Accelerating the Structure Exploration of Diverse Bi–Pt Nanoclusters via Physics-Informed Machine Learning Potential and Particle Swarm Optimization

Bimetallic Bi-Pt nanoclusters exhibit diverse structural motifs, including core-shell, Janus, and mixed alloy configurations, due to the unique bonding characteristics be- tween Bi and Pt atoms. Using density functional theory (DFT) refinements from ChIMES physically machine-learned potential and CALYPSO particle swarm opti- mization global searches, we systematically classified 34 Bi-Pt nanoclusters based on coordination number and radial distribution function analysis, achieving 87% accuracy compared to manual labels. Our results reveal that Bi atoms predominantly occupy surface sites, driven by charge transfer effects. Cohesive energy trends alone proved in- sufficient for structure differentiation, necessitating a data-driven approach employing principal component analysis (PCA) and K-means clustering. Furthermore, vibra- tional, electronic, and infrared (IR) spectral analyses provided additional insights into structure-property relationships. PCA applied to IR, density of states (DOS), and vi- brational DOS (VDOS) spectra identified key features distinguishing structural classes. The first two principal components of the IR, VDOS, and DOS datasets strongly cor- related with Bi-Pt vibrational modes in the range 114–170 cm-1, highlighted important low-frequency modes (smaller than 100 cm-1), and reflected electronic delocalization near the Fermi level, respectively, further distinguishing structural categories. The trained ChIMES model, incorporating force and energy terms, enabled reliable op- timization simulations. Our findings offer an original framework for the automated classification and analysis of bimetallic nanoclusters, enhancing the understanding of their stability and functional properties.