Classifying and predicting the electron affinity of diamond nanoparticles using machine learning

Abstract

Using a combination of electronic structure simulations and machine learning we have shown that the characteristic negative electron affinity (NEA) of hydrogenated diamond nanoparticles exhibits a class-dependent structure/property relationship. Using a random forest classifier we find that the NEA will either be consistent with bulk diamond surfaces, or much higher than the bulk diamond value; and using class-specific random forest regressors with extra trees we find that these classes are either size-dependent or anisotropy-dependent, respectively. This suggests that the purification or screening of nanodiamond samples to improve homogeneity may be undertaken based on the negative electron affinity.