Unprecedentedly bright X-ray flaring in Cygnus X-1 observed by INTEGRAL

TL;DR

This paper reports the first observation of extremely bright, short-duration X-ray flares in Cygnus X-1, revealing unprecedented energetic events likely caused by jet activity or interactions with stellar wind.

Contribution

It presents the first detection of such intense flaring in Cygnus X-1 despite extensive prior monitoring, highlighting a new extreme behavior in this well-studied black hole binary.

Findings

Flares reached peak luminosity of 1.1-2.6×10^38 erg/s within seconds.

Flares observed across all INTEGRAL instruments with no significant spectral change.

No corresponding radio flux increase detected during flares.

Abstract

We study three extraordinarily bright X-ray flares originating from Cyg X-1 seen on 2023 July 10 detected with INTEGRAL. The flares had a duration on the order of only ten minutes each, and within seconds reached a 1-100 keV peak luminosity of $1.1-2.6\times10^{38}$ erg/s. The associated INTEGRAL/IBIS count rate was about ${\sim}$10x higher than usual for the hard state. To our knowledge, this is the first time that such strong flaring has been seen in Cyg X-1, despite the more than 21 years of INTEGRAL monitoring, with almost ${\sim}$20 Ms of exposure, and the similarly deep monitoring with RXTE/PCA that lasted from 1997 to 2012. The flares were seen in all three X-ray and $\gamma$-ray instruments of INTEGRAL. Radio monitoring by the AMI Large Array with observations 6 h before and 40 h after the X-ray flares did not detect a corresponding increase in radio flux. The shape of the X-ray spectrum shows only marginal change during the flares, i.e., photon index and cut-off energy are largely preserved. The overall flaring behavior points toward a sudden and brief release of energy, either due to the ejection of material in an unstable jet or due to the interaction of the jet with the ambient clumpy stellar wind.

Tables

Table 1. Table 1 : Best-fit parameters for the IBIS spectra of the time outside of the flares and the three flares.

	Non-Flare	Flare 1	Flare 2	Flare 3
$T_{\exp }$	57488	302	260	242
cutoffpl
$\Gamma$	${2.17}_{-0.20}^{+0.19}$	$1.7\pm 0.4$	$1.5\pm 0.4$	$1.6\pm 0.4$
$E_{\mathrm{fold}}$	${190}_{-70}^{+170}$	${65}_{-17}^{+31}$	${47}_{-9}^{+13}$	${53}_{-12}^{+18}$
$F_{40-80}$	$1.88\pm 0.04$	$14.2\pm 0.4$	$25.5\pm 0.6$	$33.3\pm 0.8$
${\chi }^2/\mathrm{dof}$	46.2/59	55.7/59	71.4/59	63.4/59
brokenpl
${\Gamma }_1$	${2.17}_{-0.43}^{+0.17}$	${2.13}_{-0.31}^{+0.23}$	${2.44}_{-0.23}^{+0.17}$	${2.21}_{-0.33}^{+0.25}$
$E_{\mathrm{break}}$	$55\pm 9$	${55}_{-6}^{+7}$	$64\pm 8$	${57}_{-7}^{+11}$
${\Gamma }_2$	$2.70\pm 0.08$	${3.06}_{-0.13}^{+0.15}$	${3.49}_{-0.17}^{+0.23}$	${3.21}_{-0.13}^{+0.20}$
$F_{40-80}$	$1.92\pm 0.05$	$14.5\pm 0.4$	$25.7\pm 0.6$	$33.7\pm 0.9$
${\chi }^2/\mathrm{dof}$	36.2/58	49.2/58	69.1/58	62.7/58
compTT
$k{T}_e$	${130}_{-70}^{+170}$	${33}_{-8}^{+22}$	${25}_{-4}^{+6}$	${28}_{-5}^{+8}$
$\tau$	${0.10}_{-0.10}^{+0.28}$	${0.9}_{-0.5}^{+0.4}$	$1.1\pm 0.4$	$1.1\pm 0.4$
$F_{40-80}$	$1.88\pm 0.04$	$14.1\pm 0.4$	$25.4\pm 0.6$	$33.1\pm 0.8$
${\chi }^2/\mathrm{dof}$	46.2/59	57.5/59	74.8/59	65.9/59

Equations

$$\frac{r}{r_{\mathrm{g}}}\simeq 1.6\times 10^4{\left(\frac{b}{30}\right)}^{-1}{\left(\frac{M_{\mathrm{BH}}}{21M_{\odot }}\right)}^{-1}\frac{t_{\mathrm{D}}}{100\mathrm{s}},$$

Full text

1 Introduction

Cyg X-1 is one of the best studied stellar-mass black hole X-ray binaries. Discovered in 1962 (Bowyer et al., 1965), the black hole has a mass of $21.2\pm 2.2\,M_{\odot}$(Miller-Jones et al., 2021) and is in an almost circular 5.6 d orbit with its donor, HDE 226868 (Bolton, 1972), at a separation of 0.24 AU and a distance of $2.22^{+0.18}_{-0.17}$ kpc from us (Miller-Jones et al., 2021). Systematic long-term monitoring in the X-rays started around 1975 with Ariel V (Holt et al., 1979), was continued with Ginga (Kitamoto et al., 2000), CGRO/BATSE (Ling et al., 1997), and the RXTE/ASM (Grinberg et al., 2013) and is currently available from MAXI and Swift/BAT. Cyg X-1 has also been the subject of detailed campaigns of pointed observations, with missions such as RXTE (Pottschmidt et al., 2003; Wilms et al., 2006; Grinberg et al., 2014), INTEGRAL (e.g., Del Santo et al., 2013; Cangemi et al., 2021), and NICER (e.g., König et al., 2024). In total, almost half a century of X-ray monitoring has shed light on the variability of Cyg X-1 on timescales from milliseconds to months. ††† deceased 17 June 2025

Cyg X-1 is a persistent X-ray source that, similar to other BHBs, can be found in two canonical states, which can be characterized through their spectral and timing properties (e.g., Grinberg et al., 2013; König et al., 2024, and references therein). The “hard state” shows an X-ray spectrum that can mainly be explained by Comptonization of soft seed photons by a hot electron plasma. In the “soft state”, thermal emission from the accretion disk dominates the spectrum, although at least some Comptonized radiation is still observed. Between the two main states Cyg X-1 transits through the so called “intermediate state”, with changes between different states typically happening on timescales of days to weeks. On short timescales of milliseconds to minutes power spectra and other timing quantities show characteristic state-dependent behavior that have been interpreted as being due to a combination of variability in the accretion flow and in the Comptonizing plasma (König et al., 2024, and references therein). Superimposed on and potentially distinct from this variability, flares with a duration of seconds to minutes have been seen (e.g., Wilms et al., 2007). Consistent with such flaring activity, the power spectra sometimes display an additional noise component below $\sim$0.01 Hz (e.g., Vikhlinin et al., 1994).

Cyg X-1 has also shown variable but persistent radio emission, which is due to the presence of a radio jet (Gallo et al., 2005; Rushton et al., 2012). The radio variability has been tracked during various years-long campaigns with, e.g., the Ryle telescope, AMI, or MERLIN (e.g., Fender et al., 2006; Gleissner et al., 2004; Rodriguez et al., 2015b). In addition to a correlation between radio and hard X-ray flux in both the hard and the soft state (Gleissner et al., 2004; Zdziarski et al., 2020), these campaigns also showed the (rare) presence of radio flares which may be similar to bubble ejection events in blazars (e.g., Fender et al., 2006; Pooley, 2017). This includes a case where an X-ray flare was followed by a radio flare $\sim$7 min later (Wilms et al., 2007).

X-ray flaring on intermittent timescale has also been observed in BH-LMXBs during their outburst. Worth highlighting is the rich variability of GRS 1915+105, particularly for its repeatability (Belloni, 2010), while less predictable flaring was seen, e.g., in V404 Cygni (e.g., Alfonso-Garzón et al., 2018; Tetarenko et al., 2017). Flaring in the radio band is an established signature of the hard to soft-state transition of BHB outbursts (Fender & Belloni, 2004). On much shorter timescales of $\lesssim\,1\,\mathrm{s}$ flaring and variability has been attributed to changes in the inner disk region and magnetic reconnection events above it (Lyubarskii, 1997; Uttley & Malzac, 2025). Here we report on a series of exceptionally bright flares observed with INTEGRAL on 2023 July 10. During these flares, which occurred during a time interval of little more than one hour and lasted only for $\sim$5–10 minutes each, the X-ray emission of Cyg X-1 was brighter by a factor of more than 20 compared to the brightest emission seen in the more than 20 Ms of INTEGRAL observations taken since its launch in 2002. In Sect. 2 we present the light curves of the flares. We then put the time of the flares in context of the long-term spectral and state evolution of Cyg X-1 in Sect. 3 and discuss the detailed behavior of the source in Sect. 4. We discuss and summarize our results in Sect. 5.

2 Strong flaring of Cyg X-1

Cyg X-1 has been a regular target of INTEGRAL since its launch in 2002 (e.g., Pottschmidt et al., 2003; Del Santo et al., 2003; Cadolle Bel et al., 2006). Since 2013 our team organized regular monitoring observations during the two observing windows each year, given by visibility constraints, with the aim to further constrain the hard X-ray polarization found with INTEGRAL (Laurent et al., 2011; Jourdain et al., 2012; Rodriguez et al., 2015b). As part of these regular observations, on 2023 July 10, during INTEGRAL’s revolution 2661, during a $<$1 h long time interval strong flaring was discovered through the MMODA interface (Neronov et al., 2021; Ferrigno et al., 2022)111A collection of spectra, images, and lightcurves of Cyg X-1 obtained with the INTEGRAL telescope is available at https://www.astro.unige.ch/astroordas/mmoda., with a flux that strongly exceeded the fluxes detected since the launch of INTEGRAL.

For a detailed analysis we extracted spectra and light curves from all the IBIS/ISGRI (Ubertini et al., 2003; Lebrun et al., 2003) and JEM-X (Lund et al., 2003) data of Cyg X-1 taken during revolution 2661 using Offline Science Analysis (OSA) software package 11.2. The flares were confirmed in two INTEGRAL Science Windows, i.e., slightly offset pointings of the INTEGRAL satellite222The science windows are 266100120010 (MJD 60135.273–60135.311, live time 1611 s), and 266100130010 (MJD 60135.312–60135.351, live time 1943 s), with a brief slew between them. Here and elsewhere in the paper, all MJDs refer to the local INTEGRAL satellite time system and are not barycentered. Since the two Science Windows have off-axis angles of $5\aas@@fstack{\circ}6$ and $5\aas@@fstack{\circ}2$, respectively, placing Cyg X-1 barely at the edge of the field of view of JEM-X, no spectral analysis with JEM-X is possible. The minute-scale evolution of the flares is shown in Fig. 1. The IBIS and JEM-X light curves clearly show the flaring behavior (Fig. 1a), as does the total event rate measured in the SPectrometer on INTEGRAL (SPI: Vedrenne et al., 2003, see Fig. 1c). No simultaneous information is available in the optical, since Cyg X-1 was outside the ${\sim}5^{\circ}$ field of view of INTEGRAL’s optical monitor.

In Fig. 2 we compare the count rate measured during 60 s long time bins during the flares to the distribution of the count rates found in the other ${\sim}330000$ 60 s-long time bins covering the ${\sim}23$ yr of Cyg X-1 INTEGRAL monitoring (a total exposure time of ${\sim}20\,$Ms ). The flares clearly represent by far the brightest events ever seen for Cyg X-1 with INTEGRAL.

The exceptional brightness of Cyg X-1 during the flares makes it important to confirm that they are not due to some other event in the field of view, such as a background flare or a gamma-ray burst. First, we check other sources in the same field for flaring. The light curves of Cyg X-3 and 3A 1954+319, the most significantly detected sources in the field of Cyg X-1, display only a minuscule flux increase during the flaring period (Fig. 1d and e). This slight increase is likely due to effects of the deconvolution algorithm in the OSA, where a small fraction of the flux of other sources in the field of view can be misattributed to the source under study (Goldwurm et al., 2003).

In addition, we generate the light curve of Cyg X-1 without relying on deconvolution. Since IBIS utilizes a coded mask, for a given pointing direction each source in the field of view only illuminates a subset of pixels. Figure 1f shows the count rate measured in those pixels of IBIS that are illuminated by Cyg X-1 (Pixel Illumination Fraction, $\mathrm{PIF}>0.7$) and compares it with those that are not illuminated. The only clear signal originates from Cyg X-1 itself. No increase is seen in the pixels that are not illuminated by Cyg X-1. This clearly illustrates that no background activity is the origin of the flares. Since the likelihood that the flares are from a serendipitous source very close to Cyg X-1 is extremely small, we conclude that the flares must come from Cyg X-1 itself.

Finally, while no pointed observations with instruments on other spacecraft were performed during the flares, Cyg X-1 was also monitored with MAXI, which provides on-demand light curves in user defined energy bands333http://maxi.riken.jp/mxondem/index.html. These light curves are shown together with the more long-term behavior around the flare in Fig. 3. A slight increase in count rate at the time of the flares is present (Fig. 3d) and might be attributable to the flare (partially) happening during a MAXI exposure, which are generally separated by one orbit of about 90 min.

We therefore conclude that the flares are indeed intrinsic to Cyg X-1.

3 The flaring episode in context

We now turn to the behavior of Cyg X-1 in the months surrounding the flaring episode. As shown in Fig. 3a, the flares occurred during a phase when the hard flux of Cyg X-1 rose. Such a behavior is commonly seen in this source during transitions from the soft state to the hard state. Indeed, the well-established classification by Grinberg et al. (2013), based on the MAXI monitor and illustrated in Fig. 4, places Cyg X-1 close to the soft-to-intermediate-state transition during the flare. During the following four daily scans, Cyg X-1 briefly entered the intermediate state.

The general source behavior in the time surrounding the flaring episode was typical for Cyg X-1. Reviewing the INTEGRAL spectroscopy during the days surrounding the flare, no strong or sudden changes in the hard photon index are apparent, implying a stable geometry of the Comptonizing plasma without any sudden change in flux or hardness (Fig. 3b).

Some black hole X-ray binaries show increased radio flaring during such state transitions. To investigate the possibility of a correlated radio flare, we used data from the Arcminute Microkelvin Imager (AMI, AMI Consortium et al., 2008). The 15 GHz radio monitoring consists of two 10 minutes pointings for each day with appropriate observing conditions, separated by a calibration observation. The radio flux density per 10 min pointing and for one linear polarization direction, i.e., Stokes I+Q, are shown in Fig. 3c. AMI provided snapshots shortly before the flaring episode on MJD 60135.0225 (6.4 hrs before the first flare) and then 40.4 h after the last flare at MJD 60137.0145, showing that the flux increased from 7.7 mJy to 16.5 mJy, which is within the range of typical variability in the soft to soft-intermediate state (Rodriguez et al., 2015b; Lubiński et al., 2020). In the following five days, the radio flux density continuously decreased. The $\sim$2 d data gap starting just before the flares prevents us from making a statement about the presence of a simultaneous radio flare.

4 Source behavior during the flares

Figure 5 shows a zoom-in onto the IBIS light curves at 5 s resolution. The first and third flare are characterized by very fast rises of 30–50 s in duration, with very fast doubling timescales of 15 s, followed by a “flat top” with an average IBIS count rate of ${\sim}500\,\mathrm{cts}\,\mathrm{s}^{-1}$, but with very strong normalized rms variability on timescales of tens of seconds, 51% and 36% for flare 1 and 3 at a time resolution of $\Delta t=10$ s, and slightly less at a time resolution of 20 s. The variability is significantly increased with respect to the 24% at $\Delta t=10$ s we measure outside the flares444The normalized rms-variability is only an approximate estimator for the variability in the case of red noise light curves (Vaughan et al., 2003), but unfortunately the shortness of the flares precludes a more detailed characterization of the variability properties, e.g., through power spectra.. At the end of the flares Cyg X-1 quickly returns to its previous count rate, and the variability goes back to its pre-flare value.

In contrast, the second flare is characterized by a sudden increase in normalized rms variability (39% at $\Delta t=10$ s) and a much slower rise (doubling time scale: ${\sim}150\,\mathrm{s}$) to a peak count rate of about $1400\,\mathrm{cts}\,\mathrm{s}^{-1}$. Unfortunately, the decay back to the pre-flare luminosity is only partly covered due to a spacecraft slew, but overall the shape of the second flare appears to be almost triangular or Gaussian in shape. Again, with 36% the rms variability during this episode is much stronger than outside of the event.

As discussed in Sect. 3, at timescales of hours to years, flux changes in Cyg X-1 are closely connected with spectral changes, as the source moves from the hard to the soft state and back. We illustrate this in the hardness-intensity-diagram of Cyg X-1 shown in Fig. 6, which is based on 100 s resolution data from all 22 years of INTEGRAL-monitoring of the source. In contrast, despite the large flux amplitude during the flares, the spectral hardness remains almost constant during these episodes (Fig. 6, teal data points and see also Fig. 1 for a light curve of the hardness ratio).

To further quantify possible changes of the spectral shape, we extracted IBIS spectra from each of the flares as well as a spectrum for the time outside the flares (defined as the time interval outside the shaded regions in Fig. 1). Due to the off-axis angle of $>5^{\circ}$ during the flare, we did not use the JEM-X data for spectral analysis. To take into account calibration uncertainties, a systematic uncertainty of 3% of the count rate was added in quadrature to the statistical uncertainty of each spectral bin.

We modeled the four spectra with typical models for X-ray spectra of black hole X-ray binaries in the hard X-rays: a simple powerlaw, an exponential cutoff power law (cutoffpl), a broken power law (bknpow), and thermal Comptonization (comptt; Titarchuk, 1994). The latter three are all able to describe a break or cutoff in the powerlaw spectrum. The best fitting parameters are listed in Table 1. All uncertainties are given at the 90% level for one parameter of interest, and we employed ISIS666https://space.mit.edu/cxc/isis/ for spectral fitting. The simple power law is unable to provide a satisfactory fit, leading to $\chi^{2}/\mathrm{dof}$ of $88.0/60$, $145.7/60$, and $120.3/60$ for the three flares, respectively. The three models with an intrinsic turnover or break, those shown in Fig. 7, all describe the spectra well.

This is in agreement with previous work, which has shown the break in the bknpow model to describe the spectra of BHBs similarly well to the cutoff powerlaw and even more complex jet models (Markoff et al., 2003; Nowak et al., 2011).

A direct comparison between the non-flare and flare spectra, determined by dividing them by each other, shows that the flare spectra are slightly softer (Fig. 7, bottom panel). In the cutoffpl-fits this softening is reflected by a decrease of the folding energy, similarly, while the comptt models describe the softening with a decrease in the temperature of the Comptonizing plasma, $kT_{\mathrm{e}}$, together with a strong increase in the optical depth (see Table 1). We caution, however, that the data are not good enough to distinguish between these different models, and that the energy coverage of IBIS alone is not sufficient to separate the spectral continuum from the relativistic reflection hump that contributes significantly in the IBIS-band. A direct physical interpretation of the spectral parameters is therefore unadvisable.

Using the comptt model, the peak 1–100 keV luminosities are $1.1\times 10^{38}\,\mathrm{erg}\,\mathrm{s}^{-1}$, $2.6\times 10^{38}\,\mathrm{erg}\,\mathrm{s}^{-1}$, $2.2\times 10^{38}\,\mathrm{erg}\,\mathrm{s}^{-1}$, for flare 1 through 3, while the total 1–100 keV fluence contained in the three flares is $3.01\,\times 10^{40}\,\mathrm{erg}$, $4.74\,\times 10^{40}\,\mathrm{erg}$, and $3.84\,\times 10^{40}\,\mathrm{erg}$. As these luminosities rely on the extrapolation of the spectral fits towards lower energies, they are only a rough lower limit. Even though we do not have direct spectral data, however, we can infer from the JEM-X that the soft emission likely shows an increase similar to the increase of the luminosities in the 30–100 keV range, which are $2.3\times 10^{37}\,\mathrm{erg}\,\mathrm{s}^{-1}$, $4.3\times 10^{37}\,\mathrm{erg}\,\mathrm{s}^{-1}$, $3.5\times 10^{37}\,\mathrm{erg}\,\mathrm{s}^{-1}$, for the three flares, respectively.

The hard photon index $\Gamma_{1}\sim 2.1\textrm{--}2.4$, inferred from the brokenpl fits, during the three flares and the surrounding observation matches what has been seen in Cyg X-1 during the soft to intermediate state (e.g., Nowak et al., 2011; Lubiński et al., 2020). It is remarkable that the spectra during the flares, especially flare 1 and flare 3, show very little change in spectral slope. Only the spectra of flare 2 display mild softening. This becomes visible through the ratio with the non-flare spectrum, which decrease towards higher energies, as seen in Fig. 7. This contrasts with the rather vertical evolution in the hardness-intensity diagram of Fig. 6, but might result from the chosen energy bands and time binning.

5 Discussion and Conclusions

Before we discuss the possible physical origin of the flaring, we briefly summarize their main properties:

• •

the three flares represent extreme source behavior that has not been seen previously in the 21 years of monitoring of Cyg X-1 with INTEGRAL, nor in the earlier RXTE monitoring between 1997 and 2012,

• •

the flares occurred during the soft-intermediate state, when Cyg X-1 was moving towards the hard state,

• •

the flares occurred at orbital phase $\phi_{\mathrm{orb}}=0.01$ based on the ephemeris of Brocksopp et al. (1999), i.e., close to upper conjunction of the black hole,

• •

the flares have peak luminosities of 1–100 keV luminosity of $1.1\textrm{--}2.6\times 10^{38}\,\mathrm{erg}\,\mathrm{s}^{-1}$ ($4.1\%\textrm{--}9.7\%\,L_{\mathrm{Edd}}$), a dynamic flux range of ${\sim}15$, and a duration of about 400 s each, with fluences of $3\textrm{--}5\times 10^{40}$ erg each,

• •

the intensity profiles are complex with fast rise and decay times ${\sim}10\,\mathrm{s}$ for the first and third flare, and a slow rise and fast decay for the second flare,

• •

during all three flares the normalized rms variability is significantly increased,

• •

there is little spectral change in the hard X-rays, with only a slight softening $>30\,\mathrm{keV}$ (see ratio panel in Fig. 7).

Besides their timing properties mentioned above, it is the extreme peak luminosities that make the flares stand out above the typical variability seen in Cyg X-1. To our knowledge, the only comparable event in Cyg X-1 was detected by RXTE in 2005 April (Wilms et al., 2007). This event had a similar duration of ${\sim}10\,\mathrm{min}$ and coincided with a radio flare that was delayed by about 400 s with respect to the X-ray flare. Similarly to the flares observed by INTEGRAL, Cyg X-1 was in the intermediate state, transitioning towards the hard state. A notable difference is the lower dynamic range: while we observe a ${\sim}15\times$ increase in hard X-ray flux, Wilms et al. (2007) saw the radio and X-ray flux increase by only about a factor of three777Since the PCA and IBIS energy ranges differ significantly, a more precise comparison of the fluxes is difficult, in addition, the peak X-ray flux of the RXTE flare could be underestimated as RXTE only caught its decay.. Radio flares with behavior similar to that studied by Wilms et al. (2007) were also discussed by Fender et al. (2006), including a strong 140 mJy radio flare with a duration of about one hour during the intermediate state of Cyg X-1, but unfortunately these radio flares lacked simultaneous X-ray data. The orbital phase aligns closely with phase zero. However, we would primarily expect the absorption to vary as a line-of-sight effect (Grinberg et al., 2015; Szostek & Zdziarski, 2007) and not the phenomena related to the accretion flow itself. For comparison, the flare reported in Wilms et al. (2007) occurred around orbital phase 0.82. Concerning the spectral shape, the necessity of a cutoff does point more towards thermal origin of the X-ray emission, rather than synchrotron emission. Jet models such as those discussed by Markoff et al. (2005); Nowak et al. (2011); Maitra et al. (2017); Kantzas et al. (2021) can, however, produce spectral shapes more complex than a simple powerlaw. A jet origin for the observed flares can therefore not be excluded based on spectral shape.

Before investigating the possible origin of the flares, we emphasize that the flares discussed here are different from shorter-term variability in accreting black holes, which is sometimes explained in terms of shot-noise (e.g., Bhargava et al., 2022; Gierliński & Zdziarski, 2003). The latter is thought to be connected to magnetic reconnection and the evolution of plasmoids in a current sheet above the disk (e.g., Ripperda et al., 2020; El Mellah et al., 2022; Merloni & Fabian, 2001) and occurs on timescales of $10\textrm{--}100\,r_{\mathrm{g}}/c$. The variability of $\sim$10 min seen here, however, corresponds to $10^{6}\,r_{\mathrm{g}}/c$, or approximately the light crossing time between the donor and the black hole888We caution that the term “flare” is not well-defined in astronomy, and may denote events on very different timescales. For black hole X-ray binaries alone, “flare” has been used for large amplitude flux changes on time scales of seconds, minutes, days, and months, which are likely to be due to very different physics. In the following, we use the term solely for flux changes on timescales less than a few 100 minutes.. Conversely, following a simple scaling by mass as appropriate for black holes, the minute-scale variability as seen in AGNs corresponds to millisecond variability in BHBs.

Such short-term variability has previously also been attributed to the innermost region of the disk. Variable seed photons from the disk can then be upscattered in the corona to induce variability in harder energy-bands (Uttley & Malzac, 2025, and references therein). In such a scenario, the dynamic timescale is, however, of the order of $10^{-3}$ s, comparable to the Keplerian timescale, and therefore much faster than the observed flaring episodes (Lyubarskii, 1997; Done et al., 2007). An overview of the different timescales in accreting black holes is given by Kara & Garc´ıa (2025). If the variability originates from the disk, it must therefore have its origin in the outer regions, where the Keplerian and viscous timescales are much longer. In such a case, the model of propagating fluctuations could be considered a possible explanation for the flares. Under this assumption, aperiodic fluctuations in the accretion are introduced, classically by turbulent changes in the disk viscosity at all radii of the disk. These fluctuations propagate inwards towards the hotter region of the accretion disk, reproducing the observed time lags and power spectral density (see, e.g., Lyubarskii, 1997; Uttley et al., 2014; Ingram & Done, 2011). A sudden increase in the mass accretion rate originating at a radius of ${\sim}10^{2}\,r_{\mathrm{g}}$ could lead to variability on the order of 100 s, corresponding to the viscous time scale. The inwards propagation is, however, a diffusion process and smoothens out the induced variability. The rapid rise and decay time of the flares are therefore atypical for propagating fluctuations. The model is further usually only applicable to stochastic variability, not individual flares. Due to a lack of soft spectra and light curves, however, such a scenario cannot be fully ruled out.

Similarly, an explanation though the dissipation time $t_{\mathrm{D}}$ after magnetic heating, as used by Merloni & Fabian (2001) following earlier work by Haardt et al. (1994) requires a region size of

[TABLE]

with the unitless dissipation velocity $b=c/v_{\mathrm{diss}}$. This is much larger than the inner regions of the accretion disk. Accretion fluctuations alone are therefore disfavored as explanations, but could still serve as triggers for downstream changes in the jet or corona.

For examples of flaring on times scales of hours we can expand our view to other BHB systems beside Cyg X-1. Flaring has been observed in the radio band for several transient black hole X-ray binaries, where they occur around state transitions and may be correlated with changes in the X-ray timing behavior (e.g., Fender et al., 2009; Miller-Jones et al., 2012). This has relevance to our discussion as these have also been attributed to ejection events through the jet. As these state transitions, however, occur over days, they are therefore phenomena on very different timescales and the physical mechanism behind these flares could be very different. Direct observations of such hour-long radio flares are rare. Homan et al. (2020) describe a strong radio flare in MAXI J1820+070, lasting for $\sim$2.5 h and reaching $\sim$50 mJy, which was accompanied by a quasi-simultaneous weak flare in the 7–12 keV NICER band. In this case, however, the X-ray flux only increased by a few percent. Flaring on a timescale of minutes in the optical and X-rays has also been studied for the 2015 outburst of V404 Cygni (Rodriguez et al., 2015b; Tetarenko et al., 2017; Alfonso-Garzón et al., 2018). Here, these flares display a similar complexity in shape to those seen in Cyg X-1, albeit at a much lower dynamic range. Explanations for the type of minute-long flares discussed here build on the observation that, for those flares where simultaneous broad-band data exist, flaring is not limited to the X-rays alone, but seen across the electromagnetic spectrum. Time delays are common, where longer wavelengths are delayed with respect to shorter ones (e.g., Fender et al., 2023, and references therein). As suggested by Fender & Belloni (2004) and Wilms et al. (2007), a possible model explaining these events is some kind of ejection event, similar to the flaring seen in blazars first discussed by van der Laan (1966). In such models, a bubble of relativistic electrons is ejected from the accretion flow and subsequently cools down, moving the peak of the emission to longer wavelengths. The slight spectral changes seen in the hard X-rays are interpreted as a change of the Compton-$y$-parameter due to the expansion and cooling of the ejected material. See Younsi & Wu (2015) for detailed computations of plasmoid ejections that take general relativistic effects into account and suggest that strong variability and multiple reflares on timescales of tens of $r_{\mathrm{g}}/c$ are in principle possible. Such ejections have been invoked for the explanation of flares in the lightcurves of V404 Cygni with timescales of ${\sim}200\,\mathrm{s}$ by Maitra et al. (2017) on the basis of multiwavelength data. In this case, a predominant disk origin was explicitly ruled out on the basis of correlated optical variability. The very characteristic heart-beat variability and particularly large orbital period, and therefore disc make V404 Cygni a unique system (Fender & Belloni, 2004; Steeghs et al., 2013). This might limit to which we can transfer explanation for flaring in V404 Cygni to Cyg X-1, as the disc might behave very differently. Yet we hold that close to the black-hole the physics around the black hole should still be similar and V404 Cygni serves as a valuable comparison.

As suggested by Rodriguez et al. (2015a) for V404 Cygni, the amplitude of the flaring could be increased due to directly boosted radiation when the jet axis is pointed directly towards the observer. Such a boosting would imply a fairly large deviation of the jet direction during the flares with respect to its default orientation. The nominal angle between the jet axis (and the orbital angular momentum) and our line of sight at upper conjunction ($\phi_{\mathrm{orb}}=0$) is ${\sim}60^{\circ}$ (Krawczynski et al., 2022; Miller-Jones et al., 2021; Orosz et al., 2011). Such a large deviation is difficult to achieve. Models for the warping of the inner accretion disk in Cyg X-1 imply a warp of only ${\sim}30^{\circ}$ (Ibragimov et al., 2007) in order to explain the $\sim$150 d or $\sim$294 d superorbital variability in the system (Zdziarski et al., 2011; Brocksopp et al., 1999; Priedhorsky et al., 1983; Kemp et al., 1983) and the optical polarization variability (Kravtsov et al., 2023, and references therein). It is therefore unlikely that regular disk warping and variable boosting is the sole source of variability. An alternative explanation for the flares is the interaction of structures in the stellar wind of the high-mass donor star and the jet. Based on numerical modeling of clump-jet interaction, Perucho & Bosch-Ramon (2012) show that for winds with a steep clump size distribution, it is possible for large clumps to enter the jet and be completely disrupted, potentially even choking off the jet. The simulations by Perucho & Bosch-Ramon and Araudo et al. (2009) predict rare, but luminous flares with durations of $1000\,\mathrm{s}\,(R_{\mathrm{clump}}/10^{11}\,\mathrm{cm})(c_{\mathrm{c}}/10^{8}\,\mathrm{cm})$, where $R_{\mathrm{clump}}$ is the clump radius and $c_{\mathrm{clump}}=(2L_{\mathrm{jet}}/\pi v_{\mathrm{jet}}\rho_{\mathrm{clump}}R_{\mathrm{jet}}^{2})^{1/2}$ is the speed of sound in the clump. Here, $v_{\mathrm{jet}}$ is the speed of the jet, $L_{\mathrm{jet}}$ its luminosity, and $R_{\mathrm{jet}}$ its radius at the point of interaction, while $\rho_{\mathrm{clump}}$ is the density of the clump. Observations of X-ray dips in Cyg X-1 show that the wind of its donor, HDE226868, is clumpy (Hirsch et al., 2019; Grinberg et al., 2015, and references therein). Line-of-sight variations of the absorbing column are stronger close to upper conjunction of the black hole (Grinberg et al., 2015), indicating that more clumps pass through our line of sight, probably close to the black hole. This is exactly the time interval when the flare was observed, although this is probably a coincidence. We note, however, that if the flare is due to a clump destruction event it is puzzling why its spectral shape would be similar to that seen outside the flare, and why we see three flares and not just one. In addition, while clumps are normal in high-mass stellar winds (Owocki et al., 1988, and references therein), with a filling factor of $\sim$11% for Cyg X-1 (Rahoui et al., 2011), they are not expected to exist in low-mass X-ray binaries, where flaring is also observed. To conclude, we serendipitously observed an unprecedented set of ${\sim}$10 minute long, X-ray flares with fluences in excess of $10^{40}$ erg, with hard X-ray luminosities that are at least three times higher than anything seen before in the $\geq 22\,$years of INTEGRAL monitoring of Cyg X-1. The flares occurred in a state where the accretion flow of black hole binaries is hypothesized to be unstable. The strong flaring might be related to (i) some kind of ejection event, (ii) a restructuring of the outflow (“jet”) in the system, or (iii) the interaction of a clump in the stellar wind with the jet. The observations presented here illustrate the need for continued monitoring even of supposedly “well-known” sources, since it allows us to catch dramatic and very rare events in such systems.

Acknowledgements.

We especially acknowledge the crucial contribution of Katja Pottschmidt – not only to this paper but the field of Black-hole timing in general. Without her support, mentorship, and scientific insight this work would not have been possible. Her untimely passing is felt sorely. This work has been partially funded by the Bundesministerium für Wirtschaft und Klimaschutz under Deutsches Zentrum für Luft- und Raumfahrt grant 50 OR 1909. This research is supported by the DFG research unit FOR 5195 ‘Relativistic Jets in Active Galaxies’ (project number 443220636, grant number WI 1860/20-1). TB & JR acknowledge partial funding from the French Space Agency (CNES). The material is based upon work supported by NASA under award number 80GSFC24M0006. MP acknowledges support by the Spanish Ministry of Science trough Grant PID2022-136828NB-C43, and by the Generalitat Valenciana through grant CIPROM/2022/49. The research is based on observations with INTEGRAL, an ESA project with instruments and science data center funded by ESA member states (especially the PI countries: Denmark, France, Germany, Italy, Switzerland, Spain) and with the participation of Russia and the USA. This research has made use ISIS 1.6.2-51 (Houck & Denicola, 2000) and of ISIS functions (ISISscripts) provided by ECAP/Remeis observatory and MIT (https://www.sternwarte.uni-erlangen.de/isis/).