Machine learning analysis on critical structural factors of Al:ZnO (AZO) films

Machine learning (ML) has been recently approved to generalize materials design insights from many narrow-scoped individual experiments. In this study, decision tree algorithm (DT) was able to clarify critical structural factors enabling Al-doped ZnO (AZO) films with wider (or narrower) energy band gap and higher (or lower) electrical resistivity on published literature data as well as having ∼90 % average prediction accuracy. Cost-effective alternative of expensive physical models and simulations to reveal promising AZO films before physical experiments. Trained DTs were interpreted to uncover critical design insights. It was concluded that all microstructural features (crystalline size, crystallinity, dislocation density and interlamellar spacing) should be considered simultaneously through a holistic perspective in a design proposal. Al doping concentration and film thickness, as charge carrier contributors, were found as critical features for band gap and electrical resistivity, respectively. Effects on microstructural features should be considered as well as prerequisite in film design because of complex crystal growth mechanism.