Broad-band X-ray Observations on Thermonuclear (Type I) X-ray Bursts by Insight-HXMT

CHEN Yu-peng
Institute of High Energy Physics, Chinese Academy of Sciences, Beijing 100049

Abstract

Since its successful launch into orbit in 2017, the Hard X-ray Modulation Telescope (HXMT), also known as Insight-HXMT, has been operating continuously for over eight years, accumulating a vast amount of observational data. During its observations of multiple thermonuclear burst sources, it has detected more than 200 thermonuclear burst events. With its outstanding wide energy band detection capability (1–250 keV) and large effective area (>5000 cm² at 20 keV), the Insight-HXMT satellite has conducted in-depth analyses of the energy spectra and time-varying characteristics of thermonuclear bursts, particularly focusing on the radiation features in the hard X-ray band (>20 keV), and systematically revealing the interaction mechanisms between thermonuclear bursts and the accretion environment. Specific research achievements include: the first observation of the high-temperature corona cooling process caused by a thermonuclear burst in a single event, providing direct evidence for the study of the interaction between thermonuclear bursts and the corona; the first discovery and confirmation of the correlation between the anisotropy of surface radiation of neutron stars and the accretion rate, offering important clues for understanding the physical processes on the surface of neutron stars; and systematic studies on the enhanced accretion radiation effect triggered by thermonuclear bursts, as well as the obscuring effect of the accretion disk on thermonuclear bursts. These achievements not only expand our understanding of the physical processes of thermonuclear bursts but also provide new observational perspectives for the study of neutron star accretion systems.

Key words X-rays: binaries, bursts, stars: neutron, telescopes: Insight-HXMT (Insight Hard X-ray Modulation Telescope)

1. Introduction

Stars derive their energy from nuclear fusion, while celestial bodies can also influence their surroundings through gravity. The process of attracting surrounding material via gravitational forces and releasing gravitational energy is called accretion, and accretion physics is used to explain the outburst phenomena observed in X-ray binaries. In X-ray binaries where the compact object is a neutron star, there exists an observational phenomenon whose energy source is nuclear fusion—namely, thermonuclear X-ray bursts, also known as Type I X-ray bursts. Thermonuclear bursts are unstable nuclear fusion events of material (mostly hydrogen or helium) accreted onto the neutron star surface, manifesting as sudden, intense eruptions superimposed on the outburst light curve. The peak luminosity can reach the Eddington limit within seconds, with spectra generally resembling blackbody radiation. The burst timescales range from seconds to minutes, with fast rise and slow decay profiles, accompanied by spectral softening during the flux decline phase. Theoretically, the brightest thermonuclear bursts reach the Eddington luminosity, where radiation pressure can cause photospheric radius expansion (PRE). The radiating area corresponds to the neutron star's surface area, allowing constraints on the neutron star's mass, radius, distance, and spin based on observed bursts. Simultaneously, thermonuclear bursts interact with the surrounding accretion environment, enabling their use as probes to study accretion disks, coronae, and other accretion structures.

Since the first detection of thermonuclear bursts in 3A 1820–30 in 1975, 120 Galactic X-ray binaries have been observed to produce thermonuclear bursts as of November 15, 2024. These bursts are primary observational targets for X-ray telescopes such as RXTE (Rossi X-ray Timing Explorer), NICER (Neutron star Interior Composition Explorer), and Insight-HXMT. The recurrence time of thermonuclear bursts ranges from hours to days, with the ratio of accretion energy to thermonuclear burst energy release during this interval being 40–200. Assuming no outflow and that all accreted material participates in thermonuclear bursts on the neutron star surface, this value represents the ratio of gravitational energy to nuclear energy conversion efficiency, which has long served as a criterion for identifying thermonuclear bursts and determining the composition of nuclear fusion material (as hydrogen and helium have different nuclear energy conversion efficiencies), distinguishing them from Type II X-ray bursts. Type II X-ray bursts have similar light curve profiles but shorter intervals between bursts (as short as tens of seconds) and non-thermal spectra, often showing a reduction in continuum emission before and after bursts, forming dip-like structures similar to binary outbursts, generally attributed to accretion instabilities. Currently, only two sources in the Galaxy have been observed to exhibit Type II X-ray bursts: Rapid Burster (MXB 1730−335) and Bursting Pulsar (GRO J1744−28), with thousands of Type II bursts detected from these sources. The former has been observed to produce thermonuclear bursts, while the latter has not, possibly due to its stronger magnetic field (~5.3 × 10¹¹ G).

The majority of thermonuclear bursts last approximately 10–100 s. Based on their total burst energy and nuclear energy conversion efficiency, the ignition column density on the surface of a 10 km neutron star is ~10⁸ g cm⁻². Since helium fusion proceeds faster, short rise timescales (~1 s) are generally attributed to helium fusion, while long rise timescales (~10 s) are attributed to hydrogen fusion. After reaching peak flux, the decay timescale is approximately 10–100 s. The peak blackbody temperature can reach 3 keV, decreasing during the decay phase until reaching pre-burst levels. However, some strong sources do not show blackbody temperature decreases at the end of thermonuclear bursts, such as Cyg X–2, possibly due to high accretion rates causing burst photons to be scattered by the corona.

For the brightest thermonuclear bursts, the luminosity can reach the Eddington limit, where radiation pressure blows the neutron star's photosphere away from the surface, causing photospheric expansion. These photospheric radius expansion (PRE) bursts can reach heights of tens of kilometers, accompanied by spectral softening and decreased blackbody temperature. The light curves show single peaks in soft X-ray bands and double peaks in hard X-ray bands, while spectral evolution exhibits double-peaked blackbody temperatures—highest at the beginning and end of photospheric expansion (up to 3 keV) and lowest when the photospheric radius is largest (below 1 keV). PRE bursts constitute approximately 20% of all thermonuclear bursts. Theoretically, with determined accretion composition, the Eddington luminosity depends only on neutron star mass, enabling mass calculations from observed burst flux, while the measured blackbody area allows radius determination, thereby constraining the neutron star's equation of state. Recent studies indicate that burst emission spectra contain not only blackbody components but also radiation related to the accretion environment, which must be subtracted before measuring neutron star parameters.

Theoretically, in neutron star binaries, the accretion disk near the neutron star surface has Keplerian velocities much greater than the neutron star's surface rotation speed, causing accreted material to decelerate and form a transition region at the neutron star equator known as a spreading layer or boundary layer. This region reaches temperatures of several keV and heights of several thousand meters, and together with the corona, is considered the source of non-thermal components in the high/soft (banana) state. For the corona, based on spectral fitting results of neutron star binary outbursts and analogy with black hole binaries, temperatures are generally considered to be several keV to over 100 keV, with optical depths of 1–10. However, its specific morphology remains unclear, with proposed models including a disk corona covering the disk, a spherical corona covering the neutron star surface, a "lamppost" model suspended above the neutron star poles, relativistic jets, or hybrid models.

When analyzing the inner radius of accretion disks using the diskbb model in Xspec, results are affected by absorption along the line of sight, particularly when disk temperatures are low and the telescope's effective area at the lowest energies is limited. Additionally, accretion disk radiation is influenced by non-thermal coronae or boundary layers, requiring corrections to the derived inner disk radius. Since thermonuclear bursts occur on the neutron star surface, they interact with surrounding accretion structures including the disk, corona, and boundary layer. Following the discovery of thermonuclear bursts, theorists in the 1980s predicted these interactions, such as disk reflection of bursts and increased accretion rates caused by burst radiation. However, due to small telescope effective areas of only tens to hundreds of square centimeters at the time, these interactions were not observed until the launch of RXTE nearly 20 years later, which had an effective area of approximately 6000 cm².

Thermonuclear burst and accretion environment interactions include: (1) burst radiation causing coronal temperature decreases while being Compton-scattered by hot electrons in the corona; (2) burst radiation decelerating inner disk material through Poynting-Robertson drag, increasing accretion rates, while viscous processes cannot refill the cleared inner disk region quickly enough, causing decreased accretion flow after bursts; (3) burst radiation pressure blowing away the disk and corona, reducing accretion radiation while changing disk obscuration of burst emission; and (4) burst radiation reflecting off the accretion disk, forming reflection spectra such as iron fluorescence lines.

Observationally, the first evidence for burst-accretion environment interactions was found in superbursts with hour-long durations. While typical thermonuclear bursts last seconds to minutes, intermediate bursts lasting minutes to 40 minutes and superbursts lasting hours are rare long bursts thought to result from sudden fusion of helium or carbon. For bursts lasting tens of seconds, RXTE and Chandra observations of the strongest burst from the accreting millisecond pulsar SAX J1808.4–3658 revealed an additional component beyond neutron star blackbody radiation. This component had the same spectral shape as the persistent emission but with 20 times increased flux during the burst, corresponding to 60% of the burst peak flux. The spectral model, blackbody + fₐ × F_persistent (where fₐ is the accretion increase factor and F_persistent is the persistent spectral shape), indicated a 20-fold accretion rate increase during the burst. This model, known as the fₐ model, is the mainstream explanation for deviations from blackbody spectra during bursts and matches theoretical predictions from 1989 that burst radiation increases accretion rates through Poynting-Robertson drag. Assuming the inner disk touches the neutron star surface, the accretion luminosity increase is L_PR = (8/3)(1-η)L_b, where L_b is burst flux and η is the energy conversion efficiency of matter falling from infinity to the neutron star surface. For a 10 km neutron star radius, η = 0.2, giving L_PR = 0.6L_b, consistent with observed flux increases but without the predicted post-burst accretion rate decrease. An alternative explanation attributes the increased flux to burst radiation reflected by the optically thick disk and then scattered by the optically thin corona, where reflected photons have energies mostly <0.5 keV. Similar soft X-ray excesses during bursts have been found using RXTE, NuSTAR, NICER, AstroSat, and Insight-HXMT data for sources including Aql X–1, 4U 1608–52, 4U 1636–536, 4U 1730–22, and MAXI J1816–195.

From Insight-HXMT observations of approximately ten accreting neutron stars, we have identified a sample of about 200 bursts, as shown in [TABLE:1]. Benefiting from Insight-HXMT's large detection area and broad energy band, we have obtained burst spectral evolution and studied their interactions with the accretion environment.

2. Impact of Burst Radiation on Accretion: Hard X-ray Deficit During Bursts

For thermonuclear bursts lasting tens of seconds, the first evidence for burst-accretion environment interactions came in 2003 when RXTE observed a flux decrease in the 30–60 keV band during a burst from Aql X–1, though with only 2σ significance. In 2005, by combining Chandra and RXTE observations of multiple bursts from GS 1826–238, spectral fitting of 0.5–40 keV burst spectra revealed coronal temperature decreases from 20 keV to 3 keV. In 2012, by stacking dozens of bursts from IGR J17473–2721, we discovered a significant deficit in the 30–50 keV RXTE/PCA light curve with ~50% amplitude. Subsequently, in many burst sources with sufficiently bright hard X-ray persistent emission (Swift/BAT flux >100 mCrab), hard X-ray deficits have been observed during bursts using RXTE, INTEGRAL, NuSTAR, and Insight-HXMT for sources including IGR J17473–2721, Aql X–1, 4U 1636–536, GS 1826–238, KS 1731–260, 4U 1705–44, 4U 1728–34, and 4U 1724–30, as summarized in [TABLE:2].

Pre-HXMT X-ray telescopes had relatively small hard X-ray detection areas, making >5σ deficits difficult to observe in individual bursts, requiring stacking of tens to hundreds of bursts. For example, the hard X-ray deficit in 4U 1636–536 was based on stacking 36 bursts observed by RXTE/PCA. After HXMT's launch, we detected hard X-ray deficits in individual bursts from 4U 1636–536, as shown in [FIGURE:1]. This short thermonuclear burst occurred during the 2018 outburst, showing a 6.2σ deficit in the 40–70 keV band, consistent with previous RXTE/PCA stacking results. Furthermore, we performed the first broad-band spectral fitting of bursts to study their impact on accretion radiation, finding that the hard X-ray deficit appears in the accretion spectrum above 40 keV.

Theoretically, when the accretion corona cools from higher to lower temperatures during bursts, the hard X-ray deficit fraction should be larger at higher energies. For example, assuming a 10% temperature decrease in the radiation region (from 10.8 keV to 9.7 keV), the flux would decrease by 50% in 50–60 keV and 30% in 30–40 keV. This estimate matches observations of MAXI J1816–195. By stacking Insight-HXMT burst light curves, we observed a hard X-ray deficit in 30–100 keV with 15.7σ significance. As expected, the deficit fraction is energy-dependent, being larger at higher energies (40–50 keV). Notably, the deficit fraction saturates at ~50% above 50 keV, possibly indicating another hard X-ray production region—accretion columns at the neutron star poles unaffected by bursts. Similar saturation has been observed in 4U 1636–536, 4U 1608–52, and Aql X–1.

The saturation phenomenon suggests that thermonuclear bursts affect only part of the hard X-ray emission, implying either a multi-layered corona (e.g., sparse and dense components, with bursts affecting only the former) or a jet. Given MAXI J1816–195's nature as an accreting millisecond pulsar, we favor a jet origin for the unaffected hard X-ray component. Unlike 4U 1608–52, MAXI J1816–195 is an accreting millisecond pulsar. Does thermonuclear burst radiation affect the pulsed signal from polar caps? By stacking MAXI J1816–195 hard X-ray burst data—a task difficult with RXTE due to the lack of numerous Type I bursts from bright hard X-ray sources and soft X-ray band dead-time issues—we found that the accretion-powered pulsed signal from polar caps remains detectable during bursts with unchanged intensity and no phase drift. This indicates that pulsed radiation is largely unaffected by bursts, successfully decoupling the two hard X-ray components of accretion radiation: half from the hot corona and half from polar cap pulsed emission (jet). This hard X-ray deficit saturation, also observed in non-pulsating sources 4U 1608–52 and 4U 1636–536, suggests the universality of these two hard X-ray components in accretion radiation.

3. Impact of Accretion Radiation on Bursts: Scattering by the Corona

For accretion radiation in low/hard or high/soft state low-mass X-ray binaries, spectra can be fitted with an absorbed thermal Comptonization model (with diskbb seed photons), described in XSPEC using thcomp (a more precise version of nthcomp) with parameters including optical depth τ, electron temperature kT_e, and scattering/covering fraction f_sc. Since burst photons may also be affected by the accretion region, we investigated whether this model can fit burst spectra. By treating pre-burst emission as background, we used tbabsthcompbb to fit burst spectra, where tbabs accounts for neutral hydrogen absorption along the line of sight and thcomp parameters are fixed at values obtained from persistent emission fitting. This convolution thermal Comptonization model and blackbody model have the same degrees of freedom as the fₐ model but with additional parameters, allowing assessment of contributions from corona-scattered photons and directly emitted photons from the neutron star surface.

During the photospheric expansion phase of bursts from 4U 1608–52 and 4U 1730–22, this model provides the best fit with physically acceptable parameters. We found that the convolution thermal Comptonization model fits as well as the fₐ model but with more degrees of freedom, statistically outperforming the fₐ model during the photospheric expansion phase (with lowest blackbody temperatures), as shown in [FIGURE:3]. In Insight-HXMT burst observations, the convolution thermal Comptonization model successfully fits soft-state burst spectra, assuming constant coronal parameters during bursts with burst photons being corona-scattered. However, for higher-statistics observations such as NICER bursts, although the convolution model fits better than a simple blackbody, it performs worse than the fₐ model, possibly because coronal parameters may vary during bursts (e.g., temperature decreases from tens to a few keV in low/hard states, or changes in optical depth τ and covering fraction f_sc). Such parameter variations, causing density and geometric evolution of the corona during bursts, require broad-band, large-area observations (e.g., joint NICER and Insight-HXMT observations) that are currently lacking.

4. Impact of Accretion Radiation on Bursts: Disk Obscuration and Accretion-Rate-Dependent Burst Radiation Anisotropy

In observations of photospheric expansion burst spectra, the peak flux generally reaches a (local) maximum near the radius peak, as theoretically predicted. However, in PRE bursts from 4U 1608–52 and 4U 1730–22 detected by Insight-HXMT, the burst flux differs between photospheric lift-off and touchdown, with peak flux not coinciding with maximum photospheric radius. The flux deficit during the PRE rise phase is interpreted as disk obscuration of part of the neutron star surface, as illustrated in [FIGURE:4]. Without bursts, the neutron star's lower hemisphere is obscured by the accretion disk, leaving only the upper hemisphere visible; during bursts, the inner disk radius increases, revealing the previously obscured lower hemisphere.

Burst radiation anisotropy should theoretically depend on accretion state, as low/hard state disks are thought to be truncated at larger distances, but no observational evidence supported accretion-state-dependent anisotropy before 2022. Since outbursts in low/hard and high/soft states have different accretion environments (e.g., different inner disk radii), bursts in these states should exhibit different anisotropy degrees: flux deficits during PRE bursts should be observed in high accretion rate states but not in low accretion rate states. This prediction was confirmed by Insight-HXMT observations of 4U 1608–52 bursts. As shown in [FIGURE:5], high/soft state PRE bursts show disk obscuration, while low/hard state PRE bursts do not. However, with only one low/hard state PRE burst, the universality of this conclusion requires further verification.

5. Summary and Outlook

Current Insight-HXMT scientific results on thermonuclear bursts have focused primarily on broad-band spectral analysis, yielding numerous results on burst-accretion environment interactions. Timing results such as burst oscillations and millihertz quasi-periodic oscillations have been observed but remain limited. For example, Insight-HXMT detected 614 Hz burst oscillations in 4U 1608–52 bursts, but with low significance. Over the past two years, radiation-induced difficulties in background estimation and calibration below 2 keV for Insight-HXMT's low-energy telescope have emerged. Following the successful launch and normal scientific operations of the Einstein Probe (EP) satellite, joint observations with EP's follow-up telescope and Insight-HXMT will provide broad-band (0.5–250 keV), high-statistics spectral and timing results, yielding further insights into thermonuclear burst physics.

Acknowledgments: This work utilizes data and software from the Insight-HXMT mission, supported by the China National Space Administration (CNSA) and the Chinese Academy of Sciences (CAS).

References

van Paradijs J, Lewin H G. A&A, 1986, 157: L10
Chen Y P, Zhang S, Zhang S N, et al. ApJ, 2012, 752: 9
Chen Y P, Zhang S, Zhang S N, et al. ApJ, 2013, 777: 9
Chen Y P, Zhang S, Qu J L, et al. ApJ, 2018, 864: 30
Chen Y P, Zhang S, Ji L, et al. ApJ, 2022, 936: L12
Chen Y P, Zhang S, Ji L, et al. JHEAp, 2023, 40: 76
Chen Y P, Zhang S, Ji L, et al. MNRAS, 2024, 531: 2021, 920: 59
Chen Y P, Zhang S, Zhang S N, et al. JHEAp, 2019, 24: 23
Shaposhnikov N, Titarchuk L, Haberl F. ApJ, 2003, 593: L35
Shaposhnikov N, Titarchuk L. ApJ, 2004, 606: L57
In't Zand J J M, Galloway D K, Marshall H L, et al. A&A, 2013, 553: A83
Ballantyne D R, Strohmayer T E. ApJ, 2004, 602: L105
Keek L, Ballantyne D R, Kuulkers E, et al. ApJ, 2014, 797: L23
Degenaar N, Ballantyne D R, Belloni T, et al. SSRv, 2018, 214: 15
Walker M A, Meszaros P. ApJ, 1989, 346: 844
Maccarone T J, Coppi P S. A&A, 2003, 399: 1151
Thompson T W J, Rothschild R E, Tomsick J A, et al. ApJ, 2005, 634: 1261
Chen Y P, Zhang S, Torres D F, et al. A&A, 2011, 510: 9
Ji L, Zhang S, Chen Y P, et al. MNRAS, 2013, 432: 9
Güver T, Bostanci F, Boztepe T, et al. ApJ, 2022, 935: 9
Ji L, Zhang S, Chen Y P, et al. ApJ, 2014, 791: L39
Sánchez-Fernández C, Kajava J J E, Poutanen J, et al. A&A, 2020, 634: A58
Ji L, Zhang S, Chen Y P, et al. A&A, 2014, 564: A20
Kajava J J E, Koljonen K I I, Nättilä J, et al. MNRAS, 2017, 472: 78
Kashyap U, Chakraborty M. MNRAS, 2022, 512: 6180
Peng J Q, Zhang S, Chen Y P, et al. A&A, 2024, 685: 9
Ji L, Ge M, Chen Y, et al. ApJ, 2024, 966: L3
Zdziarski A A, Szanecki M, Poutanen J, et al. MNRAS, 2020, 492: 5234
Chen Y P, Zhang S, Ji L, et al. ApJ, 2022, 936: 46
Galloway D K, Muno M P, Hartman J M, et al. ApJS, 2008, 179: 360
Worpel H, Galloway D K, Price D J. ApJ, 2013, 772: 94
Worpel H, Galloway D K, Price D J. ApJ, 2015, 801: 60
Galloway D K, In’t Zand J, Chenevez J, et al. ApJS, 2020, 249: 32
Bult P, Altamirano D, Arzoumanian Z, et al. ApJ, 2021, 920: 59