Metastable materials, corresponding to local minima in the potential energy landscape, are abundant in nature and technology. However, computational crystal structure prediction methods have traditionally been designed to focus only on the prediction of stable structures corresponding to the global minimum.

A new approach for the prediction of metastable phases with specific structural features has been implemented within the XtalOpt evolutionary algorithm. This method can be employed to predict materials with specified features including the local crystalline order (e.g. the coordination number or chemical environment) and the symmetry (e.g. Bravais lattice and space group). This method is applied to various chemistries and demonstrated to find various low-energy metastable phases, some which are experimentally known.

XtalOpt is published under an open-source license, thereby contributing towards cyberinfrastructure.