Properties optimisation of nanostructures via machine learning: Progress and perspective

Abstract

Nanostructures play a vast role in the current Internet of NanoThings (IoNT) era due to remarkable properties and features that precisely impart their desired application functions in catalysis, energy and other fields. The exploration in understanding their minute features caused by the flexibility of compositional and complex atomic arrangement from the synthesis reaction widely opens for the in-depth discovery of their search space such as particle size, morphology and structures that controlled the characteristics. A wide range of possible compositions and various lattice atomic arrangements combined with small particle size distribution and large surface area create grand challenges to copy/differentiate their corresponding specific properties. Thus, the employment of machine learning (ML)-based strategies using the closed-loop experimental data from the nanostructure synthesis to help navigate and optimise for the large classes of data attributes related to the size, morphology and other properties from the trained model are reviewed. The data attributes are assisted by discussions of the selected case studies from the recent literature that highlight different condition nanostructures. The review concludes with a discussion of perspectives on the major challenges in the implementation of ML data-driven design in the field of nanostructure synthesis.