Suzaku observation of a symbiotic X-ray binary 4U 1954+319

# Spectral and Timing Nature of the Symbiotic X-ray Binary 4U 1954+319: The Slowest Rotating Neutron Star in an X-ray Binary System

Teruaki Enoto11affiliation: NASA Goddard Space Flight Center, Astrophysics Science Division, Code 662, Greenbelt, MD 20771, USA; teruaki.enoto@nasa.gov 22affiliation: High Energy Astrophysics Laboratory, RIKEN Nishina Center, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan Makoto Sasano33affiliation: Department of Physics, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, 113-0033, Japan , Shin’ya Yamada22affiliation: High Energy Astrophysics Laboratory, RIKEN Nishina Center, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan , Toru Tamagawa22affiliation: High Energy Astrophysics Laboratory, RIKEN Nishina Center, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan , Kazuo Makishima22affiliation: High Energy Astrophysics Laboratory, RIKEN Nishina Center, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan 33affiliation: Department of Physics, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, 113-0033, Japan ,
Katja Pottschmidt44affiliation: NASA Goddard Space Flight Center, Astrophysics Science Division, Code 661, Greenbelt, MD 20771, USA 55affiliation: CRESST & University of Maryland Baltimore County, Baltimore, MD 21250, USA , Diana Marcu44affiliation: NASA Goddard Space Flight Center, Astrophysics Science Division, Code 661, Greenbelt, MD 20771, USA 55affiliation: CRESST & University of Maryland Baltimore County, Baltimore, MD 21250, USA , Robin H.D. Corbet11affiliation: NASA Goddard Space Flight Center, Astrophysics Science Division, Code 662, Greenbelt, MD 20771, USA; teruaki.enoto@nasa.gov 5 5affiliationmark: , Felix Fuerst66affiliation: Cahill Center for Astronomy and Astrophysics, California Institute of Technology, Pasadena, CA 91125 , and Jörn Wilms77affiliation: Dr. Remeis-Sternwarte and Erlangen Centre for Astroparticle Physics, Universität Erlangen-Nürnberg, Sternwartstr. 7, 96049 Bamberg
###### Abstract

The symbiotic X-ray binary 4U 1954+319 is a rare system hosting a peculiar neutron star (NS) and an M-type optical companion. Its 5.4 h NS spin period is the longest among all known accretion-powered pulsars and exhibited large (7%) fluctuations over 8 years. A spin trend transition was detected with Swift/BAT around an X-ray brightening in 2012. The source was in quiescent and bright states before and after this outburst based on 60 ks Suzaku observations in 2011 and 2012. The observed continuum is well described by a Comptonized model with the addition of a narrow 6.4 keV Fe K line during the outburst. Spectral similarities to slowly rotating pulsars in high-mass X-ray binaries, its high pulsed fraction (60–80%), and the location in the Corbet diagram favor high -field ( G) over a weak field as in low-mass X-ray binaries. The observed low X-ray luminosity (), probable wide orbit, and a slow stellar wind of this SyXB make quasi-spherical accretion in the subsonic settling regime a plausible model. Assuming a  G NS, this scheme can explain the 5.4 h equilibrium rotation without employing the magnetar-like field ( G) required in the disk accretion case. The time-scales of multiple irregular flares ( s) can also be attributed to the free-fall time from the Alfvén shell for a  G field. A physical interpretation of SyXBs beyond the canonical binary classifications is discussed.

accretion — X-rays : binaries — binaries : symbiotic — magnetic fields — stars : neutron — stars: individual (3A 1954+319, 4U 1954+319)

## 1. Introduction

Symbiotic X-ray binaries (SyXBs) are a growing new subclass of X-ray binaries which consist of an X-ray bright neutron star (NS) and an M-type giant primary star. Originally SyXBs were seen as a subclass of low-mass X-ray binaries (LMXB), where accretion typically is from a K-type optical counterpart with a mass . It took more than 30 years, however, to recognize that their properties are different from LMXB and also from those of high-mass binaries (HXMB), i.e., systems where the mass donor is an O- or B-type giant with . SyXBs are now considered as a separate subclass of X-ray binaries. They are named in analogy to symbiotic binaries, which consist of a white dwarf and an M giant companion, but their nature has not yet been understood.

Some SyXBs were known as famous X-ray sources since the early X-ray observations: GX 1+4 (Davidsen et al., 1977; Makishima et al., 1988; Chakrabarty & Roche, 1997), 4U 1700+24 (Garcia et al., 1983; Masetti et al., 2002), and Sct X-1 (Kaplan et al., 2007). With the discovery of new sources such as IGR J161942810, there are now nearly 5 SyXBs and some candidates in our Galaxy; e.g., 5 objects out of 10 candidates (Table 1 in Lü et al. 2012) have been confirmed as SyXBs, while three were ruled out (see e.g., Masetti et al. 2012). Observational features of the SyXB class are i) quite long orbital periods111For many SyXB, orbital period has not yet been known nor confirmed in long-term X-ray studies (e.g., Corbet et al. 2008)., e.g., 1161 d in GX 1+4. ii) long NS spin periods over 110–18300 s as shown in Fig. 1, and iii) high X-ray variability ranging from short to long timescales (1 s to up to a year).

SyXB X-ray spectra are in general well described by an absorbed power-law with a photon index 0.5–2.0 and a high energy roll-over, which is often modified in the soft X-rays by strong absorption. This spectrum resembles that of X-ray pulsars in HMXBs, and thus, SyXBs have been interpreted in the literature as NSs with G field in HMXBs. As another interpretation, following the classical LMXB classification based on optical companions, SyXBs were also interpreted as NSs in the low luminosity low/hard state of LMXB as having weak magnetic field (-field), e.g.,  G (Nagae et al., 2008; Kitamura et al., 2013). In any of these cases, the NSs in SyXB are expected to have -fields strengths of around  G or much lower (e.g.,  G).

The canonical -fields implied by the spectral analogy with HMXB and LMXB, however, are in contradiction to the fields implied by their timing behavior. The slow NS pulsations in SyXBs are sometimes in spin equilibrium or show a large spin-up rate (Makishima et al., 1988; González-Galán et al., 2012). If interpreted in terms of standard disk accretion torque theory (Ghosh & Lamb, 1979), the slow rotation and torque reversals imply extraordinary strong dipole fields ( G), which are close to or exceed the quantum critical field,  G. Such a magnetar scenario for binary systems has been proposed, e.g., to explain the 1.6 h pulsation and the high spin-up rate of the SyXB IGR J163584726 (Patel et al., 2007). Similar scenarios have also been discussed for slowly rotating X-ray pulsars in HMXBs such as 4U 0114+65 (Li & van den Heuvel, 1999) or 4U 2206+54 (Reig et al., 2012), highly variable supergiant fast X-ray transients (SFXTs; Bozzo et al. 2008), and some slowly rotating NSs in the Small Magellanic Cloud (Ho et al., 2014; Klus et al., 2014). Except for the timing properties, however, no direct evidence for ultra-strong -fields has been found so far.

As shown by Shakura et al. (2012) and Postnov et al. (2012), the extremely strong fields can be avoided when one assumes that the accretion does not happen through an accretion disk but via quasi-spherical accretion. In these models matter is gravitationally captured, e.g., from the donor’s stellar wind. For low mass accretion rates, , gravitationally captured matter subsonically accretes onto the NS forming an extended, quasi-spherical shell around the magnetosphere, where large scale convective motions and subsonic turbulence lead to accretion onto the NS. Contrary to the canonical high luminosity Bondi-Hoyle accretion, where matter supersonically accretes and the sign of angular momentum (prograde or retrograde) determines the spin behavior (Perna et al., 2006), for quasi-spherical accretion the spin-up/spin-down is determined by the specific angular momentum of matter at the magnetospheric boundary and the angular velocity of the NS.

In this paper we discuss Suzaku observations of the SyXB 4U 1954+319 hosting the slowest rotating NS in an X-ray binary system. This source was originally discovered by Uhuru (Forman et al., 1978), and observed in the 1980s with Ariel (Warwick et al., 1981), EXOSAT, and Ginga (Tweedy et al., 1989). These observations suggested that the source was a wind accreting HMXB with a very inhomogeneous wind (Tweedy et al., 1989), although no optical counterpart was known at that time. The source had virtually been forgotten until, after twenty years of no observations, Chandra’s localization of the source position allowed the optical identification of the donor, which was found to be an M4-5 III star () at a distance of 1.7 kpc (Masetti et al., 2006b). Therefore 4U 1954+319 was re-classified as a SyXB.

Further progress in our understanding of the characteristics came with the discovery of a 5 h periodicity in Swift/BAT monitoring data by Corbet et al. (2006, 2008). This period is strongly variable (Marcu et al., 2011). So far, no orbital period has been discovered, although Mattana et al. (2006) argue that the lower limit of the orbital period is 400 d. These parameters are incompatible with typical white dwarf systems, such that the compact object in 4U 1954+319 is likely a NS. As shown in Figure 1, the extremely long spin period makes 4U 1954+319 the slowest rotating accretion-powered NS system known to date222A long periodicity, hour, was also detected from the central object 1E1613485055 in the young shell-type supernova remnant RCW 103. This object, however, has not been confirmed as an accretion-powered pulsar., and a prominent example to understand the SyXB class.

Currently, no consensus has yet been obtained on the accretion models for SyXBs, or the -field of their NSs. Revisiting and developing the quasi-spherical model further requires the spectral and timing investigations of SyXBs. In this paper, we report on the broad-band timing and spectral properties of 4U 1954+319 and discuss a possible accretion model beyond the conventional HMXB and LMXB classification.

## 2. Observation and Data Reduction

### 2.1. Suzaku observations

There are two pointed Suzaku (Mitsuda et al., 2007) observations of 4U 1954+319. The first one, in AO6, was conducted on 2011 October 23–24 for a gross on-source time of 117 ks. Nearly one year later, on 2012 October 5, MAXI monitoring detected a brightening of this source333Trigger ID 6206800002: http://maxi.riken.jp/pipermail/x-ray-star/2012-October/000187.html, with the X-ray intensity reaching 15 mCrab in the 3–10 keV band. About 27 days later, on 2012 November 1/2, a follow-up ToO observation was performed with Suzaku for a gross on-source time of 120 ks. Table 1 summarizes the two Suzaku observations.

The X-ray Imaging Spectrometer (XIS; Koyama et al., 2007) on board Suzaku, coupled to the X-ray Telescopes (XRT; Serlemitsos et al., 2007), consists of two front-illuminated (FI) CCD detectors, XIS0 and XIS3, and one back-illuminated sensor (XIS1). The CCDs are sensitive in the 0.2–12 keV range and, when operated in the full window mode, read out every 8 s. We also used data from the Hard X-ray Detector (HXD Takahashi et al., 2007), which consists of Silicon PIN diodes (HXD-PIN) and GSO scintillators (HXD-GSO), and cover the 10–70 keV and 50–600 keV bands, respectively. The HXD was operated in the nominal mode during both observations with 61 s time resolution. The bright X-ray source Cygnus X-1, which is only from 4U 1954+319, is still outside the HXD-PIN’s tightly collimated field of view (FOV, FWHM), but within the FOV of the HXD-GSO (). Thus, we did not use the HXD-GSO data in the following analyses due to a possible contamination from Cyg X-1. Based on the INTEGRAL source catalog, there is no other strong contamination source within the HXD-PIN FoV.

### 2.2. Data processing

We analyzed the XIS and HXD datasets as obtained from the standard pipeline processing, revision 2.7 and 2.8 for the 2011 and 2012 data sets, respectively. All data analyses were performed using HEADAS version 6.11 or later. Following the standard Suzaku screening criteria, X-ray events were discarded if one of the following conditions held true: (a) they were recorded during the South Atlantic Anomaly, and for the XIS and the HXD, respectively, and later than 180 s before the next passage for the HXD, (b) while the satellite was in regions of low geomagnetic cutoff rigidity (6 GV), (c) while the target was closer than to the Earth’s rim (d) if the instantaneous pointing direction deviated more than from the mean, or (e) during periods of telemetry saturation. After screening, the net exposures for the 2011 observation were 60.2 ks (XIS) and 53.7 ks (HXD), while those in 2012 were 60.8 ks (XIS) and 50.5 ks (HXD), respectively.

On-source XIS events were extracted from a region of radius centered on the source. Background events were derived from an annulus around the source. From the HXD-PIN data, we subtracted the Non X-ray Background (NXB) created with the LCFITD method (Fukazawa et al., 2009) and filtered following the same screening criteria as those used for the on-source data. The Cosmic X-ray Background (CXB) was modeled as described by Enoto et al. (2010a) with the same spectral shape as given by Moretti et al. (2009). In the following analyses, unless otherwise specified, all uncertainties are given at the 68% () confidence level.

### 2.3. Reduction of the X-ray spectra

For the spectral analysis we summed the two XIS-FI sensors and utilized XIS data from the 0.8–10 keV band and the HXD-PIN data from the 15–70 keV band. The 1.7–1.9 keV XIS data are discarded due to calibration uncertainties. Uncertainties in the flux normalization of the instruments are taken into account with a multiplicative constant. The cross-normalization factor of HXD-PIN relative to XIS-FI was fixed at 1.171 based on the XIS-nominal Crab Nebula calibration (Maeda et al., 2008), while that of XIS-FI and XIS-BI was allowed to differ by up to 1%. A 0.5% systematic uncertainty was added in quadrature to the 2012 XIS spectral bins in order to absorb instrumental calibration uncertainties. The response matrix files (RMFs) and auxiliary response file (ARFs) for the XIS were produced using the FTOOLs xisrmfgen and xissimarfgen (Ishisaki et al., 2007), while standard detector response files (epoch 11) were used for the HXD-PIN.

## 3. Suzaku Observations of 4U 1954+319

### 3.1. Long-term variability

Figure 2 shows the long-term variability of the X-ray intensity of 4U 1954+319, covering 17 years since 1994. X-ray count rates were re-binned to 6 and 12 days resolution and converted into the mCrab unit using conversion factors provided by the RXTE/ASM and Swift/BAT teams444The long-term RXTE/ASM monitoring data are provided by the RXTE/ASM team (http://xte.mit.edu), while the Swift/BAT transient monitor results provided by the Swift/BAT team (http://heasarc.gsfc.nasa.gov/docs/swift/results/transients/). The reported conversion factors, assuming a Crab-like spectrum, are and , for ASM and BAT, respectively.. As shown in the 20 d running average of the fluxes (Fig. 2, blue curve), the X-ray intensity of 4U 1954+319 is highly variable on a time scale of years, with fluxes changing from a few mCrab to at least 20 mCrab or more. In addition, faster variability on timescales of weeks or days is also seen with irregular flares.

Figure 2 also shows the spin history of 4U 1954+319 as measured with Swift/BAT. We determined the pulse periods using standard epoch folding (Leahy, Elsner, & Weisskopf, 1983) and using pulse profiles with 19 phase bins. Spin values are shown with filled circles if pulses are detected with . The measured period history is consistent with previous work covering data taken before 2010 (Marcu et al., 2011). The present work extends the analysis until the end of 2012 and shows the detection of a new change from spin-down to spin-up which occurred around 2012 March (MJD56000), when the average X-ray intensity increased to 25 mCrab. The analysis shows that the first and second Suzaku observations were taken during the spin-down and spin-up phases, respectively. The former observation was almost in quiescence (0.3 mCrab), before the outburst, while the latter was during a brighter phase after the flaring (10 mCrab) and the spin-transition.

### 3.2. Timing analysis

Figure 3 shows the XIS and HXD-PIN light curves of 4U 1954+319 from the two observations. Similar to the behavior observed on long timescales (Fig. 2), the X-ray light curve is highly variable also on short time scales. Figure 4 further illustrates the enlarged light curve. The X-ray fluctuation was moderate in 2011, while spiky flare events were seen in 2012 with 100 s scale, showing a rapid flux change by an order of magnitude or more. During the 2012 outburst (Fig. 4), the spiky flares become enhanced in the latter part of observation. An increase of the short-term fluctuation is also seen in the power spectral density (PSD) in Figure 5. The PSD continuum can be described by a power law components with best fit slopes of in 2011 and in 2012. There are no signatures of clear short-term periods or quasi periodic oscillations. No low-frequency roll over was detected, either.

Epoch folding of the 2012 XIS and HXD data yielded a spin period of  h. However, the period determination with Suzaku is not accurate enough, due to the limited exposure which corresponds to only 6 cycles of the NS rotation. Thus, we searched the Swift/BAT monitoring data (§3.1) for the spin period using intervals of  d and  d around the Suzaku pointings. The pulse periods derived from these data are  h () and  h (), in 2011 and 2012, respectively, where 15 phase bins were used for the epoch folding. Average pulse profiles, obtained at these periods, are superposed in Fig. 3 on the Suzaku light curves. Even though sporadic flares are rather strong, we can trace individual pulses in most cases.

Figure 6 shows the pulse profiles of the two Suzaku observations. The pulse profile in 2012 is approximately sinusoidal below 1 keV, with a single peak at phase 0.4. A second peak at phase 0.2 appears at higher energies. Figure 7 shows the energy dependence of the pulsed fraction of 4U 1954+319, defined as , where and are the maximum and minimum of the pulse profile. The fraction is 60–80% in the 0.5–30 keV band, and increases with energy. This is a typical trend often seen in X-ray pulsars in HMXBs (Lutovinov & Tsygankov, 2008).

### 3.3. X-ray spectrum during the flare in 2012

Figure 8a shows Suzaku spectra of 4U 1954+319 obtained on the two occasions. We superpose the modeled HXD-PIN NXB and CXB, together with the typical 1% systematic uncertainty of the NXB. The HXD-PIN detection is marginal in 2011. The ratio between the two spectra is shown in Fig. 8b. In 2012, 4U 1954+319 was brighter by nearly 1–2 orders of magnitude, and exhibited harder spectra.

We now turn to modeling the X-ray spectrum in Figure 9, and summarize the best fit parameters in Table 2. Fitting the data with a single, absorbed power law leaves a soft excess below 2 keV, an iron line at 6.4 keV, and a spectral roll over (Fig. 9a). Here we use the latest interstellar absorption model tbnew (Wilms et al., 2000) with an improved abundance wilm and cross section tables vern (Verner et al., 1996). Due to higher data quality compared to earlier INTEGRAL measurements, the broken power-law model used by Fürst et al. (2011) and Marcu et al. (2011) was not successful (), either. The exponentially cutoff power-law with an additional soft blackbody and an Fe-line, improved the fit to (Fig. 9b) with a photon index of , a cutoff energy  keV, and a blackbody temperature of  keV. The spectral parameters are similar to those found by Mattana et al. (2006) by applying the same model to the BeppoSAX data, although our fit is not satisfactory.

As alternative representations of the roll-over of the continuum, we employed two empirical models: 1) a high energy cutoff model of the form (in XSPEC notation)

where

 highecut={1for E

with the cutoff energy  keV and e-folding energy  keV. We also modeled the data using the 2) a Negative and Positive power law with an EXponential (NPEX),

which is often utilized to represent broad-band continuum from the accretion column of highly magnetized NSs (Mihara, 1995; Makishima et al., 1999; Enoto et al., 2008). The cutoff energy and the photon index of the NPEX continuum were obtained as 6.6 keV and 0.77, respectively, and the blackbody temperature as 0.14 keV. As shown in Fig. 9c, d, and Table 2, both models reproduced the broadband spectra quite well (). If we exclude the soft blackbody, becomes worse (). Replacing the black body component with a diffuse plasma emission model (apec555Astrophysical Plasma Emission Code (APEC), a collisonally-ionzied diffuse gas model http://atomdb.org/; Smith et al. 2001), did not give acceptable fits ().

Finally, as a more physically-based interpretation, we also tried an optically thick Comptonizaiton model, compTT (Titarchuk & Lyubarskij 1995). This gave an acceptable fit, , with a null hypothesis probability of 0.02 (Fig. 9e). We therefore adopt this model as a canonical continuum model of 4U 1954+319. The derived best-fit parameters are a soft photon (Wien) temperature of  keV, plasma temperature  keV, and an optical depth of . These derived and values are within the typical range of the comptt model of this source,  keV and  keV, by the multi X-ray satellite studies (Masetti et al., 2007b) and by the INTEGRAL observation (Marcu et al., 2011). Combining the previous two successful models, the broad-band spectrum is explained by the Comptonized model. Figure 10 gives a representation of the data and the model. The 2–10 keV and 1–70 keV absorbed X-ray fluxes are and , translating to absorption-corrected X-ray fluxes of and , respectively.

### 3.4. Iron line and search for CRSF

The 6.4 keV neutral Fe K line emission is obvious in Fig. 9a. In the compTT continuum fits, its center energy was found at keV with an equivalent width of  eV and a total line flux of . The line is narrow with a width  eV (), corresponding to a Doppler width .

Detailed inspection of the spectrum indicates a shallow dent at 7.1 keV, and if we let the iron abundance free, the data favor the 2.4-times larger iron abundance relative to the default improving the fit by for (F-test probability of ). Then, when we try various abundance tables (e.g., angr, Grevesse & Anders 1989; aspl, Asplund et al. 2009) with the phabs absorption model (Balucinska-Church & McCammon, 1992), the same comptt continuum changes up to by 50%. Under these uncertainties, we assume  cm, derived from the best-fit comptt model, as the default value in the discussion666If the Fe-K line is due to fluorescence from an optically thin gas which spherically surrounds the source, its equivalent width is expected to be linearly correlated with the column density, as (Inoue, 1985). The line could therefore be produced in a gas with a column density of .. Subtracting the 21 cm interstellar column density in the direction to 4U 1954+319, (Kalberla et al., 2005), the intrinsic column density in the binary system is .

We also search the HXD-PIN data for a possible presence of a cyclotron resonance scattering feature (CRSF), which provides a direct probe of the -field at the accretion column (Makishima et al., 1999; Schönherr et al., 2007). Although a shallow structure at 40 keV is visible, the depth of the absorption (cyclabs) was constrained to (1). Thus, no CRSF was found in the Suzaku data.

### 3.5. X-ray spectrum in the 2011 quiescent state

In the first Suzaku observation performed in 2011 before the flare, the source was fainter by more than an order of magnitude, with the absorbed 2–10 keV X-ray flux (comptt model). Since the HXD-PIN detection is marginal, we only took the XIS-FI and BI spectra into account for this data set. Table 2 summarizes the spectral models used to fit the data. Simple power-law and blackbody models failed with , while the diskbb, cutoffpl, pow*highecut, and compTT models were all acceptable. Using the same comptt model, the unabsorbed 1–10 keV flux was . The 1 upper-limit intensity of a narrow ( eV) Fe K line at 6.4 keV was .

### 3.6. Luminosity and mass-accretion rate

As seen above, between the quiescent phase in 2011 and the outburst phase of 2012 the time-averaged X-ray flux of 4U 1954+319 changed by a factor of about 100. Assuming a distance of 1.7 kpc, the fluxes observed in 2011 and 2012 correspond to 1–70 keV luminosities of 0.022 and 2.1 with . Correcting for the X-ray absorption increases these luminosities to 0.025 and 2.2, respectively. These values fall in the luminosity range typically seen in SyXBs, (Lü et al., 2012).

We can estimate the accretion rate onto the NS as

 ˙MNS=RNSηGMNSLX={4.7×1013gs−1in 20113.8×1015gs−1in 2012 (4)

where is the efficiency of accretion,  km is the NS radius, and is the NS mass.

Since 4U 1954+319 is a highly variable source (§3.2), we further investigated, in Figure 11, distributions of count rates in the XIS and HXD-PIN light curves, referring to Vela X-1 studies by Fürst et al. (2010). The count rate distributions closely follow a log-normal distribution in both the quiescent phase (2011, Fig. 11a) and the outburst (2012, Fig. 11a and b). Assuming the 2012 time-averaged spectral shape, we further converted these to luminosity at the top of Fig. 11a. As previously shown in Fig 8, the spectrum becomes harder as the X-ray luminosity increases from the quiescent phase (2011) to the outburst (2012). Such a hardening trend is also detected within the 2012 observation when we divide the data into three intensity-sorted spectra (Fig. 12).

## 4. Discussion

### 4.1. Summary of the observations

Combining the Swift/BAT and RXTE/ASM long-term monitoring with the Suzaku observations, we derived the following observational results which need to be explained:

1. The slow 5.4 hour NS spin period of 4U 1954+319 exhibits a 7% fluctuation over 8 years with at least four reversals in the sign of the period derivative.

2. In the 2011 October and 2012 November Suzaku observations, before and after the 2012 November X-ray outburst, the 1–70 keV X-ray luminosities were and , respectively. In the former and the latter observations, the source was in a spin-down and a spin-up phase, respectively.

3. The 0.5–70 keV X-ray pulse profile shows a main peak and a sub peak. The pulsed fraction mostly increases with energy from 60% at 3 keV to 80% at 30 keV.

4. During the 2012 outburst, the X-ray flux was highly variable on 10–1000 s time scales, with many irregular short flares each with a typical duration of 100 s. The Fourier power density spectrum is dominated by red noise, i.e., became redder from in 2011 to in 2012. The X-ray count rates follow a log-normal distribution in both observations.

5. The 0.8–70 keV phase-averaged spectra in the 2012 outburst can be represented by the Comptonization spectral models; compTT, BB+NPEX, or BB+highecut. A fluorescent Fe-K line at 6.37 keV is detected with an equivalent width of , and a line flux of . The line is narrow, with a upper limit for the Gaussian width of 50 eV. The quiescent 2011 spectrum is also well described by comptt with   keV, or also by diskbb, cutoffpl, and highecut models.

6. The 2012 XIS spectrum is harder than that measured in 2011. The intensity-sorted X-ray spectra in 2012 slightly harden with source brightness.

7. The 1–70 keV band does not contain any statistically significant cyclotron lines.

### 4.2. Are SyXBs a subclass of LMXB?

X-ray binaries hosting NSs are conventionally classified based on stellar types of the optical counterparts. Commonly, HMXBs host a strongly-magnetized NS and a high-mass companion (i.e., Be stars or OB supergiants), while LMXBs are thought to host a weakly magnetized NS and a low-mass star (e.g., K-type stars). However, this old optical identification ignores accretor properties (e.g., NS -field) which mainly determine the X-ray radiation. In fact, the conventional stereotype assuming weakly-magnetized NSs in LMXBs is now challenged by recent observational evidence of strongly magnetized NSs in LMXBs, e.g., 4U 1626-67 (Camero-Arranz et al., 2012) and 4U 1822-37 (Sasano et al., 2013).

The required overhaul of this outworn notion is more clearly shown in SyXBs. In contrast to the conventional classification which makes SyXBs the same category as LMXBs, the apparent high pulsed fraction (60–80%) of SyXBs (Fig. 7, §4.1, point 3) suggests a similarity to strongly-magnetized and highly-pulsed NSs in HMXBs (pulsed fraction 20–100%, Lutovinov & Tsygankov 2008) rather than to non-pulsed NSs in LMXBs or to weakly-pulsed millisecond X-ray pulsars (typical pulsed fraction at a few percent, Patruno & Watts 2012). In addition, if we look at the location of SyXBs in the “Corbet diagram” (Corbet, 1986) shown in Figure 13, SyXBs again resemble HMXBs due to their wide-orbit and slow NS rotation. Finally, SyXB X-ray spectra, and that of 4U 1954+319 in particular, and the other behavior summarized in §4.1 also suggest a closer similarity of 4U 1954+319 to X-ray pulsars in HMXBs with a  G -field.

Even though many of these points argue towards a closer similarity of SyXB and HMXB, such an association would be a significant challenge to our current understanding of X-ray binary evolution. Stellar evolution theory shows that the evolutionary time scale for the companion of the NS to become an M4-5 III counterpart is much longer than the typical lifetime of binary systems with OB-type donors, and thus, the 4U 1954+319 system would be old. If one dogmatically expected NS -field decay, the high -field indicated by the strong X-ray pulsation easily causes a contradiction to the assumption of an old binary system. Furthermore, it is also difficult to keep such a binary system after a supernovae explosion, since the common envelop hypothesis, which is usually assumed to make close binary systems hosting NSs in LMXBs, cannot be reconciled with the suggested large orbit of SyXBs (Fig. 13). Te conventional evolutional theory leaves us with a real conundrum.

In addition to needing to study the population synthesis of such binaries (see, e.g., Lü et al., 2012; Chakrabarty & Roche, 1997), the observed nature of SyXB requires us to investigate more exotic scenarios: e.g., a magnetar-descendent was captured by an evolved M-type giant in a close encounter via a magnetic braking interaction between the NS strong -field and a large atmosphere of the M-type giant. An alternative possibility is an accretion-induced collapse (AIC) of a white dwarf (Nomoto & Kondo, 1991) which potentially produces not only weak-field pulsars but also a magnetar (Thompson & Duncan, 1995).

In the following, we take the features of 4U 1954+319 listed in §4.1 as observational clues to the mystery of SyXB, comparing the conventional accretion model and the revisited quasi-spherical accretion model, regardless of the conventional HMXB and LMXB classification and population synthesis.

### 4.3. The X-ray Spectrum of 4U 1954+319

The first clues on the nature of 4U 1954+319 come from broad-band spectroscopy. We assume that the observed X-ray spectrum is uniquely characterized mainly by , -field, and inclination of the compact object regardless of the type of the mass-donor. As shown in §2.3 and previously suggested by Masetti et al. (2007b), spectral modeling implies that Comptonization is the dominant process forming the X-ray continuum, similar to the accretion column of the canonical X-ray pulsars in HMXBs777Note that previous observations also included a soft excess which could be described with plasma emission with  eV (Masetti et al., 2007b) which is not seen here since the Suzaku low energy data below 1 keV have uncertainties due to the contamination on the optical blocking filter.. Thus, in Figure 14, we now compare the X-ray spectrum of 4U 1954+319 with different X-ray spectral shapes of other related X-ray pulsars.

Figure 14a shows a comparison of the X-ray spectrum of 4U 1954+319 with all archival Suzaku SyXB observations: GX 1+4, 4U 170024 (Nagae et al., 2008), and IGR J161942810 (Kitamura et al., 2013). The rotational, orbital, and optical information of these X-ray sources is listed in Table 3 with the Suzaku observation records and references. To extract these spectra we use the same procedure as for 4U 1954+319 (§2.2). The spectra were obtained by fitting the data with a cutoffpl model with interstellar absorption (phabs), and, if needed, an iron line. GX 14 exhibits harder spectra peaking at 20–40 keV than 4U 1954+319. The spectral shape of IGR J161942810, for which orbital and pulse periods have not yet been detected, is similar to that of 4U 1954+319 in quiescence, and thus their accretion environment is expected to be similar.

In many accreting NSs with -fields in the  G regime, cyclotron lines at energies  keV have been seen, where is the gravitational red shift of the NS (Caballero & Wilms, 2012; Schönherr et al., 2007). In Fig. 14b we compare 4U 1954+319 with well-studied CRSF sources. The 1.24 s X-ray pulsar Her X-1 is a prototypical CRSF source with  keV ( G) with an unusual optical companion (A9–B companion with a mass of ). The HMXBs 1A 1118616 and A 0535+262 exhibit the strongest -fields among known CRSF sources, reaching  G and  G, respectively. IGR J163934643 was originally classified as a SyXB, but finally identified as a HMXB (Bodaghee et al., 2012; Pearlman et al., 2011). As shown in Fig. 14b, the spectral shape of 4U 1954+319 is similar to 1A 1118616 and A 0535262 rather than Her X-1. Although a widely accepted physical model has not yet been established to explain the X-ray continuum, it has been suggested that due to cyclotron cooling the high energy cutoff appears around the CRSF energy. This means that the X-ray continuum is expected to extend to higher energies as the -field becomes stronger (Makishima et al., 1999).

The non-detection of a CRSF in 4U 1954+319 with the XIS and HXD means that either the -field is not in the range G, or that the cyclotron line is too faint to be detected in the data. The latter is not too unlikely, with a number of sources with similar behavior being observed. A possible explanation for this is that CRSFs can be smeared out and/or filled by photon spawning, i.e., the emission of photons close to the resonance energy during their de-excitation from higher Landau levels (Schönherr et al., 2007; Fürst et al., 2011). If, on the other hand,  G, according to Makishima et al. (1999) we would expect a harder X-ray continuum than that of HMXB with a cyclotron line.

In Fig. 14c we compare the spectrum of 4U 1954+319 with those of some HMXBs which have not shown a clear CRSF: HMXB 4U1909+07, 4U 2206+54, and 4U 0114+65. The phase-averaged spectral shapes of these sources are quite similar to that of 4U 1954+319. From a spectral view point they can be regarded as triplets. 4U 1909+07 is a HMXB with an OB companion hosting a 604-sec X-ray pulsar in a 4.4 d orbit. Its pulse period is reported to show a random walk like behavior (Fürst et al., 2011), which also resembles the period history of 4U 1954+319, and possibly suggesting wind type accretion rather than the stable accretion disk. The other two X-ray pulsars, 4U 2206+54 and 4U 0114+65, exhibit very long rotational periods of 5500 s and 9700 s, respectively (Finley et al., 1992; Reig et al., 2009; Masetti et al., 2006a). Thus, we expect that such a long rotational period is strongly related to the spectral shape in Fig. 14c of these sources.

Let us finally compare, in Fig. 14d, the spectrum with the low luminosity () low/hard state of LMXBs as motivated by Kitamura et al. (2013) who tried to search for similarities between SyXBs and low-field LMXB-NSs (e.g., Aql X-1) rather than high-field NSs, based on the optical classification. Although the -field of such LMXBs starts to dominate the accretion flow and an optically thin spherical flow would realize a similar geometry as the SyXB, as we already discussed, SyXBs should be different from other LMXBs due to 1) slow pulsation with a large pulsed fraction (Fig. 6) suggesting a high -field, 2) the large orbital period (Fig. 13) suggesting wind-type accretion, 3) the lack of clear evidence on the accretion disk, and 4) no observations of LMXB-like strong soft blackbody spectra, and 5) the strong narrow 6.4 keV iron fluorescence line which is more typically found in HMXBs but rare (10%) in standard LMXBs (Torrejón et al., 2010). In addition, LMXBs are usually characterized by highly ionized 6.6–6.9 keV lines.

In conclusion, the broad-band 4U 1954+319 spectrum best resembles that of X-ray pulsars in HMXBs, especially slowly rotating pulsars (e.g., 4U 0114+65 or 4U1909+07) without an apparent accretion disk as well as another quiescent SyXB, IGR J161942810. Therefore, a strong field or higher is favored rather than a low field as in non-pulsating or weakly-pulsed NSs in LMXBs.

### 4.4. The long spin period of 4U 1954+319

Let us now investigate the timing behavior of the source. As already discussed previously (e.g., Marcu et al., 2011; Corbet et al., 2008, and references therein), the 5.4 h period of 4U 1954+319 can only be explained as being the rotation period of the NS. Both an orbital modulation or an M-type star pulsation, can be ruled out. The long-term pulse history implies at least four spin-torque transitions between a spin-up and a spin-down phase. The period evolution indicates that outside the outbursts the NS is gradually slowing down, while during the outbursts it enters a spin-up trend correlated with high accretion rates. This behavior suggests that it is reasonable to assume that the source is near an equilibrium period of accretion momentum transfer.

In the standard disk accretion model (Ghosh & Lamb, 1979), the equilibrium period is Eq. (76) in Shakura et al. (2012)

 Peq≈7μ6/730˙M−3/716 s, (5)

where is the mass accretion rate, and is the neutron star’s magnetic dipole moment. Inserting the values for 4U 1954+319 measured in 2012, i.e., and  s (5.4 hour) yields an extremely high, magnetar-like -field reaching  G. This evaluation would hold even if considering the long-term history, since the average mass-accretion rate is close to the value in the 2012 Suzaku observation. This is clearly shown in Figure 15 where the long-term distribution of mass-accretion rates is illustrated estimated from Fig. 2.

While a magnetar-like -field has been suggested for the SyXB IGR J163584726 (Patel et al., 2007), the field derived above is an order of magnitude higher than the strongest magnetar field (SGR 180620, with ; Kouveliotou et al., 1998; Enoto et al., 2010b). 4U 1954+319 also does not show any other magnetar-like properties, such as short bursts or giant flares (Enoto et al., 2009). In addition, it is unclear whether standard disk theory can be applied to SyXBs. For example, González-Galán et al. (2012) show that the spin and X-ray intensity behavior of the SyXB GX 1+4 cannot be explained either by standard disk accretion theory or even by more advanced ideas such as accretion from a retrograde disk. We therefore have to also consider alternative accretion models.

As the Corbet diagram (Fig. 13) implies large orbital periods of SyXBs (e.g.,  d for GX 1+4; Hinkle et al. 2006), it is reasonable to assume that 4U 1954+319 does not accrete via Roche lobe overflow but is a wind-fed accretor. The large orbital period leads to a slow NS orbital velocity888 Although the orbital radius of 4U 1954+319 has not yet been detected, assuming a circular orbit with its orbital period at 500 d, we will derive the semimajor axis at cm via the Kepler’s law and masses of the NS and M giant. Then the orbital velocity becomes 30 km s. . Combining this with a slow stellar wind compared to HMXBs and with low mass-accretion rate, small angular momentum differences are an intrinsic importance of the SyXBs which makes the mass-accretion more quasi-spherical.

As recently shown by Shakura et al. (2013), the X-ray luminosity of 4U 1954+319 actually puts the source in the regime of quasi-spherical, subsonic setting accretion in the magnetosphere. Figure 16 shows a sketch of this alternative wind accretion scenario for 4U 1954+319, which is possible for and . Similar to the standard Bondi-Hoyle-Lyttleton accretion, in this scenario a bow shock is formed in the wind at a characteristic distance () from the NS and matter entering this region is accreted onto the compact object. Unlike the classical Bondi-Hoyle-Lyttleton accretion at higher luminosities, for the low of 4U 1954+319, Compton cooling becomes ineffective and the matter can settle in the magnetosphere with a subsonic velocity through a quasi-spherical hot shell (§1). If this happens, the angular momentum can be transferred via large-scale convective motions or turbulence in this quasi-static shell. This results in an equilibrium spin period of (Postnov et al., 2012)

 Peq≃13000 μ12/1130˙M−4/1116v48(Porb100d)s, (6)

where ) is the relative velocity between the stellar wind and the NS.

The most uncertain parameter in Eq. (6) is . The wind velocity strongly affects , but it is not well known for the late-type M giants due to lack of observations. Existing measurements for wind speeds are mainly available for massive OB or Wolf-Rayet stars with high mass-loss rates (). These radiatively-driven winds are very fast (up to ; Lamers & Cassinelli 1999). The situation is different for SyXB systems. Here no strong evidence for fast winds exists. Late F–M stars are usually thought to have smaller mass-loss rates () with much slower terminal velocities (; Espey & Crowley, 2008). These slower speeds are consistent with the lower escape velocities of the stars. For example, if the donor in 4U 1954+31 has and a radius of as suggested by Mattana et al. (2006), then the escape velocity from the companion star would be only . To our knowledge, the only direct measurement of the wind speed in a SyXB was performed for the M6-giant V2116 Oph, the donor star of the SyXB GX 1+4, where the wind speed was found to be 100–200 (Chakrabarty et al., 1998; Hinkle et al., 2006).

Assuming that the wind in 4U 1954+319 is comparable to that in GX 1+4, Figure 17 shows the dependence of the equilibrium period for several plausible different combinations of , , , and as calculated with Eq. (6). If 4U 1954+319 has a typical accreting NS with a -field of  G, then its equilibrium period can only be explained by a fast M-star wind (e.g., 300 km s, probably even faster than that observed in GX 1+4) and the long orbital period (e.g., 500 d).

### 4.5. Time scales of Irregular flares

A further handle on the accretion mechanism at work in 4U 1954+319 comes from the short term variability of the source. The quiescent state in 2011 shows a steady weak accretion (in Fig. 4 a,b), while the outburst in 2012 exhibits many irregular flare on time scales of 100 s (Figs. 3 and 4 c, d). This highly variable fluctuation indicates a blobby mass accretion (Fig. 16). The X-ray light curves show that the X-ray variation is mainly due to short-time scale spiky flares. Similar spike-like structures, also called “shots”, were studied previously in analyses of the light curves of accreting black holes such as Cygnus X-1 (Negoro et al., 1994; Yamada et al., 2013). In these objects the spectrum gradually softens during the peak of a shot, followed by an immediate hardening. Figure 18 shows a similar analysis for 4U 1954+319. The shots in the light curves are represented by a shape with nearly symmetric rise and decay (Fig. 18 a, b, c) on a timescale of 50 s (at which the count rate drops below the average) with signs of a slight hardening in the early phase of the shot-like behavior which is indicated in (Fig. 18 d, e).

A possible explanation for such shots in 4U 1954+319 are individual blobby structures in the accretion flow. To see whether this interpretation is reasonable, we need to take a look at the typical size and time scales of this system (see Fig. 17 right panel for a visualization). In the model by Postnov et al. (2012) the stellar wind is gravitationally captured by the NS inside the Bondi radius,

 RB=2GMNSv2=3.78×1010v−28cm. (7)

For , i.e., a stellar wind velocity of , the Bondi radius  cm. Inside this radius a quasi-static shell is formed above the magnetosphere. This is the reservoir of matter from which the NS accretes. In this model the co-rotation radius is (Postnov et al., 2012)

 Rc=(GMP24π2)1/3=1.21×1011cm(P5.40h)2/3. (8)

Even though the co-rotation radius is very large, it is still inside the Bondi radius if the stellar wind is slower than (Fig. 17 right). This sets an upper limit for the wind speed in the system.

The size of the magnetosphere, characterized by the Alfvén radius where the gas pressure becomes equal to the magnetic pressure (Ghosh & Lamb, 1979), is modified for quasi-spherical accretion at low mass accretion rates and becomes (Postnov et al., 2012; Shakura et al., 2013)

 RA∼1.6×109cm(μ330˙M16)2/11. (9)

With we find  cm for  G, and  cm for G, respectively. This means that the Alfvén radius is smaller by 1–2 orders of magnitudes than and .

This estimate for the size of the Alfvén radius is roughly consistent with that found from spectral analysis. While the 2012 spectrum is Comptonized, during the 2011 quiescent state Comptonization does not play a strong role. Approximating the quiescent spectrum by a blackbody and using the Stefan-Boltzmann law, the hot spot size on the NS is . Since the hot spot size is determined by the location where matter couples onto the -field, we can use this measurement to estimate the size of the Alfvén radius (Lamb et al., 1973),

 RA≲105km(Rhot100m)−2∼5.9×109cm. (10)

The measured hot spot radius thus yields a -field in the  G range (Fig. 17).

Depending on the ability of the plasma to enter the magnetosphere, matter from the magnetosphere funnels onto the NS poles along the -field lines. The free fall time is given by

 tff=51s(r1010cm)3/2. (11)

The free-fall timescale from the Alfvén radius for  G is 30 s, close to the typical time scale of individual shots (Fig. 18). Thus, it is possible to interpret the shot-like flares as produced during the free-fall of the blobs from the shell above the magnetosphere. Even though the cooling time scale of the matter is complicated when considering different cooling processes, if we assume Compton cooling as an effective process (Shakura et al., 2012), the cooling time,

 tcool=103 s⋅1˙M16(r1010 cm)2 (12)

becomes 1200 s at G and . This is longer than the typical free-fall time () which satisfies the assumption of a low-luminosity quasi-spherical accretion.

The log-normal distribution of the X-ray light curve count rates (Fig. 11) is also consistent with the picture of quasi-spherical magnetospheric accretion. As suggested by Fürst et al. (2010), such a log-normal distribution implies that the underlying physical process should be multiplicative rather than additive, since addition of many independent emissions will produce a Gaussian distribution due to the central limit theorem. Such a process could be induced via a positive feedback loop, where blob accretion induces subsequent accretion, e.g., via enhanced Compton cooling, potentially creating positive feedback until the matter at the boundary of the magnetosphere is exhausted. Alternatively, the structures leading to a log-normal distribution of the accretion rate could also be due to a highly inhomogeneous and clumpy stellar wind from the M-giant companion or due to turbulent processes in the spherical shell itself.

### 4.6. Absorption in the spherical shell

Finally let us compare the above accretion scheme with the absorption properties of the circumstellar material. These come from both the absorption modeling of the X-ray continuum and the analysis of the narrow Fe K line at 6.4 keV. The latter was detected with Suzaku in the 2012 observation, and has been previously reported with BeppoSAX ( eV) and RXTE ( eV; Masetti et al. 2007b). A narrow iron line feature is characteristic to HMXB-NSs rather than LMXB-NSs.

As a plausibility check, we assume that is directly related to the properties of the shell in the low-luminosity wind accretion model. Using the Suzaku-derived , the number density () inside the spherical shell () becomes (), corresponding to a mass density of where is the proton mass. If the NS is fed by blobs of matter which have the density of the shell at the Alfén radius, then in order to reach the average mass accretion rate where and are the typical blob size and shot time scale, we find  cm with and  s. It is intriguing that this initial blob radius is comparable to the Alfvén radius (e.g., cm at G) and much smaller than the quasi-spherical shell size . Thus, the shell can store enough mass to generate the observed multiple flares and sustain accretion for longer time intervals.

Finally, as a cross check we note that the ionization parameter of the shell at the Bondi radius is , i.e., the outer part of the shell is at most moderately ionized, making it possible to produce the 6.4 keV line in the shell.

## 5. Conclusion

We analyzed two Suzaku data sets, obtained in 2011 and 2012, of the SyXB 4U 1954+319 consisting of the slowly rotating 5.4 h X-ray pulsar and a late type primary star (M4-5 III). Our main results are as follows;

1. We reconfirmed the slow rotation period and the large period fluctuations. We also identified several additional features of 4U 1954+319, including a hard continuum without a clear CRSF, a narrow Fe-K line, and sporadic flares. The continuum was harder when the source was brighter.

2. From spectral and timing properties, 4U 1954+319 is suggested to have a strong -field of  G. This is supported by the high pulsed fraction of 60–80%, the slow rotation, the hard featureless continuum, and the narrow Fe-K line.

3. The NS is expected to be close to equilibrium because of the four spin transitions and large period fluctuations. Such an evolution can be explained by a quasi-spherical accretion in a subsonic accretion regime which is considered to apply to 4U 1954+319 because of its low luminosity ( erg s), presumably wide orbital period, and rather low wind velocity from the M-type giant.

4. Recurrent irregular flares during the outburst have a typical time scale of 50 s, which is interpreted as intermittent accretion from the Alfvén radius of the  G NS. The log-normal distribution of the light curve indicates an underlying multiplicative process (e.g., positive feedback) in the accretion.

5. The extreme magnetar-like extreme -field ( G) derived from the standard disk accretion in Eq. (5) is not required if we assume a low luminosity and quasi-spherical accretion.

6. The presence of a system like 4U 1954+319 which is very likely to involve a high-field NS, provides another example which urges us to challenge the conventional classification of NS-binaries into LMXBs and HMXBs.

The authors would like to express their thanks to Suzaku team for their prompt observation during the 2012 flaring activity. TE was supported by JSPS KAKENHI, Grant-in-Aid for JSPS Fellows, 24-3320. We thank Pranab Ghosh and Hiromitsu Takahashi for useful discussions on this source, Kunugawa Tomoya and Kenta Hotokezaka for comments on the binary evolution. We thank the Deutsches Zentrum für Luft- und Raumfahrt for partial funding under DLR grant number 50 OR 1207.

## References

• Asplund et al. (2009) Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481
• Balucinska-Church & McCammon (1992) Balucinska-Church, M., & McCammon, D. 1992, ApJ, 400, 699
• Bodaghee et al. (2007) Bodaghee, A., Courvoisier, T. J.-L., Rodriguez, J., et al. 2007, A&A, 467, 585
• Bodaghee et al. (2012) Bodaghee, A., Rahoui, F., Tomsick, J. A., & Rodriguez, J. 2012, ApJ, 751, 113
• Bozzo et al. (2008) Bozzo, E., Falanga, M., & Stella, L. 2008, ApJ, 683, 1031
• Caballero & Wilms (2012) Caballero, I., & Wilms, J. 2012, Mem. SA It, 83, 230
• Camero-Arranz et al. (2012) Camero-Arranz, A., Pottschmidt, K., Finger, M. H., et al. 2012, A&A, 546, A40
• Chakrabarty & Roche (1997) Chakrabarty, D., & Roche, P. 1997, ApJ, 489, 254
• Chakrabarty et al. (1998) Chakrabarty, D., van Kerkwijk, M. H., & Larkin, J. E. 1998, ApJ, 497, L39
• Corbet (1986) Corbet, R. H. D. 1986, MNRAS, 220, 1047
• Corbet et al. (2006) Corbet, R., Barbier, L., Barthelmy, S., et al. 2006, ATEL 797
• Corbet et al. (2008) Corbet, R. H. D., Sokoloski, J. L., Mukai, K., Markwardt, C. B., & Tueller, J. 2008, ApJ, 675, 1424
• Davidsen et al. (1977) Davidsen, A., Malina, R., & Bowyer, S. 1977, ApJ, 211, 866
• Enoto et al. (2008) Enoto, T., Makishima, K., Terada, Y., et al. 2008, PASJ, 60, 57
• Enoto et al. (2009) Enoto, T., Nakagawa, Y. E., Rea, N., et al. 2009, ApJ, 693, L122
• Enoto et al. (2010a) Enoto, T., Nakazawa, K., Makishima, K., et al. 2010a, PASJ, 62, 475
• Enoto et al. (2010b) Enoto, T., Nakazawa, K., Makishima, K., et al. 2010b, ApJ, 722, L162
• Espey & Crowley (2008) Espey, B. R., & Crowley, C. 2008, in ASP Conf. Ser. 401, 166
• Farrell et al. (2008) Farrell, S. A., Sood, R. K., O’Neill, P. M., & Dieters, S. 2008, MNRAS, 389, 608
• Finley et al. (1992) Finley, J. P., Belloni, T., & Cassinelli, J. P. 1992, A&A, 262, L25
• Forman et al. (1978) Forman, W., Jones, C., Cominsky, L., et al. 1978, ApJS, 38, 357
• Fukazawa et al. (2009) Fukazawa, Y., Mizuno, T., Watanabe, S., et al. 2009, PASJ, 61, 17
• Fürst et al. (2010) Fürst, F., Kreykenbohm, I., Pottschmidt, K., et al. 2010, A&A, 519, A37
• Fürst et al. (2011) Fürst, F., Marcu, D. M., Pottschmidt, K., et al. 2011, arXiv:1106.2708
• Fürst et al. (2011) Fürst, F., Kreykenbohm, I., Suchy, S., et al. 2011, A&A, 525, A73
• Garcia et al. (1983) Garcia, M., Baliunas, S. L., Elvis, M., et al. 1983, ApJ, 267, 291
• Ghosh & Lamb (1979) Ghosh, P., & Lamb, F. K. 1979, ApJ, 232, 259
• González-Galán et al. (2012) González-Galán, A., Kuulkers, E., Kretschmar, P., et al. 2012, A&A, 537, A66
• Grevesse & Anders (1989) Grevesse, N., & Anders, E. 1989, in AIP Conference Proceedings 183, 1
• Ikhsanov (2007) Ikhsanov, N. R. 2007, MNRAS, 375, 698
• Hickox et al. (2004) Hickox, R. C., Narayan, R., & Kallman, T. R. 2004, ApJ, 614, 881
• Hinkle et al. (2006) Hinkle, K. H., Fekel, F. C., Joyce, R. R., et al. 2006, ApJ, 641, 479
• Ho et al. (2014) Ho, W. C. G., Klus, H., Coe, M. J., & Andersson, N. 2014, MNRAS, 437, 3664
• Inoue (1985) Inoue, H. 1985, Space Sci. Rev., 40, 317
• Ishisaki et al. (2007) Ishisaki, Y., Maeda, Y., Fujimoto, R., et al. 2007, PASJ, 59, 113
• Kalberla et al. (2005) Kalberla, P. M. W, Burton, W. B., Hartmann, D. et al. 2005, å, 440, 775
• Kaplan et al. (2007) Kaplan, D. L., Levine, A. M., Chakrabarty, D., et al. 2007, ApJ, 661, 437
• Kitamura et al. (2013) Kitamura, Y., Takahashi, H., & Fukazawa, Y. 2014, PASJ, in press (arXiv:1308.5803)
• Klus et al. (2014) Klus, H., Ho, W. C. G., Coe, M. J., Corbet, R. H. D., & Townsend, L. J. 2014, MNRAS, 437, 3863
• Kouveliotou et al. (1998) Kouveliotou, C., Dieters, S., Strohmeyer, T. 1998, Nature, 393, 253
• Koyama et al. (2007) Koyama, K., Tsunemi, H., Dotani, T., et al. 2007, PASJ, 59, 23
• Lamb et al. (1973) Lamb, F. K., Pethick, C. J., & Pines, D. 1973, ApJ, 184, 271
• Lamers & Cassinelli (1999) Lamers, H. J. G. L. M., & Cassinelli, J. P. 1999, Introduction to Stellar Winds, by Henny J. G. L. M. Lamers and Joseph P. Cassinelli, pp. 452. ISBN 0521593980. Cambridge, UK: Cambridge University Press, June 1999.,
• Leahy, Elsner, & Weisskopf (1983) Leahy, D. A., Elsner, R. F., & Weisskopf, M. C. 1983, ApJ, 272, 256
• Li & van den Heuvel (1999) Li, X.-D., & van den Heuvel, E. P. J. 1999, ApJ, 513, L45
• Liu et al. (2006) Liu, Q. Z., van Paradijs, J., & van den Heuvel, E. P. J. 2006, A&A, 455, 1165
• Liu et al. (2007) Liu, Q. Z., van Paradijs, J., & van den Heuvel, E. P. J. 2007, A&A, 469, 807
• Lü et al. (2012) Lü, G.-L., Zhu, C.-H., Postnov, K. A., et al. 2012, MNRAS, 424, 2265
• Lutovinov & Tsygankov (2008) Lutovinov, A., & Tsygankov, S. 2008, American Institute of Physics Conference Series, 1054, 191
• Maeda et al. (2008) Maeda et al. 2008, JX-ISAS-SUZAKU-MEMO-2008-06 http://www.astro.isas.jaxa.jp/suzaku/doc/suzakumemo/suzakumemo-2008-06.pdf
• Makishima et al. (1988) Makishima, K., Ohashi, T., Sakao, T., et al. 1988, Nature, 333, 746
• Makishima et al. (1999) Makishima, K., Mihara, T., Nagase, F., & Tanaka, Y. 1999, ApJ, 525, 978
• Marcu et al. (2011) Marcu, D. M., Fürst, F., Pottschmidt, K., et al. 2011, ApJ, 742, L11
• Masetti et al. (2002) Masetti, N., Dal Fiume, D., Cusumano, G., et al. 2002, A&A, 382, 104
• Masetti et al. (2006a) Masetti, N., Orlandini, M., dal Fiume, D., et al. 2006a, A&A, 445, 653
• Masetti et al. (2006b) Masetti, N., Orlandini, M., Palazzi, E., Amati, L., & Frontera, F. 2006b, A&A, 453, 295
• Masetti et al. (2007a) Masetti, N., Landi, R., Pretorius, M. L., et al. 2007a, A&A, 470, 331
• Masetti et al. (2007b) Masetti, N., Rigon, E., Maiorano, E., et al. 2007b, A&A, 464, 277
• Masetti et al. (2012) Masetti, N., Nucita, A. A., & Parisi, P. 2012, A&A, 544, A114
• Mattana et al. (2006) Mattana, F., Götz, D., Falanga, M., et al. 2006, A&A, 460, L1
• Mihara (1995) Mihara, T. 1995, Ph.D. Thesis, University of Tokyo
• Mitsuda et al. (2007) Mitsuda, K., Bautz, M., Inoue, H., et al. 2007, PASJ, 59, 1
• Moretti et al. (2009) Moretti, A., Pagani, C., Cusumano, G., et al. 2009, A&A, 493, 501
• Morrison & McCammon (1983) Morrison, R., & McCammon, D. 1983, ApJ, 270, 119
• Nagae et al. (2008) Nagae O., Takahashi H., Shirai H., Fukazawa Y., Proc. VII Microquasar Workshop: Microquasars and Beyond, PoS(MQW7) (Trieste: SISSA). 2008. p. 102.
• Negoro et al. (1994) Negoro, H., Miyamoto, S., & Kitamoto, S. 1994, ApJ, 423, L127
• Nespoli et al. (2010) Nespoli, E., Fabregat, J., & Mennickent, R. E. 2010, A&A, 516, A94
• Nomoto & Kondo (1991) Nomoto, K., & Kondo, Y. 1991, ApJ, 367, L19
• Patel et al. (2007) Patel, S. K., Zurita, J., Del Santo, M., et al. 2007, ApJ, 657, 994
• Patruno & Watts (2012) Patruno, A., & Watts, A. L. 2012, arXiv:1206.2727
• Pearlman et al. (2011) Pearlman, A. B., Corbet, R. H. D., Pottschmidt, K., & Skinner, G. K. 2011, AAS/High Energy Astrophysics Division, 12, #42.06
• Perna et al. (2006) Perna, R., Bozzo, E., & Stella, L. 2006, ApJ, 639, 363
• Postnov et al. (2010) Postnov, K., Shakura, N., González-Galán, A., et al. 2010, Proceedings of the 8th INTEGRAL Workshop ”The Restless Gamma-ray Universe”, id.15
• Postnov et al. (2012) Postnov, K., Shakura, N. I., Kochetkova, A. Y., & Hjalmarsdotter, L. 2012, Proceedings of the 9th INTEGRAL Workshop ”An INTEGRAL view of the high-energy sky”, id.22
• Reig et al. (2009) Reig, P., Torrejón, J. M., Negueruela, I., et al. 2009, A&A, 494, 1073
• Reig et al. (2012) Reig, P., Torrejón, J. M., & Blay, P. 2012, MNRAS, 425, 595
• Sakurai et al. (2012) Sakurai, S., Yamada, S., Torii, S., et al. 2012, PASJ, 64, 72
• Sasano et al. (2013) Sasano, M., Makishima, K., Sakurai, S., Zhang, Z., & Enoto, T. 2014, in press, arXiv:1311.4618
• Schönherr et al. (2007) Schönherr, G., Wilms, J., Kretschmar, P., et al. 2007, A&A, 472, 353
• Serlemitsos et al. (2007) Serlemitsos, P. J., Soong, Y., Chan, K.-W., et al. 2007, PASJ, 59, 9
• Shakura et al. (2012) Shakura, N., Postnov, K., Kochetkova, A., & Hjalmarsdotter, L. 2012, MNRAS, 420, 216
• Shakura et al. (2013) Shakura, N. I., Postnov, K. A., Kochetkova, A. Y., & Hjalmarsdotter, L. 2013, Physics Uspekhi, 56, 321
• Shakura et al. (2013) Shakura, N. I., Postnov, K. A., Kochetkova, A. Y., & Hjalmarsdotter, L. 2014, European Physical Journal Web of Conferences, 64, 2001
• Smith et al. (2001) Smith, R. K., Brickhouse, N. S., Liedahl, D. A., & Raymond, J. C. 2001, ApJ, 556, L91
• Smith et al. (2012) Smith, D. M., Markwardt, C. B., Swank, J. H., & Negueruela, I. 2012, MNRAS, 422, 2661
• Sugizaki et al. (2010) Sugizaki, M., Mihara, T., Nakagawa, Y. E., et al. 2010, The Astronomer’s Telegram, 2702, 1
• Takahashi et al. (2007) Takahashi, T., Abe, K., Endo, M., et al. 2007, PASJ, 59, 35
• Thompson & Duncan (1995) Thompson, C., & Duncan, R. C. 1995, ASP Conference Series in 72. 301
• Titarchuk & Lyubarskij (1995) Titarchuk, L., & Lyubarskij, Y. 1995, ApJ, 450, 876
• Torrejón et al. (2010) Torrejón, J. M., Schulz, N. S., Nowak, M. A., & Kallman, T. R. 2010, ApJ, 715, 947
• Tweedy et al. (1989) Tweedy, R. W., Warwick, R. S., & Remillard, R. 1989, Two Topics in X-Ray Astronomy, Volume 1: X Ray Binaries.  Volume 2: AGN and the X Ray Background, 296, 661
• Verner et al. (1996) Verner, D. A., Ferland, G. J., Korista, K. T., & Yakovlev, D. G. 1996, ApJ, 465, 487
• Warwick et al. (1981) Warwick, R. S., Marshall, N., Fraser, G. W., et al. 1981, MNRAS, 197, 865
• Wilms et al. (2000) Wilms, J., Allen, A., & McCray, R. 2000, ApJ, 542, 914
• Yamada et al. (2013) Yamada, S., Negoro, H., Torii, S., et al. 2013, ApJ, 767, L34
You are adding the first comment!
How to quickly get a good reply:
• Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
• Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
• Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters