Discovery of 442-Hz Pulsations from an X-ray Source in the Globular Cluster NGC 6440
Fotis P. Gavriil1 , Tod E. Strohmayer, Jean H. Swank, Craig B. Markwardt2,3
9v1 [astro-ph] 6 Aug 2007
NASA Goddard Space Flight Center, Astrophysics Science Division, Code 662, X-ray Astrophysics Laboratory, Greenbelt, MD 20771, USA ABSTRACT We report on the serendipitous discovery of a 442-Hz pulsar during a Rossi X-ray Timing Explorer (RXTE) observation of the globular cluster NGC 6440. The oscillation is detected following a burst-like event which was decaying at the beginning of the observation. The time scale of the decay suggests we may have seen the tail-end of a long-duration burst. Low-mass X-ray binaries (LMXBs) are known to emit thermonuclear X-ray bursts that are sometimes modulated by the spin frequency of the star, the so called burst oscillations. The pulsations reported here are peculiar if interpreted as canonical burst oscillations. In particular, the pulse train lasted for ?500 s, much longer than in standard burst oscillations. The signal was highly coherent and drifted down by ?2×10?3 Hz, much smaller than the ?Hz drifts typically observed during normal bursts. The pulsations are reminiscent of those observed during the much more energetic “superbursts”, however, the temporal pro?le and the energetics of the burst suggest that it was not the tail end nor the precursor feature of a superburst. It is possible that we caught the tail end of an outburst from a new ‘intermittent” accreting X-ray millisecond pulsar, a phenomenon which until now has only been seen in HETE J1900.1?2455 (Galloway et al. 2007). We note that Kaaret et al. (2003) reported the discovery of a 409.7 Hz burst oscillation from SAX J1748.9?2021, also located in NGC 6440. However, Chandra X-ray Observatory imaging indicates it contains several point-like X-ray sources, thus the 442 Hz object is likely a di?erent source. Subject headings: — stars: neutron — X-rays: bursts — globular clusters: individual (NGC 6440)
1 2 3
NPP Fellow; Oak Ridge Associated Universities, Oak Ridge, TN CRESST Department of Astronomy, University of Maryland, College Park, MD 20742.
–2– 1. Introduction
The discovery of millisecond spin periods of neutron stars in Low Mass X-ray Binaries (LMXBs) with the Rossi X-ray Timing Explorer (RXTE) has helped elucidate the nature of these sources. Neutron star LMXBs consist of a neutron star accreting from a low mass companion. As material (mostly H and He) is accreted onto the star and gets compressed, it eventually ignites and burns unstably (see Strohmayer & Bildsten 2007). This phenomenon is observed as a Type I X-ray Burst. Type I X-ray bursts have been observed from over ?70 LMXBs (see Liu et al. 2006, and references therein). The recurrence time of these bursts varies but in some cases it can be as frequent as every few hours (see Galloway & Muno 2007, for examples). Occasionally it is possible to observe the spin of the neutron star modulating the burst emission – the so called “burst oscillations” (Strohmayer et al. 1996). Burst oscillations have been observed from ?18 LMXBs (see Galloway & Muno 2007, for examples). X-ray bursts from LMXBs have been discovered which are ?1000 times longer, and thus, that much more energetic, than canonical Type I X-ray bursts. They are aptly named “superbursts”. Superbursts are believed to occur by the unstable burning of carbon (Strohmayer & Brown 2002; Cumming & Bildsten 2001). Strohmayer & Markwardt (2002) discovered highly coherent pulsations during a superburst from 4U 1636?53. The pulse train lasted for ?800 s, as opposed to the ?10 s long pulse trains observed in Type I X-ray bursts. Superbursts are much rarer than Type I X-ray bursts. Thus far ten have been observed from eight LMXBs (see in’t Zand et al. 2004).
Between 2005 March 7 and 2005 July 21 the PCA Galactic Bulge Scan Survey4 discovered an outburst from the direction of the Globular Cluster NGC 6440. A follow up 1.8-ks long pointed RXTE observation (observation identi?cation number 91050-03-07-00) was initiated on 2005 June 14. The data presented here were acquired from the Proportional Counter Array (PCA) on board RXTE. The PCA consists of ?ve identical and independent proportional counter units (PCUs). Each PCU is a 90% Xenon/10% Methane gas ?lled proportional counter. Each PCU has a collimated 1? ×1? ?eld-of-view, 256 spectral channels in the 2–60 keV range, and a limiting temporal resolution of ?1 ?s. Only two PCUs (PCU 0 and PCU 2) were operational throughout the span of the observation. The data were
–3– taken in E 125us 64M 0 1s mode, which returns events to a limiting resolution of 125 ?s and with moderate (64 channels as opposed to the full 256 channels) spectral resolution. This mode was used because it is not as susceptible to bu?er over?ows during high countrate data as compared to the less restrictive modes (e.g. GoodXenon). Using the pointing of the s observation (17h 48m 52.8, -20? 21′32′′ ) and the planetary ephemeris DE200, the times of raw events were corrected to the solar system barycenter. Binning the events into a 1-s resolution lightcurve reveals a quickly decaying burst (Fig. 1).
Analysis and Results
This burst was reminiscent of canonical Type I X-ray bursts from LMXBs, therefore we searched this event for burst oscillations. We rebinned the raw events using the full spectral band pass into a time series with 0.5/1024 s resolution, which yields an equivalent Nyquist frequency of 1024 Hz. A Leahy normalized (see Leahy et al. 1983) power density spectrum (PDS) is displayed in Figure 1 (inset). A prominent peak is clearly seen at 442 Hz. The probability of this peak occurring by chance after accounting for the number of trials (the total number of frequency bins in our PDS) is ?2 × 10?9 .
The NGC 6440 Field
NGC 6440 harbors the bright X-ray transient SAX J1748.9?2021, which exhibited a 409.7 Hz burst oscillation (Kaaret et al. 2003). However, there are many other X-ray sources in the cluster (see Fig 2). Chandra X-ray Observatory (CXO) imaging by Pooley et al. (2002) revealed 24 X-ray sources in NGC 6440, and they concluded that 4–5 of these sources are likely quiescent LMXBs. Thus, the phenomenon we discovered is most likely from a di?erent object in the cluster.
To study the frequency evolution of the pulse we calculated a dynamic Z 2 statistic .The Z 2 statistic is analogous to the Fourier power spectrum, with the advantage that the data need not be binned, thus allowing us to oversample our data. The Z 2 statistic is de?ned as:
2 = Nγ
1800 Leahy Normalized Power 0 200 1700 1600 1500 Counts 1400 1300 1200 1100 1000 900 800 400 600 Time (s) 800 1000 60 50 40 30 20 10 0 400 410 420 430 440 450 460 470 480 Frequency (Hz)
Fig. 1.— Burst lightcurve binned with 1-s resolution and using the full PCA bandpass. Inset: “Leahy” normalized (Leahy et al. 1983) power spectrum of the entire observation. Notice the signi?cant peak at 442 Hz.
Fig. 2.— Archival CXO image of NGC 6440. Notice that the cluster contains many X-ray point sources.
–6– where Nγ is the total number of photons in each interval, Nharm is the total number of harmonics that one deems signi?cant, νj is the frequency searched over, and ti is the event time. The factor in front of the summations normalizes the Z 2 statistics in an analogous way to Leahy normalizing a PDS. We calculated the dynamic Z 2 power spectrum using a 200-s long window, which was translated across the observation with a step size of 16 s. Our dynamic Z 2 power spectrum is displayed in Fig. 3. Notice that the pulsations where highly signi?cant for 576 s. As is common for canonical X-ray bursts, the frequency drifts; however, unlike the large ?1-2 Hz drifts seen in those, the pulsation here drifts down in frequency only by ?2.1×10?3 Hz in 576 s. To further quantify this frequency evolution we determined how the phase of the pulsa2 tions varies as a function of time. For each interval with Z1 > 16 we generated a pulse pro?le by folding the events in that interval on the frequency determined from the Z 2 statistic. We then cross correlated these pulse pro?les with a sinusoid of ?xed phase to determine the pulse phase as a function of time. We can model the pulse phase (φ) at a given time (t) by the following Taylor expansion: 1 φ(t) = φ(t0 ) + ν(t0 )(t ? t0 ) + ν(t0 )(t ? t0 )2 + · · · , ˙ 2 (2)
where ν is the barycentric frequency, ν is the frequency derivative, and t0 is some ref˙ erence epoch. We were able to whiten our phase residuals with just a frequency derivative (see Fig. 3, bottom). We ?nd ν = 442.33850(5) Hz, ν = 3.7(2) × 10?6 Hz s?1 at ˙ t0 =53535.463215 MJD (UTC). The phase residuals after subtracting the model given by Eq. 2 are shown in Fig. 3 (bottom panel). Notice that the pulse train lasts for 576 s; this is quite long when compared to those of Type I X-ray bursts. We folded all the events within the 576 s pulse train using our best ?t ν and ν, and the resulting pulse pro?le is shown ˙ in Fig. 4. Notice that the pulse pro?le is highly sinusoidal. Fitting a sinusoid to the pulse pro?le returned a reduced χ2 value of 0.76 for 13 degrees of freedom, a?rming the absence of any harmonic content. From the ?t we obtain a fractional pulse amplitude of 2.1±0.1%. It is possible that the frequency drift we have observed is entirely due to orbital motion. The observed spin frequency of a neutron star in a circular orbit is given by ν = ν0 1 ? vns sin i sin c 2πt + φ0 T , (3)
where, ν0 is the barycentric frequency at time t = t0 , vns sin i is the projected velocity of the neutron star, T is the orbital period, and φ0 is the orbital phase at time t = t0 . Unfortunately we do not have a long enough data set to place interesting constraints on T or vns sin i; however, we can determine whether orbital modulation is consistent with the observed frequency drift. For example, if we assume, for solely demonstrative purposes, that this system
1500 1000 442.35 442.345 442.34 442.335 442.33 0.05 0 ?0.05 0 500 1000
Time (in seconds from start of observation)
Fig. 3.— Pulse Frequency Evolution. Top: The burst lightcurve as shown in Fig. 1. Middle: Dynamic Z 2 power spectrum. Notice that the pulsations remain highly signi?cant 2 (with Z1 > 16) for 576 s. The solid line represents our best-?t frequency model consisting of just a frequency and a frequency derivative term. Bottom: Phase residuals after subtracting our best ?t frequency model. Notice that the pulsations are highly coherent.
33500 0 0.5 1 Phase (Cycles) 1.5 2
Fig. 4.— The pulse pro?le obtained by folding all events in the 576-s long pulse train (i.e., the interval where the pulsations were highly signi?cant as determined from the Z 2 power spectrum) on the ephemeris determined in our timing analysis. The pulse shape is highly sinusoidal. The solid line represents the best-?t sinusoid. The ?t yielded χ2 =0.76 for 13 degrees of freedom, a?rming the absence of any harmonic content.
–9– has a projected velocity comparable to the one found by Strohmayer & Markwardt (2002) for 4U 1636?53 (vns sin i = 136 km ?1 ), then ?tting Eq. 3 to our frequency time series yields a good ?t with a reasonable orbital period, i.e. T =7±1 hr. Hence, binary motion can account for the observed frequency drift. Observing another outburst would be essential in ultimately constraining the orbital parameters of this system.
The timing analysis seems to suggest that this event is in fact very di?erent from a canonical burst oscillation. To study it further we performed a spectral analysis. We analyzed the spectral evolution of the burst by breaking the burst up into ten intervals. The exposure of each interval was selected by demanding that each one contained an equal number of photons. Using the last interval as an estimator of the persistent emission, we ?t each interval to a simple photoelectrically absorbed blackbody model while holding the column density ?xed to the value found by Pooley et al. (2002) for the optical reddening. We ?nd only subtle evidence for spectral softening, signi?cant only at the 1-σ level. Using the distance estimate to the cluster found by Ortolani et al. (1994), d = 8.5 ± 0.4 kpc, we were able to calculate the luminosity of the burst. At the start of the observation the ?ux was ?2×1037 erg s?1 , which is ?0.1LEdd , where LEdd is the Eddington luminsosity for a neutron star. The burst lightcurve was well ?t by an exponential with a decay timescale of 22 s. Unfortunately the observation only caugth the tail end of the burst-like event, we could not therefore characterize the peak luminosity of the burst.
Discussion Super Burst?
The timing properties of the pulsations are reminiscent of those observed during superbursts. In particular the highly coherent pulsations and the long pulse trains. The superburst interpretation would be very exciting given the rarity of such events. However, we could not unambiguously determine this is the case given that we only caught the tail end of the burst. To study this phenomenon further we extrapolated our luminosity model determined in our spectral analysis before the start time of the observation. We ?nd that the burst reaches the Eddington luminosity ?50 s before the start of the observation. Now, if we assume that we caught the tail of the outburst it is very surprising that it dropped from ?LEdd down to ?.01LEdd in ?77 s. This is much too fast for a superburst-like event. If, on the other
– 10 – hand, this is a precursor events, such as those seen in superbursts, than it is unclear why the persistent ?ux did not rise back to ?LEdd shortly after as observed in other superbursts. Thus, the spectral analysis of this event suggests it is not a superburst.
The timing properties of the pulsations share all the hallmarks of those observed during superbursts, but the spectral analysis makes this interpretation unlikely. If this event is not a superburst, then could it be that these highly coherent pulsations have eminated from an accretion-powered millisecond pulsar? Recently Galloway et al. (2007) discovered “intermittent” pulsations from the accretion-powered millisecond pulsar (AMP) HETE J1900.1?2455. They found that the properties of this pulsar di?ered from those of the other six known AMPs. For example, the pulsations were only present in the ?rst few months of its outburst as opposed to other AMPs which show pulsations throughout (see Wijnands 2004, for a review). In addition, the pulse amplitude increased at three points in time that were almost coincident with the times of thermonuclear bursts. The pulsed amplitude subsequently decaded after the bursts. Thus, it is plausible that we have observed a similar phenomenon. Following the theoretical work of Cumming et al. (2001) and Payne & Melatos (2006), Galloway et al. (2007) suggest that the reason why the pulsations in HETE J1900.1?2455 sometimes “switch o?” is because the magnetic ?eld is burried by accreted material, thus it cannot channel the ?ow which gives rise to the pulsations in the ?rst place If the event we have discovered is another example of these intermittent pulsations from an unidenti?ed source in NGC 6440, then a future outburst should reveal more long pulse trains, and if they happen to be correlated with the times of Type-I X-ray bursts then this would solidify the connection between this source and HETE J1900.1?2455.
We discovered a 442 Hz pulsation during a burst from an X-ray source in NGC 6440. We could not establish the energetics of the burst as RXTE only caught the tail end of the event, but based on the timing properties of the pulsation we conclude that it was not a canonical burst oscillation. In particular we ?nd that: The pulse train lasted for 576 s, as opposed to the ?10 s long pulse trains observed in Type I X-ray bursts. The signal was highly coherent and drifted down only by 2.1 × 10?3 Hz, as opposed to Type-I X-ray burst which exhibit drifts which are three orders of magnitude larger. The pulsations share all
– 11 – the properties of those observed during superbursts; however, the complete energetics of this event make this interpretation implausible. We conclude that the long pulse train is most likely an “intermittent” pulsation from an accreting millisecond pulsar such as those seen only in HETE J1900.1-2455 thus far (Galloway et al. 2007). We note that NGC 6440 is known to harbor a transient X-ray source with a 409.7 Hz spin frequency (Kaaret et al. 2003) However, X-ray imaging of the cluster by Pooley et al. (2002) has shown that NGC 6440 contains many X-ray sources. Thus, the burst presented here most likely emanated from a di?erent object in the cluster. FPG is supported by the NASA Postdoctoral Program administered by Oak Ridge Associated Universities at NASA Goddard Space Flight Center. This research has made use of data obtained through the High Energy Astrophysics Science Archive Research Center Online Service, provided by the NASA/Goddard Space Flight Center.
REFERENCES Cumming, A. and Bildsten, L. 2001, ApJ, 559, L127 Cumming, A., Zweibel, E., & Bildsten, L. 2001, ApJ, 557, 958 Galloway, D. K., Morgan, E. H., Krauss, M. I., Kaaret, P., Chakrabarty, D. 2007, ApJ, 654, L73 Galloway, D. K., Muno, M. P. 2007, ApJ, in press (arXiv:astro-ph/0608259) Kaaret, P., Zand, J. J. M. i., Heise, J., & Tomsick, J. A. 2003, ApJ, 598, 481 in’t Zand, J. J. M., Cornelisse, R., & Cumming, A. 2004, A&A, 426, 257 Leahy, D. A.,Darbro, W.,Elsner, R. F., Weisskopf, M. C., Kahn, S., Sutherland, P. G., & Grindlay, J. E. 1983, ApJ, 266, 160 Liu, Q. Z., van Paradijs, J., & van den Heuvel, E. P. J. 2006, A&A, 455, 1165 Ortolani, S., Barbuy, B., Bica, E. 1994, A&AS, 108, 653 Payne, D. J. B., Melatos, A. 2006, ApJ, 652, 597 Pooley, D., Lewin, W. H. G., Verbunt, F., Homer, L., Margon, B., Gaensler, B. M., Kaspi, V. M., Miller, J. M., Fox, D. W., & van der Klis, M. 2002, ApJ, 573, 184
– 12 – Strohmayer, T. E. & Bildsten, L. 2003, http://xxx.lanl.gov/abs/astro-ph/0301544 Strohmayer, T. E. & Brown, E. F. 2002, ApJ, 566, 1045 Strohmayer, T. E. & Markwardt, C. B. 2002, ApJ, 577, 337 Strohmayer, T. E., Zhang, W., Swank, J. H., Smale, A., Titarchuk, L., Day, C., & Lee, U. 1996, ApJ, 469, L9 Wijnands, R. 2004, Nucl. Phys. B, 132, 496
A This preprint was prepared with the AAS L TEX macros v5.0.