Broadband X-ray Spectral and Timing Analyses of the Black Hole Binary Candidate Swift J1658.2-4242: Rapid Flux Variation and the Turn-on of a Transient QPO

TL;DR

This study analyzes rapid flux changes and transient QPOs in the black hole candidate Swift J1658.2-4242, revealing intrinsic accretion flow changes and weak disk reflection, advancing understanding of black hole accretion dynamics.

Contribution

It presents the first detailed spectral and timing analysis of Swift J1658.2-4242 during a rapid flux decrease and QPO emergence, highlighting accretion disk instabilities as the cause.

Findings

Rapid 45% flux decrease in 40 seconds.

Transient 6-7 Hz QPO appears during flux change.

Weak relativistic disk reflection compared to previous observations.

Abstract

We report results from joint NuSTAR, Swift and XMM-Newton observations of the newly discovered black hole X-ray binary candidate Swift J1658.2-4242 in the intermediate state. We observe a peculiar event in this source, with its X-ray flux rapidly decreasing by $\sim$45\% in $\sim$40~s, accompanied by only subtle changes in the shape of the broadband X-ray spectrum. In addition, we find a sudden turn-on of a transient QPO with a frequency of $6-7$~Hz around the time of the flux change, and the total fractional rms amplitude of the power spectrum increases from $\sim$2\% to $\sim$10\%. X-ray spectral and timing analyses indicate that the flux decrease is driven by intrinsic changes in the accretion flow around the black hole, rather than intervening material along the line of sight. In addition, we do not significantly detect any relativistic disk reflection component, indicating it is much weaker than previously observed while the source was in the bright hard state. We propose accretion disk instabilities triggered at a large disk radius as the origin of the fast transition in spectral and timing properties, and discuss possible causes of the unusual properties observed in Swift J1658.2-4242. The prompt flux variation detected along with the emergence of a QPO makes the event an interesting case for investigating QPO mechanisms in black hole X-ray binaries.

Tables

Table 1. Table 1 Power Spectra Properties

	$\nu$qpo	rmsqpo	$Q$	rmstot	${\chi }^2/\nu$
	(Hz)	($\%$)	($\nu$/FWHM)	($\%$)
High-flux Epoch
NuSTAR	$6.83\left(\mathrm{fixed}\right)$	¡2.4	$3.9\left(\mathrm{fixed}\right)$	${2.4}_{-0.8}^{+1.1}$	$0.94$
Low-flux Epoch
XMM	$6.47\pm 0.15$	$4.7\pm 0.4$	$4.3\pm 0.7$	$9.0\pm 0.2$	$1.00$
NuSTAR	$6.83\pm 0.09$	$5.9\pm 0.4$	$3.9\pm 0.7$	$10.8\pm 0.2$	$1.05$

Table 2. Table 2 Spectral Fitting Results: Part I

Component	Parameter	High-flux Epoch	Low-flux Epoch
Model 1: TBnew*(diskbb+cutoffpl) [NuSTAR]
tbnew	$N_H$ ($\times {10}^{23}{\mathrm{cm}}^{-2}$)	$1.60\pm 0.04$	$1.48\pm 0.03$
diskbb	$k{T}_{\mathrm{in}}$ (keV)	$1.46\pm 0.02$	$1.28\pm 0.01$
	Norm	$74\pm 5$	$84\pm 3$
cutoffpl	$\Gamma$	$2.09\pm 0.03$
	$E_{\mathrm{cut}}$ (keV)	$52\pm 4$
	Norm	$3.1\pm 0.2$	$1.7\pm 0.1$
	${\chi }^2/\nu$	$2151/1698=1.27$
Model 2: TBnew*gabs*(diskbb+cutoffpl+Gaussian) [Swift+NuSTAR]
tbnew	$N_H$ ($\times {10}^{23}{\mathrm{cm}}^{-2}$)	$1.75\pm 0.03$	$1.41\pm 0.02$
gabs	$E_{\mathrm{gabs}}$ (keV)	$7.09\pm 0.04$
	Norm	$\left(1.3\pm 0.4\right)\times {10}^{-2}$	$\left(1.3\pm 0.2\right)\times {10}^{-2}$
diskbb	$k{T}_{\mathrm{in}}$ (keV)	$1.44\pm 0.01$	$1.29\pm 0.01$
	Norm	$86\pm 4$	$78\pm 3$
cutoffpl	$\Gamma$	$2.07\pm 0.03$
	$E_{\mathrm{cut}}$ (keV)	$50\pm 4$
	Norm	$3.0\pm 0.2$	$1.6\pm 0.1$
Gaussian	$E_{\mathrm{gauss}}$ (keV)	$6.48\pm 0.04$
	Norm		$\left(9\pm 1\right)\times {10}^{-4}$
	${\chi }^2/\nu$	$2900/2580=1.12$
Model 3: TBnew*gabs*(diskbb+relxilllp+Gaussian) [Swift+NuSTAR]
tbnew	$N_H$ ($\times {10}^{23}{\mathrm{cm}}^{-2}$)	$1.84\pm 0.03$	$1.55\pm 0.02$
gabs	$E_{\mathrm{gabs}}$ (keV)	$7.09\pm 0.03$
	Norm	$\left({2.0}_{-0.4}^{+0.5}\right)\times {10}^{-2}$	$\left(1.8\pm 0.3\right)\times {10}^{-2}$
diskbb	$k{T}_{\mathrm{in}}$ (keV)	$1.38\pm 0.02$	$1.17\pm 0.03$
	Norm	${83}_{-8}^{+6}$	${103}_{-10}^{+7}$
relxilllp	$\Gamma$	${2.32}_{-0.03}^{+0.04}$
	$E_{\mathrm{cut}}$ (keV)	${91}_{-14}^{+17}$
	$h$ ($r_g$)	$18\pm 8$
	$a^{\ast }$ ($c$J/GM2)	$0.998\left(\mathrm{fixed}\right)$
	$R_{\mathrm{in}}$ ($r_g$)
	$A_{\mathrm{Fe}}$ (solar)	$0.91\left(\mathrm{fixed}\right)$
	$\theta$ (∘)	$64\left(\mathrm{fixed}\right)$
	log $\left(\xi \right)$	${3.7}_{-0.3}^{+0.4}$
	$R_{\mathrm{ref}}$	$0.4\pm 0.3$
	Norm (${10}^{-2}$)	${6.5}_{-1.1}^{+1.3}$	$3.5\pm 0.6$
Gaussian	$E_{\mathrm{gauss}}$ (keV)	$6.45\pm 0.06$
	Norm		$\left(7\pm 1\right)\times {10}^{-4}$
	${\chi }^2/\nu$	$2837/2576=1.10$
$F_{\mathrm{disk}}$ (erg cm-2 s-1)a		$5.66\times {10}^{-9}$	$3.09\times {10}^{-9}$
$F_{\mathrm{powerlaw}}$ (erg cm-2 s-1)a		$1.19\times {10}^{-8}$	$6.51\times {10}^{-9}$
$F_{\mathrm{total}}$ (erg cm-2 s-1)a		$1.75\times {10}^{-8}$	$9.60\times {10}^{-9}$

Table 3. Table 3 Spectral Fitting Results: Part II

Component	Parameter	High-flux Epoch	Low-flux Epoch
Slab corona geometry: TBnew*gabs*(diskbb+compPS+Gaussian)
tbnew	$N_H$ ($\times {10}^{23}{\mathrm{cm}}^{-2}$)	$1.64\pm 0.02$	$1.33\pm 0.02$
gabs	$E_{\mathrm{gabs}}$ (keV)	$7.11\pm 0.04$
	Norm	$\left(1.3\pm 0.4\right)\times {10}^{-2}$	$\left(1.2\pm 0.2\right)\times {10}^{-2}$
diskbb	$k{T}_{\mathrm{in}}$ (keV)	$1.38\pm 0.01$	$1.23\pm 0.01$
	Norm	$108\pm 5$	$96\pm 3$
compPS	$\tau$	$1.6\pm 0.1$
	$k{T}_e$ (keV)	${24}_{-1}^{+2}$
	Norm	${169}_{-9}^{+16}$	${150}_{-8}^{+10}$
Gaussian	$E_{\mathrm{gauss}}$ (keV)	$6.48\pm 0.04$
	Norm		$\left(1.0\pm 0.1\right)\times {10}^{-3}$
	${\chi }^2/\nu$	$2901/2580=1.12$
Spherical corona geometry: TBnew*gabs*(diskbb+compPS+Gaussian)
tbnew	$N_H$ ($\times {10}^{23}{\mathrm{cm}}^{-2}$)	$1.63\pm 0.02$	$1.33\pm 0.02$
gabs	$E_{\mathrm{gabs}}$ (keV)	$7.11\pm 0.04$
	Norm	$\left(1.3\pm 0.4\right)\times {10}^{-2}$	$\left({1.3}_{-0.1}^{+0.2}\right)\times {10}^{-2}$
diskbb	$k{T}_{\mathrm{in}}$ (keV)	$1.38\pm 0.01$	$1.24\pm 0.01$
	Norm	$110\pm 5$	$98\pm 3$
compPS	$\tau$	$2.1\pm 0.2$
	$k{T}_e$ (keV)	$25\pm 2$
	Norm	$60\pm 5$	${54}_{-4}^{+5}$
Gaussian	$E_{\mathrm{gauss}}$ (keV)	${6.48}_{-0.02}^{+0.04}$
	Norm		$\left(1.0\pm 0.1\right)\times {10}^{-3}$
	${\chi }^2/\nu$	$2899/2580=1.12$

Full text

1. INTRODUCTION

Most galactic black hole X-ray binaries are discovered as transients, when they go into bright outbursts in the X-ray band. During typical outbursts, these systems are known to undergo undergo a transition from the low/hard to the high/soft state through relatively short-lived intermediate states, with these states exhibiting distinct spectral and timing properties (see Remillard & McClintock, 2006; Belloni & Motta, 2016, for reviews). Swift J1658.2–4242 is a new black hole X-ray binary candidate, first reported by Swift-BAT on 2018 February 16 (Barthelmy et al., 2018). Subsequent observations exhibiting dips in the light curve indicate that it is highly absorbed and is viewed at a high inclination angle (Beri et al., 2018; Lien et al., 2018; Xu et al., 2018b). Relativistic disk reflection features, including a broad asymmetric Fe K$\alpha$ line peaking at 6–7 keV and Compton reflection hump around 30 keV, were detected by NuSTAR when Swift J1658.2–4242 was in the low/hard state (Xu et al., 2018b). Relativistic reflection features in the X-ray spectrum of Galactic black hole binaries are believed to arise from reprocessing of hard X-ray emission from the corona by the inner accretion disk (Fabian et al., 1989; Reynolds, 2014). Detailed modeling of the reflection spectra of Swift J1658.2–4242 suggests that the central object in the system is a rapidly spinning black hole ($a^{*}>0.96$), and also supported the conclusion that we view the system at high inclination, finding $\theta=64^{+2\circ}_{-3}$ for the inner disc (Xu et al., 2018b).

Variability on a wide range of timescales is characteristic of the accretion process around both stellar-mass and supermassive black holes (e.g., Ulrich et al., 1997; van der Klis, 2006; McHardy, 2010). Rapid flux variations are commonly found in the X-ray light curves of black hole binaries, and are believed to probe the properties of the inner accretion flow in the vicinity of stellar-mass black holes. The power spectrum generated from the X-ray light curve of a black hole X-ray binary is typically characterized by a variable broadband noise component along with transient and discrete peaks superimposed on top of the continuum (e.g., van der Klis, 1989, 2006). The peaks are termed quasi-periodic oscillations (QPOs). Low-frequency QPOs observed in black hole binary systems, with typical frequencies from a few mHz to 30 Hz, can be generally classified into three main types (type-A, B and C), based on their characteristics in the power density spectrum (see Wijnands et al., 1999; Casella et al., 2004, 2005, for details). Type-A QPOs are characterized by a weak and broad peak in the power spectrum. The most common type of low-frequency QPOs are type-C QPOs, which are usually strong and span the frequency range of $\sim$0.1–30 Hz. Type-B QPOs are rarer, limited to a narrow frequency range (typically $\sim$5-6 Hz), and can be distinguished by the fractional amplitude and underlying noise shape. Type-B QPOs are usually observed during the time of transitions between hard-intermediate (HIMS) and soft-intermediate (SIMS) states (or transitions between hard and soft states), and have been proposed to be associated with major jet ejection events (e.g., Soleri et al., 2008; Fender et al., 2009). There is evidence that type-B QPOs are stronger in systems that are viewed close to face-on, while type-C QPOs are stronger in high-inclination systems (Motta et al., 2015), supporting the idea that the two types have different physical origins. The rapid transitions of QPOs have been found in a few black hole binaries or black hole candidates via time-resolved analysis: GS 1124–684 (Takizawa et al., 1997), XTE J1859+226 (Casella et al., 2004; Belloni et al., 2005), GX 339–4 (Nespoli et al., 2003), XTE J1550–564 (Sriram et al., 2013), XTE J1859+226 (Sriram et al., 2016), and also recently in MAXI J1535–571 (Huang et al., 2018), in most cases type-B QPOs are identified.

Despite the richness in observational phenomenology, the physical origins of QPOs are still highly uncertain. For low-frequency QPOs, the frequency is much longer than the dynamical timescales in the strong gravity regime of stellar-mass black holes. Different theoretical models have been put forward to provide a physical explanation for low-frequency QPOs: some models invoke geometric effects, e.g., Lense-Thirring precession of a radially extended section of the inner accretion flow (e.g, Stella et al., 1999; Ingram et al., 2009; Motta et al., 2018), which is a General-Relativity frame-dragging effect; other models consider intrinsic instabilities in the accretion flow, e.g., oscillations caused by standing shocks (Chakrabarti & Molteni, 1993) or magneto-acoustic wave propagation (Titarchuk & Fiorito, 2004; Cabanac et al., 2010). Similarly, various attempts have also been made to understand the broadband aperiodic noise continuum in the power spectrum (e.g., Lyubarskii, 1997; Arévalo & Uttley, 2006; Ingram & Done, 2010), with the propagation of fluctuating accretion being a popular model, however there is still no consensus on the origin. X-ray spectral-timing studies are believed to be a promising method to investigate the nature of QPOs.

In this work, we present X-ray spectral and timing analyses of the new black hole binary candidate Swift J1658.2–4242 in the intermediate state of its 2018 outburst, observed by NuSTAR, Swift and XMM-Newton. The paper is structured as follows: in Section 1, we describe details of the observations used in this work and the data reduction procedures; results from our X-ray timing and spectral analyses are present in Section 3 and Section 4, respectively; finally, we discuss the physical implications from the observations in Section 5, and summarize the paper in Section 6.

2. Observations and DATA REDUCTION

Nuclear Spectroscopic Telescope Array (NuSTAR; Harrison et al. 2013) and XMM-Newton (Jansen et al., 2001) jointly observed Swift J1658.2–4242 on 2018 February 25 (MJD 58175). The outburst of Swift J1658.2–4242 was also monitored daily by the X-ray Telescope (XRT; Burrows et al. 2005) on the Neil Gehrels Swift Observatory (Swift; Gehrels et al. 2004). The joint NuSTAR and XMM-Newton observations caught Swift J1658.2–4242 shortly after significant spectral softening occurred during the rising phase of outburst, close to the time of the hard-to-soft state transition (see Swift monitoring light curves in Figure 1(a)). Based on the hardness-intensity diagram (HID, Figure 1(c)), Swift J1658.2–4242 is consistent with being in an intermediate state, transitioning from a canonical hard to a canonical soft state, at a phase of an outburst when some black hole binaries are known to display repeated fast transitions back and forth involving HIMS and SIMS (e.g., Belloni & Motta, 2016).

2.1. NuSTAR

We reduced the NuSTAR data of Swift J1658.2–4242 (OBSID: 80301301002) using NuSTARDAS pipeline v.1.6.0 and CALDB v20180419. After standard data filtering with NUPIPELINE, the exposure times are 31.5 ks and 31.8 ks for the two focal plane modules, FPMA and FPMB, respectively. The source spectra and light curves were extracted from a circular region with the radius of 180$\arcsec$ centered on the location of Swift J1658.2–4242 using NUPRODUCTS. Background was estimated from source-free areas using polygonal regions. The NuSTAR spectra were grouped to have a signal-to-noise ratio (S/N) of at least 30 per bin.

2.2. XMM-Newton

The XMM-Newton observation of Swift J1658.2–4242 (OBSID: 0802300201) were processed using the XMM-Newton Science Analysis System (SAS) v17.0.0 following standard procedures. The EPIC-MOS1 and EPIC-MOS2 cameras were closed during the observation. The prime instrument we use is EPIC-pn (Strüder et al., 2001). The first part of the observation was carried out in the EPIC-pn Timing Mode, and was later switched to the Burst Mode (a special flavour of the Timing Mode with low live time) due to the unanticipated high count rate. We selected events with Pattern $\leqslant$ 4 (singles and doubles) and Quality Flag = 0. The source region was chosen as rows 20 $\leqslant$ RAWX $\leqslant$ 55 (Timing Mode), 25 $\leqslant$ RAWX $\leqslant$ 50 and RAWY $\leqslant$ 142 (Burst Mode). Corresponding background was extracted from rows 3 $\leqslant$ RAWX $\leqslant$ 6. The resulting exposure times after data filtering are 41 ks for the Timing Mode, and 630 s for the Burst Mode. The source light curves were corrected for dead time, PSF and other instrumental effects using the epiclccorr tool. Due to an apparent disagreement of XMM-Newton data in spectral slope with NuSTAR and Swift-XRT, we do not use XMM-Newton spectra for spectral analysis in this work. The spectral slope disagreement is likely to be associated with pile-up issues along with possible calibration uncertainties for bright targets observed by XMM-Newton in the Timing Mode, which cannot be completely solved by excising data from the core of the XMM-Newton PSF.

2.3. Swift

Two Swift-XRT observations of Swift J1658.2–4242 (OBSID: 00010571004, 00010571005) were taken close to the time of the joint NuSTAR and XMM-Newton observations, on 2018 February 25 and 26, respectively. The data were taken in the Windowed Timing mode. We extracted source spectra from a circular region with the radius of 70$\arcsec$, and the background was extracted from an annulus area with the inner and outer radii of 200$\arcsec$ and 300$\arcsec$. The averaged Swift-XRT count rates are $\sim$38 ct s*-1* and $\sim$23 ct s*-1* for the first and the second observation, respectively, thus the data are not affected by pile-up distortions. After standard data filtering, the exposure times are 857 s and 795 s for the first and second Swift-XRT observations. The Swift-XRT spectra were rebinned to have a S/N of least 5 per energy bin.

3. Variability

A dramatic decrease in the flux of Swift J1658.2–4242 was simultaneously observed by NuSTAR and XMM-Newton at 07:00:14 UTC on 2018 February 25 (around 1.27$\times 10^{4}$ s in Figure 2). The source count rate decreased by $\sim$35% in the XMM-Newton band (0.3–10 keV), and $\sim$40% in the NuSTAR band (3–79 keV) within only $\sim$40 s. The flux levels before and after this sudden flux variation were relatively stable, although increased variability can be seen in the XMM-Newton light curve right before the large flux drop around $1.27\times 10^{4}$ s (see Figure 2, right panel). The timing of the sharp count rate drop is aligned very well in the NuSTAR and XMM-Newton light curves, showing no noticeable energy-dependent time delay. We henceforth refer to the time intervals before and after the large flux drop as the high- and the low-flux epoch, respectively. The same trend of flux variation is also reflected in the long term Swift light curves of Swift J1658.2–4242 (Figure 1), where there is a flux difference of about 50% between the two Swift-XRT exposures taken shortly before and after our joint NuSTAR and XMM-Newton observations.

Despite the dramatic change in flux level, there is only subtle variation in the broadband X-ray spectral energy distribution. We extracted source light curves taken by NuSTAR and XMM-Newton in four different energy intervals, and calculated hardness ratios based on count rates in the corresponding energy bands, defined as: ${\rm HR1={C_{3-6~{}keV}/C_{0.3-3~{}keV}}}$, ${\rm HR2={C_{6-10~{}keV}/C_{3-6~{}keV}}}$ and ${\rm HR3={C_{10-79~{}keV}/C_{6-10~{}keV}}}$. The sudden drop in count rates around $1.27\times 10^{4}$ s is evident in all four energy bands, and the changes observed in hardness ratios are small. As shown in Figure 2, HR1 and HR2 decrease by $\sim$20% and $\sim$12%, respectively, at the time of the flux decrease, while HR3 remains constant. The small decrease in hardness ratios below 10 keV (HR1 and HR2) indicates that the source energy spectrum turns slightly softer when the flux is lower, ruling out the possibility that the flux variation is caused by photoelectric absorption from material temporarily obscuring the line of sight. In addition, the constant hardness ratio (HR3) in the hard X-ray band throughout the observations suggests that the non-thermal spectral shape of Swift J1658.2–4242 remains roughly the same despite of the large flux change.

For X-ray timing analysis, we generated power spectra of Swift J1658.2–4242 from NuSTAR and XMM-Newton data. We first applied barycenter corrections to the event files, transferring the photon arrival times to the barycenter of the Solar system using JPL Planetary Ephemeris DE-200. The NuSTAR power spectra were produced using MaLTPyNT in the energy band of 3–79 keV, with a light curve binning size of $2^{-6}$ s, and was averaged in 256 s intervals. The power spectrum generated by MaLTPyNT is the cross-power density spectrum of FPMA and FPMB, which helps to reduce deadtime distortions (Bachetti et al., 2015). We generated an XMM-Newton EPIC-pn power spectrum using the powspec tool in the XRONOS package (Stella & Angelini, 1992). The EPIC-pn power spectrum was calculated in the 0.3–10 keV band with a light curve time resolution of 0.01 s, and was averaged from 8 intervals. Both power spectra were generated using the root-mean-square (rms) normalization, and were geometrically rebinned by a factor of 1.03 to reach nearly equally spaced frequency bins in the logarithmic scale.

The dynamical NuSTAR power spectrum is displayed in Figure 3. The first two orbits of data lack any significant periodicity. A QPO appears in the power spectrum at \sim$$6-7 Hz from $\sim$1.3$\times 10^{4}$ s, coinciding with the time of the rapid flux decrease found in Figure 2. For further analysis, we generate NuSTAR and XMM-Newton power density spectra for the high- and low-flux epochs separately.

We fit the time-averaged power spectra of Swift J1658.2–4242 during the two epochs with a unity response file in XSPEC v12.9.0n (Arnaud, 1996). In this work, we perform all spectral fitting in XSPEC using $\chi^{2}$ statistics, and all parameter uncertainties are reported at the 90% confidence level for one parameter of interest unless otherwise clarified. The power density spectra of Swift J1658.2–4242 can be adequately fitted with a multi-Lorentzian model, which is commonly used for black hole binaries (Belloni et al., 2002). We use three Lorentzians to fit for the noise continuum, with the centroid frequency fixed at zero; one Lorentz function for the QPO, and one for the possible sub-harmonic with the frequency linked with half the fundamental QPO frequency.

Comparing the the time-averaged NuSTAR power spectra at the high- and low-flux epochs (see Figure 4), it is evident that the source variability increased significantly after the rapid flux drop: the total fractional rms variability (rmstot, integrated in the frequency range of 0.1–20 Hz) in the NuSTAR band increases from $2.4^{+1.0}_{-0.8}$% to $10.8\pm 0.2$%, and a QPO peak emerges in the power spectra along with increased noise continuum above 0.1 Hz.

The time-averaged QPO frequency measured at the low-flux epoch is $\nu_{\rm 3-79~{}keV}=6.83\pm 0.09~{}{\rm Hz}$ in the NuSTAR band, and is $\nu_{\rm 0.3-10~{}keV}=6.47\pm 0.15~{}{\rm Hz}$ in the XMM-Newton band. The quality factor $Q~{}(\nu/{\rm FWHM})$ of the QPO is $4.3\pm 0.7$ measured by NuSTAR, and is $3.9\pm 0.7$ measured by XMM-Newton, with the fractional rms amplitude, rmsqpo, of $5.9\pm 0.4\%$ and $4.7\pm 0.4\%$, respectively (see details in Table 1). The characteristics of the QPO are similar to those of type-B or type-C low-frequency QPOs in black hole binaries (e.g., Casella et al., 2004, 2005; Motta et al., 2015). Type-C QPOs are usually strong, but can also be weak when appear in the soft state (fractional rms amplitude $\sim$1–25%). They are variable in the frequency range of $\sim$0.1–30 Hz, and have a strong flat-top noise continuum in the power spectrum. Type-B QPOs are typically weaker (fractional rms amplitude $\sim$1–10%), characterized by a weak red noise continuum, and are usually detected in a narrow frequency range around 5–6 Hz. We note that considering that the QPO peak lies on top of a strong noise continuum in the power spectra in Figure 4, the low-frequency QPO detected in Swift J1658.2–4242 is most likely to be a type-C QPO. The lack of a type-B QPO here is unusual, as a switch between a type-B QPO and a type-A/type-C QPO or noise is expected during fast transitions between HIMS and SIMS (e.g., Belloni, 2010; Belloni & Motta, 2016). The QPO frequency and strength increase at higher energies, as can been seen by comparing the QPO properties in the two different instrument bands, which is typical for QPOs observed in black hole binaries (e.g., Rodriguez et al., 2002; Casella et al., 2004).

The QPO is absent in the high-flux epoch. By using a Lorentzian function to fit for the possible presence of a QPO in the high-flux epoch, with the centroid and width of the Lorentzian fixed at the low-flux values, we can put an upper limit on the fractional rms amplitude of the QPO of ¡ 2.4% (90% confidence level, in the 3–79 keV band of NuSTAR). We can rule out the possibility that the non-detection of QPO in the high-flux epoch is simply due to dilution by increased X-ray photons: the rms amplitude of the QPO in the low-flux epoch is 5.9%, the value would be 3.5% if diluted by extra 40% of non-variable photons, which exceeds the upper limit obtained. Therefore, the non-detection of a QPO during the start of the observation when the count rate is high is because of the intrinsic weakness of the QPO signal.

4. SPECTRAL ANALYSIS

For spectral modeling, we separately extracted NuSTAR energy spectra accumulated before and after the flux change. The NuSTAR exposure times during the high- and the low-flux epoch are $\sim$3.7 ks and $\sim$27.4 ks, respectively. We first fit the NuSTAR spectra in the two epochs jointly with a simple absorbed cutoff power-law model plus a thermal disk blackbody component, TBnew*(diskbb+cutoffpl) (Model 1), in XSPEC notation. The cutoffpl model is widely used to fit the non-thermal X-ray emission in black hole binaries. It is used as a phenomenological spectral description of the Comptonized emission generated in the corona in the vicinity of a black hole. Neutral absorption is accounted for by using the TBnew absorption model, with the cross-sections from Verner et al. (1996) and abundances from Wilms et al. (2000). In the case of Swift J1658.2–4242, neutral absorption is mostly intrinsic to the source, as the absorption column density, $N_{\rm H}$, measured by modeling the X-ray spectrum, greatly exceeds the Galactic value of $N_{\rm H,Gal}=1.55\times 10^{22}$ cm*-2* (Kalberla et al., 2005). All abundances are fixed at the Solar value in the TBnew model. Changes in the shape of the non-thermal spectra of Swift J1658.2–4242 before and after the large flux variation are minimal, as revealed by the constant values of HR3 (see Figure 2(e)). Therefore, we link the parameters of the cutoffpl model between epochs, including the power-law index, $\Gamma$, and the exponential high-energy cutoff, $E_{\rm cut}$, while allow the normalization of the cutoffpl model to vary independently.

As shown in Figure 5, Model 1 fits the spectral continuum well, with the reduced chi-squared of $\chi^{2}/\nu=2151/1698=1.27$ ($\nu$ is the number of degrees of freedom). Allowing the parameters of $E_{\rm cut}$ and $\Gamma$ to have different values at the two epochs brings no significant improvement to the fit. The prominent spectral residuals of Model 1 are a narrow Fe K$\alpha$ emission line at $\sim$6.4–6.5 keV, and a narrow absorption feature at $\sim$7.1 keV (see Figure 5). The narrow emission line is only seen after the drop in flux, while the absorption line is seen in both the high- and low-flux epochs.

In order to achieve broadband X-ray coverage, we further include two contemporaneous Swift-XRT spectra in the spectral modeling, so that the parameters sensitive to the soft X-ray band (i.e., absorption column density, $N_{\rm H}$, and disk blackbody temperature, $T_{\rm in}$) can be better constrained. We use NuSTAR spectra in the energy range of 4–79 keV and Swift-XRT spectra in 1.0–10.0 keV, following Xu et al. (2018b). We allow the cross-normalization constants to vary freely for NuSTAR/FPMB and Swift-XRT, and fix the value at unity for NuSTAR/FPMA. To improve the fit, we add a Gaussian absorption line model, gabs and a Gaussian emission line model, Gaussian, to account for the spectral residuals of Model 1 shown in Figure 5, with the total model set up in XSPEC as: TBnewgabs(diskbb+cutoffpl+Gaussian) (Model 2). For simplicity, we fix the line width, $\sigma$, of the gabs model at 0.1 keV and the width of the Gaussian model at 0.2 keV. During the joint spectral fitting, we link the centroid energies of gabs and Gaussian between epochs, while allow their corresponding strength to vary independently. The addition of these two extra model components reduces the reduced chi-squared of the fit ($\chi^{2}/\nu=2900/2580=1.12$), and leaves no visually evident spectral residuals (see Figure 6).

The shape of the broadband X-ray continuum is consistent with black hole binaries in the intermediate state (Remillard & McClintock, 2006; Belloni & Motta, 2016). The spectral fitting measures a soft power-law index ($\Gamma=2.07\pm 0.03$) and a high disk blackbody temperature ($kT_{\rm in}\simeq 1.2-1.4$ keV, the value varies between epochs), whereas the contribution of the thermal disk to the total unabsorbed flux in $0.1-500$ keV is only $\sim$32%, indicating that Swift J1658.2–4242 was yet to enter a canonical soft (thermal dominant) state. The changes in the shape of the broadband X-ray spectrum after the large flux decrease are subtle (see Figure 6(a)), which are reflected by the similar values of the best-fit parameters measured for the two epochs (see Table 2). The values of the inner disk temperature, $kT_{\rm in}$, are measured to be $1.44\pm 0.01$ keV and $1.29\pm 0.01$ keV for the high- and the low-flux epoch, respectively. The absorption column density, $N_{\rm H}$, slight decreases from $(1.75\pm 0.03)\times 10^{23}$ cm*-2* to $(1.41\pm 0.02)\times 10^{23}$ cm*-2* after the flux drop. The variation in hardness ratios described in Section 3 is caused by this simultaneous decrease in the inner disk temperature and the absorption column density. Via modeling the broadband X-ray spectra, we can confidently rule out increased photoelectric absorption as the origin of the large decrease in X-ray flux.

One key difference when comparing the NuSTAR spectra of Swift J1658.2–4242 in the intermediate state with that from an earlier observation in the hard state (Xu et al., 2018b), besides spectral softening, is the disappearance of strong relativistic reflection features (see Figure 5 for a comparison). Strong relativistic reflection features, including a broad and asymmetric Fe K$\alpha$ line and Compton reflection hump, are commonly found in bright black hole binaries during intermediate and soft states, enabling a measurement of the black hole spin in those systems (e.g., Tomsick et al., 2014; Walton et al., 2016; Miller et al., 2018). In the case of Swift J1658.2–4242, although relativistic disk reflection features were clearly detected in the bright hard state in Xu et al. (2018b), there is no clear indication for such a component on top of the Comptonization continuum in the intermediate state. Considering that the flux of the hard X-ray coronal emission in the intermediate state is comparable to that in the hard state, reflection features should be easily detectable in the intermediate state as long as the reflection fraction is comparably high. Therefore, the non-detection of relativistic reflection features in this work is because of the intrinsic weakness of the relativistic reflection component rather than the source being too faint.

Lacking prominent relativistic reflection features in the data, we cannot obtain a robust measurement of the inner accretion disk parameters via reflection spectral modeling. Thus, we simply calculate the upper limit of the strength of any reflection component that comes from the inner disk. To do so, we add a relativistic disk reflection component modeled by the relxilllp model (relxill v1.0.2; Dauser et al., 2014; García et al., 2014) to the fit, with the total model set up in XSPEC as: TBnewgabs(diskbb+relxilllp+Gaussian) (Model 3). The relxilllp model assumes a lamp-post geometry for the corona, which parameterizes the disk emissivity profile by the height of the corona, $h$, above the accretion disk. A cutoff power-law continuum is included in the relxilllp model, and is used as input for the reflection spectrum. To reduce the number of free parameters, we fix the black hole spin, $a^{*}=0.998$, the inclination of the accretion disk, $\theta=64^{\circ}$, and the iron abundance in the disk, $A_{\rm Fe}=0.91$, according to the best-fit values found in Xu et al. (2018b), as $a^{*}$ and $A_{\rm Fe}$ do not change between states and the $\theta$ is also usually assumed to remain constant. The best-fit reflection fraction measured by Model 3 is low, $R_{\rm ref}=0.4\pm 0.3$, confirming the lack of a strong relativistic reflection component in the spectra (see also in Figure 6(a), the contribution of the putative relativistic reflection component is weak when compared to the total spectra).

We note that although the addition of a relativistically blurred reflection component improves the fit slightly with four extra parameters by $\Delta\chi^{2}/\Delta\nu=63/-4$, we do not consider it as significantly detected. As shown in Figure 6, the addition of a relxill component in Model 3 only slightly reduces the residuals between 10 keV and 20 keV, which are likely to come from small spectral curvature not perfectly accounted for in Model 2, rather than being related to relativistic reflection features. We stress that when fitting spectra with the relxill model, the constraint on the inner disk radius can only be confidently achieved in cases where the profile of the broad Fe K$\alpha$ line is clearly visible. Since there is no broad Fe K$\alpha$ line present in the spectral residuals in Figure 5, the small inner disk radius, $R_{\rm in}$, measured (see Table 2) could be driven by the lack of prominent blurred reflection features in the spectra. Thus the model tends to artificially make the line broad to mimic part of the continuum in the Fe K$\alpha$ band. Therefore, we consider the best-fit reflection fraction obtained in Model 3 as a crude upper limit, and do not discuss physical implications of the rest of the reflection parameters to avoid over-interpretation of the data.

The high S/N of the NuSTAR spectra of Swift J1658.2–4242 used in this work allows a comparison of the strength of secondary spectral features in the two epochs: the strength of the narrow absorption line at \sim$$7.1 keV modeled by gabs can be considered as constant within errors; whereas the absolute flux of the narrow Fe K$\alpha$ emission line at $6.4-6.5$ keV modeled by Gaussian clearly increases after the flux drop (the emission line is not detected in the high-flux epoch, see best-fit parameters in Table 2). Both features were also observed in Swift J1658.2–4242 with NuSTAR early on during its 2018 outburst in the hard state, and were discussed in detail in Xu et al. (2018b).

Absorption lines in the Fe K band of the X-ray spectrum of black hole binaries are commonly associated with blueshifted absorption lines from highly ionized iron (e.g., Fe XXV and Fe XXVI) generated in an accretion disk wind, which are frequently found when the accretion rate is high in the soft state (e.g., Miller et al., 2008; Ponti et al., 2012; Miller et al., 2016). The constant strength we measured of the narrow absorption line implies an invariant disk wind component during the high- and low-flux epochs.

Narrow Fe K$\alpha$ emission lines in the spectra of black hole binaries are believed to be produced by distant reprocessing of the hard X-ray photons from the corona, which may arise from distant reflection of the coronal emission by the outer edge of a flared accretion disk, or re-emission in a disk wind. We can get an estimate of the location of the narrow Fe K$\alpha$ line emission in terms of the distance from the black hole based on the Fe K$\alpha$ line width. We replace the simple Gaussian emission line model in Model 3 with a second relxilllp model component (by setting ${\rm relf\_frac}=-1$, only the reflected part is used). As the line energy is consistent with being neutral, we fix the ionization parameter111The ionization parameter, $\xi$, is defined as $\xi=4\pi F_{\rm x}/n$, where $F_{\rm x}$ is the ionizing flux, and $n$ is the gas density., log($\xi$), of this relxilllp model component at 0, allow the parameter of the inner disk radius, $R_{\rm in}$, to vary freely, and link all other parameters with those of the relativistic disk reflection component. We can get a lower limit of the inner radius of this relxilllp component to be 406 $r_{\rm g}$ (where $r_{\rm g}\equiv{\rm GM}/c^{2}$, is the gravitational radius), which we use to represent the disk radius responsible for the narrow Fe K$\alpha$ line. The emergence of a narrow Fe K$\alpha$ emission line in the low-flux epoch indicates the appearance of reprocessing material at a large distance from the central black hole after the flux drop.

5. DISCUSSION

We have presented analyses of the coordinated NuSTAR and XMM-Newton observations with contemporaneous Swift-XRT data of the new black hole binary candidate Swift J1658.2–4242, which caught the source in the intermediate state during its 2018 outburst. A rapid decrease in the source flux is observed by both telescopes, accompanied by the turn-on of a transient low-frequency QPO. The dramatic variation in flux and timing properties together with only minor changes in the broadband X-ray spectra are unusual for black hole X-ray binaries, and the physical driver of the event is uncertain. We discuss possible causes of the uncommon properties observed in Swift J1658.2–4242 based on results from our X-ray spectral and timing analyses.

5.1. Invariance of Coronal Properties

The power-law component extending to high energies in the X-ray spectra of black hole binaries is believed to originate from the so-called corona in the vicinity of black holes. Hot electrons in the corona up-scatter soft disk photons into the hard X-ray band. The parameters characterizing the spectral shape of the coronal emission is the photon-index, $\Gamma$, and exponential cutoff at the high energy end, $E_{\rm cut}$, which are associated with physical properties of the corona, its optical depth and the electron temperature (Lightman & Zdziarski, 1987; Petrucci et al., 2001).

In order to measure the physical properties of the corona, we replace the cutoffpl component in Model 2 with the Comptonization model, compPS (Poutanen & Svensson, 1996). We assume a thermal electron distribution in the compPS model, and perform spectral fitting assuming a slab and a spherical geometry for the corona. We link the seed photon temperature in the compPS model with the disk blackbody temperature, $kT_{\rm in}$, in the diskbb model. The physical model, compPS, fits the data equally well as the phenomenological model, cutoffpl. We cannot distinguish between the two coronal geometries based on the spectral modeling. The optical depth, $\tau$, and electron temperature, $kT_{\rm e}$, of the corona can be well constrained (see Table 5.1 for the best-fit parameters). We find $\tau=1.6\pm 0.1$ and $kT_{\rm e}=24^{+2}_{-1}$ keV assuming a slab geometry, and $\tau=2.1\pm 0.2$ and $kT_{\rm e}=25\pm 2$ keV assuming a spherical geometry, which is similar to the typical values reported in other black hole X-ray binaries (e.g., Del Santo et al., 2013; Sánchez-Fernández et al., 2017). Allowing $\tau$ and $kT_{\rm e}$ to have different values for the two epochs does not bring significant improvement to the fit, indicating that these parameters which define the coronal properties are invariant in spite of the large flux variation.