3.1. Pre-periastron Flare Spectra

The spectra were obtained using the MAXI/GSC on-demand process¹⁰; all of them were analysed and modelled with the XSPEC (^{Arnaud 1996}) package and the energy range used for spectral fitting was (2.0-20.0) keV. Both phenomenological and physical models commonly applied to accreting X-ray pulsars have been tested. Models with power law components have been developed to HMXBs, such as Cen X-3 (^{Ebisawa et al. 1996}) and GX 301-2 (^{Fürst et al. 2011}; ^{Ji et al. 2021}). First of all, traditional models like power law with a high energy cut-off or with a Fermi-Dirac cut-off were explored. Although these models reasonably well reproduce the observed flare spectra between (2.0−20.0 keV) (χr2 ≈ [0.96−1.32]), they cannot offer a good statistical and/or physical solution for all spectra because the parameters of the high energy cut-off and the photon index (Γ = [−0.3,0.6]) could not be constrained well. As the principal aim is to describe with a consistent model all spectra these models were rejected. Then, the orbital phase-averaged spectra were described using a simple model with two power laws described by Equations (4) and (5). A fast charged particle traversing a region containing a strong magnetic field will change direction because the magnetic field exerts a force perpendicular to the direction of motion. Because the velocity vector changes, the charged particle is accelerated and consequently emits electromagnetic energy, called synchrotron radiation. Observed polarisation in the radiation is usually a proof of synchrotron emission (^{Giménez Cañete & Castro Tirado 2005}). The usual spectra form can be described by a power lawF(E)=BE ^−Γ, whereBrepresents a constant and Γ is the photon index. The more positive the value of the Γ, the softer is the spectrum. The scattering of X-rays by interstellar dust softens the spectrum byE ⁻²relative to the source spectrum, and soft X-rays are scattered more strongly than hard X-rays. In the energy range (0.1−10) keV, the main interaction between X-rays and matter is the photoelectric effect, whose cross-section varies with energy as ≈ Z ³ E ⁻³. Consequently, absorption is greatest at low energies and in high-Z materials. Thus, the energy spectra have been fitted by two power-law components, where each photon index indicates the source spectrum and the scattering or absorption spectrum, respectively. Initially, the orbital phase-resolved spectra were fitted with a composite model of two absorbed power laws:

Although both models described above fitted the averaged spectra between (2.0−20.0 keV) with a similar statistical confidence (Equation 4, χr2 = [0.86−1.10]; Equation 5, χr2 =[0.89−1.18]), they could not offer a consistent astrophysical solution to all the resolved spectra. For example, in the flare spectra NH2values were above 200×1022 cm−2 in eight spectra (Equation 4), NH2 ​≈ (231−106) × 10²²cm⁻²; Equation 5), NH2 ≈ (266−106)×10²²cm⁻²), which were not consistent with theN _H ​ values (10−80) × 10²²cm⁻²obtained in the MAXI/GSC analysis of ^{Islam & Paul (2014)}.

During flare episodes the surface temperature of the neutron star can reach several million degrees and, therefore, it will emit blackbody radiation with photons in the X-ray range. The blackbody component has been used to describe the soft energy band spectra of HMXBs like Cen X-3 using data from XMM-Newton (^{Sanjurjo-Ferrín et al. 2021}) and MAXI/GSC (^{Torregrosa et al. 2022}). The overall MAXI/GSC spectra of the source was modelled with an absorbedbbodycomponent. The spectral shape showed evidence for an Fe K-shell absorption edge at ≈ 7.1 keV (Saraswat et al. 2006; ^{Endo et al. 2002}; ^{Ji et al. 2021}); therefore, an edge component fixed at this energy was added. The following model was applied to describe all observational data:

where the Gaussian line was also included to account for the iron fluorescent emission line at ≈ 6.4 keV, if present. The model gave a good statistical description (pre-flare, χr2​=0.91−1.22; flare, χr2​=0.89−1.07; post-flare, χr2=0.85−1.18). In Figures 9-11 the ten MAXI/GSC spectra for each pre-periastron section, the best-fitting model, and residuals to the best-fitting model are shown.

Some spectra showed a low-energy excess below ≈4.0 keV as can be seen in Figure 9, images (b), (c), (g), (i) and (j); Figure 10, image (g); and Figure 11, images (g), (h) and (j). This soft excess could be produced by a transitory structure in the line of sight because it was not a permanent effect in all X-ray spectra. Therefore, a possible explanation may be the presence of a transitory disc which enhances the accretion of matter. X-ray pulsars such as GX 301-2 tend to be spinning up because the accreted material from the optical star has an angular momentum which is eventually transferred to the compact object. The higher the X-ray luminosity, the more material is accreted by the pulsar, and therefore the faster it will spin. The strongest spin-up event of GX 301-2 so far (between December 2018 and March 2019) pointed to an accretion due to both direct accretion from the stellar wind and a temporary accretion disc (^{Nabizadeh et al. 2019}). This scenario was also supported by the observations taken with theInsight-Hard X-ray Modulation Telescopeduring the initial part of this spin-up episode (^{Liu et al. 2021}). On the other hand, the low-energy excess could also be explained by scattering in the gas stream around the neutron star and in the stellar wind of the B-type companion star (^{Saraswat et al. 1996}).

Figures 12 to 19 show the evolution of the best-fit model parameters during the pre-periastron and apastron passages. In Figures 13-19 filled black squares represent the pre-flare parameter values, filled red triangles the flare parameter values, and open blue circles the post-flare parameter values.

Fig. 12 Evolution of the bbody norm for the pre-periastron flare and apastron outbursts versus the extraction number (MAXI/GSC data, model defined by Equation 6). Top panel: pre-flare. Second panel: flare. Third panel: post- flare. Bottom panel: apastron out-bursts. Filled red squares: bbody norm calculated by using L ₃₉/D102. Open blue circles: bbody norm fit value. The color figure can be viewed online. 

Fig. 13 Evolution of the radius of the emission zone for the pre-periastron flare and apastron outbursts versus the extraction number (MAXI/GSC data, model defined by Equation 6). Filled black squares: pre-flare. Filled red triangles: flare. Open blue circles: post-flare. Filled dark grey stars: apastron outbursts. The color figure can be viewed online. 

Fig. 14 Evolution of the bbody temperature for the pre-periastron flare and apastron outbursts versus the extraction number (MAXI/GSC data, model defined by Equation 6). Filled black squares: pre-flare. Filled red triangles: flare. Open blue circles: post-flare. Filled dark grey stars: apastron outbursts. The color figure can be viewed online. 

Fig. 15 Evolution of the unabsorbed flux for the pre-periastron flare and apastron outbursts (in units of 10^-9 erg cm^-2 s^-1) versus the extraction number (MAXI/GSC data, model defined by Equation 6). Filled black squares: pre-flare. Filled red triangles: flare. Open blue circles: post-flare. Filled dark grey stars: apastron outbursts. The color figure can be viewed online. 

Fig. 16 Evolution of the luminosity for the pre-periastron flare and apastron outbursts versus the extraction number (MAXI/GSC data, model defined by Equation 6). Filled black squares: pre-flare. Filled red triangles: flare. Open blue circles: post-flare. Filled dark grey stars: apastron outbursts. The color figure can be viewed online. 

Fig. 17 Evolution of the optical depth of the edge for the pre-periastron flare and apastron outbursts versus the extraction number (MAXI/GSC data, model defined by Equation 6). Filled black squares: pre-flare. Filled red triangles: flare. Open blue circles: post-flare. Filled dark grey stars: apastron outbursts. The color figure can be viewed online. 

Fig. 18 The parameters of the Fe Kα emission line versus the extraction number (MAXI/GSC data, model defined by Equation 6). Top panel: line energy. Second panel: line width. Third panel: equivalent width. Bottom panel: intensity. The unit of the line flux I is photons s^-1 cm^-2. Filled red triangles: flare. Open blue circles: post- flare. The color figure can be viewed on-line. 

Fig. 19 Evolution of the column density for the pre-periastron, flare, and apastron outbursts versus the extraction number (MAXI/GSC data, model defined by Equation 6). Filled black squares: pre-flare. Filled red triangles: flare. Open blue circles: post-flare. Filled dark grey stars: apastron outbursts. The color figure can be viewed online. 

Taking the uncertainties into account, the results for thebbody normalisation(Figure 12) obtained with XSPEC (open blue circles) are consistent with the values obtained usingL ₃₉​/D102​ (filled red squares).

The radius of the emission zone (Figure 13) is of the order of 1 km, which is consistent with a hot spot on the NS surface and with the results of XMM-Newton data. The temperature of the blackbody (kTin keV, Figure 14) has a constant value between (3.0−3.5) keV in the pre-flare spectra, but it is enhanced in the sixth extraction, which is reflected in a drop of the emission radius (Figure 13). In the flare spectra, it is enhanced in the (5.5-6.5) keV range indicating an increase in the accretion rate which is reflected in a higher X-ray luminosity compared to pre-and post-flare. In the post-flare the temperature is approximately constant, (3.8-4.1) keV, between the third and the tenth extraction, but there is a large increase in the first and second extractions with a value ≈ 6.0 keV which is reflected in a clear decrease of the emission zone between ≈ (0.2-0.3) km.

In the flare the unabsorbed flux (Figure 15) has no significant evolution in the fitting energy range (2.0-20.0) keV, showing that the accretion rate is quite stable. The unabsorbed fluxes show opposite trends between the pre-flare (the flux increases) and the post- flare (the flux decreases) showing that the accretion rate has a different behaviour along the pre-periastron passage. A possible explanation for this would be that the absorbing matter is located in the line of sight between the observer and the neutron star in the post- flare (N _H ≳ 2 10²³ cm^-2) and away from the line of sight in the pre-flare (N _H < 2 10²³ cm^-2). Thus, it is confirmed that the distribution of circumstellar matter around the compact object is rather inhomogeneous during the pre-periastron passage (^{Saraswat et al. 1996}; ^{Islam & Paul 2014}).

The X-ray luminosity (Figure 16) is consistent with a constant value (L _X ≈1.3 × 10³⁷ erg s^-1) in the flare spectra (which agrees with the observations ID 0555200301 and ID 0555200401 from XMM-Newton) and is greater than in the pre-flare (L _X = [4 - 7] 10³⁶ erg s^-1) and in the post-flare (L _X = [3 - 8] 10³⁶ erg s^-1) spectra. This scenario, in which the wind mass-loss rate is insufficient to power the source entirely, indicates that an additional mechanism is needed to give the neutron star the fuel it needs. Wind-fed HMXBs are powered by accretion of the radiatively driven wind of the luminous component on the compact object with typical X-ray luminosities ≈10³⁶ erg s^-1 (see ^{Martínez-Nuñez et al. 2017}; ^{Kretschmar et al. 2019}, for instance), i.e. one order of magnitude smaller than observed in the flare event. A possible mechanism for enhancing the X-ray emission is thought to be the presence of a gas stream trailing the neutron star (^{Saraswat et al. 1996}; ^{Islam & Paul 2014}). Another cause could be that an accretion disc has formed around the neutron star (^{Nabizadeh et al. 2019}; ^{Liu et al. 2021}) or a combination of both possibilities (^{Saraswat et al. 1996}). Spin-up episodes are usually characterised by an increase in X-ray luminosity, associated with an enhancement of the accreted matter as a consequence of the formation of a temporary accretion disc around the neutron star (^{Nabizadeh et al. 2019}; ^{Manikantan et al. 2023}).

This source is seen through a column density in the range N _H = [10-40] 10²² cm^-2, and all X-rays below [2-3] keV are absorbed. At these column densities a feature due to the K edge of Fe at 7.1 keV should be present. Figure 17 shows the optical depth of the iron edge (τ _Kedge) best-fit parameter that lies in the range [0.3 - 0.8]. This represents an ionisation state of nearly neutral iron Fe i-v (if values reported by ^{Saraswat et al. 1996}; ^{Endo et al. 2002}; ^{Ji et al. 2021}, are taken into account).

The long-term averaged spectra in the flare section present a fluorescent iron emission line energy consistent with the Fe Kα line and a constant value of ≈ (6.35±0.08) keV, as can be seen in Figure 18, top panel, which implies an ionisation state of iron up to Fe XVII. It was only detected in the first two spectra in the post-flare section and it was totally absent in the pre- flare section. However, ^{Islam & Paul (2014)} performed an orbital phase-resolved spectral analysis of GX 301-2 and they reported that the iron emission line was detected in all the orbital phases. The fitted values of the line width, the equivalent width and the intensity of the Fe Kα emission line are plotted in Figure 18: second, third and bottom panels, respectively.

If the emission line luminosity is produced by a thin spherical shell of matter, i.e. a shell of gas surrounding a point source of continuum radiation with uniform ionisation, composition and density, the line equivalent width should be proportional to the column density (^{Kallman et al. 2004}; ^{Torrejón et al. 2010}) EW _line(keV) ≃ 0.3 N _H (10²⁴ cm^-2). The absorbing component varies in the range ≈ [10-30] × 10²² cm^-2 (see second panel from the top of Figure 19), which implies an equivalent width of Fe Kα in the range EW _line ≈ [30-90] eV. On the other hand, for an isotropic surrounding cosmic fluorescing plasma, ^{Inoue (1985)} obtained a relationship between the expected equivalent width of the iron Kα line and the hydrogen column density as EW _expected(eV) = 100 N _H (10²³ cm^-2) which is valid for N _H < 10²⁴ cm^-2 and a photon index of the power-law spectrum of 1.1 (see also ^{Endo et al. 2002}, for example). They also concluded that GX 301-2 corresponds to this case and, therefore, EW _line ≈ [100-300] eV. However, MAXI/GSC observations gave equivalent widths of the iron emission line in the range EW_line ≈ [600-1100] eV, which implies large deviations from the linear correlation. The largest value of the equivalent width was obtained with the smallest column density. The fact that the Fe Kα is not a single line but a superposition of different Kα lines of differently strongly ionised iron together with the Compton shoulder at ≈6.24 keV and the wind speed could explain the large equivalent widths found in this study (^{Watanabe et al. 2003}; ^{Fürst et al. 2011}; ^{Ji et al. 2021}). Taking the uncertainties into account, the best-fit parameters in the pre-periastron orbital phase are consistent with previous studies (^{Islam & Paul 2014}; ^{Manikantan et al. 2023}). As a consequence, during the pre-periastron passage the fluorescent iron emission line is not emitted from a spherically symmetric distribution of matter surrounding the neutron star. Although this study has focused on flare episodes, a very small part of the orbit, these results are consistent with those obtained by ^{Islam & Paul (2014)}.

Figure 20 shows the equivalent width of the Fe Kα against the absorption column density during the pre-periastron passage compared with the theoretical predictions for an isotropically distributed gas and for a spherical shell of gas surrounding the source. In fact, it seems that there is a moderate anti-correlation between them during the pre-periastron passage and a Pearson correlation coefficient r = -0.63 was found. This result is in contrast to other studies, such as ^{Makino et al. (1985)}, ^{Endo et al. (2002)}, ^{Fürst et al. (2011)}, ^{Ji et al. (2021)}, where they found a linear correlation between these two parameters which suggests that the accretion material near the neutron star is spherically distributed. Nevertheless, ^{Ji et al. (2021)} also found deviations from this linear correlation when the column density was higher than 1.7 × 10²⁴ cm^-2. They pointed out that this could be due to the formation of an accretion disc.

Fig. 20 Variability of the equivalent width as a function of the column density (also known as the curve of growth). Filled black triangles: flare. Open black circles: post-flare. The dotted line represents the relation for an isotropic surrounding cosmic fluorescing plasma, EW _expected(eV) = 100 N _H (10²³ cm^-2) (^{Inoue 1985}). The solid line shows the relation for the luminosity produced by a thin spherical shell of matter EW _line(keV) ≃ 0.3 N _H (10²⁴ cm^-2) (^{Kallman et al. 2004}; ^{Torrejón et al. 2010}). 

In Figure 21 the intensity of the Fe Kα emission line is shown versus the unabsorbed flux of the source in the [2-20] keV energy band (left plot) and the optical depth of the iron K-edge absorption versus the equivalent width of the iron Kα line (right plot). A moderate linear correlation (Pearson correlation coefficient r = 0.63) can be seen, which should be consistent with the expected line intensities for the hydrogen column densities derived from the model described by Equation 6 (see fig.6b in ^{Haberl 1991b}, where the incident spectrum was assumed to be a power-law with photon index Γ = 0.55). In contrast, from the plot τ _edge versus EW, an anticorrelation relationship (Pearson correlation coefficient r = -0.75) seems to be present between these parameters during the pre-periastron passage. The solid line represents the linear fit found by ^{Ji et al. (2021)} where they suggested that the reprocessed material reached an optical depth of unity for EW ≈ 400 eV (see also ^{Torrejón et al. 2010}, where they also found a linear correlation).

Fig. 21 Left panel: Log-log plot of the Fe Kα intensity as a function of the unabsorbed flux in the (2-20) keV energy band. The solid line represents a linear fit. Right panel: Plot of the optical depth of the iron K-edge absorption as a function of the equivalent width of the iron Kα line (in keV). The solid line is the linear fit found by ^{Ji et al. (2021)}. An apparent anti-correlation can be seen during the pre-periastron passage. The larger the EW of the Fe Kα, the smaller the optical depth of the iron K-edge absorption. Filled black triangles: flare. Open black circles: post-flare. 

3.2. Apastron Flare Spectra

The folded light curve of GX 301-2 shows two flare-like features at binary orbital phases ≈ 0.26, i.e. before apastron passage (^{Haberl 1991b}; ^{Saraswat et al. 1996}), and at ≈ 0.45 near-apastron passage (as can be seen in Figure 3, see also ^{Pravdo et al. 1995}; ^{Pravdo & Ghosh 2001}, for instance). In general, periodic near-apastron outbursts show lower intensity than the pre-periastron flare, although X-ray emission can sometimes be as low as ≈ 10³⁶ erg s^-1, i.e. it cannot be distinguished from the baseline X-ray intensity along the orbit. Thus, the criterion to identify an apastron outburst was that the unabsorbed X-ray flux was greater than ≈ 2 × 10^-9 erg s^-1 cm^-2. The orbital phase of the apastron outbursts was also obtained with the parameters reported by ^{Koh et al. (1997)}. Then, the Good Time Intervals (GTIs) were derived and the corresponding MJDs are listed in the caption of Figure 22. Finally, the spectra were extracted following the process explained in § 3.1.

Fig. 22 MAXI/GSC spectra of GX 301-2 in the [2.0-20.0] keV band corresponding to the apastron flare. MJDs from left to right for each spectrum: (a) [55376.5-55380.5], (b) [55710.0-55713.0], (c) [55870.0.0-55873.0], (d) [56163.5-56164.5], (e) [57285.0-57288.0], (f) [57454.0-57457.0], (g) [58029.5-58041.5], (h) [58480.5-58495.5] and (i) [59119.0-59122.0]. Units of the x-axis: Energy (keV). Units of the y-axis: Normalized counts s^-1 keV^-1. 

In this analysis, all apastron spectra were fitted with the same model used for the pre-periastron flare spectra. The range of the χr2 values was [0.7-1.2] for all ts. Nevertheless, it should be noted that no Gaussian component has been included here because the fluorescent iron emission line was not detected by MAXI/GSC. Figure 22 shows the data, the absorbed blackbody modified by the iron K-edge absorption best-fit model (top panels), and the residuals for the model (bottom panels). The best-fit model parameters are plotted in Figures 12-17 and 19, bottom panels.

The unabsorbed fluxes and the derived radii of the blackbody emission region are shown in Figures 15 and 13, respectively. The results for the radii are consistent with a hot spot on the NS surface (^{Sanjurjo-Ferrín et al. 2021}; ^{Torregrosa et al. 2022}) (see Table 4, Column 3).

Using the definition of the bbody normalisation K _bb = L ₃₉/D102, we have derived its value in the nine spectra (see Table 4, Column 4). These values are consistent with those obtained from fitting the model F (E) = tbnew × bbody × edge, taking the uncertainties into account (see Figure 12, bottom panel).

The first apastron outburst spectrum observed by MAXI/GSC is shown in Figure 22 (a) whose X-ray luminosity, see Table 4, Column 6, L _X(a) ≈ 2 × 10³⁷ was unusually brighter than that of pre-periastron flare (see Figure 16 and compare second and bottom panels). This flare was also detected by Fermi/GBM and could be associated with a rapid spin-up episode of GX 301-2 (^{Finger et al. 2010}). According to the long-term light curve obtained with Fermi/GBM, rapid spin-up began on June 23, 2010 (MJD 55 370.84) and finished on July 22, 2010 (MJD 55 399.07), as can be seen in Figure 8. Pulse timing measurements for the interval June 23.8-July 22.1 showed a spin-up frequency rate of ν˙_spin = (3.89±0.08) 10^-12 Hz/s, giving a spin-up time scale of ≈12 yr. This extracted spectrum covered only 1/7 of the spin-up episode with an spin-up frequency rate of ν˙_spin = (6.21±0.21) × 10^-12 Hz/s. The following apastron flare with an X-ray luminosity similar to the pre-periastron flare is shown in Figure 22 (d) and was not associated with any significant changes in the spin period of the neutron star (the spin frequency decreased from ν ₁ = (1.46861±0.00017) 10^-3 Hz on MJD 55 611.1160 to ν ₂ = (1.46818±0.00018) × 10^-3 Hz on MJD 55 616.9968). The last one studied in this work corresponded to the eighth extraction, Figure 22 (h), where the source presented a new rapid spin-up event (^{Abarr et al. 2020}; ^{Liu et al. 2021}; ^{Manikantan et al. 2023}). In this case, during the observation the spin frequency increase fromν ₁ = (1.467459±0.000011) × 10^-3 Hz on MJD 58 485.1001 to ν ₂ = (1.472308±0.000007) × 10^-3 Hz on MJD 58 494.9950 at a rate of ν˙_spin = (5.75 ± 0.02) ×10^-12 Hz/s, which is near to the spin frequency increase over 18 days (MJD 58485.1001-58 502.9107) ν˙_spin = (5.70 ± 0.02) × 10^-12 Hz/s. Then, the spin frequency rose over 24 days (MJD 58 502.9107-58 553.1933) at a rate of ν˙_spin = (2.37 ± 0.18) × 10^-12 Hz/s.

Another rapid spin-up episode seen with Fermi/GBM from MJD 54 830.9617 to MJD 54 855.1007 showed a spin-up frequency rate of ν˙_spin = (2.97±0.12) × 10^-12 Hz/s (^{Finger et al. 2010}). Moreover, similar spin-up events were detected with the Burst And Transient Source Experiment (BATSE) and reported by ^{Koh et al. (1997)}; ^{Bildsten et al. (1997)}. They found a spin frequency growth over 23 days at a rate of ν˙_spin = 4.5 ×10^-12 Hz/s and over 15 days at a rate of ν˙_spin = 3.0 × 10^-12 Hz/s.

The rest of the apastron flares shown in Figures 22 (b), (c), (e), (f), (g) and (i) had an X-ray emission slightly lower than pre-periastron flares which were not associated with any significant changes in the spin period of the neutron star. From MJD 55 708.9521 to MJD 55 714.8995, the spin-up rate was ν˙_(b) = (1.1 ± 0.6) × 10^-12 Hz/s; from MJD 55 860 to 55 875 only two pulse timing measurements were taken, MJD 55 871.1393 and 55 873.1066, which gave a spin-down rate of ν˙_(c) = (-1.9 ± 1.4) × 10^-12 Hz/s; from MJD 57 265.2712 to MJD 57 290.7870, the spin-up rate was ν˙_(e) = (9.8 ± 0.8) × 10^-13 Hz/s (although this time interval had a gap without measurements MJD 57 273 - 57 286); from MJD 57 449.0139 to 57 458.9447 the spin-down rate was ν˙_(f) = (-1.1 ± 0.3) × 10^-12 Hz/s; from MJD 58 028.9070 to MJD 58 036.7733, the spin-up rate was ν˙_(g) = (1.0 ± 0.7) × 10^-12 Hz/s (but before and after there was a gap of 10 days without pulse measurements); and from MJD 59 117.1258 to MJD 59 121.0794, the spin-up rate was ν˙_(g) = (2.4 ± 1.1) × 10^-12 Hz/s.

As far as it is known from the rapid spin-episodes in GX 301-2, during the apastron passage the source becomes as bright as a pre-periastron flare, as a consequence of the formation of a transitory accretion disc. Thus, the mass transfer from the companion star to the neutron star is a rather irregular process during this orbital phase (see X-ray luminosities in Table 4). The strongest events show that the material is accreted from the stellar wind, possibly from a gas stream (^{Leahy & Kostka 2008}) and probably through a temporary accretion disc (^{Koh et al. 1997}; ^{Nabizadeh et al. 2019}; ^{Abarr et al. 2020}; ^{Liu et al. 2021}; ^{Manikantan et al. 2023}). However, the fourth apastron flare [see extraction number 4 in Table 4 and Figure 22 (d)] did not show a spin-up of the neutron star and no transitory disc was formed. Therefore, the material should be accreted by the stellar wind and possibly from a gas stream.

It should be noted that the uorescence iron emission line at ≈6.4 keV has not been detected in any spectrum of the apastron flare. Sensitive X-ray observatories such as ASCA (^{Endo et al. 2002}), XMM-Newton (^{Giménez-García et al. 2015}), and Chandra (^{Torrejón et al. 2010}) detected and resolved the iron line complex in GX 301-2. In contrast, MAXI/GSC does not have enough sensitivity to distinguish between a weak, broad iron K emission line and a bright X-ray continuum (^{Rodes-Roca et al. 2015}; ^{Torregrosa et al. 2022}). Nevertheless, other long-term orbital phase resolved spectroscopy studies reported the presence of the Fe Kα line in all orbital phases (^{Islam & Paul 2014}; ^{Manikantan et al. 2023}).