Spectral properties of MXB 1658–298 in the low/hard and high/soft state

ABSTRACT

1 INTRODUCTION

Low mass X-ray binaries (LMXBs) consist of a neutron star (NS) or a black hole (BH) with a low mass (⁠|${\lesssim } 1 \, \mathrm{M}_{\odot}$|⁠) companion orbiting around it. The compact object accretes matter from its companion and the accretion rate is usually inferred from the position of the sources in their X-ray colour–colour diagrams (CDs) or hardness–intensity diagrams (HIDs; Lewin & van der Klis 2006). The low-luminosity sources trace out an atoll-like shape, with the different branches referred to as the banana branch (high inferred accretion rate) and the island state (low inferred accretion rate; Hasinger & van der Klis 1989). They have luminosity in the range of 0.001–0.5L_Edd (van der Klis 2006).

The NS LMXBs display spectral states and hysteresis pattern similar to that observed in BH binaries, indicating similarity between these two kinds of sources (Muñoz-Darias et al. 2014). NS spectral emission is composed of three main components: disc blackbody from the accretion disc and Comptonization component due to Comptonization of soft X-rays in the corona as of the BH spectra (Done, Gierliński & Kubota 2007), with the addition of a blackbody component for emission from the NS surface or boundary layer (Lin, Remillard & Homan 2007, 2009; Armas Padilla et al. 2017). During the soft state, the emission is usually dominated by the thermal components, with characteristic temperatures of kT ∼ 0.5–2.0 keV (Bloser et al. 2000a; Sakurai et al. 2012; Armas Padilla et al. 2017). The Comptonized component is weak with low temperature and large optical depth. In the hard state, the spectra are dominated by a hard Comptonized component with temperatures of a few tens of keV and low optical depths. The thermal components are observed at low temperatures (kT < 1 keV) and with significantly lower luminosity (Raichur, Misra & Dewangan 2011; Sakurai et al. 2012; Zhang et al. 2016; Armas Padilla et al. 2017).

Narrow absorption features from highly ionized Fe and other elements have been observed in large number of X-ray binaries. In particular, Fe xxv (He-like) or Fe xxvi (H-like) absorption lines near 6.6–7.0 keV have been reported from several NS LMXBs (e.g. Sidoli et al. 2001; Díaz Trigo et al. 2006; Hyodo et al. 2009; Ponti, Muñoz-Darias & Fender 2014; Ponti et al. 2015; Raman, Maitra & Paul 2018) that are almost all viewed at fairly high inclination angles; most of them being dippers (Boirin et al. 2004, 2005; Díaz Trigo et al. 2006). This indicates that the highly ionized plasma probably originates in an accretion disc atmosphere or wind, which could be a common feature of accreting binaries but primarily detected in systems viewed close to edge-on (Díaz Trigo & Boirin 2013, 2016).

MXB 1658–298 is a transient atoll NS LMXB system (Wijnands et al. 2002) which shows eclipses, dips, and thermonuclear bursts in its X-ray light curves. It was discovered in 1976 with the third Small Astronomy Satellite (SAS-3; Lewin et al. 1976). It has an orbital period of ∼7.1 h and its eclipse lasts for ∼15 min (Cominsky & Wood 1984). This active phase lasted for about 2 yr. The second outburst was observed in 1999–2001 (in ’t Zand et al. 1999b) during which coherent burst oscillations at frequency of ∼567 Hz were reported (Wijnands, Strohmayer & Franco 2001). The 0.5–30 keV BeppoSAX spectrum obtained from observations made during this outburst was modelled by a combination of a soft disc blackbody and a harder Comptonized component (Oosterbroek et al. 2001). The residuals to this fit suggested the presence of emission features due to Ne–K/Fe–L and Fe–K in the spectrum. The observations made with XMM–Newton revealed the presence of narrow resonant absorption features due to O viii, Ne x, Fe xxv, and Fe xxvi, together with a broad Fe emission feature (Sidoli et al. 2001). The third outburst from the source was detected in the year 2015 (Bahramian, Heinke & Wijnands 2015; Negoro et al. 2015), which lasted for ∼1.5 yr (Parikh et al. 2017). The evolution of the orbital period over about 40 yr of data indicates the presence of a circumbinary planet around this binary system (Jain et al. 2017).

In this work, we report results of spectral study of the persistent emission of MXB 1658–298 by using the Swift/X-Ray Telescope (XRT) and Nuclear Spectroscopic Telescope Array(NuSTAR) observations made during 2015 and 2016. We also report the detection of state-dependent absorption features due to highly ionized materials.

2 OBSERVATIONS AND DATA REDUCTION

The NuSTAR (Harrison et al. 2013) mission features two telescopes, focusing X-rays between 3 and 79 keV on to two identical focal planes (usually called focal plane modules A and B, or FPMA and FPMB). During its 2015–2017 outburst, MXB 1658–298 was observed twice with NuSTAR. The observation details are given in Table 1. We have used the most recent NuSTAR analysis software distributed with heasoft version 6.20 and the latest calibration files (version 20170120) for reduction and analysis of the NuSTAR data. The calibrated and screened event files have been generated by using the task nupipeline. A circular region of radius 100 arcsec centred at the source position was used to extract the source events. Background events were extracted from a circular region of same size away from the source. The task nuproduct was used to generate the light curves, spectra, and response files. Using grppha, the spectra were grouped to give a minimum of 200 counts bin⁻¹. The FPMA/FPMB light curves were background corrected and summed using lcmath. The persistent spectra were extracted by excluding the dips, eclipses, and bursts.

MXB 1658–298 was also monitored by Neil Gehrels Swift Observatory (Gehrels et al. 2004) during its 2015–2017 outburst. For this work, we have focused only on those observations (listed in Table 1) that coincided with the NuSTAR observations. The Swift/XRT (Burrows et al. 2005) data were analysed using standard tools incorporated in heasoft version 6.20. The 2015 Swift/XRT observation was obtained in the Photon Counting (PC) mode wherein the effects of photon pile-up have been corrected.¹ The 2016 observation was obtained in the Windowed Timing (WT) mode and it was not affected by pile-up as the XRT count rate was <100 count s⁻¹ (Romano et al. 2006). We have used xselect to extract source events from a circular region of 70.8 arcsec (30 pixels) radius for WT mode; and annulus region of inner radius 10 arcsec and outer radius 100 arcsec for PC mode. Background events were obtained from the outer regions of the CCD. For the WT mode, we have extracted background from a region with same size of the source and for the PC mode, circular region of radius 150 arcsec was used. Exposure maps were used to create ancillary response files with xrtmkarf, to account for hot pixels and bad columns on the CCD. The latest response matrix file was sourced from the calibration data base (version 20160609). Spectra were grouped to contain a minimum of 20 counts bin⁻¹.

We have used xspec (Arnaud 1996) version 12.9.1 for the spectral fitting. The persistent spectra extracted from Swift/XRT, NuSTAR-FPMA, and NuSTAR-FPMB observations were fitted simultaneously. We have added a constant to account for cross-calibration of different instruments. The value of constant for NuSTAR-FPMA was fixed to 1 and set free for others. Because of the low-energy spectral residuals in the WT mode, the WT mode data were fitted in the energy range 0.5–10 keV²; and 0.3–10 keV energy range was used for PC mode data. The NuSTAR data in 3–70 and 3–35 keV energy range were used for spectral fitting for 2015 and 2016 observations, respectively. The photoelectric absorption cross-section of Verner et al. (1996) and abundance of Wilms, Allen & McCray (2000) have been used throughout. All the spectral uncertainties and the upper limits reported in this paper are at 90 per cent confidence level. We have assumed the source distance to be equal to 10 kpc (Muno et al. 2001; Sharma et al. 2018).

3 LIGHT CURVE AND SPECTRAL STATES

Fig. 1 shows the background-subtracted FPMA+FPMB light curves of MXB 1658–298 binned at 20 s in energy range of 3–78 keV extracted from NuSTAR observations of 2015 and 2016. During the 2015 observation, the background-subtracted average count rate during the persistent phase was ∼6 count s⁻¹. One Type-I thermonuclear X-ray burst was also observed where burst count rate reached >500 count s⁻¹ (Sharma et al. 2018). During 2016, the background-subtracted average persistent count rate increased to ∼40 count s⁻¹ and the bursts observed were weaker and more frequent than 2015 observation.

Figure 1.

Background-subtracted light curve of MXB 1658–298 extracted from 2015 (left) and 2016 (right) NuSTAR observations binned at 20 s in energy band of 3–78 keV.

Open in new tabDownload slide

The CD and HID for MXB 1658–298 are shown in Fig. 2. They have been created after removing data during the bursts, dips, and the eclipses. The soft colour corresponds to the ratio of count rate between 5–8 and 3–5 keV and the hard colour is the ratio between 12–30 and 8–12 keV. In Fig. 2, the data points of 2015 and 2016 observations are denoted by black and red colours, respectively. In this figure, two branches are visible: the lower branch (red) can be identified with the banana branch and the upper (black) with the island state in CD (left-hand panel of Fig. 2). These different spectral branches depend on the accretion rates (Hasinger & van der Klis 1989). Observation of 2015 shows low accretion state (island) and of 2016 is showing high accretion state (banana), as shown in HID (right-hand panel of Fig. 2). For a quick comparison, the persistent spectrum for both observations is shown in Fig. 3. It clearly shows that the spectrum of 2015 was low/hard and of 2016 was high/soft.

Figure 2.

Colour–colour diagram (left) and hardness–intensity diagram (right) of MXB 1658–298. The time resolution of the data is 200 s and the count rate is in the energy range of 3–30 keV.

Open in new tabDownload slide

Figure 3.

Persistent spectrum extracted from NuSTAR (FPMA and FPMB) observations of 2015 and 2016. The spectrum of 2015 was hard and had low flux emission, while spectrum of 2016 was soft and had high flux emission. (The black and red colour represent the FPMA and FPMB data of 2016 observation, and blue and green colour represent FPMA and FPMB data of 2015 observation.)

Open in new tabDownload slide

4 SPECTRAL ANALYSIS AND RESULTS

The persistent emission spectra of MXB 1658–298 have been modelled with two component models (Oosterbroek et al. 2001; Sidoli et al. 2001; Díaz Trigo et al. 2006; Sharma et al. 2018). During the 1999–2001 outburst, MXB 1658–298’s broad-band spectrum was studied with BeppoSAX and was modelled with combination of disc blackbody and Comptonization component (Oosterbroek et al. 2001). For the same outburst, the spectrum obtained from the observations made with XMM–Newton was modelled with combination of blackbody and cut-off power law (Sidoli et al. 2001). In this work, we have modelled the persistent emission spectrum obtained from the observation of 2015 (low/hard) and 2016 (high/soft) with different combinations of continuum. This kind of comprehensive spectral analysis is being done for the first time in MXB 1658–298. We have tried single-component models for the fitting the spectra of low/hard and high/soft states. But they do not provide statistically adequate fit in both the cases.

4.1 The hard state emission

We started with single-temperature blackbody (bbodyrad) plus exponential cut-off power law (cutoffpl) to model the spectrum of 2015. tbabs has been used for modelling the interstellar absorption. The model tbabs × (bbodyrad + cutoffpl) did not give a good fit, χ² = 735 for 644 degrees of freedom (dof).

We also examined the continuum with another thermal Comptonization model comptt (Titarchuk 1994) along with the same thermal components. comptt model contains as free parameters: the temperature of the Comptonizing electron kT_e, the plasma optical depth τ, and the input temperature of the soft photon (Wein) distribution kT_seed. A spherical geometry was assumed for the Comptonizing region (corona). The model tbabs × (bbodyrad  + comptt) provided us a marginally better fit, χ²/dof = 709/643 (model A2, see Table 2).

We finally fit the spectrum with the three-component model, composed by two thermal components (bbodyrad and diskbb) and Comptonization component, nthcomp[bb] (model A3) or nthcomp[diskbb] (model A4; Lin et al. 2007; Armas Padilla et al. 2017). We obtained comparable chi-square values in the both cases. The best-fitting parameters obtained from the above described models are reported in Table 2. Fig. 4 shows the best-fitting spectrum of MXB 1658–298 in the hard state from all the four best-fitting models explained above. We used the cflux convolution model in xspec to estimate the unabsorbed flux in the 0.5–10 and 0.1–100 keV energy range. The X-ray luminosities calculated from unabsorbed flux in the 0.1–100 keV energy range are reported in Table 2.

Figure 4.

Spectrum of MXB 1658–298 extracted from 2015 observation (hard state) fitted with combinations of thermal and non-thermal components (model is mentioned on the top-right of each figure; DBB – diskbb, BB – bbodyrad). The NuSTAR data have been rebinned for representation purpose. The black points represent the Swift/XRT, red points represent NuSTAR-FPMA, and green points represent NuSTAR-FPMB spectra.

Open in new tabDownload slide

The inner disc radius R_in can be estimated from diskbb normalization Norm_disc using R_in = ακ²(Norm_disc/cos i)^1/2D_10 kpc km, where D_10 kpc is the distance to the source in units of 10 kpc, i is the inclination angle of disc, and α = 0.41 and κ = 1.7 are correction factors (Kubota et al. 1998). The inner disc radius of |$11^{+5}_{-3}$| and |$9^{+8}_{-2}$| km is estimated for models A3 and A4, respectively, assuming the inclination angle of 80° (as source is eclipsing binary; Frank, King & Lasota 1987). The calculated inner disc radius is nearly the order of NS radius or close to NS. However, the value of the inner disc radius is uncertain, subjected to the uncertainty in the disc inclination angle i. The structure of Comptonized corona was same for both the cases and gave y-parameter ∼2.7 and optical depth of ∼4.5.

4.2 Soft state emission

During the observation of 2016, MXB 1658–298 was in the high/soft state and significantly detected above background up to 35 keV only. We have modelled the persistent spectra of 2016 observation with absorbed cut-off power law with a thermal blackbody component, tbabs×(bbodyrad + cutoffpl). The fit obtained was unacceptable with χ²/dof = 1290/847. This model described the 0.5–35 keV continuum well except between 6 and 10 keV. The structure of residuals showed the presence of an absorption line near 6.7 keV and two absorption edges at 7.7 and 9 keV (see Fig. 5). So we added a Gaussian absorption line model (gauss) and two absorption edge models (edge). This resulted in an acceptable fit with χ²/dof = 857/840 (fit improvement by adding an edge at 9 keV: −Δχ² = 121 for two additional parameters; another edge at 7.7 keV: −Δχ² = 143.7 for two additional parameters; and Gaussian absorption line at 6.7 keV: −Δχ² = 167.7 for three additional parameters). The best-fitting parameters of this model (B1) are given in Table 3. With this model, the photon index obtained was hard (∼0.91). Similar hard photon index during soft state was found by Bloser et al. (2000a,b) with cut-off power-law model. If cutoffpl is replaced with powerlaw, the spectrum up to 10 keV will give photon index of ∼1.8, consistent with the value measured in the soft state of NS LMXBs (Boirin et al. 2005; Díaz Trigo et al. 2006; Raman et al. 2018).

Figure 5.

Residuals obtained after modelling persistent spectrum of 2016 observation with absorbed blackbody and cut-off power law. For simplicity we have shown only NuSTAR data in 3–20 keV. These residuals indicate the absorption feature in 6–10 keV (absorption line and two absorption edges) mainly due to Fe K.

Open in new tabDownload slide

In the next approach, we replaced the cut-off power-law component with Comptonization component nthcomp. The model tbabs×(bbodyrad + nthcomp[bb or diskbb]) also showed the presence of same spectral feature of absorption in 6–10 keV. So, we added the same Gaussian absorption line and two absorption edge models. The model with nthcomp[diskbb] resulted in unphysical values of emission radius of disc photons ∼2−3 km and kT_BB < kT_disc (Lin et al. 2007). When input seed photon type changed to blackbody (nthcomp[bb]; model B2), an acceptable fit was obtained (Table 3).

We replaced the bbodyrad with diskbb in model B2. This gave a marginally poor fit but it describes the continuum well. This model (B3) represents the blackbody emission from NS surface/boundary layer with |$kT_{\rm BB}=1.29^{+0.21}_{-0.24}$| keV and emission radius of R₀ ∼ 3−4 km is completely Comptonized and emission from accretion disc with |$kT_{\rm disc}=0.94^{+0.11}_{-0.12}$| keV and inner disc radius of |$R_{\rm in}=15.8^{+3.7}_{-2.3}$| km is directly visible. We have also found that adding a third thermal component, diskbb in model B2 and bbodyrad in B3 did not improve the fit. So, the continuum during high/soft state can be described with combination of blackbody or disc blackbody and thermally Comptonization component where input seed photons are provided by blackbody (NS surface/boundary layer). Fig. 6 shows the best-fitting spectrum of MXB 1658–298 in the soft state from above described best-fitting models B1, B2, and B3. The model tbabs × (bbodyrad + nthcomp[bb]), i.e. model B2 reproduce the data best as it improve the residuals in the soft band. The best-fitting parameters and the X-ray luminosities calculated from unabsorbed flux in the 0.1–100 keV energy range are reported in Table 3. The estimated X-ray luminosity during 2016 was a factor of ∼4 higher than 2015.

Figure 6.

Spectrum of MXB 1658–298 extracted from 2016 observation (soft state) fitted with combinations of thermal and non-thermal components. The model is mentioned on the top-right of each figure. The colour scheme as described in Fig. 4. The figures have been rebinned for representation purpose.

Open in new tabDownload slide

4.2.1 Absorption features

The absorption line observed during soft state at 6.77 ± 0.03 keV showed line width of ∼0.1 keV and was broader than the results of the previous observations, when single absorption lines from Fe xxv and Fe xxvi were detected (Sidoli et al. 2001). This line can be due to blend of Fe xxv Kα (6.70 keV) and Fe xxvi Kα (6.97 keV) absorption lines. Accordingly, we applied two Gaussian absorption lines at 6.70 and 6.97 keV by freezing the line energy parameter, instead of single Gaussian absorption line in the model B2. Both lines obtained were unresolved, we found the upper limits on the line widths of <0.12 and <0.14 keV for Fe xxv and Fe xxvi, respectively. The equivalent width (EW) of Fe xxv Kα and Fe xxvi Kα lines obtained were −42 ± 8 and −18 ± 7 eV, respectively. The two absorption edges were observed at 7.72 ± 0.05 and 9.06 ± 0.07 keV with optical depth of ∼0.15 and 0.11, respectively. The absorption edge at 9 keV is due to highly ionized Fe xxv/xxvi K edge. Similar, absorption feature as observed in MXB 1658–298 were also observed in Galactic jet source GRO J1655–40 (Yamaoka et al. 2001). Since absorption line features of Fe K ions are detected, the accompanying absorption edge structures from the Fe ions in the same ionization states would also appear in the spectra. Hence following Yamaoka et al. (2001) and Sidoli et al. (2002), we applied two absorption edges fixed at 8.83 keV (Fe xxv) and 9.28 keV (Fe xxvi) instead of the single 9 keV edge detected above. The best-fitting parameters for single and double absorptions are shown in Table 4. The obtained optical depth for 8.83 and 9.28 keV edges was nearly the same ∼0.06, shows both edges arise from nearly same region. Fig. 7 shows the spectra with double absorption lines from Fe xxv Kα and Fe xxvi Kα and their respective absorption edges.

Figure 7.

Spectrum of MXB 1658–298 during high/soft state when modelled with double absorption. For representation purpose, figure has been rebinned and only NuSTAR data are shown.

Open in new tabDownload slide

The absorption lines due to Fe xxv Kβ and Fe xxvi Kβ at 7.88 and 8.26 keV, respectively, had been observed in NS LMXB AX J1745.6–2901 (Hyodo et al. 2009; Ponti et al. 2015) and GX 13+1 (Sidoli et al. 2002). The observed absorption edge at 7.7 keV can be due to Fe xxv/xxvi Kβ absorption lines with contribution from highly ionized Ni Kα absorptions lines (Yamaoka et al. 2001) or(/and) Fe K emission as absorption edge can mimic the broad Fe K emission line (D’Aí et al. 2006; Egron et al. 2011; Mondal et al. 2016).

4.3 Photoionized absorption model

We have also modelled the observed absorption features in the soft state with a photoionized absorption component (zxipcf; Reeves et al. 2008) and assumed to be totally covering the X-ray source (covering fraction was fixed to 1). We modelled the broad-band spectrum with model tbabs×zxipcf×(bbodyrad + nthcomp[bb]). We note that zxipcf model was able to reproduce Fe K absorption lines between 6 and 7 keV and absorption edge at 9 keV (as shown in Fig. 8). However, residuals at 7–8 keV were still present. The fit showed positive residuals around 7 keV and negative residuals at 8 keV (see lower panel of Fig. 8). The fit obtained was not good, χ²/dof = 927/844. We first modelled these 7–8 keV features with an absorption edge. The fit improved to χ²/dof = 876.5/842 with column density of N_H ∼ 10²⁴ cm⁻² and ionization parameter of log (ξ) ∼ 4.2 of ionized absorber. The absorption edge was observed at 7.53 ± 0.09 keV with optical depth of 0.07 ± 0.01. In the second approach, we modelled this feature with emission line using Gaussian model and fit improved to χ²/dof = 878/841. The line energy observed was |$6.78^{+0.06}_{-0.09}$| keV with line width of |$0.38^{+0.11}_{-0.04}$| keV. The best-fitting column density of ionized layer was |$N_\mathrm{ H} = 6.7^{+1.2}_{-2.4} \times 10^{23}$| cm⁻², having an ionization parameter log |$(\xi) = 4.06^{+0.09}_{-0.14}$|⁠. The EW of Fe K emission line obtained was |$61^{+27}_{-20}$| eV. The edge at 7.5 keV can mimic the Fe K emission feature which was also found in 4U 1728–34 (D’Aí et al. 2006; Egron et al. 2011) and 4U 1820–30 (Mondal et al. 2016). Broad Fe K emission line was also observed during previous outburst with EW of |$41^{+17}_{-9}$| eV in BeppoSAX spectra (Oosterbroek et al. 2001) and |$160^{+60}_{-40}$| eV in XMM–Newton spectra (Sidoli et al. 2001).

Figure 8.

Spectrum when modelled with tbabs×zxipcf×(bbodyrad + nthcomp[bb]). For representation purpose, figure has been rebinned and only NuSTAR data are shown.

Open in new tabDownload slide

4.3.1 Absorption in hard state

Since no absorption feature was detected during persistent spectrum of 2015, therefore we added the Gaussian absorption line in model A1 to estimate upper limits. We found an absorption line at |$6.63^{+0.13}_{-0.12}$| keV (line width fixed to 10 eV) with EW of −15 ± 6 eV with marginal fit improvement of Δχ² = 6 for two additional parameters with F-test probability of 0.07. The addition of absorption line did not improve the fit significantly. So, this absorption feature is not much significant in this 2015 observation. We also estimated the upper limit on the EW of 6.4 keV Fe emission line of 16.5 eV (line width fixed to 0.3 keV) and on 6.97 keV Fe xxvi absorption line of −9 eV (line width fixed to 10 eV). We have also modelled the persistent spectrum of 2015 with photoionization absorption model zxipcf. We used the resultant model tbabs × zxipcf × (bbodyrad  + nthcomp[diskbb]). The covering fraction and redshift was fixed to 1 and 0, respectively. This improved the fit with Δχ² = 4 for two additional parameters with F-test probability of 0.16, which is not significant. So during observation of 2015, no absorption features were detected significantly as observed in the observation of 2016.

5 DISCUSSION AND CONCLUSION

This work shows the comparison of the broad-band spectrum of MXB 1658–298 in its soft and hard states. This kind of analysis has been done for the first time in this source. During its recent outburst, it was in low/hard state during the 2015 observation and in high/soft states during the 2016 observation. Many of the NS LMXBs have shown the transitions between the spectral states (Muñoz-Darias et al. 2014). Transient atoll source Aql X–1 has been observed in both hard and soft states during its 2007 outburst (Raichur et al. 2011; Sakurai et al. 2012). During the outburst of 2011, Aql X–1 is reported to show a spectral transition from hard to soft state (Ono et al. 2017). The transient LMXB AX J1745.6–2901 is also known to show both the spectral states during an outburst (Ponti et al. 2018).

During low/hard state, the combined broad-band spectrum from NuSTAR and Swift/XRT observations can be well described with combination of thermal component (blackbody) and Comptonized disc component. However, combination of two thermal components (blackbody and disc blackbody) and Comptonized component cannot be ruled out (see Table 2). During 2016 observation when source was in high/soft state, its broad-band spectrum can be described with combination of blackbody or disc blackbody and a Comptonization component (see Table 3). In the soft state, our study supports the NS surface (or boundary layer) as the dominant source for the Comptonization seed photons yielding the observed weak hard emission, while in the hard state inner accretion disc is favoured but with three-component model solutions, either the disc or the NS surface/boundary layer is equally favoured. The spectral shape during both observations was different and was according to the spectral state of the source. The electron temperature of Comptonizing plasma decreased to ∼4 keV in high/soft state from ∼19 keV in low/hard state. During low/hard state, Comptonizing region had optical depth of ∼4.1 and during high/soft state, plasma became optically thick with optical depth of ∼7. This type of variations in electron temperature and plasma optical depth has been observed during the hard and soft spectral state of several atoll LMXBs (e.g. Done et al. 2007; Raichur et al. 2011; Sakurai et al. 2012; Ono et al. 2017).

Like MXB 1658–298, the broad-band spectrum of Aql X–1, a typical non-dipping LMXB, and 4U 1915–05, a typical dipping LMXB, was also modelled with diskbb + nthcomp[bb] model in the soft state (Sakurai et al. 2012; Zhang et al. 2014). They showed blackbody emission of R_BB ∼ 3−4 km, which implies that the blackbody emission arises from an equatorial belt-like region on the NS surface. They also commonly indicate that the inner edge of the accretion disc is close to the NS surface (R_in ∼ 10−20 km). The measured coronal temperature kT_e of MXB 1658–298, ∼4 keV, is somewhat higher than that of Aql X–1 (2−3 keV), but is significantly lower than 4U 1915–05 (∼9 keV; Sakurai et al. 2012; Zhang et al. 2014). The dipping LMXBs in soft state (banana state) have systematically harder spectra than the normal LMXBs (Gladstone, Done & Gierliński 2007). The strength of Comptonization is normally evaluated by the y-parameter. Following Zhang et al. (2014), for LMXBs in the soft state, the difference between kT_e and kT_seed becomes small and they employed a new definition of the y-parameter as 4(τ + τ²/3)(kT_e − kT_seed)/m_ec². The newly defined y-parameter of MXB 1658–298 is ∼0.5 similar to 4U 1915–05 in the soft state (Zhang et al. 2014). This value appears at the upper limit of the recalculated y-parameters of normal LMXBs in the soft state by Zhang et al. (2014). Thus similar to 4U 1915–05, MXB 1658–298 is inferred to show stronger Comptonization effects than low and medium inclination LMXBs. We have obtained y-parameter ∼2.5 in the hard state (as defined in Section 4.1), a value higher than EXO 0748–676 (y ∼ 1.4) and other low and medium inclined LMXBs (y < 1.4; Zhang et al. 2016). The large difference in the value of y-parameter between MXB 1658–298 and EXO 0748–676 (both dipping LMXBs) can be due to the spectral modelling. EXO 0748–676 was modelled with diskbb + nthcomp[bb] (Zhang et al. 2016), while MXB 1658–298 was modelled with bbodyrad + nthcomp[diskbb]. Our model assume the disc is fully covered by Comptonizing corona and the high value of y-parameter suggests that corona is flattened over the disc. So, our study suggests that the Comptonizing corona has an oblate shape and the seed photons have to pass through the corona with a longer path resulting in systematically stronger Comptonization (Zhang et al. 2014, 2016).

We have observed absorption lines due to highly ionized Fe K in MXB 1658–298 during high/soft state. The Fe xxv Kα (6.70 keV) and Fe xxvi Kα (6.97 keV) absorption lines were observed with EW of |$-43^{+8}_{-7}$| eV and −18.6 ± 7.5 eV, respectively, with their corresponding absorption edges with optical depth of ∼0.06. For the first time absorption edges due to highly ionized Fe are observed in MXB 1658–298. The intense Fe K absorption features during soft states have been found in both NS and BH systems (e.g. Neilsen & Lee 2009; Ponti et al. 2012, 2014, 2016, 2018; Degenaar et al. 2014, 2016; Bozzo et al. 2016). The Fe K absorber is related to the presence of accretion disc absorber. In nearly 70 per cent of the known NS LMXBs, absorbing material is known to be bound to the accretion disc atmosphere. There is no shift in their absorption line energies. The rest of systems show blueshift in their absorption line energies (Díaz Trigo & Boirin 2013). For example, MXB 1658–298 (previous outburst), 4U 1323–62, and EXO 0748–676 did not show significant energy shifts, indicating that the absorption might occur in the disc atmosphere and does not necessarily require an outflow (Sidoli et al. 2001; Boirin et al. 2005; Díaz Trigo et al. 2006; Ponti et al. 2014). So, the observed absorption lines and edges in MXB 1658–298 are probably due to accretion disc atmosphere (or wind) as the source is viewed close to edge-on. Since, the observed K absorption lines and K absorption edges due to Fe ions were absent or not significantly detected during low/hard state. So, this suggests a connection between Fe K absorber and spectral state of the source. The connection between spectral state and presence of accretion disc wind or atmosphere has been found in both BH systems (Neilsen & Lee 2009; Ponti et al. 2012) and high inclination NS LMXBs: EXO 0748–676 (Ponti et al. 2014) and AX J1745.6–2901 (Ponti et al. 2015, 2018). Our study presents an overall picture of high inclination LMXBs (BH and NS) and their spectral states/disc absorber connection.

ACKNOWLEDGEMENTS

This research has made use of data obtained from the High Energy Astrophysics Science Archive Research Center (HEASARC), provided by NASA’s Goddard Space Flight Center. This research has made use of the NuSTAR Data Analysis Software (nustardas) jointly developed by the ASI Science Data Center (ASDC, Italy) and the California Institute of Technology (USA). RS and AJ acknowledge the financial support from the University Grants Commission (UGC), India, under the SRF and MANF-SRF schemes, respectively.

Footnotes

REFERENCES