An Extreme X-Ray Variability Event of a Weak-line Quasar

We list in Table 1 the two X-ray observations of SDSS J1539+3954. These Chandra observations were performed with the Advanced CCD Imaging Spectrometer spectroscopic array (ACIS-S; Garmire et al. 2003) in VFAINT mode. The new Chandra Cycle 21 observation is a part of our program to obtain deeper X-ray coverage for a set of WLQs (PI: W. N. Brandt).

Table 1.  X-Ray Observations and Derived Properties of SDSS J1539+3954

Observation	Observation	Exposure	Soft-band	Hard-band	Band	Γeff	${f}_{2\mathrm{keV}}$	${\alpha }_{\mathrm{ox}}$	${\rm{\Delta }}{\alpha }_{\mathrm{ox}}$(σ)
ID	Start Date	Time	Counts	Counts	Ratio
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
14948	2013 Dec 13	5.3	<2.4	<2.5	⋯	⋯	<0.83	<−2.18	<−0.46 (3.17)
22528	2019 Sep 12	7.3	${29.1}_{-5.5}^{+6.6}$	${15.2}_{-4.1}^{+5.3}$	${0.52}_{-0.14}^{+0.21}$	${2.0}_{-0.4}^{+0.4}$	16.58	−1.68	$0.04\,(0.26)$

Note. (1) Chandra observation ID. (2) Observation start date. (3) Background-flare cleaned effective exposure time in the 0.5–8 keV band in units of ks. (4)/(5) Aperture-corrected source counts in the soft (0.5–2 keV)/hard (2–8 keV) band. An upper limit at a 90% confidence level is given if the source is not detected. (6) Ratio between the soft-band and hard-band counts. "⋯" indicates that the source is undetected in both bands. (7) 0.5–8 keV effective power-law photon index. "⋯" indicates that ${{\rm{\Gamma }}}_{\mathrm{eff}}$ cannot be constrained. (8) Rest-frame 2 keV flux density in units of 10⁻³² erg cm⁻² s⁻¹ Hz⁻¹. (9) Measured ${\alpha }_{\mathrm{ox}}$ values. (10) Difference between the measured ${\alpha }_{\mathrm{ox}}$ and the expected ${\alpha }_{\mathrm{ox}}$ from the Just et al. (2007) ${\alpha }_{\mathrm{ox}}$–${L}_{2500\mathring{\rm A} }$ relation. The statistical significance of this difference measured following Table 5 of Steffen et al. (2006) is given in parentheses.

Download table as:  ASCIITypeset image

We processed the Chandra data using the Chandra Interactive Analysis of Observations (CIAO) tools (Fruscione et al. 2006), following the steps in Luo et al. (2015) and Ni et al. (2018). We first used the CHANDRA_REPRO script and the DEFLARE script to create the cleaned event file, and then created images in the 0.5–2 keV (soft), 2–8 keV (hard), and 0.5–8 keV (full) bands. Source positions are determined by WAVDETECT. In case of non-detection (see Section 3 of Ni et al. 2018 for the definition of non-detection), we adopt the SDSS position. We performed aperture photometry in both the soft band and hard band: source counts were extracted from a 2'' radius circular aperture centered on the source position, and background counts were extracted from a source-free annular region with a 10'' inner radius and a 40'' outer radius. When the source is undetected, the upper limits on the source counts were derived using the 90% confidence-level table in Kraft et al. (1991). As we can see in Figure 1(a), SDSS J1539+3954 was not detected in the first 5.3 ks Chandra observation, and was detected with ≈44 counts in total in the second 7.3 ks Chandra observation.

Zoom In Zoom Out Reset image size

We list in Table 2 all the optical spectroscopic observations of SDSS J1539+3954. SDSS J1539+3954 was observed by SDSS-I/II on 2004 June 15, and the Baryon Oscillation Spectroscopic Survey of SDSS-III (BOSS; Dawson et al. 2013) on 2012 April 29. The reduced SDSS and BOSS spectra were downloaded directly from the SDSS data archive.

After SDSS J1539+3954 was found to exhibit large X-ray variability, we promptly observed this object again with the Low-resolution Spectrograph-2 (LRS2; Chonis et al. 2014) on the Hobby–Eberly Telescope (HET; Ramsey et al. 1998). The observation with the blue arm of LRS2 (LRS2-B) was performed 8 days (≈2.7 days in the rest frame) after the Chandra observation, and the observation with the red arm of LRS2 (LRS2-R) was performed 12 days (≈4.1 days in the rest frame) after the Chandra observation. The LRS2 spectra were reduced with the HET pipeline panacea.¹⁶ The current version of the HET pipeline cannot correct for telluric absorption, and there are still some channel discontinuities. The flux calibration is estimated to have an uncertainty around 20% (G. Zeimann 2019, private communication). However, it is still sufficient for us to probe the basic properties of the UV continuum and emission lines. In Figure 2, we display all the spectra obtained for SDSS J1539+3954 together. They do not show noticeable variations.

Zoom In Zoom Out Reset image size

For the new Chandra detection (≈44 counts) of SDSS J1539+3954, we calculated the band ratio (the ratio of hard-band counts to soft-band counts) and its uncertainty with the code BEHR (Park et al. 2006). Assuming a power-law spectrum modified by Galactic absorption, we use the Portable, Interactive, Multi-mission Simulator (PIMMS)¹⁷ to derive the 0.5–8 keV effective power-law photon index (Γ_eff) and its uncertainty from the band ratio. We have also performed a power-law fit with XSPEC (Arnaud 1996) using the Cash statistic (Cash 1979) and obtained consistent results (see Figure 1(b) for the X-ray spectrum). The results are listed in Table 1. The Γ_eff value of SDSS J1539+3954 is consistent with that of typical luminous radio-quiet quasars (Γ_eff = 1.8–2.0; e.g., Reeves et al. 1997; Shemmer et al. 2005; Just et al. 2007; Scott et al. 2011). We then derived the unabsorbed soft-band flux (which covers rest-frame 2 keV) with PIMMS from the soft-band net count rate and Γ_eff, and calculated the rest-frame 2 keV flux density (f_{2 keV}) from the flux and Γ_eff. In the case of X-ray non-detection, we could set an upper limit on ${f}_{2\mathrm{keV}}$ following the same method using the upper limit for the soft-band net count rate and Γ_eff = 2.0 adopted from the case of X-ray detection. As can be seen in Table 1, f_2 keV varied by a factor of ≳20 between the two epochs. We note that different reasonable assumptions for the Γ_eff value (${{\boldsymbol{\Gamma }}}_{\mathrm{eff}}\approx 1.6$–2.4) in the case of X-ray non-detection will not substantially change the results, and the variation in ${f}_{2\mathrm{keV}}$ is always extreme (by a factor of ≳17–24).

In Figure 2, we can see that the UV continuum level and emission-line profiles of SDSS J1539+3954 remain generally unchanged from 2004 to 2019, considering the flux uncertainty of the HET spectrum. Our measurement of the redshift z = 1.935 ± 0.004 based on the Mg ii line in the SDSS and BOSS spectra is consistent with the value reported in Luo et al. (2015). Here, the adopted redshift is the average of two measurement results (from the two spectra); the uncertainty is the difference between two measurements. We measured the C iv REW, blueshift, and FWHM in the three different epochs, as C iv properties are found to be linked with X-ray weakness among typical quasars (e.g., Gibson et al. 2008; Timlin et al. 2020). For each spectrum, we performed both the local and global continuum fits following Yi et al. (2019). For the local continuum fit, we fit a power-law function to the two "anchor" regions located on the blue/red-wing ends of the C iv emission line; for the global continuum fit, we model the continuum with a reddened power-law function to fit spectral windows that are relatively free of emission/absorption features. After subtracting the fitted continuum, we fit the C iv emission with two Gaussian components to measure its REW, blueshift, and FWHM. The REW is calculated from the fitted profile normalized by the continuum; the blueshift is converted from the wavelength bisecting the cumulative total flux of the C iv emission; the FWHM is derived from the combination of the two fitted Gaussian components. The reported C iv REW, blueshift, and FWHM values are the mean values generated by the two different continuum-fitting methods, and the reported uncertainties are the combination of the uncertainty from Monte Carlo simulations and the difference between the two methods. The measurement results are shown in Table 2. We can see that the C iv emission-line properties do not have significant variations in the three different epochs. While the X-ray flux of SDSS J1539+3954 has experienced a dramatic increase, it remains a WLQ.

In Figure 3, we display the multiwavelength SED of SDSS J1539+3954, showing photometric data collected by the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010), Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006), SDSS, Galaxy Evolution Explorer (GALEX; Martin et al. 2005), and Chandra, as well as the Catalina Real-time Transient Survey (CRTS; Drake et al. 2009), Zwicky Transient Facility (ZTF; Bellm et al. 2019), and Near-Earth Object Wide-field Infrared Survey Explorer project (NEOWISE; Mainzer et al. 2014). CRTS monitored SDSS J1539+3954 from 2005 July 1 to 2013 September 28 (which is ≈2.5 months before the first Chandra observation of SDSS J1539+3954), and the results are reported in the V band (≈1870 Å in the rest frame); the Data Release 2 of ZTF includes the light curves of SDSS J1539+3954 from 2018 March to 2019 June in both the g band and the r band (≈1610 Å and 2160 Å in the rest frame); the 2019 data release of NEOWISE contains the light curves of SDSS J1539+3954 from 2014 to 2018 at 3.4 μm and 4.6 μm (≈1.1 and 1.6 μm in the rest frame). No remarkable flux variation is detected by CRTS, ZTF, or NEOWISE, though we do observe modest variability in the ZTF and NEOWISE light curves (e.g., the ZTF light curves show variability of ≲0.1 mag).¹⁸ The IR-to-UV SED of SDSS J1539+3954 is similar to those of typical quasars. The X-ray luminosity of SDSS J1539+3954 from the first Chandra observation is much lower than that of the composite SED of luminous quasars (Richards et al. 2006); the X-ray luminosity from the recent Chandra observation is roughly consistent with the composite SED.

Zoom In Zoom Out Reset image size

We measured the ${\alpha }_{\mathrm{ox}}$ parameter for SDSS J1539+3954 in the two different X-ray epochs (see Footnote 1 for the equation). As discussed before, the UV luminosity of SDSS J1539+3954 does not exhibit significant variability during the long-term monitoring, and the spectroscopic observations of SDSS J1539+3954 in three different epochs show a roughly consistent UV continuum level. Thus, when calculating ${\alpha }_{\mathrm{ox}}$, we adopt the ${f}_{2500\mathring{\rm A} }$ value measured from the SDSS spectrum (Shen et al. 2011) for the two different X-ray epochs consistently. With the ${f}_{2\mathrm{keV}}$ value/upper limit obtained in Section 3.1, we can calculate the ${\alpha }_{\mathrm{ox}}$ value/upper limit for the two epochs. From the empirical ${\alpha }_{\mathrm{ox}}$–L_2500 Å relation for typical quasars (e.g., Just et al. 2007), we derived an expected value of ${\alpha }_{\mathrm{ox}}$ for SDSS J1539+3954. Then, we measured the difference between the observed ${\alpha }_{\mathrm{ox}}$ and the expected ${\alpha }_{\mathrm{ox}}$ as ${\rm{\Delta }}{\alpha }_{\mathrm{ox}}$. The ${\alpha }_{\mathrm{ox}}$ and ${\rm{\Delta }}{\alpha }_{\mathrm{ox}}$ results are listed in Table 1. ${\rm{\Delta }}{\alpha }_{\mathrm{ox}}$ provides a measurement of X-ray weakness compared with the expected X-ray flux, as the ratio between the observed X-ray flux and expected X-ray flux is simply ${10}^{{\rm{\Delta }}{\alpha }_{\mathrm{OX}}/0.384}$. The first Chandra observation of SDSS J1539+3954 demonstrates that it is X-ray weak by a factor of ≈16, and the second Chandra observation shows that SDSS J1539+3954 now follows the empirical ${\alpha }_{\mathrm{ox}}$–${L}_{2500\mathring{\rm A} }$ relation well (lying ≈ 0.3σ above the relation).¹⁹

The extreme X-ray variability discovered from SDSS J1539+3954 is the second example of extreme X-ray variability among WLQs. PHL 1092 is a z = 0.40 radio-quiet quasar with log L_bol ≈ 46.65, and it was the first WLQ found to exhibit extreme X-ray variability (by a factor of ≈260; e.g., Miniutti et al. 2012). As summarized in Liu et al. (2019), at high luminosities (L_bol > 10⁴⁶ erg s⁻¹), PHL 1092 was the only radio-quiet non-BAL quasar known to show such extreme X-ray variability, and like SDSS J1539+3954 it varied between an X-ray weak state and an X-ray normal state. SDSS J1539+3954 has log L_bol ≈47.17 (Shen et al. 2011), which is even more luminous (by a factor of ≈3) than PHL 1092. The discovery of extreme X-ray variability from SDSS J1539+3954, combined with the PHL 1092 results, suggests that weak UV emission lines may be a good indicator for finding extreme X-ray variability events among luminous radio-quiet quasars.

We have proposed a thick inner accretion-disk model to explain the multiwavelength properties of WLQs (e.g., Figure 1 of Ni et al. 2018), which also has the potential to explain this extreme X-ray variability event. Simulations and analytical models suggest that for quasars with high Eddington ratios, geometrically thick inner accretion disks with high column densities are expected (e.g., Abramowicz et al. 1988; Jiang et al. 2014, 2019; Wang et al. 2014). Dense outflows arise with these thick disks in simulations (e.g., Jiang et al. 2014, 2019), and we implicitly include any associated dense outflow within the term "thick disk" throughout. The thick inner accretion disk can prevent ionizing X-ray/EUV photons from reaching an equatorially concentrated high-ionization broad emission-line region (BLR), while the UV/optical photons from the disk itself remain unobscured. Thus, weak high-ionization emission lines are observed. When our line of sight (to the central X-ray source) intercepts the thick inner accretion disk for a given WLQ, we observe an X-ray weak WLQ; when it misses this shield, we observe an X-ray normal WLQ.

As predicted in Section 6.2 of Luo et al. (2015), in the context of this model, the observed extreme X-ray variability event could be caused by a slight change in the thickness of the disk that moved across our line of sight.²⁰ This could arise due to rotation of a thick inner disk that is somewhat azimuthally asymmetric, or alternatively due to small changes in the inner-disk structure itself (e.g., see Figure 3 of Jiang et al. 2019). If the change in the disk thickness is slight, there will be no significant change in the BLR illumination and UV/optical photon emission to affect the UV emission-line and continuum properties.²¹

Ross & Fabian (2005) have also proposed a reflection model that has the potential to explain this extreme X-ray variability event, where the X-ray variability is driven by gravitational light-bending effects. Such effects can change the amount of X-ray emission reaching an observer when the distance between the primary X-ray source and the central black hole changes, causing the observed X-ray variability (Miniutti & Fabian 2004). However, in the context of this reflection model, it is not clear why extreme X-ray variability would have any particular link with WLQs.