# Identification of hidden order and emergent constraints in frustrated magnets using tensorial kernel methods

## TL;DR

This paper introduces a machine learning method using tensorial kernels and spectral graph analysis to identify and characterize hidden order parameters and emergent constraints in frustrated magnetic systems, including spin liquids.

## Contribution

It presents a novel, interpretable machine learning approach that automatically learns phase diagrams and identifies order parameters and gauge constraints in classical frustrated spin models.

## Key findings

- Successfully distinguishes between hidden biaxial spin nematic and spin liquid phases.
- Provides a systematic hierarchy of disordered regimes.
- Establishes the physical relevance of local constraint susceptibilities.

## Abstract

Machine-learning techniques have proved successful in identifying ordered phases of matter. However, it remains an open question how far they can contribute to the understanding of phases without broken symmetry, such as spin liquids. Here we demonstrate how a machine learning approach can automatically learn the intricate phase diagram of a classical frustrated spin model. The method we employ is a support vector machine equipped with a tensorial kernel and a spectral graph analysis which admits its applicability in an effectively unsupervised context. Thanks to the interpretability of the machine we are able to infer, in closed form, both order parameter tensors of phases with broken symmetry, and the local constraints which signal an emergent gauge structure, and so characterize classical spin liquids. The method is applied to the classical XXZ model on the pyrochlore lattice where it distinguishes---among others---between a hidden biaxial spin nematic phase and several different classical spin liquids. The results are in full agreement with a previous analysis by Taillefumier \emph{et al.} [Phys. Rev. X 7, 041057 (2017)], but go further by providing a systematic hierarchy between disordered regimes, and establishing the physical relevance of the susceptibilities associated with the local constraints. Our work paves the way for the search of new orders and spin liquids in generic frustrated magnets.