Timing and spectral study of the Be XRB IGR J11305-6256: Swift discovers the orbital period and a soft X-ray excess
V. La Parola, A. D'Aì, G. Cusumano, A. Segreto, N. Masetti, A. Melandri
aa r X i v : . [ a s t r o - ph . H E ] M a y Mon. Not. R. Astron. Soc. , 1–5 (2013) Printed 30 October 2018 (MN L A TEX style file v2.2)
Timing and spectral study of the Be XRB IGR J11305–6256: Swiftdiscovers the orbital period and a soft X-ray excess.
V. La Parola , A. D’A`ı , G. Cusumano , A. Segreto , N. Masetti , A. Melandri INAF - Istituto di Astrofisica Spaziale e Fisica Cosmica, Via U. La Malfa 153, I-90146 Palermo, Italy Dipartimento di Fisica e Chimica, Universit`a di Palermo, via Archirafi 36, 90123, Palermo, Italy INAF - Istituto di Astrofisica Spaziale e Fisica Cosmica di Bologna, via Gobetti 101, 40129, Bologna, Italy INAF - Brera Astronomical Observatory, via Bianchi 46, 23807, Merate (LC), Italy
ABSTRACT
IGR J11305-6256 is one of the numerous sources discovered through the INTEGRAL scan ofthe Galactic Plane. Thanks to the Swift-BAT survey, that allows the frequent sampling of anysky region, we have discovered in the hard X-ray emission of this source a modulation witha period of . d. The significance of this periodic modulation is ∼ ∼ R ⊙ corre-sponding to ∼ times the radius of the companion star. The broadband XRT-BAT (0.3 − Key words:
X-rays: binaries – X-rays: individual: IGR J11305-6256.Facility:
Swift
During the last decade, the IBIS/ISGRI telescope (Ubertini et al.2003) on board the INTEGRAL gamma-ray satellite (Winkler et al.2003) has discovered several new Galactic sources, thanks to thehard X-ray window of the telescope (20 −
80 keV) that allows tounveil the presence of highly absorbed objects that appear toofaint in the soft X-ray energies, and thanks to the wide field ofview ( ∼ o ) and to the scan monitoring program of the Galac-tic plan that favored the capture of transient episodes from pre-viously unknown objects. Moreover, since November 2004 theSwift satellite (Gehrels et al. 2004), with its Burst Alert Telescope(BAT, Barthelmy et al. 2005) has been performing a continuousmonitoring of the hard X-ray sky with different observational fea-tures with respect to ISGRI: the observation of large sky areasthanks to a field of view of 1.4 steradian (half coded) and sev-eral switches of the satellite pointing direction within a day. Thesecharacteristics allow to observe up to ∼ of the entire skyevery day, with a duty cycle of 10 to 20% for each direction.This resulted of paramount importance to monitor the variabil-ity of the sources and to achieve good statistics for broad bandspectral analysis. In particular for the high mass X-ray binaries(HXMBs) class, it has led to a sensible increase of orbital perioddetections (e.g. Corbet & Krimm 2009; Corbet et al. 2010a,b,c,d,e;Corbet & Krimm 2010; Cusumano et al. 2010; La Parola et al. 2010; D’A`ı et al. 2011a), and to discover absorption features intheir hard X-ray spectra (D’A`ı et al. 2011b).In this Letter we analyze the soft and hard X-ray data collectedby Swift on IGR J11305-6256.IGR J11305-6256 is a transient source discovered by theIBIS/ISGRI telescope in the Carina region, with a flux of 8 mCrabin the 20–60 keV band (Produit et al. 2004). A Swift-XRT ob-servation allowed the accurate localization of IGR J11305-6256(RA J = 11h 31m 06.5s, Dec J = − ◦ ′ ′′ . , error ra-dius: 6 ′′ ) and confirmed the association with the blue giant star HD100199 (of spectral type B0 IIIe, Garrison et al. 1977), located ata distance of ∼ J = 11h 31m 06.95s, Dec J = − ◦ ′ ′′ . , with position un-certainty ≃ . ′′ , Tomsick et al. 2008). The Chandra data showeda weakly absorbed ( N H = 3 . +2 . − . × cm − ) flat power law( Γ = 0 . +0 . − . ) spectrum, with a 0.3-10 keV unabsorbed flux of +20 − × − erg cm − s − , corresponding to a luminosity of . +2 . − . × erg s − , at a distance of 3 kpc. This Letter is or-ganised as follows. Section 2 describes the Swift data reduction.Section 3 reports on the timing analysis. Section 4 describes thebroad band spectral analysis. In Section 5 we briefly discuss ourresults. c (cid:13) V. La Parola et al.
1E 1145.1-61412E 2448
IGR J11305-6256 30’
Figure 1.
Left panel: 0.2 −
10 keV XRT image with superimposed the position of the optical counterpart, marked with a cross, and the BAT error circle of0.86’ (magenta circle). Right panel: 15 −
150 keV BAT significance map in the sky region around IGR J11305-6256.Obs Obs ID T start T elapsed T exp P h
Orb rate(MJD) (s) (s) c/s1 35098001 53627.01 81703 6007 0.46 0.332 35224002 53636.33 12118 1048 0.54 0.383 35224003 53690.03 82708 15899 0.98 0.434 35098002 53725.67 23856 3658 0.28 0.145 35098003 53728.00 19014 5127 0.30 0.13
Table 1.
Log of Swift-XRT observations. T elapsed is the observationlenght; T exp is the net exposure time; P h
Orb is the orbital phase referredto the profile in Figure 2b.
Swift-XRT (Burrows et al. 2005) observed IGR J11305-6256 fivetimes between November and December 2005. The source was al-ways observed in photon counting mode (Hill et al. 2004). Table1 reports the log of these observations with the most relevant de-tails. XRT data were calibrated, filtered and screened using thestandard procedures included in the XRTDAS package HEAsoft6.0.4. For each observation we extracted source events and spectrafrom a circular region of 20 pixel radius (1 pixel = 2.36”) centeredon the source position as determined with
XRTCENTROID , adopt-ing standard grade filtering 0 −
12. Fig.1 (left) shows the 0.2 − BARYCORR . Thebackground spectra were built using the same grade selection andselecting events in an annular region centered on the source withradii of 40 and 70 pixels to avoid contamination from the sourcepoint spread function wings. The source spectra of each observa-tion were summed to obtain a single spectrum, and the same wasdone for the background spectra. The ancillary files were combinedusing
ADDARF weighting them by the exposure times of the rele-vant spectra. Finally, spectra were re-binned with a minimum of20 counts per energy channel to allow the use of the χ statistics.The spectra were analyzed using the spectral redistribution matri-ces and the ancillary response file v.011 (suitable for data collectedwith CCD substrate voltage V ss = 0V ).The raw BAT survey data of the first 88 months of the Swift mission were retrieved from the HEASARC public archive andprocessed with the BATIMAGER software (Segreto et al. 2010), thatperforms screening and mosaicking of the survey data and producesbackground subtracted spectra and light curves for each detectedsource. Fig.1 (right) shows the 15 −
150 keV significance sky map(exposure time of 17.7 Ms) centered on IGR J11305-6256, wherethe source is detected with a maximum significance of 26.7 stan-dard deviations. The light curve for timing analysis was extractedin the same energy range. The time tag of each bin was corrected tothe solar system barycentre (SSB) by using the task
EARTH SUN .The official BAT spectral redistribution matrix was used for spec-tral analysis. Quoted errors are given at 90% confidence level for asingle parameter, unless otherwise stated. In order to search for long periodicities in the emission ofIGR J11305-6256, we exploited the 15 −
150 keV BAT light curveusing the folding technique (Buccheri & Sacco 1985): we searchedin the 0.5 −
500 days period range, with a period resolution ofP / ( N ∆ T BAT ) , where P is the trial period, N = 16 is the numberof phase bins used to build the profile and ∆ T BAT ∼ Ms is thedata time span. The average count rate in each profile bin has beenevaluated weighting the rates by the inverse square of their statisti-cal error, which is appropriate when dealing with a large spread inthe error values. This happens in the BAT data because the sourceis typically observed at several off-axis angles, thus introducing alarge spread in the source signal-to-noise ratio (SNR), uncorrelatedwith the source count rate.Fig. 2 (a) shows the periodogram where a strong feature atP = 120 . ± . d (the error is the period resolution at P )emerges with χ ∼ . . Other relevant peaks are visible forperiods corresponding to multiples of P . The pulsed profile evalu-ated at P with T epoch = 54658 . MJD is shown in Figure 2 (b).The periodogram is characterised by an increasing strong red noisefor higher period values, caused by the long term variability ofthe source. For this reason the significance of the periodicity at P http://heasarc.gsfc.nasa.gov/docs/archive.html http://heasarc.gsfc.nasa.gov/docs/heasarc/caldb/data/swift/bat/index.htmlc (cid:13) , 1–5 χ Period (days) (a) P × − − . × − C oun t s s − p i x e l − Phase (b) −50 0 50 F r equen cy o f z v a l ue s z(c) Figure 2. a : Periodogram of
Swift -BAT (15 −
150 keV) data forIGR J11305-6256. b : Light curve folded at a period P = 120 . day, with16 phase bins. The vertical arrows mark the phases corresponding to theXRT observations. c : Distribution of z = χ − F χ values extracted inthe period range between 20 and 200 days, excluding the z values around P . The continuous line is the best fit obtained with an exponential modelapplied to the tail of the distribution above z > . cannot be evaluated relying on the χ statistics, and an alternativemethod shall be used. We proceeded according to the followingsteps. • We modeled the ascending trend of the χ distribution fittingit with a 2nd order polynomial and we created a new distribution(z) subtracting the best fit F χ from the χ distribution. The z valueat P is ∼ . . • We built the histogram of the z distribution (Figure 2, c) ex- × − − . × − R a t e ( c oun t/ s / p i x e l s ) Time (MJD)
Figure 3. P = 120 . day. The vertical arrows mark the epoch of the XRTobservations. The first and the third bars correspond to XRT observations1-2 and 4-5 respectively. tracting the z values only from the period range 60–200 days andexcluding the interval around P . • We fitted the histogram values for z > with an exponentialfunction. The resulting best fit model is shown in Fig. 2 (c). • We evaluated the area under the histogram: we summed thearea of each single bin from its left boundary up to z = 20 ; beyond z = 20 we integrated the best fit exponential model up to infinity. • We evaluated the integral of the best-fit exponential functionbeyond 173.4 and normalised it to the total area of the histogram.The result ( . × − ) is the probability of chance occurrenceof a z value equal to or larger than 405 (or a χ equal to or largerthan 261.8) and it corresponds to a significance for the detectedfeature of ∼ . standard deviations in Gaussian statistics.Fig. 3 shows the BAT 15 −
150 keV light curve sampled at timeintervals equal to P . The source shows a strong long term variabil-ity. Significant X-ray emission was observed up to mid 2007 ( ∼ MJD 54400), then the source weakened to an intensity level aver-aged over P roughly consistent with the background.IGR J11305-6256 has been always detected in the XRTpointed observations. Table 1 reports the source count rate andthe orbital phase of each observation evaluated with respect toP and T epoch . The source intensity varies between 0.1 and 0.4counts/s and these variations result uncorrelated to the P modu-lation shown in Fig. 2 (b). The timing analysis on the XRT datahas to cope with two main issues: the read-out time of the XRT-PCdata ( δT XRT =2.5073 s) and the fragmentation of the observationinto several snapshots with different duration and time separation.These effects introduce systematic features in the results of the tim-ing analysis that can mask the significance of any real source mod-ulation. Therefore we performed a folding analysis for each XRTobservation on the events arrival times randomized within the XRT-PC time resolution bin. Moreover, we produced the periodogramrelevant to each snapshot with exposure time higher than 500 ssearching in the period range δT XRT − s: the periodogramsobtained from snapshots belonging to the same observation werefinally summed. We did not find any significant feature in the result-ing five periodograms. We repeated this analysis selecting snapshotlasting more than 1000 s and searching for modulations up to 250s. Again, no significant features emerged above the white noise. c (cid:13) , 1–5 V. La Parola et al.
In order to have the best available SNR spectrum, we merged allthe XRT spectra from the observations listed in Table 1 and we ac-cumulated a hard X-ray BAT spectrum over the 88-month of moni-toring. We preliminarily checked that the variability of the spectralparameters was lower than the constraints derived by a fit error de-termination on the single observations of Table1. To this aim, thebackground subtracted spectra of the XRT observations were fit-ted simultaneously with a model consisting of an absorbed power-law. Forcing the spectral parameters to common values for the fivedatasets and allowing only a multiplicative constant to weight fordifferent fluxes, we obtained similar residuals.Several BAT spectra were produced selecting the data in dif-ferent time intervals (MJD intervals 53470 − − − − − − . ± . .The broad band average spectrum of IGR J11305-6256 (XRT:0.3 −
10 keV; BAT: 15 −
150 keV) is plotted in Fig. 4 (a). We in-troduced a multiplicative factor in the model to account for thenon-simultaneity of the BAT and XRT spectra and for any inter-calibration uncertainty. The factor is kept fixed to 1 for the XRTdata and left free to vary for the BAT data. First we tried an ab-sorbed power-law with a high-energy cut-off (model 1). The re-duced χ is 504.4 (with 389 dof) with residuals below 1 keVstrongly suggesting the presence of other continuum components.Therefore, we added to the power-law emission a black-body com-ponent at a temperature of ∼ χ = 371.5 for387 dof). In order to evaluate the statistical significance of this im-provement we built 100000 simulated XRT spectra using the bestfit parameters of model 1 and the rate uncertainties of the observedXRT data. The simulated spectra were fitted together with the BATspectrum both with model 1 and model 2 obtaining a value of ∼ as the highest difference in the χ values for the two models. Thisresults corresponds to a chance probability to obtain the decrementin χ as measured in the observed XRT+BAT source data lowerthan . × − , that corresponds to a significance larger than ∼ . standard deviations in Gaussian statistics. In the 0.5 − × − and 4.0 × − erg cm − s − , respectively (assumed with respect to the XRT data). Alter-natively, we also explored a scenario where the continuum X-rayemission (modeled with only a cut-off power-law) is partially ab-sorbed by local neutral matter ( pcfabs in Xspec; model 3). Thismodel gave a statistically similar result (reduced χ = 0.97 for 386dof), compared to the blackbody + cutoff powerlaw . Wepresent in Table 2, the best-fitting parameters for the spectral mod-els that have been considered, while data and residuals in units ofsigmas are shown in Fig.4. We have presented in this work the spectral and timing results fromthe complete set of Swift observations of IGR J11305-6256. Thesource is an accreting Be XRB, whose X-ray activity has turnedto quiescence in recent years. Analyzing the long-term BAT light
Figure 4.
IGR J11305-6256 broad band spectrum.
Panel (a) : XRT and BATdata, best-fitting phabs*cutoffpl) model (red line).
Panel (b)
Resid-uals in units of σ for phabs*cutoffpl) model. Panel (c) : Residuals inunit of σ for the phabs*(bbody+cutoffpl) model. Panel (d) : Resid-uals in unit of σ for the phabs*pcfabs*(cutoffpl) model. curve, we found evidence at ∼ σ for the orbital period of thesystem to be 120.8 days. The third Kepler’s law allows us to derivethe semi-major axis of this binary system: a = ( GP ( M ⋆ + M X ) / π ) / ≃ R ⊙ = 19 R ⋆ (1)where M X = 1 . M ⊙ is the mass of the neutron star, M ⋆ ≃ M ⊙ is the mass expected for the spectral type of the companion starand R ⋆ ≃ R ⊙ is its expected radius (Lang 1992). Such a largeorbital separation is indeed common in Be XRBs. Considering thewhole class of Be XRB, the long orbital period would suggest alsovery long spin periods and high eccentricity (Reig 2011). However,the folded light curve shows a smooth modulation with the orbitalphase, and suggests that accretion may not be clocked with a peri-astron passage, as typical for classical highly-eccentric Be XRBs.The broadband data collected by pointed Swift observationsand by the long-term BAT monitoring indicated that the spec-trum cannot be described with an absorbed power-law with a high-energy cut-off as in the case of most low-luminosity accreting high-mass/Be X-ray binaries. The fit is sensibly improved when a ther-mal black-body component is added to the model, or when the X-ray emission is only partially absorbed. In the first scenario, wemay be observing thermal emission from the accreting magneticcaps of the neutron star, the derived black-body radius being com-patible with a small portion of the surface of the NS. The black-body temperature (1.6 keV) is also within the expected range forother similar Be XRB sources for which this component was clearlydetected (La Palombara et al. 2013), but it is sensibly lower thanthe typical temperatures for soft excesses found in other HMXBs, c (cid:13) , 1–5 [htp] Table 2.
Best-fitting spectral parameters. † Covering fraction for NH 2. ‡ Constant factor to be multiplied to the model in the BAT energy range in order tomatch the BAT count rate. Parameter cutoff bb+cutoff pcfabs*cutoff (model 1) (model 2) (model 3)NH 1 ( × cm − ) 0.32 ± ± ± × cm − ) 2.6 ± F † n H . +0 . − . Γ ± ± ± E cutoff (keV) 12.0 ± +10 − ± N powerlaw ( × − ph keV − cm − s − at 1 keV) 1.5 ± ± (4 . +1 . − . ) kT BB (keV) 1.60 ± ± C ‡ BAT ± +1 . − . . +0 . − . χ / dof 503 / 388 369 / 386 374 / 386 where the emission may be related to the reprocessed emission atthe magnetospheric boundary or by diffuse gas around the X-raysystem (Hickox et al. 2004). However, all the Be XRBs that haveshown presence of this component are also X-ray pulsars with ahigh modulated fraction. Lack of evidence in the Swift XRT dataof any pulsation below 250 s may point to very long periods thatcan be detected only accumulating observations with very long ex-posures.Another scenario, that is statistically equivalent, explains thedeviations in the soft part of the spectrum as due to the presenceof a neutral absorber that partially covers the X-ray emission. Thecovered fraction is very high ( f > ) and the relative columndensity (2.6 × cm − ) is sensibly higher than the whole columndensity due to the ISM (0.26 × cm − ). This scenario requiresthat the X-ray source may be embedded (but not completely) in adense stellar wind of the companion star, or that part of the outeratmosphere of the companion is stripped, rapidly cooled and dis-persed along the NS orbit. In many HMXBs this scenario has beenoften invoked (see e.g. Naik et al. 2011), but rarely for Be XRBs,due to the fact that Be stars are less efficient in driving strongwinds. Another important finding that would corroborate this sce-nario, would be the detection of a strong fluorescent iron line, thatshould be imprinted in the circum-stellar matter. To this aim, a highsignal-to-noise at the 6 − ACKNOWLEDGMENTS
This work has been supported by ASI grant I/011/07/0.
REFERENCES
Barthelmy, S. D., et al. 2005, Space Science Reviews, 120, 143Buccheri, R., & Sacco, B. 1985, Data Analysis in Astronomy, 15Burrows D. N., et al., 2005, SSRv, 120, 165Corbet, R. H. D., & Krimm, H. A. 2009, The Astronomer’s Tele-gram, 2008, 1Corbet, R. H. D., Krimm, H. A., & Skinner, G. K. 2010, The As-tronomer’s Telegram, 2559, 1Corbet, R. H. D., Krimm, H. A., Barthelmy, S. D., et al. 2010, TheAstronomer’s Telegram, 2570, 1Corbet, R. H. D., Barthelmy, S. D., Baumgartner, W. H., et al.2010, The Astronomer’s Telegram, 2588, 1 Corbet, R. H. D., Barthelmy, S. D., Baumgartner, W. H., et al.2010, The Astronomer’s Telegram, 2598, 1Corbet, R. H. D., Barthelmy, S. D., Baumgartner, W. H., et al.2010, The Astronomer’s Telegram, 2599, 1Corbet, R. H. D., & Krimm, H. A. 2010, The Astronomer’s Tele-gram, 3079, 1Cusumano, G., La Parola, V., Romano, P., et al. 2010, MNRAS,406, L16D’A`ı, A., La Parola, V., Cusumano, G., et al. 2011, A&A, 529,A30D’A`ı, A., Cusumano, G., La Parola, V., et al. 2011, A&A, 532,A73Garrison, R. F., Hiltner, W. A., & Schild, R. E. 1977, ApJS, 35,111Gehrels, N., et al. 2004, ApJ, 611, 1005Hickox, R. C., Narayan, R., & Kallman, T. R. 2004, ApJ, 614, 881Hill, J. E., Burrows, D. N., Nousek, J. A., et al. 2004, SPIE, 5165,217Lang, K. R. 1992, Astrophysical Data I. Planets and Stars, X, 937pp. 33 figs.. Springer-Verlag Berlin Heidelberg New York,La Palombara, N., Mereghetti, S., Sidoli, L., Tiengo, A., & Espos-ito, P. 2013, arXiv:1301.5120La Parola, V., Cusumano, G., Romano, P., et al. 2010, MNRAS,405, L66Masetti, N., Pretorius, M. L., Palazzi, E., et al. 2006, A&A, 449,1139Naik, S., Paul, B., Kachhara, C., & Vadawale, S. V. 2011, MN-RAS, 413, 241Produit, N., Ballet, J., & Mowlavi, N. 2004, The Astronomer’sTelegram, 278, 1Reig, P. 2011, Ap&SS, 332, 1Segreto, A., Cusumano, G., Ferrigno, C., La Parola, V., Mangano,V., Mineo, T., & Romano, P. 2010, A&A, 510, A47Tomsick, J. A., Chaty, S., Rodriguez, J., Walter, R., & Kaaret, P.2008, ApJ, 685, 1143Ubertini, P., Lebrun, F., Di Cocco, G., et al. 2003, A&A, 411,L131Winkler, C., Courvoisier, T. J.-L., Di Cocco, G., et al. 2003, A&A,411, L1This paper has been typeset from a TEX/ L A TEX file prepared by theauthor. c (cid:13)000