SDSS J075101.42+291419.1: A Super-Eddington Accreting Quasar with Extreme X-ray Variability

TL;DR

This paper reports the discovery of extreme X-ray variability in a super-Eddington accreting quasar, SDSS J075101.42+291419.1, highlighting its unique spectral and variability properties and exploring possible physical explanations.

Contribution

It presents the first detailed analysis of extreme X-ray variability in a super-Eddington quasar, proposing physical scenarios like disk reflection and partial covering absorption.

Findings

X-ray flux dropped by a factor of ~22 between observations.

The quasar shows steep X-ray spectra with photon index ~3.

Estimated 15-24% of super-Eddington AGNs exhibit such extreme variability.

Abstract

We report the discovery of extreme X-ray variability in a type 1 quasar: SDSS J$075101.42+291419.1$. It has a black hole mass of $1.6\times 10^7~\rm M_\odot$ measured from reverberation mapping (RM), and the black hole is accreting with a super-Eddington accretion rate. Its XMM-Newton observation in 2015 May reveals a flux drop by a factor of $\sim 22$ with respect to the Swift observation in 2013 May when it showed a typical level of X-ray emission relative to its UV/optical emission. The lack of correlated UV variability results in a steep X-ray-to-optical power-law slope ($\alpha_{\rm OX}$) of -1.97 in the low X-ray flux state, corresponding to an X-ray weakness factor of 36.2 at rest-frame 2 keV relative to its UV/optical luminosity. The mild UV/optical continuum and emission-line variability also suggest that the accretion rate did not change significantly. A single power-law model modified by Galactic absorption describes well the $0.3-10$ keV spectra of the X-ray observations in general. The spectral fitting reveals steep spectral shapes with $\Gamma\approx3$. We search for active galactic nuclei (AGNs) with such extreme X-ray variability in the literature and find that most of them are narrow-line Seyfert 1 galaxies and quasars with high accretion rates. The fraction of extremely X-ray variable objects among super-Eddington accreting AGNs is estimated to be $\approx 15-24\%$. We discuss two possible scenarios, disk reflection and partial covering absorption, to explain the extreme X-ray variability of SDSS J$075101.42+291419.1$. We propose a possible origin for the partial covering absorber, which is the thick inner accretion disk and its associated outflow in AGNs with high accretion rates.

Tables

Table 1. Table 1 Observation Log

		Observation	Exposure Time	Bandpass or
Observatory and Instrument	Date	ID	(s)	Effective Wavelength
Swift XRT (PC mode)	2013–05–27	00039549001	4110	0.3–10 keV
Swift UVOT (UVW1)	2013–05–27	00039549001	4065	2600 Å
Swift UVOT (UVW2)	2014–01–25	00039550001	101	1928 Å
Swift XRT (PC mode)	2014–09–04	00039550002	2015	0.3–10 keV
Swift UVOT (U)	2014–09–04	00039550002	2106	3465 Å
Swift UVOT (UVW1)	2015–03–18	00039550003	158	2600 Å
Swift UVOT (UVW2)	2015–05–28	00039550004	901	1928 Å
Swift UVOT (UVW1)	2015–09–14	00039550005	626	2600 Å
Swift UVOT (U)	2016–05–20	00039550007	394	3465 Å
XMM-Newton PN	2015–05–04	0761510101	10650	0.3–12 keV
XMM-Newton MOS1	2015–05–04	0761510101	12543	0.5–10 keV
XMM-Newton MOS2	2015–05–04	0761510101	12649	0.5–10 keV
XMM-Newton OM (U)	2015–05–04	0761510101	8000	3440 Å
XMM-Newton OM (UVW1)	2015–05–04	0761510101	8000	2910 Å
XMM-Newton OM (UVM2)	2015–05–04	0761510101	4000	2310 Å

Table 2. Table 2 X-ray and UV/optical properties

Observatory	Date	$f_{2\mathrm{keV}}$	$f_{\mathrm{W2}}$	$f_{\mathrm{M2}}$	$f_{\mathrm{W1}}$	$f_U$	$f_{2500\text{Å}}$	${\alpha }_{\mathrm{OX}}$	$\Delta {\alpha }_{\mathrm{OX}}$	$f_{\mathrm{weak}}$
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
Swift	2013–05–27	$9.13$	–	–	$6.30$	–	$6.57$	$-1.48\pm 0.02$	$-0.09$	${1.8}_{-0.2}^{+0.2}$
	2014–01–25	–	$4.74$	–	–	–	$5.37$	–	–	–
	2014–09–04	$2.86$	–	–	–	$7.04$	$6.23$	$-1.67\pm 0.05$	$-0.28$	${5.4}_{-1.5}^{+2.0}$
	2015–03–18	–	–	–	$3.89$	–	$4.06$	–	–	–
	2015–05–28	–	$4.83$	–	–	–	$5.47$	–	–	–
	2015–09–14	–	–	–	$5.44$	–	$5.67$	–	–	–
	2016–05–20	–	–	–	–	$6.98$	$6.18$	–	–	–
XMM-Newton	2015–05–04	$0.36$	–	$4.60$	$5.12$	$5.80$	$5.03$	$-1.97\pm 0.03$	$-0.60$	${36.2}_{-5.3}^{+6.0}$

Table 3. Table 3 Spectral Fitting Results

Observatory	Observation	Band	Total	Background	${\Gamma }_s$	${\Gamma }_h$	C-stat/dof	$F_{0.5\text{–}2\mathrm{keV}}$	$F_{2\text{–}10\mathrm{keV}}$
	Date	(keV)	Counts	Counts				(${10}^{-13}\mathrm{erg}{\mathrm{cm}}^{-2}{s}^{-1}$)	(${10}^{-13}\mathrm{erg}{\mathrm{cm}}^{-2}{s}^{-1}$)
Swift	2013–05–27	0.3–10	176	3.4	${2.89}_{-0.13}^{+0.14}$	–	$108.7/103$	${8.47}_{-0.43}^{+0.71}$	${3.03}_{-0.46}^{+0.62}$
		1.8–10	22	1.8	–	${2.36}_{-0.69}^{+0.72}$	$22.9/22$	${6.70}_{-5.67}^{+2.87}$	${5.15}_{-0.66}^{+2.70}$
Swift	2014–09–04	0.3–10	34	2.0	${3.07}_{-0.37}^{+0.37}$	–	$20.5/30$	${3.01}_{-0.45}^{+0.55}$	${0.83}_{-0.33}^{+0.64}$
XMM-Newton	2015–05–04	0.3–10	295	72.1	${3.06}_{-0.17}^{+0.18}$	–	$126.7/178$	${0.39}_{-0.03}^{+0.04}$	${0.11}_{-0.03}^{+0.04}$

Table 4. Table 4 AGNs with Extreme X-ray  Variability

Object	$z$	log($M_{\mathrm{BH}}/M_{\odot }$)	Note aaMethod for estimate of BH mass. M: reverberation mapping method; X: X-ray excess variance method; R: scaling relation of Vestergaard & Peterson (2006).	FWHM(H$\beta$)	$L_{5100\text{Å}}$	log $L_{\mathrm{Bol}}$	log $\dot{ℳ}$	${\lambda }_{\mathrm{Edd}}$	References
				$\mathrm{km}{s}^{-1}$	$\mathrm{erg}{s}^{-1}$	$\mathrm{erg}{s}^{-1}$
Quasars
SDSS $\mathrm{J0751}+2914$	0.121	7.20	M	1679	44.21	45.22	1.45	0.7	Du et al. (2018)
PG $1211+143$	0.085	7.87	M	2012	44.73	45.72	0.84	0.5	Du et al. (2015); Danehkar et al. (2018)
PHL 1092	0.396	8.48	X	1800	45.45	46.65	0.65	1.11	Dasgupta et al. (2004); Nikołajuk et al. (2009); Miniutti et al. (2012)
PG $0844+349$	0.064	7.66	M	2694	44.22	45.4	0.50	0.36	Vasudevan & Fabian (2009), Du et al. (2015)
Selected Lower-luminosity Counterparts
1H $0707-495$	0.0411	6.60	R	980	43.52	44.47	1.57	0.6	Done & Jin (2016)
IRAS $13224-3809$bbThe BH-mass estimates of IRAS $13224-3809$ in previous studies have large uncertainties. Listed in this table are the possible ranges of estimated BH mass and Eddington ratio. Its $L_{5100\text{Å}}$ value was derived from an interpolation of the $B$- and $V$-band photometric data (Ojha et al., 2009).	0.0667	6–7	–	650	43.94	44.53	–	$0.3\text{–}3$	Jiang et al. (2018)
Mrk 335	0.0258	6.93	M	1707	43.76	44.73	1.27	0.42	Du et al. (2015)
NGC 4051	0.00234	5.42	M	851	41.96	42.91	1.59	0.21	Du et al. (2015)

Equations

$$L\propto P_{\mathrm{var}}^3\times \prod _{i=1}^{40}[P_{\mathrm{var}}(1-D{)}^{n_{\mathrm{obs},i}}+(1-P_{\mathrm{var}})],$$

Full text

1111affiliationtext: National Astronomical Observatories of China, Chinese Academy of Sciences, 20A Datun Road, Beijing 100020, China1212affiliationtext: School of Astronomy and Space Science, University of Chinese Academy of Sciences, 19A Yuquan Road, Beijing 100049, China

1. Introduction

Active galactic nuclei (AGNs) are in general powerful X-ray sources in the Universe. X-ray emission from AGNs mainly originates from a corona of hot electrons in the vicinity of the black hole (BH). Thermal UV/optical photons emitted from the accretion disk are inverse-Compton scattered by the hot electrons to X-ray energies, forming a power-law continuum (e.g., Sunyaev & Titarchuk, 1980; Haardt & Maraschi, 1993). The detailed physics of the corona is not well understood. Observationally, except for radio-loud AGNs and other AGNs with potential X-ray absorption, such as broad absorption line (BAL) quasars, the X-ray emission of type 1 AGNs is closely related to the UV/optical emission from the accretion disk. This is generally quantified as a tight correlation between the X-ray-to-optical power-law slope parameter ($\alpha_{\rm OX}$111$\alpha_{\rm OX}=0.384{\rm log}(f_{\rm 2~{}keV}/f_{\rm 2500~{}{\textup{\AA}}})$, where ${\sc f_{\rm 2~{}keV}}$ and $f_{\rm 2500~{}{\textup{\AA}}}$ are the rest-frame 2 keV and $2500~{}{\textup{\AA}}$ flux densities. It is an X-ray loudness parameter that is used to compare the UV/optical and X-ray luminosities.) and the $2500~{}{\textup{\AA}}$ monochromatic luminosity (e.g., Strateva et al., 2005; Steffen et al., 2006; Just et al., 2007; Grupe et al., 2010; Lusso et al., 2010). This relation is often used to assess the level of X-ray weakness for AGNs. When the observed $\alpha_{\rm OX}$ value of an AGN is lower than that expected from the $\alpha_{\rm OX}$–$L_{\rm 2500~{}{\textup{\AA}}}$ relation ($\Delta\alpha_{\rm OX}=\alpha_{\rm OX}-\alpha_{\rm OX,exp}<0$, where more negative values indicate more extreme X-ray weakness relative to the $\rm 2500~{}{\textup{\AA}}$ monochromatic luminosity), it is X-ray weak by a linear factor of $f_{\rm weak}=10^{-\Delta\alpha_{\rm OX}/0.384}$.

X-ray luminosity variability is a characteristic property of AGNs, which is generally related to instabilities of the corona or fluctuations of the accretion flow (e.g., Nandra, 2001; McHardy et al., 2006; MacLeod et al., 2010; Yang et al., 2016; Zheng et al., 2017). The intrinsic X-ray variability of AGNs contributes partially to the dispersion of the $\alpha_{\rm OX}$–$L_{\rm 2500~{}{\textup{\AA}}}$ relation (e.g., Vagnetti et al., 2010; Gibson & Brandt, 2012; Vagnetti et al., 2013; Chiaraluce et al., 2018). Compared to AGN variability at longer wavelengths, AGN X-ray variability often displays larger amplitudes on timescales of years down to minutes (e.g., Ulrich et al., 1997; Peterson, 2001). The typical long-term X-ray flux variability amplitude of AGNs is $\sim 20\textrm{--}50\%$, and seldom exceeds $200\%$ (e.g., Yuan et al., 1998; Grupe et al., 2001; Paolillo et al., 2004; Mateos et al., 2007; Vagnetti et al., 2011; Gibson & Brandt, 2012; Soldi et al., 2014; Yang et al., 2016; Middei et al., 2017; Maughan & Reiprich, 2018). There are several types of AGNs that may have excess X-ray variability compared to typical AGNs. Radio-loud AGNs are known to have X-ray emission contributed from the jets (e.g., Worrall et al., 1987; Wilkes & Elvis, 1987; Worrall & Birkinshaw, 2006; Miller et al., 2011; Komossa, 2018), and they often show strong X-ray variability related to the jets (e.g., Gliozzi et al., 2007; Carnerero et al., 2017; Zhu et al., 2018). Strong X-ray variability has also been observed in several BAL quasars, likely due to X-ray absorption variability (e.g., Gallagher et al., 2004; Saez et al., 2012; Kaastra et al., 2014; Mehdipour et al., 2017). Recently, a small number of ”changing-look” quasars have been discovered (e.g., LaMassa et al., 2015; Parker et al., 2016; Mathur et al., 2018; Oknyansky et al., 2019). Such an AGN likely has different amounts of X-ray emission in different states, although its $\alpha_{\rm OX}$ values are not necessarily abnormal as the X-ray and UV/optical emission might vary in a coordinated manner. We do not consider these types of AGNs in our following discussion of AGN X-ray variability.

Besides radio-loud AGNs, BAL quasars, and changing-look quasars, only a small number of extremely X-ray variable AGNs have been reported in the past two decades; in this study we consider extremely X-ray variable AGNs as those varying in X-rays by factors of larger than 10 ($\gtrsim 2.5\sigma$ deviation from the expected $\alpha_{\rm OX}$–$L_{\rm 2500~{}{\textup{\AA}}}$ relation; Table 5 of Steffen et al. 2006). These AGNs all vary between the X-ray normal and weak states, with X-ray weakness factors of $f_{\rm weak}>10$ ($\Delta\alpha_{\rm OX}<-0.384$) in the low state. Most of these objects are narrow-line Seyfert 1 galaxies (NLS1s), which are a well-studied group of AGNs that generally have small BH masses ($\lesssim 10^{7}~{}\rm M_{\odot}$) and large Eddington ratios ($\lambda_{\rm Edd}\gtrsim 0.1$; e.g., Boller et al. 1996; Leighly 1999a, b; Grupe et al. 2001, 2004b; Gallo 2018; Komossa 2018 and references therein). We give a few examples of this phenomenon observed in NLS1s. Mrk 335 fell into a historically low X-ray flux state in 2007 with a flux drop by a factor of 30, and it has ever since shown persistent X-ray variability and occasional intense flares with variability factors of $\approx 10$ in nearly 11 years of monitoring observations (e.g., Grupe et al., 2007, 2012; Gallo et al., 2018). NGC 4051 was observed in an extremely dim state in 1998, with the X-ray flux about 20 times fainter than its historical average value (e.g., Guainazzi et al., 1998; Uttley et al., 1999; Peterson et al., 2000). 1H $0707-495$ and IRAS $13224-3809$ exhibit very rapid X-ray variability with variability amplitudes up to $\approx 100\textrm{--}200$ within only a few hours (e.g., Boller et al., 2002; Fabian et al., 2009, 2012, 2013; Ponti et al., 2010; Jiang et al., 2018).

At higher luminosities, only three radio-quiet non-BAL quasars have been found to show extreme X-ray variability (by factors of larger than 10), which are PHL 1092 (e.g., Miniutti et al., 2009, 2012), PG $0844+349$ (e.g., Gallagher et al., 2001; Gallo et al., 2011), and PG $1211+143$ (e.g., Bachev et al., 2009). In this study, we consider quasars as AGNs with the rest-frame $5100~{}\textup{\AA}$ luminosity larger than $10^{44}~{}\rm erg~{}s^{-1}$, which typically have BH masses of $\gtrsim 10^{7}\rm~{}M_{\odot}$. These three extremely X-ray variable quasars are narrow-line (NL) type 1 quasars that possess narrower than typical $\rm H\beta$ lines in their optical spectra similar to NLS1s, indicating relatively small BH masses and high Eddington ratios compared to typical quasars. Another common feature among these extremely X-ray variable NLS1s and NL type 1 quasars is that there is no coordinated UV/optical continuum or emission-line variability with their X-ray variability, suggesting that the accretion rates of these objects did not change significantly (e.g., Bachev et al., 2009; Gallo et al., 2011; Grupe et al., 2012; Miniutti et al., 2012; Robertson et al., 2015; Buisson et al., 2018). Most of the extremely X-ray variable AGNs mentioned above have been discussed in early studies of $\alpha_{\rm OX}$ variability in Seyfert 1 galaxies, and it has been suggested that extremely variable $\alpha_{\rm OX}$ values are preferentially observed in NLS1s (e.g., Gallo, 2006; Vasudevan et al., 2011).

The rarity of extremely X-ray variable AGNs discovered so far may be related to the limited numbers of X-ray observations available for individual objects, and we probably have not found the extremely X-ray weak states for most of the NLS1s and NL type 1 quasars. Using the published data of two extremely X-ray variable AGNs, Mrk 335 and PHL 1092, we roughly estimated that the duty cycles of their extremely X-ray weak states ($f_{\rm weak}>10$, $\Delta\alpha_{\rm OX}<-0.384$) are $30\%$ and $60\%$, respectively (see the analysis in Section 4.2 below). If extreme X-ray variability is a common feature for NLS1s and NL type 1 quasars, with a similar $\approx 30\textrm{--}60\%$ duty cycle for the extremely X-ray weak state, we would expect that among a large sample of these AGNs, $\approx 30\textrm{--}60\%$ will deviate significantly from the $\alpha_{\rm OX}$–$L_{\rm 2500~{}{\textup{\AA}}}$ relation (with $\Delta\alpha_{\rm OX}<-0.384$). However, such large fractions of X-ray weak outliers have not been found in previous studies of the $\alpha_{\rm OX}$–$L_{\rm 2500~{}{\textup{\AA}}}$ relation (e.g., Grupe et al., 2010; Vasudevan et al., 2011; Gibson & Brandt, 2012). The rarity of extreme X-ray variable AGNs is likely intrinsic, and they only constitute a small fraction of NLS1 or NL type 1 quasar population (see Section 4.2 below for detailed discussion).

Here we report the discovery of extreme X-ray variability in another quasar, SDSS J$075101.42+291419.1$ (hereafter SDSS J$0751+2914$). The source has a redshift of $z=0.1208$, and it is a radio-quiet (radio loudness $R=1.29$)222The radio loudness parameter is defined as $R=f_{\rm 6~{}cm}/f_{\rm 2500~{}\textup{\AA}}$, where $f_{\rm 6~{}cm}$ and $f_{\rm 2500~{}\textup{\AA}}$ are flux densities at 6 cm and $2500~{}\textup{\AA}$, respectively (e.g., Jiang et al., 2007). The value is adopted from Shen et al. (2011). quasar that possesses spectroscopic characteristics similar to NLS1s; e.g., a relatively narrow $\rm H\beta$ line ($\rm FWHM=1679~{}km~{}s^{-1}$), strong optical Fe II emission, and weak [O III] lines. It is one of the targets in the reverberation mapping (RM) campaign targeting super-Eddington accreting massive black holes (SEAMBHs; Du et al., 2015, 2016, 2018). Its virial BH mass constrained from the $\rm H\beta$ RM is $1.6\times 10^{7}~{}\rm M_{\odot}$, and its dimensionless mass accretion rate derived from the standard thin disk model (Shakura & Sunyaev, 1973) is $\dot{\rm\mathscr{M}}=\dot{M}c^{2}/{L_{\rm Edd}=20.1(\ell_{44}}/{\cos i})^{3/2}(M_{\rm BH}/10^{7}M_{\odot})^{-2}=28.2$, where $\dot{M}$ is the mass accretion rate, $L_{\rm Edd}=1.5\times 10^{38}(M_{\rm BH}/M_{\odot})~{}\rm erg~{}s^{-1}$ is the Eddington luminosity for solar composition gas, $\ell_{44}=L_{5100~{}\textup{\AA}}/10^{44}~{}\rm erg~{}s^{-1}$ is the rest-frame $5100~{}\textup{\AA}$ luminosity in units of $10^{44}~{}\rm erg~{}s^{-1}$, $M_{\rm BH}$ is the BH mass, and i (adopted as cos ${\it i}=0.75$) is the inclination angle of the disk (e.g., Du et al., 2015, 2018). The Eddington ratio and $\dot{\rm\mathscr{M}}$ follow the relation: $\lambda_{\rm Edd}=L_{\rm Bol}/L_{\rm Edd}=\eta\dot{\rm\mathscr{M}}$, where $L_{\rm Bol}$ is the disk bolometric luminosity, and $\eta$ is the mass-to-radiation conversion efficiency. According to the criterion of $\dot{\rm\mathscr{M}}>3$ used for identifying SEAMBH candidates (Wang et al., 2014b; Du et al., 2015), SDSS J$0751+2914$ can be classified as a SEAMBH candidate. We caution that there could be substantial systematic uncertainties on the the measured BH mass and accretion rate of SDSS J$0751+2914$, because for an AGN with a high accretion rate, the broad line region (BLR) may no longer be virialized (e.g., Marconi et al., 2008, 2009; Netzer & Marziani, 2010; Krause et al., 2011; Pancoast et al., 2014; Li et al., 2018) and the accretion disk may deviate from the standard thin disk (e.g., Abramowicz et al., 1980, 1988; Wang & Netzer, 2003; Ohsuga & Mineshige, 2011; Wang et al., 2014a; Jiang et al., 2016, 2017). SDSS J$0751+2914$ has archival Swift (Gehrels et al., 2004) and XMM-Newton (Jansen et al., 2001) X-ray observations, for which simultaneous UV/optical photometric observations by the same satellites are available. We serendipitously discovered its extreme X-ray variability during systematic analyses of the X-ray properties for SEAMBHs (H. Liu et al, in preparation).

The paper is organized as follows. In Section 2, we present the mutiwavelength observations and describe the data analysis processes. The main results are presented in Section 3. In Section 4, we describe the common properties of extremely X-ray variable AGNs, estimate the occurrence rate of extreme X-ray variability among AGNs with high accretion rates, and discuss possible physical mechanisms for extreme X-ray variability in AGNs. We summarize and present future prospects in Section 5. Throughout this paper, we use J2000 coordinates and a cosmology with $H_{0}=67.4$ km s*-1* Mpc*-1*, $\Omega_{\rm M}=0.315$, and $\Omega_{\Lambda}=0.686$ (Planck Collaboration et al., 2018).

2. Multiwavelength Observations and Data Reduction

2.1. Swift Observations

SDSS J$0751+2914$ has been observed by Swift on seven occasions since 2013. These observations were performed simultaneously with the X-ray Telescope (XRT; Burrows et al. 2005) and UV-Optical Telescope (UVOT; Roming et al. 2005). For five observations, the exposure times are less than 1 ks, and we did not use their XRT data since no useful constraints can be derived. We analyzed the XRT data of the other two observations, which were performed on 2013 May 27 and 2014 September 4, respectively. The observation log is given in Table 1. The XRT was operated in Photon Counting (PC) mode (Hill et al., 2004) and the data were reduced with the task xrtpipeline version 0.13.2, which is included in the HEASOFT package 6.17. The spectral extraction was performed using the task xselect version 2.4. Source photons were selected from a circular region centered on the optical position of SDSS J$0751+2914$ with a 47″ radius. The corresponding background photons were extracted from a nearby source-free circular region with a 100″ radius. The ancillary response function files (arfs) were generated by xrtmkarf, and the standard photon redistribution matrix files (rmfs) were obtained from the CALDB. There are 176 and 34 photons in the 0.3–10 keV band in the spectra of the 2013 and 2014 observations, respectively. Considering the small numbers of counts, we grouped the spectra using grppha such that each bin contains at least 1 photon count.

Each of the seven UVOT observations was performed using only one filter (U, UVW1, or UVM2), which is reported in Table 2. After aspect correction, the exposures were co-added for each segment in each filter using the task uvotimsum. Source counts were extracted from a circular region with a 5″ radius centered on the source position determined by the task uvotdetect, and the background counts were extracted from a nearby source-free circular region with a radius of 20″. The magnitudes and fluxes in these UVOT bands were then computed using the task uvotsource. These data were corrected for Galactic extinction at the source position ($E_{\rm B-V}=0.042$; Schlegel et al. 1998) following the dereddening approach of Cardelli et al. (1989) and O’Donnell (1994).

2.2. XMM-Newton Observation

SDSS J0751+2914 was observed with XMM-Newton for $\sim 11$ ks on 2015 May 4 using simultaneously the European Photon Imaging Camera (EPIC) PN (Strüder et al., 2001) and MOS (Turner et al., 2001) detectors, and the Optical Monitor (OM; Mason et al., 2001). The observation information is reported in Table 1. This observation was presented in Castelló-Mor et al. (2017), while we reprocessed the observational data to make comparisons to the Swift observations. The EPIC observations were operated in Full Window mode. The data were processed using the XMM-Newton Science Analysis System (SAS v.16.0.0) and the latest calibration files. We only used the EPIC PN X-ray data, which were reduced with the task epproc. Only single and double events were selected, and bad pixels and high background flares were filtered from the calibrated event lists based on the standard selection criteria, which resulted in a final cleaned exposure time of $5.3$ ks. We extracted the source spectrum using a circular region with a radius of 35″centered on the source position determined by the task edetect-chain. The background spectrum was extracted from a nearby source-free circular region of the same size in the same CCD chip. Spectral response files were generated using the tasks rmfgen and arfgen. The source spectrum contains 295 photons in the 0.3–10 keV energy band. We grouped the spectrum using the task specgroup with a minimum of one photon count per energy bin.

The OM has similar filters to those of the Swift UVOT, although the effective wavelengths of these filters are somewhat different (see Table 1). The OM observation was reduced with the task omchian, which generated five exposures for three filters (U, UVW1, and UVM2). The photometric results of every exposure are recorded in SWSRLI files, and we extracted the magnitudes and fluxes of our target from these files. We adopted the mean magnitudes and fluxes of all the exposure segments for each filter, which were then corrected for Galactic extinction.

2.3. Lijiang Observations

SDSS $\rm J0751+2914$ was observed repeatedly with the Lijiang 2.4 m telescope at the Yunnan Observatories of the Chinese Academy of Sciences during 2013 November–2014 May and 2016 October–2017 June (Du et al., 2015, 2018). It was also observed with the Lijiang telescope simultaneously with the XMM-Newton observation on 2015 May 4. The details of the observations and the data reduction were reported in Du et al. (2014, 2015, 2018). The light curves of the $5100~{}{\textup{\AA}}$ continuum flux density and the $\rm H\beta$ emission-line flux in the two RM monitoring periods are presented in Section 3.4 below. We gathered three Lijiang spectra, for which the observation dates are closest to those of the X-ray observations. The first spectrum was observed on 2013 November 13, 170 days after the 2013 Swift observation. The second spectrum was observed on 2014 May 17, 110 days before the 2014 Swift observation. Another spectrum is the one observed simultaneously with the 2015 XMM-Newton observation.

3. RESULTS

3.1. X-ray Spectral Analysis

All of the X-ray spectral fitting was performed with XSPEC (v12.9.1; Arnaud, 1996). Due to the small numbers of counts in the spectra, the Cash statistic (CSTAT; Cash 1979) 333The W statistic was actually used in the XSPEC spectral fitting when background spectra are included. see https://heasarc.gsfc.nasa.gov/xanadu/xspec/manual/ XSappendixStatistics.html for details. was used in parameter estimation as it is based on the Poisson distribution. We first used a simple power-law model modified by Galactic absorption (wabs*zpowerlw) to fit the 0.3–10 keV spectra. The Galactic neutral hydrogen column density was fixed at $N_{\rm H}=3.55\times 10^{20}\rm~{}cm^{-2}$ (Kalberla et al., 2005). The best-fit results are presented in Figure 1 and the model parameters are reported in Table 3. The error bars plotted in Figure 1 and the parameter errors listed in Table 3 are at a 68% ($1\sigma$) confidence level. In general, these models provide reasonable fits to the three spectra. The three power-law photon indices ($\rm\Gamma_{s}$ in Table 3) indicate steep spectral shapes, which are consistent with those of NLS1s (e.g, Brandt et al., 1997; Leighly, 1999b).

Soft X-ray excess emission can steepen the power-law slope of an X-ray spectrum. It is common in NLS1s, and also in quasars in general (e.g., Boller et al., 1996; Leighly, 1999b; Porquet et al., 2004; Grupe et al., 2010; Marlar et al., 2018). Thus, we also attempted to fit the 1.8–10 keV (rest-frame energies above 2 keV) hard X-ray spectra with the same power-law model to isolate the intrinsic X-ray emission from the corona. This analysis is only feasible for the 2013 Swift observation, since there is no photon in the hard X-ray spectrum of the 2014 Swift observation, and the hard X-ray spectrum of the 2015 XMM-Newton observation is dominated by the $\approx 42$ background photons (leaving $\approx 8$ net source photons). Due to the limited signal-to-noise ratio of the 2013 Swift hard X-ray spectrum, its power-law photon index ($\rm\Gamma_{h}$ in Table 3) derived from the fitting has considerable uncertainties. Considering the large uncertainties, the $\rm\Gamma_{h}$ value is consistent with the $\rm\Gamma_{s}$ value of the entire $0.3\textrm{--}10$ keV spectrum, and the rest-frame 2 keV flux density derived from the $1.8\textrm{--}10$ keV spectral fitting is also consistent with that derived from the $0.3\textrm{--}10$ keV spectral fitting. Since the spectra appear to be dominated by soft-band counts, we also fit the same single power-law model to the $0.3\textrm{--}1.8$ keV spectra of the 2013 Swift and 2015 XMM-Newton observations. The fitting revealed two steep power laws of $\Gamma\approx 3$, consistent with the $0.3\textrm{--}10$ keV spectral fitting results. In the discussion below, we thus adopt the 2 keV flux densities and $0.5\textrm{--}2$ keV fluxes derived from the $0.3\textrm{--}10$ keV spectral fitting.

3.2. X-ray Variability

Figure 2(a) depicts the long-term variability of the 0.5–2 keV flux for SDSS $\rm J0751+2914$. With the flux dropping by a factor of $21.6\pm 2.4$ (factors of $2.8\pm 0.5$ and $7.7\pm 1.5$ successively from 2013 to 2014 and 2014 to 2015), SDSS $\rm J0751+2914$ has gradually fallen into a low X-ray flux state. In addition, we examined the short-term X-ray variability of SDSS $\rm J0751+2914$ within the 2013 normal-flux state and 2015 low-flux state. Figure 3 shows the $0.3\textrm{--}2$ keV light curve with time bins of 1 ks for the 2015 XMM-Newton observation. The average count rate in this observation is 0.046 $\rm Cts~{}s^{-1}$, with a root-mean-square (rms) variability of 0.011 $\rm Cts~{}s^{-1}$ and a fractional rms variability amplitude (e.g., Equation 10 of Vaughan et al., 2003) of $8\%$. For the 2013 Swift observation, we extracted a $0.3\textrm{--}2$ keV light curve with time bins of 500 s. The measured average count rate is 0.055 $\rm Cts~{}s^{-1}$, with a rms variability of 0.020 $\rm Cts~{}s^{-1}$ and a fractional rms variability amplitude of $32\%$.

With the simultaneous X-ray and UV observations, we can compute reliably the X-ray-to-optical power-law slope ($\alpha_{\rm OX}$) of SDSS $\rm J0751+2914$. We first measured a UV/optical spectral slope using the data of the 2015 XMM-Newton OM observation, which is the only observation having multiple filters (see Table 1). Fitting a single power-law model to the U, UVW1, and UVM2 data points revealed a spectral slope of $\alpha_{\nu}=-0.57$. This slope is consistent with those of typical NLS1s (e.g., Grupe et al., 2010). We then determined the $2500~{}{\textup{\AA}}$ flux density of the 2015 XMM-Newton observation from the best-fit model of the OM data. For each of the seven Swift UVOT observations, only one UV filter was used. The $2500~{}{\textup{\AA}}$ flux densities were extrapolated from the flux densities of the available filters, adopting the same power-law slope of $\alpha_{\nu}=-0.57$. If we adopt a spectral slope of $\alpha_{\nu}=-0.44$ in the extrapolation which is the average value for typical quasars (e.g., Vanden Berk et al., 2001), the resulting $\alpha_{\rm OX}$ values would only change slightly (by less than 0.01) and our following analyses and discussions would not be affected. After obtaining the $\alpha_{\rm OX}$ values (see Table 2), we calculated the difference ($\Delta\alpha_{\rm OX}$) between the observed $\alpha_{\rm OX}$ and that ($\alpha_{\rm OX,exp}$) expected from the $\alpha_{\rm OX}$–$L_{2500~{}{\textup{\AA}}}$ relation, which indicates the level of X-ray weakness. These X-ray and UV/optical properties of SDSS $\rm J0751+2914$ are listed in Table 2.

Figure 4 shows the $\alpha_{\rm OX}$ versus $L_{2500~{}{\textup{\AA}}}$ values of the three X-ray observations for SDSS $\rm J0751+2914$. Typical AGNs in the sample of Steffen et al. (2006) are also presented for comparison. The 2013 Swift data point of SDSS $\rm J0751+2914$ is close to the red line that represents the $\alpha_{\rm OX}$–$L_{2500~{}{\textup{\AA}}}$ relation in Steffen et al. (2006), indicating a normal X-ray emission level at this time. However, the quasar became extremely X-ray weak in the 2015 XMM-Newton observation with $\Delta\alpha_{\rm OX}=-0.60\pm 0.03$ ($f_{\rm weak}=36.2^{+6.0}_{-5.3}$), corresponding to a $3.6\sigma$ deviation from the $\alpha_{\rm OX}$–$L_{2500~{}{\textup{\AA}}}$ relation (see Table 5 of Steffen et al. 2006). The $\Delta\alpha_{\rm OX}$ variability is also shown in Figure 2(c); the decreasing of the $\Delta\alpha_{\rm OX}$ value from 2013 to 2015 is coordinated with the drop of the X-ray flux.

3.3. UV Variability

Presented in Figure 2(b) is the light curve of the $2500~{}{\textup{\AA}}$ flux density derived from the UVOT and OM photometric data. The gray dashed lines indicate the dates of the simultaneous X-ray and UV observations. Combined with the results reported in Table 2, we find that the UV flux density has a much smaller variability amplitude compared to the 2 keV flux density. The $2500~{}{\textup{\AA}}$ flux density varied by just a factor of 1.3 between the high X-ray flux state and the low X-ray flux state. Such little variation of the UV flux suggests that the physical mechanism leading to the strong X-ray variability largely does not affect the UV emission.

We have no UV spectrum of SDSS $\rm J0751+2914$ to identify whether it is a BAL quasar, in which case it may be affected by absorption associated with outflows (e.g., Murray et al., 1995; Matthews et al., 2016). BAL quasars with extreme X-ray variability often show significant UV/optical continuum and BAL variability coordinated with the X-ray variability (e.g., Gallagher et al., 2004; Saez et al., 2012; Kaastra et al., 2014; Mehdipour et al., 2017). SDSS $\rm J0751+2914$ lacks significant UV/optical variability coordinated with its X-ray variability. It is thus probably not a BAL quasar with strong X-ray variability. A UV spectroscopic observation of SDSS $\rm J0751+2914$ is required to confirm this notion.

3.4. Optical Spectrum and Light Curves

Three Lijiang spectra in the rest-frame $4200\textrm{--}5200~{}{\textup{\AA}}$ range are shown in Figure 5, which also includes the SDSS spectrum observed on 2002 December 28. The continuum and emission lines did not vary significantly in general with a $\sim 10\%$ percent variability amplitude among the four observations. The emission-line profiles in the SDSS spectrum appear relatively sharp because of the better spectral resolution of the SDSS observation.

Figure 6 shows the light curves of the $5100~{}{\textup{\AA}}$ continuum flux density ($F_{5100~{}{\textup{\AA}}}$) and the $\rm H\beta$ emission-line flux ($F_{\rm H\beta}$) during the 2013–2014 and 2016–2017 RM monitoring periods (Du et al., 2015, 2018). We also added a data point measured from the 2015 May 4 Lijiang spectrum following the approach in Du et al. (2014). During 2013–2015, the $F_{5100~{}{\textup{\AA}}}$ and $F_{\rm H\beta}$ values of SDSS $\rm J0751+2914$ have maximum variability amplitudes of $\sim 30\%$, which are only mild compared to its extreme X-ray variability. This indicates that the accretion rate of SDSS $\rm J0751+2914$ did not change significantly during this period and the extreme X-ray variability should be driven by some other mechanisms. The $F_{5100~{}{\textup{\AA}}}$ and $F_{\rm H\beta}$ values in the 2016–2017 period have increased in general compared to those in the 2013–2015 period, and the maximum variability amplitudes of the two parameters are $\sim 70\%$ among all the RM observations. This variability is more significant than that of the other SEAMBHs which generally have maximum variability amplitudes of $\sim 20\textrm{--}30\%$ (see Du et al., 2014, 2015, 2016, 2018). The X-ray observations of SDSS $\rm J0751+2914$ were performed in 2013–2015, and there is no coordinated optical continuum and emission-line variability with the X-ray variability of SDSS $\rm J0751+2914$ during this period.

We also investigated the $V$-band light curve of SDSS $\rm J0751+2914$ obtained from the Catalina Real-Time Transient Survey (CRTS; Drake et al. 2009). The monitoring period is between 2005 April and 2013 September. During this period, the $V$-band magnitude of SDSS $\rm J0751+2914$ varied between $15.89\textrm{--}16.27$ (a 42% maximum variability amplitude in flux), with a mean value of 16.06.

3.5. Mutiwavelength Spectral Energy Distribution

We gathered infrared (IR)-to-UV photometric data to construct the rest-frame spectral energy distribution (SED) for SDSS $\rm J0751+2914$, which is shown in Figure 7. The data were collected from the public catalogs of the Wide-field Infrared Survey Explorer (WISE; Wright et al., 2010), Two Micron All Sky Survey (2MASS; Skrutskie et al., 2006), SDSS, and Galaxy Evolution Explorer (GALEX; Martin et al., 2005). We added the UVOT and OM photometric data, and the corresponding 2 keV and 10 keV luminosities to the SED. Also, the $5100~{}{\textup{\AA}}$ monochromatic luminosities calculated from the three Lijiang spectra in Figure 5 were added. We caution that most of these photometric data are not contemporaneous. All the SED data were corrected for the Galactic extinction at the source position. The mean SED of typical SDSS quasars with luminosities of log$(\nu L_{\nu}|_{\lambda=2500~{}\textup{\AA}}/\rm erg~{}s^{-1})\leq 45.41$ in Krawczyk et al. (2013), scaled to the mean $5100~{}{\textup{\AA}}$ luminosity of SDSS $\rm J0751+2914$, is shown in Figure 7 for comparison. We note that the IR-to-UV SED of SDSS $\rm J0751+2914$ is consistent with those of typical quasars, except for the GALEX FUV data (shown as the brown point in Figure 7) at rest-frame $1364.2~{}{\textup{\AA}}$ ($\rm log~{}\nu_{rest}=15.34$), which lies below the mean SED. The date of this GALEX observation is 2006 December 29, which is prior to the three X-ray observations and the RM observations, and thus we cannot determine whether this particular feature is related to the extreme X-ray variability. A far UV spectrum is required to examine if there is any UV absorption. The X-ray data points indicate the soft X-ray spectral shapes, which have not changed significantly between the three X-ray observations. In spite of the extreme X-ray flux variability, the optical-to-UV SED did not change significantly.

We estimated the bolometric luminosity of SDSS $\rm J0751+2914$ by integrating the scaled SED template of Krawczyk et al. (2013) shown in Figure 7. Most of the IR radiation ($\sim 1\textrm{--}30~{}\rm\mu m$) is produced in the large-scale ”dust torus” beyond the accretion disk. It is considered to be the reprocessed emission and should not be included in the computation of the bolometric luminosity (e.g., Krawczyk et al., 2013, and references therethin). However, super-Eddington accreting quasars are expected to produce much stronger extreme UV radiation than typical quasars (e.g., Wang et al., 2014a; Castelló-Mor et al., 2016), but this portion of the SED is not observable, and it is not represented by the Krawczyk et al. (2013) mean quasar SED. Thus we included the IR SED in the integration to compensate somewhat for the uncertain extreme UV emission. The resulting bolometric luminosity is $\rm 1.66\times 10^{45}~{}erg~{}s^{-1}$ ($\rm 1.03\times 10^{45}~{}erg~{}s^{-1}$ if not including the $1\textrm{--}30~{}\rm\mu m$ SED). We note that the X-ray spectrum in the SED template was included in the integration of the bolometric luminosity. Since the $2\textrm{--}10$ keV template luminosity contributes only a small fraction ($\approx 2\%$) of the bolometric luminosity, the result would only change slightly if we adopt the $2\textrm{--}10$ keV luminosities determined from the observational data. We also caution that the observed X-ray luminosities likely do not represent the intrinsic X-ray luminosity (see Section 4.3).

We also estimated the bolometric luminosity using the bolometric correction from the $3~{}\rm\mu m$ monochromatic luminosity (Gallagher et al., 2007), which was derived from the WISE photometric data. The resulting bolometric luminosity is $\rm 1.61\times 10^{45}~{}erg~{}s^{-1}$, consistent with the integrated luminosity of the $30~{}\rm\mu m\textrm{--}10~{}keV$ SED. Given that the BH mass of SDSS $\rm J0751+2914$ is $1.6\times 10^{7}~{}M_{\odot}$, we obtained an Eddington ratio of 0.7. This high Eddington ratio suggests that SDSS $\rm J0751+2914$ is indeed accreting at a high accretion rate. We caution that the BH mass and subsequently the Eddington ratio may have substantial uncertainties (see Section 1).

4. DISCUSSION

4.1. Extremely X-ray Variable AGNs

SDSS $\rm J0751+2914$ is another NL type 1 quasar with extreme X-ray variability. It varied in X-rays by a factor of $21.6$ on a timescale of three years. In the same period there was no coordinated UV/optical continuum or emission-line variability, indicating that the accretion rate of SDSS $\rm J0751+2914$ is almost constant and there are some other factors instead of a change of accretion rate driving the extreme X-ray variability. These features are similar to the other three extremely X-ray variable NL type 1 quasars reported, PG $0844+349$ (e.g., Gallo et al., 2011), PG $1211+143$ (e.g., Bachev et al., 2009), and PHL 1092 (e.g., Miniutti et al., 2012). Thus SDSS $\rm J0751+2914$ is a new member of the extremely X-ray variable quasar population. The basic characteristics of these four extremely X-ray variable quasars are listed in Table 4.

Table 4 also lists a few representative extremely X-ray variable NLS1s (with $L_{\rm 5100~{}\textup{\AA}}<10^{44}~{}\rm erg~{}s^{-1}$) selected from the literature. We note that the BH-mass estimates and the computed bolometric luminosities have large uncertainties. Thus there are substantial uncertainties on the derived $\rm\dot{\mathscr{M}}$ and $\lambda_{\rm Edd}$ values (see Section 1). In general, these extremely X-ray variable AGNs have high accretion rates. According to the criterion of $\rm\dot{\mathscr{M}}>3$ used for identifying SEAMBH candidates (Du et al., 2014, 2015), all of these extremely X-ray variable AGNs (except IRAS $13224-3809$) can be considered as SEAMBH candidates. This result implies a connection between extreme X-ray variability and high accretion rates in AGNs.

The steep X-ray spectral shapes of SDSS $\rm J0751+2914$, whether in the high or low X-ray flux state, are similar to those of PHL 1092 (Miniutti et al., 2009, 2012). Considering the large uncertainties, the spectral shape ($\rm\Gamma_{s}$ in Table 3) of SDSS $\rm J0751+2914$ did not change significantly. However, a flattening of the hard ($>2~{}\rm keV$) X-ray spectral shape from the normal to low state is often observed in other extremely X-ray variable AGNs (PG $0844+349$: Gallo et al. 2011; PG $1211+143$: Bachev et al. 2009; Mrk 335: Grupe et al. 2012; Gallo et al. 2015; 1H $0707-495$: Fabian et al. 2012; IRAS $13224-3809$: Jiang et al. 2018). Their low-state X-ray spectra show substantial curvature in the $\approx~{}2\textrm{--}6$ keV band, which is generally interpreted as blurred reflection arising within a few gravitational radii of the BH or partial covering absorption (see discussion in Section 4.3 below). For SDSS $\rm J0751+2914$ and PHL 1092 that are at relatively high redshifts and have relatively low X-ray fluxes, the current observations are probably not sufficiently sensitive to detect this hard X-ray component. Their low-state spectra are likely dominated by the soft X-ray excess component (see Figure 6 of Miniutti et al. 2012 and Figure 1), and there is no apparent curvature nor hardening emerging in their $\approx 2\textrm{--}10$ keV spectra. Hard X-ray observations with NuSTAR (Harrison et al., 2010) and Suzaku (Mitsuda et al., 2007) of some extremely X-ray variable AGNs suggest that their $>10$ keV X-ray fluxes and spectral shapes are less variable (e.g., Gallo et al., 2015; Kara et al., 2015; Jiang et al., 2018). An additional spectral curvature at higher energies ($\approx 20\textrm{--}30$ keV) is also usually observed in different flux states, and it is interpreted as the Compton-reflection hump. A hard X-ray observation with NuSTAR or a deep XMM-Newton or Chandra observation on SDSS $\rm J0751+2914$ in its low X-ray flux state is required to investigate whether it possesses a hard/flat X-ray spectrum similar to the other extremely X-ray variable AGNs.

4.2. Occurrence Rate of Extreme X-ray Variability among

AGNs with High Accretion Rates

Inspired by the possible connection between extreme X-ray variability and high accretion rates, we investigated the occurrence of extremely variable X-ray sources among AGNs with high accretion rates ($\lambda_{\rm Edd}\gtrsim 0.1$). We first target NLS1s that generally have high accretion rates. In the soft X-ray selected Seyfert sample of Grupe et al. (2010), there are 49 broad line Seyfert 1 galaxies (BLS1s) and 43 NLS1s (19 of the 43 can be considered as quasars, with $5100~{}\textup{\AA}$ luminosities exceeding $\rm 10^{44}~{}erg~{}s^{-1}$). All the NLS1s in this sample are high accretion rate AGNs with $\lambda_{\rm Edd}\gtrsim 0.1$ (Grupe et al., 2010). Only three NLS1s (RX J$2217.9-5941$: Grupe et al. 2004a; Mrk 335: Grupe et al. 2007; PG $1211+034$: Bachev et al. 2009) have been found to vary in X-rays by factors of more than 10 between multiple Swift observations. There is no BLS1 in this sample found to show extreme X-ray variability.

The fraction of extremely X-ray variable AGNs ($P_{\rm var}$) among NLS1s should be larger than $3/43$ ($7\%$), as some of them were probably not identified due to the limited number of observations available. This fraction $P_{\rm var}$ also represents the probability of a NLS1 being extremely X-ray variable. We estimated $P_{\rm var}$ using the observations and $\Delta\alpha_{\rm OX}$ values of the NLS1 sample presented in Grupe et al. (2010). We computed the likelihood ($L$) of observing three extremely X-ray variable objects among the 43 NLS1s as a function of $P_{\rm var}$, which can be expressed as

[TABLE]

where $n_{{\rm obs},i}$ is the total number of observations for each object (with a range of 1–9 from Grupe et al. 2010), and D is the duty cycle of the extremely X-ray weak state ($\rm\Delta\alpha_{OX}<-0.384$). The first term of Equation (1) corresponds to the likelihood of observing three objects being extremely X-ray variable. The second term corresponds to the likelihood of observing 40 objects being X-ray normal in all $n_{{\rm obs},i}$ observations (either being extremely X-ray variable but not observed in the X-ray weak state or being non-variable). With the published data, we estimated the duty cycles of the extremely X-ray weak state of two extremely X-ray variable AGNs, PHL 1092 and Mrk 335 (Grupe et al., 2012; Miniutti et al., 2012). Mrk 335 was in the extremely X-ray weak state in about $30\%$ of the about four-year long continuous monitoring observations with a total exposure time of $\approx 301$ ks (see Grupe et al., 2012). PHL 1092 was in the extremely X-ray weak state in about $60\%$ of the observations (see Miniutti et al., 2012)444This fraction is an overestimate of the duty cycle of the extreme X-ray weak state for PHL 1092, as some X-ray observations of PHL 1092 were follow-up observations triggered by its X-ray weak state.. We thus adopted a $30\textrm{--}60\%$ range for the duty cycle of the extremely X-ray weak state for every extremely X-ray variable AGN.

The distributions of L as a function of $P_{\rm var}$ are shown as the green (for a duty cycle of $30\%$) and blue (for a duty cycle of $60\%$) curves in Figure 8, which show two peaks at $P_{\rm var}\approx 11\%$ and $8\%$, respectively, indicating that the most probable $P_{\rm var}$ value is $11^{+9}_{-3}\%$ ($8^{+7}_{-2}\%$) when the duty cycle of the extremely X-ray weak state is $30\%$ ($60\%$); the $1\sigma$ uncertainties on the $P_{\rm var}$ values were derived from the L distributions. Thus the fraction of extremely variable X-ray sources among NLS1s is estimated to be $\approx(8\textrm{--}11)\%$. The above estimate of $P_{\rm var}$ depends on the uncertain estimate of the duty cycle (D) of the extremely X-ray weak state, but the dependence is not very strong. For example, in the extreme cases of $D=10\%$ and $D=100\%$, which are unlikely given the current observations of the limited sample, the corresponding $P_{\rm var}$ values are $24\%$ and $7\%$, respectively. Thus, likely only a small fraction of NLS1s are extremely X-ray variable.

Subsequently, we investigated the fraction of extremely variable X-ray sources among the SEAMBHs in the RM campaign conducted by Du et al. (2014, 2015, 2018), which is the parent sample including SDSS $\rm J0751+2914$. There are 24 SEAMBHs in this RM campaign, of which 20 have archival X-ray observations. One object, Mrk 486 (PG 1535+547), is an X-ray weak quasar that shows extreme X-ray spectral variability (e.g., Schartel et al., 2005; Ballo et al., 2008). It was classified as a mini-BAL quasar (Brandt et al., 2000; Sulentic et al., 2006), so that its extreme X-ray behavior is likely related to the outflowing wind (e.g., Giustini, 2016). We thus excluded it from this SEAMBH sample. Four objects (Mrk 335, Mrk 142, Mrk 493, and Mrk 1044) of the remaining 19 objects are also in the NLS1 sample of Grupe et al. (2010) discussed above, and we adopted their X-ray analysis results. Only Mrk 335 among these four objects has been found to show extreme variability in X-rays. We then derived the X-ray properties of the other 15 AGNs (including 14 quasars and one NLS1). The details of the data analysis will be presented in H. Liu et al (in preparation). Among these 15 AGNs, three objects (SDSS $\rm J0751+2914$, Mrk 382, and IRASF $12397+3333$) have three, five, and three observations, respectively, and the other 12 objects have been observed only once. SDSS $\rm J0751+2914$ is the only object among these 15 SEAMBHs found to show extreme X-ray variability. The other two objects with multiple observations show little X-ray variability, and we adopted their mean $\alpha_{\rm OX}$ values. Figure 9 shows the $\alpha_{\rm OX}$ versus $L_{2500~{}{\textup{\AA}}}$ distribution for the 15 objects. Thus two objects (Mrk 335 and SDSS $\rm J0751+2914$) in the sample of 19 SEAMBHs have ever shown extreme X-ray variability. Based on a process similar to that described above, we estimated that the most probable $P_{\rm var}$ value is $24^{+24}_{-8}\%$ ($15^{+16}_{-5}\%$) when the duty cycle of the extremely X-ray weak state is $30\%$ ($60\%$). Thus the fraction of extremely variable X-ray sources among SEAMBHs is $\approx(15\textrm{--}24)\%$. We note that this fraction is not very consistent with the fraction ($8\textrm{--}11\%$) for the NLS1 sample above, probably due to the different selection criteria of the SEAMBH and NLS1 samples which did not yield consistent populations of AGNs with high accretion rate. In the extreme cases of $D=10\%$ and $D=100\%$, the corresponding $P_{\rm var}$ values for the SEAMBH sample are $58\%$ and $11\%$, respectively.

We investigated if extremely X-ray variable AGNs also are outliers in terms of other physical properties. We first compared the dimensionless accretion rates of the two extremely variable AGNs discovered (Mrk 335 and SDSS $\rm J0751+2914$) to those of other AGNs in the SEAMBH sample. The $\rm log~{}\dot{\mathscr{M}}$ values of these SEAMBHs span a range of 0.55–2.98 (see Du et al., 2014, 2015, 2016, 2018). Mrk 335 and SDSS $\rm J0751+2914$ both have moderate dimensionless accretion rates ($\rm log~{}\dot{\mathscr{M}}=1.28~{}and~{}1.45$) among all the SEAMBHs. They do not have extreme BH masses, $\rm H\beta$ FWHMs, or optical luminosities either. However, we caution that the estimated BH masses and accretion rates may have large uncertainties (see Section 1). These results and the $\approx(15\textrm{--}24)\%$ occurrence rate of extreme X-ray variability among SEAMBHs suggest that, although the high accretion rate may be a key factor for the extreme X-ray variability in AGNs, there should be some factor, other than the BH mass, $\rm H\beta$ FWHM, or optical luminosity, that also influences the X-ray variability.

4.3. Possible Scenarios for Extreme X-ray Variability

The extreme X-ray variability of SDSS $\rm J0751+2914$ and other similar AGNs is unlikely the intrinsic variability of typical AGNs which rarely exceeds a variability factor of 200% (see Section 1). There are probably physical causes for the observed extreme phenomenon. Here we discuss two popular scenarios, the reflection and partial covering absorption models, frequently adopted to explain the extreme X-ray variability of AGNs.

The reflection model proposed by Ross & Fabian (2005) has been applied to explain the X-ray properties of many extremely variable NLS1s (e.g., 1H $0707\textrm{--}495$: Fabian et al. 2004, 2012; Mrk 335: Grupe et al. 2007, 2008; Gallo et al. 2015; NGC 4051: Ponti et al. 2006; IRAS $13224-3809$: Ponti et al. 2010; Jiang et al. 2018) and luminous quasars (e.g., PG $0844+349$: Gallo et al. 2011; PHL 1092: Miniutti et al. 2012). It proposes that in addition to the continuum emission observed directly, part of the primary X-ray emission from the corona is reflected to the line of sight by the accretion disk. The observed X-ray variability is the result of changes in the height of the corona (simplified as a point source above the BH in the ”lamppost” geometry; e.g., Fabian & Vaughan 2003; Miniutti et al. 2003). As the X-ray point source approaches the BH, the gravitational light bending gets stronger and it reduces the number of X-ray photons reaching the observer (Miniutti & Fabian, 2004). More primary X-ray power-law photons are reflected by the inner disk, so that the ratio of the reflection flux to the observed power-law continuum flux is larger. In the low state, the $\rm<10~{}keV$ spectrum is dominated by a soft-excess component and a broad Fe K$\alpha$ emission line at $\sim 6\textrm{--}7~{}$keV, with a flat spectral shape and curvature emerging in the $\rm\approx 2\textrm{--}6~{}keV$ band. The observed X-ray and multiwavelength properties of SDSS $\rm J0751+2914$ are generally consistent with this scenario, although the low-state spectrum is likely dominated by the soft excess component and we did not observe the broad Fe K$\alpha$ emission due to the limited photon counts.

Although the reflection model usually describes well the spectra of extremely variable AGNs, one main caveat is that the disk thickness is assumed to be very small and negligible compared to the height of the corona. However, the vertical structure of the accretion disk should not be neglected for AGNs with $\lambda_{\rm Edd}\gtrsim 0.1$, at which the accretion flow may become advection dominated in the radial direction, and the inner disk becomes geometrically thick (e.g., Abramowicz et al., 1980, 1988; Wang & Netzer, 2003; Ohsuga & Mineshige, 2011; Wang et al., 2014a; Jiang et al., 2016, 2017). Such a thick disk may produce a reflected spectrum different from that from a thin disk, and it is also likely to obscure the X-rays from the corona when the inclination angle is large (e.g., Luo et al. 2015; Ni et al. 2018; Taylor & Reynolds 2018; and references therein). Therefore, for extremely X-ray variable AGNs that typically have high accretion rates ($\lambda_{\rm Edd}\gtrsim 0.1$), the geometry of the accretion disk probably needs to be considered in the modeling of the X-ray variability.

The partial covering absorption scenario (e.g., Tanaka et al., 2004; Turner et al., 2009; Miniutti et al., 2012) depicts that the observed X-ray variability is attributed to the variation of the covering factor, ionization, and column density of the absorber. Under this scenario, the unabsorbed part of the primary X-ray emission dominates the $\lesssim 2$ keV soft X-rays, and the absorbed component dominates the hard X-rays. This model can also explain the X-ray variability of SDSS $\rm J0751+2914$. The origin and physics of the partial covering absorption material are not well understood. Since there is no coordinated UV/optical variability in general, the partial covering absorber must be located within the BLR and close to the BH, otherwise it may absorb UV/optical photons.

4.4. Partial Covering Absorption by A Thick Disk/Outflow

Since the extreme X-ray variability of AGNs appears to be associated with high accretion rates (Section 4.1), we consider that a geometrically thick accretion disk and its associated dense outflow may serve as the partial covering absorber for blocking the central X-ray emission. For AGNs with high accretion rates, the inner accretion disk is expected to be geometrically thick, and such an accretion disk is likely to produce a strong outflowing wind (e.g., Ohsuga & Mineshige, 2011; Takeuchi et al., 2014; Jiang et al., 2016, 2017). The thick disk and its associated outflow can absorb partially the X-ray emission when the inclination angle is large (see Figure 18 in Luo et al. 2015 and Figure 1 in Ni et al. 2018). As the height/size of the corona changes like in the reflection scenario, the covering factor of the thick disk/outflow with respect to the X-ray corona changes accordingly, resulting in the observed X-ray variability. The thick disk/outflow does not affect the observed UV/optical continuum or emission lines. This model can explain the X-ray and multiwavelength properties of SDSS $\rm J0751+2914$ and other AGNs with extreme X-ray variability. It also naturally explains the small occurrence rate of extremely X-ray variable AGNs with high accretion rates, as only AGNs with a line of sight close to the edge of the thick disk/outflow may experience variable partial covering X-ray absorption when the corona height/size changes. We note that if the absorber is the outflow associated with the inner disk, it must be relatively compact and cling to the disk; otherwise, the fraction of extremely X-ray variable AGNs would be much larger.

4.4.1 Connections to Weak Emission-Line Quasars

Our proposed scenario above shares the same basic nature as that for weak emission-line quasars (WLQs) in Luo et al. (2015), and WLQs are generally considered to have high accretion rates (e.g., Luo et al., 2015; Ni et al., 2018; Marlar et al., 2018). Therefore, WLQs and high accretion-rate AGNs with extreme X-ray variability are probably connected. As proposed by Luo et al. (2015), the inner puffed-up disk in a rapidly accreting WLQ could block the nuclear ionizing emission from reaching the the BLR, which results in the observed weak high-ionization UV emission lines (e.g., C IV). We investigated the UV emission lines of the extremely X-ray variable NLS1s and quasars listed in Table 4. PHL 1092 exhibits a weak, blueshifted, and asymmetric C IV emission line similar to those in WLQs (e.g., Miniutti et al., 2012). Two NLS1s, IRAS $13224-3809$ and 1H $0707-495$, also show weak C IV emission lines with equivalent widths (EW) less than $15~{}{\textup{\AA}}$ (Leighly & Moore, 2004), satisfying the C IV EW criterion for WLQs (Ni et al., 2018). The other three objects (PG $1211+143$, PG $0844+349$, and Mrk 335) do not show weak C IV lines (e.g., Baskin & Laor, 2005; Wu et al., 2009; Tang et al., 2012). As the emission-line strength is influenced by many factors, including anisotropic continuum and line emission, gas metallicity, and BLR geometry (e.g., Luo et al., 2015, and references therein), it is probably not surprising to observe typical C IV line strengths in a significant fraction of high accretion rate AGNs with thick inner accretion disks. A UV spectrum is needed to check if SDSS $\rm J0751+2914$ has a weak C IV emission line similar to those of PHL 1092 and WLQs.

Considering that WLQs and AGNs with extreme X-ray variability likely share the same nature, we expect that some WLQs with large inclination angles would also vary extremely in X-rays. Specifically, all the WLQs that were observed to be extremely X-ray weak could be extremely X-ray variable. Unfortunately, most of the extremely X-ray weak WLQs have only been observed once and we cannot assess their variability. Moreover, if the variability timescale scales with the BH mass, it would take a much longer time to detect X-ray variability in WLQs with BH masses that are typically one order of magnitude larger than those of the extremely X-ray variable quasars listed in Table 4. Among the 32 WLQs in the representative sample of Ni et al. (2018), there are two extremely X-ray weak ($f_{\rm weak}>10$, $\Delta\alpha_{\rm OX}<-0.384$) WLQs plus 10 X-ray undetected WLQs that could also be extremely X-ray weak. The fraction ($2/32\textrm{--}12/32$) is in general consistent with the fraction of extremely X-ray variable AGNs among high accretion rate AGNs, supporting a common origin for these extreme phenomena.

5. summary and future work

In this paper, we report the discovery of extreme X-ray variability in a type 1 quasar: SDSS J$0751+2914$. It is powered by a super-Eddington accreting BH with a mass of $\sim 1.6\times 10^{7}~{}M_{\odot}$. Based on archival observations, we have constrained its X-ray and UV/optical properties in different epochs. SDSS J$0751+2914$ shows extreme X-ray variability by a factor of larger than 10, and it lacks significant UV/optical variability coordinated with its X-ray variability. We also investigated other extremely X-ray variable AGNs with similar properties in the literature. Our main results are as follows:

In general, a single power-law model modified by Galactic absorption describes well the 0.3–10 keV spectra of the three X-ray observations. The spectral fitting yielded three steep power-law photon indices, $\rm\Gamma_{s}=2.89^{+0.14}_{-0.13}$, $3.07^{+0.37}_{-0.37}$, and $3.06^{+0.18}_{-0.17}$, for the high, intermediate, and low X-ray flux states, respectively. See Section 3.1. 2. 2.

Between 2013 May 27 and 2015 May 4 the observed 0.5–2 keV flux of SDSS $\rm J0751+2914$ dropped by a factor of $21.6\pm 2.4$. Since its UV flux shows little change in this period, it became extremely X-ray weak in 2015 May with a steep X-ray-to-optical power-law slope ($\alpha_{\rm OX}$) of $-1.97$, corresponding to an X-ray weakness factor of $36.2$ at rest-frame 2 keV. See Sections 3.2 and 3.3. 3. 3.

The optical continuum and emission lines of SDSS J0751+2914 show little change between the high and low X-ray flux states, which indicates an almost constant accretion rate. See Section 3.4. 4. 4.

Most of the extremely X-ray variable AGNs reported in the literature are NLS1s and NL type 1 quasars that have high accretion rates. But only a small fraction of such objects are extremely X-ray variable by factors of more than 10. The fractions of extremely variable X-ray sources among NLS1s (Grupe et al., 2010) and SEAMBHs (Du et al., 2014, 2015, 2018) are estimated to be $\approx 8\textrm{--}11\%$ and $\approx 15\textrm{--}24\%$, respectively. See Sections 4.1 and 4.2 5. 5.

We reviewed the reflection and partial covering absorption models, frequently applied to explain the extreme X-ray variability of NLS1s. Either model can explain the overall observational data for SDSS J$0751+2914$. We further propose that a thick accretion disk and its associated outflow can serve as the absorber in the partial covering absorption scenario. This model can explain the X-ray and multiwavelength properties of SDSS J$0751+2914$ and other AGNs with extreme X-ray variability. It also explains naturally the small fraction of extremely X-ray variable AGNs among AGNs with high accretion rates, as only AGNs with a line of sight close to the edge of the thick disk/outflow may experience variable partial covering X-ray absorption when the height/size of the corona changes. We also discuss the connections between extremely X-ray variable AGNs and WLQs. See Sections 4.3 and 4.4.

We tried to piece together a complete picture of the population of extremely X-ray variable AGNs by exploring the archival data and results for SDSS J$0751+2914$ and other NLS1s and NL type 1 quasars showing similar properties. However, previous observations and studies were mainly focused on individual objects, and there are a lot of uncertainties when we tried to understand the nature of these objects as a unified population. For example, the duty cycle of the extremely X-ray weak state and the occurrence rate of extreme X-ray variability among AGNs with high accretion rates that we estimated in the current paper have substantial uncertainties, which affect our interpretation of the physical nature. We consider several possible future efforts below that may help constrain better the properties of SDSS J$0751+2914$ and other similar AGNs, which would ultimately help us understand better the central engine of accreting BHs.

Multi-epoch monitoring observations with XMM-Newton or Chandra are required to obtain a longer term X-ray light curve of SDSS J$0751+2914$, for the purpose of constraining its duty cycle of the extremely X-ray weak state. Also, considering that its optical flux has significantly increased in the 2016–2017 RM monitoring period, it is of interest to investigate its current X-ray state. In addition, we estimate that a deeper XMM-Newton observation on SDSS J$0751+2914$ with $\approx 70$ ks exposure time could reveal a flat/hard $>2$ keV X-ray spectrum similar to the other extremely X-ray variable AGNs (see Section 4.1), if it is still in the extremely X-ray weak ($f_{\rm weak}\sim 36$) state. A NuSTAR observation is also required to examine if there is a hump in the $\approx 20\textrm{--}30~{}\rm keV$ spectrum (see Section 4.1).

We proposed a connection between extremely X-ray variable AGNs and WLQs. A UV spectrum of SDSS J$0751+2914$ will be able to determine whether it has any weak UV emission lines (e.g., C IV). Moreover, a UV spectroscopic survey of extremely X-ray variable AGNs (including exploring archival data) will allow us to examine such a connection systematically. A UV spectrum of SDSS J$0751+2914$ will also help us to rule out the possibility that it is a BAL quasar.

We investigated the X-ray variability of the NLS1s in Grupe et al. (2010), and we adopted their X-ray analysis results to estimate the occurrence rate of extreme X-ray variable NLS1s. New archival observations have become available for some of these NLS1s since Grupe et al. (2010). These data will help to constrain the fraction of extremely X-ray variable AGNs with greater certainty. In addition, a systematic multi-epoch X-ray survey on the SEAMBHs is required to discover more extremely variable X-ray sources and constrain better the occurrence rate of extreme X-ray variability among this population.

We thank the referee for the helpful comments and suggestions. We thank Yanmei Chen, Qiusheng Gu, and Zhiyuan Li for helpful discussions. We acknowledge financial support from the National Natural Science Foundation of China grant 11673010 (H.L., B.L.), National Key R&D Program of China grant 2016YFA0400702 (H.L., B.L.), National Thousand Young Talents program of China (B.L.). W.N.B. acknowledges financial support from NASA ADP grant 80NSSC18K0878 and CXC grant GO6-17083X.