Zeolite structure determination using genetic algorithms and geometry optimisation

The recently presented software zeoGAsolver is discussed, which is based on genetic algorithms, with domain-dependent crossover and selection operators that maintain the size of the population in successive iterations while improving the average fitness. Using the density, cell parameters, and symmetry (or candidate symmetries) of a zeolite sample whose resolution can not be achieved by analysis of the XRD (X-ray diffraction) data, the software attempts to locate the coordinates of the T-atoms of the zeolite unit cell employing a function of fitness' (F), which is defined through the different contributions to the penalties' (P) as F = 1/(1 + P). While testing the software to find known zeolites such as LTA (zeolite A), AEI (SSZ-39), ITW (ITQ-12) and others, the algorithm has found not only most of the target zeolites but also seven new hypothetical zeolites whose feasibility is confirmed by energetic and structural criteria.