A computational study of the reconstruction of amorphous mesoporous materials from gas adsorption isotherms and structure factors via evolutionary optimization

A general method for the three-dimensional reconstruction of mesoporous materials by evolutionary optimization against target data is developed. The method is applied specifically in reconstruction of amorphous material models using gas adsorption data, structure factor data, or a combination of both. A recently introduced lattice-gas approach is used to model adsorption in these calculations, and a high-pass limited Fourier representation is used to facilitate evolution of large-scale structures during the optimization. Reconstructions are made of several material models which mimic real materials obtained either by phase separation and etching or by sol-gel processing. Analysis of the reconstructions provides considerable insight into the type and quantity of structural information probed by gas adsorption and small-angle scattering experiments. We find that reconstructions based only on structure factors tend to underestimate the mean pore size. We also find that in many cases excellent reconstructions can be obtained using only adsorption-branch data, and that in all cases reconstructions based jointly on both types of data are superior to those based only on one, suggesting that these measures contain "complementary" information. It is also found that in most cases the use of desorption data is not warranted, and that the use of adsorption data taken at many temperatures will not improve reconstructions. The reproducibility of the method is shown to be satisfactory. The method can be computationally expensive if gas adsorption data are used, but it is easily parallelized, and therefore results can still be obtained in reasonable time. Finally, the possible application of this approach to real systems, including templated porous materials, is discussed.