Predicting polymeric crystal structures by evolutionary algorithms

The recently developed evolutionary algorithm USPEX proved to be a tool that enables accurate and reliable prediction of structures for a given chemical composition. Here we extend this method to predict the crystal structure of polymers by performing constrained evolutionary search, where each monomeric unit is treated as one or several building blocks with fixed connectivity. This greatly reduces the search space and allows the initial structure generation with different sequences and packings using these blocks. The new constrained evolutionary algorithm is successfully tested and validated on a diverse range of experimentally known polymers, namely polyethylene (PE), polyacetylene (PA), poly(glycolic acid) (PGA), poly(vinyl chloride) (PVC), poly(oxymethylene) (POM), poly(phenylene oxide) (PPO), and poly (p-phenylene sulfide) (PPS). By fixing the orientation of polymeric chains, this method can be further extended to predict all polymorphs of poly(vinylidene fluoride) (PVDF), and the complex linear polymer crystals, such as nylon-6 and cellulose. The excellent agreement between predicted crystal structures and experimentally known structures assures a major role of this approach in the efficient design of the future polymeric materials.