Unexpected thermal conductivity enhancement in aperiodic superlattices discovered using active machine learning

This work demonstrates a general-purpose adaptive ML-accelerated search process that discovers unexpected lattice thermal conductivity (κ) enhancement in aperiodic superlattices (SLs) compared to periodic superlattices. Molecular dynamics simulations and a convolutional neural network (CNN) are used to predict κ for numerous structures. The method iteratively identifies aperiodic SLs with enhanced thermal transport, adding them to the CNN training data. The identified structures show increased coherent phonon transport due to closely spaced interfaces.

Prabudhya Roy Chowdhury, Xiulin Ruan