Tracing the outburst decay of soft X-ray transients Aql X-1 and 4U 1608-52 with XSPECT

Abstract: XSPECT instrument on-board XPoSat mission is a soft X-ray spectrometer sensitive in the energy band 0.8$-$15 keV. XSPECT has observed several bright neutron star low mass X-ray binaries since launch. Two well known sources, Aql X-1 and 4U 1608-52 which are soft X-ray transients, were observed by XPoSat during the decay phase of their recent outbursts in September 2024 and February 2025 respectively. During XSPECT observations, 4U 1608-52 exhibited a superburst which is a long duration thermonuclear burst, believed to be triggered by carbon burning. We carry out a detailed spectro-temporal analysis of the superburst, tracing its onset, rise, and decay over the next several hours. Using time-resolved spectroscopy, we probe the spectral evolution of the source and find that the persistent emission is suppressed during the superburst and the emission can be described by a gradually cooling blackbody component. The superburst was preceded by a precursor burst which is a normal type-I X-ray burst. We also observe a type-I burst $\sim 5$ days after the superburst, indicating resumption of burst activities which is typically quenched after a superburst. Aql X-1 also exhibited a type-I burst during XSPECT observations. The persistent emission of both the sources can be fitted using a combination of blackbody and disk blackbody emission or, alternatively, using a disk Comptonized by an optically thick plasma. Using the latter model, we find a clear flux dependence of the Comptonization parameters, with both the sources exhibiting harder spectra at higher accretion rates.

• The paper demonstrates XSPECT’s ability to monitor neutron-star transient outbursts, detecting three type-I bursts and completely capturing the third known superburst from 4U 1608-52.
• Both systems remained on the soft banana branch while their spectra hardened at higher flux, driven by increasing coronal optical depth and Comptonization efficiency rather than an additional direct blackbody component.
• The 4U 1608-52 superburst cooled from approximately 1.9 to 0.8 keV, released about 3.8 × 10^42 erg in model estimates, suppressed persistent emission, and was followed by renewed bursting after only 5.4 days.

Overview

This paper presents a spectro-temporal study of two well-known neutron star soft X-ray transients, Aql X-1 and 4U 1608-52, using the XSPECT instrument on board XPoSat. XSPECT is a collimated spectrometer employing swept charge devices over the $0.8-15$ keV band, and its long, quasi-continuous observations of bright sources make it well suited to tracking outburst decay and burst phenomena. The authors triggered Target of Opportunity observations during the decay phases of recent outbursts: Aql X-1 during 2024 October 4–10 and 4U 1608-52 during 2025 March 8–10 and March 17–26. The observations yielded three type-I thermonuclear bursts (one from Aql X-1, two from 4U 1608-52) and — most notably — a complete capture of a superburst from 4U 1608-52 on 2025 March 19–20, only the third known from this source.

Observations and data reduction

The data were reduced with the standard XSPECT pipeline (v2.2.1), applying nominal screening, flare filtering, and event merging per segment. For 4U 1608-52, four detectors (SCDs 4–7) showed optical-light contamination near eclipse ingress/egress and were excluded, leaving 11 of 15 SCDs; all 15 were used for Aql X-1. Spectra were extracted day-wise for persistent emission, fitted in XSPEC over $0.8-11$ keV with optimal binning, a 1% systematic uncertainty, Al/Si absorption edges at $\sim$1.5 and $\sim$1.8 keV, a cross-calibration constant between the two fields of view ($\sim$1.09–1.10), and gain fitting. Hardness-intensity diagrams were constructed with soft color $(2.5-4.5)/(1-2.5)$ keV and hard color $(6-11)/(4.5-6)$ keV. The average count rate decayed from $\sim$70 to 55 cps (Aql X-1) and $\sim$63 to 35 cps (4U 1608-52) over the observation spans, with hardness remaining essentially flat — both sources stayed on the soft (banana) branch throughout, at luminosities of roughly $0.08-0.14\,L_{\rm Edd}$ (4U 1608-52) and $0.8-11$0 (Aql X-1).

Persistent emission spectroscopy

Three continuum models provided statistically equivalent fits ($0.8-11$1): (i) diskbb + bbodyrad ("model 1"), (ii) nthComp, and (iii) thcomp × diskbb ("model 3"). All models required a Gaussian line at $0.8-11$26.7 keV (F-test probability $0.8-11$3), consistent with disk reflection as previously reported by NuSTAR and AstroSat/NICER studies. The authors prefer model 3 because thcomp better reproduces Monte Carlo Comptonization spectra than nthComp.

Under model 1, Aql X-1 shows an inner disk temperature of $0.8-11$40.9 keV with color-corrected inner radius $0.8-11$540 km, plus a blackbody at $0.8-11$61.65 keV with radius $0.8-11$75.5 km — smaller than the neutron star surface. For 4U 1608-52, the disk temperature is $0.8-11$81 keV with $0.8-11$920 km, and the blackbody radius shrinks from $\sim$03.6 to $\sim$12.5 km. Under model 3, seed photon temperatures drop to $\sim$20.55 keV (Aql X-1) and $\sim$30.6–0.7 keV (4U 1608-52), with inferred inner radii of $\sim$4100–120 km and $\sim$530–60 km respectively — illustrating the strong model dependence of disk parameters when Comptonization is included. Both coronae are optically thick ($\sim$6) with Compton $\sim$7 parameters of 1.7–3.1.

A key result concerns flux dependence: for both sources, optical depth and the Compton $\sim$8 parameter increase with flux, while the photon index decreases — i.e., the spectra harden at higher accretion rates. The authors attribute this to enhanced Comptonization efficiency outweighing soft-photon cooling in the boundary layer. Notably, under model 3 there is no statistical evidence for an additional direct blackbody component (improvement p-values of $\sim$915–20%), implying the corona intercepts most seed photons.

Timing results

No QPOs were detected in power density spectra of either the persistent or bursting emission, across the range probed (up to 500 Hz). The three type-I bursts show rise times of 1–2 s and e-folding times of 9.4 s (Aql X-1) and $\sim$06 s (4U 1608-52). Peak fluxes correspond to $\sim$10.9, $\sim$21.1, and $\sim$30.7 $\sim$4 respectively; the first 4U 1608-52 burst reached the Eddington limit, consistent with the published touchdown flux of $\sim$5 erg cm$\sim$6 s$\sim$7, which also supports the adopted distance of $\sim$8 kpc as a conservative upper limit. Ignition column depths are $\sim$9 g cm$\sim$0.

The superburst of 4U 1608-52

XSPECT captured the full superburst despite orbital gaps of $\sim$160–70 min. The event was preceded by a precursor type-I burst at $\sim$223:05 UTC on March 19, peaking at $\sim$3270 cps and reaching the Eddington limit. Rather than decaying to the pre-burst level, the flux remained elevated and then brightened abruptly by a factor of 2 within 7 minutes. The superburst rise could not be fit with a single linear ramp; three piecewise linear segments yield a total rise time of $\sim$413.6 min, with peak at 23:28:37 UTC. The decay has an e-folding time of 3.05 h ($\sim$57 h), with the hard band peaking 2.6 min earlier and decaying faster ($\sim$6 h) than the soft band (3.6 h) — consistent with envelope cooling shifting the blackbody peak to lower energies, analogous to KS 1731-260.

Time-resolved spectroscopy used the pre-burst model scaled by an $\sim$7 factor plus a blackbody component. The blackbody temperature declines monotonically from $\sim$81.9 keV to $\sim$90.8 keV over six intervals, with blackbody radii of $(2.5-4.5)/(1-2.5)$05–7 km. Two findings stand out:

• Persistent emission suppression: unlike normal type-I bursts, where $(2.5-4.5)/(1-2.5)$1 (attributed to Poynting-Robertson drag), $(2.5-4.5)/(1-2.5)$2 throughout the superburst, indicating quenching of the accretion flow. Freeing the disk parameters yields a lower inner disk temperature and larger inner radius, supporting disruption of the inner disk. This mirrors the near-complete Comptonization quenching reported for the 4U 1820-30 superburst, though it contrasts with Keek et al.'s finding of increased persistent flux during the 4U 1636-536 superburst.
• Rapid resumption of bursting: a type-I burst occurred only $(2.5-4.5)/(1-2.5)$35.4 days after the superburst — an order of magnitude shorter than the upper-limit quenching times of 99.8 d (2005) and 58.98 d (2020) previously reported for this source. Under the Cumming & Bildsten framework, this implies heat influx from deeper layers stabilized H/He burning only briefly, suggesting a comparatively shallow or less energetic ignition.

Applying the Keek et al. prescription to the rise, the luminosity evolution shows two slopes ($(2.5-4.5)/(1-2.5)$4 then 0.5), encoding the initial temperature profile $(2.5-4.5)/(1-2.5)$5 with depth. Fitting the decay with the Cumming cooling model gives $(2.5-4.5)/(1-2.5)$6 and $(2.5-4.5)/(1-2.5)$7, consistent with prior superbursts from this source, corresponding to a total energy release of $(2.5-4.5)/(1-2.5)$8 erg and a burnt carbon mass fraction of $(2.5-4.5)/(1-2.5)$920%. However, the directly measured X-ray fluence (%%%%6$0.8-15$6%%%%1 erg cm$(6-11)/(4.5-6)$2, i.e., $(6-11)/(4.5-6)$3 erg at the surface) accounts for only $(6-11)/(4.5-6)$416% of the modeled energy release, with the remainder attributed to neutrino losses and conduction to deeper layers. The X-ray fluence implies an ignition depth $(6-11)/(4.5-6)$53–4 times shallower than the cooling-model value — a tension the paper notes but does not resolve. The superburst released $(6-11)/(4.5-6)$6350 times more X-ray energy than the normal type-I bursts.

Limitations and open questions

Several caveats bear directly on these results. The small effective area of XSPECT prevented fine time-resolved spectroscopy of the type-I bursts, so photospheric radius expansion signatures could not be tested, and hints of a secondary peak in the precursor burst remain unmodeled due to limited statistics. Orbital observing gaps mean the superburst decay-to-pre-burst time ($(6-11)/(4.5-6)$79 h) and the burst-quenching interval are upper limits, and the "pre-burst" baseline is defined only $(6-11)/(4.5-6)$84.5 h before the event because intervening orbits were data-dump passes. The $(6-11)/(4.5-6)$9 method cannot capture changes in spectral shape of the persistent component; freeing the disk parameters instead yields poorly constrained values. The discrepancy with Boztepe et al.'s report of a hotter, closer-in disk after the 2020 superburst remains unexplained beyond appeals to different superburst energetics and sampling. Finally, the physical interpretation of the suppressed persistent emission — whether inner disk disruption, corona modification, or spectral reshaping — cannot be uniquely determined from these data alone.

Conclusion

This work demonstrates XSPECT's capability for pile-up-free, multi-day monitoring of bright NS-LMXBs through outburst decay. The principal contributions are the complete light curve and time-resolved spectroscopy of the third superburst from 4U 1608-52 — showing a cooling blackbody, suppressed persistent emission ($\sim$0), an energetic Eddington-limited precursor, and unusually rapid resumption of type-I bursting — together with evidence that both sources exhibit harder-when-brighter behavior driven by increasing coronal optical depth. The short post-superburst quenching time relative to previous events from the same source raises a specific open question: whether superburst energetics, ignition depth, or accretion state at ignition govern the duration of H/He burning stabilization, a distinction that requires comparable coverage of future events.