SAX J1808.4-3658 has a 2.5 millisecond neutron star rotation period and exhibits X-ray pulsations due to its rotating hot spot. Here we present an analysis of the pulse shapes of SAX J1808 during its 1998 outburst. The modeling of the pulse shape includes several effects, including gravitational light-bending, doppler effects and two spectral components with different emissivity. In addition we include the new effects of light-travel time-delays and the neutron star's oblate shape. We also consider two different data sets, with different selection in time period (1 day versus 19 days of data combined) and different energy binning and time resolution. We find that including time-delays and oblateness results in stronger restriction on allowed masses and radii. A second result is that the choice of data selection strongly affects the allowed masses and radii. Overall, the derived constraints on mass and radius favor compact stars and a soft equation of state.

We present a simple method for including the oblateness of a rapidly rotating neutron star when fitting X-ray light curves. In previous work we showed that the oblateness induced by rotation at frequencies above 300 Hz produces a geometric effect which needs to be accounted for when modelling light curves to extract constraints on the neutron star's mass and radius. In our model X-rays are emitted from the surface of an oblate neutron star and propagate to the observer along geodesics of the Schwarzschild metric for a spherical neutron star. Doppler effects due to rotation are added in the same manner as in the case of a spherical neutron star. We show that this model captures the most important effects due to the neutron star's rotation. We also explain how the geometric oblateness effect can rival the Doppler effect for some emission geometries.

Raytracing computations for light emitted from the surface of a rapidly rotating neutron star are carried out in order to construct light curves for accreting millisecond pulsars. These calculations are for realistic models of rapidly rotating neutron stars which take into account both the correct exterior metric and the oblate shape of the star. We find that the most important effect, comparing the full raytracing computations with simpler approximations currently in use, arises from the oblate shape of the rotating star. Approximating a rotating neutron star as a sphere introduces serious errors in fitted values of the star's radius and mass if the rotation rate is very large. However, for lower rotation rates acceptable mass and radius values can be obtained using the spherical approximation.

We present raytracing computations for light emitted from the surface of a rapidly-rotating neutron star in order to construct light curves for X-ray pulsars and bursters. These calculations are for realistic models of rapidly-rotating neutron stars which take into account both the correct exterior metric and the oblate shape of the star. We find that the most important effect arising from rotation comes from the oblate shape of the rotating star. We find that approximating a rotating neutron star as a sphere introduces serious errors in fitted values of the star's radius and mass if the rotation rate is very large. However, in most cases acceptable fits to the ratio M/R can be obtained with the spherical approximation.

We demonstrate that fitted values of stellar radius obtained by fitting theoretical light curves to observations of millisecond period X-ray pulsars can significantly depend on the method used to calculate the light curves. The worst-case errors in the fitted radius are evaluated by restricting ourselves to the case of light emitted and received in the equatorial plane of a rapidly-rotating neutron star. First, using an approximate flux which is adapted to the one-dimensional nature of such an emission region, we show how pulse shapes can be constructed using an exact spacetime metric and fully accounting for time-delay effects. We compare this to a method which approximates the exterior spacetime of the star by the Schwarzschild metric, inserts special relativistic effects by hand, and neglects time-delay effects. By comparing these methods, we show that there are significant differences in these methods for some applications, for example pulse timing and constraining the stellar radius. In the case of constraining the stellar radius, we show that fitting the approximate pulse shapes to the full calculation yields errors in the fitted radius of as much as about +/- 10%, depending on the rotation rate and size of the star as well as the details describing the emitting region. However, not all applications of pulse shape calculations suffer from significant errors: we also show that the calculation of the soft-hard phase lag for a 1 keV blackbody does not strongly depend on the method used for calculating the pulse shapes.