\LETTERLABELNinjaSat monitoring of Type-I X-ray bursts from the clocked burster SRGA J144459.2$-$604207

Type-I X-ray bursts (hereafter, X-ray bursts) are explosive transients in low-mass X-ray binaries (LMXBs), characterized by recurrent rapid increases in luminosity by order of magnitude, with emission continuing for several seconds to hours. These energetic bursts are triggered by nuclear burning in the accreting layer of a neutron star (NS) (for a review, see [Galloway & Keek (2021)]). Key observables of X-ray bursts, such as rise time, peak flux, duration, and recurrence time, depend on the properties of the accreted matter, particularly the mass accretion rate and composition (e.g., [Galloway et al. (2008)]). Among over 115 known X-ray bursters (Galloway et al., 2020), only six sources exhibit notably regular burst recurrence times, called clocked bursters. GS 1826$-$24, which was first identified by Ginga X-ray satellite (Makino, 1988), is a prominent example of clocked bursters of which the periodic features were observationally confirmed by BeppoSAX and RXTE (e.g., Ubertini et al. (1999); Galloway et al. (2004)).

Clocked bursters are valuable for constraining theoretical models, as well as understanding the properties of the NS and companion star (Heger et al. (2007); Meisel (2018); Dohi et al. (2020, 2021); Hu et al. (2021); Lam et al. (2022); Dohi et al. (2022)). Due to their regular burst recurrence times, one can investigate the relationship between the mass accretion rate and the physical conditions on the NS surface. The relation between the burst recurrence time $\Delta t_{\rm rec}$ and the persistent flux $F_{\rm per}$ with a constant $C$:

serves as a powerful diagnostic. Assuming a critical fuel mass for burst ignition, a power-law index $\eta=1$ is derived, where the persistent flux $F_{\rm per}$ is proportional to the mass accretion rate. Previous studies have found $\eta$ values ranging from approximately 1 to values greater than 1.¹¹1For instance, $\eta=1.05\pm{0.02}$ for GS 1826$-$24 (Galloway et al., 2004), $\eta=0.95\pm{0.03}$ for MXB 1730$-$335 (Bagnoli et al., 2013), $\eta\gtrsim 1$ for IGR J17480$-$2446 (Linares et al., 2012), and $\eta>1.35$ for 1RXS J180408.9$-$342058 (Dohi et al., 2024a). Theoretical studies using various X-ray burst models (e.g., Lampe et al. (2016); Dohi et al. (2024a)) suggest that $\eta$ depends on the physical properties of binary systems, although this remains under-explored. To further constrain the $\Delta t_{\rm rec}$–$F_{\rm per}$ relation, long-term monitoring of X-ray bursts is essential.

Due to their flexibility, CubeSats operating within the nanosatellite framework are well-suited for such long-term monitoring of X-ray bursts. NinjaSat, a 6U-size ($10\times 20\times 30~{}{\rm cm}^{3}$) CubeSat X-ray observatory, was launched into a sun-synchronous orbit at an altitude of approximately 530 km on 2023 November 11 (Enoto et al., 2020; Tamagawa et al., 2024). Equipped with two Xe-based proportional counters (Gas Multiplier Counters; GMCs) covering the 2–50 keV energy range, NinjaSat’s total effective area of approximately 32 cm² at 6 keV is over two orders of magnitude larger than the X-ray detectors on previously launched CubeSats (Takeda et al., 2023). This significant area allows for monitoring of persistent flux and detection of X-ray bursts, even within a CubeSat mission. NinjaSat enables agile follow-up observations of bright X-ray transients and facilitates extended long-term monitoring.

X-ray bursts from the new LMXB SRGA J144459.2$-$604207 (hereafter, SRGA J1444) are suitable targets for NinjaSat. SRGA J1444 was discovered in outburst by the Mikhail Pavlinsky ART-XC telescope onboard the Spectrum Roentgen Gamma (SRG) observatory on 2024 February 21 (Molkov et al., 2024), and later confirmed by MAXI (Mihara et al., 2024) and Swift (Chandra, 2024). Prompt follow-up observations by NICER discovered the pulsation at 447.9 Hz and X-ray bursts, along with an orbital period at 5.22 h, identifying SRGA J1444 as an accretion-powered millisecond X-ray pulsar (Ng et al., 2024). X-ray bursts from SRGA J1444 were also reported by Swift (Mariani et al., 2024), MAXI (Negoro et al., 2024), INTEGRAL (Sanchez-Fernandez et al., 2024a), Insight-HXMT (Li et al., 2024), and IXPE (Papitto et al., 2024). A quasi-periodic burst occurrence ranging from $\sim 1.7~{}{\rm hr}$ to $\sim 2.9~{}{\rm hr}$ was initially recognized in the INTEGRAL observations (Sanchez-Fernandez et al., 2024b), thus identifying SRGA J1444 as the sixth clocked burster. Subsequent IXPE observations reported the increase of recurrence time up to $\sim 7.9~{}{\rm hr}$ and the decrease of the persistent X-ray intensity, resulting in a significant low power-law index in equation (1), with $\eta\sim 0.8$ (Papitto et al., 2024). NinjaSat also observed SRGA J1444 and detected 12 X-ray bursts—the first detection of X-ray bursts by a CubeSat (Takeda et al., 2024).

In this letter, we report on the long-term monitoring campaign of SRGA J1444 conducted with the newly launched CubeSat X-ray observatory NinjaSat. Throughout this letter, all errors are given at the 1$\sigma$ confidence level unless otherwise specified.

The GMC is a non-imaging gas X-ray detector equipped with a space-proven gas electron multiplier (Tamagawa et al., 2009). Each GMC fits into a compact 1U-size (10 $\times$ 10 $\times$ 10 cm³) with a mass of 1.2 kg, making it suitable for CubeSats. The time-tagging resolution for each X-ray photon is 61 $\mu$s. NinjaSat is capable of conducting the pointing observation of X-ray sources with an accuracy of less than $0\fdg 1$, utilizing an X-ray collimator with a $2\fdg 1$ field of view (full-width at half-maximum) equipped with each GMC.

Because of the high charged-particle background, we limit the GMC operation to a low-background region, with latitudes between $\sim 30^{\circ}$ south and $\sim 40^{\circ}$ north, excluding the South Atlantic Anomaly. Currently, the astronomical operational area covers $\sim$ 37% of the total, although the observation efficiency for each source is also affected by the Earth occultation and the battery charging of the satellite. In the payload commissioning phase, we observed the Crab Nebula for the detector calibration and successfully detected the pulsation at 33.8 ms, confirming that the absolute time is correctly assigned to each X-ray photon with an accuracy of at least sub-milliseconds (Tamagawa et al., 2024).

NinjaSat observed SRGA J1444 for one day on 2024 February 23 (MJD 60363) and started a monitoring campaign on February 26 (MJD 60366), which continued through 2024 March 18 (MJD 60387) until the end of the outburst. The observation was conducted with one GMC (GMC1), whose detector calibration was more advanced at the end of the initial observation phase. We also observed the Crab Nebula before and after the monitoring campaign of SRGA J1444, from 2024 February 23 to February 26 and on March 19. The detector calibration with the Crab Nebula indicates that using background data from the same observation period is more appropriate than using blank sky data from different periods under the current background model. Therefore, we estimated the background level using data from the SRGA J1444 observation period when the satellite was not pointed at either SRGA J1444 or the Earth. We got effective exposures of 197.5 ks, 104.7 ks, and 16.6 ks for SRGA J1444, background data, and the Crab Nebula, respectively. Photon arrival times were corrected to the solar system barycenter with FTOOLS barycen using the DE-405 planetary ephemeris with source coordinates R.A. $=221\fdg 24558$, Decl.$=-60\fdg 69869$ obtained by Chandra (Illiano et al., 2024).

The main background component in the GMC data is the non-X-ray background, which includes charged particle events and electrical noise events. The GMC has two readout pads: a circular inner readout electrode with a radius of 25.0 mm and an annular outer electrode with radii between 25.1 mm and 33.5 mm. Signals from each readout pad are digitized by a 12-bit analog-to-digital converter at a sampling rate of 25 MHz, followed by onboard analysis to extract waveform parameters—such as pulse height, rise time, and the Pearson correlation coefficient between the inner and outer waveforms $R$—which are subsequently downlinked. Charged particle events entering perpendicular to the readout pad can be rejected because their signal rise time ($\sim$ 1 $\mu$s) is longer than that of X-ray events ($\sim 400$ ns) due to the difference in the distributed length of the electron cloud along the drift direction. On the other hand, events from the parallel direction leave signals on both pads, resulting in a relatively high correlation coefficient $R$. This allows them to be distinguished from X-ray events, except when X-rays enter between the pads. Additionally, the event cut based on the correlation coefficient $R$ is also useful for the common-mode electrical noise simultaneously triggered in both channels, which typically has a value of $R\sim 0.95$.

In this letter, we only use the X-ray count rate for the following analysis without the response for the spectral discussions. The persistent count rate in the 2–10 keV band is converted to X-ray intensity in units of mCrab, i.e., a flux referenced to the count rate of the Crab Nebula. Because the detector calibration is still ongoing, we employed a tentative event selection criterion: selecting events with signal rise times in the range of 200–800 ns and correlation coefficients $R$ of less than 0.4. The average raw background rates of the inner and outer regions in the 2–10 keV band are 6.0 counts s^-1 and 10.6 counts s^-1, respectively. After applying the event cut, the rates are reduced to 0.294 $\pm$ 0.003 counts s^-1 and 2.26 $\pm$ 0.02 counts s^-1, respectively. The event cut based on the correlation coefficient $R$ between the inner and outer waveforms is sensitive to events that are simultaneously triggered in both regions. Consequently, it functions similarly to the anti-coincidence method, making it more effective in the inner region, which is surrounded by the outer region. Thus, we analyzed only event data of the inner region for the persistent flux evaluation to obtain a higher signal-to-noise ratio, while we used event data extracted from both regions for the X-ray burst analysis. The background-subtracted rate of the Crab Nebula in the 2–10 keV band is estimated to be 11.89 $\pm$ 0.03 counts s^-1. The time variation in the 3.0-hr binned background light curve shows a slight periodicity at approximately 1 and 6 days and follows a Gaussian distribution with a 1$\sigma$ width of 0.025 counts s^-1, corresponding to 2.5 mCrab.

To search for X-ray bursts, we extracted 10-s binned light curves from all screening data and clearly detected 12 X-ray bursts (IDs 1–12), as listed in table 3. These bursts exhibit a peak X-ray intensity of $\sim 1$ Crab and lasting $\sim$ 20 s (Takeda et al., 2024), consistent with observations reported by NICER and SRG (Ng et al., 2024; Molkov et al., 2024).

Figure 1 shows the persistent light curve of SRGA J1444 in the 2–10 keV band observed with NinjaSat, compared with that from MAXI (Matsuoka et al., 2009). The 2–10 keV X-ray intensity observed with MAXI was calculated using publicly available data.²²2http://maxi.riken.jp/top/slist.html The outburst of SRGA J1444 reached the maximum X-ray intensity at $\sim$ 100 mCrab at MJD 60361 and then gradually decayed for nearly 30 days to the background level. The NinjaSat monitoring campaign covered almost the entire outburst decay phase until MJD 60387. The flux evolution observed with NinjaSat is consistent with that of MAXI, achieving comparable statistical errors with 3.0-hr bins to those of the MAXI daily light curve. NinjaSat enables us to track finer variations in the persistent X-ray intensity.

In this letter, we present the long-term monitoring of SRGA J1444 conducted with NinjaSat over a period of approximately 25 days. We found that SRGA J1444 exhibited X-ray bursts with a fast rise time of $<$5 s and a short duration of $\tau=18.1\pm 1.2$ s (IDs 1–11), the latter of which is consistent with the values derived by the spectral analysis from other satellites, such as IXPE ($\tau=16.8\pm 1.6$ s, Papitto et al. (2024)) and SRG ($\tau\sim 16$ s, Molkov et al. (2024)), with an accuracy of $\sim$10%. The fast rise time and short duration are characteristic features of sources with relatively He-rich accreted fuel. This is because He burns during the burst via the triple-$\alpha$ reaction, which proceeds on a much shorter time scale than the H burning via the hot-CNO cycle, $rp$, and $\alpha p$ processes. The observed burst duration in SRGA J1444 is longer than in pure-He bursters but shorter than sources with the solar composition (figure 5). Given that no photospheric radius expansion burst has been observed from SRGA J1444, including with observations from other satellites, it is reasonable to assume that SRGA J1444 has a relatively He-enhanced accreted fuel compared to the solar composition. The He-enhanced scenario with $X/Y=1.5$ and $Z_{\rm CNO}=0.015$ is further supported by the theoretical predictions from the HERES model (Dohi et al., 2024b).

We show that the recurrence time in SRGA J1444 is roughly inversely proportional to the persistent X-ray intensity with a power-law index $\eta=0.84^{+0.02}_{-0.01}$. We also noted that $n_{i}$ are closely matched to integer values within $\pm$5%, except for $n_{10}$. The deviations from integers could be attributed to the average variation in $C$ between each detected burst and/or to limitations of the simplified burst model expressed in equation (3). Contributing factors may include incomplete observational coverage, potential variations in persistent flux on timescales of 3.0 hr or shorter, which corresponds to the time resolution of the light curve (figure 1), and long-term spectral variations associated with the outburst. Our results are consistent with the IXPE observation reported by Papitto et al. (2024), which shows $\eta\sim 0.8$ with recurrence time deviations ranging from a few % to roughly 10%⁵⁵5The variation of the recurrence time is reflected in the deviation of $n_{i}$ from their nearest integer values in our formulation expressed in equation (3). (see also Fu et al. 2024).

The estimated index $\eta=0.84^{+0.02}_{-0.01}$ of SRGA J1444 is the lowest value observed among X-ray bursters (section 1). Dohi et al. (2024a) investigated the $\eta$ dependence on the equation of state (EOS) and NS masses in the range of $1.1M_{\odot}$ to $2.0M_{\odot}$. They found that compacted NS models tend to have lower values of $\eta$.⁶⁶6Note that their results were obtained for the case of H-rich accreted fuel, unlike the He-enhanced scenario we expected. However, because no significant correlation between the power-law index $\eta$ and the composition of the accreted fuel was found (Lampe et al., 2016), their conclusion holds. The value of $\eta=0.84^{+0.02}_{-0.01}$ for SRGA J1444 does not match the model predictions for NS masses below $2.0M_{\odot}$, suggesting that it is a more massive NS. A comparison of observed value $\eta$ in SRGA J1444 with theoretical models that encompass a broader range of masses above $2.0M_{\odot}$ could provide valuable insights for constraining the mass and the EOS of NSs.

The maximum observed value of the burst recurrence time can be used to constrain the composition of the accreted fuel. Both the NinjaSat and IXPE observations indicate that the $\Delta t_{\rm rec}$–$F_{\rm per}$ relation with $\eta\sim 0.8$ remained valid up to at least $7.9$ hr. This implies that during this period, H burned stably via the hot CNO cycle between each burst without depletion, followed by a mixed H/He burst (Case 1; Fujimoto et al. (1981)). In this bursting regime, the accreted H is depleted in time of $t_{\rm CNO}=9.8\ {\rm hr}\ (X/0.7)(Z_{\rm CNO}/0.02)^{-1}$ (Lampe et al. (2016)). Using the maximum recurrence time of $\Delta t_{\rm rec}=7.9$ hr observed in SRGA J1444, we obtain a constraint 7.9 hr $\leq(1+z)t_{\rm CNO}$, where $1+z$ represents the gravitational redshift at the photosphere. Assuming a solar hydrogen abundance of $X=0.74$, along with a canonical NS with a mass $M_{\rm NS}=1.4M_{\odot}$ and the radius $R_{\rm NS}=11.2$ km (giving $1+z=1.259$), we derive an upper limit for the CNO mass fraction as $Z_{\rm CNO}\leq 0.033$. Similarly, for a massive NS with $M_{\rm NS}=2.0M_{\odot}$ and the radius $R_{\rm NS}=11.2$ km, the upper limit becomes $Z_{\rm CNO}\leq 0.038$. Note that for the He-enhanced scenario with $X/Y=1.5$ and $Z=0.015$, as inferred by Dohi et al. (2024b), $t_{\rm CNO}$ is estimated to be 11.0 hr, satisfying the constraint $7.9~{}\text{hr}<(1+z)t_{\rm CNO}$ regardless of the NS mass.

This study demonstrated that CubeSat pointing observations can provide valuable astronomical X-ray data. Even a compact detector, with an effective area of several tens of cm² onboard a CubeSat, can successfully observe both burst and persistent emissions. Given the rarity of X-ray bursts, which occur during less than 1% of the total observation time (Galloway et al., 2020), CubeSats offer a complementary approach to large canonical observatories, providing long-term, flexible observations critical for detecting these transient events.

This project was supported by JSPS KAKENHI (JP23KJ1964, JP17K18776, JP18H04584, JP20H04743, JP20H05648, JP21H01087, JP23K19056, JP24H00008, JP24K00673). T.T. was supported by the JSPS Research Fellowships for Young Scientists. T.E. was supported by “Extreme Natural Phenomena” RIKEN Hakubi project. N.N. received support from the RIKEN Intensive Research Project (FY2024–2025).