The Changing-look Optical Wind of the Flaring X-Ray Transient Swift J1858.6-0814

In addition to a large variety of accretion-related observables, mostly seen in X-rays (McClintock & Remillard 2006; van der Klis 2006; Belloni et al. 2011), low-mass X-ray binaries (LMXBs) also show a complex outflow phenomenology. This initially included synchrotron radio emission from jets, either in the form of compact sources or discrete ejections (e.g., Mirabel & Rodriguez 1999; Fender et al. 2004), and subsequently X-ray winds of highly ionized material (e.g., Miller et al. 2006; Ponti et al. 2012; Díaz Trigo & Boirin 2016). These outflows can be the dominant source of power released and mass consumed/expelled by the system during some accretion phases (Fender & Muñoz-Darias 2016), and as such represent a fundamental part of the entire accretion process onto stellar-mass black holes (BHs) and neutron stars (NSs).

Furthermore, intense and sensitive spectroscopic campaigns carried out over the last few years have unveiled the presence of optical winds in several BH transients. P-Cyg profiles have been discovered in V404 Cyg (Muñoz-Darias et al. 2016, 2017; Mata Sánchez et al. 2018; see also Casares et al. 1991 for detections during the 1989 outburst) and MAXI J1820+070 (Muñoz-Darias et al. 2019b), while an archival search showed the presence of conspicuous wind signatures in several outbursts of V4641 Sgr (Muñoz-Darias et al. 2018; see Lindstrøm et al. 2005 and Chaty et al. 2003 for earlier reports). In addition, the classical systems GRO J1655-40 and GX 339-4 showed complex emission line profiles (Soria et al. 2000; Rahoui et al. 2014, respectively), while the intriguing optical dips of Swift J1357.2-0933 (Corral-Santana et al. 2013) have recently been found to be related to disk outflows seen at high orbital inclination (Charles et al. 2019; Jiménez-Ibarra et al. 2019b). All the above, together with the near-infrared P-Cyg profiles witnessed in at least one luminous NS system (GX 13+1; Bandyopadhyay et al. 1999; see also Homan et al. 2016), indicate that cold accretion disk winds (i.e., those detected at optical and infrared wavelengths) are a relatively common feature—perhaps ubiquitous—in the LMXB accretion phenomena. For the best studied case of V404 Cyg, these are found to have a severe impact on the accretion process and outburst evolution, with an associated mass outflow rate greatly exceeding the accretion rate, albeit this system might represent an extreme case (Muñoz-Darias et al. 2016; Casares et al. 2019).

Swift J1858.6-0814 was discovered by the BAT monitor on board the Neil Gehrels Swift Observatory (Gehrels et al. 2004) on 2018 October 25, and soon after cataloged as a new galactic X-ray binary transient (Krimm et al. 2018). Since the early phase of the outburst, it became a target of special interest owing to its remarkable flaring behavior at both X-rays (Ludlam et al. 2018) and optical wavelengths (Baglio et al. 2018; Paice et al. 2018; Vasilopoulos et al. 2018). This triggered the comparison with the BH transients V404 Cyg and V4641 Sgr by Ludlam et al. (2018), which was reinforced by telegrams reporting the presence of intrinsic X-ray absorption (Reynolds et al. 2018) and optical winds (Muñoz-Darias et al. 2019a). In this Letter, we present multiepoch, high-cadence optical spectroscopic observations of Swift J1858.6-0814 showing that this system displays optical features indicating the presence of an accretion disk wind.

We obtained optical spectroscopy using OSIRIS (Cepa et al. 2000) attached to the Gran Telescopio Canarias at the Observatorio del Roque de los Muchachos in La Palma, Spain. The target was observed in eight different epochs over a time lapse of five months within 2019 March–August. We obtained between 7 and 22 individual spectra per epoch with the grisms R1000B (4200–7400 Å) and R2500R (5575–7685 Å) depending on the night (see Table 1). This resulted in a total of 90 spectra with a time cadence of ∼five minutes and a velocity resolution of ∼350 or ∼160 $\mathrm{km}\,{{\rm{s}}}^{-1}$ depending on the grism (measured from the FWHM of sky lines; i.e., assuming that resolution is dominated by the 10 slit and not the seeing). Weather conditions were good throughout the campaign, with seeing around ∼1'' and clear skies. Only epoch 1 was observed on a bright night at high airmass (∼1.7) due to the very limited visibility window of the target at that time of the year. Across the entire campaign the slit was rotated to include a brighter field star placed ∼10'' southwest from the target. In order to carry out a relative flux calibration, spectra from this object were treated in the same way as those of Swift J1858.6-0814. Data were reduced, extracted, and wavelength calibrated using IRAF tools, while molly and custom python routines were used for the analysis. From our g-band acquisition images (typically three per epoch and taken before the spectra) we derive magnitudes in the range of 15.4–16.5 across the entire campaign (calibrated against Pan-STARRS).

Table 1.  Observing Log

Epoch	Observing Window (UT)	Grisma and Exposures	g-band Magnitude
1	24 Mar (05:14–05:49)	LR (7 × 280 s)	16.1
2	14 Apr (05:30–06:16)	LR (9 × 280 s)	15.4–15.7
3	30 Apr (04:32–05:22)	LR (10 × 280 s)	15.9–16.2
4	12 May (03:36–04:32)	LR (1 × 280 s) + HR (10 × 280 s)	15.7–16.4
5	9 Jun (03:52–04:48)	LR (1 × 280 s) + HR (10 × 280 s)	16.2–16.5
6	1 Jul (04:03–04:59)	LR (1 × 280 s) + HR (10 × 280 s)	16.2–16.4
7	6 Aug (00:00–01:52))	HR (21 × 280 s) + LR (1 × 280 s)	16.3
8	18 Aug (21:57–22:43)	HR (9 × 284 s)	15.7–16.4

Notes. All the data were taken in 2019.

^aLR and HR indicate lower (∼350 $\mathrm{km}\,{{\rm{s}}}^{-1}$) and higher (∼160 $\mathrm{km}\,{{\rm{s}}}^{-1}$) resolution grism (R1000B and R2500R, respectively).

Download table as:  ASCIITypeset image

We focus our analysis on the 5500–7200 Å spectral range covered by both the lower (R1000B) and higher resolution (R2500R) grisms. It includes the emission lines He i at 5876 Å (He i–5876), Hα (6563 Å), He i–6678, and He i–7065. The R1000B spectra (epochs 1–7) also include the Bowen blend (mainly N iii at 4641 Å) and He ii at 4686 Å. These emission lines are detected across epochs 2–7, while the limited data quality of epoch 1 only allows us to study Hα in great detail.

Figure 1 shows the trailed spectra (top panels) corresponding to He i–5876 (left) and Hα (plus He i–6678; right). Intensity is indicated by a color scale covering from 0.92 (deep blue) to 1.08 (bright yellow) times the continuum level. Each spectral region was independently normalized by fitting the adjacent, local continuum with a first-order polynomial. This was done by considering relatively broad continuum regions at each side of the emission lines (∼±10,000 $\mathrm{km}\,{{\rm{s}}}^{-1}$), masking the lines themselves and their closest continuum, as well as other contaminant features (e.g., He i–6678 for Hα). By repeating this analysis with slightly different mask configurations we estimate that the normalization process introduces an uncertainty in the continuum level of ∼0.3% (i.e., much lower than the ±8% intensity scale used in the trailed spectra). The blue wing of the He i–5876 emission profile, typically the most sensitive to the presence of optical winds (e.g., Muñoz-Darias et al. 2016), show blueshifted absorptions in virtually every observing epoch. These absorptions (represented by deep blue traces) make the blue part of the line disappear completely during a large part of the observing campaign. We interpret this as the signature of an accretion disk wind, similar to those previously witnessed in some BH transients.

Zoom In Zoom Out Reset image size

A closer look at the trailed spectra reveals that the blueshifted He i–5876 absorptions are more conspicuous between epochs 3 and 8, while Hα shows them only on epoch 5 and more weakly on epoch 7 (Figure 1, top right panel). Likewise, He i–6678 behaves in a consistent way with He i–5876, albeit it is significantly less intense than the latter, and therefore any observational feature is expected to be less marked.

Figure 2 (left panel) shows the evolution of He i–5876 using epoch 2–8 nightly average spectra. We can distinguish two groups of data by looking at the evolution of the blue wing of the emission profile (i.e., negative velocities). Epochs 3, 5, and 7 (and also 2 to some extent) show standard P-Cyg profile shapes (i.e., blueshifted absorption and redshifted emission), with the most conspicuous case (epoch 5) reaching a terminal velocity¹⁰ of 1700 $\mathrm{km}\,{{\rm{s}}}^{-1}$ and an absorption depth at the core of the profile of 95% the continuum level. An usual approach to determine the terminal velocity in a systematic way is to perform Gaussian fits to the blueshifted absorption (e.g., Muñoz-Darias et al. 2016). However, in this case the shapes are clearly non-Gaussian and hence the velocity was simply determined by visual inspection. Given the high signal-to-noise ratio of our data, we estimate this to be accurate within ∼100 $\mathrm{km}\,{{\rm{s}}}^{-1}$ (see, e.g., insets in Figure 3). The second group of data is formed by epochs 4, 6, and 8. They show odd profiles characterized by strong and slightly blueshifted absorption troughs reaching down to 92% the continuum level in some cases (epochs 4 and 8). However, the corresponding epoch-averaged Hα lines are dominated by standard double-peaked profiles during these times (Figure 2; right panel). The combination of a wind-induced blueshifted absorption at low velocity, together with an underlying double-peaked profile (observed in Hα) offers a viable explanation for the behavior of He i–5876 during these epochs. Under this interpretation, the emission bumps at ∼−500 $\mathrm{km}\,{{\rm{s}}}^{-1}$ would be associated with a partial absorption (by the wind) of the double-peaked disk component. If this is the case, the blue peak would appear weaker than the red one even when Hα shows the opposite behavior (e.g., epoch 4). This dimming of the blue emission suggests that even though the blue edge of the absorption indicates wind velocities of just a few hundred kilometers per second, a higher velocity component is likely present.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It is important to remark that while the Hα profile seems to be unaffected during these low-velocity wind phases, it is clearly asymmetric or even shows blueshifted absorptions during the remaining epochs. As can be seen in the trailed spectra (Figure 1), the most clear Hα wind detection also occurs on epoch 5. The wind signature is particularly strong in the last three individual spectra of this window (indicated by a dashed-lined, orange rectangle in the trailed spectra), whose averaged spectrum is shown in Figure 3. Profound P-Cyg profiles are detected in He i–5876, and especially Hα with a blueshifted absorption reaching 90% the continuum level and a blue edge velocity of ∼2400 $\mathrm{km}\,{{\rm{s}}}^{-1}$. This is larger than ∼1700 $\mathrm{km}\,{{\rm{s}}}^{-1}$ observed in He i–5876 (see insets in Figure 3).

3.1. Continuum Flux Variability

3.2. Search for Broad Emission Line Wings

Swift J1858.6-0814 is, after the BH systems V404 Cyg, V4641 Sgr, and MAXI J1820+070, the fourth transient LMXB showing optical P-Cyg profiles. Buisson et al. (2020) have recently reported the detection of several X-ray flares consistent with being type I X-ray bursts. If confirmed, this would make Swift J1858.6-0814 the first NS transient showing optical wind signatures, reinforcing the similarities between the outflow phenomenology seen in BH and NS transients (e.g., Ponti et al. 2014 for X-ray winds; Miller-Jones et al. 2010 for jets). In addition, the system has displayed profound, periodic dips/eclipses implying a high inclination and strongly suggesting a relatively long orbital period of 21.8 hr (Buisson et al. 2020).

As in the previous cases, the emission line that displays the most conspicuous wind signatures is He i–5876. This is also one of the best optical wind markers in massive stars and accreting white dwarfs (e.g., Prinja & Fullerton 1994; Kafka & Honeycutt 2004). Our 90 spectra, taken over eight epochs in a time lapse of five months, show that the optical wind is active during a large fraction of the time, and continuously detected in our data from (at least) epoch 2 (Figures 1 and 2). Epoch 5 is arguably the most interesting window, with strong wind signatures in both He i–5876 and Hα. The blueshifted absorptions become particularly conspicuous in the last three spectra of this epoch, showing also a significant evolution toward high velocities. The inset in the bottom left panel of Figure 1 shows the continuum flux evolution during epoch 5. It can be seen that a flare occurs in the middle of the window. Interestingly, the blueshifted Hα absorption almost disappears during the peak (top right panel in Figure 1), while the last three spectra (orange dots) correspond to the decay of the flare. This behavior strongly resembles that of V404 Cyg (e.g., Figure 2 in Muñoz-Darias et al. 2016), which showed the strongest P-Cyg profiles during a low luminosity epoch following a flare. However, the terminal velocity was not observed to vary in that case.

In V404 Cyg, the presence and variability of the P-Cyg profiles was found to be correlated with the ionization state of the outer disk, which is traced by the relative strength of He ii to Hβ emission. In particular, the strongest P-Cyg profiles were found at low ionization; a conclusion that was also supported by the analysis of MAXI J1820+070 (Muñoz-Darias et al. 2019b). Here, we have also computed the flux ratio of He ii to Hβ (as well as the EWs of He ii and Bowen blend) in the 30 spectra taken with the lower resolution grism. Epochs 2 and 3 show variability within a factor of 2, roughly following the evolution of the continuum flux, while the single value obtained for epoch 5 is in the lower end of the sample. However, as in the case of the continuum flux variability, we do not find a clear correlation with the presence/absence of wind features. Nevertheless, given the lack of high-cadence coverage in He ii and the similarities between epoch 5 and the behavior seen in V404 Cyg, we cannot rule out that ionization effects are a key factor in determining the visibility and properties of the wind.

4.1. Comparison with Other LMXBs

Table 2 displays the main wind observables as well as some fundamental system parameters of the four transient LMXB with optical P-Cyg detections. As recently discussed in Hare et al. (2020), the strong flaring activity and variable X-ray absorption of Swift J1858.6-0814 (see also Ludlam et al. 2018; Reynolds et al. 2018) resembles the behavior of V404 Cyg and V4641 Sgr (see, e.g., Gallo et al. 2014; Morningstar et al. 2014; Kimura et al. 2016; Motta et al. 2017; Muñoz-Darias et al. 2018). These two objects display nonstandard outbursts characterized by sharp rising phases followed by significant luminosity drops and the absence of steady soft states (e.g., Casares et al. 2019 for V404 Cyg). The discovery outburst of Swift J1858.6-0814 is still ongoing and it is beyond the scope of this paper to discuss the outburst evolution. Nevertheless, the source has been active in radio during the entire optical campaign (J. van den Eijnden et al. 2020, in preparation), which is one of the standard observables associated with BH (and NS to some extent) hard and intermediate states (e.g., Fender & Muñoz-Darias 2016). This is consistent with the behavior seen in V404 Cyg, V4641 Sgr, and MAXI J1820+070.

Table 2.  Transient LMXBs with Optical P-Cyg Profiles

Transient	Period (hr)	Inclination (deg)	P-Cyg Depth	Terminal Velocity ($\mathrm{km}\,{{\rm{s}}}^{-1}$)	Outburst	References
V404 Cyg	155.3	60–70	30%	3000	Nonstandard	1, 2
V4641 Sgr	67.6	60–70	40%	1600 (3000)a	Nonstandard	3, 4
Swift J1858.6-0814	21.8 (?)	Dipping/eclipsing	10%	2400	Ongoing	5, this work
MAXI J1820+070	16.9	60–81	2.5%	1800	Standard	6, 7, 8

Notes.

^aThe maximum P-Cyg terminal velocity is ∼1600 $\mathrm{km}\,{{\rm{s}}}^{-1}$. However, broad wings reaching ∼3000 $\mathrm{km}\,{{\rm{s}}}^{-1}$ were also observed.

References. (1) Casares & Jonker (2014), (2) Muñoz-Darias et al. (2016), (3) Orosz et al. (2001), (4) Muñoz-Darias et al. (2018), (5) Buisson et al. (2020), (6) Torres et al. (2020), (7) Atri et al. (2020), (8) Muñoz-Darias et al. (2019b).

Download table as:  ASCIITypeset image

Table 2 also reports the maximum terminal velocity and blue-absorption depth observed in the four transients with optical P-Cyg detections. These values are likely biased since (among other things) the observing campaigns (e.g., number and frequency of observations) were significantly different and the wind parameters might also change from outburst to outburst. For instance, the 1989 outburst of V404 Cyg is characterized by lower wind velocities than the 2015 event (Mata Sánchez et al. 2018). Nevertheless, the numbers suggest that, from a purely observational point of view, the wind signatures of Swift J1858.6-0814 are reminiscent but not as extreme as those found in V404 Cyg and V4641 Sgr, while they significantly exceed those of MAXI J1820+070. This last source displayed a regular outburst evolution (e.g., Shidatsu et al. 2019) and has an orbital period of ∼17 hr (Torres et al. 2019). A tempting possibility that arises from Table 2 is that larger accretion disks (i.e., ∼long orbital periods) might produce stronger winds, that carry away more mass and/or angular momentum. This could impact the observed outburst evolution, which at least for the cases of V404 Cyg and V4641 Sgr deviates from the standard patterns typically seen in LMXBs (see, e.g., Dunn et al. 2010; Muñoz-Darias et al. 2014 for global studies). Clearly, more observations are required to confirm this speculation. A key system on this matter could be GRS 1915+105—the BH–LMXB with the longest orbital period (e.g., Corral-Santana et al. 2016)—which has shown some of the best examples of X-ray winds and radio jets (e.g., Neilsen & Lee 2009). Although this very extinguished source cannot be observed at optical wavelengths, sensitive, infrared spectroscopy might be able to shed light on this topic.

Finally, it is worth discussing the role of the orbital inclination in the wind detectability, since it has been found to be a key parameter for X-ray winds, which are best seen at high inclination (Ponti et al. 2012; Díaz Trigo & Boirin 2016). Table 2 shows how optical P-Cygs have only been detected in sources with relatively high inclination (≳60°; see Higginbottom et al. 2019 for a discussion on the wind geometry and detectability as a function of the line of sight). To this list one could add the BH transient Swift J1357.2-0933. This high inclination system has not shown standard P-Cyg profiles, but its characteristic optical dips have recently been found to be associated with broad, blueshifted absorptions indicating the presence of an outflow (Charles et al. 2019; Jiménez-Ibarra et al. 2019a).

We have detected clear optical features indicating the presence of an accretion disk wind in the X-ray transient Swift J1858.6-0814. The observational properties of the wind are similar to those observed in V404 Cyg and V4641 Sgr. These systems also share other observables with Swift J1858.6-0814, such as the presence of frequent flares and variable X-ray absorption. As is the case for other LMXBs with optical wind detections, the outflow is contemporaneous with the radio jet. This work provides additional support for systematic and sensitive optical spectroscopic studies of active X-ray binaries in order to unveil the occurrence rate, observational properties, and impact of these cold accretion disk winds.

We acknowledge support by the Spanish MINECO under grant AYA2017-83216-P. T.M.D. and M.A.P.T. acknowledge support via Ramón y Cajal Fellowships RYC-2015-18148 and RYC-2015-17854. M.A.P. is funded by the Juan de la Cierva Fellowship Programme (IJCI–2016-30867). D.M.S. acknowledges support from the ERC under the European Unions Horizon 2020 research and innovation programme (grant agreement No. 715051; Spiders). J.v.d.E. and N.D. are supported by an NWO Vidi grant awarded to N.D. F.F. and D.A. acknowledge support from the Royal Society International Exchanges The first step for High-Energy Astrophysics relations between Argentina and UK. D.A. and D.J.K.B. acknowledge support from the Royal Society. M.O.A. acknowledges support from the Royal Society through Newton International Fellowship program. N.C.S. acknowledge support by the Science and Technology Facilities Council (STFC), and from STFC grant ST/M001326/1. F.M.V. acknowledges support from STFC under grant ST/R000638/1. molly software developed by Tom Marsh is acknowledged.