3. Pulsar Geometry

The swapping of the two peaks in the pulse profile could be a geometrical effect. As the pulsar orbits the companion one magnetic pole comes into view while the other is obscured and is seen in reflection (medium component). However it is clear that there are at least two components required to describe the spectrum.

The ``flare'' might be due to the pulsar encountering a region of enhanced density in the stellar wind, causing an increase in the mass accretion rate. However the short duration of the ``flare'' requires that this accretion occurs directly to the pulsar and not through the accretion disk. This is due to the small radial drift velocities expected in the thin, optically thick disk model of Shakura and Sunyaev (1973). An increase in the mass accretion rate feeding the disk would be expected to be smeared out over a timescale of days. Another possibility is an enhancement in the rate of thermonuclear burning on the surface. While the instabilities that cause type I X-ray bursts on non-magnetic neutron stars are ruled out by Cen X-3's strong magnetic field (e.g. Lewin et al. (1995)), it might be possible for unstable carbon burning to to cause a flare lasting minutes to hours (Bildsten and Brown (1997); Bildsten (1995,1997)). However, given the much lower efficiency of thermonuclear fusion, the ``flare'' is easier to explain as an increase in the mass accretion rate.

The most likely explanation is that the ``flare'' is due to an impulsive increase in the accretion rate caused by a clump of matter in the inner accretion disk. The existence of such a clump might be expected to be accompanied by a QPO if the beat frequency mass accretion model is correct. QPOs are detected in the pre-eclipse data and, with marginal confidence, in the post-flare data. In the pre-eclipse data the QPO's centroid frequency is 47.4$\pm$0.2 mHz and its width is 1.3$\pm$0.9 mHz. In the case of the post-flare data there is a possible feature around 60 mHz and a knee at $\sim$ 34 mHz. QPOs with frequencies above around 40 mHz have not been seen in Cen X-3 before. The QPO frequency during the BBXRT observation was 40.7$\pm$1.1 mHz. The BBXRT unabsorbed luminosity was 3.29 x 10³⁷ erg s^-1 in the 2-10 keV band, while the mean luminosity during the ASCA observation was 5 x 10³⁶ erg s^-1. As the luminosity, and hence the mass accretion rate, decrease the Keplerian rotation frequency at the magnetosphere $\Omega_{K}^{}$(r_A) should also decrease. One then expects the QPO frequency ( $\omega$ = $\Omega_{K}^{}$(r_A) - $\Omega_{0}^{}$ if $\Omega_{K}^{}$(r_A) > $\Omega_{0}^{}$) to decrease. If the possible QPOs seen in the ASCA data are real, they would imply that the QPO frequency has increased with a decrease in luminosity. There are several ways this could happen. If $\Omega_{K}^{}$(r_A) < $\Omega_{0}^{}$, the QPO frequency will be $\omega$ = $\Omega_{0}^{}$ - $\Omega_{K}^{}$(r_A). However, the centrifugal barrier would tend to inhibit accretion through the propeller effect and switch off the source. Another possibility is that the QPOs are harmonics of the fundamental beat frequency. It is also possible that these features are produced by some mechanism other than the beat frequency mass accretion model.