Monte Carlo computer simulation of the XRF intensity dependence on the propagation plane inclination

When a multicomponent sample is irradiated with a collimated beam of X-rays, the absorption by the photoelectric effect give place to a characteristic X-ray emission of each atom class, which is known as X-ray fluorescence (XRF). If this outgoing radiation is detected by collimating into a sharp direction, a strong geometrical dependence on the incident and take-off angles (measured in relation to the normal vector to the irradiated surface) is found. A particular parameterization [1] of both angles which relates them to the so called propagation plane (the plane that contains both X-ray collimated beams or, equivalently, both the incidence and the take-off directions of the X-rays), made it possible for every class of atom to separate the radiation they emitted as consequence of the absorption of the incident excitation spectrum (primary process) from which they produced by absorption of characteristic radiation emitted by neighbor atoms of different classes, whose X-rays have enough energy to be absorbed by the observed atoms (secondary process). Both types of XRF emission, despite of their different origins, have exactly the same energy and can not be individuated by other means. When detection occurs, the whole intensity is accumulated at every energy. The above mentioned parameterization was found following a theoretical model [1], which permitted us to predict that the calculated intensity due to the secondary process should vanish while the primary remains unchanged, when the α angle of the propagation plane tilt increases toward π/2. For the limit case α = π/2 this property make it possible to separate the radiation emitted by the different effects, because then only the primary remains. Computer simulation was necessary to have a first principles checking of what was predicted theoretically. Monte Carlo simulation confirmed the previous results and helped us to understand why this happened. By dividing the sample into thin slabs and looking at the accumulated intensity in every one of them, it is possible to explain the invariance of the primary emission and the vanishing of the secondary one [2]. Careful application of variance reduction techniques renders feasible to implement the photon transport simulation on a single user microcomputer powered with a floating-point numerical coprocessor, while the simulation times are quite lower than the reported CPU time of a VAX/780 timeshared computer. © 1989.