Classification of Metal Nanoclusters Using Convolutional Neural Networks

Abstract

Catalysis happens only at the surface of materials, this makes nanoparticles of particular interest in the field of catalysis because of their high surface-to-volume ratio. The exact atomic structure of nanoparticle surfaces is of particular importance in catalysis, and the expression of surface facets is largely governed by their overall structure. Typically, small metal nanoparticles will take one of three major structural isomers: decahedron, icosahedron or cuboctahedron (Figure 1). Determination of the structural isomer of a nanoparticle can be performed using high-angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) [1,2].

Charting Nanocluster Structures via Convolutional Neural Networks

“…Machine learning (ML) is emerging as an invaluable analysis tool in the field of nanoclusters, as it allows efficient navigation of the complexity of the structural landscape by extracting meaningful patterns from large collections of data. ML has already found application in microscopy image recognition, , dimensionality reduction and exploration of potential energy surfaces, structural recognition, − characterization of the local atomic environment, , and machine learning force fields for metals. , …”