Machine Learning Magnetic Parameters from Spin Configurations.

Area of Science:

• Condensed matter physics
• Materials science
• Computational physics

Background:

• Hamiltonian parameter estimation is vital but often time- and cost-intensive.
• Extracting parameters from high-resolution images is challenging due to large Hilbert spaces.

Purpose of the Study:

• To develop a novel, efficient protocol for Hamiltonian parameter estimation using machine learning and imaging.
• To accurately predict material properties from experimental images.

Main Methods:

• A machine learning (ML) architecture was trained to map spin configurations to Hamiltonian parameters using simulated images.
• The trained ML model was applied to experimental images for parameter estimation.
• A physical model was used to predict material properties based on estimated parameters.

Main Results:

• The ML protocol successfully reproduced experimental spin configurations.
• Key material properties, including coercive field, saturation field, and specimen volume, were accurately predicted.
• The approach demonstrated stability and efficiency in parameter estimation.

Conclusions:

• The proposed ML-based protocol offers a stable and efficient solution for Hamiltonian parameter estimation.
• This method accelerates the analysis of condensed matter systems and material property prediction.
• It opens new avenues for leveraging imaging data in physics and materials science.