A Chandra observation of the millisecond X-ray pulsar IGR J175113057
IGR J175113057 is a low mass X-ray binary hosting a neutron star and is one of the few accreting millisecond X-ray pulsars with X-ray bursts. We report on a 20 ksec Chandra grating observation of IGR J175113057, performed on 2009 September 22.
We determine the most accurate X-ray position of IGR J175113057, =17 51 08.66,
= 30 57 41.0(90% uncertainty of 0.6).
During the observation, a 54 s long type-I X-ray burst is detected. The persistent (non-burst) emission has an absorbed 0.5–8 keV luminosity of 1.710 (at 6.9 kpc) and can be well described by a thermal Comptonization model of
soft, 0.6 keV, seed photons up-scattered by a hot corona. The type-I X-ray burst spectrum, with average luminosity over the 54 sec duration =1.610 , can be well described by a blackbody with 1.6 keV and 5 km. While an evolution in temperature of the blackbody can be appreciated throughout the burst (average peak = keV to tail = keV), the relative emitting surface shows no evolution.
The overall persistent and type-I burst properties observed during the Chandra observation are consistent with what was previously reported during the 2009 outburst of IGR J175113057.
Subject headings:accretion, accretion disks – X-rays: binaries – X-rays: bursts – stars: neutron – pulsars: individual: IGR J175113057
Low Mass X-ray Binaries containing a neutron star (hereafter NS LMXBs) are very old systems (10 yrs), with a NS magnetic field that is believed to have decayed to about 10 G. It is believed that since the NS spends a substantial fraction of its life accreting gas via an accretion disk, it is finally spun-up to millisecond levels (tauris06). This belief is supported by the fact that in 23 cases (out of more than 150 known LMXBs, liu07) the NS spin frequency has been detected at the millisecond level (see patruno10, for a complete list). These detections support the scenario that LMXBs are the progenitors of millisecond radio pulsars with a low magnetic field.
When pulsations occur during surface thermonuclear explosions, known as type-I X-ray bursts (see strohmayer06, for a review), the sources are known as nuclear powered X-ray pulsars (hereafter NPXP). Up to now, pulsations during bursts (called burst oscillations) have been detected in 15 sources (altamirano10). During these events, the accumulated nuclear fuel first ignites at the point of the neutron star surface where it reaches the critical ignition column density and then spreads to all adjacent areas on the surface. When nuclear burning occurs uniformly over the surface, no “hot spot” is created and the neutron star spin will still be hidden. But in some cases a “patchy” burning process can occur, making the neutron star spin period visible. With the decrease of the X-ray burst flux, the non-uniformity fades out and so do the pulsations.
In other cases, pulsations occur in the “persistent” X-ray emission (i.e., not during type-I X-ray bursts), and the sources are known as accreting millisecond X-ray pulsars (hereafter AMXP). Up to now, 14 such sources have been detected (all transient X-ray sources, patruno10b; papitto11) and it is believed that matter from the accretion disk is channeled by the magnetic field lines onto the magnetic poles, forming a hot spot visible in X-rays. An important detection for our comprehension of the pulsating mechanism in AMXPs (versus the non pulsating majority of LMXBs) was achieved with the discovery of pulsations that were not detected throughout the outburst, but only intermittently, e.g., as HETE J1900.12455 (kaaret06), Aql X1 (casella08) and SAX J1748.92021 (altamirano08). These sources are important because they may be the intermediate link between persistent AMXPs and non pulsating neutron star LMXBs.
Of the currently known 14 AMXPs, only five, including IGR J175113057, belong to both the NPXP and AMXP classes, i.e., show pulsations during type-I X-ray bursts and during the persistent (non-burst) emission (altamirano10).
On 2009 September 12 (MJD 55087) INTEGRAL discovered a new hard X-ray source, IGR J175113057 (baldovin09), detected during the INTEGRAL Galactic bulge monitoring program (kuulkers07). Shortly thereafter, we reported the best X-ray position of the source from a preliminary analysis of our Chandra data: =17 51 08.66, = 30 57 41.0 (90% uncertainty of 0.6, nowak09). Near infrared follow-up observations identified within the Chandra error box the counterpart at a magnitude of =18.00.1 (torres11a; torres11b), but no radio counterpart was detected with a 3 upper limit of 0.10 mJy (miller09).
Shortly after the discovery, pulsations at 245 Hz were reported (markwardt09, using RXTE data), as well as the first type-I X-ray burst (bozzo09, Swift data), and burst oscillations very close to the neutron star spin frequency (watts09, RXTE data), making IGR J175113057 the fifth LMXB hosting a neutron star belonging to both the AMXP and NPXP classes.
Similarly to other AMXPs, IGR J175113057 can be classified as an atoll source based on its timing and spectral characteristics (bozzo09; papitto10; kalamkar11; ibragimov11; falanga11).
The source faded beyond RXTE detection limit after 2009 October 8 (MJD 55113, markwardt09), with coherent pulsations detected throughout all the outburst. Recently, possible twin kHz quasi-periodic oscillations (QPO) have been reported (kalamkar11). During the whole outburst, a total of 18 type-I X-ray bursts have been detected, marked as vertical arrows in Fig 1: ten by RXTE, three by Swift (one of which in common with RXTE), two by XMM-Newton, three by INTEGRAL, and one by Chandra (see falanga11, for a complete list). With the exception of the Chandra one, all the type-I X-ray bursts have been previously studied and reported (bozzo09; altamirano10; papitto10; falanga11; riggio11).
In this paper we focus on the unpublished Chandra/HETG observation of IGR J175113057 (2009 September 22, MJD 55096) that we triggered as part of our approved Chandra target of opportunity program.
2. The data
We observed IGR J175113057 for 20 ks with Chandra on 2009 September 22, from 07:40:39 UT until 13:32:03 UT with the High Energy Transmission Grating Spectrometer, HETGS (canizares00) collecting high resolution spectral information with the High Energy Grating, HEG 0.8–10 keV, and Medium Energy Grating, MEG 0.4–8.0 keV. The data were analyzed in a standard manner, using the CIAO version 4.3 software package and Chandra CALDB version 4.4.6. The spectra were analyzed with the ISIS analysis system, version 1.6.1 (houck02). For pileup correction in the presence of high fluxes (such as the type-I X-ray burst), we used the S-lang script simple_gpile2, within the ISIS fitting package, as described in nowak08 and hanke09. The Chandra -order spectrum was not used in the spectral analysis as it severely suffers from pile-up, especially in the burst phase. Given the source brightness and the intrinsic low Chandra background, no background removal was applied.
To develop a feeling of the overall flux evolution of IGR J175113057, we reduced
available RXTE data (same dataset of altamirano10). Standard filtering criteria were applied (see, e.g., rodriguez08)
and the average count rate was obtained from the top layer of PCU2 for
each individual observation. Light curves from the Crab
nebula and pulsar from the two closest observations were also extracted and used to renormalize
the PCA source count rate to the Crab one.
3.1. The X-ray position of IGR J175113057
We extracted the X-ray position of IGR J175113057 from the -order image obtaining =17 51 08.66, = 30 57 41.0, consistent with what we had reported in nowak09.
Fig. 2 shows the zeroth order image at (upper panel) and (lower panel) binning (burst plus persistent emission). There is apparent structure in these images due to pileup, especially during the burst portion of the lightcurve. For this reason, the source location was determined by intersecting the readout streak (visible in the image) with the grating arms (outside of the field of view covered in the image, but wide boxes along the arm positions are shown). This was accomplished with the findzo algorithm, which is standardly used for determining the zeroth order position when a readout streak is strongly detected (and hence pileup is affecting the zeroth order image). This is discussed more extensively in huenemoerder11, who discuss its use in the Chandra Transmission Gratings Catalog (TGCAT). Its estimated positional uncertainty is . We note that cross-correlation of Sloan Digital Sky Survey (SDSS) source positions with those obtained from the Chandra Source Catalog (CSC) – which relies upon Chandra absolute astrometry, as we use here – require only a 1- systematic correction (i.e., at 90% confidence level) to bring the SDSS and CSC positions into statistical agreement. (see Fig. 22 of primini11).
The error we obtain is, however, significantly less than the 90% confidence level uncertainty claimed for Chandra absolute astrometry111http://cxc.harvard.edu/cal/ASPECT/celmon/ when no other sources are present in the field of view for refined registration of the field, as is the case for this observation. We, therefore, attribute to the position found a 90% uncertainty of 0.6.
3.2. The type-I burst profile
An overview of the outburst of IGR J175113057 obtained from RXTE data is shown in Fig. 1. The Chandra data presented here occurred on 2009 September 22 (MJD 55096 in the plot).
A zoom in the lightcurve of IGR J175113057 as obtained during our Chandra observation, including the type-I X-ray burst detected at 2009 September 22 12:54:56 UTC, and lasting for about 54 s, is shown in Fig. 3. At the time of the burst the source was not being observed by RXTE.
Fitting the burst profile with a Fast Rise Exponential Decay (hereafter FRED) function, we obtain: a start time of the burst =(178611) s, relative to the first photon arrival time of the observation, which translates in 2009 September 22 12:54:56 UTC; a rise time of the burst, , within the range of (3.6–6.3) s and a decay time of the burst =(141) s.
In order to minimise the offset of photon arrival times due to the fact that the CCD chips are read out quasi serially in the timed exposure (TE) mode, the above results on the FRED properties were obtained using the S3 chip alone. Rectangular regions along the arms (2090 pixel boxes) and a circular region of 30 pixels around the -order (excising the innermost 16 pixel radius to avoid pile-up in the non-dispersed photon region) were used, together with a two-frame bin time, i.e., 3.68 s (1.84 s2 frames).
Given the very sharp flux increase of the burst
(Fig. 3 and 1.2 s for other reported bursts from the source, altamirano10), it is clear that the binning time used for the fit is likely to affect the result, especially as far as the rise time is concerned. Indeed, a fit with a one-frame bin time (1.84 s) tends to give a shorter rise =(1.6–3.6) s, but clear residual structures
appear, making any attempt to further constrain the rise-time inconclusive with the current Chandra TE mode.
3.3. Persistent and type-I burst spectral analysis
We extracted the spectra of the persistent (non-burst) emission as well as of the type-I X-ray burst, hereafter burst-all. Furthermore, we divided the burst in three segments that we call hereafter rise (about the first 4 s), peak (the next 13 s) and tail (the next 26 s)222The final 10 s of the burst have not been studied separately due to the extremely poor statistics.. These times were chosen also taking into account the more natural CCD-related read-out time frame, since the timed exposure (TE) mode configuration is not designed to accurately sample fast variability.
For each of the five parts (persistent, burst-all, rise, peak, tail), we extracted the first order dispersed spectra ( for HEG and MEG) and to increase the signal-to-noise ratio, we merged the two HEG () and MEG () spectra into one combined spectrum, for a total of five spectra (one per part)333The Chandra -order spectrum was not used in the spectral analysis as it severely suffers from pile-up, especially in the burst phase.. Final binning, starting at 0.8 keV, was chosen to have a signal to noise ratio higher than 5 and a minimum of 16 MEG channels per bin.
Within the type-I X-ray burst, while we are confident that the burst-all, peak and tail portions (54, 13 and 26 s, respectively) include no time-tag shift due to the CCDs read-out, there may potentially be a problem when integrating something varying as fast as 4 s (rise). Hence, though we have extracted its spectrum for visualization purposes, we have decided not to perform spectral fitting of the rise bit.
Figure 4 shows four of the five spectra we obtain: persistent (black crosses, the dimmest one), peak (blue circles, the brightest one), tail (brown triangles), and the shortest, 4 s exposure, spectrum of the rise with widest energy binning (green squares).
|(a)(a)In the fit we have used an improved model for the absorption of X-rays in the interstellar medium by wilms00.||(b)(b)Assuming a distance of 6.9 kpc (altamirano10).||Average flux(c)(c)Absorbed 0.5–8 keV flux.||Average luminosity(d)(d)Absorbed 0.5–8 keV luminosity, assuming a distance of 6.9 kpc.||DoF|
|Burst peak (13 s)||[1.02]||[0.55]||[1.79]||||51||110||5.93|
|Burst tail (26 s)||[1.02]||[0.55]||[1.79]||||5.30.8||18.9||1.02|
|Burst-all (54 s)||[1.02]||[0.55]||[1.79]||||1.60.1||5.20.5||28.6||1.6|
|Note. – Errors bars are 90% confidence level for one
parameter. The input seed photons to the nthComp model are blackbody in shape.
The fit of the persistent spectrum with a single non-Comptonised component, be it blackbody or disk blackbody, was very poor, with clear structured residuals. Hence we used a thermal Comptonization model (nthComp in XSPEC terminology, zdziarski96; zycki99), where soft seed photons of temperature are up-scattered by a thermal population of electrons at a temperature of .
Since the hot electrons up-scatter the seed photons, there are few photons remaining at energies below the typical seed photon energies, making it significantly different from a power-law below this energy. However, the spectrum can be parameterized by an asymptotic power-law index () that is also a parameter in the model, together with , and the model normalization.
While the low energy roll-over of the spectrum, related to , can be well appreciated in the Chandra energy range, the higher energy one, related to , is known to be out of the current range (e.g., papitto10). Since our spectral fits are not sensitive to its value, we choose to freeze it to =50 keV in the spectral fit. In the model, the seed photons can be blackbody or disk blackbody but since we cannot discriminate between the two with the current data, we choose to report only the blackbody shape case.
Figure 4 shows the best fit we obtained with the absorbed thermal Comptonization model, while Table 1 shows the obtained parameters. In this case, and of Table 1 are not applicable, since no additional thermal component is required in the persistent spectrum (see section 4.1).
The fits of the burst-all, peak and tail segments were obtained adding to the above persistent model, fixed and considered as the continuum, an additional blackbody component (bbodyrad in XSPEC terminology). The obtained best fit, temperatures and radii (assuming a distance of 6.9 kpc, altamirano10) can also be seen in Table 1.
We note that in all the spectral fitting of Table 1 the additional function simple_gpile2 was applied to the spectra in order to correct for pile-up distortions, as explained in nowak08 and hanke09.
4.1. The persistent emission
The broad-band spectra of AMXPs can normally be described as the composition of an accretion disk emission (peaking below 2 keV), a blackbody originating from the hot spot and a hard X-ray emission originated by thermal Comptonisation contributing to the whole broad-band 1-200 keV emission (e.g., see Figure 1 in ibragimov11, and references therein). As an example of the temperatures involved, the joint XMM-Newton-RXTE spectrum of IGR J175113057 as observed by papitto10 could be modeled by those three components that were interpreted, from the softest to the hardest, as a multicolored disk emission (0.2 keV), thermal emission from the neutron star surface (= keV) and thermal Comptonisation emission of hotter seed photons (= keV) by a hot plasma of electrons (= keV, =)444The authors obtain the well constrained Comptonization parameters using simultaneous XMM-Newton-RXTE data..
In our 20 ksec Chandra observation, due to lower statistics, the presence of more than one thermal emission component is not required by the data. Indeed, we obtain a single soft population of =0.550.03 keV as seed photons for the Comptonisation, most likely the non disentangled combination of accretion disk and neutron star surface/halo thermal populations, with no additional thermal component required. A 90% upper limit on the normalization of an additional thermal component, e.g., bbodyrad model (with temperature fixed to the Comptonization component of Table 1) gives an absorbed 0.5–8 keV flux of , to be compared with our persistent emission of . This is consistent with what papitto10 obtain using a 70 ksec XMM-Newton observation555We refer to Table 2 in papitto10, model A, the closest to the one we have here., i.e., a total thermal absorbed 0.5–8 keV flux of about , besides the Comptonised nthComp component. Furthermore, adding a diskline component and freezing its values (except normalization) to the best fit from papitto10, we obtain a 90% upper limit on the equivalent width of 50 eV, consistent with that measured by XMM-Newton, 43.90.06 eV.
The 0.5–8 keV spectrum of IGR J175113057 is energetically dominated by a power-law which is equivalent to a broader Comptonised emission on a limited bandwidth between and . In the case of nthComp, the code provides, as the best-fit parameters, the seed photon temperature , the electron temperature and the power-law spectral index .
Since Chandra limits us to studying the 0.8–8 keV range, the temperature of the Comptonising plasma was fixed to =50 keV, while the photon index of the power-law, that dominates the spectrum, was obtained by the fit as = . Since no high energy cutoff appears in our Chandra spectrum of IGR J175113057, the choice of fixing the parameter to 50 keV does not affect our results. An assumed temperature of, e.g., 100 keV yields comparable results and value.
Once and are provided, it is possible to infer the Thomson optical depth through the relation:
(see, e.g., lightman87). We obtain values of optical depth =(1–2) for =(50–100 keV), similarly to what obtained by papitto10.
Using the persistent model obtained in Table 1, we obtain =1.710 (at 6.9 kpc) and an extrapolated unabsorbed =710 . This is consistent with Figure 4 in altamirano10, where a persistent (7–8)10 is expected from the source at 6.9 kpc in the time lasting between the fifth and sixth RXTE burst, when the Chandra observation occurred.
A more detailed analysis of the persistent spectrum is not justified by the data. No low energy features are visible and since the distribution of the residuals does not show any systematic trend, we believe that the best fit model we obtain (Fig.4 and Table 1) is the simplest and most coherent description of the data, with results compatible with what found in literature, albeit subject to uncertainties due to model extrapolations and comparison of different mission calibrations.
4.2. The type-I burst
During our Chandra observation, a type-I X-ray burst was observed (Fig.3). The burst, fit with a FRED function resulted in a rise time =(3.6–6.3) s (but see section 3.2) and decay =(141) s. This is “slow” if compared to the ten type-I X-ray bursts of IGR J175113057 observed in the RXTE/PCA data (altamirano10), where all the bursts reached their maxima within 1.2 s and with decay times in the range of 5–8 s. Although we cannot exclude that this burst is slightly longer than the RXTE reported ones, we note that a more accurate comparison is hampered by the limitations of the Chandra timed exposure (TE) mode, where each chip is exposed for approximately 2 s, to be compared with the 0.1 s time resolution of the RXTE data of altamirano10 (see section 3.2).
The low statistics obtained in the burst prevented us from doing an accurate phase resolved spectroscopy as done in the case of, e.g., the brightest RXTE burst (altamirano10; falanga11) or in the XMM-Newton ones (papitto10). In our case it was only possible to split the burst in three segments (rise, peak, tail), fitting only the latter two because of instrumental limitations (see section 3.3). As shown in Table 1, the burst emission could be well fit by a single blackbody with the temperature decaying from = keV to = keV, likely indicating the cooling of the neutron star surface after the burst ignition. The related emitting area has in both cases a radius comparable with 5 km that is consistent with that found by, e.g., falanga11 and papitto10 for the other bursts from IGR J175113057. We note however that direct comparison with the other X-ray bursts is to be taken with caution, because unlike in the other cases, where a detailed phase resolved analysis was possible, in our study the obtained quantities of Table 1 are averaged on large portions of the burst (e.g., 13 s for the peak).
Nevertheless, a comparison of the overall properties of our type-I burst with the ones previously reported using RXTE and INTEGRAL can be attempted. Using the peak model obtained in Table 1 (with and the nthComp normalisation set to 0), we obtain a peak unabsorbed luminosity 1.510 for the source at 6.9 kpc. This is a 13 s average value and to compare it with the non averaged RXTE ones, we should estimate our “real peak” value, obtaining it from the FRED function that best fits our burst profile. This results in 2.310 , to be compared to Figure 4, middle panel, in altamirano10, where a bolometric peak (2.5–3.5)10 is expected from the source at 6.9 kpc, during the Chandra observation, between the fifth and sixth RXTE burst.
To obtain an estimate of the overall type-I burst total energy release () and fluence (), to be compared with the results of the RXTE bursts by altamirano10 and of the RXTE-INTEGRAL ones by falanga11, we consider the burst-all spectrum of Table 1. Setting and the nthComp normalisation to 0, we obtain a bolometric unabsorbed luminosity of 210 that results in a total energy release of 1.110 erg in 54 s. This is slightly lower than the range obtained by altamirano10, (2.5–3)10 erg. Similarly, a fluence of =210 is obtained, to be compared to (3.2–4.2)10 of falanga11. In both cases we are dimmer than the previously reported X-ray bursts. Indeed, our burst could be intrinsically dimmer, however considering that we are subject to uncertainties in the model extrapolations beyond the Chandra energy domain, as well as to uncertainties in the mission cross-calibrations, it is reasonable to conclude that we have no strong evidence for the type-I X-ray burst observed by Chandra from IGR J175113057 to be inconsistent with the previously reported bursts. Furthermore, the time averaging issue reported above and in Section 3.2 due to the timed exposure (TE) mode of the observation may be the dominant source of discrepancy. Indeed a limited number of type-I X-ray bursts have been studied up to now with Chandra grating; most observations were done in continuous clocking mode (CC), for which the ACIS-S CCDs are read out continuously, providing a 3 msec timing, at the expense of one dimension of spatial resolution. An example is, e.g., the study of radius-expansion burst spectra from 4U 172834 (galloway10). Out of the 25 bursts detected, time-resolved spectroscopy of the summed signal from the four brightest bursts (with summed bolometric peak flux of about ) was carried out. A clear photospheric radius expansion in these bursts could be seen, well sampled over seven time bins on a total of 12 sec of burst duration. Stacking data-sets from several Chandra burst intervals for a more detailed spectral study and evolution has also been the approach of thompson05 for GS 1826238. Similarly to our case, TE mode had been used and indeed time bins of about 10 sec (minimum) were extracted for a spectral study on the six averaged type-I bursts detected from the source (lasting about 150 sec).