Study of the Berezinskii-Kosterlitz-Thouless transition: An unsupervised machine learning approach

Abstract

The Berezinskii-Kosterlitz-Thouless (BKT) transition in magnetic system is an intriguing phenomena and an accurate estimation of the BKT transition temperature has been a long-standing problem. In this work we explore the anisotropic classical Heisenberg XY and XXZ models with ferromagnetic exchange on a square lattice and antiferromagnetic exchange on a triangular lattice using an unsupervised machine learning approach called principal component analysis (PCA). In earlier studies of the BKT transition, spin configurations and vorticities calculated from Monte Carlo method are used to determine the transition temperature $T_{BKT}$, but those methods fail to give any conclusive results by analyzing the principal components in the PCA approach. In this work vorticities are used as initial input to the PCA and curve of the first principal component with temperature is fitted with a function to determine an accurate value of $T_{BKT}$. This procedure works well for anisotropic classical Heisenberg with ferromagnetic exchange on square lattice as well as for frustrated antiferromagnetic exchange on a triangular lattice. The classical anisotropic Heisenberg antiferromagnetic model on the triangular lattice has two close transitions; the BKT at $T_{BKT}$ and Ising like phase transition for chirality at $T_c$ and it is difficult to separate these transition points. It is also noted that using the PCA method and manipulation of their first principal component, not only separation of transition points are possible but also transition temperature can be determined accurately.