The emission of Cygnus X-1: observations with INTEGRAL SPI from 20 keV to 2 MeV
aa r X i v : . [ a s t r o - ph . H E ] S e p Received 2011 August 2; Accepted 2011 September 8
Preprint typeset using L A TEX style emulateapj v. 8/13/10
THE EMISSION OF CYGNUS X-1: OBSERVATIONS WITH
INTEGRAL
SPI FROM 20 KEV TO 2 MEV
E. Jourdain, J. P. Roques and J. Malzac
Universit´e de Toulouse; UPS-OMP; IRAP; Toulouse, FranceCNRS; IRAP; 9 Av. colonel Roche, BP 44346, F-31028 Toulouse cedex 4, France
Received ; accepted
Received 2011 August 2; Accepted 2011 September 8
ABSTRACTWe report on Cyg X-1 observations performed by the SPI telescope onboard the INTEGRAL missionand distributed over more than 6 years. We investigate the variability of the intensity and spectralshape of this peculiar source in the hard X-rays domain, and more particularly up to the MeV region.We first study the total averaged spectrum which presents the best signal to noise ratio (4 Ms ofdata). Then, we refine our results by building mean spectra by periods and gathering those of similarhardness.Several spectral shapes are observed with important changes in the curvature between 20 and 200keV, even at the same luminosity level. In all cases, the emission decreases sharply above 700 keV,with flux values above 1 MeV (or upper limits) well below the recently reported polarised flux (Laurentet al. 2011), while compatible with the MeV emission detected some years ago by CGRO/COMPTEL(McConnell et al., 2002).Finally, we take advantage of the spectroscopic capability of the instrument to seek for spectralfeatures in the 500 keV region with negative results for any significant annihilation emission on 2 ksand days timescales, as well as in the total dataset.
Subject headings: radiation mechanisms: general— Gamma-rays:individual (Cyg X-1) — gamma rays:observations INTRODUCTION
Cyg X-1 is an unavoidable target for any high energyinstrument. Being one of the most luminous sources (upto the MeV range), it represents an ideal lab to study themechanism at work in the direct environment of a blackhole. There, the accretion flow is thought to form an op-tically thick disk and/or an optically thin corona, whilethe observed radio jets could originate from the samearea (see e.g.the review by Done, Gierli´nski & Kubota2007). The high energy radiation provides insights onthe physical processes at work in this region. Thanksto its persistent high flux ( ∼ > ∼ INTE-GRAL mission contains instrument exploring the sameenergy domain. Recently, Laurent et al. (2011), stackedall the
INTEGRAL/IBIS data available for Cygnus X-1and detected the presence of a non-thermal power lawcomponent between 400 keV and 2 MeV. Interestinglythey found that contrary to the thermal Comptoniza-tion component present below 400 keV, the non-thermalpower-law emission appears to be strongly polarized.This non-thermal component appears to have a flux thatis stronger than that measured by CGRO by a factor 5-10 and a much harder spectral slope Γ ≃ .
6. Here, weuse another
INTEGRAL instrument operating in this en-ergy range, the spectrometer SPI, to investigate the highenergy spectral shape of Cyg X-1 and test the presenceof non-thermal high energy excess.Our first goal is to take advantage of the sensitivityachieved with the SPI detector and the large amountof data and perform a detailed analysis of the energy ex-tent of the hard X-ray emission together with its spectralvariability.Moreover, the spectroscopic capability of the Germaniumcrystals allow us to seek for the presence of spectral fea-tures linked to the annihilation process.Hereafter, we present briefly the instrument, data setand the method followed for the analysis. Then, we re-port on our results and start by examining the total Jourdain et al.mean spectrum to determine the emission above a fewhundreds of keV, where scarcity of photons imposes ex-posure as long as possible. In a second step, we analysethe source behavior during individual revolutions, andbuild several averaged spectra, following some hardnesscriteria which can be considered as characteristic of thespectral state of the source. We conclude with a com-parison with previous results. INSTRUMENT, OBSERVATIONS AND DATA ANALYSIS
SPI is a spectrometer aboard the INTEGRAL obser-vatory operating between 20 kev and 8 MeV. The de-scription of the instrument and its performance can befound in Vedrenne et al. (2003) and Roques et al. (2003).The main features of interest for our study are a largeFoV (30 ◦ ) with an angular resolution of 2 . ◦ (FWHM)based on a coded aperture mask. The Germanium cam-era, beyond an excellent spectroscopic capability, offersa good sensitivity over more than 2 decades in energywith a unique instrument. It is surrounded by a 5-cmthick BGO shield (ACS, Anti-Coincidence Shield) whichmeasures the particle flux. This latter can be used as agood tracers of the background level.During a 3-day orbit, the usual dithering strategy(Jensen et al., 2003) consists of a hundred of 30-40min exposures (also called scw for ’science window’),with a given region scanned by 2 ◦ steps following pre-determined patterns. The recommended pattern for SPIobservations is a grid of 5X5 around the chosen target.Unfortunately, in order to content more proposers, ex-cept a few exceptions, most of the Cyg X-1 data hasbeen obtained through ’amalgamated’ observations, iewith the pattern center somewhere between Cyg X-1 andCyg A region. As a consequence, Cyg X-1 appears onlyin one side of the FoV, reducing the mean efficient area(the source is partially coded), with some interruptionsin the observation sequences, when the source goes outfrom the field of view. For our analysis, we have selectedin the whole INTEGRAL observation plan, those revo-lutions in which Cyg X-1 is included in the ± ◦ FoVduring more than 20 scw (50 ks). This gives a total of 42revolutions. These observations encompass 4 Ms of effec-tive time, from 2003, June to 2009 December. They weregrouped together according to temporal proximity into13 periods. Exposures with high background level (en-try/exit of radiation belts, solar activity) or large sourceoff-angle (source/pointing axis angle beyond 13 ◦ ) havebeen removed from our dataset. Table 1 gives some de-tails about the observation pattern, beginning, end anduseful duration, for each of the defined periods.We follow the analysis method described in Jourdain& Roques (2009), based on a sky model fitting, througha χ minimization. This methods makes use of the PulseShape Discrimination (PSD) system, a second electronicchain operating in parallel, to eliminate spurious eventsoccurring in the MeV region. In the 650 keV-2.2 MeVenergy range, we use only the PSD tagged events, whilebelow 650 keV, all the single events are analysed. Thisprocedure has been validated with the Crab Nebula ob-servations.In this work, we consider a common sky model for all theobservations, which consists in 4 sources, namely: CygX-1, Cyg X-3, EXO2023+375 and GRS1915+105. Tokeep the number of degrees of freedom to a minimum, we consider that the three latter have a constant fluxover the revolution timescale. For Cyg X-1, we considera variability timescale of 1 scw ( ∼ ∼ xspec
11 (Arnaud 1996). Due to some uncertainties on theenergy response in the lowest channels and threshold ef-fects, we exclude the first channels (
E < STUDY OF THE SPECTRAL SHAPE
The total mean spectrum
The total mean spectrum gathers the whole set of avail-able clean data (4 Ms of observation distributed overmore than 6 years). It has been built by extracting onespectrum per revolution then summing them. Due tothe impressive signal to noise ratio at low energy, we areunable to ensure that the response matrices are knownwith a sufficient precision. We have thus added 0.5% ofsystematic errors to the data during the fit procedure.The data have been first fitted with a simple analyti-cal cutoff power law and a Comptonization + reflectionmodel (reflect*comptt in xspec ). Residuals clearly showdeviation from these models above 200-300 keV wherean excess of emission appears above the model predic-tion. To go further, we keep a Comptonization law (+its reflected component) in the low energy part and fo-cus on the high energy emission. Following Laurent et al.(2011), we first try to model this additional componentby a single power law. The fit converges toward a pho-ton index of ∼ . ± χ value remains unacceptable,with a huge contribution of the last channels. Indeed, thehigh energy part doesnot follow such a power-law : in-stead, the emission presents a rather sharp decrease after700 keV and the source is not detected above 1 MeV. Abetter result is obtained when modeling this high energycomponent by a second Comptonization region with atemperature around 100 keV. As several couples (kT, τ )can reproduce the data, we fixed τ to 0.5 and obtaineda best fit value of kT=123 keV (see Table 2). Fig. 1presents the observed spectrum with one and two Comp-tonization models. Even though the set of parametersproposed in Table 2 is only a possible solution, the twoComptonization model provides a good description of thedata and more specifically of the curvature observed upto 1 MeV.ygnus X-1 emission with INTEGRAL
SPI 3
Flux, hardness and hardness versus flux evolution
The MeV emission may depend on the source spec-tral state or intensity. To follow potential changes in thesource emission, we have analysed in details each revo-lution and used the information to group observationscorresponding to similar states.Fig. 2 displays the temporal evolution of Cyg X-1 in30-70 keV energy range, with the fluxes averaged byrevolution (100-200 ks timescale) and Fig. 3 the evo-lution of the hardness (defined in the 30-120 keV rangeby F − keV /F − keV ). In Fig. 4 and Fig. 5, wepresent the same flux and hardness but detailed on timescales of a few days (with a resolution of one scw ie ∼ χ tests and is strongly rejected for all peri-ods (reduced χ always greater than 5.9 for a numberof degrees of freedom ranging from 100 to 300). Thevariability appears to be chaotic but in both cases overa limited amplitude: the source is always detected andvaries by a factor not greater than 3. We can recognizethe usual temporal behavior of Cyg X-1 (see e.g. Ling etal. 1987 and Zdziarski et al. 2002). Note that the firstperiod, which corresponds the lowest and softest state inour sample, has been analysed by Malzac et al. (2006).They identified the (rather unusal) source behavior asa mini or failed transition between soft and hard state,in a so called ”intermediate state”. However, no robust(that is, lasting more than a few hours) incursion in thesoft state can be reported during our observations andwe conclude that Cyg X-1 was always in a hard (LH) orintermediate states.Fig. 6 displays the hardness as a function of the 30-70 keV source flux (revolution averaged values). Thehardness intensity plane can be divided in three regionscorresponding to the main clusters of data points. Thosethree regions are outlined in Fig. 6. The first two regionsgather the points with hardnesses respectively below 0.24and between 0.24 and 0.28, and, incidentally, correspondto the first part of the INTEGRAL mission (revolutions79 to 261 ie June 2003 - November 2004).In the third group, the flux levels span a broader range,extending toward higher values, but the mean hardnessesnever decrease below 0.29. During this period, which cov-ers more than 3 years (mid 2006 up to end of 2009), thesource evolution consists of an intensity variation withina factor of ∼ ∼
45 keV (in the middle of the studied en-ergy band, by construction). It could be attributed toan increase of both temperature and optical depth of theComptonising medium. When combined, the two vari-ability modes give rise to a global complex behavior inthe hard X-ray domain, although the source remains inthe HS (and intermediate states).To study in more details the spectral shape and itsevolution in the high energy part, we have to accumulatedata corresponding to the same hardness.
Comparison of different averaged spectra
Based on the three hardness levels repered in Fig. 6, wehave built the corresponding averaged spectra, hereafterrefered to as ’low hardness’, ’mid hardness’ and ’highhardness’ samples, respectively. They are displayed onFig. 8, while best fit parameters are given in Table 3.Similarly to the total spectrum, the data are modeledby a Comptonization law plus its reflection component(required when looking at residuals) and a second hot-ter Comptonization. The evolution of the slope in thelow energy part (in direct relation with the hardness)is clear, with an increase of the peak energy from ∼ ∼
150 keV. This behavior is not related to thereflection component but appears as an evolution of themacroscopic parameters (kT and τ ) of the Comptonizingmedium, which both increase from the soft to the hardsample (see Table 3).Looking now to the high energy part, no significant emis-sion is detected above 700 keV. A degeneracy between pa-rameters being unavoidable, we choose to fix the secondComptonizing medium temperature and optical depth to130 keV and 0.6 respectively for all spectra, but free nor-malisation factors.Even though it represents only a possible description ofour data, this model, involving a second (hotter andthiner) Comptonization medium with constant param-eters kT e and τ , provides an acceptable interpretation ofthe Cyg X-1 behavior in the hard X-ray domain. Even ifother models could reproduce the data, such two Comp-tonizing regions (or more generally, variable kT e and/or τ ) scenario resembles those already applied in previousworks to Cyg X-1 and several BHBs (see for examplethe Suzaku observations analysed by Makishima et al,2008, and references therein). Beyond the specific sets ofbest fit parameter formally obtained, the inadequacy ofthe single comptonization region suggests spatial and/orrapid temporal variation of the temperature or opticaldepth of the Comptonization region. ANNIHILATION FEATURE
Cygnus X-1 is one of the brightest Galactic sources ofhard X-rays up to several hundreds of keV and is there-fore a good candidate for positron production. The ex-cellent energy resolution of the SPI germanium detectormakes it the best instrument to seek for potential anni-hilation signatures in the observed emission.We have thus looked for any emission feature, narrow(10 keV FWHM) or broad (80 keV FWHM), transient(scw ie ∼ σ upper limits of 2-3 × − ph cm − s − and 0.5-1 × − ph cm − s − for anarrow and broad line, respectively. On the revolution Jourdain et al.timescale, no significant excess above the continuum con-tribution are found, with upper limits ranging between3-6 × − ph cm − s − and 0.7-1.1 × − ph cm − s − according to the revolution duration. Finally, consid-ering the whole set of data (4 Ms), persistent emissionfeatures, if any, should be below 6 × − ph cm − s − and 1.3 × − ph cm − s − . COMPARISON WITH OTHER INSTRUMENTS
We now compare our observations with the availabledata in the MeV region. From Fig. 9, we can see thatour ”high hardness sample” observations has a spectralshape that is very close to that of the HS mean spec-trum reported by McConnell et al. (2002) from GRO/COMPTEL and OSSE instruments.To compare the whole set of data (INTEGRAL/SPI +CGRO/Comptel+OSSE), we use a model consisting in acutoff power law plus a broken power-law with the firstindex fixed to -1 ( to avoid any contribution in the lowenergy part).Imposing the same photon index for the cutoff power-laws, the fit procedure converges toward slightly differentcutoff energies ( ∼
140 and 160 keV, for SPI and OSSEdata respectively, no contribution in COMPTEL range).We also impose a common second photon index in thebroken power law and obtained 3.4 ± ± ±
25) keV in the SPI data,it allows us to recover the additional emission above thelow energy component, the very soft index limiting itsextension toward high energies.In conclusion, even if the SPI data do not show significantemission above 1 MeV, the upper limits are marginallyconsistent with the non-thermal tail observed at severalMeV by CGRO/COMPTEL. We note however that thisemission is not taken into account by the two-zone ther-mal Comptonization model proposed above which wouldrequire an additional component to fit the COMPTELdata. The overall good agreement between the SPI andOSSE spectra indicate that the average spectral proper-ties in the hard state have remained constant betweenthe two epochs.Recently, Laurent et al. (2011) presented a stackedspectrum of Cygnus X-1 obtained using the IBIS dataduring nearly the same observation period as ours. IBISis composed of two position sensitive detector layers IS-GRI (CdTe, 15-1000 keV; Lebrun et al. 2003), and PIC-sIT (CsI, 200 keV-10 MeV; Labanti et al. 2003). Thesetwo detectors are usually used independently to producespectra and light curves. Laurent et al. (2011) usedthe IBIS/ISGRI data up to ∼
400 keV and IBIS/PICsiTat higher energies. Fig. 10 compares our averaged SPIspectra with the results of Laurent et al. 2011. We notethat most of the published results from IBIS actually useonly the ISGRI detector. With the recent versions of thepublic software, the results from IBIS/ISGRI are gener-ally compatible with those of SPI. This good agreementis confirmed by our comparison of the stacked SPI andIBIS/ISGRI spectra. At higher energies however, the re-sults are clearly different: the SPI fluxes are lower thanthe IBIS/PICsiT points by a factor of about 5 at least.The origin of this disagreement is unclear. We were not able to go further in the analysis of the discrepancy asthere are very few published results from IBIS/PICsiT. SUMMARY AND CONCLUSIONS
We used 4 Ms of
INTEGRAL
SPI data to study theemission of Cyg X-1 from 20 keV to the MeV region.While the source has been essentially detected in theHS, its presents complex variability on all timescales. Wehave studied the luminosity and hardness evolution of thesource above 20 keV, on the scw (2000 s) and revolution(1-2 days) timescales. The revolution averaged data givea nice picture of the long term source behavior. A changeby a factor of 2-3 in luminosity can be accompanied eitherby a significant softening in the 20-150 keV domain (seeFig. 8) or by the same spectral shape just moved up anddown.Then, in order to improve the photon statistics at highenergies we combined observations to produce long ex-posure time average spectra. The analysis of the stackedspectra reveals that the emission of Cyg X-1 extends upto ∼
700 keV but presents a sharp cutoff around that en-ergy. Whatever the criterion we used to built averagedspectra (low/mid/high hardness, low/high intensity, alldata), no emission can be detected above 1 MeV.Nevertheless, a single Comptonization model does notprovide a good description of the overall spectral shape.We have shown that a two temperature model providesa good fit of the SPI data, with a minimum of free pa-rameters ( τ and kT of the second Compton componentcan be considered as constant along the time).In a final step to investigate the high energy emission,we compare all the data available above 300 keV for CygX-1 in the last 2 decades and conclude that while ourSPI upper limits are marginally compatible with the softpowerlaw reported by COMPTEL in the 90’ (McConnellet al, 2002), they clearly disagree (lower by a factor ∼ ACKNOWLEDGMENTS
The
INTEGRAL
SPI project has been completed un-der the responsibility and leadership of CNES. We aregrateful to ASI, CEA, CNES, DLR, ESA, INTA, NASAand OSTC for support. We thank A. A. Zdziarski forproviding us with GRO/COMPTEL and OSSE data.
REFERENCES ygnus X-1 emission with
INTEGRAL
SPI 5
Bassani, L., Dean, A. J., Di Cocco, G., Perotti, F. and Stephen,J. B., 1989, ApJ, 343, 313Done, C., Gierli´nski, M., & Kubota, A., 2007, A&A Rev., 15, 1Gierlinski, M., Zdziarski, A.A., Done C. et al., 1997, MNRAS,288, 958Gierlinski, M., Zdziarski, A.A., Poutanen, J.. et al., 1999,MNRAS, 309, 496Jensen, P.L., Clausen, K., Cassi, C. et al., 2003, A&A, 411, L7Jourdain, E. & Roques, J. P., 2009, ApJ, 704, 17Laurent, P., Rodriguez, J., Wilms, J et al., 2011, Science, 332,438Labanti, C., Di Cocco, G., Ferro, G., et al., 2003, A&A, 411,L149Lebrun, F., Leray, J. P., Lavocat, P., et al., 2003, A&A, 411, L141Liang, E. P. and Nolan, P. L., 1984, SSRv, 38, 353Ling J.C., Mahoney, W. A., Wheaton W. A. & Jacobson A. S.,1987, ApJ, 321, L117 Makishima, K., Takahashi, H., Yamada, S. et al., 2008, PASJ,60, 585Malzac, J., Petrucci, P.O., Jourdain, E. et al., 2006, A&A, 448,1125McConnell, M. L., Zdziarski, A. A., Bennett, K. et al., 2002,ApJ, 572, 984Roques J.P., Schanne S., Von Kienlin A., et al, 2003, A&A, 411,L91Vedrenne, G., Roques, J.P., Schonfelder, V. et al, 2003, A&A,411, L63Zdziarski, A. A., Poutanen, J., Paciesas, W. S. and Wen L.,2002, ApJ, 578, 357
Jourdain et al.
Table 1
Log of the
INTEGRAL
SPI observations of Cyg X-1 used in this paper.revol Start End usefulnumber duration (ks)79-80 (5x5) 2003-06-07 00:59 2003-06-12 03:35 293210-214 (A) 2004-07-03 00:01 2004-07-17 00:25 709251-252 (A) 2004-11-03 14:23 2004-11-07 16:26 176259 & 261 (H) 2004-11-26 12:28 2004-12-03 15:43 143470 (EXO, H) 2006-08-19 09:19 2006-08-21 16:02 159486 (EXO, H) 2006-10-06 00:11 2006-10-08 07:55 160498-505 (GP) 2006-11-11 19:31 2006-12-04 06:20 535628-631 (A) 2007-12-04 19:05 2007-12-15 21:08 388673 (A) 2008-04-18 17:41 2008-04-19 22:09 54682-684 (A) 2008-05-14 08:13 2008-05-22 19:54 304739-746 (A) 2008-11-01 02:14 2008-11-24 05:25 551803-806 (A) 2009-05-11 08:27 2009-05-22 11:32 371875(H*) & 877(H) 2009-12-12 16:18 2009-12-19 20:57 160
Note . — In the first column, the letter after the revolution number indicates the dithering strategy used: (5x5) for the standart 5X5 pattern(see section 2); (A) for a pointing strategy centered between Cyg X-1 and Cyg A region; (H) for the hexagonal pattern and (GP) for a GalacticPlane scan. (H*) During the rev 875, the pointing strategy follows a pattern proposed by Wilms et al in their AO-7 proposal. All this informationis available on the dedicated ESA site web http://integral.esa.int/isocweb.
Table 2
Fit parameters for the total averaged spectra of Cyg X-1 presented in Fig. 2Model Ω kT τ α or kT τ χ red keV keV (DoF)Refl*Comptt 0.90 ± ± ± ± ± ± ± ± ± ±
10 0.5 (fixed) 1.7 (36)
Note . — Parameters obtained for the mean Cyg X-1 spectrum (all publicly available observations corresponding to 4 Ms of useful duration).0.5% of systematic have been added to the data. Two parameters have been fixed in the second model to overcome some degeneracy.
Table 3
Fit parameters for 3 averaged spectra of Cyg X-1 (see Fig. 8)Sample/Model Ω kT τ kT τ χ red keV keVLow hardness sample 0.55 ± ± ± ± ± ± ± ± ± Note . — Best fit parameters for the three mean Cyg X-1 spectra (grouped by same hardness). The model consists in two Comptonizationcomponents, the second one with fixed parameters (except normalisation). 0.5% of systematic have been added to the data. ygnus X-1 emission with
INTEGRAL
SPI 7
Figure 1.
Cyg X-1 averaged spectra for the whole data set (4 Ms of useful duration between 2003 and 2009). Approximations withComptonization + reflection and two Comptonization models are represented by solid lines. 0.5% systematic errors have been added tothe data. Upper limits are at 2 σ level. See Table 2 for models parameters. Jourdain et al.
Figure 2.
Light curves of Cyg X-1 in the 30-70 keV energy range from
INTEGRAL
SPI since the beginning of the mission. Each pointrepresents a (part of) revolution. The straight line represents the mean flux.
Figure 3.
Hardness ratio evolution of Cyg X-1. Each point represents a (part of) revolution. ygnus X-1 emission with
INTEGRAL
SPI 9
Figure 4.
The same as Fig. 2, detailed by periods. Each point represents a science windows (scw, ∼ Figure 5.
The same as Fig. 3, detailed by periods. Each point represents a science windows (scw, ∼ Figure 6.
Hardness versus flux evolution. Each point represents a (part of) revolution. The solid straight line joins points in thechronological order (from the triangle to the diamond symbols). The squares with numbers correspond to revolutions used in the nextfigure. Dashed lines materialize the limits used to build different subsets of data (see text and Figure 8). ygnus X-1 emission with
INTEGRAL
SPI 11
Figure 7.
Individual spectra illustrating the spectral evolution. The 3 middle spectra illustrate the pivoting of the shape at constant flux,while the lowest ( N
1) and the highest ( N
5) correspond to a change in intensity with similar spectral shape.
Figure 8.
Cyg X-1 averaged spectra for the three hardness levels. Solid lines represent a model with 2 Comptonization models. Dottedlines correspond to the first component (see Table 3 for the parameters) and dashed lines to the second one (with kT fixed to 130 keV and τ fixed to 0.6). Upper limits are at a 2 σ level. ygnus X-1 emission with INTEGRAL
SPI 13
Figure 9.
Cyg X-1 composite spectrum from CGRO (OSSE in red, COMPTEL in green, averaged Hard state, from McConnell et al.,2002) and INTEGRAL/SPI (in blue, high hardness sample averaged spectrum). For each instrument, the dotted curve corresponds to acutoff power law component, the dashed curve to a broken power law with first photon index fixed to -1, solid line to the total (See Section5).