MULTI-WAVELENGTH COVERAGE OF STATE TRANSITIONS IN THE NEW BLACK HOLE X-RAY BINARY SWIFT J1910.2–0546

Swift J1910.2–0546 was observed with Chandra for ≃30 ks starting on 2012 September 22 at UT 23:22 (MJD 56192, ObsID 14634). The incident flux was dispersed onto the ACIS-S CCDs using the High Energy Transmission Grating (HETG) with the ACIS-S array operated in the "CC33$\_$GRADED" mode. The data were processed using the ciao software (ver. 4.4). The positive and negative diffraction orders of the High-Energy Grating (HEG) and Medium-Energy Grating (MEG) spectra were combined using add_grating_orders. The spectral data were fitted in XSpec (ver. 12.7; Arnaud 1996).

Swift J1910.2–0546 was monitored during 47 Swift observations of ≃1 ks that were performed between 2012 August 7 and November 23 (target ID 32521). These provide simultaneous coverage in X-rays and UV via the X-ray Telescope (XRT) and the Ultra-violet/Optical Telescope (UVOT), respectively. The Swift data were reduced using heasoft (ver. 6.13).

All XRT observations were carried out in Windowed Timing mode and processed using the task xrt$\_$pipeline. The source was detected with raw XRT intensities varying between ≃6 and 250 counts s⁻¹. After examining the data for possible pile-up, we chose to conservatively exclude the central 3 pixels for count rates of ≃150–200  counts s⁻¹ and 5 pixels for ≃200–250 counts s⁻¹ (see also Reynolds & Miller 2013).

Using XSelect, we extracted XRT spectra from a box of 60 pixels long and 20 pixels wide centered on the source. Background events were obtained from neighboring regions using a box of the same dimensions. The corresponding arfs were created using the tool xrtmkarf, and the rmfs (ver. 14) were sourced from the calibration data base. The spectral data were grouped to a minimum of 20 photons per bin. We also investigated the spectral evolution by determining the X-ray hardness ratio for each observation. For the purpose of the present work, we defined this as the ratio of counts in the 1.5–10 keV and 0.6–1.5 keV energy bands.

All UVOT images were obtained using the uvm2 filter (λ_c ≃ 2246 Å). Magnitudes and flux densities were extracted with the tool uvotsource using a standard aperture of 5''. A nearby, source-free aperture of 15'' served as a background reference.

We monitored Swift J1910.2–0546 at optical and nIR wavelengths using the ANDICAM detector mounted on the SMARTS 1.3-m telescope at the Cerro Tololo Inter-American Observatory in Chile. Between 2012 August 1 and October 17 the source was observed 42 times using a selection of B, V, R, I, J, H, and K filters.

Images of 250 s were obtained for the B and V filters, while the R and I band exposures were 300 s. The nIR data consisted of 7 dithered images of 40 s each in J, 7 dithered images of 40 s each in H, and 13 dithered images of 25 s each in K. These dithered frames were flat-fielded, sky subtracted, aligned, and average-combined using an in-house iraf script.

Two nearby Two Micron All Sky Survey stars were taken as references to perform optical and nIR differential photometry with respect to Swift J1910.2–0546. Their optical magnitudes were taken from the NOMAD and USNO-B1 catalogs. The night-to-night variability in the reference stars is accounted for in the magnitude errors of Swift J1910.2–0546.

The Chandra/HETG observation was obtained during a soft state (see Section 4.2). However, the HEG and MEG spectra do not show X-ray absorption lines that would evidence the presence of an ionized disk wind (see Figure 1). We fitted the MEG data between 0.5 and 2.0 keV using an absorbed disk blackbody (diskbb; Mitsuda et al. 1984) to constrain the hydrogen column density. For this purpose we adopted the tbnew prescription with the vern cross-sections and wilm abundances (Verner et al. 1996; Wilms et al. 2000).

Zoom In Zoom Out Reset image size

The spectral fit shown in Figure 1 yielded a hydrogen column density of N_H = (3.5  ±  0.1) × 10²¹ cm⁻², and a disk temperature of kT_in = 0.42 ± 0.01 keV. Allowing the abundances of O, Ne, Mg and Fe to vary did not indicate any prominent deviations from Solar composition.

To obtain X-ray fluxes and to characterize the outburst evolution, all XRT spectra were fitted to a combined disk blackbody (diskbb) and a power law (powerlaw). Interstellar absorption was taken into account by using the tbabs model with the hydrogen column density fixed to N_H = 3.5 × 10²¹ cm⁻² (Section 4.1; for a justification, see Miller et al. 2009). X-ray fluxes were calculated in the 0.6–10 keV range. The results of our spectral analysis are illustrated in Figure 2.

Zoom In Zoom Out Reset image size

Our Swift observations started on 2012 August 7 (MJD 56146), shortly after Swift J1910.2–0546 had entered a hard state (Nakahira et al. 2012; Bodaghee et al. 2012; King et al. 2012). The XRT spectra were initially dominated by a hard power law with an index of Γ ≃ 2, but thermal emission from a relatively cool accretion disk with a temperature of kT_in ≃ 0.25 keV contributed ≃40% to the total 0.6–10 keV flux (Figure 2).¹²

Around September 5 (MJD 56175), the X-ray spectrum softened considerably. The disk temperature increased and the thermal emission dominated the 0.6–10 keV flux from September 10 onward (MJD 56180), indicating that Swift J1910.2–0546 had entered a soft state (Figure 2).¹³ Notably, even at the peak of the soft state when the thermal emission accounted for ≃99% of the total flux, the disk remained relatively cool with kT_in ≃ 0.5 keV.

After October 2 (MJD 56202), the thermal flux started to decrease while the non-thermal flux increased. The disk cooled to a temperature of kT_in ≃ 0.2 keV and the power law hardened to Γ ≃ 1.8.¹⁴ During the last phase of our monitoring campaign (October 25 through November 23, MJD 56225–56254), the X-ray spectrum is completely dominated by the hard power law, demonstrating that the source was back in a hard state (Figure 2). Swift J1910.2–0546 thus traced out the typical black hole X-ray spectral states (see also Reis et al. 2013), and made two state transitions during our monitoring campaign.

Figure 3 displays the nIR, optical, UV, and 0.6–10 keV X-ray light curves of Swift J1910.2–0546, as well as the evolution of the nIR (J − K) and optical (B − R) colors, and the X-ray hardness ratio. A striking feature is a dip in intensity that occurs ≃95 days into the outburst. At all wavelengths the flux steadily decreases toward a minimum on a timescale of ≃10 days, after which it recovers on a similar timescale. The optical and nIR magnitudes faded by ≃ 0.9 mag, which corresponds to a ≃55% drop in flux. The UV magnitude faded by ≃1.3 mag, or ≃70% in flux. The soft (disk) and hard (power law) X-ray flux decreased by ≃80% and 90%, respectively.

Zoom In Zoom Out Reset image size

The measured X-ray flux reached a minimum on 2012 September 5 (MJD 56175; ObsID 32521012), whereas the point of lowest UV flux was recorded two observations earlier on September 1 (MJD 56171; ObsID 32521010). This is suggestive of a time delay between the different wavelengths (see also Krimm et al. 2013). In an attempt to quantify this delay, we fitted the light curves to a simple broken exponential to find the time of minimum flux in each band between August 23 and September 12 (days 85–105 in Figure 3). The results are listed in Table 1 and illustrated in Figure 4. Although the error margins are large, it appears that the dip first occurred in R, I, and J around day 92 (MJD 56169), whereas both longer (H and K) and shorter (B, V, and UV) wavelengths are delayed by ≃ 1–4 days, followed by X-rays 6 days later (around day 98).

Zoom In Zoom Out Reset image size

Table 1. Time Lags in Swift J1910.2–0546

Waveband	Lag
	(days)
	Flux Dip	Full Light Curve
UV–X-ray	4.06 ± 0.71	$7.72^{+0.80}_{-0.71}$
B–X-ray	4.84 ± 0.74	8.24 ± 0.43
B–UV	0.78 ± 0.95	$2.72^{+1.00}_{-0.58}$
B − V	−0.34 ± 1.33	−0.13 ± 0.26
B − R	−2.04 ± 0.72	−0.52 ± 0.18
B − I	−1.53 ± 1.06	−0.52 ± 0.22
B − J	−1.48 ± 1.12	$-0.38^{+0.28}_{-0.24}$
B − H	−0.17 ± 1.75	$-0.13^{+0.49}_{-0.44}$
B − K	2.27 ± 2.47	⋅⋅⋅

Notes. The waveband listed first is taken as the reference, i.e., the lag given is with respect to that. When considering the full light curve no lag could be determined for the K band, as it was poorly correlated with the other light curves.

Download table as:  ASCIITypeset image

We also searched for and quantified any time lags using the standard cross-correlation analysis techniques widely used for optical reverberation in active galactic nuclei. We used a linear-interpolation cross correlation (White & Peterson 1994) and determined the uncertainties in the lags using a flux randomization and random subset sampling Monte Carlo method (Peterson et al. 1998, 2004). There were not enough points to apply this technique to a subset of the data covering the time of the flux dip only. Assuming that the variability in each waveband and each spectral state are caused by the same (or otherwise different but causally related) physical processes, we applied the method to the full data set. This analysis suggests that in the overall light curves there is a negligible time difference between the optical and nIR bands, but that the UV and X-rays lag the longer wavelengths by ≃3 and 8 days, respectively. These lags are not exactly the same as those estimated from the flux dip only (see Table 1), which may be the result of different emission mechanisms operating in the hard and soft spectral states.

Notably, Swift J1910.2–0546 entered a soft state ≃1 week after the intensity dip was seen in the X-rays (Section 4.2). This suggests a possible causal connection between the flux dip and the state transition (see Section 5). This is corroborated by the fact that around the time of the flux dip the nIR color suddenly became much bluer, e.g., the J − K color abruptly dropped from ≃1 to 0.5 around day 95 in Figure 3. In contrast, the optical bands continued to vary together, with little change in color, as illustrated by the evolution of B − R. The nIR colors remained blue during the soft state, only to become redder again when the H and K bands brightened and the source transitioned back into a hard state (around day 138; see also Section 5.3).

Whereas the SMARTS optical/nIR observations stopped at the end of the soft state, X-ray/UV monitoring with Swift continued. This revealed that the X-rays steadily decayed into the hard state, whereas the UV emission started to rise ≃ 2 weeks after the soft to hard state transition (around day 158). At the same time the X-ray spectrum considerably hardened, as illustrated by the strong increase in XRT hardness ratio (Figure 3).

To investigate the apparent anti-correlation seen between the UV and X-rays in the hard state (Section 4.3), we calculated UV fluxes (F_uvm2) by multiplying the measured UVOT flux densities with the FWHM of the uvm2 filter (≃ 498 Å; Poole et al. 2008). In Figure 5 we plot the relation between the UV and the 0.6–10 keV X-ray flux. Above F_X ≃ 10⁻⁹ erg cm⁻² s⁻¹, the two are positively correlated (Spearman rank coefficient ρ = 0.52, corresponding to a ≃3σ significance), whereas at lower fluxes a very significant negative correlation is apparent (ρ = 0.96; ≃7σ).

Zoom In Zoom Out Reset image size

To quantify these relations, we fitted both sides separately to a power law of the form $F_{\mathrm{uvm2}} \propto F_{X}^{\beta }$. This yielded β = 0.15 ± 0.05 and −0.45 ± 0.03 for the positive and negative correlation, respectively (dotted lines in Figure 5). A fit with a broken power law yields similar indices (β ≃ 0.22 and −0.45), and a break located at F_X ≃ 1.1 × 10⁻⁹ erg cm⁻² s⁻¹.

To investigate the time evolution along the X-ray/UV relations, we divided the outburst into five contiguous intervals of 9/10 observations: These are indicated by different symbols in Figure 5. This illustrates that the UV/X-ray relation evolved from a positive into a negative correlation once the source had moved from the soft to the hard state (around day 138). The anti-correlation thus mostly arises during the last part of the light curves when the UV emission rises while the X-rays decrease (Figure 3).

It is plausible that the observed anti-correlation is introduced due to the presence of a time lag between the X-ray and UV emission. To gauge this possibility, we investigated the relation after applying a time lag of 7.72 days (Section 4.3). Figure 5 (right) shows that a negative correlation arising from the last set of observations persists, although it is less significant (ρ = 0.68; ≃2σ). The positive correlation is now stronger (ρ = 0.59; ≃4σ). A fit with a broken power law yields the same indices as before applying the time lag (β ≃ 0.15 and −0.45), and a break located at F_X ≃ 7 × 10⁻¹⁰ erg cm⁻² s⁻¹.

The observed (anti-)correlation is thus sensitive to the presence of a time lag. We determined a 7.72 days time delay between the X-ray and UV emission using the full light curves. However, since different emission mechanisms could be contributing to the UV flux in the hard and soft spectral states, different delays may be expected and hence affecting the observed X-ray/UV relation (see also Section 5.3). We note that the Swift/BAT and MAXI monitoring light curves show that the outburst of Swift J1910.2–0546 continued after our pointed XRT and UVOT observations stopped (due to Sun constraints). The X-ray/UV relation may have eventually returned to a positive correlation, tracing out a similar path as the X-ray/nIR relation seen in other black holes (e.g., Coriat et al. 2009). However, it is not clear if the underlying physical processes should be the same.

A Chandra/HETG spectrum obtained when Swift J1910.2–0546 was in the soft state appeared featureless, despite the fact that ionized disk winds are expected to be ubiquitous in this spectral state (e.g., Miller et al. 2008, 2012). If such winds are mostly equatorial, a possible explanation for the lack of ionized absorption features is that Swift J1910.2–0546 is viewed at relatively low inclination (Ponti et al. 2012). This would be consistent with the analysis of disk reflection features, which suggested an inclination of i ≲ 20° (Reis et al. 2013).

Geometrical effects could perhaps also account for the fact that the accretion disk of Swift J1910.2–0546 peaked at kT_in ≃ 0.5 keV, which is much cooler than typically seen for black hole LMXBs in the soft state (kT_in ≳ 1 keV; e.g., Dunn et al. 2011; Reynolds & Miller 2013). As shown by Muñoz-Darias et al. (2013), disks look cooler when viewed at lower inclination, which can be understood in terms of relativistic effects. Swift J1910.2–0546 could thus fit into this picture. Alternatively, the low temperature could be the result of a small disk size, e.g., due to a short orbital period (see Section 5.2), a truncated disk or a retrograde black hole spin (Reis et al. 2013).

Our multi-wavelength light curves revealed a prominent flux dip that first appeared in the optical band, only to be followed by X-rays ≃6 days later. A comparable X-ray lag (≃8 days) is suggested by cross-correlating the light curves over the full time range covered by our monitoring campaign. A natural explanation for the observed delay may be a (small-scale) mass-transfer instability that originated at the outer edge of the accretion disk and then propagated inward. The time delay could then represent the viscous timescale of the disk, $t_{{\rm visc}}(r) = ({2}/{3\alpha })({h}/{r})^{-2}\Omega _{K}^{-1}$. Here α is the viscosity parameter, h the scale height of the disk, and Ω_K the Keplerian frequency at radius r.

Assuming α = 0.1, h/r = 0.01 (i.e., a geometrically thin disk), and M = 8 M_☉, a viscous timescale of t_visc = 6 days would yield a disk radius of r ≃ 3400 GM/c² (≃ 4 × 10⁹ cm). This is relatively small, perhaps supporting the fairly short ≃2–4 hr orbital period hinted by its rapid optical flaring (Lloyd et al. 2012, but see Casares et al. 2012). On the other hand, a comparable time delay of several days between the nIR and X-rays has also been inferred for black hole LMXBs that have (much) longer orbital periods (e.g., LMC–X3, GX 339–4, and 4U 1957+11; Brocksopp et al. 2001; Homan et al. 2005; Russell et al. 2010).

The interpretation of the dip as a mass-transfer instability could naturally explain why in all bands the flux recovered to the pre-dip level. However, this picture may not account for the fact that the flux appeared to hit a minimum in the H and K bands later than in the optical/UV (albeit still before the X-rays; Figure 4). This may suggest that multiple (competing) emission mechanisms are operating, or that the flux dip is the result of a different physical process altogether (Section 5.3). This is perhaps hinted by the fact that there appears to be a causal connection between the dip and the transition to the soft state ≃ 1–2 weeks later. This connection would not be obvious to explain in terms of a small-scale mass-transfer instability.

Interestingly, a very similar flux dip as we observed for Swift J1910.2–0546 was reported for GX 339–4 during its 2010 outburst (Yan & Yu 2012). In that source the UV flux decreased by ≃60% in ≃2 weeks time, which was associated with a drop in the optical/nIR (by ≃85%) as well as the radio flux. Within ≃10 days after this drop the source transitioned from a hard to soft state, and the UV flux dip was therefore associated with jet quenching (Yan & Yu 2012).

In Swift J1910.2–0546, support for change in accretion morphology (perhaps connected to a jet) is provided the observation that the nIR color suddenly became bluer right around the time of the flux dip (Figure 3). It remained as such until the end of the soft state, which started ≃ 10 days after the color change. This behavior was not observed in the optical bands, suggesting a relative suppression of the nIR emission during the soft state, i.e., when jets are known to be quenched.

When the source transitioned back to the hard state, the K and H band flux started to rise and the nIR color become redder again (around day 138 in Figure 3). Although our SMARTS monitoring stopped at this time, the hinted nIR brightening may be similar to that seen in other black hole LMXBs entering the hard state (e.g., Jain et al. 2001; Buxton & Bailyn 2004; Buxton et al. 2012; Kalemci et al. 2005, 2013; Russell et al. 2010, 2013; Dinçer et al. 2012; Corbel et al. 2013). The optical/nIR secondary maxima seen in these sources have been ascribed to the revival of a jet (e.g., Dinçer et al. 2012; Kalemci et al. 2013).

Continued Swift monitoring revealed that in Swift J1910.2–0546 the UV emission started to increase ≃20 days after the transition back into the hard state (around day 158 in Figure 3), whereas the X-rays continuously decayed. This caused the relation between the UV and 0.6–10 keV flux to move from a positive correlation (with a slope of β ≃ 0.15) into a negative correlation (β ≃ −0.45). This turnover in the UV/X-ray relation occurred at F_X ≃ 10⁻⁹ erg cm⁻² s⁻¹ (0.6–10 keV). The distance toward Swift J1910.2–0546 is unknown, but if the peak of the outburst (as observed by MAXI) did not exceed the Eddington limit (L_EDD), the transition flux corresponds to ≲ 0.03 × L_EDD. The UV brightening was accompanied by a dramatic increase in the X-ray hardness ratio, which again suggests a causal connection with changes in the accretion morphology.

This rise in UV emission is reminiscent of the secondary nIR/optical maxima seen in other black hole LMXBs. In a systematic study of a number of different sources and outbursts, Kalemci et al. (2013) showed that the nIR peaks ≃ 5–15 days after the transition from the soft to the hard state. In Swift J1910.2–0546, the UV appears to peak ≃ 10 days after it starts rising, or ≃ 30 days after the state transition (around day 168 in Figure 3). This is thus considerably later than the reported nIR peaks (Kalemci et al. 2013). In a jet interpretation it might be expected that the UV increases after the nIR (Corbel et al. 2013; Russell et al. 2013), but it is not immediately clear if this delay could be as large as ≃ 20 days.

An alternative interpretation for both the multi-wavelength flux dip and the UV peak may be provided in terms of the collapse and recovery of a hot inner flow (Veledina et al. 2011, 2013). Such a radiatively inefficient flow might replace the inner accretion disk at low mass-accretion rates (e.g., Esin et al. 1997). Analysis of disk reflection features suggested that the disk in Swift J1910.2–0546 could indeed be subject to evaporation (Reis et al. 2013). This model predicts that the outer regions of the hot flow (traced by the optical/nIR emission) collapses several days before an X-ray state transition is observed (causing a decrease in flux; Veledina et al. 2013). This is consistent with the time between the dip and the soft state transition observed for Swift J1910.2–0546 (≃1–2 weeks). Moreover, the hot flow model predicts that the optical/nIR spectrum hardens (i.e., becomes bluer) in this transition, consistent with the drop in J − K color that we observed before the state transition (Figure 3).

Veledina et al. (2011) interpret the optical/nIR secondary maxima seen after hard state transitions as the recovery of a hot inner flow (see also Kalemci et al. 2013; Veledina et al. 2013). The same mechanism may account for the UV peak (and corresponding X-ray/UV anti-correlation) discussed in this work. However, in the picture of a recovering hot flow, the UV emission should rise before the nIR (Veledina et al. 2013), whereas in Swift J1910.2–0546 there are indications that the nIR started to rise first. Moreover, a delay as long as ≃30 days between the state transition and the UV peak may be difficult to account for (see also Kalemci et al. 2013).

It is worth noting that although UV monitoring is often hampered by substantial interstellar extinction, anti-correlated UV and X-ray emission has also been observed for the neutron star Cyg X-2 (Rykoff et al. 2010). On the other hand, such an anti-correlation was not seen for the black hole LMXBs XTE J1817–330 and Swift J1357.2–0933, despite dense UV/X-ray coverage over a large range of fluxes (Rykoff et al. 2007; Armas Padilla et al. 2013). The latter source may have never entered a soft state (Armas Padilla et al. 2013; see also Corral-Santana et al. 2013 and Shahbaz et al. 2013), which could possibly explain the absence of an anti-correlation given that the UV rise appears to be connected to a state transition. However, XTE J1817–330 was monitored for well over a month after it transitioned from the soft to the hard state, but its X-ray and UV emission remained tightly correlated (Rykoff et al. 2007). Future X-ray/UV monitoring efforts of suitable targets are warranted to further understand the role of the UV emission in state transitions.