Hard-state Optical Wind during the Discovery Outburst of the Black Hole X-Ray Dipper MAXI J1803–298

In order to determine the X-ray state of J1803 during the optical observations, we analyzed the publicly available data from the Neutron star Interior Composition Explorer instrument (NICER; Gendreau et al. 2012), following a standard prescription for the data reduction ⁶ (see, e.g., Cúneo et al. 2020a). We carried out a phenomenological fit of the observations between days MJD 59336 and MJD 59511 in XSPEC (v.12.12.0; Arnaud 1996) using a multicolor disk (DISKBB; Mitsuda et al. 1984) plus a Comptonization component (NTHCOMP; Zdziarski et al. 1996; Zycki et al. 1999). We used the TuebingenBoulder interstellar medium absorption model (TBABS in XSPEC) with the cross-sections of Verner et al. (1996) and abundances of Wilms et al. (2000) to account for interstellar absorption. We assumed an equivalent hydrogen column N_H = 3.2 × 10²¹ cm⁻² (Bult et al. 2021). From the fits, we derived fluxes in the 2–6 and 6–10 keV energy bands to construct the hardness intensity diagram (HID; Homan et al. 2001). This allows for a direct comparison with the BH sample presented in Muñoz-Darias et al. (2013). Since the distance is unknown, we assumed a standard value of 8 kpc to compute the X-ray luminosity in Figure 1. We note that a detailed X-ray study of the NICER data is beyond the scope of this Letter.

Zoom In Zoom Out Reset image size

The resulting HID (Figure 1) shows a q-shaped hysteresis pattern, a behavior typically seen in BH (e.g., Fender et al. 2004) and neutron star (Muñoz-Darias et al. 2014) LMXBs. The HID has triangular shape as well as a spike during the hard-to-soft transition (a.k.a. a plume), both distinctive features of high-inclination BH transients (Muñoz-Darias et al. 2013). This is in agreement with the detection of strong X-ray dips, which imply i > 60 deg (Frank et al. 1987). We note that the hardness value reached during the soft state is similar to that shown by some mid-to-low-inclination BHs in Muñoz-Darias et al. (2013). However, the latter study uses RXTE data, whose high-energy coverage (>10 keV) and sensitivity enables an accurate modeling of the Comptonization component. This is not the case for the soft-state observations used here, whose hard X-ray band fluxes are most likely underestimated due to NICER's lower sensitivity at high energies. Thus, the actual value of the X-ray color during the soft state has to be taken with caution.

Figure 1 shows the location of the spectroscopic epochs in the HID. GTC-1 and GTC-2 (see Table 1) were obtained during the raise of the outburst in the hard state. Epochs GTC-3 to GTC-6 (including VLT-1) were obtained during either the bright hard state or the hard-intermediate state (HIMS), as shown by the detection of type-C quasi-periodic oscillations (QPOs; Chand et al. 2021; Xu & Harrison 2021). The discovery of type-B QPOs coincident with epoch GTC-7 (Ubach et al. 2021) is a defining trait of a soft-intermediate state (SIMS; Belloni et al. 2011). The remaining spectra correspond to the soft state.

We normalized the spectra to their continuum level. We used different masks (in order to avoid emission and telluric lines), explored a range of polynomial orders, and normalized both the full spectra and the particular regions of interest separately. While these resulted in a reliable normalization (see, e.g., Figures 2 and 3), we note that the continuum shape was not smooth. The presence of relatively deep, known diffuse interstellar bands (DIBs, such as ∼6613 Å) and other unidentified features hampered a better characterization of the spectral continuum, although we stress that this has no impact on the results in this paper.

Zoom In Zoom Out Reset image size

The VLT-1 spectrum covers the widest wavelength range, from near-ultraviolet to the NIR (Figure 2). We identify hydrogen emission lines corresponding to the Balmer, Paschen, and Bracket series. He i and ii transitions are also detected in emission, with the strongest examples at 10830.17 Å and 4685.75 Å, respectively. Additionally, high-excitation lines produced by fluorescence, such as the Bowen blend (composed of a mixture of C, O, and N lines), are also identified.

The GTC spectra, while covering a more limited wavelength range, allow us instead to analyze the evolution of line profiles during the outburst. In particular, we focus our analysis on Hα, a reference line present during the whole event, and also discuss the profile of the adjacent He i 6678.15 Å line when detected. In order to provide quantitative results on the profile variability, we perform a multi-Gaussian fit to Hα. We fit to each spectrum two emission Gaussians of the same width (W), with variable heights (H_blue and H_red) and centroid positions, which are conveniently defined in terms of the peak-to-peak separation (DP) and the centroid of the double-peaked structure (γ). Due to the appearance of broad absorption components during certain epochs, we also included when necessary a third Gaussian in absorption. All reported uncertainties correspond to a statistical 1σ value.

We find dramatic changes in the Hα profile across the outburst event (Figure 3), the most remarkable examples being those observed during the hard state (GTC-1 and GTC-2). In GTC-1 the profile is double-peaked with the red peak clearly stronger by a factor 2. The profile also shows a broad, blue wing extending down to ∼−2000 km s⁻¹ , while the red peak reaches the continuum level much more abruptly (∼1000 km s⁻¹). The observed peak-to-peak separation is DP = 1160 ± 30 km s⁻¹, and the profile is blueshifted, with the centroid at γ = −109 ± 15 km s⁻¹. During GTC-2, the profile shows an opposite behavior: the double peak is symmetric, while the peak-to-peak separation has significantly shrunk to DP = 753 ± 10 km s⁻¹, and the center of the two-Gaussian structure is now redshifted at γ = 173 ± 9 km s⁻¹. The striking difference between the two profiles is even clearer when we plot both epochs together (Figure 4). The two red peaks match each other, but the inter-peak depression on GTC-1 coincides with the blue peak in GTC-2. A P-Cygni profile dipping ∼2% below the continuum level is also shown in GTC-2, with velocities of ∼1000 km s⁻¹ and ∼1250 km s⁻¹ for the core and its blue-edge velocity, respectively. The red wing extends now to ∼2200 km s⁻¹. A weaker blueshifted absorption is also observed in the nearby He i 6678.15 Å line, albeit with a slightly lower blue-edge velocity of ∼1000 km s⁻¹.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Epochs GTC-3 to GTC-10, including VLT-1, show double-peaked profiles with peak-to-peak separation (DP = 500–700 km s⁻¹) and centroids (γ = 40–140 km s⁻¹) closer to that exhibited in GTC-2. They also show a broad absorption, where the emission line is embedded, and whose depth varies between epochs (with the deepest example being GTC-3, reaching 8% below the continuum level), but always constrained within ±2000 km s⁻¹. Similar broad absorptions have been previously observed in a handful of LMXBs (e.g., MAXI J1807+132; Jiménez-Ibarra et al. 2019b), but their origin is not well understood yet. Dubus et al. (2001) proposed they are produced in optically thick regions of the disk, mimicking the behavior of a stellar atmosphere. However, this scenario favors the observation of such features in low-inclination systems, which is at odds with their detection in relatively high-inclination LMXBs such as XTE J1118+480 (Khargharia et al. 2013), and now J1803. The relative intensity of the red and blue emission peaks is also variable, partly due to the changing depth of the underlying broad absorption, but their height ratio (H_red/H_blue = 0.9–1.4) is never comparable to that seen in GTC-1 (H_red/H_blue = 2.0). The base of the emission component is restricted to the ±1000 km s⁻¹ range across all epochs, again highlighting the striking difference with GTC-1 and GTC-2.

The latest epoch (GTC-11), taken 3 months after the outburst peak, shows a symmetric profile with larger double-peak separation than in previous observations (DP = 980 ± 20 km s⁻¹) but not as large as that observed in GTC-1. While this is still a soft-state spectrum, the profile is broader than those observed at higher luminosity and similar to typical LMXB quiescent profiles. No obvious absorption component is present at this time.

To date, up to six LMXBs have been found to exhibit P-Cygni profiles in their NIR and/or optical spectra: GX 13+1 (Bandyopadhyay et al. 1997), V404 Cyg (Casares et al. 1991; Muñoz-Darias et al. 2016, 2017; Mata Sánchez et al. 2018), V4641 Sgr (Muñoz-Darias et al. 2018), Swift J1858.6−0814 (Muñoz-Darias et al. 2020), MAXI J1820+070 (Muñoz-Darias et al. 2019), and GRS 1716−249 (Cúneo et al. 2020b). Our spectroscopy of J1803 showed P-Cygni profiles during GTC-2 in two emission lines (Hα and He i 6678 Å). The P-Cygni profiles have (blue-edge) terminal velocities of 1000–1250 km s⁻¹, which falls within the typical range observed in other LMXBs (∼1000–3000 km s⁻¹). The detection of these optical wind features in J1803, a BH with a proposed orbital period as low as P_orb ∼ 7 hr (Xu & Harrison 2021), shows that these can develop over a wide range of accretion disk sizes (see, e.g., Muñoz-Darias et al. 2020).

Another significant feature traditionally associated with outflows is the presence of broad emission wings. These have been previously observed in a handful of LMXBs, sometimes simultaneously with the P-Cygni profiles (e.g., Mata Sánchez et al. 2018). They appear as either symmetric, broad wings, typically detected after a bright flare and accompanied by an enhanced Balmer decrement (Muñoz-Darias et al. 2016; Mata Sánchez et al. 2018), or as a nonsymmetric blueshifted emission wing (e.g., MAXI J1820+070; Sánchez-Sierras & Muñoz-Darias 2020). Regarding J1803, the most promising example is shown by the GTC-1 spectrum, with a blue wing extending to ∼−2000 km s⁻¹, similar in absolute value to the red wing of the GTC-2 P-Cygni spectrum. The Hα profile in GTC-1 is particularly unusual, as proven by its large FWHM and DP, as well as its asymmetry. A possible interpretation involves the presence of a strong, blueshifted absorption at low velocity (∼−500 km s⁻¹) which cancels the blue peak, thus producing the observed profile. Low-velocity P-Cygni profiles are intrinsically harder to identify, as they produce a dent on the broad emission from the accretion disk rather than a clear dip below the continuum. The combination of all these features in the GTC-1 J1803 spectrum leads us to conclude that outflows are present since the onset of the outburst, consistent with preliminary reports by Buckley et al. (2021).

Finally, other possible outflow signatures include blueshifted absorption components during optical dips (e.g., Swift J1357.2−0933; Jiménez-Ibarra et al. 2019a), broad-wing emission components only detected at the end of the outburst (e.g., Rahoui et al. 2014; Panizo-Espinar et al. 2021), and complex profiles such as flat-top or triangular profiles (MAXI J1820+070; Sánchez-Sierras & Muñoz-Darias 2020). In this regard, VLT-1 reveals flat-top and triangular profile shapes in NIR lines such as Paβ or Paγ (saw-tooth profile), but also flat-top lines like He ii 4686 Å or Pa9. This leads us to propose that outflows are also present during the soft state, and are preferentially traced by the NIR spectrum. A similar conclusion was derived from the study of MAXI J1820+070 (Sánchez-Sierras & Muñoz-Darias 2020). They suggested that this is likely the result of ionization effects rather than intrinsic changes in the outflow rate or geometry.

J1803 has been considered as a candidate to harbor a BH attending to the displayed X-ray properties during the outburst. However, only dynamical studies of the source during quiescence enable the determination of the compact object mass, thus confirming its nature. In this regard, Casares (2015) introduced a new method to determine the companion star radial velocity semiamplitude (K₂) from a scaling with the FWHM of the Hα emission line. Nevertheless, this correlation only holds for spectra obtained during true quiescence, a state occurring from months to years after the outburst (see, e.g., Casares 2015). While the study of this parameter during the earlier stages of the event is hampered by the complexity of the profiles (i.e., due to the presence of outflows), soft-state spectra are expected to evolve smoothly toward the quiescence level as the disk shrinks to its original size.

The Hα line during the J1803 outburst does indeed show significant variability in FWHM between ∼900 and 1600 km s⁻¹, an expected behavior as the accretion disk expands. In this regard, our latest observed spectrum (GTC-11, 3 months after the outburst onset) shows the largest FWHM = 1570 ± 100 km s⁻¹, favoring this interpretation for J1803. In an attempt to assess the true quiescence FWHM from the aforementioned value, we inspected the literature to assess typical quiescence-to-outburst correction ratios (FWHM_ratio = FWHM_quies/FWHM_out). We note that larger corrections are required if hard-state or early soft-state epochs are considered, due to a combination of a smaller accretion disk and the influence of outflows. For this reason, we focused on optical spectra at low X-ray luminosity and as late in the event as possible (such as our J1803 GTC-11 epoch). We found the best comparison with the recent outburst of MAXI J1820+070, a high-inclination transient BH for which we take FWHM_quies = 1793 ± 101 km s⁻¹ (Torres et al. 2020) and a range of FWHM_out = 1180–1360 km s⁻¹, from measurements in Muñoz-Darias et al. (2019) outburst spectra of 1220 ± 40 km s⁻¹ (epoch 32) and 1340 ± 20 km s⁻¹ (epoch 33). This corresponds to a conservative range of FWHM_ratio ∼ 1.2–1.6, consistent with that observed in other transient BHs, e.g., FWHM_ratio ∼ 1.2–1.3 for Swift J1357.2−0933 (Torres et al. 2015), and FWHM_ratio ∼ 1.4 for V404 Cyg (Mata Sánchez et al. 2018).

We use the above estimate to construct a uniform distribution of FWHM_ratio to be combined with a normal distribution for the measured FWHM, yielding FWHM_quies ∼ 2000–2500 km s⁻¹ (1800–2700 km s⁻¹) at the 68% (95%) confidence level. Following Casares (2015), we derive K₂ ∼ 460–570 km s⁻¹ (410–620 km s⁻¹). This, together with P_orb ∼ 7 hr, yields a lower limit to the mass of the compact object of ≳ 3 M_⊙, providing further support to its BH nature. In addition, by considering the constraints on the orbital inclination (i > 65 deg) and assuming a typical range for the binary mass ratio (0.01 ≤ q ≤ 0.2) we obtain M_BH ∼ 4–8 M_⊙ (3–10 M_⊙). This mass range is consistent with that of the known population of BHs in LMXBs (Casares & Jonker 2014).