Theoretical Prediction of Novel Materials with the XtalOpt Evolutionary Algorithm

The XtalOpt evolutionary algorithm for crystal structure prediction has been extended to enable the prediction of materials with specific properties. A fitness function has been implemented wherein the user can denote the percent contribution that enthalpy and the property (e.g. Vickers hardness obtained via a macroscopic hardness model and the shear modulus as determined via machine learning, percentage of hydrogen atoms that do not form H-H bonds, or density of states at the Fermi level) have on the fitness function. We have used XtalOpt to search for hard and stable carbon allotropes. Several novel carbon allotropes that are superconducting or posess super-long sp3-sp3 bonds were found. We also discovered novel hydrides that could potentially be conventional superconductors.