Algorithms for the Markov Entropy Decomposition

TitreAlgorithms for the Markov Entropy Decomposition
Type de publicationJournal Article
Nouvelles publications2013
AuteursFerris, AJ, Poulin, D
JournalPhys. Rev. B
Mots clésBelief Propagatio, Markov entropy
The Markov entropy decomposition {(MED)} is a recently-proposed, cluster-based simulation method for finite temperature quantum systems with arbitrary geometry. In this paper, we detail numerical algorithms for performing the required steps of the {MED}, principally solving a minimization problem with a preconditioned Newton's algorithm, as well as how to extract global susceptibilities and thermal responses. We demonstrate the power of the method with the spin-1/2 {XXZ} model on the {2D} square lattice, including the extraction of critical points and details of each phase. Although the method shares some qualitative similarities with exact-diagonalization, we show the {MED} is both more accurate and significantly more flexible.
Texte complet