High Throughput Discovery of Quantum Cluster Motifs in Correlated Electron Materials

Quantum clusters which are finite assemblies of transition-metal atoms with molecular orbitals extending over multiple sites, provide a platform for emergent electronic and magnetic phenomena, including tunable multiferroicity, magnetic anisotropy, and metal-insulator transitions. Trimer-based Nb₃​Br₈​, which hosts out-of-plane ferroelectricity and S =1/2 spins per trimer, recently enabled a field-free Josephson diode, while Nb₃​Cl₈​ has been explored as a candidate spin liquid. We introduce a custom graph-based algorithm that integrates symmetry and bonding considerations with high-throughput computational data to systematically identify such cluster materials. Starting from large-scale DFT databases, we apply physical filters such as short transition-metal separations and anomalous magnetic responses, construct graph representations to detect discrete clusters, and classify these with point-group symmetry analysis and dimensionality classification using principal-component analysis based on singular-value decomposition. The workflow is designed for automation and scalability, enabling efficient screening of candidate compounds. We will present a database of identified cluster materials, their classifications, and associated analysis tools. This approach highlights the potential of graph-based representations, combined with high-throughput computation, to accelerate the discovery of correlated electron materials with novel functionality.