, Rachel Kuzio de Naray, Wesley Peters, Matthew D. Anderson, Jeremy Jones, Kathryn Lester, Camilo Machuca, J. Robert Parks, Crystal L. Pope, Mitchell Revalski, Caroline A. Roberts, Dicy Saylor, R. Andrew Sevrinsky, and Clay Turner Department of Physics and Astronomy, Georgia State University, Atlanta, GA 30303, USA; [email protected] Received 2016 June 30; revised 2016 August 10; accepted 2016 August 11; published 2016 October 20 ABSTRACT We present the results of a recent reverberation mapping campaign for UGC 06728, a nearby low-luminosity Seyfert 1 in a late-type galaxy. Nightly monitoring in the spring of 2015 allowed us to determine an Hβ time delay of t = 1.4  0.8 days. Combined with the width of the variable Hβ line profile, we determine a black hole mass of MBH = (7.1  4.0) ´ 10 5 M. We also constrain the bulge stellar velocity dispersion from higher-resolution longslit spectroscopy along the galaxy minor axis and find s = 51.6  4.9 km s−1. The measurements presented here are in good agreement with both the RBLR –L relationship and the MBH –s relationship for active galactic nuclei. Combined with a previously published spin measurement, our mass determination for UGC 06728 makes it the lowest-mass black hole that has been fully characterized, and thus an important object to help anchor the low-mass end of black hole evolutionary models. Key words: galaxies: active – galaxies: nuclei – galaxies: Seyfert spins requires high-X-ray luminosities that are only found in AGNs (e.g., Reynolds 2014 and references therein), so the study of active black holes is an important key to unraveling the growth and evolution of cosmic structure. Unfortunately, bright AGNs are relatively rare in the local universe, leading to a disconnect in our current understanding of nearby black holes compared to those observed at larger look-back times. In particular, we are lacking direct comparisons of black hole mass constraints through multiple independent techniques in the same galaxies. There are a handful of published comparisons of reverberation masses and gas dynamical masses (e.g., Hicks & Malkan 2008), including the low-mass Seyfert NGC 4395 (Peterson et al. 2005; den Brok et al. 2015). The agreement is generally quite good, though the number of galaxies studied is small. Stellar dynamics, on the other hand, is a good check against reverberation masses because it relies on modeling a noncollisional system, unlike gas dynamics, where the AGN may be expected to inject energy on resolvable spatial scales. However, only two such comparisons currently exist for black hole from reverberation mapping and stellar dynamical modeling: NGC 4151 (Bentz et al. 2006a; Onken et al. 2014) and NGC 3227 (Denney et al. 2009a; Davies et al. 2006). While the techniques give roughly consistent masses for these two examples, there are caveats and limitations to both reverberation mapping and dynamical modeling, and a larger comparison sample is needed to fully assess the consistency of the local and the cosmological black hole mass scales. We have therefore undertaken a program to identify and monitor local AGNs, where it might be possible to obtain both a reverberation and a stellar dynamical mass constraint. Both techniques are time-and resource-intensive, and there are very few broadlined AGNs within z  0.01, where the spatial resolution provided by 8–10m class telescopes would be likely to resolve the black hole’s gravitational influence on the nuclear stellar dynamics, but we hope to increase the sample of mass comparisons by a factor of a few. We currently have stellar dynamical modeling underway for two other local 1. INTRODUCTION Supermassive black holes are now believed to inhabit the nuclei of all massive galaxies. Furthermore, the active galactic nucleus, or AGN, phase is generally understood to be a shortterm event in the life of a typical black hole, triggered either by a merger event or secular processes in the host galaxy (seethe review of Heckman & Best 2014 and references therein). Tight scaling relationships between the observed properties of black holes and their host galaxies point to a symbiotic relationship between the two (e.g., Magorrian et al. 1998; Ferrarese & Merritt 2000; Gebhardt et al. 2000; Gültekin et al. 2009; Kormendy & Ho 2013; van den Bosch 2016), in which the growth of structure and the evolution of galaxies across cosmic time is fundamentally linked to supermassive black holes. Understanding this link requires an understanding of black hole demographics, not just in the local universebut also at higher redshift, where we can witness the growth of structure occurring. Black holes, as opposed to galaxies, are incredibly simple objects that can be fully characterized with only two fundamental measurements: mass and spin. In the Milky Way, years of astrometric monitoring of stars in the central ∼0.01 parsec have led to an extremely precise determination of the mass of our own supermassive black hole (Ghez et al. 2000, 2008; Genzel et al. 2000). Unfortunately, all other galaxies are too distant for this same technique to be employed, and different techniques must be used to understand the masses of a population of central black holes. For galaxies out to ∼100 Mpc, spatially resolved observations of the bulk motions of stars or nuclear gas disks can be combined with dynamical modeling to constrain the central black hole mass (seethe reviews of Ferrarese & Ford 2005and Kormendy & Ho 2013). Reverberation mapping (Blandford & McKee 1982; Peterson 1993), on the other hand, takes advantage of AGN flux variability to constrain black hole masses through timeresolved, rather than spatially resolved, observations, thus obviating any distance limitations. Furthermore, the most widely used technique to constrain supermassive black hole 1 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. AGNs, and we describe here the reverberation results for an additional local AGN in our sample, UGC 06728. 2. OBSERVATIONS UGC 06728 is a low-luminosity Seyfert 1 located at α=11:45:16.0, δ=+79:40:53, z=0.00652 in a late-type galaxy that is highly inclined to our line of sight. It was monitored nightly over the course of two months in the spring of 2015. Optical spectroscopy and photometry were obtained at Apache Point Observatory (APO) in New Mexico, with additional supporting photometry obtained at Hard Labor Creek Observatory in Georgia. We describe the details below. 2.1. Spectroscopy Spectrophotometric monitoring of UGC 06728 was carried out at APO with the 3.5 m telescope from 2015 April 15 to May 30 (UT dates here and throughout). Our monitoring program was scheduled for the first hour of almost every night during this time period, coincident with evening twilight. We employed the Dual Imaging Spectrograph (DIS), which uses a dichroic to split the incoming beam into a red arm and a blue arm, with the low-resolution (B400/R300) gratings centered at 4398 and 7493 Å. The B400 and R300 gratings, when used together, cover the entire optical bandpass between the atmospheric cutoff and 1 μm, with a nominal dispersion of 1.8 Å pix−1 and 2.3 Å pix−1 respectively. Spectra were obtained through a 5″ slit rotated to a position angle of 0° (oriented north–south) and centered on the AGN. On each visit, a single spectrum with an exposure time of 600 s was acquired at a typical airmass of 1.5. Observations of the spectrophotometric standard star Feige 34 were also acquired with each visit. All spectra were reduced with IRAF1 following standard procedures. An extraction width of 12 pixels was adopted, corresponding to an angular width of 5″ and 4 8 for the blue and red cameras, respectively. The desire to minimize sampling gaps and maximize temporal coverage means that ground-based reverberation campaigns must rely on spectroscopy obtained under nonphotometric conditions. While a spectrophotometric standard star can help correct the overall shape of the spectrum for atmospheric effects, as well as those from the telescope and instrument optics, an additional technique is required to achieve absolute flux calibrations of all the spectra. Fortuitously, the narrow emission lines do not vary on short timescales of weeks to months, so they can serve as convenient “internal” flux calibration sources. We utilize the van Groningen & Wanders (1992) spectral scaling method, which accounts for small differences in wavelength calibration, flux calibration, and resolution (from variations in the seeing). The method compares each spectrum to a reference spectrum built from the best spectra (identified by the user) and minimizes the differences within a specified wavelength range. The method has been shown to result in relative spectrophotometry that is accurate to ∼2% (Peterson et al. 1998a). We restricted the scaling algorithm to focus on the spectral region containing the [O III] λλ4959, 5007 doublet. Additionally, we adopted an overall flux scale based on the integrated [O III] λ5007 flux Figure 1. Mean (top) and root mean square (bottom) of all the blue-side spectra obtained from APO during the monitoring campaign. measured from the nights with the best observing conditions of fl5007 = 41.6 ´ 10-15 erg s−1 cm−2. The red-side spectra showed only Hα emission smoothly blended with [N II] λλ6548, 6583. Emission from [S II] λλ6716, 6730 and [O I] λλ6300, 6363 is extremely weak and difficult to detect above the continuum. With no suitable narrow lines available, we were unable to accurately intercalibrate the red-side spectra and we do not consider them further. Figure 1 displays the final mean and root mean square (rms) of all the calibrated spectra acquired throughout the campaign. The rms spectrum displays the variable spectral components, of which Hβ, He II λ 4686, and Hγ are apparent, as is the AGN continuum. 2.2. Photometry Broadband g and r images were obtained at APO with the imaging mode of the DIS spectrograph each night directly after acquiring the spectra. The dual-arm nature of the spectrograph allowed both images to be obtained simultaneously. The typical exposure time was 30 s, and a single image in each filter was obtained per visit. Images were reduced in IRAF following standard procedures. The DIS imaging mode provides a relatively small field of view (∼4′×7′), but there were a handful of convenient bright stars in all of the images (see Figure 2). We carried out aperture photometry employing circular apertures with radii of 3 78 in g and 3 6 in r, and sky annuli of 6 3−7 56 and 6 0−7 2 respectively. Calibrated g- and r-band magnitudes for three field stars were adopted from APASS (the AAVSO Photometric All-Sky Survey; Henden & Munari 2014) and set the photometric zeropoints. Photometric monitoring was also carried out with the 24 inch Miller Telescope at Hard Labor Creek Observatory (HLCO), owned and operated by Georgia State University in Hard Labor Creek State Park near Rutledge, GA. V-band images were acquired with an Apogee 2048×2048 detector, spanning a field of view of 26 3×26 3 with a pixel scale of 0 77. On a typical night, three exposures were obtained at an airmass of ∼1.5, each with an exposure time of 300 s. The wide field of view of the HLCO images included a large number of field stars, allowing us to derive a V-band light curve for UGC 06728 by employing image subtraction techniques. 1 IRAF is distributed by the National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under cooperative agreement with the National Science Foundation. 2 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. Figure 3. Spectroscopic continuum and photometric light curves (left panels) and the cross-correlation of each light curve relative to the spectroscopic continuum light curve (right panels). No apparent time delays are detected, except perhaps in the r band, and the light curve features are quite similar. course of the campaign allows us to determine these light curves directly from the spectra without carrying out any spectral modeling or decomposition, which has the potential to introduce artificial features into light curves. In Figure 3, we show the spectroscopic continuum light curve relative to the V-band residual light curve and the g and r photometric light curves (tabulated in Table 1). The V-band residual light curve does not contain significant emission from any broad emission lines, so we combined it with the continuum light curve determined from our spectra to improve the time sampling, especially in the first half of the campaign. We selected pairs of points from the two light curves that were contemporaneous within 0.5 days and fit for the best multiplicative and additive factors to bring the V-band residual fluxes into agreement with the measured continuum flux densities. These best-fit factors account for the differences in host-galaxy background light, average AGN flux level, and bandpass. The V-band light curve was scaled according to the best-fit parameters and merged with the continuum light curve. We then examined the g-band light curve from the APO photometry and found that there was no significant time delay relative to the merged continuum+V light curve, so we merged it as well by again finding the multiplicative and additive scale factors necessary to bring it into agreement with contemporaneous points in the continuum+V light curve. Our final merged continuum light curve was binned to a0.5 day sampling to improve the accuracy. The overall shape of the r-band light curve agrees with the other photometric light curves and the continuum light curve, but the variability level is somewhat damped by additional host-galaxy flux and there is possibly a slight delay in the light curve, so we did not merge the rband with the other light curves. A detectable delay in r is not unexpected, given that the filter bandpass is centered on Hα. While g is centered on Hβ, the overall contribution of Hβ to the total filter bandpass is much smaller than for Hα and r. In particular, Hβ contributes only 2% of the g-band flux, with the variable component of Hβ accounting for only 10% of the total Hβ contribution, or 0.2% of the total g-band flux. On the other hand, Hα contributes 15% of the total r-band flux. Figure 4 displays the final merged and binned continuum light curve and the broademission-line light curves (tabulated in Table 2). The variability statistics for each of the light curves Figure 2. Example r-band image acquired with the imaging mode of the DIS spectrograph at APO. The field stars used to set the magnitude zeropoint are marked with circles. The scale of the image is 3 9×6 7 and is oriented with north up and east to the right. We first registered all the images to a common alignment with the Sexterp package (Siverd et al. 2012). We then carried out the image subtraction analysis with the ISIS package (Alard & Lupton 1998; Alard 2000). ISIS builds a reference frame from the best images (specified by the user) and then uses a spatially varying kernel to convolve the reference frame to match each individual image in the dataset. Subtraction of the two results in a residual image in which all constant components have disappeared and only variable flux remains. In the case of UGC 06728, the hostgalaxy and the average AGN brightness are subtracted from all the residual images, leaving behind only the brightness of the AGN relative to its mean level. Aperture photometry is then employed to measure this variable flux, which may be positive or negative, at the location of the target of interest in each residual image, providing a V-band residual light curve. 3. LIGHT CURVE ANALYSIS Light curves for the broad emission lines Hβ, He II l4686, and Hγ were derived directly from the scaled spectra. We fit a local, linear continuum below each emission line and then integrated the flux above this continuum to determine the total emission-line flux. This includes the contribution from the narrow component of each emission line, which is simply a constant flux offset. We also determined a continuum light curve from the spectra at 5100 ´ (1 + z ) Å, which has the merit of being completely uncontaminated by emission lines. The strong continuum and emission-line variability over the 3 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. Table 1 Photometric Light Curves HJD (days) 7127.6059 7130.6023 7131.6043 7134.6484 7134.6493 7135.6058 7141.6148 7142.6108 7143.6100 7144.6099 7149.6224 7150.6151 7151.6161 7152.6159 7153.6223 7156.6189 7159.6214 7160.6215 7162.6218 7165.6511 7166.6239 7168.6555 7169.6224 7170.6218 7171.6249 g (AB mag) r (AB mag) HJD (days) V (resid. cts/10,000) 15.546±0.006 15.544±0.008 15.510±0.008 15.676±0.006 15.677±0.005 15.580±0.009 15.638±0.007 15.671±0.010 15.630±0.011 15.605±0.011 15.612±0.006 15.608±0.009 15.561±0.008 15.556±0.008 15.507±0.006 15.460±0.010 15.479±0.007 15.461±0.007 15.491±0.007 15.540±0.008 15.494±0.008 15.448±0.005 15.400±0.014 15.495±0.011 15.444±0.009 14.823±0.005 14.840±0.006 14.820±0.006 14.941±0.021 14.989±0.020 14.869±0.006 14.916±0.006 14.931±0.007 14.919±0.007 14.903±0.007 14.882±0.005 14.898±0.006 14.866±0.006 14.867±0.006 14.820±0.005 14.795±0.006 14.794±0.005 14.793±0.005 14.798±0.006 14.862±0.007 14.819±0.005 14.780±0.005 14.785±0.007 14.807±0.006 14.784±0.006 7134.7311 7135.8689 7136.7193 7142.6382 7144.6886 7145.6625 7146.8040 7147.6872 7148.5865 7150.6653 7151.7214 7152.7171 7153.6520 7156.7405 7158.6132 7159.6255 7160.6317 7162.6928 7164.6410 7165.6504 7166.7008 7167.7025 7172.6709 7173.6752 1.216±0.024 0.133±0.025 0.309±0.024 1.202±0.060 0.205±0.042 1.093±0.034 0.865±0.038 0.180±0.045 0.789±0.021 0.746±0.024 0.351±0.026 0.036±0.020 −0.319±0.022 −0.204±0.042 −0.410±0.025 −0.170±0.026 −0.346±0.009 −0.118±0.021 0.514±0.025 0.280±0.025 0.109±0.024 0.226±0.021 −0.661±0.037 −0.801±0.030 where s 2 is the variance of the fluxes, d 2 is their mean-square uncertainty, and áF ñ is the mean flux. Furthermore,column (8) is the ratio of the maximum to the minimum flux in the light curve, Rmax. We employed the interpolated cross-correlation function (ICCF) methodology (Gaskell & Sparke 1986; Gaskell & Peterson 1987) with the modifications of White & Peterson (1994) to search for time delays of the emission lines relative to the continuum. The ICCF method calculates the crosscorrelation function (CCF) twiceby firstinterpolatingone light curve and then the otherand averages the two results together to determine the final CCF. The CCF can be characterized by its maximum value (rmax ), the time delay at which the maximum occurs (tpeak ),and the centroid (tcent ) of the points around the peak above some value (typically 0.8rmax ). CCFs for each light curve relative to the continuum are displayed in Figure 4 (right panels). For the continuum light curve, this is the autocorrelation function. To quantify the uncertainties on the time delay measurements, tcent and tpeak , we employ the Monte Carlo “flux randomization/random subset sampling” method of Peterson et al. (1998b, 2004). This method is able to account for the measurement uncertainties as well as the effect of including or excluding any particular data point. The “random subset sampling” is implemented such that, from the N available data points within a light curve, N points are selected without regard to whether a point has been previously chosen. For a point that is sampled 1  n  N times, the uncertainty ofthat point is scaled by a factor of n1 2 . The typical number of points that is not selected in any specific realization is ~1 e. The “flux randomization” component takes the newly sampled light curve and modifies the flux values by a Gaussian deviation of the flux uncertainty. These modified light curves are then cross- Figure 4. Merged continuum light curve and emission-line light curves (left panels). The right panels display the cross-correlation of each light curve relative to the continuum, and the red histograms (arbitrarily scaled) display the cross-correlation centroid distributions. are tabulated in Table 3. Column (1) lists the spectral feature and column (2) gives the number of measurements in the light curve. Columns (3) and (4) list the average and median time separation between measurements, respectively. Column (5) gives the mean flux and standard deviation of the light curve, and column (6) lists the mean fractional error (based on the comparison of observations that are closely spaced in time). Column (7) lists the excess variance, computed as Fvar = s2 - d2 , áF ñ (1 ) 4 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. Table 2 Spectroscopic Light Curves HJD (days) 7127.60131629 7130.59772753 7131.59975985 7134.64244313 7135.60113042 7141.6102368 7142.60623451 7143.6053809 7144.60529841 7149.61781551 7150.61051989 7151.61149262 7152.61133458 7153.61769047 7156.61430815 7159.61682615 7160.61690956 7162.61714264 7163.62052311 7165.63733328 7166.61939046 7168.65095798 7169.61788092 7170.61722538 7171.62035657 7172.62633782 5100 ´ (1 + z ) Å (10−15 erg s−1 cm−2 Å−1) Hβ (10−15 erg s−1 cm−2) Hγ (10−15 erg s−1 cm−2) He II (10−15 erg s−1 cm−2) 2.146±0.013 2.235±0.018 2.151±0.016 1.953±0.007 2.317±0.023 2.019±0.011 1.792±0.045 2.227±0.023 2.253±0.033 1.852±0.021 1.941±0.031 2.101±0.016 2.051±0.016 2.256±0.011 2.268±0.021 2.317±0.013 2.445±0.019 2.216±0.013 2.317±0.051 2.098±0.010 2.203±0.013 2.415±0.008 2.515±0.026 2.434±0.037 2.325±0.014 2.627±0.014 76.605±0.039 65.745±0.064 66.059±0.053 59.745±0.009 60.694±0.101 55.835±0.025 54.629±0.464 55.439±0.101 58.688±0.236 56.921±0.089 53.080±0.185 60.457±0.054 57.825±0.050 63.765±0.025 63.760±0.086 67.404±0.033 73.758±0.067 72.343±0.031 65.958±0.505 64.085±0.019 70.381±0.032 66.380±0.012 67.836±0.125 76.008±0.228 66.832±0.036 72.460±0.035 45.467±0.091 31.085±0.126 34.986±0.098 29.567±0.011 30.367±0.206 27.859±0.045 31.498±1.061 31.572±0.221 26.036±0.502 31.359±0.145 25.462±0.354 36.374±0.109 34.132±0.107 32.758±0.042 38.456±0.175 34.757±0.066 31.379±0.128 37.067±0.058 32.747±1.106 32.687±0.026 35.662±0.059 35.326±0.015 35.631±0.273 36.909±0.465 42.273±0.072 39.271±0.062 12.614±0.092 7.226±0.158 11.465±0.127 8.501±0.014 5.223±0.263 3.583±0.056 7.873±1.207 2.160±0.262 L 4.714±0.176 8.538±0.455 8.983±0.131 12.753±0.123 19.423±0.055 1.949±0.212 14.462±0.080 19.432±0.164 15.755±0.076 11.655±1.309 9.924±0.032 10.006±0.076 15.805±0.019 14.877±0.332 19.810±0.575 13.417±0.088 22.804±0.080 time delays, corrected for a factor of 1 + z , are formally the same within the uncertainties. correlated with the ICCF method described above, and the whole process is repeated many times (N = 1000). From the large set of realizations, we build distributions of tcent and tpeak . The median of each distribution is taken to be the measurement value, and the uncertainties are set such that they mark the upper 15.87% and lower 15.87% of the realizations (corresponding to 1s for a Gaussian distribution). The red histograms in the Figure 4 depict the cross-correlation centroid distribution for each emission line. To further check that combining the various photometric and spectroscopic light curves has not affected our measured time delays, we also determined the time delay of Hβ relative to each of the individual continuum, V-band, and g-band light curves. Each of these light curves is slightly undersampled relative to the combined continuum light curve, but the CCFs and recovered Hβ time delays agree within the measurement uncertainties. We also investigated the time delays with the JAVELIN package (Zu et al. 2011). JAVELIN fits the continuum variations with a damped random walk model. It then assumes a top hat model for the reprocessing function, and determines the best-fit shifting and smoothing parameters for the emissionline light curves by maximizing the likelihood of the model. Uncertainties on each of the model parameters are assessed through a Bayesian Markov Chain Monte Carlo method. We denote time delays from JAVELIN as tjav . Given the extremely short time delays, we were unable to fit a single model while including all the emission lines simultaneously, so we instead modeled each emission line separately relative to the continuum (see Figure 5). Time delay measurements are listed in Table 4. While each of the measurements is an observed time delay, the rest-frame 4. LINE WIDTH MEASUREMENTS The widths of the broad emission lines in AGN spectra are interpreted as the line-of-sight velocities of the bulk motion of the gas. The narrow emission lines, however, are known to emit from gas that is not participating in the same bulk motion. Therefore, good practice is to isolate the broad emission from the narrow emission when quantifying the line width. In the spectrum of UGC 06728, however, it is not clear what part of the Hβ line is narrow emission (seeFigure 1). Furthermore, the narrow lines contribute almost no signal to the rms spectrum, demonstrating that our internal spectral scaling method has minimized their apparent variability from changing observing conditions throughout the monitoring campaign. As it is the variable part of the emission line (the rms profile) that we are most interested in, we do not attempt any narrow line subtraction for this object. We measured the widths of the broad Hβ, He II λ4686, and Hγ emission lines in both the mean and the rms spectra and we report two different line width characterizations: the full width at half the maximum flux (FWHM) and the second moment of the line profile (sline ). Line widths were measured directly from the spectra, with each line profile defined as the flux above a local linear continuum. Uncertainties in the emission-line widths were determined using a Monte Carlo random subset sampling method. From a set of N spectra, a subset of N spectra were selected without regard to whether they had been previously chosen. The mean and rms of the subset were created, from which the FWHM and sline of an emission line were determined and recorded. Distributions of line width 5 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. Table 3 Light Curve Statistics Time Series N áT ñ (days) Tmedian (days) 5100 Å V g r Hβ Hγ He II 26 24 25 25 26 26 25 1.8±1.4 1.7±1.3 1.8±1.4 1.8±1.4 1.8±1.4 1.8±1.4 1.9±1.5 1.0 1.1 1.0 1.0 1.0 1.0 1.0 áF ña ásF F ñ Fvar Rmax 2.21±0.20 −0.22±0.56 2.83±0.21 3.19±0.17 64.3±6.7 33.9±4.6 11.3±5.7 0.009 0.050 0.008 0.007 0.002 0.007 0.033 0.090 −2.56 0.072 0.052 0.105 0.135 0.500 1.466±0.038 −0.659±0.028 1.290±0.017 1.213±0.023 1.443±0.005 1.786±0.025 11.7±1.3 Note. 5100 Å, g-bandand r-band flux densities are in units of 10−15 erg s−1 cm−2 Å−1. V-band residual flux is in units of 10,000 counts. Emission-line fluxes are in units of 10−15 erg s−1 cm−2. a therefore constrain Dl disp , from high-quality, high-resolution observations of the narrow emission lines in the literature. However, we have previously monitored other AGNs with this same instrumental setup, and so we adopt the value of Dl disp = 14.1 Å that we determined for NGC 5273 from a spring 2014 monitoring campaign (Bentz et al. 2014). Our final resolution-corrected line width measurements are listed in Table 5. 5. BLACK HOLE MASS All of the time delays measured for UGC 06728 are very short, which is to be expected given the low luminosity of the AGN. The time delays determined for Hβ are the only ones that are not formally consistent with zero within the measurement uncertainties, so Hβ is the only emission line we will consider for the determination of the black hole mass. However, Hβ is also the emission line for which we have the largest number of reverberation results (seeBentz & Katz 2015 for a recent summary), so it is also the most reliable emission line for determining MBH . The black hole mass is generally determined from reverberation-mapping measurements as Figure 5. Continuum and Hβ light curves with interpolated JAVELIN light curves drawn from the distribution of acceptable models. Table 4 Time Lags Feature Hβ Hγ He II tcent (days) tpeak (days) tjav (days) +0.7 1.40.8 +1.0 0.0-1.3 +0.9 -0.21.1 +0.6 1.10.6 +2.5 -0.70.7 +1.8 -0.70.7 +0.2 1.30.7 +0.1 -1.50.7 +0.2 -1.40.1 MBH = f (3 ) where τ is the time delay for a specific emission line relative to variations in the continuum, and V is the line-of-sight velocity width of the emission line, with c and G being the speed of light and gravitational constants, respectively. The emissionline time delay is interpreted as a measure of the responsivityweighted average radius of the broadline region for that specific emission feature (e.g., Hβ). The scaling factor f accounts for the detailed geometry and kinematics of the broadline-region gas, which is unresolvable. In practice, the multiplicative factor, á f ñ, which is found to bring the MBH –s relationship for AGNs with reverberation masses into agreement with the MBH –s relationship for nearby galaxies with dynamical black hole masses (e.g., Gültekin et al. 2009; Kormendy & Ho 2013; McConnell & Ma 2013) is used as a proxy for f. In this way, the population average factor provides an overall scale for reverberation masses that should be unbiased, but the mass of any particular AGN is expected to be uncertain by a factor of two to threebecause of object-toobject variations. The value of á f ñ has varied in the literature from 5.5 (Onken et al. 2004) to 2.8 (Graham et al. 2011), measurements were built up over 1000 realizations. We take the mean and the standard deviation of each distribution as the measurement and its uncertainty, respectively. Following Peterson et al. (2004), we corrected the emissionline widths for the dispersion of the spectrograph. The observed emission-line width, Dl obs, can be described as Dl 2obs » Dl 2true + Dl 2disp, c tV 2 , G (2 ) where Dltrue is the intrinsic line width and Dl disp is the broadening induced by the spectrograph. The employment of a wide spectrograph slit for reverberation campaigns means that the spectrograph dispersion cannot be determined from night sky emission lines or from arc lamp lines—the unresolved AGN point source, even under poor seeing conditions, will not fill the spectrograph slit. Given the relative obscurity of this particular AGN, we were unable to estimate Dltrue , and 6 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. Table 5 Line Widths Mean Feature Hβ Hγ He II FWHM (km s−1) 1144.5±58.3 2333.6±80.3 2626.2±593.7 rms sline (km s−1) FWHM (km s−1) sline (km s−1) 758.3±19.4 821.8±21.8 1124.7±127.7 1309.7±182.2 2492.3±1704.7 4016.7±912.9 783.7±92.3 919.9±70.4 1605.6±157.8 depending on which objects are included and the specifics of the measurements. We adopt the value determined by Grier et al. (2013) of á f ñ = 4.3  1.1. Combining the time lag (tcent ) and line width (sline ) measurements for Hβ and scaling by á f ñ, we determine MBH = (7.1  4.0) ´ 10 5 M. 6. DISCUSSION The extremely rapid response of the broad emission lines to variations in the continuum flux in UGC 06728 means that our daily sampling was not fine enough to resolve time delays for all the broad optical recombination lines. The time delay of Hβ is the only one that is not formally consistent with zero delay, and it is only marginally resolved at that. However, while we were not able to resolve the time delays for Hγ and He II, we can examine them in light of the expected virial relationship for BLR gas that is under the gravitational dominance of the black hole. In particular, we would expect that R µ V -2 . This relationship has been shown to be a good description of observations when reverberation results from multiple emission lines have been recovered (e.g., Peterson et al. 2004; Kollatschny 2003; Bentz et al. 2010). Figure 6 shows the measurements for the optical recombination lines in UGC 06728, with the expected relationship scaled to match the measurements for Hβ. There is generally good agreement with the expected relationship within the measurement uncertainties, such that we would not expect to resolve the responses of these emission lines with our current sampling. A monitoring campaign with finer temporal resolution (Dt = 0.25–0.5 days) would be needed to further improve upon these constraints. Figure 6. Line width vs. time delay as measured from the broad optical recombination lines in the spectrum of UGC 06728. The dotted line shows the expected relationship of R µ V -2 and is scaled to match the measurements for Hβ. Even though the time delays are quite short, and unresolved in the case of Hγ and He II, the measurements are in relatively good agreement with the expected relationship. hardly comparable to the quality afforded by HST, the DIS images do allow us to place some rough constraints on the starlight contribution to the flux density at 5100 ´ (1 + z ) Å. We aligned and stacked several of the g-band images to increase the signal-to-noise in the combined image. Using the two-dimensional surface brightness fitting program GALFIT (Peng et al. 2002, 2010), we created a model of the pointspread function (PSF) of the stacked image by fitting multiple Gaussian components to the profile of a field star in a restricted portion of the image. We then employed this model PSF while fitting the full frame, including a background sky gradient, a PSF for the AGN and the nearby star, an exponential profile for the disk of the galaxy, and a Sérsic profile for the bulge. The bulge profile, in particular, is very compact with a half-light radius of 1.7 pix (0 7), and likely degenerate with the AGN PSF, so we caution that our estimate of the starlight contribution is probably more like a lower limit. Figure 7 displays a 2 5×2 5 region of the stacked g-band image, our best-fit model from GALFIT, and the residuals after subtracting the model from the image. As described earlier, calibrated g-band photometry for three field stars from APASS (the AAVSO Photometric All-Sky Survey; Henden & Munari 2014) was used to set the overall flux scale of the image. We also account for a slight flux scaling factor, due to the difference in effective wavelength of the g filter compared to 5100 ´ (1 + z ) Å, using Synphot and a template galaxy bulge spectrum (Kinney et al. 1996). Our estimate of the host-galaxy contribution to the spectroscopic flux density is Removing fgal = (1.09  0.22) ´ 10-15 erg s−1 cm−2 Å−1. 6.1. Consistency with the RBLR –L Relationship Furthermore, we can examine the location of UGC 06728 on the AGN RBLR –L relationship to further assess the Hβ time delay measurement. For very nearby galaxies, like UGC 06728, however, one complication is the large fraction of host-galaxy starlight that contributes to the continuum emission at restframe 5100 Å through the large spectroscopic slit (∼5″) employed in a reverberation mapping campaign. The usual method to correct for this contamination is to carry out twodimensional surface brightness modeling of a high-resolution image of the galaxy (usually from the Hubble Space Telescope to maximize the image quality), thereby isolating the hostgalaxy starlight components from the unresolved AGN point source. Using the modeling results to create an “AGN-free” image allows the starlight contribution to be directly constrained (Bentz et al. 2006b, 2009, 2013). Unfortunately, there are no HST images of UGC 06728. The highest resolution optical images available are the APO DIS g-band images discussed above, with a pixel scale of 0 42 pixel−1. While 7 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. Figure 7. Stacked g-band image of a 2 5×2 5 region centered on UGC 06728 (left) with the white rectangle showing the geometry of the ground-based spectroscopic monitoring aperture. The best-fit model determined from GALFIT is displayed in the middle panel, and the right panel shows the residuals after subtraction of the model from the image. All images are oriented with north up and east to the right. this contribution results in an AGN-only continuum flux density of fAGN = (1.12  0.23) ´ 10-15 erg s−1 cm−2 Å−1. Assuming a luminosity distance of DL=27 Mpc and correcting for Galactic absorption along the line of sight (Schlafly & Finkbeiner 2011), we derive log lLl = 41.83  0.24 erg s−1. Figure 8 displays the RBLR –L relationship for nearby AGNs based on reverberation mapping of Hβ (Bentz et al. 2013). The filled circle shows the location of UGC 06728 with the Hβ time delay we have derived here and the luminosity after correction for the estimated starlight contribution. The agreement between UGC 06728 and its expected location based on its estimated luminosity is extremely good considering the barely resolved nature of the time delay and the caveats in the luminosity determination. Furthermore, we can expect that the agreement is actually somewhat better than depicted, given the likelihood that the starlight correction to the luminosity is underestimated as described above. Taking our galaxy decomposition at face value, we can estimate the bulge-to-total ratio as B T » 0.2, which suggests that the Hubble type of the galaxy is ∼Sb (Kent 1985). We also estimate the color of the galaxy as g - r » 0.9, which suggests M Lg » 6 (Zibetti et al. 2009). The total stellar mass of the galaxy is M » 7.5 ´ 109 M, which also agrees with the hostgalaxy being Sb−Sc in type. Figure 8. Hβ time delay for UGC 06728 and estimated AGN luminosity (filled point) compared to the radius–luminosity relationship for other reverberationmapped AGNs (Bentz et al. 2013). calibration, as well as spectra of HD 125560 (spectral type K3III) and HD 117876 (spectral type G8III) to provide velocity templates with the same wavelength coverage and dispersion as the galaxy. All spectra were reduced with IRAF following standard procedures. An extraction width of 40 pixels (corresponding to 16″ on the blue camera and 16 8 on the red camera) was adopted to maximize the galaxy signal in the resultant spectra. Following flux calibration of the spectra, we employed the pPXF (Penalized Pixel Fitting) method of Cappellari & Emsellem (2004) to extract the stellar kinematics. The Mgb absorption signature was not detected in the galaxy spectra, but the Ca II triplet features were detected, so we focused on fitting the red spectra only. During the fitting process, we restricted the wavelength region to 8525−8850 Å and determined the best-fit parameters (velocity, velocity dispersion, h3, and h4) using first one velocity template star and then the other. The best fits to the spectrum of UGC 06728 are displayed in Figure 9: HD125560 (red line) provided a best-fit velocity dispersion of 56.5 km s−1and HD117876 (blue line) provided a best fit of 46.7 km s−1. We take the average of these as the bulge stellar velocity dispersion, s = 51.6  4.9 km s−1. 6.2. Consistency with the MBH –s Relationship To further explore the reverberation results for UGC 06728 within the context of the larger reverberation sample, we obtained supplemental observations on 2016 May 13 with the DIS Spectrograph on the APO 3.5-m telescope with the intent of constraining the bulge stellar velocity dispersion. The highresolution B1200 and R1200 gratings were employed, providing nominal dispersions of 0.62 and 0.58 Å pix−1 and wavelength coverages of 1240 and 1160 Å, respectively. The blue grating was centered at 4900 Å to target the Mgb stellar absorption signature, and the red grating was centered at 8500 Å for the Ca II triplet absorption. The 0. 9 slit was rotated to a position angle of 150° east of north, approximately along the minor axis of the galaxy. Given the high inclination of the galaxy, we specifically avoided the major axis of the galaxy to mitigate the effects of rotational broadening from the disk within the one-dimensional extracted spectra. Two 1200 s exposures were obtained through patchy clouds and with marginal seeing at an airmass of 1.6. Spectra of the standard star, Feige 34, were also obtained to assist with the flux 8 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. Figure 9. Spectrum of UGC 06728 in the wavelength region around the Ca II triplet absorption lines. The red and blue lines show the best-fit models to the stellar absorption lines based on HD 125560 and HD 117876, respectively. We take the average of the solutions provided by the two template stars as our measurement of the bulge stellar velocity dispersion in UGC 06728. Figure 10. UGC 06728 (filled point) and the AGN MBH –s relationship from Grier et al. (2013). black hole mass measurements with current and near-future technology, UGC 06728 could potentially be a worthwhile target for dynamical modeling. With this constraint on the bulge stellar velocity dispersion in UGC 06728, we can explore its location on the AGN MBH –s relationship. Figure 10 displays the AGN MBH –s relationship from Grier et al. (2013; open points and line), with the location of UGC 06728 shown by the filled circle. The scatter at the low-mass end of the MBH –s relationship for AGNs with reverberation masses seems to be much smaller than that found for megamaser host galaxies (Greene et al. 2010). Läsker et al. (2016) also found the megamaser host galaxies to have a high scatter relative to the MBH –L bulge and MBH –Mbulge relationships. Each sample of direct black hole masses, whether dynamical, reverberation, or masering, has its own set of biases and assumptions that are independent of the other techniques, so further exploration into this apparent disagreement is likely to shed light on the reliability of black hole mass measurements as they are currently applied. Furthermore, we can estimate the black hole sphere of influence (rh) in the nucleus of UGC 06728. Generally defined as rh = GMBH , s 2 6.3. Mass and Spin Implications Walton et al. (2013) analyzed Suzaku observations of UGC 06728 and determined that it was a “bare” AGN, with minimal intrinsic absorption. Fitting the X-ray spectrum with a relativistic reflection model, and assuming an accretion disk inclination of i=45°, they determined a dimensionless spin parameter of a > 0.7, indicating thatthe black hole is spinning rapidly. Combined with our mass contraint of MBH = (7.1  4.0) ´ 10 5 M, UGC 06728 is one of a small number of massive black holes that are completely characterized. A few other low-mass black holes have both mass and spin constraints, and they appear to agree with the properties derived for UGC 06728. MCG-06-30-15 is only slightly more massive with MBH = (1.6  0.4) ´ 106 M (Bentz et al. 2016) and is spinning near maximally (a > 0.9; Brenneman & Reynolds 2006; Chiang & Fabian 2011; Marinucci et al. 2014). NGC 4051 is another example, with MBH = (1.3  0.4) ´ 106 M (Denney et al. 2009b) and a > 0.99 (Patrick et al. 2012). Black hole evolutionary models have only recently begun to treat black hole spin in addition to mass. Depending on the model, it is not clear if the properties of the black hole in UGC 06728 are expected or surprising. For example, the model of Volonteri et al. (2013) predicts that black holes with MBH » 106 M in gas-rich galaxies at z < 0.5 (including AGNs) should have slowly rotating black holes with dimensionless spin parameters of a < 0.4 . This model is based on many observational constraints, including the MBH –s relationship, with which we have shown UGC 06728 to be in agreement. One caveat to the evolutionary model of Volonteri et al. (2013) is that it does not account for black hole feeding through disk instabilities, which could be a reason for the apparent discrepancy here. Disk instability accretion events would likely be correlated and serve to spin up a black hole. The evolutionary models of Sesana et al. (2014) attempt to include this effect by linking the gas dynamics of the extended galaxy to the central black hole. Their models predict that local black holes with MBH » 106 M should tend to be spinning (4 ) rh is often employed as a convenient metric for determining the probability of success for constraining MBH from spatially resolved stellar dynamics. However,Gültekin et al. (2009) argue that a strict reliance on resolving rh is not necessary for useful constraints on black hole masses. Combining our measurements of MBH and s and again assuming a luminosity distance of DL=27 Mpc, we estimate rh=0 01 for UGC 06728. While this angular size is smaller than the achievable spatial resolution of integral field spectrographs on the largest ground-based telescopes today, it is interesting to note that it is not much smaller than rh for NGC 3227. Davies et al. (2006) were able to constrain the black hole mass of NGC 3227 through stellar dynamical modeling, even though the reverberation mass and bulge stellar velocity dispersion predict rh=0 018. Given the very limited number of AGNs where it will be possible to carry out a direct comparison of reverberation-based and stellar dynamical-based 9 The Astrophysical Journal, 831:2 (10pp), 2016 November 1 Bentz et al. near maximally, and that accreting black holes in spiral galaxies should also tend to have near-maximal spins. Interpretation of black hole spin measurements is still somewhat debated as well. Bonson & Gallo (2016) argue that black hole spins tend to be overestimated in many cases, though they state thatthis is likely not the case for the most maximally spinning black holes (a > 0.8). Furthermore, there is a very strong selection bias inherent in the sample of AGNs with spin measurements. Rapidly spinning black holes have significant boosts to their X-ray flux through increased radiative efficiency, and the current sample of AGNs with spin constraints is based on observations of the brightest X-ray sources, so the current sample will strongly favor rapidly spinning black holes (Brenneman et al. 2011; Vasudevan et al. 2016). In any case, UGC 06728 is an important addition to the sample. As the least massive central black hole that has been fully described, it will help to anchor future studies, both observational and theoretical, of central black hole demographics. Bentz, M. C., Horenstein, D., Bazhaw, C., et al. 2014, ApJ, 796, 8 Bentz, M. C., & Katz, S. 2015, PASP, 127, 67 Bentz, M. C., Peterson, B. M., Netzer, H., Pogge, R. W., & Vestergaard, M. 2009, ApJ, 697, 160 Bentz, M. C., Peterson, B. M., Pogge, R. W., Vestergaard, M., & Onken, C. A. 2006b, ApJ, 644, 133 Bentz, M. C., Walsh, J. L., Barth, A. J., et al. 2010, ApJ, 716, 993 Blandford, R. D., & McKee, C. F. 1982, ApJ, 255, 419 Bonson, K., & Gallo, L. C. 2016, MNRAS, 458, 1927 Brenneman, L. W., & Reynolds, C. S. 2006, ApJ, 652, 1028 Brenneman, L. W., Reynolds, C. S., Nowak, M. A., et al. 2011, ApJ, 736, 103 Cappellari, M., & Emsellem, E. 2004, PASP, 116, 138 Chiang, C.-Y., & Fabian, A. C. 2011, MNRAS, 414, 2345 Davies, R. I., Thomas, J., Genzel, R., et al. 2006, ApJ, 646, 754 den Brok, M., Seth, A. C., Barth, A. J., et al. 2015, ApJ, 809, 101 Denney, K. D., Peterson, B. M., Pogge, R. W., et al. 2009a, ApJL, 704, L80 Denney, K. D., Watson, L. C., Peterson, B. M., et al. 2009b, ApJ, 702, 1353 Ferrarese, L., & Ford, H. 2005, SSRv, 116, 523 Ferrarese, L., & Merritt, D. 2000, ApJL, 539, L9 Gaskell, C. M., & Peterson, B. M. 1987, ApJS, 65, 1 Gaskell, C. M., & Sparke, L. S. 1986, ApJ, 305, 175 Gebhardt, K., Bender, R., Bower, G., et al. 2000, ApJL, 539, L13 Genzel, R., Pichon, C., Eckart, A., Gerhard, O. E., & Ott, T. 2000, MNRAS, 317, 348 Ghez, A. M., Morris, M., Becklin, E. E., Tanner, A., & Kremenek, T. 2000, Natur, 407, 349 Ghez, A. M., Salim, S., Weinberg, N. N., et al. 2008, ApJ, 689, 1044 Graham, A. W., Onken, C. A., Athanassoula, E., & Combes, F. 2011, MNRAS, 412, 2211 Greene, J. E., Peng, C. Y., Kim, M., et al. 2010, ApJ, 721, 26 Grier, C. J., Martini, P., Watson, L. C., et al. 2013, ApJ, 773, 90 Gültekin, K., Richstone, D. O., Gebhardt, K., et al. 2009, ApJ, 698, 198 Heckman, T. M., & Best, P. N. 2014, ARA&A, 52, 589 Henden, A., & Munari, U. 2014, CoSka, 43, 518 Hicks, E. K. S., & Malkan, M. A. 2008, ApJS, 174, 31 Kent, S. M. 1985, ApJS, 59, 115 Kinney, A. L., Calzetti, D., Bohlin, R. C., et al. 1996, ApJ, 467, 38 Kollatschny, W. 2003, A&A, 407, 461 Kormendy, J., & Ho, L. C. 2013, ARA&A, 51, 511 Läsker, R., Greene, J. E., Seth, A., et al. 2016, ApJ, 825, 3 Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115, 2285 Marinucci, A., Matt, G., Miniutti, G., et al. 2014, ApJ, 787, 83 McConnell, N. J., & Ma, C.-P. 2013, ApJ, 764, 184 Onken, C. A., Ferrarese, L., Merritt, D., et al. 2004, ApJ, 615, 645 Onken, C. A., Valluri, M., Brown, J. S., et al. 2014, ApJ, 791, 37 Patrick, A. R., Reeves, J. N., Porquet, D., et al. 2012, MNRAS, 426, 2522 Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H. 2002, AJ, 124, 266 Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2010, AJ, 139, 2097 Peterson, B. M. 1993, PASP, 105, 247 Peterson, B. M., Bentz, M. C., Desroches, L. B., et al. 2005, ApJ, 632, 799 Peterson, B. M., Ferrarese, L., Gilbert, K. M., et al. 2004, ApJ, 613, 682 Peterson, B. M., Wanders, I., Bertram, R., et al. 1998a, ApJ, 501, 82 Peterson, B. M., Wanders, I., Horne, K., et al. 1998b, PASP, 110, 660 Reynolds, C. S. 2014, SSRv, 183, 277 Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103 Sesana, A., Barausse, E., Dotti, M., & Rossi, E. M. 2014, ApJ, 794, 104 Siverd, R. J., Beatty, T. G., Pepper, J., et al. 2012, ApJ, 761, 123 van den Bosch, R. 2016, ApJ, in press (arXiv:1606.01246) van Groningen, E., & Wanders, I. 1992, PASP, 104, 700 Vasudevan, R. V., Fabian, A. C., Reynolds, C. S., et al. 2016, MNRAS, 458, 2012 Volonteri, M., Sikora, M., Lasota, J.-P., & Merloni, A. 2013, ApJ, 775, 94 Walton, D. J., Nardini, E., Fabian, A. C., Gallo, L. C., & Reis, R. C. 2013, MNRAS, 428, 2901 White, R. J., & Peterson, B. M. 1994, PASP, 106, 879 Zibetti, S., Charlot, S., & Rix, H.-W. 2009, MNRAS, 400, 1181 Zu, Y., Kochanek, C. S., & Peterson, B. M. 2011, ApJ, 735, 80 7. SUMMARY We present an Hβ time delay and a reveberation-based black hole mass for the nearby, low-luminosity Seyfert UGC 06728. With t = 1.4  0.8 days and MBH = (7.1  4.0) ´ 10 5 M, UGC 06728 is at the low end of observed properties within the reverberation mapping sample. The time delay and estimated AGN luminosity agree with the RBLR –L relationship for other reveberation-mapped AGNs, and a measurement of s = 51.6  4.9 km s−1 from long-slit spectroscopy shows that the black-hole mass agrees with the AGN MBH –s relationship. With MBH < 106 M, UGC 06728 is currently the lowest-mass central black hole that is fully described by both direct mass and spin constraints. We thank the referee for thoughtful comments that improved the presentation of this paper. M.C.B.gratefully acknowledges support from the NSF through CAREER grant AST-1253702. This research is based on observations obtained with the Apache Point Observatory 3.5 meter telescope, which is owned and operated by the Astrophysical Research Consortium. We heartily thank the staff at APO for all their help with this program. This research has made use of the AAVSO Photometric All-Sky Survey (APASS), funded by the Robert Martin Ayers Sciences Fund. This research has made use of the NASA/IPAC Extragalactic Database (NED),which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration and the SIMBAD database, operated at CDS, Strasbourg, France. REFERENCES Alard, C. 2000, A&AS, 144, 363 Alard, C., & Lupton, R. H. 1998, ApJ, 503, 325 Bentz, M. B., Cackett, E. M., Crenshaw, D. M., et al. 2016, ApJ, in press (arXiv:1608.01229) Bentz, M. C., Denney, K. D., Cackett, E. M., et al. 2006a, ApJ, 651, 775 Bentz, M. C., Denney, K. D., Grier, C. J., et al. 2013, ApJ, 767, 149 10