From Synthesis to Property: Expanding the Horizons of Machine Learning in Nanomaterials

Nanotechnology involves the manipulation of matter at dimensions ranging from 1 to 100 nanometers. Prediction of synthesis parameters, structure, properties and applications is a cascade process in nanomaterials research, each of these stages being interconnected and having a correlative influence on one another. Traditionally, the "trial and error" approach in nanomaterial research has several limitations, including time-consuming, laborious and resource-intensive. With the rise and vigorous development of artificial intelligence technology as the fourth paradigm of materials research, machine learning offers a significant research prospect for the accelerated new materials design, synthesis optimization and property prediction. In this review, the three key elements of machine learning including data, descriptors and machine learning methods for nanomaterial research are discussed. An overview of the applications of machine learning in nanomaterial research is provided, particularly focusing on various synthesis methods of single nanomaterials and property prediction of nanocomposites, through the framework of synthesis-structure-property-application relationships. Finally, the potential of this fast-growing field is highlighted, as well as the formidable challenges it faces.