Detection of possible burst oscillation in the neutron star low-mass X-ray binary 4U 1323-62

Abstract: Burst oscillations observed during thermonuclear X-ray bursts arise from asymmetric brightness patterns on the neutron star surface and provide a direct probe of the neutron star spin frequency. We present a detailed timing analysis of the neutron star low-mass X-ray binary 4U 1323-62 using 2024 observations with XMM-Newton and NuSTAR observatories. We identify nine thermonuclear X-ray bursts in the XMM-Newton/EPIC-pn data, along with eclipsing dips in the light curve. One of the XMM-Newton bursts exhibits a rare doublet structure. In addition, NuSTAR detects six bursts, four of which occur simultaneously with those observed with XMM-Newton. We identify a possible burst oscillation signal at $\sim$611.5 Hz in the XMM-Newton data. The strongest oscillation, detected during the primary burst of the doublet burst, reaches a maximum $Z_1<sup>{2}$ power of $\sim35$. An analytical estimate accounting for the searched frequency and time intervals gives a significance of $\sim3.0σ$, whereas independent Monte Carlo simulations yield a more robust global significance of only $\sim2.4σ$. We therefore interpret the signal as a tentative detection of burst oscillation. The folded pulse profile in the 0.5-10 keV band is well described by a sinusoid, with a fractional rms amplitude of $\sim30\pm6$%. The oscillation frequency corresponds to a neutron-star spin period of $\sim$1.635 ms, suggesting that 4U 1323-62 may harbor a rapidly rotating millisecond neutron star.

• The paper reports a tentative 611.5 Hz burst oscillation during the decay of a doublet burst, corresponding to a possible neutron-star spin period of approximately 1.635 ms.
• Using coordinated 2024 XMM-Newton and NuSTAR observations, the authors identified 15 thermonuclear bursts and found the signal with an approximately 30% fractional rms amplitude in one XMM-Newton burst.
• Monte Carlo testing yielded a conservative global significance of about 2.4σ, so the candidate requires confirmation through additional bursts and independent high-cadence observations.

The neutron star low-mass X-ray binary 4U 1323-62 (XB 1323-619) has long been known as a dipping, burst-producing source with a poorly constrained neutron-star spin. Mandal & Naik present a timing analysis of a coordinated 2024 XMM-Newton (135 ks, EPIC-pn timing mode) and NuSTAR (90 ks) campaign, reporting 15 thermonuclear (Type I) X-ray bursts—nine with XMM-Newton, six with NuSTAR, four detected simultaneously—and a tentative burst oscillation at approximately 611.5 Hz, which, if confirmed, would identify 4U 1323-62 as a millisecond pulsar with a spin period of about 1.635 ms (2608.14010).

Observations and burst sample

The XMM-Newton observation was conducted in EPIC-pn timing mode on 7 August 2024, with source events extracted from a 12-pixel strip centered at RAWX = 36 and barycenter-corrected using the JPL DE405 ephemeris and the Gambino et al. source position. NuSTAR data from both FPM modules were processed with standard NUSTARDAS pipelines. The XMM-Newton bursts have peak count rates of 220–270 counts s⁻¹, while the NuSTAR bursts peak at 140–170 counts s⁻¹.

A notable morphological result is the doublet burst (burst B3) in the XMM-Newton data, in which the primary and secondary peaks are separated by approximately 20 minutes; the secondary peaks at about 170 counts s⁻¹. Doublet bursts are rare, previously reported in only a handful of sources (EXO 0748-67, 4U 1636-536, 4U 1608-522), and are generally interpreted as incomplete fuel consumption in the primary burst followed by re-ignition of residual material at greater depth. The authors note that such short recurrence times are difficult to accommodate within standard ignition models, since the interval is insufficient for significant fresh fuel accumulation. The primary peak of this doublet was not covered by NuSTAR, which constrains the broadband characterization of the key event.

The burst oscillation search

Because no spin frequency had been reported for 4U 1323-62, the authors performed a blind search over 100–900 Hz using both Leahy-normalized FFT power spectra and the $Z_1^2$ (and $Z_2^2$) statistic computed directly from photon arrival times, with sliding-window dynamic power spectra (4 s window, 0.25 s step, 0.25 Hz resolution). A signal was found only during the decay phase of the primary burst of doublet B3, at approximately 611.5 Hz, with a maximum $Z_1^2$ power of approximately 34.4–35. The same signal is recovered with an independent 3 s sliding window and in a fixed 4 s segment, where the Leahy-normalized FFT and $Z_1^2$ analyses give consistent powers of 37.6–38.5. The $Z_2^2$ statistic did not exceed $Z_1^2$, indicating negligible harmonic content and a sinusoidal pulse profile.

The significance estimates deserve careful attention, because they differ by method:

Method	False-alarm probability	Significance
Analytical, trial-corrected (4 s window)	$p_{\rm multi} \approx 1.9 \times 10^{-3}$	~2.9σ
Analytical, trial-corrected (3 s window)	$p_{\rm multi} \approx 1.5 \times 10^{-3}$	~3.0σ
Monte Carlo, 50,000 realizations, 100–900 Hz	$p_{\rm emp} = 8.9 \times 10^{-3}$	~2.4σ

The Monte Carlo procedure preserves the burst morphology and Poisson noise properties, redistributing photon arrival times according to the observed count-rate profile and applying the identical search pipeline; the observed maximum power of 35 lies in the extreme tail of a simulated distribution peaking at $Z_1^2 \approx 26.9$. The authors explicitly acknowledge that the overlapping sliding windows introduce correlations, so the analytical trial correction—assuming approximately 5.4 × 10⁴ independent frequency-time trials—likely overestimates the number of independent tests. The more conservative empirical global significance of approximately 2.4σ is the basis for their classification of the result as a tentative detection rather than a firm one. This is the central epistemic claim of the paper, and it is stated with appropriate caution.

Oscillation properties

The fractional rms amplitude is estimated independently by two methods that agree well: from the noise-corrected signal power ($Z_2^2$0 over 317 photons) via the standard $Z_2^2$1 prescription, yielding $Z_2^2$2; and from a sinusoidal fit to the folded profile ($Z_2^2$3 counts s⁻¹, $Z_2^2$4), yielding $Z_2^2$5. This amplitude is high for the burst decay phase. In the standard picture, the burning layer should have spread azimuthally by late times, erasing the rotational modulation; persistent late-time oscillations are instead usually attributed to hydrodynamic instabilities or surface ocean modes. The authors concede that limited photon statistics could partly drive the large amplitude, and the decay-phase detection itself is atypical relative to the more common rise-phase oscillations associated with spreading hot spots.

No significant frequency drift is observed during the burst, unlike the few-Hz drifts seen in 4U 1728-34, 4U 1636-536, KS 1731-260, and X 1658-298, which are attributed to angular momentum conservation in the expanding and contracting burning layer. A stable frequency is consistent with a rotational modulation origin, but with only one burst and modest significance this cannot be tested rigorously.

Relation to previous work and interpretation

The 611.5 Hz candidate is not the first from this source: Bilous & Watts reported two RXTE-era candidates at 415 and 656 Hz with comparable powers (35.9 and 32.2). The present frequency differs from both, so the new detection does not confirm those earlier candidates, and the multiplicity of sub-threshold candidates across instruments underscores the difficulty of establishing a spin for this source. If the 611.5 Hz signal is real, it would place 4U 1323-62 among the rapidly rotating accreting neutron stars, comparable to 4U 1608-522 (~620 Hz) and below 4U 1820-30 (~720 Hz).

The association between the oscillation and the doublet morphology is suggestive but not established. The authors note that complex burst profiles (doublets, double-peaked bursts) in sources such as 4U 1636-536 and 4U 1705-44 have coincided with burst oscillations, plausibly because non-uniform flame spreading or stalling fronts create the surface brightness asymmetries required for detectable modulation. They also emphasize the broader empirical fact that burst oscillations are intermittent: detected in only 1 of 14 bursts in 4U 1916-053, 3 of 15 in 4U 1636-536, and a single NICER burst out of 16 in 4U 1730-22. The single-burst detection in 4U 1323-62 is consistent with this pattern but also limits the physical inferences that can be drawn.

Limitations and open questions

The paper is candid about its statistical footing. The global significance of approximately 2.4σ from Monte Carlo simulations is well below the conventional discovery threshold, and the analytical 3.0σ estimate rests on an independence assumption the authors themselves identify as invalid for overlapping windows. The detection is confined to a single XMM-Newton burst; NuSTAR did not cover the relevant event, so there is no independent instrumental confirmation. The decay-phase occurrence and the high (~30%) fractional rms amplitude are both anomalous relative to the standard hot-spot spreading picture, and photon statistics may be a contributing factor. The relationship of the doublet burst to the oscillation mechanism remains unclear, as does the question of why the secondary burst of the doublet shows no oscillation. Finally, the paper leaves open whether the 611.5 Hz signal corresponds to the true spin frequency or to a pattern speed offset from it, a distinction that cannot be resolved without either a coherent pulsation detection or additional burst detections at the same frequency.

Conclusion

This work provides the most sensitive timing search for burst oscillations in 4U 1323-62 to date, exploiting soft X-ray coverage below 3 keV that prior RXTE, EXOSAT, AstroSat, and NuSTAR studies lacked. It reports a rare doublet burst and a tentative ~611.5 Hz oscillation with a ~30% rms amplitude during the decay of the doublet's primary peak, implying a candidate millisecond spin period of ~1.635 ms. The result is appropriately labeled tentative: the empirical global significance is ~2.4σ, the detection rests on one burst in one instrument, and both the decay-phase occurrence and the large amplitude challenge the conventional interpretation. Confirmation will require additional high-cadence observations of future bursts from this source, ideally with simultaneous coverage across instruments.