Monte Carlo simulation of linearly polarized photons

The collision routine actually used in computer codes for Monte Carlo simulation of (partially) linearly polarized photons follows essentially one intuitive approach which simulates the unpolarized fraction of the beam by generating many polarized photons with their electric field vectors randomly oriented on the polarization plane. Clearly, this approach is more inefficient for simulating unpolarized than polarized radiation, and produces biased results which propagate on the multiple scattering terms. We propose a new scheme which is deduced from the vector transport model, valid for photons with arbitrary states of polarization. The proposed scheme avoids the splitting (and therefore the bias), thus improving the Monte Carlo variance reduction. In addition, the full polarization state after the collision can be straightforwardly determined from the incidence state and scattering geometry, allowing an update of the polarization state at every point in the phase space, which gives macroscopic information on the X-ray optics and helps to determine the contribution of each single collision chain to the final state.