Toward predicting surface energy of rutile TiO2 with machine learning

Control and design of reaction conditions to acquire desirable morphology is a complex and difficult process. In principle, one can always predict the equilibrium morphologies of nanoparticles once the specific surface energies of exposed crystallographic facets become available. However, the surface energies can be easily changed by environmental conditions, and are rarely measured. As a result, we employ the k-nearest neighbors (KNN) model to predict surface energies from experimentally observed equilibrium morphologies of rutile TiO2, which may provide guidelines for the rational design and synthesis of rutile TiO2 micro-/nanocrystals with desired morphologies.