Hard X-ray Tail Discovered in the Clocked Burster GS 1826-238
HHard X-ray Tail Discovered in the Clocked Burster GS − J. Rodi , , E. Jourdain , , and J. P. Roques , Universit´e de Toulouse; UPS-OMP; IRAP; Toulouse, France CNRS; IRAP; 9 Av. Colonel Roche, BP 44346, F-31028 Toulouse cedex 4, France
ABSTRACT
The LMXB NS GS 1826 −
238 was discovered by
Ginga in 1988 September. Due to thepresence of quasi-periodicity in the type I X-ray burst rate, the source has been a frequenttarget of X-ray observations for almost 30 years. Though the bursts were too soft to be detectedby
INTEGRAL /SPI, the persistent emission from GS 1826 −
238 was detected over 150 keVduring the ∼
10 years of observations. Spectral analysis found a significant high-energy excessabove a Comptonization model that is well fit by a power law, indicating an additional spectralcomponent. Most previously reported spectra with hard tails in LMXB NS have had an electrontemperature of a few keV and a hard tail dominating above ∼
50 keV with an index of Γ ∼ − −
238 was found to have a markedly different spectrum with kT e ∼
20 keV and a hardtail dominating above ∼
150 keV with an index of Γ ∼ .
8, more similar to BHXRB. We reporton our search for long-term spectral variability over the 25 −
370 keV energy range and on acomparison of the GS 1826 −
238 average spectrum to the spectra of other LMXB NS with hardtails.
Subject headings:
X-rays: general — X-rays: binaries — Neutron Stars: individual (GS 1826 −
1. Introduction
GS 1826 −
23 was discovered by
Ginga on 1988September 8 (Makino et al. 1988) and was initiallyconsidered to be a black hole candidate becauseits hard spectrum and flickering were similar toCyg X-1 (Tanaka 1989). Not until Ubertini et al.(1997) were X-ray bursts reported from the sourceto reveal the compact object as a neutron star(NS). Homer et al. (1998) confirmed this resultby detecting optical bursts. They also measureda 2-day orbital period for GS 1826 − ∼ BeppoSAX exposure time, Ubertini etal. (1999) observed 70 X-ray bursts and found aquasi-periodicity in the burst recurrence time of5.76 h, which suggests a stable accretion rate.Observations by the ROSAT-PSPC identifiedthe optical counterpart, determining that the sys-tem is a low mass X-ray binary (LMXB) (Motchet al. 1994; Barret et al. 1995). From Barret etal. (1995) the estimated distance to the source is4 −
10 kpc, while in’t Zand et al. (1999) and Konget al. (2000) place an upper limit of 8 kpc on the distance.LMXB NS systems can be broadly classified aseither Z-sources or atoll sources. Z-sources traceout a ’Z’ shape in a color-color diagram while atollsources trace out a ’banana’ shape (Church et al.2014). Muno et al. (2002) tentatively classified GS1826 −
238 as an atoll source though the source hadshown very little variability in its color-color dia-gram over the 5 years of observations they studied.Previous X-ray observations of GS 1826 − INTE-GRAL allow for a study of the high-energy spec-trum of this source above 100 keV.In this work, we looked at the hard X-ray/softgamma-ray spectrum of GS 1826 −
238 out to 370keV having analyzed ∼
11 Ms of SPI data over1 a r X i v : . [ a s t r o - ph . H E ] D ec he span of ∼ −
238 spectrum withthe spectra of other LMXB NS that have beenreported to have hard tails, specifically the atollsource 4U 1728-34.
2. Instrument and Observations
On 2002 October 17, the
International Gamma-ray Astrophysics Laboratory ( INTEGRAL ) waslaunched from Baikonur, Kazachstan. The satel-lite has an eccentric ∼ . ◦ (Jensen et al. 2003). Thespectrometer INTEGRAL /SPI spans the 20 keV − − INTEGRAL hasobserved the Galactic Center region at approxi-mately 6 month intervals and thus has created along observational baseline for studying sources inthis region.We have utilized this long baseline by analyz-ing 223
INTEGRAL revolutions spanning MJD52719 − − )using only the revolutions where GS 1826 − ◦ of the SPI pointing direction forat least 10 science windows ( ∼ INTE-GRAL revolutions during each observation period,the beginning and ending MJD of the period andthe exposure time during the period.Because GS 1826 −
238 is in the Galactic Centerregion, there will be some contribution to the de-tector count rate due to diffuse positronium emis-sion above ∼
300 keV, which could potentially re-sult in an artificial high-energy spectral tail forthe source. To mitigate this effect, the expectedincident count-rate due to diffuse positronium wasmodeled based on parameters reported by Bouchetet al. (2008) and then subtracted from the rawdata before analyzing the data with SPIDAI.In the spectral analysis, the SPI data weregrouped into 50 energy bins spanning 22 − Publicly available interface developed at IRAPto analyze SPI data. Available at http://sigma-2.cesr.fr/integral/spidai. See description in Burke etal. (2014) keV. The first two energy channels were ignoredin spectral fitting because of uncertainties in theenergy response (Jourdain & Roques 2009) thusreducing the energy range to 25 −
650 keV. Nosignificant long-term emission was detected above370 keV thus the spectral analysis has been limitedto 25 −
370 keV.
3. Results3.1. Temporal Variability
GS 1826 −
238 exhibits frequent type I X-raybursts with a quasi-periodicity of 5.76 h (Ubertiniet al. 1999). The duration of these bursts are typ-ically ∼
100 s in the 2 −
10 keV energy range (Jiet al. 2014) and are soft enough to have a negligi-ble effect >
25 keV. GS 1828 −
238 also undergoeslonger duration variability, which can be seen inFigure 1. The SPI light curve is shown for fourbroad energy bands (25 −
50 keV, 50 −
150 keV,150 −
450 keV, and 450 −
650 keV) covering MJD52700 − INTEGRAL revolu-tion. In each panel the solid line denotes a flux of 0mCrab (except for the top panel), and the dashedline denotes the long-term average flux in that en-ergy band. For each panel, the Period numbershave been plotted above the corresponding data.The total 25 −
50 keV significance is 265 . σ , the50 −
150 keV significance is 65 . σ , the 150 − . σ , and the 450 −
650 keVsignificance is 1 . σ . GS 1826 −
238 shows rela-tively little variability within a period, apart fromPeriod 3. During this period, the 25 −
50 keV fluxdecreased from ∼
120 mCrab to below 20 mCrabin ∼ ∼
90 mCrab inapproximately 40 days.GS 1826 −
238 has undergone two large dips af-ter this work (during
INTEGRAL revolutions notcurrently publicly available). A dip began on MJD56816 (2014 June 8) when
MAXI detected a de-crease in the hardness ratio from ∼ . −
10 keV flux increasing from roughly 50mCrab to 140 mCrab on MJD 56822 (2014 June14) (Nakahira et al. 2014). During this time the15 −
50 keV
Swift /BAT flux decreased, as can beseen in the publicly available light curve . Based http://swift.gsfc.nasa.gov/results/transients/Ginga1826-238/ INTEGRAL /SPI observations of GS 1826 − INTEGRAL
Revolutions Time Exposure Time(MJD) (ks)1 0053 − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − −
238 spanning MJD 52700 − −
650 keV. The dashed line in each panel corresponds to the average flux, and the solid line correspondsto a flux of 0 mCrab. The Period number has been plotted above the corresponding data for each panel.on the BAT data, the flux decreased from ∼ ∼ ∼ MJD 56850. Afterwhich, the flux began to increase over the spanof about 40 days, and finally recovered to ∼ ∼ ∼
100 mCrab to fluxes consistent with0 mCrab. The source spent only a few days inthis low flux state before recovering to the initialflux level over the course of about 30 days until ∼ MJD 57210 (2015 July 7). The first two dips show similar timescales of afew days for the flux decreases. The third dip be-gins with a slow decline over ∼
20 days, beforestarting a fast decline lasting a few days. All ofthe dips reached a minimum flux of near or con-sistent with 0 mCrab. Also, the timescales for theflux increases are roughly 30 −
40 days long. Thedurations of the dips also show a difference, withthe second burst having lasted a few 10’s of daysnear its minimum flux level while the first and lastdips spent only a few days at their minimum fluxlevels.4ig. 2.—
Left Column: ( Top ) Plot of cutoff energy for each period when the photon index is held fixed atΓ = 1 .
59. Solid line denotes best-fit cutoff energy from average spectrum ( E cut = 57 keV). ( Bottom ) Plot ofsignificance of period cutoff energy from average spectrum. Solid line denotes deviation of 0 σ , and dashedlines denote deviations of +3 σ and − σ . Right Column: ( Top ) Plot of photon index for each period whenthe cutoff energy is held fixed at E cut = 57 keV. Solid line denotes the best-fit photon index from averagespectrum (Γ = 1 . Bottom ) Plot of significance of period photon index from average spectrum. Solidline denotes deviation of 0 σ , and dashed lines denote deviations of +3 σ and − σ . Due to a low signal-to-noise ratio during a sin-gle revolution, a search for spectral variability ona period timescale was performed. To do so, anaverage spectrum was generated using all periodsexcept Period 3, which shows the large flux vari-ability. The average spectrum was poorly fit bya power law model with a best-fit photon indexof Γ = 2 . ± .
01 with χ /ν = 11 .
80 ( ν = 36),but was significantly better fit by a cutoff powerlaw with Γ = 1 . ± .
05 and a cutoff energy of E cut = 57 ± χ /ν = 1 .
56 ( ν = 35) withresiduals suggesting a high-energy excess above ∼
150 keV. To compare the period spectra withthe average spectrum, each period spectrum wasfit once with the photon index fixed to the aver-age value (Γ = 1 . E cut = 57 keV), and thephoton index was allowed to vary.In the left column of Figure 2, the top panelshows the cutoff energy for each period when thephoton index is fixed with the solid line mark-ing the cutoff energy from the average spectrum.The bottom panel shows how significantly the cut-off energies deviate from the long-term average.The solid line denotes a value of 0 σ away fromthe average value while the dashed lines denote+3 σ and − σ away from the average value. Thetop panel in the right column shows the photon in-dex when the cutoff energy is fixed with the bot-tom panel again showing the significance of thedeviation from the long-term average.When the photon index is fixed, only two pe-riods have a cutoff energy > | σ | away from theaverage. The two outlying periods are 9 and 18,5ig. 3.— Top : Plot of 25 −
50 keV SPI light curvefor Period 3 revolutions.
Bottom : Plot of photonindex for Period 3 revolutions. Solid line denotesthe photon index of the average spectrum withΓ = 2 .
52. Dashed vertical lines in both panelsdenote start and stop times of potential soft statebased on ASM data.and both are ∼ σ below the average. Similarly,when the cutoff energy is fixed, these two peri-ods again show a large deviation from the average( ∼ . σ above the average). Also, a third periodis ∼ σ below the average, Period 12. When thephoton index is fixed, Period 12 has a deviation ofa little under +3 σ . As Periods 9, 12, and 18 showdeviations > | σ | away from the average with ei-ther the photon index fixed or the cutoff energyfixed, these periods have been removed from thelong-term analysis and analyzed separately. (Seebelow.) Fig. 4.— Average 25 −
370 keV spectrum.
CompTT model shown with dashed line.
CompTT + powerlaw model shown with solid line. Based on the top plot in the right column of Fig-ure 2, Periods 9 and 18 have photon indexes thatare relatively soft compared to the other periodswhile the photon index for Period 12 is harder thanall of the other periods. When the cutoff energy isfixed to 57 keV, the photon indexes for Periods 9,12, and 18 are 1 . ± .
06 ( χ /ν = 1 . . ± . . . ± .
09 (1 . . ± .
05 ( χ /ν = 0 . −
50 keV light curve (Figure 1),has fit values consistent with the average spectrumin both cases. Such a dip is suggestive of a hard-to-soft state transition. Analysis of the 2 − RXTE /All-Sky Monitor (ASM)shows a possible hard-to-soft transition on roughlyMJD 53061.5 with a possible soft-to-hard transi-tion on about MJD 53066.5. During this 6 daytime period, the ASM data are ∼ . σ above thelong-term average flux, but the source variabilityseen outside this MJD range and the relatively lowsignificance make a confident detection of a statechange difficult.The interpretation of a short lived soft state isnot inconsistent with the SPI data. In this sce-nario, the hard-to-soft transition occurs ∼ −
50 keV light curve for Period 3 while the bot-tom panel shows the best-fit photon index for eachrevolution. In the bottom panel, the solid line de-notes the index of the long-term average spectrumΓ = 2 . ∼ . ∼ The updated spectrum with Periods 9, 12, and18 and revolution 0168 from Period 3 removedhad an exposure time of 9.5 Ms. As before, acutoff power law model resulted in a large χ /ν (1.41) with residuals above ∼
150 keV, suggest-ing a high-energy component. Thus the data werefit with a cutoff powerlaw + powerlaw model.This model greatly decreased the χ /ν to 0.81with parameters Γ = 1 . ± . E cut = 35 ±
10 keV, and Γ = 1 . ± .
98. An F-test re-sulted in a probability of chance improvement of3 . × − , which corresponds to 4 . σ .For a physical interpretation, the data were fit to a CompTT
Comptonization model (Titarchuk1994) assuming spherical accretion geometry re-sulting in model parameters with an electron tem-perature of kT e = 26 ± τ = 3 . ± .
16 with χ /ν = 2 .
12 ( ν =35) with the high-energy residuals still present.Adding a high-energy power law component yieldsfit parameters of kT e = 18 ± τ = 4 . ± .
42, and Γ = 1 . ± .
61 with χ /ν = 0 .
87. Froman F-test, the probability of chance improvementwas 1 . × − , which corresponds to 5 . σ . Thusthe addition of a high-energy component is statis-tically significant.The data were also fit to CompTT model con-volved with a reflection component (Magdziarz &Zdziarski 1995) with the reflection fraction fixedto 1. The fit with a reflect(compTT) model re-sulted in fit parameters with kT e = 34 ± τ = 2 . ± .
21 ( χ /ν = 1 . ν = 35). The elec-tron temperature is significantly higher than the CompTT alone, but, like the
CompTT model alone,this model also fails to adequately fit the high-energy tail.Figure 4 shows the average spectrum plottedwith the
CompTT model shown with a dashed lineand the
CompTT + powerlaw model shown witha solid line. From the plot, it is clear that the CompTT model fails to describe the data above ∼
150 keV. As the positronium emission has beenremoved from the data, the hard tail is likely theresult of emission from GS 1826 −
4. Discussion4.1. Comparison with Previous Observa-tions
The interesting bursting behavior of GS 1826 −
238 means that it has been observed in X-rays of-ten since its discovery. This allowed for severalcomparisons with the SPI results. Cocchi et al.(2010) studied GS 1826 −
238 in the 3 −
200 keVenergy range using
INTEGRAL /JEM-X and IS-GRI over three time periods from 2003 April to2006 November. These observations were during
INTEGRAL revolutions 0061 − − −
200 keVflux to be roughly constant when comparing rev-7lutions 0061 − − − − ∼
30% lower. The science windows within 5 ◦ ofthe SPI pointing direction were used to match thefield of view of JEM-X and thus use the same ob-servations as Cocchi et al. (2010). Analysis of the25 −
200 keV SPI data found results in agreementwith those from Cocchi et al. (2010).The SPI data during these intervals were fit toa cutoff power law model over the 25 −
200 keV en-ergy range to best match the Cocchi et al. (2010)energy range. Figure 5 shows the contour plotsfor each interval with the best fit parameters by ablack ’X’ for SPI and by a black square for Coc-chi et al. (2010). The contours mark the 68 . − − .
05 and E cut = 32 keV ( χ /ν = 1 . ν =28), Γ = 1 .
61 and E cut = 55 keV ( χ /ν = 0 . .
11 and E cut = 27 keV ( χ /ν = 0 . . in’t Zand et al. (1999) analyzed the 0 . − BeppoSAX data covering 1997 April 6 . − . . ± .
03, and the cutoff energy foundwas E cut = 51 . ± .
03 keV. When the long-term average SPI spectrum was fit to the 25 − .
56 with a E cut = 56 ± χ /ν = 1 .
22 ( ν = 28).Figure 6 shows the confidence regions for the SPIfit using the same contour levels as in Figure 5.The fit parameters from in’t Zand et al. (1999)are shown by a black square and are well out-side the 99% confidence region, indicating spectralvariability between the SPI observations and the BeppoSAX observations. Additionally, in’t Zand et al. (1999) fit the 60 −
150 keV data to a powerlaw to compare with OSSEobservations (Strickman et al. 1996) in the 60 − BeppoSAX data were best fit by aphoton index of Γ = 3 . ± . . ± .
5. The 60 −
370 keV SPIdata are well fit by a photon index of Γ = 3 . ± .
07, in agreement with both results, indicatingthat the hard X-ray/soft gamma-ray spectrum hasremained relatively constant for almost 20 yearsdespite variability at lower energies.
Since the discovery of a hard tail in GX 5-1 with
Ginga (Asai et al. 1994), observations of LMXBNS have found that similar hard tails are not un-common (e.g. Sco X-1 (Di Salvo et al. 2006; D’Ai et al. 2007; Revnivtsev et al. 2014), GX 17+2(Di Salvo et al. 2000), 4U 1636 −
53 (Fiocchi et al.2006), and 4H 1820 −
30 (Tarana et al. 2007)) andhave been seen in both Z-sources (Sco X-1 and GX17+2) and atoll sources (4U 1636-53 and 4H 1820-30). A majority of the spectra showing tails can becharacterized by a soft state with a Comptoniza-tion component with a temperature of kT e ∼ ∼ −
50 keVwith a slope of Γ ∼ − −
238 is considerably different with the sourcein a hard state with a Comptonization compo-nent extending out to ∼
150 keV after which thehard tail becomes the dominant component withΓ = 1 .
77. Due to the large error, the photon indexis consistent with the lower portion of the Γ ∼ − −
34 in the hard state withspectra similar to GS 1826 − CompTT + powerlaw model,the electron temperatures for these spectra were kT e ∼ .
7, 9 .
1, and 10 . ∼ .
5, 1 .
9, and 1 .
8. The errors on thetails with Γ < ∼ −
3. Thus the hard tails ob-served in both the soft and hard states are likelyto be the same with the hard tail “hidden” bythe Comptonization component when the electrontemperature increases.8ig. 5.— Confidence regions in the E cut − Γ parameter space for
INTEGRAL revolutions 0061 − left ),0119 − center ), and 0495 ( right ) with contours for the 68 .
3% (black), 90% (red), and 99% (green)confidence regions. The best fit parameters from SPI are plotted with a black ’X’ and the best fit parametersfrom Cocchi et al. (2010) are plotted with a black square.Fig. 6.— Contour plot showing the confidence re-gions for the SPI long-term average spectrum fitto a cutoff power law in the 25 −
200 keV energyrange. Contours mark the 68 .
3% (black), 90%(red), and 99% (green) confidence regions. The
BeppoSAX fit parameters from in’t Zand et al.(1999) are marked by a black square. Tarana et al. (2011) found that the hard statesof 4U 1728 −
34 could be well fit using a hybridthermal/non-thermal
CompPS model (Poutanen &Svensson 1996). The fit parameters from the threeobservations were: (1) kT e ∼
27 keV, τ y ∼
3, and p ∼ .
2, (2) kT e ∼
24 keV, τ y ∼ .
8, and p ∼ . kT e ∼
20 keV, τ y ∼ .
4, and p ∼ .
6. The GS1826-238 average spectrum was fitted to the samemodel, resulting in best fit parameters of kT e =31 ± τ y = 2 . ± . p = 0 . ± . χ /ν = 0 .
82 ( ν = 34). The GS 1826 − −
5. Conclusion
In this work, we presented ∼
11 Ms of GS1826 −
238 observations from
INTEGRAL /SPIthat span ∼ . − −
50 keV flux aroundMJD 53000 with a possible short-lived transitionto a soft state before recovering to near pre-diplevels. Spectral analysis showed little variabilityoutside of Periods 9, 12, and 18 and revolution0168 in Period 3. (See Section 3.2.1.) This al-lowed us to look at 9.5 Ms of exposure time in theaverage spectrum which spanned the 25 −
370 keV9nergy range.The spectrum was dominated by a Comptoniza-tion component up to ∼
150 keV with a high-energy excess extending up to ∼
400 keV. Fit-ting the data to either a cutoff power law or a
CompTT model required a power law tail to achievean acceptable χ /ν . For the cutoff power lawmodel the data were well described by Γ =1 . E cut = 35 keV, and Γ = 1 .
80. In the caseof the
CompTT model, the best fit parameters were kT e = 18 keV, τ = 4 .
39, and Γ = 1 .
77. The spec-trum was also well fit by a hybrid thermal/non-thermal
CompPS model with kT e = 31 keV, p =0 .
65, and τ y = 2 . BeppoSAX results from in’tZand et al. (1999) found a significantly harder pho-ton index than the SPI index in the 25 −
200 keVrange, but found a similar cutoff energy for a cut-off power law model. Also, the comparison of thespectra >
60 keV between OSSE,
BeppoSAX , andSPI found all in agreement with the data well fitby a power law of Γ ∼ .
1. These results indicatethat the spectral shape remained roughly constantaround 100 keV while varying at lower energiesover the span of nearly 20 years.A comparison of the GS 1826 −
238 spectrum tothe spectra of other NS with hard tails found GS1826 −
238 to have a significantly hotter electrontemperature, resulting in the hard tail not becom-ing the significant component until ∼
150 keV.Similarities to 4U 1728 −
34 were found as bothsources show a high energy component extendingout above at least ∼
100 keV. Both sources couldbe well fit using the hybrid thermal/non-thermalComptonization model,
CompPS , which indicatesthe presence of a non-thermal process at work athigh energies.The presence of a non-thermal hard tail is ofteninterpreted as being related to a jet (Markoff et al.2005; Migliari et al. 2010; Tarana et al. 2011). Cor-relations between radio and X-ray fluxes have beenreported for 4U 1728 −
34 (Migliari et al. 2003)and GX 17 + 2 (Migliari et al. 2007), which sup-ports that interpretation. As of yet no detectionof GS 1826 −
238 has been reported, but Fender& Hendry (2000) estimate the radio emission inthe hard state to be at least 5 −
10 fainter thanthe soft state. With the exception of the possibleshort-lived soft state around MJD 53000, the onlyreported soft state observations of GS 1826 −
238 have been by MAXI in 2014 June (Nakahira et al.2014). (See Section 3.1.) Thus hard tails mightbe common in NS hard states, but detecting themrequires long exposure times with the current in-strument sensitivities.With detections of high-energy tails in NS hardstates, their hard X-ray/soft gamma-ray spectralshapes in the hard state show another similar-ity to BH spectra, as hard tails have been re-ported for a number of sources (e.g. Cyg X-1,GRS 1915+105, GX 339-4). These commonali-ties suggest that the same physical mechanismsare at work in both types of sources. An inter-esting difference, though, is at radio wavelengthswhere emission is high compared to the soft statefor BH while for NS the emission is low comparedto the soft state.
Acknowledgments
The
INTEGRAL
SPI project has been com-pleted under the responsibility and leadership ofCNES. We are grateful to ASI, CEA, CNES, DLR,ESA, INTA, NASA and OSTC for support.
REFERENCES
Asai, K., Dotani, T., Mitsuda, K. et al. 1994,PASJ, 46, 479Barret, D., Motch, C., & Pietsch, W. 1995, A&A,303, 526Barret, D., Olive, J. F., Boirin, L. et al. 2000, ApJ,533, 329Bouchet, L., Del Santo, M., Jourdain, E. et al.2009, ApJ, 693, 1871Bouchet, L., Jourdain, E., Roques, J. P. et al.2008, ApJ, 679, 1315Burke, M. J., Jourdain, E., Roques, J. P., &Evans, D. A. 2014, ApJ, 787, 50Cocchi, M., Farinelli, R., & Paizis, A 2011, A&A,529, 155Cocchi, M., Farinelli, R., Paizis, A., & Titarchuk,L. 2010, A&A, 509, A2Church, M. J., Gibiec, A., & Baluci´nska-Church,M. 2014, MNRAS, 438, 278410’ Ai, A., ´Zycki, Di Salvo, T. et al. 2007, ApJ,667, 411Di Salvo, T., Stella, L., Robba, N. R. et al. 2000,ApJ, 544, 119Di Salvo, T., Iaria, R., Robba, N., Burderi, L.2006, ChJAS, 6, 183Fender, R. P. & Hendry, M. A. 2000, MNRAS,317, 1Fiocchi, M., Bazzano, A., Ubertini, P., & Jean, P.2006, ApJ, 651, 416Homer, L., Charles, P. A., & O’Donoghue, D.1998, MNRAS, 298, 497in’t Zand, J. J. M.. Heise, J., Kuulkers, E. et al.1999, A&A, 347, 891Jensen, P. L., Clausen, K., Cassi, C. et al. 2003,A&A, 411, L7Ji, L., Zhang, S., Chen, Y. et al. 2014, ApJ, 782,40Jourdain, E. & Roques, J. P. 1994, ApJ, 426, L11Jourdain, E. & Roques, J. P. 2009, ApJ, 704, 17Kong, A. K. H., Homer, L., Kuulkers, E. et al.2000, MNRAS, 311, 405Magdziarz, P. & Zdziarski, A. A. 1995, MNRAS,273, 837Makino, F. & Ginga Team 1988, IAU Circ. 4653Markoff, S., Nowak, M. A., & Wilms, J. 2005, ApJ,635, 1203Migliari, S., Fender, R. P., Rupen, M. et al. 2003,MNRAS, 342, 67Migliari, S., Miller-Jones, J. C. A., Fender, R. P.et al. 2007, ApJ, 671, 706Migliari, S. Tomsick, J. A., Miller-Jones, J. C. A.et al. 2010, ApJ, 710, 117Motch, C., Barret, D., Pietsch, W. et al. 1994,IAU Circ. 6101Muno, M. P., Remillard, R. A., & Chakrabarty,D. 2002, ApJ, 568, 35 Nakahira, S., Mihara, T., Sugizaki, M. et al. 2014,ATel 6250Pottschmidt, K., Chernyakova, M., Lubi´nski, P. etal. 2008, in Proceedings of the 7th INTEGRALWorkshop,An INTEGRAL View of CompactObjects. p. 98Poutanen, J. & Svensson, R. 1996, ApJ, 470, 249Revnivtsev, M. G., Tsygankov, S. S, Churazov, E.M, & Krivonos 2014, MNRAS, 445, 1205Roques, J. P., Schanne, S., von Kienlin, A. et al.2003, A&A, 411, L91Strickman, M., Skibo, J., Purcell, W. et al. 1996,A&AS, 120, 217Tanaka, Y. 1989, ESASP, 296, 3Tarana, A., Bazzano, A., Ubertini, P., &Zdziarski, A. A. 2007, ApJ, 654, 494Tarana, A., Belloni, T., Bazzano, A., M´endez, M.,& Ubertini, P. 2011, ApJ, 416, 873Titarchuk, L. 1994, ApJ, 434, 570Thompson, T. W. J., Rothschild, R. E, Tomsick,J. A., & Marshall, H. L. 2005, ApJ, 634, 1261Ubertini, P., Bazzano, A., Cocchi, M. & Natalucci,L. 1997, IAU Circ. 6611Ubertini, P., Bazzano, A., Cocchi, M. et al. 1999,ApJ, 514, 27