A thousand looks at X Persei: X-ray spectroscopy at high time resolution

Abstract: X Persei is a classical Be/X-ray binary composed of an O9.5III-B0V Be-type star and a neutron star (NS). The NS exhibits coherent pulsations with a spin period of ~837 s, orbiting the Be star with a ~250 d period. X Persei is notable for its exceptionally hard X-ray emission extending beyond 100 keV. Due to the mild eccentricity of the orbit, ~0.11, the orbital separation varies between roughly 35 R_star at periastron and 44 R_star at apastron. In this work, we analyze five targeted observations obtained with the XMM-Newton and Chandra observatories taken over a 10-year time span, with the aim of investigating the structure and variability of the circumstellar disk surrounding the Be star, in particular the presence of over-dense areas known as clumps. We performed spectral and timing analyses, including average and NS spin-resolved spectroscopy for three of the observations, producing individual spectra at intervals as short as 210 s, corresponding to different epochs of the NS spin, resulting in approximately 1200 spectra. This detailed analysis aimed at resolving emission-line features otherwise diluted in averaged spectra and the evolution of continuum components along the NS spin. The observed spectra are accurately modeled by a two-component continuum comprising a high-temperature blackbody and a power-law component. Phase-resolved spectroscopy reveals transient Fe K alpha emission linked to clumps in the circumstellar disk during high-density epochs, with an 8-9% prevalence. Dips in the X-ray light curve are tied to clump passages. The disk-density modeling, based solely on X-ray data, suggests a compact and dense disk, with a radial density exponent alpha of 2.4-3.3, and a high inner disk density, rho_0 ~ (6-20) x 10^-10 g cm^-3, in agreement with previous studies conducted in the optical and infrared bands.

• The paper demonstrates that 210-second X-ray spectra reveal transient neutral Fe Kα emission in only 8–9% of spectra, tracing dense, intermittently illuminated clumps in X Persei’s circumstellar disk.
• The analysis combines Chandra and XMM-Newton observations, spin-resolved spectroscopy, dip and spike detection, disk modeling, and photoionization calculations to connect absorption variability with clumpy accretion and Rayleigh–Taylor instabilities.
• The paper finds a dense inner decretion disk with inferred densities of roughly (6–20) × 10⁻¹⁰ g cm⁻³ and a ∼3 × 10⁴-second spin-torque modulation, supporting quasi-spherical settling accretion onto a highly magnetized neutron star.

The system and the motivation

X Persei (4U 0352+309) is a wide, low-eccentricity Be/X-ray binary (BeXB) hosting an O9.5III–B0V donor and a neutron star (NS) with a spin period of $\sim$837 s and an orbital period of $\sim$250 d ($e \approx 0.11$, $i \simeq 26^\circ$–$33^\circ$), placing the NS at roughly 35–44 stellar radii from the donor. The source is notable for its exceptionally hard spectrum extending beyond 100 keV, a broad $\sim$27 keV feature attributed to direct cyclotron emission rather than a CRSF, and torque arguments favoring a magnetar-strength field of order $10^{14}$ G. The paper's objective is to probe the structure and clumpiness of the circumstellar disk using X-ray data alone, exploiting the fact that the slow spin and high brightness permit spectra to be extracted on timescales as short as one quarter of the NS spin cycle (~210 s).

Observations and methods

Five archival observations spanning a decade are analyzed: three Chandra datasets (HRC-S/LETG, ACIS-I, ACIS-S/LETG) and two XMM-Newton EPIC-pn observations. Pile-up in the XMM data was mitigated by excising central pixels, leaving the pn camera as the sole usable detector. Spectral fitting was performed in ISIS with TBnew absorption, a partial-covering local absorber (covering fraction CF used as a clumping proxy), a powerlaw plus blackbody continuum, and Gaussian lines. Line significance was assessed with Monte Carlo false-alarm probabilities from 1000 line-free simulations per observation. Feature detection in light curves used the publicly available dipspeaks package, which builds synthetic light curves matching the noise properties to flag genuine dips and spikes.

Average spectroscopy

All five spectra are well described by the two-component continuum ($\chi^2_r \approx 1.1$–1.2). The local column density varies strongly between epochs, from $$N_{\rm H}^{\rm LOC} \approx 1.3\times10^{21}\,\mathrm{cm^{-2}}$$ (Chandra-2, the lowest-luminosity state) up to $7.4\times10^{23}\,\mathrm{cm^{-2}}$ (Chandra-3). Luminosities span roughly two orders of magnitude, from %%%%1$\sim$1%%%%1 to $\sim$2$\sim$3, with the blackbody emitting radius ranging from 0.16 km (low state) to 1.34 km. The consistency between the fitted blackbody normalizations and those implied by the adopted Gaia DR3 distance of 0.69 kpc supports both the distance and the hot-spot interpretation. The authors note that the Gaia parallax-based distance carries a re-normalized unit weight error of 1.5, slightly above the recommended acceptance threshold of 1.4 — a caveat on all luminosity-derived quantities.

Spin-resolved spectroscopy and transient Fe K$\sim$4

Dividing each observation into four NS spin phases yields about 1200 individual spectra. Both continuum components pulsate with the spin, peaking when the hot spot faces the observer; pulse fractions reach 0.61 (hard band) and component-level pulse fractions up to 0.5 for the powerlaw. This phase locking of both components points to a common origin near the NS surface or accretion column.

The central result is the detection of transient Fe K$\sim$5 emission in short (210 s) spectra that is absent from phase-averaged spectra. In the XMM-1 and XMM-2 datasets, 11 and 35 features respectively fall within the Fe K$\sim$6 window, against 5 and 3 in line-free simulations; binomial tail probabilities are $\sim$7 and $\sim$8. When detected, the line has remarkably stable properties: centroid $\sim$96.40–6.42 keV, equivalent width $e \approx 0.11$070 eV, flux %%%%2$\sim$2%%%%2 photons s$e \approx 0.11$3 cm$e \approx 0.11$4, with a prevalence of only 8–9%. The implication is that neutral, dense material is illuminated only intermittently — consistent with clumps whose recombination times ($e \approx 0.11$5–30 s for $e \approx 0.11$6) wash out in averaged spectra but occasionally surface in short exposures.

Dips, spikes, and accretion phenomenology

Using simultaneous soft- and hard-band criteria, the analysis identifies 20, 3, 203, and 0 high-probability dip (clump candidate) events in XMM-1, Chandra-2, XMM-2, and Chandra-3, corresponding to rates of 3, 0.2, 6, and 0 clumps per hour. Spikes in the hard band — interpreted as Rayleigh–Taylor instability (RTI) entries of matter into the magnetosphere — number 42, 10, 310, and 32 (rates of 5, 0.7, 9, and 1 per hour). The spike rate increases monotonically with the fluxes of both pulsating continuum components, although the authors concede that five observations do not establish a statistically robust correlation.

Disk structure from X-ray absorption alone

A viscous decretion disk model is folded through the system geometry to predict the line-of-sight column density, which is then matched to the measured $e \approx 0.11$7 via particle swarm optimization. The parameters $e \approx 0.11$8 and $e \approx 0.11$9 are strongly degenerate, a limitation the paper states explicitly. Within priors informed by optical/IR work ($i \simeq 26^\circ$0 between $i \simeq 26^\circ$1 and $i \simeq 26^\circ$2), the fits yield $i \simeq 26^\circ$3–3.3 and $i \simeq 26^\circ$4–$i \simeq 26^\circ$5, i.e., a dense inner disk of $i \simeq 26^\circ$6–$i \simeq 26^\circ$7. Notably, Chandra-2 — the lowest-luminosity epoch — shows the steepest $i \simeq 26^\circ$8 ($i \simeq 26^\circ$9), suggesting a more compact disk configuration during low states. These values agree broadly with Telting et al.'s optical/IR inference of a dense inner disk, though that study derived a much steeper exponent ($33^\circ$0); the discrepancy is partly attributable to differing assumed inclinations (nearly pole-on versus $33^\circ$1).

Combining the disk model with XSTAR photoionization calculations, the authors reconstruct where Fe K$33^\circ$2 can form. XMM-1 and XMM-2 show extended regions satisfying the emission conditions, whereas Chandra-3 does not — matching the observational absence of the line there. The co-occurrence of Fe K$33^\circ$3 detections, clump candidates, and higher covering fractions in XMM-1/XMM-2, against their joint absence in Chandra-3, forms a self-consistent picture in which clumps are rapidly ionized by the NS radiation field, producing short-lived neutral-emission windows.

Spin-period modulation and quasi-spherical settling accretion

A sliding-window Lomb–Scargle analysis reveals a quasi-sinusoidal variation of the spin period around $33^\circ$4 s with characteristic time $33^\circ$5 s, implying torque reversals with $33^\circ$6 and specific torque $33^\circ$7$33^\circ$8. The authors interpret this within the quasi-spherical settling accretion framework: at $33^\circ$9, a hot shell forms above the magnetosphere ($\sim$0 cm for $\sim$1 G), and matter enters via RTI. The observed timescale matches the magnetic diffusion time $\sim$2: with spike durations averaging ~600 s (range 120–1600 s) as the RTI time, $\sim$3 and $\sim$4 give $\sim$5 s, as observed. This agreement is presented as quantitative support for the settling-accretion picture at equilibrium, though it rests on order-of-magnitude estimates of the turbulence parameters.

Limitations and open questions

Several caveats bear directly on the results. The disk-fit parameters suffer from intrinsic $\sim$6–$\sim$7 degeneracy, so the quoted ranges should be read jointly rather than independently. The Fe K$\sim$8 reconstruction assumes a smooth disk, neglects time-dependent ionization, and uses simplified geometry, so the emission maps are qualitative. The spike–flux trend lacks statistical significance given only five epochs. The distance uncertainty (RUWE above the Gaia acceptance criterion) propagates into all luminosity-based quantities. Finally, the spin-modulation interpretation depends on assumed values of the settling factor and eddy-size parameter; whether these hold across luminosity states remains untested. A specific open question is whether the quasi-sinusoidal torque modulation persists at higher accretion rates, where radiative cooling would shorten $\sim$9.

Conclusion

By resolving X-ray spectra on 210 s timescales, this work demonstrates that transient Fe K$10^{14}$0 emission — present in only 8–9% of short spectra yet statistically robust — traces clumpy material in the Be disk at the NS orbital distance, a signature invisible in conventional averaged spectroscopy. The X-ray-only disk modeling yields a dense, compact decretion disk consistent with optical/IR constraints, and the quasi-sinusoidal spin-period modulation on a $10^{14}$1 s timescale provides quantitative support for quasi-spherical settling accretion onto a slowly rotating, highly magnetized neutron star.