Probing hidden spin order with interpretable machine learning

Various types of multipolar order are ubiquitously found in frustrated spin systems, yet their identification is difficult and often causes confusion with spin liquids. In this work, by amalgamating the theory of orientational tensors and the strong interpretability of a particular machine-learning technique, the authors introduce an algorithm to detect the occurrence and the order parameter of hidden multipolar order and demonstrate its advantage over traditional methods. This approach provides a new scheme in the search for novel magnets as well as a comprehensive way to scrutinize spin liquid candidates.