2002MNRAS.331..141G

We present new results based on RXTE observations of the millisecond pulsar SAX J1808.4-3658, carried out during the decay of the 1998 April outburst. The X-ray spectrum can be fitted by a two-component model. We interpret the soft component as blackbody emission from a heated spot on the neutron star, and the hard component as coming from Comptonization in plasma heated by the accretion shock as the material collimated by the magnetic field impacts on to the neutron star surface. The hotspot is probably the source of seed photons for Comptonization. The hard component illuminates the disc, giving rise to a reflected spectrum. The amount of reflection indicates that the disc is truncated at fairly large radii (20-40R_g), consistent with the lack of relativistic smearing of the spectral features. The inferred evolution of the inner radius is not consistent with the magnetic field truncating the disc. Instead, it seems more likely that the inner disc radius is set by some much longer time-scale process, most probably connected to the overall evolution of the accretion disc. This disc truncation mechanism would then have to be generic in all low-mass accretion rate flows both in disc accreting neutron stars and black hole systems.

The phase-resolved spectra show clearly that the blackbody and hard Comptonized spectra pulse independently. This obviously gives an energy-dependent phase-lag. Full general relativistic effects are not required to explain this. The soft blackbody component is optically thick, so its variability is dominated by its changing projected area, while the Doppler shifts (which are maximized 90° before the maximum in projected area) are somewhat stronger for the translucent column.

We do not detect Compton reflection from the neutron star surface, though we predict that it should be present in the X-ray spectrum. This would give an unambiguous observational measure of M/R if there is any iron on the neutron star surface which is not completely ionized.