Discovery of Correlated Electron Molecular Orbital Materials using Graph Representations

TL;DR

This paper introduces a graph-based framework for systematically discovering correlated electron molecular orbital (CEMO) materials in inorganic solids, enabling high-throughput identification and classification of transition-metal clusters.

Contribution

The authors develop a novel graph representation method and high-throughput screening approach to discover and classify CEMO materials, providing an open dataset and tools for the scientific community.

Findings

Identified 5,306 cluster-containing materials from 34,548 compounds.

Discovered 2,627 stable or metastable compounds with isolated clusters.

Revealed symmetry and element-dependent trends in cluster formation.

Abstract

Correlated electron molecular orbital (CEMO) materials host emergent electronic states built from molecular orbitals localized over clusters of transition metal ions, yet have historically been discovered sporadically and generally been treated as isolated case studies. Here we establish CEMO materials as a systematically discoverable class and introduce a graph-based framework to identify, classify, and organize transition-metal cluster motifs in inorganic solids. Starting from crystal structures in the Materials Project, we construct transition metal connectivity graphs, extract cluster motifs using a bond-cutting algorithm, and determine cluster point groups, effective cluster sublattice dimensionality, and translational symmetry. Applying this approach in a high-throughput screen of 34,548 compounds yields 5,306 cluster-containing materials, including 2,627 stable or metastable compounds with isolated clusters and 984 materials featuring mixed-metal clusters. The resulting dataset reveals symmetry and element-dependent trends in cluster formation. By integrating cluster classification with flat-band lattice topology and battery-relevant information, we provide further relevant information to multiple scientific communities. The accompanying open dataset, Cluster Finder software, and interactive web platform enable systematic exploration of cluster-driven electronic phenomena and establish a general pathway for discovering correlated quantum materials and functional materials with cluster-based or extended metal-metal bonding in inorganic solids.