Learning phase transitions in ferrimagnetic GdFeCo alloys

TL;DR

This paper uses machine learning, specifically the 'learning by confusion' scheme, to identify and characterize phase transitions in ferrimagnetic GdFeCo alloys, aligning well with traditional methods but offering potential speed advantages.

Contribution

The study introduces a machine learning approach to detect phase transitions in GdFeCo alloys, revealing a triple phase transition point with a universal $W$-shape, and demonstrates its effectiveness compared to numerical minimization.

Findings

Identified triple phase transition point in GdFeCo alloys.

Machine learning results agree with thermodynamical minimization.

Potential for faster phase transition detection.

Abstract

We present results on the identification of phase transitions in ferrimagnetic GdFeCo alloys using machine learning. The approach for finding phase transitions in the system is based on the `learning by confusion' scheme, which allows one to characterize phase transitions using a universal $W$-shape. By applying the `learning by confusion' scheme, we obtain 2D $W$-a shaped surface that characterizes a triple phase transition point of the GdFeCo alloy. We demonstrate that our results are in the perfect agreement with the procedure of the numerical minimization of the thermodynamical potential, yet our machine-learning-based scheme has the potential to provide a speedup in the task of the phase transition identification.