Broad-band spectral analysis of LMXB 2S 0921-63 with Suzaku

ABSTRACT

We present the broad-band spectral analysis of the low-mass X-ray binary 2S 0921-63 by using the Suzaku archival data covering the orbital phase between 0.31 and 1.16 during four close observations. It is the first time that a broad-band spectral analysis of 2S 0921-63 has been done up to 25 keV. The 0.5–10 keV XIS count rate varied between ∼1 and ∼5 counts s⁻¹ during the observations. A partial X-ray eclipse and broad post-eclipse intensity dip were observed during the observations. The X-ray emission hardened marginally during the intensity dip. We have modelled the source spectra by simultaneously fitting the XIS and HXD-PIN spectra for each of the four observations. The broad-band spectra of the source can be described by a model comprising a very hot blackbody having temperature, kT_BB ≈ 1.66–2.13 keV, a high-energy cut-off power law, and an Fe emission line at E_line ∼ 6.7 keV. A second model, accounting for the Comptonization of the thermal emission from accretion disc along with an Fe emission line, describes the broad-band spectra of 2S 0921-63 equally well.

1 INTRODUCTION

2S 0921-63 is an eclipsing low-mass X-ray binary (LMXB) discovered with SAS-3 (Li et al. 1978). The low-mass binary companion is an optical star of spectral type K0III (Branduardi-Raymont et al. 1983; Shahbaz et al. 1999). From photometric and spectroscopic studies, Branduardi-Raymont et al. (1981, 1983) and Cowley, Crampton & Hutchings (1982) determined an orbital period of about 9 d. Combined with EXOSAT X-ray observations, Mason et al. (1987) refined the orbital period to 9.0115(5) d. Intensity dips and partial eclipses lasting for about 1.3 d have been observed in the optical (Branduardi-Raymont et al. 1983; Mason et al. 1985, 1987) as well as X-ray energies (Mason et al. 1987).

The nature of the compact object is still unclear, mainly due to the absence of coherent pulsation or type-I X-ray bursts, typical to neutron star (NS) LMXBs. The initial mass estimate of 2.0–4.3 M_⊙ for the primary suggested the possibility of a massive NS or a low-mass black hole in the system (Shahbaz et al. 2004; Jonker et al. 2005). Later, Shahbaz & Watson (2007) and Steeghs & Jonker (2007) refined the primary mass limits to lower values consistent with the canonical NS mass, thus supporting the idea of an NS primary.

2S 0921-63 exhibits an X-ray to optical flux ratio (L_X/L_opt) of the order of 1, unusually less compared to L_X/L_opt > 100 for other sources (Mason et al. 1985, 1987). The possibility of an extended X-ray source obscured from direct view by an accretion disc at high inclination was proposed to explain the observed partial eclipses and low ratio of X-ray to optical flux (Mason et al. 1985; Krzeminski & Kubiak 1991).

The 0.6–10 keV spectrum of 2S 0921-63 obtained with data from the Einstein satellite was described with a thermal bremsstrahlung model having kT ∼ 13 keV, moderately absorbed by a column density, N_H ∼ 5.7 × 10²¹ cm⁻² (Branduardi-Raymont et al. 1983). The spectral analysis for the 1985 Einstein data suggested the softening of spectrum during the eclipse (Mason et al. 1987). The spectrum showed similarity to the unsaturated Comptonized spectra, typical to other bright sources (e.g. Sco X-1, Sco X-2, GX 17 + 2, X 1705-440; White, Peacock & Taylor 1985; White et al. 1986). These results were inconclusive due to contamination of data at low energies.

The ASCA spectrum showed significant hardening compared to the low-inclination LMXBs along with the presence of Fe emission line at 6.75 keV (Asai et al. 2000). Kallman et al. (2003) used the high-resolution data from Chandra and XMM and found the emission lines from O, Ne, Mg, Si, S, and Fe in the source spectrum.

In this paper, we present the broad-band spectral study of LMXB 2S 0921-63 by analysing the Suzaku archival data obtained in 2007 August. We aim to study the properties of the entire spectrum covering an energy range of 0.5–25 keV.

2 OBSERVATIONS

The X-ray Imaging Spectrometer (XIS; Koyama et al. 2007) onboard the fifth Japanese X-ray observatory Suzaku (Mitsuda et al. 2007) comprises of four units numbered from 0 to 3. The three units are front-illuminated CCDs (XIS0, XIS2, and XIS3), and one is a back-illuminated CCD (XIS1). XIS provides an energy coverage in the range of 0.2–12 keV. The Hard X-ray Detector (HXD-PIN; Takahashi et al. 2007) extends the energy range to about 60 keV. Suzaku observed 2S 0921-63 four times between 2007 August 23 and 2007 August 30 in normal modes and collected data in 3 × 3 and 5 × 5 pixel format (except for the second observation where only 3 × 3 mode data were present) for a combined exposure of 172.4 ks by using the XIS. HXD-PIN collected data for a net exposure of 125.8 ks. Table 1 lists the details of the observations used in this work. We have utilized the data from XIS0, XIS1, and XIS3 detectors in our analysis. We do not use XIS2 data owing to damage due to micro-meteorites in 2006.

2.1 XIS spectra

The unfiltered XIS event files have been screened and filtered utilizing the task aepipeline provided with the heasoft version 6.29a. We used the calibration database (CALDB) version 20181010 on processed XIS event files for the calibration. We used the xselect tool to extract the product files for each CCD individually, by combining the 3 × 3 and 5 × 5 pixel format cleaned event files. For each XIS detector, we used a circular region of 180 arcsec centred at the source to extract the source spectra. A similar-sized circular region was used to extract the background spectra situated away from the source photons. We generated the response files for each XIS detector by using the xisrmfgen tool. We used source coordinates RA =140.6444° and Dec. =63.2948° (Gaia Collaboration 2020) to produce the ancillary response files by using the script xissimarfgen.

The spectra from all the detectors for each observation have been re-binned for at least 20 counts per energy bin by using the ftool grppha. We used the spectra from all three XIS detectors for the simultaneous fitting in the energy range 0.5–10 keV of all four observations. We removed the energy bins between 1.7 and 2.3 keV to avoid the calibration uncertainties due to the artificial structures around Si and Au edges (Jiang et al. 2021).

2.2 HXD-PIN spectra

We have utilized the HXD-PIN data extending the energy coverage above 10 keV for the first time for the source. We extracted the HXD-PIN spectra for each observation following the standard procedure described in the Suzaku ABC Guide.¹ The raw event files were processed by using the HXD CALDB version 20110913. We generated the HXD-PIN spectra for each observation by using the hxdpinxbpi routine. The task uses pseudo-events files distributed with the archival data and performs dead time correction to the spectrum. We used observation-specific tuned PIN background event files to account for the contribution from the non-X-ray background and cosmic X-ray background (CXB). The hxdpinxbpi task simulates and models the CXB contribution and generates the total background spectrum following the model given by Boldt (1987). We used the response file ae_hxd_pinxinome4_20080129.rsp for HXD-PIN spectral fitting specific to the epoch of the observation.

HXD-PIN detected the source up to 25 keV above the background during all four observations (Fig. 1). However, due to the low statistics below 13 keV, we ignored the energy bins below 13 keV and have used the 13–25 keV energy range for the spectral analysis.

2.3 Light curves

The XIS and HXD-PIN light curves in the respective energy bands of 0.5–10 and 13–25 keV are shown in Fig. 2. The top and middle panels show the XIS and HXD-PIN light curves with corresponding hardness ratio in the bottom panel. The observations cover the orbital phases of 0.31–0.44, 0.56–0.68, 0.86–0.98, and 1.04–1.16, respectively, based on the ephemeris of Ashcraft et al. (2012) with phase zero at the centre of the eclipse. X-ray flux is variable across the four epochs. XIS count rate varies from 2.86 ± 0.03 to 3.74 ± 0.02 counts s⁻¹ between the starting and ending of the first observation while the HXD-PIN count rate remains constant at ∼0.3 counts s⁻¹.

During the second epoch, count rates increase relatively and remain considerably constant at ∼4 and ∼0.32 counts s⁻¹ for XIS and HXD-PIN, respectively. The orbital phase for the third epoch coincides with the eclipse of the primary X-ray source. Gradual decrease and a dip near the end of the observation are evident from the XIS light curve. The XIS count rate gradually decreases from 2.86 ± 0.03 to 2.40 ± 0.03 counts s⁻¹ and then to 2.00 ± 0.04 counts s⁻¹ near the dip. It then recovers to 2.92 ± 0.05 counts s⁻¹ near the end of the third observation. HXD-PIN collected data only for the first 15 ks and did not record any significant variation in the count rate near the dip.

The last epoch shows significant variation with a broad dip in the XIS energy band consistent with the previously reported dips (Mason et al. 1987; Krzeminski & Kubiak 1991). A marginal variation in the HXD-PIN count rate is also evident, resulting in the increased hardness near the centre of the dip. Similar hardening during the intensity dips has been reported for several eclipsing (e.g. X 2127 + 119, Ioannou et al. 2002; XTE J1710-281, Younes, Boirin & Sabra2009) and dipping sources (e.g. 4U 1323-62, Boirin et al. 2005; X 1254-690, Díaz Trigo et al. 2009). For a clearer picture, we generated a hardness intensity plot by using the light curves to compare the variation during different observation epochs (Fig. 3). While the source emission was variable during the span of the observations, though the variation in the hardness ratio was not significant. Thus, we used the time-averaged spectra from each observation for our analysis.

3 RESULTS

We have used xspec (Arnaud 1996) v12.12.0 for simultaneous spectral fitting of XIS and HXD-PIN spectra for each observation individually. We introduced a cross-normalization factor to account for the different instrumental calibrations. We fixed its value at 1 for the XIS0 and XIS3 data sets and left it free for XIS1. The cross-normalization factor was fixed at 1.158 for HXD-PIN (Kokubun et al. 2007). We have adopted the updated photoionization cross-sections of Verner et al. (1996) and solar abundances by Wilms, Allen & McCray (2000) for our spectral analysis. We tied all the model parameters across XIS0, XIS3, XIS1, and HXD-PIN spectra for spectral fitting.

The X-ray spectrum of 2S 0921-63 has been studied only up to 10 keV. The spectrum is generally modelled with an absorbed simple power law or cut-off power law along with emission features from different elements (Mason et al. 1987; Kallman et al. 2003). Following this, we tried to fit the spectra by using the high-energy cut-off power-law model. We included the tbabs component to account for the absorption medium. This model failed to provide an adequate fitting to the spectra with large systematic residuals below 1.5 keV and between 6 and 7 keV for all four epochs. The addition of a Gaussian component (gaus) at 6.7 keV gave a statistically better fit. But a large residual below 1.5 keV still existed along with a narrow feature near 7 keV.

We then used a two-component model, typical for LMXBs, i.e. a soft thermal component with a power law component. We added the bbody (Mitsuda et al. 1984) component to the existing model to account for any contribution from the NS. It improved the fittings significantly for all the four epochs. The best-fitting returned blackbody temperature between 1.7 and 1.8 keV, photon index in the range 1.6–1.7, cut-off energy (E_cut) ∼3 keV, and Fe emission line energy ∼6.7 keV. Although the new model provided a reasonable good fitting to the spectra, some features could not be modelled. Some residual persisted around 0.7, 1, and 7 keV. While we identified the latter ones as possible absorption edges from Ne (1.196 keV) and Fe (7.117 keV), we used a Gaussian absorption component to model the 0.7 keV feature in the spectra from all epochs except the third epoch, where it was not statistically favoured. Contrary to our claim of an absorption edge at 1.196 keV, Kallman et al. (2003) identified the 1.17 keV feature in Chandra and XMM data as the Ne x Ly β emission line. But due to the limited resolution of Suzaku data and model uncertainties, it is not possible to determine the actual nature of this feature.

We, thus, define tbabs*egde*edge*(bbody + powerlaw*highecut + gaus  + gaus) in xspec as our first model, M1. The model provided an acceptable fitting to the spectra with a χ²/ν of 5986.3/5643, 5694.7/5606, 5317.1/5367, and 5473.5/5304 for the four observations, respectively. Fig. 4 shows the best-fitting unfolded spectra for the first epoch with model M1 and its components. The lower panels show the residuals with respect to model M1 of the four observations. All the best-fitting parameters are reported in Table 2.

Our attempts to model the spectra using a cut-off power law instead of highecut power law failed as the fitting resulted in substantial residuals near the lower and higher energy end of the spectra. When we replaced bbody component with the multicolour accretion disc blackbody, diskbb, the fitting returned unreasonably high disc temperatures ∼3 keV with very low disc normalization values thus, we did not include the results for these models in our analysis.

The typical two-component model successfully modelled the broad-band Suzaku spectra of 2S 0921-63. From literature, it is known that the spectrum of 2S 0921-63 can be described as an unsaturated Comptonized spectrum, typical to X-ray sources (e.g. Sco X-1, Mason et al. 1987; Sco X-2, GX 17 + 2, Krzeminski & Kubiak 1991).

Utilizing the broad-band coverage of Suzaku, we check for the Comptonized emission from the source. We fit the spectra by using the thermal Comptonization model nthcomp (Zdziarski, Johnson & Magdziarz 1996; Życki, Done & Smith 1999) in xspec. We included two Gaussian components and two edges to the model as before and set the seed photon source to NS blackbody. The model provided a reasonable fitting to the spectra however the fits returned low values of kT_seed (∼0.2 keV) and a large emission radius (>40 km). We then changed the seed photon source to the accretion disc. This model, tbabs*edge*edge*(nthcomp[dbb] + gaus  + gaus) (M2, hereafter) provided an acceptable fitting to the spectra of all four observations. The best-fitting spectra and residuals are shown in Fig. 5, with derived parameters listed in Table 2. We have computed the unabsorbed bolometric flux in 0.1–100 keV for the model and components by using the convolving model cflux.

4 DISCUSSION

We have analysed the data of LMXB 2S 0921-63 obtained with Suzaku in 2007 August. We have performed a broad-band spectral analysis of the source by using about 172 ks of data. We have used the XIS data covering 0.5–10 keV for this purpose. We have also, for the first time, extended the energy coverage for the spectral analysis to 25 keV by using the HXD-PIN (13–25 keV) data. We have modelled the spectra with a two-component model and have used a Comptonization model to describe the spectra.

We have presented the XIS (0.5–10 keV) and HXD-PIN (13–25 keV) light curves covering the orbital phases between 0.31 and 1.16. The light curve covers the partial eclipse (0.86–0.98) and includes the post-eclipse dipping phase. While the eclipse is not completely covered by HXD-PIN, it is evident from the hardness ratio plot that the emission hardens marginally during the dip.The smooth shape of the dip during the last observation and its proximity to the predicted eclipse makes it difficult to establish if the intensity dip during the third observation is the original eclipse. Similarly, Krzeminski & Kubiak (1991) reported the presence of many identical eclipses in the light curve of 2S 0921-63 during one cycle. Until a new X-ray orbital ephemeris is established, it will be difficult to identify the real partial eclipse confidently.

We have modelled the broad-band spectra of 2S 0921-63 by using a single temperature blackbody to check for the presence of NS blackbody emission (Lyu et al. 2014; Zhang et al. 2016) and a high-energy cut-off power-law component to account for the non-thermal contribution. For all four observations, we obtained a high blackbody temperature of |$1.85_{-0.06}^{+0.08}$|⁠, 1.94 ± 0.04, |$2.05_{-0.07}^{+0.08}$|⁠, and 1.72 ± 0.06 keV, respectively. We interpret this hot blackbody component as the NS surface/boundary layer between disc (Popham & Sunyaev 2001; Cackett et al. 2010; Sharma et al. 2020). The best-fitting photon indices decrease, suggesting that the spectra harden from the first observation towards the last observation (Table 2). The cut-off energy (E_cut) attains a low and constant value ∼3 keV during all observations except for the third observation, where it falls significantly to |$1.04_{-0.06}^{+0.07}$| keV. The low value of E_cut suggests a low corona temperature (Zhang et al. 2016). The best-fitting confirmed a low neutral column density with 0.27 ± 0.02, 0.30 ± 0.02, |$0.15_{-0.06}^{+0.03}$|⁠, and 0.29 ± 0.02 × 10²² cm⁻² for four observations, respectively. These values are in agreement with the previously reported values of column density (Mason et al. 1987; Kallman et al. 2003).

The spectra of LMXBs are often described with the Comptonization model as it provides more information than a simple power-law description (Iaria et al. 2005; Sakurai et al. 2012; Zhang et al. 2014, 2016; Sharma et al. 2020). We used the broad-band coverage of Suzaku and modelled the entire spectrum of 2S 0921-63 with a Comptonization model for the first time. A single-component disc-seed Comptonization model provided a satisfactory good fitting to the spectra of all four observations. The photon index values agreed with the model M1 results favouring a hard spectral state. A low corona temperature, kT_e ∼ 2 keV, is again consistent with the low E_cut value from model M1. We obtained a low seed photon temperature of 0.20 ± 0.06 and 0.20 ± 0.03 keV for the first two observations with an upper limit of 0.39 and 0.41 keV for the last two observations. For the Comptonizing medium, we have estimated a relatively high optical depth of ∼19 (Zdziarski et al. 1996). This high value is possibly due to the Comptonized component being close to the Wien spectrum (Cackett et al. 2010). Similar low corona temperature of the order ∼ 2–3 keV with high optical depth (11–13) have been reported for LMXB 4U 1758-20 (White et al. 1986).

The size and location of the Comptonized component is not known convincingly. Some theories claim an inner compact region (≤100 km) of the binary (Kluzniak & Wilson 1991; Popham & Sunyaev 2001) while others support a large extended region (≈50 000 km) (White & Holt 1982; Church 2001; Church & Bałucińska-Church 2004).

For a clearer picture of the ADC and Comptonization region, we estimated the size for the seed photon emission region by using the equation from in ’t Zand et al. (1999). From the first two observations, we obtained a seed photon region of ≈45 and 50 km and a lower limit of 10 km from the last two observations. Clearly, the size of the Compton seed region is consistent with the source of seed photons being the inner disc region.

Moreover, a high value of optical depth (∼19) implies that the Comptonized component originates from an optically thick corona extending over the accretion disc regions at least up to ∼50 km. A similar corona geometry has been reported for the ADC sources X 1624-490 (Iaria et al. 2007) and 4U 1822-371 (Anitra et al. 2021). For high-inclination systems, a flat Comptonizing corona geometry can explain the large optical depths as up-scattered photons travel longer path along the line of sight (Zhang et al. 2014, 2016).

Following the radius–luminosity relation for ADC, |$r_{\rm ADC}\ = \ L_{\rm X}^{0.88}$| by Church & Bałucińska-Church (2004), we estimate r_ADC ≈ 10⁹ cm. Again, using their equation (3) for radius r_c, for corona with temperature kT_e to remain in hydrostatic equilibrium gives, r_c ∼ 10⁹ cm. Thus, both the estimates agree with ADC being extended and covering the disc up to distance of ∼10⁹ cm. However, the Comptonization component that we observe in the spectra dominantly originates from the inner, optically thick region of ADC. Based on these calculations, Fig. 6 shows a schematic diagram for a clearer picture of this ADC source.

The presence of broad Fe emission lines is reported often in LMXBs (Bhattacharyya & Strohmayer 2007; Cackett et al. 2010; Miller et al. 2013; Degenaar et al. 2014). We found a broad emission line around 6.7 keV in the spectra of all four observations. We used a Gaussian component to model this feature with both the models M1 and M2. The best-fitting line energy for the feature is consistent with the highly ionized Fe xxv (He-like) emission line. The origin of Fe emission lines in LMXBs is generally attributed to the recombination in the ionized matter around the NS (White et al. 1985, 1986; Hirano et al. 1987; Kitamoto et al. 1987). Kallman et al. (2003) also reported the presence of Fe xxv emission line along with neutral Fe and Fe xxvi lines in the spectra of 2S 0921-63. We have estimated upper limits of 11, 17, 13, and 11 eV for the neutral Fe line and 48, 36, 69, and 76 eV for Fe xxvi line on the equivalent widths, respectively for the four observations. While the equivalent width of the Fe xxv emission line is high compared to the previously reported value, the value is consistent across the four observations.

We observed softening of the emission during the third observation covering the eclipse phase. The blackbody temperature increased as compared to the first two observations and reached about 2 keV. While the photon index and cut-off energy decrease, the thermal blackbody emission dominates with about ∼60 per cent contribution to the total emission flux. During the last observation, the blackbody temperature decreases to ∼1.72 keV with a hard photon index of ∼1.72 and cut-off energy ∼3.8 keV. The non-thermal component dominates with a contribution of ∼67 per cent to the total flux. Typical of the variation in the light curve, the unabsorbed bolometric flux in 0.1–100.0 keV was maximum during the second observation and minimum during the highly variable last observation. Net flux during the third and fourth observations show a significant decrease as compared to the first two epochs, consistent with the eclipse and dip.

We have found an absorption-like feature around 1 keV in the spectra from all observations. We used an absorbed Gaussian component to model this feature around 0.7–0.8 keV in all the spectra except for the third observation, where it was not statistically favoured. Similar dips near 1 keV have been reported for other dipping sources (e.g. XB 1916-053, EXO 0748-676, X 1254-690, MXB 1658-298, Díaz Trigo et al. 2006; XTE J1710-281, Younes et al. 2009). It is possible that the feature arises from the blending of several lines and edges produced by the ionized absorber around the source. It could also be due to the known contamination of the XIS detectors, primarily because of the Oxygen below 2 keV. A similar feature was found in the Suzaku spectra of XTE J1710-281 around 0.6 keV (Sharma et al. 2020). We also found an absorption edge at 7.1 keV, implying no or very low ionization of the absorbing matter (Singh & Apparao 1994). Such features may arise from the reflection of hard X-rays off the cold and optically thick material from the outer regions of the accretion disc (Gondoin et al. 2001).

Although the partial covering models did not provide an acceptable fitting to the current Suzaku data, the presence of intensity dips does not rule out the possibility of an additional medium around the primary source. Future targeted observation of 2S 0921-63 near dipping phases may prove useful in understanding its properties better.

ACKNOWLEDGEMENTS

We have used the data obtained with Suzaku, a JAXA/ISAS space mission, and used software provided by the High Energy Astrophysics Science Archive Research Center (HEASARC) online service maintained by the NASA Goddard Space Flight Center. We thank the anonymous referee for the useful suggestions that have helped in improving the results. PS acknowledges the financial support from the Council of Scientific & Industrial Research (CSIR) under the Senior Research Fellowship (SRF) scheme.

DATA AVAILABILITY

The underlying work has used the data obtained with Suzaku and is available at the High Energy Astrophysics Science Archive Research Center (HEASARC) online service. The archival data can be accessed at https://heasarc.gsfc.nasa.gov/cgi-bin/W3Browse/w3browse.pl.

Footnotes

REFERENCES