Selective energy dispersive diffraction peak fitting by using genetic algorithm

Accurate and precise estimates of X-Ray diffraction peak parameters is mandatory, when small dynamic changes of lattice parameters have to be quantitatively analyzed. To follow in real time such changes, a large set of patterns must be usually collected, so that the position of certain peaks of interest can be tracked. To calculate the positions, a fitting procedure of the peaks is required and several algorithms are reported in the literature for this purpose. However, these algorithms are mainly focused on the determination of parameters based on a model of the cell geometry. Here, we present a new algorithm allowing to carry out the fitting procedure on a portion only of the pattern, with neither tight constraints on the dataset, nor restrictive hypotheses on the sample structure. In our case, a coarse estimate of the detector resolution and of the positions of the peaks to fit are the only initial conditions required. This method can be regarded as a hybrid technique, as it makes use of a genetic algorithm approach, mixed with an intensive multiple random generation of the population, that makes it similar to a Monte Carlo technique. Moreover, adaptive genetic operators have been implemented in the data processing code. These properties result in a fast and efficient algorithm, a fundamental requirement when, as in the present case, the Energy Dispersive X-ray Diffraction method is applied to observe structural changes, which implies the acquisition of many patterns in a relatively short time. The result of this application is shown by some practical examples.