Neural evolution structure generation: High entropy alloys

We propose a method of neural evolution structures (NESs) combining artificial networks (ANNs) and evolutionary algorithms (EAs) to generate High Entropy Alloys (HEAs) structures. Our inverse design approach is based on pair distribution functions atomic properties allows one train model smaller unit cells then larger cell. With speed-up factor approximately 1000 with respect the SQSs, NESs dramatically reduces computational costs time, making possible generation very large (over 40,000 atoms) in few hours. Additionally, unlike same can be used multiple fractional composition.