Europhysics Letters (EPL)
Phase diagrams of a magnet with competing types of anisotropy: Entropy modeling and machine learning
V.V. Khitrintseva, Dmitry A Druzev, Pavel Vladimirovich Prudnikov
Abstract
Abstract We investigate critical behavior of the Heisenberg model with competing exchange and single-ion crystalline anisotropy with parameter D . We reconstruct the complete phase diagram using the joint density of states, which was calculated by the Wang-Landau algorithm. A hybrid approach involved entropy modeling with unsupervised machine learning methods is proposed for automatic identification of magnetic ordering phases and determination of bicritical point coordinates . V-shaped phase coexistance boundaries, corresponding to competition between XY -like and Ising-like order were reveald for with three types of transitions: classic second-order transitions between paramagnetic and ordered phases, first-order transition line between XY and Ising phases and crossover from Ising to paramagnetic phase at non-zero external magnetic field. The proposed approach allows for the precise detection of phase boundaries and multicritical points.