Abstract:We consider short-range classical spin glasses, or other disordered systems, consisting of Ising spins. For a low-temperature Gibbs state in infinite size in such a system, for given random bonds, it is controversial whether its decomposition into pure states will be trivial or non-trivial. We undertake a general study of the overall structure of this problem, based on metastates, which are essential to prove the existence of a thermodynamic limit. A metastate is a probability distribution on Gibbs states, for given disorder, that satisfies certain covariance properties. First, we prove that any metastate can be decomposed as a mixture of indecomposable metastates, and that all Gibbs states drawn from an indecomposable metastate are alike macroscopically. Next, we consider stochastic stability of a metastate under random perturbations of the disorder, and prove that any metastate is stochastically stable. Using related methods and older results, we prove that if the pure-state decomposition of any Gibbs states drawn from an indecomposable metastate is countably infinite, then the weights in the decomposition follow a Poisson-Dirichlet distribution with a fixed value of the single parameter describing such distributions, and also that if the overlap takes a finite number of values, then the pure states are organized as an ultrametric space, as in $k$-RSB. Dynamically-frozen states play a role in the analysis of Gibbs states drawn from a metastate, either as states or as parts of states. Using a mapping into real Hilbert space, we prove further results about Gibbs states, and classify them into six types. Metastate-average states are studied, and can be related to states arising dynamically at long times after a quench from high temperature, under some conditions.

Conditional Independence of 1D Gibbs States with Applications to Efficient Learning

Entanglement and Quantum Phase Transitions in Quantum Mixed Spin Chains

Learning Spin Liquids on a Honeycomb Lattice with Artificial Neural Networks.

From decay of correlations to locality and stability of the Gibbs state

Strong decay of correlations for Gibbs states in any dimension

Gibbs state sampling via cluster expansions

On Mixing of Markov Chains: Coupling, Spectral Independence, and Entropy Factorization

Disordered Gibbs measures and Gaussian conditioning

Infinite-Volume States with Irreducible Localization Sets for Gradient Models on Trees

High-Temperature Gibbs States are Unentangled and Efficiently Preparable

Structural analysis of Gibbs states and metastates in short-range classical spin glasses: indecomposable metastates, dynamically-frozen states, and metasymmetry

Geometric aspects of a spin chain

Efficient learning of ground and thermal states within phases of matter

Quantum concentration inequalities and equivalence of the thermodynamical ensembles: an optimal mass transport approach

Rapid mixing of global Markov chains via spectral independence: the unbounded degree case

Inference algorithm for finite-dimensional spin glasses: belief propagation on the dual lattice

Hierarchical exchangeability of pure states in mean field spin glass models

High-temperature Expansions and Message Passing Algorithms

Efficient Hamiltonian learning from Gibbs states