Capturing dynamical correlations using implicit neural representations

Area of Science:

• Condensed Matter Physics
• Materials Science
• Computational Physics

Background:

• Collective excitations in materials are key to understanding many-body physics.
• Dynamical structure factor (S(Q, ω)) is typically measured using inelastic neutron or X-ray scattering.
• Analysis involves comparing experimental data with theoretical predictions.

Purpose of the Study:

• To develop a data-driven analysis tool for spectrographic measurements.
• To efficiently extract unknown parameters from experimental data using automatic differentiation.
• To enable precise parameter extraction for ordered magnetic systems.

Main Methods:

• Utilizing neural implicit representations tailored for spectrographic data.
• Employing linear spin wave theory simulations for model training.
• Applying automatic differentiation for parameter refinement.

Main Results:

• Precise extraction of exchange parameters from inelastic neutron scattering data.
• Successful application to the square-lattice spin-1 antiferromagnet La₂NiO₄.
• Demonstration of a machine learning platform for advanced model refinement.

Conclusions:

• The developed tool offers a viable pathway for automatic refinement of models in ordered magnetic systems.
• Neural implicit representations show promise for analyzing complex material excitation spectra.
• This data-driven approach enhances the understanding of magnetic phenomena in materials.