1. Motivation

Makishima et al. (1992) describe cyclotron scattering resonance features (CSRF) in nine X-ray binary pulsars. The discovery of seven of these was made possible by the improved energy resolution and effective area of the Ginga LAC Makino and the ASTRO-C Team (1987). A 110 keV CSRF has also been observed in the spectrum of the transient X-ray binary pulsar A 0535+262 Grove et al. (1995); Kendziorra et al. (1994). Even if this is the first harmonic of a less significant feature at $\sim$ 55 keV the magnetic field ( $\sim$ 4.5 x 10¹² G) is higher than any other derived for an accreting neutron star by CSRF observations. Detection of this feature was made possible by the high energy response of the TTM and HEXE instruments on board the Mir space station and the OSSE instrument on board the Compton Gamma Ray Observatory. It was expected that the large effective area, energy resolution, and high energy response of the RXTE PCA and HEXTE detectors combined would similarly increase the number of observed CSRF in accreting neutron stars from the ten detected so far.

Direct measurements of the magnetic fields of X-ray binaries are important because they provide information on the evolution of the systems. Additional measurements of pulsar magnetic fields are required to determine if these fields decay. A measurement of the field of the relatively young pulsar Cen X-3 would help settle the question of neutron star field decay and a CSRF observation is the only method which can make this measurement directly from X-ray observations. A direct measurement of the magnetic field from a CSRF would also place constraints on the parameters of the neutron star and the structure of the accretion column. The width of the CSRF would provide a measure of the spatial extent of the accretion column.

Mihara (1995) successfully modeled the spectra of several X-ray binaries with a continuum of the form $I(E^{-\alpha_1} +E^{\alpha_2})\exp(-{E\over kT})$
where $\alpha_2$
was fixed at the value 2. The first term represents the usual power law which is dominant below 10 keV and the second term has the form of blackbody radiation. This composite model approximates an unsaturated Comptonized spectrum. However he could not obtain a good fit to the Ginga data on Cen X-3 with a continuum model of this form. The fit could be improved either by allowing $\alpha_{2}^{}$ to vary or by including a CSRF with E_Cy $\sim$ 40 keV. In any case it was clear that, whatever the choice of continuum model, there are unexplained features in the spectrum in Cen X-3 cannot be described well by a simple exponential cut-off. In the OSO 8 spectra of Cen X-3 there was an apparent flattening of the spectrum at the pulse maximum White et al. (1983). It is possible that this was due to variations in cyclotron absorption.

There had been no direct observation of a CSRF in the spectrum of Cen X-3 apart from the tentative detections at 30 keV with Ginga Nagase et al. (1992) and at $\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}20\rm\ keV$
in the EXOSAT GSPC data (see Chapter 6). In each case the cyclotron line energy is suspiciously close to the upper limit of the detector's bandpass. Nevertheless, there are other reasons to believe that a CSRF may be present at that energy. In Chapter 5 I used the beat frequency mass accretion model (Shibazaki and Lamb (1987) and references therein) to explain the 40 mHz QPO in the BBXRT observation of Cen X-3 and derived a surface magnetic field of $\sim$ 2.6 x 10¹² G. I thus predicted that the observer frame energy of the associated CSRF would be $\sim$ 29 keV. For a luminous source such as Cen X-3 it is expected that there will be a stand-off radiative shock above the neutron star's surface Wang and Frank (1981) so that the magnetic field at the emission region will be smaller. Thus this value is an upper limit to the expected fundamental cyclotron resonance energy. There is thus reason to believe that Cen X-3 has a stronger magnetic field than the average ($\sim$ 1- 2 x 10¹² G), placing a CSRF near the upper end of the Ginga LAC's 1-37 keV energy range. This, coupled with a possible broadening of the feature to $\sim$ 20 keV due to a column accretion geometry, may explain why a CSRF has not been detected in Cen X-3 before. Makishima et al. (1992) found a proportionality between the cyclotron feature energy E_Cy and the cut off energy E_c derived from a model in which the high energy roll off is due to an exponential cut off. This suggests that the high energy roll off is due mainly to cyclotron opacity.

Using the relation E_Cy $\approx$ 2E_c of Makishima et al. we can estimate the magnetic fields of pulsars in which cyclotron lines have not been detected and thus predict where the cyclotron lines might be found. The apparent cut off energy E_c for Cen X-3 was found to vary with luminosity between 8 and 11 keV (White et al. (1983); see Chapter 6), suggesting E_Cy $\sim$ 2E_c $\sim$ 16- 22 keV. However, the luminosity dependence of E_c suggests that parameters of the neutron star atmosphere such as temperature and optical depth have some effect on the apparent cut off energy. Thus these values of E_c may not be inconsistent with E_Cy $\sim$ 20-30 keV.

Although Nagase et al. were able successfully to model the cut-off in the Ginga data with cyclotron scattering, the available bandpass did not allow them to detect the upturn in the spectrum that would be present if there is in fact a cyclotron absorption feature at 30 keV. The situation was worse with the EXOSAT GSPC. It seemed that the RXTE PCA and HEXTE in combination would be well suited to examine the 30-60 keV region. Detection of this upturn would confirm that cyclotron opacity is largely responsible for the cut-off.