Introduction

Accretion onto supermassive black holes through an accretion disk of ionized gas powers active galactic nuclei (AGNs) at the centers of massive galaxies. AGNs accreting at typical rates (a few percent of the Eddington limit) have a geometrically thin but optically thick disk-the "thin-disk" model (Shakura & Sunyaev 1973). However, theoretical models predict a notably different structure for the AGN with high accretion rates significantly above the Eddington limit-super-Eddington AGN (e.g., Abramowicz et al. 1988). The occurrence of such AGNs is likely higher during the peak era of supermassive black hole growth during cosmic noon (redshifts, z = 1-3; Brandt & Alexander 2010;Shen et al. 2020). Understanding the structure of the accretion system in high-Eddington AGNs remains an open issue in accretion physics.

Although models exist for slim-disk systems, observational tests of the structure of the accretion flow in super-Eddington AGN are rare. At high accretion rates, radiation pressure is expected to dominate, causing the inner disk to inflate verticallynow called a "slim" (rather than thin) disk-with a scale height, H R, where R is the disk radius (e.g., Abramowicz et al. 1988). Photons are trapped in the fast-flowing matter, eventually falling into the black hole. Given that not all photons escape, the disks in super-Eddington AGNs are underluminous relative to the accretion rates as compared to thin disks (Jaroszyński et al. 1980). Begelman (2002) proposed an alternative scenario where the "photon-bubble instability" principle can cause the disks in super-Eddington AGNs to become inhomogeneous at scales much smaller than the disk scale height.

Reverberation mapping (RM; Blandford & McKee 1982;Peterson 1993) provides a way to observationally study the slimdisk model and broad-line region (BLR) in super-Eddington AGNs. RM takes advantage of the observed continuum variability of AGNs on many timescales (from several days to weeks and years; e.g., Peterson et al. 1982). The accretion-disk emission illuminates the BLR on larger scales, and sets the ionization structure and thus the location of the gas generating the broad emission lines (e.g., Hβ). An increase in continuum emission from the accretion disk results in an increase in broad emission-line flux after a time lag set by the sum total of the light travel time between the continuum-emitting region and the BLR (Peterson 2014), and the recombination timescale, where the latter is much smaller than the former for typical BLR densities (and therefore ignored in the time-lag calculations). RM converts this time lag into a spatial distance, the size of the BLR. Thus, applying RM to highaccretion-rate AGNs gives an observational method to test the structure of the accretion flow and BLR in these systems, and place super-Eddington AGNs on the radius-luminosity (R-L) relationship for AGNs (Kaspi et al. 2005;Bentz et al. 2013).

The Narrow-Line Seyfert 1 (NLS1) class of AGNs are considered to have high accretion rates, and typically display narrow broad emission lines (e.g., the Hβ line has an FWHM  2000 km s -1 ) in comparison to other broad-line objects (and broader than the narrow lines seen in type 2 objects), strong Fe II emission lines, and weak [O III] lines (e.g., Osterbrock & Pogge 1987;Boroson & Green 1992;Boller et al. 1996;Véron-Cetty et al. 2001) in their spectra. The Super-Eddington Accreting Massive Black Holes (SEAMBH) campaign has been performing photometric and spectroscopic monitoring over the past 9 yr of high accretion-rate AGNs that display spectral characteristics of NLS1s (e.g., Du et al. 2014Du et al. , 2015Du et al. , 2016aDu et al. , 2016bDu et al. , 2018;;Wang et al. 2014a;Hu et al. 2015;Li et al. 2018Li et al. , 2021)). Du et al. (2016b) showed that the BLRs in super-Eddington AGNs are smaller than those with sub-Eddington accretion rates. In the context of the slim-disk model, the smaller BLR sizes can be explained as a consequence of the increased scale height of the inner accretion disk that shields the BLR from the central ionizing flux (Wang et al. 2014b). Hβ, a marker of the hydrogen ionization front in the BLR, can thus exist at smaller radii than in thin accretion-disk systems. Fonseca Alvarez et al. (2020) offered an alternative explanation. In their correlation analysis of the physical and spectral properties of the Sloan Digital Sky Survey (SDSS) RM AGNs, Fonseca Alvarez et al. (2020) found that the R-L offset (defined as the ratio of the measured Hβ time lag to the expected time lag from the best-fit R-L such as that given by Bentz et al. 2013) is positively correlated to the [O III] λ5008 to Hβ luminosity ratio, which is often used as a proxy for the number of ionizing photons (e.g., Baldwin et al. 1981). The smaller BLR sizes are therefore likely a result of the changes in the shape of the ultraviolet (UV)/optical spectral energy distribution (SED) of AGNs (Fonseca Alvarez et al. 2020).

As the most promising SEAMBH object-a bright target with an extremely super-Eddington accretion rate (   = M M 250;

Edd Li et al. 2018) and a well-measured Hβ lag-Markarian 142 (Mrk 142 or PG 1022+519, R.A. = 10 h 25 m 31 20, decl. = +51°4 0′34 87, z = 0.045) is the target of our study to probe the structure of its BLR. In the 2012 SEAMBH campaign, Mrk 142 was highly variable with a fractional variability amplitude of F var = 8.1% at 5100 Å over a period of 6 months. Its variable nature makes it amenable to RM studies of both accretion-disk structure (from X-ray/UV/optical continuum time-lag studies) and the BLR structure (from continuum-emission line time lags). Accretion-disk RM applies the same principle as BLR RM to the inner and outer regions of the accretion disk to determine its size and temperature profile (Cackett et al. 2007). The more energetic X-ray/UV radiation from the inner disk illuminates the disk at larger radii where the optical photons are generated. Therefore, the lower-energy emission will respond with a positive time lag to changes in the high-energy radiation giving rise to correlated continuum light curves. Mrk 142 has a total Hβ time lag (τ) with respect to the 5100 Å continuum emission of -+ 7.9 1.1

1.2 days (Du et al. 2015) and a black hole mass of

6.23 0.45 0.26 (Li et al. 2018). In this paper, we present Mrk 142 time-lag measurements from two ground-based, optical spectroscopic RM campaigns of Mrk 142 concurrent with the photometric monitoring of the target with the Neil Gehrels Swift Observatory (Swift) in a UV band; and the Las Cumbres Observatory (LCO), Dan Zowada Memorial Observatory (Zowada), and Liverpool Telescope (Liverpool; Steele et al. 2004) in an optical band. With our joint campaign, we performed, for the first time, simultaneous measurements of the inner accretion disk and BLR size in a super-Eddington AGN. This paper is organized as follows. In Section 2, we provide details of the observations, and in Section 3, we explain the process of data reduction. In Section 4, we describe our spectral modeling followed by lightcurve analysis in Section 5. In Section 6, we outline and discuss our results in the context of previous studies. Section 7 provides closing remarks. Throughout this work, we use the standard cosmology with H 0 = 67 km s -1 Mpc -1 , Ω Λ = 0.68, and Ω M = 0.32 (Planck Collaboration et al. 2014).

Observations

We obtained concurrent observations of Mrk 142 with multiple telescopes to perform RM analysis of the accretion disk and BLR simultaneously. Figure 1 shows the continuum light curves of Mrk 142, highlighting the simultaneous coverage with different telescopes.

Gemini North Telescope

We obtained new observations of Mrk 142 long-slit spectra taken with the Gemini Multi-Object Spectrograph (GMOS; Hook et al. 2004) on the 8.1 m Gemini North Telescope (Gemini) on Maunakea, Hawai'i with 33 epochs from 2019 February 6 through June 1. These spectral observations are concurrent with the Mrk 142 photometric data from the Swift telescope comprising 180 epochs of 1 ks exposures at X-ray, UV, and optical wavelengths from 2019 January 1 through April 30 (P.I.: E. Cackett) as well as with the photometric gband data from LCO (P.I.: R. Edelson), Zowada (P.I.: E. Cackett), and Liverpool (P.I.: M. Goad); the photometric data are presented in Cackett et al. (2020). The Swift observations had a twice-daily cadence until March 19, and the cadence was decreased to daily from March 20 onward. We required observations from Gemini in early 2019 with considerable overlap with the Swift campaign to allow, for the first time, simultaneous measurements of the UV-emitting accretion disk and the BLR of a super-Eddington AGN. The cadence of the Gemini observations was set to 1 day. We obtained data for only two, sparsely separated epochs in February during the beginning of the observing period due to weather interruptions. However, observations were more frequent in March and May, and the daily cadence was achieved in the first week of April.

The spectra were taken with the GMOS-North Hamamatsu detector and a single grating, B600 with two different slits, 0 75 (narrow slit) and 5 00 (wide slit), in the two-target acquisition mode, where Mrk 142 and a comparison star were observed in the same slit. The GMOS-North Hamamatsu detector comprises three ∼2048 × 4176 pixel chips (full detector size of 6278 × 4176 pixels, mosaicked) arranged in a row with pixel size of 0 0807 with two chip gaps 4 88 wide. The choice of the grating was made to obtain the broad emission line of interest, Hβ λ4861, in the spectra. The narrow 0 75 slit was selected to obtain a spectral resolution of ∼1125 (narrow-slit data) required to study the velocity structure of Hβ. Accuracy in spectrophotometric calibration is a key for RM studies, and therefore, we used the wide slit at a resolution of ∼170 (wide-slit data) to correct for slit losses due to the narrow slit. To satisfy this calibration requirement, Mrk 142 and a comparison star for flux calibration (hereafter, calibration star) were placed simultaneously in the same slit. We achieved this for all observations by fixing the position angle of the slits at 155°. 20 east of north such that Mrk 142 appeared at the center of the slit. The selected G-type calibration star (R.A. = 10 h 25 m 36 37, decl. = +51°38′52 18, r-band magnitude = 15.9 from the SDSS catalog) has a well-calibrated spectrum and was used for previous LJT campaigns. Flatfield images were taken for every object (science target and calibration star) with the Gemini Facility Calibration Unit (GCAL) in the sequence FLAT-OBJECT-OBJECT-FLAT with both slits. The on-target exposures were 90 s long. We also took daytime arc-lamp spectra with the CuAr lamp, again, for both slits. Binning of 1 in the spectral (X) direction and 2 in the spatial (Y) direction (1 × 2) was used for all data except for the wide-slit arc-lamp spectra, which used the binning of 1 × 1. A summary of the GMOS-North science observations is provided in Table 1. The object spectra from all epochs except the narrow-slit spectra from epoch 30 were assigned a Pass ("P") flag.

Lijiang Telescope

To complement the short observing period of 33 epochs with Gemini, we incorporated supporting observations of Mrk 142 for our study. We observed Mrk 142 with the Yunnan Faint Object Spectrograph and Camera on the Lijiang 2.4 m Telescope (LJT; Wang et al. 2019) in the two-target acquisition mode with the same calibration star as used for the Gemini observations. We followed the same observing procedure as for previous SEAMBH campaigns (e.g., Du et al. 2014Du et al. , 2015)). We obtained long-slit spectra of the target at 69 epochs from 2018 November 1 through 2019 June 21, contemporaneously with the Swift, LCO+Zowada+Liverpool, and Gemini observing campaigns. Two exposures of 1200 s each were taken for each epoch, with Grism 14 and a long slit with a projected width of 2 5. The yielded spectra cover a wavelength rage of 3800-7200 Å, with a dispersion of 1.8 Å pixel -1 . The final instrumental broadening is roughly 695 km s -1 in FWHM. Bias, dome flats, and arc-lamp spectra were taken each night for calibrations, and spectrophotometric standards were observed in several nights of good weather conditions.

Table 2 provides a summary of the overlapping photometric and spectroscopic programs.

Spectral Reduction

Gemini Spectral Reduction

The spectral reduction process for all Gemini epochs included four stages (in the order of appearance below) with the Gemini Image Reduction and Analysis Facility (Gemini IRAFfoot_2 ) reduction package: (1) baseline calibrations with GCAL flats, 2D arc-lamp spectra, and bias frames; (2) cleaning of 2D spectra followed by the wavelength calibration and extraction of 1D science and calibration-star spectra (in the same slit); (3) preparing 1D spectra for analysis with PrepSpec (see introduction to PrepSpec in Section 4.1); and (4) flux calibration of the 1D science spectra. For each epoch, we first sorted the data into lists of bias frames, GCAL flats, arc-lamp spectra, and object spectra for both the narrow and the wide slits. We then used the same reduction script with different parameter settings for processing the data taken with the two slits.

Baseline Calibrations

Baseline calibrations comprised creating a masterbias image, generating a dispersion solution with the narrow-slit arc-lamp spectra, and constructing masterflat images with both the narrow-and wide-slit flatfield images. For individual observing nights, we used bias frames with the binning of 1 × 2 and a full-frame readout from the Gemini Observatory Archive. 11 We applied an overscan noise correction to all bias images, for a given night, before combining them into a masterbias image. We then reduced the narrow-slit arc-lamp spectra with bias subtraction turned off and used them to generate 2D dispersion solutions with the task gswavelength. Generating dispersion solutions was a two-step process-fitting the 1D wavelength solution in the spectral direction and fitting any distortions in the spatial direction. The reference wavelengths for the arc-lamp spectra were used from the Gemini IRAF package. Because the re-binned, wide-slit arc-lamp spectrabinned from 1 × 1 to 1 × 2 to match the binning of the corresponding GCAL flats and object spectra-were unable to provide a nondistorted wide-slit dispersion solution, we used the narrow-slit solution to wavelength calibrate the wide-slit data. For a given epoch, we combined the two GCAL flats (including a quantum efficiency correction for each) taken with the two slits to create a masterflat corrected for the uneven illumination along the GMOS detector in the long-slit mode.

Cleaning, Wavelength Calibration, and Extraction

We corrected the 2D object spectra affected by cosmic-ray hits and performed their wavelength calibration to then extract the 1D science and calibration-star spectra. With the task gscrrej, we first selected a fixed square region surrounding the cosmic-ray affected pixels above a specified threshold and then replaced them with interpolated values from local noise levels. However, this method did not correct for all cosmic rays. We applied an additional correction to the affected pixels that remained uncorrected in the next stage of the reduction process. We applied the derived narrow-slit dispersion solutions to both the narrow-and the wide-slit object spectra.

For a given epoch, we extracted 1D science and calibrationstar spectra separately from individual exposures with the task gsextract. We selected a considerable swath of background for subtraction from both sides of each trace during the extraction process. The subtraction of bright skylines from the extracted 1D spectra resulted in some sharp spikes in the spectra owing to residual noise. We applied an additional correction to remove the sharp features in the 1D spectra in the next stage of reduction (see Section 3.1.3 for details).

A few of the extracted science and calibration-star spectra showed flat regions (zero flux values) on the shorterwavelength (or blue) end (∼3355 to ∼4325 Å) that do not match the true shape of the continuum, while some spectra

Table 1 (Continued) Epoch UT a Date MJD b Start Time Airmass (YYYY-MM-DD) 0 75 Slit 5 00 Slit 0 75 Slit 5 00 Slit 32 2019-05-28 58631.251 58631.257 1.246 1.261 58631.252 58631.258 1.250 1.265 33 2019-06-01 58635.255 58635.262 1.287 1.307 58635.257 58635.263 1.292 1.312 Notes. Observations were done with the GMOS-North Hamamatsu detector in the two-target acquisition mode (Mrk 142 and a comparison star in the same slit) positioning the slit at 155°. 20 east of north, with the B600 grating (covering the broad Hβ emission line at ∼4862 Å) and two slits, 0 75 (narrow slit) and 5 00 (wide slit). Two exposures were taken with every grating/slit combination, each 90 s long. A data quality flag (DQF) of "P" (passable) or "U" (usable) was assigned to all data at the time of observing. Unless stated otherwise, all science spectra were assigned a DQF of "P." SpectroPhotometric CALibration Flag (SPCALF) indicates whether the science spectra were calibrated ("1") or not calibrated ("0") during spectral reduction (see Section 3.1.4 for more details). SpectroPhotometric CALibration Grade (SPCALG) indicates the grade assigned to the spectrophotometric calibration based on the epoch and exposure of the calibration star spectrum used for calibrating the science spectra (see Section 3.1.4 for more details). All science spectra were assigned an SPCALF of 1 and an SPCALG of "A" unless indicated otherwise. a UT: universal Time dates. b MJD: Modified Julian Date recorded at the start of the observations for individual exposures. c Science spectrum assigned SPCALF (see Note below) = 0. d Science spectrum calibrated with the narrow-slit standard star spectrum from exposure 1 and hence assigned SPCALG (see Note below) = B. e Science spectrum likely had a calibration issue and hence was not used for further analysis (see Appendix for details). f Science spectrum assigned DQF (see Note below) = U.

showed bump-like features. The flat regions were a consequence of the slit position angle not aligned along the parallactic angle, whereas the bump-like features likely resulted from the flat-fielding process, where a higher-order spline was used to create the masterflat to appropriately trace detector sensitivity near the chip-gap regions and avoid discontinuities in the calibrated spectra near the chip edges. We corrected the spectra containing flat blue ends or bumpy features individually before attempting flux calibration (see Section 3.1.3 for details).

Additional Corrections to 1D Spectra-Preparing Data for PrepSpec

To prepare the spectra for PrepSpec, it was important that each spectrum had no gaps. Before the flux calibration stage, we trimmed the blue ends of the spectra shorter than ∼4325 Å in the rest frame because they were very noisy and not required for the purposes of this study. We further processed the 1D spectra for: (1) flat blue ends (due to the slit position angle) or bump-like features (from the flat-fielding process) appearing in some spectra; and (2) spectral regions affected by artifacts from cosmic-ray removal and sky subtraction as well as chip gaps with no flux. This additional processing was important for the initial stage of modeling spectra with PrepSpec, the software tool that corrects spectra for relative calibration differences (see Section 4.1 for details).

We developed a script to correct flat and bump-like regions in the 1D spectra in Pythonfoot_4 v3.6.5. A spline function fit to a reference spectrum modeled the true shape of the affected region. We then modeled the flux over the affected pixels assuming a Gaussian distribution of data points with standard deviation equal to the measured standard deviation at the same location in the reference spectrum. The reference spectrum used for recovery was typically the spectrum from another exposure taken on the same night (see the Appendix for exceptions).

To correct for spectral regions affected by artifacts and chip gaps, we developed another Python script to replace the regions with affected data points by local median values or interpolated and simulated data. In a given window of affected points: (1) if the number of pixels was <5, the algorithm replaced every data point by the median value of a range of 5 pixels on either side of that point with the noise equal to the local median noise; and (2) if the number of pixels was 5, the algorithm first linearly interpolated across the affected region and then replaced the interpolated points with simulated data assuming a Gaussian distribution with a standard deviation equal to twice the noise in the interpolated data. The uncertainties for the corrected pixel regions were assigned to be twice as much as the standard deviation of the unaffected individual pixel values in the region.

We used the wide-slit science and calibration-star spectra to correct for the wavelength-dependent slit losses in the narrowslit spectra with a PyRAF (IRAF with Python wrapper) script. We employed the IRAF task curfit to fit a spline function to the ratios of the narrow-slit to the reference spectra. We used a single reference spectrum: the mean of the bright, wide-slit spectra. Finally, we updated the starting pixel value of the wavelength scale in the FITS file headers of the slitloss corrected spectra to generate the appropriate wavelength grid for the trimmed spectra.

The Python scripts for performing the above corrections to prepare spectra for PrepSpec analysis are publicly available on GitHub.foot_5

Flux Calibration

The flux calibration process included two steps-fitting a sensitivity curve of the detector response to the flux standard with gsstandard, and applying the sensitivity solution to the science spectra with the task gscalibrate. For flux calibration, we used the calibration star captured in the same slit as the science target except for a handful of spectra for which we used the star from another exposure of the same epoch (see the Appendix for details). Accordingly, we assigned a SpectroPhotometric CALibration Flag (SPCALF) of 1 (0) for calibrated (noncalibrated) science spectra (stated in Table 1). Based on the epoch and exposure of the standard star spectrum used for calibration, we further assigned a SpectroPhotometric CALibration Grade (SPCALG; see Table 1) to the science spectra as follows.

1. SPCALG "A": science spectrum calibrated with the standard star spectrum from the same exposure. 2. SPCALG "B": science spectrum calibrated with the standard star spectrum from the same epoch but different exposure.

The Appendix outlines special cases of spectral reduction that were treated separately.

LJT Spectral Reduction

We first reduced the LJT spectra with IRAF, following the standard procedures for bias subtraction, flatfield correction, and wavelength calibration. The spectra of both the target and the calibration star were extracted in a uniform aperture of 8 5. For those nights with good weather conditions, the spectra of the calibration star were flux-calibrated using the spectrophotometric standards. We combined these fluxcalibrated spectra to generate a fiducial spectrum of the calibration star. Then, for each exposure, a sensitivity function was obtained by fitting the fiducial spectrum to the extracted spectrum of the comparison star. Finally, we performed flux calibration of the target spectrum (see Li et al. 2021 for more details).

Comparison between Gemini and LJT Spectra

Gemini spectra from 33 epochs and LJT spectra from 69 epochs provided 102 epochs of Mrk 142 spectral observations overlapping with the Swift and LCO+Zowada+Liverpool photometric campaigns. A mean spectrum allows us to visualize spectral features with a high signal-to-noise ratio (S/N) from the combined observations, while an rms spectrum signifies the variability in the spectral features. Figure 2 displays the mean and rms of the Gemini (top) and LJT (bottom) spectra. The higherresolution Gemini mean spectrum shows sharper emission-line profiles (Hβ, [O III], and He I) as compared to the LJT mean. At lower resolution, LJT spectra are affected by instrumental broadening, which results in the narrow emission lines, e.g., [O III], appearing broader than in the Gemini mean. The instrumental broadening effect also blurs the Fe II emission (shaded in faint blue) and the coronal lines (high-ionization forbidden transitions shaded in brown) in the LJT mean spectrum. In contrast to the Gemini mean, the Fe II features at ∼4925 Å and ∼5030 Å in the LJT mean appear blended with the Hβ wings on the longer-wavelength (red) side and [O III] λ5008, respectively. The rms of the Gemini spectra shows a noisy region blueward of 4750 Å, likely dominated by calibration noise. It is worth noting, however, that the finer wavelength sampling of the Gemini spectra (owing to the narrow-slit observations) makes that region appear even noisier. On the other hand, the region toward the blue end of the LJT rms spectrum shows clear evidence of variability in the He II λ4687 line although it is heavily contaminated with Fe II in the surrounding region. Variability in Hβ is revealed by both the Gemini and the LJT rms spectra. Although no variability in He I is evident from the Gemini rms, the LJT rms shows a weak signature of variability in broad He I. A very low, broad wave appears from ∼5250 to ∼5450 Å and from ∼5650 to ∼5950 Å in the LJT rms spectrum, likely resulting from calibration. The GMOS chip-gap region in the Gemini spectrum extends from ∼5350 to ∼5410 Å, which also appears as a low bump in the rms spectrum.

Spectral Analysis

To measure the Hβ and He I emission lines in the calibrated spectra, we first corrected any discrepancies in the calibrations of the Gemini and LJT spectra, independently with Pre-pSpec, and then modeled their spectral features with Sherpa. For PrepSpec modeling of Gemini spectra, we used the spectral region from ∼4430 to ∼6300 Å. For LJT spectra, we kept the spectral region from ∼3390 to ∼6300 Å.

PrepSpec Modeling

We independently modeled the 64 narrow-slit Gemini spectra and the 69 LJT spectra with PrepSpecfoot_6 (developer: K. Horne) to correct for any relative deviations in the calibrated wavelength and flux scales. PrepSpec models spectra by fitting the continuum and emission lines with a composite model through an iterative process. We included the following model components for fitting the Mrk 142 spectra: (1) [A] verage spectrum (specified by "A")-mean of the input spectra;

(2) [C]ontinuum-variations in the continuum emission from the accretion disk modeled as a polynomial defined by l log with time-dependent coefficients; (3) [W]avelength jitterinter-spectra shifts in the wavelength scales; (4) [F]lux jittertime-dependent photometric corrections to minimize the scatter of narrow emission-line fluxes relative to their median; and (4) [B]road-line variations-variability in the broad emission-line features. Modeling emission lines in PrepSpec takes into account the velocity window half-widths of the broad as well as the narrow lines, whose initial values were set to 3000 km s -1 and 500 km s -1 , respectively. We set the broad Hβ λ4861 and He I λ5877 as variable lines for Gemini spectra, and Hγ λ4342, He II λ4687, Hβ, and He I λ5877 as variable for LJT spectra. The software uses the I Zwicky 1 (I Zw 1) template model (Véron-Cetty et al. 2001) to fit Fe II emission in the mean spectrum. PrepSpec is not designed to handle gaps in spectra or extremely large flux values, e.g., from cosmic-ray hits. Therefore, chip gaps and artifacts from cosmic-ray correction or sky subtraction in the Gemini spectra were replaced by median or simulated data (see Section 3.1.3 for details) during spectral reduction.

In the PrepSpec modeling stage, we first corrected the Gemini and LJT spectra for pixel shifts relative to the [O III] λ5008 line and then modeled the spectra with a composite model. We observed small pixel shifts (<6 pixels) while aligning the spectra along the wavelength axis. The model components were jointly fit starting with a single component and then adding components up to the ACWFB composite model for both the Gemini and the LJT spectra. PrepSpec determines the bestfitting model by accessing the Bayesian information criterion and reduced χ 2 (c n 2 , where ν stands for degrees of freedom) statistics. The goal of the fitting process is to use the fewest possible parameters to describe the data while penalizing the model for the number of parameters used. A good model yields c ñ 1 2 . Figure 3 displays the final model (dark blue curve) passing through the black mean spectrum (panel (a)) and the model (dark gray curve) to the residual rms (rmsx) spectrum (panel (b)) for the 64 narrow-slit Gemini spectra. The rms spectrum shows that the spectra are noisier at the bluer end.

Figure 4 shows the final model (panel (a)) along with the residuals (in the units of σ; panel (b)) in grayscale for the Gemini spectra. The best-fit model yielded a c n 2 value of 0.782, which indicates overfitting of the data, possibly indicating inaccurate error bars larger than the scatter in the data. The dark regions in the model highlight the prominent emission-line features of Hβ λ4861 and [O III] λλ4960, 5008. The weak fluctuations blueward of ∼4700 Å indicate more noise in that region as compared to the red end of the spectra. The residuals in grayscale display horizontal wiggles that are strongly evident in some spectra. We noted that the wiggles appear in the spectral regions replaced by simulated data to correct for residual features either from cosmic-ray correction or sky subtraction. The replacement with simulated data may have resulted in a lower performance of the model in those regions. Another probable reason for the wiggles is the use of a higher-order spline during flat-fielding in the spectral reduction process (refer Section 3.1.3 for details). ) and LJT (bottom; velocity resolution of 695.2 ± 3.9 km s -1 ) spectra highlighting the Regions of Interest-Hβ λ4861 and [O III] λλ4960, 5008 (region shaded in green), and He I λ5877 (region shaded in pink). Labels enclosed in dashed (dotted) boxes indicate broad (narrow) lines. The Gemini mean spectrum shows sharp Fe II features (shaded in faint blue), which appear blended with the red wings of Hβ and [O III] λ5008 (blue shaded bars) in the LJT mean. Owing to the high signal-to-noise of the Gemini spectra, the peculiar shape of the He I λ5877 line is clearly evident. The high-ionization coronal lines (shaded in brown) also appear sharp in contrast to the LJT spectrum, as a result. The LJT rms spectrum shows clear variability in He II (shaded in blue). Both rms spectra indicate variability in the broad Hβ. However, no significant variability is evident in He I over the timescale of Gemini+LJT observations. The yellow-shaded region in the Gemini spectra indicates the GMOS chip gap from ∼5350 to ∼5410 Å. (The reduced and calibrated Mrk 142 Gemini spectra are available in a machine-readable form.) (The data used to create this figure are available.) However, we visually inspected all spectra processed through PrepSpec and observed no anomalous behavior in the regions with wiggles. Therefore, the spectra were considered valid for further analysis.

PrepSpec modeling of LJT spectra yielded nearly even residuals with a c n 2 value of 0.791. The region redward of 6300 Å in LJT spectra comprises several blended narrow-line features, which resulted in a suboptimal performance of the PrepSpec model. Therefore, we excluded the red side of the LJT spectra (λ > 6300 Å) during PrepSpec processing.

Spectral Modeling in Sherpa

We modeled the continuum and emission lines in the Gemini and LJT spectra in Sherpafoot_7 (Freeman et al. 2001;Burke et al. 2018) v4.10.0 with a Python wrapper script. We first corrected the Gemini and LJT spectra for Galactic reddening using E(B -V ) = 0.015 (Schlafly & Finkbeiner 2011). Averaging the two narrow-slit Gemini exposures from every night into a single spectrum per epoch (with exceptions for spectra from epochs 11, 25, and 28, where we only used single exposures) yielded a total of 33 Gemini spectra. Together with the 69 LJT spectra, we modeled a total of 102 Mrk 142 spectra.

We developed a composite model with the goal of performing a clean extraction of the Hβ and He I emission lines from the Gemini and LJT spectra. We included a powerlaw fit to the continuum, three Gaussians to model each of the Hβ, He I, and He II emission lines, and a single Gaussian for each of the [O III] doublet lines. We adopted the I Zw 1 template model from Boroson & Green (1992) as a pseudocontinuum to trace the Fe II emission-line features. We also experimented with the Fe II template from Véron-Cetty et al. (2001). However, it failed to suitably trace the sharp Fe II features in Mrk 142. With the Boroson & Green (1992) Fe II template model, the fits yielded lower (c n 2 ) values than with the Véron-Cetty et al. (2001) template. Following the procedure in Hu et al. (2015), we added single Gaussian profiles for each of the six coronal lines (Fe VII λ5160, Fe VI λ5177, Ca V λ5311, Fe VII λ5278, Fe VII λ5722, and Fe VII λ6088; see Figure 2). In addition, we included the host-galaxy template with 11 Gyr at z = 0.05 from the 2013 updated version of Bruzual & Charlot (2003) galaxy templates. The host-galaxy template, affecting the redder part of the spectrum more than the bluer, contributed greatly in producing a good fit to the He I emission-line region. The fit in the Hβ region was less sensitive to host-galaxy emission. We referred to the Vanden Berk et al. (2001) rest wavelengths for setting the positions of all emission lines.

Gemini Spectral Analysis

Our goal of spectral fitting was to accurately estimate the Hβ, [O III], and He I profiles in the Gemini spectra. We aimed at finding a robust and flexible set of parameters that fit the structure in the spectra over all epochs. Figure 5 shows the composite model fit to a single-epoch Gemini spectrum.

We describe the fitting process as follows. The Gaussian used for each of the 3 lists the emission-line parameters along with their settings as used during spectral fitting.

We followed the Fe II template fitting procedure described in Hu et al. (2015), where the Fe II emission is defined by a convolution of the Boroson & Green (1992) template with a Gaussian. We applied the Gaussian as a 1D point-spread 1.183 2 Modeling individual coronal lines in the spectra considerably improved the fit in the Fe II emission region from ∼5150 to ∼5350 Å. In this region, the Fe coronal lines appeared to be slightly redshifted (0.003) with respect to their rest wavelengths. We set the coronal-line widths to 1.5 times the [O III] λ5008 line width and their flux values to specific fractions of the [O III] λ5008 flux, as determined from the fit to the Gemini mean spectrum.

From spectral modeling, we derived, at each epoch, the total FWHM and flux values of the Hβ, He I, and [O III] lines. We measured the FWHM of the broad and total (including both the broad and the narrow components) Hβ and He I lines empirically by subtracting all other model components from the spectra including the narrow lines. To calculate the contribution from the broad-line and total (again, including both the broad-and the narrow-line) flux in the Hβ and He I emission profiles, we simply added the contribution from each of their components. Tables 4-6 provide emission-line measurements for the Gemini spectra from 33 epochs. For epoch 11, the model failed to constrain the broad Hβ emission as the region blueward of the Hβ line appeared noisier compared to the other epochs. We therefore excluded epoch 11 from further analysis. Also, due to improper flux calibration at the location of the Hβ line in epoch 25, the line appeared unusually broader and brighter than at the other epochs. We therefore excluded the spectrum from epoch 25 as well.

LJT Spectral Analysis

We fit the 69 LJT spectra with the same goal of modeling the Hβ, [O III], and He I lines accurately. Figure 6 shows the fit to a single-epoch LJT spectrum.

We adopted the same model as for the Gemini spectra with small modifications to the ratios of certain fixed parameters. We determined the flux ratios for the coronal lines relative to the [O III] λ5008 line flux from the fit to the mean LJT spectrum. The width of the fixed broad Hβ Gaussian was fixed at a factor of 2.5 (instead of 2 for the Gemini spectra). A factor of 2 for this second Hβ Gaussian in LJT spectra was insufficient to trace the broad wing of Hβ, which also affected the fit to the blended Fe II feature at ∼4923 Å. Therefore, a broader Hβ component was required to generate a good fit in that region. This indicates an interplay between the broad Hβ and Fe II line emission in the fitting process. Similarly, the [O III] λ5008 appears to be blended with the Fe II emission feature at its red wing. This is caused by the instrumental broadening in LJT spectra that further resulted in wider 3). Tables 7-9 provide emission-line measurements for the LJT spectra from 69 epochs.

Light-curve Analysis

We used the Mrk 142 Gemini and LJT spectral measurements to generate light curves for the broad Hβ and He I emission-line profiles. We obtained the total broad-line light curves by integrating the flux under the two broad components (see green dashed Gaussians for Hβ and pink dashed Gaussians for He I in Figures 5 and 6) for the two emission lines. We scaled the broad Hβ light curve from Gemini to the broad Hβ light curve from LJT to generate an inter-calibrated light curve. The Hβ light curves from Gemini and LJT were offset by ∼25% from each other although they displayed similar fluctuations in their patterns. The offset can be attributed to various factors-different seeing conditions at Gemini and LJT or even the difference in the calibrations from the two telescopes. Because we are interested in measuring a time shift in the pattern with reference to the continuum variations, scaling and combining the light curves is valid for our purpose. Figure 7 shows the scaled broad Hβ light curve plotted with the original Gemini and LJT light curves. We then inter-calibrated the original Hβ light curve from LJT and the scaled Hβ light curve from Gemini with PyROA (see Section 5.1) to use the combined light curve to determine the time lag between the continuum and emission-line variability.

Cross-correlation Time Lags

We cross-correlated the broad Hβ Gemini+LJT intercalibrated light curve with the UVW2 light curve from Swift to measure the reverberation lag of the BLR response to the continuum variability from the accretion disk, which is at smaller size scales than the BLR. Following Cackett et al. (2020), we chose the UVW2 because we aim to measure the response of Hβ line-emitting gas to the UV continuum and the UVW2 was the shortest wavelength available from the photometric monitoring of Mrk 142. During cross-correlation, we also included the LCO+Zowada+Liverpool/g, the intercalibrated 5100 Å continuum from Gemini and LJT data, and the LJT broad Hβ light curves to use maximum available information for a reliable measurement of the variability pattern.

We employed Python-based running optimal average (PyROA; Donnan et al. 2021) to calculate cross-correlation time lags. PyROA 16 uses a running optimal average (ROA) calculated with a window function (defined by a Gaussian by default) of a certain width to estimate light-curve behavior while fitting all input light curves simultaneously. The width of the window function controls the flexibility of the model in deriving the driving light curve-a narrower window function 8.14 0.80 0.82 days for LJT/Hβ) with reference to the Swift/UVW2 band. With respect to the shorter-wavelength UVW2 emission, we expect to measure positive lags for the longer-wavelength emission in the g band, at 5100 Å, and for the Hβ emission line. We thus modeled the distribution of time lags as a log-Gaussian function that imposes positive lags with reference to UVW2, whose lag is fixed at 0.00 day (see Figure 8). In addition to measuring the time shift in the lightcurve pattern, the width of the Log-Gaussian model also accounts for the amount of blurring applied to the reference light curve (here, UVW2) to match the response in the echo light curves. This becomes important for BLR RM, where the emission-line variations, emerging farther away from the central engine and from a more spatially extended structure (size scale ∼1 pc) than the accretion disk, are smoother compared to the continuum variations closer to the center. In PyROA, the width of the time-lag distribution quantifies the blurring determined for each of the echo light curves:

-+ 0.28 0.18 0.19 days for LCO+Zowada+Liverpool/g, -+ 0.88 0.48 0.62 days for Gemini+LJT/5100 Å, -+ 4.88 0.90 1.16 days for LJT/Hβ), and -+

5.47 0.89 1.06 days for Gemini+LJT/Hβ. We also performed light-curve analysis with the interpolated cross-correlation

He I Light Curves

The peculiar, asymmetrical shape of the He I line-narrowline emission and a broad, asymmetrical component (modeled by a Gaussian six times the width of the broad Hβ line in our spectra fitting procedure)-is clearly evident in the high-S/N Gemini mean spectrum. The asymmetry in the broad component due to the stronger blueshifted emission feature likely indicates a wind component in the BLR. However, the cause of such a disk-wind component is not clear. Leighly (2004) performed CLOUDY simulations to model 10 high-and lowionization emission lines observed in NLS1s. She suggested that the blueshifted emission evident in the high-ionization lines in NLS1s arises from a wind moving toward us.

Interestingly, the plausible broad, blueshifted component for He II in the spectral model may also be a result of such a wind emission. Further analysis of the He II line is needed to draw firm inferences in this regard. In addition to the blueshifted wind, Leighly (2004) found that the high-ionization Lyα was dominated by emission in the accretion-disk atmosphere or at the low-velocity base of the broad-line wind. The very broad, flattened emission feature in He I may be indicative of a diskwind feature as noted for Lyα. Furthermore, Leighly (2004) derived a small covering fraction for the BLR. She argued that in an object with a small black hole mass, as in the case of Mrk 142 (M • = 3.89 × 10 6 M e 17

), a small covering fraction can result from an emission-line region closer to the plane of the disk. Li et al. (2018) performed velocity-resolved RM of Mrk 142, where the authors concluded that the two-zone BLR model (Wang et al. 2014b) best fit the Mrk 142 BLR with an opening angle (θ) of 10°-30°(representing a disk-like BLR). Estimating the covering fraction from the opening angle as Notes. The second and the third columns providing the FWHM and flux values, respectively, for the broad ("b") He I component include contributions from both the broad Gaussians defined for the line. The FWHM of the narrow ("n") He I is equal to the FWHM of the [O III] λ5008 (see Table 4, third column). The total ("t") FWHM and flux include contributions from both the broad and the narrow components

a Excluded from further analysis. See corresponding note in Table 5. b Excluded from further analysis. See corresponding note in Table 5.

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

17 1 M e = 1 Solar mass.

θ/90°yields a value of 0.1-0.3 for the covering fraction; however, these small values do not confirm the presence of a disk wind in Mrk 142. To understand the components that form the atypical Mrk 142 BLR system or the kind of BLR geometry in a super-Eddington that can show a broad, blueshifted emission feature similar to the He I line feature, we would need further investigation of BLR models and data for super-Eddington AGNs.

Although the LJT rms spectrum of Mrk 142 shows a weak variable feature for He I, the variability is not usefully quantifiable given the timescale and S/N of our current Gemini+LJT spectroscopic campaigns. Figure 9 displays the individual and total broad-line components of He I emission from both the Gemini and the LJT observations. The offset observed in the narrower broad component is similar to that observed in the broad Hβ, where the light curve from Gemini appears at higher flux values than the LJT light curve. Interestingly, the broader broad component is brighter in LJT than in Gemini likely resulting from the blueshifted disk-wind component broadened due to the wider slit used for LJT data.

Results and Discussion

We present the first results on the lag of the broad Hβ line with respect to the UV continuum in Mrk 142 from optical spectroscopic observations from Gemini+LJT with simultaneous monitoring in the photometric Swift/UVW2 and LCO +Zowada+Liverpool/g bands.

We applied a spectral model with same number of components but different parameter settings to the Gemini and LJT spectra to derive Hβ and He I light curves. We noted that a profile composed of one narrow + two broad Gaussians sufficiently traced both the Hβ and the He I lines, where the widths and positions of the narrow components were tied to those of the 1.711 2 . The smaller residuals indicate an overall good fit to the spectrum. The model performance drops significantly at both ends of the spectrum although it does not impact measurements in the Regions of Interest. Eddington AGN, where self-shadowing effects due to a slimdisk structure in super-Eddington AGNs could result in the broad-line profiles appearing more Lorentzian than Gaussian. We recognize that our Mrk 142 spectral fitting model does not align with that of Wang et al. (2014b); however, our results are not sufficient evidence to rule out the self-shadowing hypothesis. A detailed analysis of spectral line profiles for a larger sample of super-Eddington AGNs is required to build a robust understanding of how the accretion-disk behavior affects the observed broad-line profiles in these objects. We employed fixed narrow-line flux ratios F Hβ /F [O III] λ5008 and F He I /F [O III] λ5008 for the LJT spectral model (determined from flexible flux ratios for the Gemini spectra) due to the instrumental broadening in the LJT spectra affecting emission-line measurements. Although the LJT rms spectrum shows a weak feature at the He I location, PrepSpec modeling and light-curve analysis suggested that there is no adequate variability in the line that is measurable with the current Gemini+LJT data. However, we acknowledge that the He I emission line shows a peculiar profile evident from the high S/N of the Gemini spectra. The broader Gaussian used for the line indicates stronger blueshifted emission than the redshifted side of the line. Also, the He II line specifically required a broad, blueshifted component to accurately trace the emission in that region. Followed by spectral analysis, we empirically measured the FWHM values as well as calculated the narrowand broad-line flux values of the Hβ and He I lines to obtain their light curves.

Applying PyROA, we performed cross-correlation analysis with continuum (Swift/UVW2, LCO+Zowada+Liverpool/g, and Gemini+LJT/5100 Å) and broad Hβ light curves (LJT and Gemini+LJT inter-calibrated) with a goal of determining reverberation time lag for the Gemini+LJT inter-calibrated Figure 8. Time-lag measurements with reference to the Swift/UVW2 band (top row) with lag distributions modeled as log-Gaussians. Left: the top three panels show continuum light curves-Swift/UVW2 (blue), LCO+Zowada+Liverpool/g (green), and inter-calibrated Gemini+LJT/5100 Å-in units of 10 -15 erg s -1 cm -2 Å -1 . The bottom two panels show the broad Hβ light curves-LJT only (pink) and Gemini+LJT inter-calibrated (olive)-in units of 10 -14 erg s -1 cm -2 Å -1 . Right: timelag distributions (colored histograms for the mean lag and gray shaded uncertainty estimates from a log-Gaussian lag distribution) of the light curves on the left with reference to the Swift/UVW2 band, which has a fixed lag of 0.00 day. Insets in the second and the third panels show a close view of the time-lag distributions for LCO +Zowada+Liverpool/g and Gemini+LJT/5100 Å light curves, respectively. For each of the echo light curves, the black solid vertical line marks the lag measurement, given by the median lag of the colored distribution and the black dashed vertical lines (on either sides of the solid line) mark the corresponding uncertainty. It is interesting to note that the broader He I light curve from the LJT spectra is brighter than the broader He I component from the Gemini spectra likely due to the blue wing of He I (traced primarily by the broader component) in the LJT data affected by instrumental broadening. Hβ light curve. PyROA provided an improvement in quantifying the uncertainties compared to previous studies. Most early RM studies have extensively applied the ICCF method for time-lag measurements, which makes it a good comparison standard. However, ICCF struggles with nonuniformly sampled data, which is true for our Gemini+LJT campaigns similar to most other studies, and uses linear interpolation to estimate the light-curve behavior in the regions with data gaps. Consequently, the uncertainties reported for ICCF-based measurements are typically conservative compared to JAVE-LIN and PyROA, as noted in this work for the 5100 Å continuum as well as the Hβ emission from LJT and Gemini +LJT inter-calibrated data (see Table 10). In the context of the uncertainties on time-lag measurements, JAVELIN has been shown to perform better. For instance, Edelson et al. (2019) reported uncertainties from ICCF to be twice as large as those from JAVELIN, which is also evident from the results in this work (see Table 10). JAVELIN uses damped random walk (DRW) to estimate the light-curve pattern in the regions where data are not available. DRW closely characterizes the variability observed in AGNs, plausibly leading to smaller uncertainties in the final lag measurements. However, JAVE-LIN requires a good estimation of uncertainties in data. For suboptimally calibrated uncertainties, JAVELIN can sometimes fail to deliver reliable lag measurements (Donnan et al. 2021). This is likely the reason the LJT/Hβ and Gemini+LJT/ Hβ emission-line light curves, which have larger calibrated uncertainties than the continuum light curves, show time-lag measurements differing from those reported by PyROA.

PyROA offers an improvement over JAVELIN-the ROA along with a robust error treatment not only prevents the outlier points from disrupting the estimation of the driving light curve but also applies a valid algorithm for resolving data gaps.

We measured a time lag of -+

8.68 0.72 0.75 days for the Gemini +LJT inter-calibrated Hβ emission with reference to the UVW2 continuum. We also obtained a lag of -+ 0.79 0.29 0.27 days for the 5100 Å continuum with reference to the UVW2 band (that is consistent, within uncertainties, with the time lag-wavelength relationship in Cackett et al. 2020), and a lag of 7.89 ± 0.80 days for the Hβ emission with respect to the 5100 Å continuum. From here, we report a black hole mass of

= 6.28 ± 0.29 derived using Equation (1),

where we used a ( )

0.36 0.54 0.33 from the dynamical modeling of the Mrk 142 BLR by Li et al. (2018). It is possible that the value of ( ) f log has a large uncertainty, which depends on the BLR's dynamical model or the calibration using the M-σ * relation for classical bulges and pseudobulges (e.g., Ho & Kim 2014;Li et al. 2018;Yu et al. 2020). For V FWHM , we used the mean Hβ FWHM of 1680 ± 14 km s -1 from the Gemini spectra. Here, we chose the Gemini spectra due to their higher resolution providing a more reliable measurement of the narrower Hβ broad-line profile than the LJT spectra. We also considered the mean FWHM from the spectra as against the Hβ rms profiles as the rms spectra from Gemini were noisier blueward of the blue wing of the Hβ line. We further measured mean luminosities of

( )=  L log 43.832 0.001 UVW 2 , ( ) =  L log 43.643 0.002 5100 , and ( ) =  b L log 41.621 0.002 H

. For the L 5100 and L Hβ measurements, we adopted the mean flux value from our Gemini+LJT inter-calibrated light curves as their respective Gemini light curves alone were insufficient to provide a reliable flux scale due to the shorter observing timescale.

Our results agree with previously published measurements of Mrk 142 (Du et al. 2015;Li et al. 2018). From the previous 6 month SEAMBH campaign, Du et al. (2015) reported a time lag of -+ 7.9 1.1 1.2 days for Hβ with reference to 5100 Å. We measured an optical lag of 7.89 ± 0.80 days for Hβ in agreement with the Du et al. (2015) value within uncertainties. Furthermore, the derived black hole mass for Mrk 142 in this work,  ( ) M M log • = 6.28 ± 0.29, agrees with the value reported in the recent velocity-resolved RM analysis by Li et al. (2018), -+ 6.23 0.45 0.26 , within uncertainty limits. Table 11 summarizes the measured quantities in this work and shows their comparison with the values from previous studies.

To visualize our results in the broader context of reverberation-mapped AGNs, we placed the measured size of the Hβ line-emitting region on various R-L scaling relations. Comparisons are discussed below. , where ν indicates 1071 degrees of freedom, gives the model statistic for individual epochs. b This spectrum appeared very noisy likely due to some disturbance in the field of view at the time of observation. Therefore, we excluded this epoch from further analysis.

(This table is available in machine-readable form.) time lag (along the vertical axis) relative to the standard deviation of the departures of the typical RM sample from both best-fit relations (s departure ). While the departure of Mrk 142 from the Bentz et al. (2013) relation (15.7 lt-day) is s 1.15 departure (where s = 13.7 departure lt-day), its departure from the Kaspi et al. (2005) relation (4.08 lt-day) is =1s departure (where s = 26.8 departure lt-day). The latter implies that R Hβ -L Hβ is a tighter relationship for AGNs than R Hβ -L 5100 . The position of Mrk 142 in the R Hβ -L 5100 relationship (panel (a)), considerably below the best-fit line, reiterates the characteristic of super-Eddington AGNs exhibiting smaller BLR sizes in contrast to the sub-Eddington population at the same luminosities (Du et al. 2016b). Du et al. (2015) tested this deviation of high accretion-rate AGNs from the R Hβ -L 5100 relationship. Studying the differences in the BLR sizes for AGNs with low (   < M M 3 Edd ) and high (    M M 3 Edd

) mass-accretion rates, Du et al. (2015) inferred that  M influences the size scales observed in super-Eddington AGNs, while such a correlation is absent in the low mass-accretion-rate objects.

Figure 11 shows the R Hβ -L 5100 and R Hβ -L 1350 scaling relations for NGC 5548 over time with optical lag measurements for Hβ, and luminosities at 5100 and 1350 Å (L 1350 ) from Eser et al. (2015). Again, the red star in both panels represents the Mrk 142 point from this work. Eser et al. (2015) formulated a conversion from L 5100 to L 1350 for NGC 5548 (see their Equation (4)) from all RM campaigns of the object from 1988-2008(Peterson et al. 2002;;Bentz et al. 2007Bentz et al. , 2009b;;Denney et al. 2010). We applied that conversion to calculate L 1350 for NGC 5548 and generated the R Hβ -L 1350 plot (Figure 11 Table 8 (Continued) Epoch FWHM Hβ,b F Hβ,b FWHM Hβ,n F Hβ,n FWHM Hβ,t F Hβ,t (km s -1 ) ( 10 -15 erg s -1 cm -2 ) ( km s -1 ) ( 10 -15 erg s -1 cm -2 ) ( km s -1 ) ( 10 -15 erg s -1 cm -2 ) 64 1852 ± 36 77.7 ± 1.7 660 ± 11 4.82 ± 0.07 1653 ± 37 82.5 ± 1.7 65 1893 ± 33 81.0 ± 1.1 727 ± 12 5.11 ± 0.07 1682 ± 46 86.2 ± 1.1 66 1930 ± 47 82.6 ± 1.5 681 ± 12 5.01 ± 0.09 1706 ± 32 87.6 ± 1.5 67 1877 ± 35 83.7 ± 2.2 709 ± 14 4.98 ± 0.09 1695 ± 56 88.7 ± 2.2 68 1838 ± 47 81.5 ± 2.4 756 ± 3 5.11 ± 0.09 1651 ± 43 86.6 ± 2.4 69 1893 ± 67 81.3 ± 3.0 752 ± 25 5.05 ± 0.11 1696 ± 35 86.3 ± 3.0

Notes. The second and the third columns providing the FWHM and flux values, respectively, for the broad ("b") Hβ component include contributions from both the broad Gaussians defined for the line. The FWHM of the narrow ("n") Hβ is equal to the FWHM of the [O III] λ5008 (see Table 7, third column). The total ("t") FWHM and flux include contributions from both the broad and the narrow components (t = b + n).

a Excluded from further analysis. See note a in Table 7.

(This table is available in machine-readable form.) (2015). The SEAMBH objects (black squares; Du et al. 2014Du et al. , 2015;;Wang et al. 2014a;Hu et al. 2015;Li et al. 2018Li et al. , 2021;;Zhang et al. 2019) appear on the lower right of the gray solid lines, which represent the best-fit R-L relations from Bentz et al. (2013) with L 5100 (panel (a)) and Kaspi et al. (2005) with L Hβ (panel (b)), indicating a smaller size for the BLR in highly accreting AGNs compared to the more typical, sub-Eddington AGNs mapped in previous studies (green circles, panel (a) and gray circles, panel (b); Stirpe et al. 1994;Santos-Lleó et al. 1997;Collier et al. 1998;Dietrich et al. 1998Dietrich et al. , 2012;;Peterson et al. 1998Peterson et al. , 2002Peterson et al. , 2014;;Kaspi et al. 2000Kaspi et al. , 2005;;Santos-Lleó et al. 2001;Bentz et al. 2006Bentz et al. , 2007Bentz et al. , 2009aBentz et al. , 2009bBentz et al. , 2013Bentz et al. , 2014;;Collin et al. 2006;Denney et al. 2006Denney et al. , 2010;;Grier et al. 2012;Barth et al. 2013;Pei et al. 2014 (2022) obtained an α ox of -1.30 ± 0.04 from simultaneous X-ray/UV/optical observations. A less negative α ox for NGC 5548 compared to Mrk 142 suggests a harder ionizing SED for the normal Seyfert 1 galaxy, likely due to the soft X-ray excess observed in the object (e.g., Mehdipour et al. 2015), than the super-Eddington AGNs. However, the soft excess in NGC 5548 was not evident in 2013, when an obscurer heavily absorbed its soft X-ray flux (Mehdipour et al. 2015).

Most Seyferts/quasars have some sort of soft X-ray excess; the debate for years has been over the origin (ionized disk reflection, warm Comptonized emission, a mixture, or both).

The implication from Tortosa et al. (2023) is that for super-Eddington AGNs, ionized disk reflection can model the soft excess in this class of AGNs well. It is interesting to witness that despite their distinct types, the SED shapes of Mrk 142 and NGC 5547 are similar, as their α ox values agree within uncertainties-a plausible explanation for the tighter R Hβ -L 1350 correlation than R Hβ -L 5100 noted in both the objects. Future accretion-disk modeling efforts can help understanding such comparisons of R-L relationships between normal and super-Eddington AGNs.

If UV emission is closer to the driving continuum as seen in Figure 11 Notes. The second and the third columns providing the FWHM and flux values, respectively, for the broad ("b") He I component include contributions from both the broad Gaussians defined for the line. The FWHM of the narrow ("n") He I is equal to the FWHM of the [O III] λ5008 (see Table 7, third column). The total ("t") FWHM and flux include contributions from both the broad and the narrow components (t = b + n).

a Excluded from further analysis. See note a in Table 7.

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

Table 10 Time-lag Measurements Time Lag (days) PyROA a ICCF JAVELIN UVW2-to-g -+ 0.73 0.10 0.10 -+ 0.7 0.2 0.2 -+ 0.54 0.08 0.08 UVW2-to-5100 Å -+ 0.79 0.29 0.27 -+ 1.7 1.2 1.8 -+ 0.78 0.38 0.42 UVW2-to-Hβ, LJT -+ 8.14 0.80 0.82 -+ 8.8 3.5 2.5 -+ 6.95 0.46 0.69 UVW2-to-Hβ, Gemini+LJT -+ 8.68 0.72 0.75 -+ 8.7 9.1 4.2 -+ 11.38 4.49 0.51

Note.

a We consider these as the most robust time-lag measurements of the three methods. See Section 6 for further details. Mrk 142 with data from Swift, LCO, Zowada, Liverpool, and other ground-based observatories, simultaneous to the Gemini +LJT data taken as a part of the same broader RM campaign. Cackett et al. (2020) pointed that if the UVW2 band represents the driving continuum, then the black hole mass derived from the Hβ optical lag is underestimated by ∼10%. We used the UVW2 to Hβ time lag result from this work to calculate the black hole mass in Mrk 142. We obtained a mass of  ( ) M M log • = 6.32 ± 0.29 based on the UVW2 to Hβ time lag. This value is ∼10% greater than the black hole mass derived by assuming 5100 Å as the driving continuum band, and thus verifies the discrepancy estimated by Cackett et al. (2020). This discrepancy would be as high as ∼40% if X-rays, instead of UV, were the driving continuum (Cackett et al. 2020). However, our work does not propose any new implications to the accretion-disk structure of Mrk 142. A robust Hβ lag measurement with reference to the X-ray continuum from future studies will help understand how X-rays play a role in driving the continuum variability in Mrk 142 and shape its inner accretion disk.

Conclusion

We performed BLR RM analysis of Mrk 142 with mediumand low-resolution optical spectra from Gemini and LJT, simultaneous to the Swift and LCO+Zowada+Liverpool photometric campaigns reported by Cackett et al. (2020) to measure the UV lag for Hβ emission line. With PrepSpecanalysis, we corrected calibration discrepancies for both the Gemini and the LJT spectra individually. From spectral modeling in Sherpa, we measured FWHM and fluxes for [O III] λλ4960, 5008; Hβ λ4861; and He I λ5877 emission lines. To combine the 5100 Å and Hβ light curves from Gemini and LJT, we inter-calibrated the respective light curves from the two telescopes in PyROA. Applying PyROAfor time-lag analysis, we measured a UV time lag for Hβ and further derived refined black hole masses. Placing our results on various R-L scaling relations, we verified that our results are consistent with previously published values for Mrk 142. We summarize our main findings below.

1. PyROA, using the Bayesian information criterion to evaluate model performance along with a rigorous treatment of uncertainties, provided a robust method for measuring cross-correlation time lags. This project is one of the early works employing the PyROA technique for measuring RM time lags with real data. In this process, the longer timescale of LJT spectra nicely complemented the gaps in the Gemini observations. 2. We measured, for the first time, a UV time lag of -+ 8.68 0.72 0.75 days for Hβ in Mrk 142, with simultaneous photometry in the Swift/UVW2 band and optical spectroscopy with Gemini and LJT. Assuming the UV continuum as the primary driver of the observed variability, we derived a black hole mass of  ( ) M M log • = 6.32 ± 0.29. 3. We obtained a 5100 Å to Hβ time lag of 7.89 ± 0.80 days, consistent with the measured value from previous SEAMBH campaigns (Du et al. 2015). From this lag measurement, we also derived a black hole mass for Mrk 142 of  ( ) M M log • = 6.28 ± 0.29, in agreement with the mass reported by Li et al. (2018). 4. We placed the 5100 Å to Hβ time lag with measured L 1350 on the R Hβ -L 1350 relation for NGC 5548 (Eser et al. 2015). Mrk 142 falls closer to the R Hβ -L 1350 scaling relation than the R Hβ -L 5100 relation indicating that the UV is closer to the "true" driving continuum as opposed to the 5100 Å band.

In addition, we also recorded supplementary results. Our spectral analysis indicated blueshifted, broad components for the He I and He II emission lines suggestive of wind components in these higher-ionization lines. To infer the cause of such disk+wind components, we need more higherresolution data and BLR modeling efforts for super-Eddington AGNs. We intend to study the He I and He II lines in further detail in future work. Furthermore, BLR RM analysis with the concurrent X-ray data available from Swift can better inform our understanding of the measured Hβ time lags with respect to the UV continuum. We aim to explore this in future study.