Transient Phenomena in Anomalous X-ray Pulsars
GianLuca Israel, Federico Bernardini, Marta Burgay, Nanda Rea, Andrea Possenti, Simone Dall'Osso, and Luigi Stella
aa r X i v : . [ a s t r o - ph ] N ov Transient Phenomena in Anomalous X–ray Pulsars
G.L. Israel ∗ , F. Bernardini † , M. Burgay ∗∗ , N. Rea ‡ , A. Possenti ∗∗ , S. Dall’Osso ∗ andL. Stella ∗ ,§ ∗ INAF - Astronomical Observatory of Roma, Via Frascati 33, 00040 Monte Porzio Catone, Italy † University of Rome “La Sapienza”, Department of Physics, P.le A. Moro 5, 00185, Roma, Italy ∗∗ INAF - Astronomical Observatory of Cagliari, localitá Poggio dei Pini, strada 54, 09012 Cagliari, Italy ‡ SRON Netherlands Institute for Space Research, Sorbonnelaan, 2, 3584CA, Utrecht, The Netherlands § on behalf of a larger team (the complete list is reported in the acknowledgment section) Abstract.
In 2003 a previously unpulsed Einstein and ROSAT source cataloged as soft and dim (L X of few × ergs − )thermal emitting object, namely XTE J1810-197, was identified as the first unambiguous transient Anomalous X-ray Pulsar.Two years later this source was also found to be a bright highly polarized transient radio pulsar, a unique property among bothAXPs and radio pulsars. In September 2006 Swift
Burst Alert Telescope (BAT) detected an intense burst from the candidateAXP CXOU J164710.2 − − Keywords:
Pulsars; Magnetars; X-rays; Variability
PACS:
INTRODUCTION
At the beginning of the X–ray astronomy era the study ofthe X-ray variability (both of the flux and timing proper-ties) of Cen X–3 allowed astronomers to unambiguouslyassess the binary nature of the source and to identify theaccretion of mass, flowing from the companion onto arotating neutron star, as the main mechanism to produceX–rays [1, 2]. Since then, several classes of high energysources hosting a compact object have been identifiedand their variability, observed on different timescales,used to test models and/or to study the emission mech-anisms as a function of flux (while keeping fixed otherparameters, such as the distance and the geometry of thesystem). For the class of accreting neutron stars this ap-proach was successfull in confirming, among others, thepresence of a centrifugal barrier to accretion, testing thedependence of the period derivative with respect to thesource luminosity, and studying the change in the pho-ton propagation direction as a function of the accretionrate by means of the pulse shape changes [fan and pen-cil beam model; as an example see 3, 4, 5]. Isolatedneutron stars are relatively constant sources, making theabove approach unreliable. However, there are two smallclasses of isolated neutron stars which show spectacularevents during which their luminosity may change up to10 orders of magnitude on timescales down to few mil- liseconds. These objects are better known as AnomalousX–ray Pulsars (AXPs; 10 objects plus 1 candidate) andSoft g -ray Repeaters [SGRs; 4 objects plus 2 candidates;for a review see 6]. It is believed that AXPs and SGRs arelinked at some level, owing to their similar timing prop-erties (spin periods in the 2-12 s range and period deriva-tives ˙P in the 10 − –10 − s s − range). Both classeshave been proposed to host neutron stars whose emis-sion is powered by the decay of their extremely strongmagnetic fields [ > G; 7, 8].Different types of X-ray flux variability have been dis-played by AXPs. From slow and moderate flux changes(up to a factor of few) on timescales of years (virtu-ally all the object of the class), to moderate-intense out-bursts (flux variations of a factor up to 10) lasting for 1-3years (1E 2259 + − − ergs) on sub-second timescales [4U 0142 + +
586 and 1E 1048.1 − + ∼
10 persistent fluxenhancement in an AXP was followed (or proceeded) bythe onset of a bursting activity phase during which thesource displayed more than 80 short bursts [10, 11]. Thetiming and spectral properties of the sources changed sig-ificantly and recovered the pre-bursting activity phasevalues within few days, likely due to the relatively highluminosity DC level ( ∼ erg s − ). However, it wasonly in 2003 that the first transient AXP was discov-ered, namely XTE J1810-197, which displayed a fac-tor of >
100 persistent flux enhancement with respectto the unpulsed pre-outburst quiescent luminosity level[ ∼ erg s − ; 12, 13, 14, 15]. Unfortunately, the ini-tial phases of the outburst were missed and we do notknow whether a bursting activity phase, similar to that of1E 2259 + − > XMM-Newton monitoring observations of XTE J1810-197 asit approached to quiescence, (ii) the comparison of X-ray and radio emission from XTE J1810-197 by meansof two ∼ XMM-Newton and Parkes, (iii) and the results of the first 6months monitoring of CXOU J164710.2 − Swift , Chandra and
XMM-Newton in the X-rays and Parkes inthe radio band.
XTE J1810–197: FROM OUTBURST TOQUIESCENCE
Since the very first
XMM-Newton ± BB ≈ ± BB ≈ X ∼ × erg s − in the 0.5-10 keV range) was sig-nificantly different from that serendipitously recordedby ROSAT in 1992 (one BB with kT ≈
160 eV andR BB ≈
10 km; extrapolated luminosity in the 0.5-10 keVrange of L X ∼ × erg s − and for a distance of3.3 kpc). Moreover, the source showed a 5.54 s pulsationwith a pulsed fraction of nearly 45% during outburst,while an upper limit of 24% was inferred from the ROSAT data. The above issues originated a numberof important questions awaiting for an answer: Is thesoft BB component detected by XMM-Newton evolvinginto the quiescent BB component seen by ROSAT ?Alternatively, is the emission from the whole surfacealways present ? What happens to the higher temperatureBB component as the source approaches to quiescence? Which is the pulsed fraction level of the source inquiescence (if detectable) ? Does the outburst changedpermanently the timing/spectral properties (such as thepulsed fraction, the flux and temperature or size of thequiescent BB component) of the source ?In order to try answering to the above questions we re-duced all the archival (6) and still proprietary (2)
XMM-Newton observations [for the details see 20, 21] and fit-ted the eight spectra all together. All the spectra havebeen rebinned in order to ensure that each background-subtracted spectral channel has at least 25 counts, andthat the instrumental energy resolution is not oversam-pled by a factor larger than 3 [22, ; indeed the correctapplication of the above rules prevents artificially low(good) reduced c s.]. In particular, we can outline theobtained results as follows: By extending the spectral recipe outlinedby [13, 23] we applied the two BB spectral fit analy-sis to the fading phases of XTE J1810-197 until March2007 when the flux source was ∼ c ∼ H =0.58 ± × cm − ). While thesoft BB component smoothly approaches to that in qui-escence (see Figure 1, left panel, 2 nd and 3 rd plots), wenote a number of ambiguities difficult to account for bymeans of simple assumptions. The hard component BBradius is not monotone and it increases after 2.5 years ofsmooth decrease (left panel, 4 th plot) while the tempera-ture approaches to that of the soft BB in 2003 (left panel,5 th plot). Moreover, none of the spectral parameters orcomponents is able to account for the flattening, at the25% level, showed by the pulsed fraction evolution (leftpanel, 1 st plot) [21]. The addition of a further BB com-ponent gives a better fit (reduced c ∼ H =0.70 ± × cm − ; F-test probabilitygives 7.3 s ) though not yet satisfactory. Notably, the fitgives parameters and flux (for the coldest BB) whichare virtually equal to those inferred in quiescence. Evenmore interesting, the hottest BB components show anearly constant evolution of the temperatures (see Fig-ure 1, right panel, 3 rd ad 5 th plots), leaving the radiias the only variable parameters to account for the de-caying phases of the outburst (right panel, 2 nd and4 th plots). Since September 2006 the hottest BB isnot anymore needed to fit the spectra (upper limit of ∼ × − erg cm − s − ). In this scenario, the alreadymentioned flattening of the pulsed fraction might be eas- IGURE 1.
Evolution of the spectral parameters for the 2BB (left panel) and 3BB (right panel) models together with the pulsedfraction as a function of the observed 1-10 keV band flux. Central Panel: schematic view of the assumed scenario with marked theemission regions of the hottest BB components. ily accounted for by the disappearance of the hot BB.A pulse phase spectroscopic analysis do not show anyphase lag between the two highest temperature BBs [21].
Further components:
During the first two XMM ob-servations (2003-2004) the spectral fit residuals clearlysuggest (at a 3.2 s confidence level) the likely presenceof an additional hard component above 7–8 keV whichwe are not able to characterize due to poor statistics inthis band. We can only speculate that might be somewhatrelated to the presence of a hard power-law-like compo-nent detected in other AXPs [24, 25] by INTEGRAL andwhich extends up to (at least) 200keV [26]; X–ray and radio campaigns:
The simultaneous ra-dio and X–ray observations of XTE J1810-197 carriedout in September 2006 and March 2007 showed that thepulse alignment between the two bands is high and stable(see Figure 2), while the pulse width is relatively small( ∼ ∼ . . N o r m . I n t en s i t y Phase
FIGURE 2.
XTE J1810-197
XMM-Newton
PN+MOS 0.5-10keV and Parkes 20cm light curves, folded to the spin pe-riod, carried out during the September 2006 campaign. Super-imposed to the X–ray folded light curve is the best sinusoidalfit [adapted from 27]. (three BBs) is a first attempt to infer the evolution of anumber of physical quantities while making use of wellassessed and reliable components. The three BB modelhas the advantage of being model-independent, and ofminimizing the number of variable parameters during
10 100 − − − . C oun t s s − k e V − Energy (keV)
FIGURE 3.
Left Panel: phases of the
Swift
XRT,
XMM-Newton and
Chandra observations of CXOU J164710.2 − XMM-Newton point (at day -5) would be at the reported phaseonly in the hypothesis that the pre- and post-glitch parameters are similar [for more details see 14]. Right Panel: the 15-150 keV
Swift
BAT spectrum of the 20ms long burst detected from CXOU J164710.2 − the outburst (only the radii are changing and both de-creasing). With respect to other model recently devel-oped [see for example 29] we carried out a cross checkwith the timing properties and rejected all those mod-els/components not able to reproduce the pulsed frac-tion evolution. All the above inferred information areimportant in the effort we are currently making in de-veloping and using more detailed and complex (but nec-essarily model-dependent) modelizations. In particular,the simultaneous fit of the spectra and energy resolvedfolded light curves might provide a tool to infer the ge-ometry of the surface temperature distribution and to in-dependently check the goodness of the assumed spectralmodel(s). Finally, we note that the three BB model fit im-plies that the BB component accounting for the emissionfrom the whole sourface is almost unpulsed (pulsed frac-tion of 9% ± CXOU J164710.2–455216: FROMQUIESCENCE TO OUTBURST
On 2006 September 21, the candidate AXP CXOUJ164710.2 − ∼ erg s − )and short (20ms) burst promptly detected by the Swift
BAT. Together with the burst, large changes in the tim-ing and spectral properties of the persistent componentwere detected and seen evolving during the subsequentweeks. In particular, the
Swift
XRT monitoring (plus twoproprietary
XMM-Newton and two archival
Chandra ob-servations ) during the first six months since the outburst allowed to infer the following characteristics:
The BAT burst:
The prompt event recorded by
Swift
BAT has an exponential time decay t of 3.3 ± s confidence level) and the spectrum can be fitted withboth a blackbody with kT of 9.9 ± . . keV and a G of1.8 ± . ≈ − erg cm − corresponding to a total en-ergy of ∼ × ergs (for a fiducial distance of 5kpc).Compared with the properties of the previously detectedAXP bursts, the current burst has a duration within 1 s from the log-normal distribution average value inferredfor 1E 2259 + ∼
80 detected from 1E 2259 + − The phase-coherent timing and the glitch:
Thepulse phase evolution is consistent with the occurrenceof a large glitch (
D n / n ∼ − ), the largest ever de-tected from a neutron star . We also detected a quadraticcomponent in the pulse phases corresponding to a ˙P = . ( ) × − s s − and implying a magnetic fieldstrength of 10 G (see Figure 3, left panel). The first 1- The glitch detection was obtained by minimizing the number ofvariable peaks in the pulse profile. A less significant timing solutionis feasible and requesting only a ˙P component. We consider the latterunlike since it would imply high variability for all the peaks; see [19]for details.
Swift
XRT spectrum was carried out ∼
13 hoursafter the burst detection and showed, in addition to a kT ∼ .
65 keV blackbody ( R BB ∼ . G ∼ . kT s = . ± .
05 keV with R BBs = . ± . kT h = . ± . . keV with R BBh = . ± . The flux and pulsed fraction evolution:
The fluxdecay of CXOU J164710.2 − F (cid:181) t a , with index a of –0.28 ± +
586 burst-active phase).Moreover, we found that the PL component decays morerapidly (index a of –0.38 ± a of –0.14 ± ∼
80% (as recorded by an
XMM-Newton observationfew days before the burst) to ∼
10% few hours after theBAT event. The spectral and timing analysis clearly showthat only the blackbody component is responsible for thepulsed flux. In particular, the signal fractional rms as afunction of time is well fitted by a power-law with index a of +0.38 ± The quiescent properties:
Archival
Chandra dataanalysis revealed that the modulation in quiescence is100% pulsed at energies above ∼ Radio observations:
Since the onset of the outburst,CXOU J164710.2 − mJy at 1400 MHz [30].All these results confirmed unambiguously thatCXOU J164710.2 − − − Swift mission.
ACKNOWLEDGMENTS
The complete list of the co-authors is: S. Campana, A.Corongiu, J. Cummings, M. Dahlem, M. Falanga, P. Es-posito, D. Götz, S. Mereghetti, P. M. Muno, R. Perna,R. Turolla and S. Zane. This work is partially supportedat OAR through ASI, MIUR, and INAF grants. Weacknowledge financial contribution from contract ASI-INAF I/023/05/0.
REFERENCES