2. The continuum spectrum

There are significant, systematic variations in the continuum spectrum with pulse phase. The shape of the continuum spectrum in these data may be explained satisfactorily by either a black body plus thermal bremsstrahlung or a Comptonization model. In the former case the black body may approximate the spectrum of radiation from the accretion column and the thermal bremsstrahlung component may be due to the extended stellar wind or possibly a shock above the neutron star's surface. The canonical cut-off power law does not fit the data well. The best-fitting model is a power law with a cyclotron scattering cut-off. This does not necessarily mean that a CSRF has been detected. The inferred cyclotron line energy is suspiciously close to the upper limit of the energy range used in fitting the spectra. This was also noted by Nagase et al. (1992). When the data are fitted to a power law with cyclotron absorption the apparent cyclotron energy varies with pulse phase. The variation of the cyclotron energy with pulse phase is determined by the structure of the accretion column. If most of the accretion takes place on the polar caps the luminosity of Cen X-3 can be close to the Eddington limit. Thus the accretion flow is dominated by radiation pressure. It is likely that the accreting matter is decelerated in a collisionless shock some distance above the surface of the neutron star. This would lead to a fan-beam emission pattern, with the pulse maximum occurring for viewing directions normal to the magnetic field. One would then expect the maximum depth of the cyclotron resonance to coincide with the pulse minimum.

The maxima of the cyclotron depth $\tau_{\rm Cy}$
and the cyclotron energy E_Cy appear to coincide with a bump in the falling part of the pulse profile's main peak. We could thus interpret this third peak as the phase when we see both poles at 90^o.