Machine learning analysis of dimensional reduction conjecture for nonequilibrium Berezinskii-Kosterlitz-Thouless...

Area of Science:

• Condensed matter physics
• Statistical mechanics
• Machine learning applications

Background:

• The dimensional reduction conjecture proposes equivalence between static snapshots of driven disordered systems and space-time trajectories of lower-dimensional pure systems.
• This conjecture implies that the three-dimensional (3D) random field XY model, when driven out of equilibrium, may exhibit characteristics of the Berezinskii-Kosterlitz-Thouless (BKT) transition.
• Verifying this conjecture offers insights into the behavior of complex physical systems under non-equilibrium conditions.

Purpose of the Study:

• To investigate and verify the dimensional reduction conjecture in driven disordered systems.
• To explore the potential connection between 3D driven random field XY models and 2D pure XY models via the conjecture.
• To apply machine learning techniques for analyzing system configurations and testing theoretical predictions.

Main Methods:

• Utilizing a machine learning approach, specifically neural networks, to analyze system configurations.
• Training neural networks to distinguish between static snapshots of a 3D driven random field XY model and space-time trajectories of a 2D pure XY model.
• Employing image recognition capabilities of neural networks to detect subtle similarities or differences between the two system types.

Main Results:

• Neural networks trained on system configurations were unable to differentiate between the 3D driven random field XY model snapshots and the 2D pure XY model trajectories.
• The inability of the machine learning model to distinguish the two suggests a deep underlying equivalence as proposed by the conjecture.
• This provides strong evidence supporting the dimensional reduction conjecture in the context of driven disordered systems.

Conclusions:

• The dimensional reduction conjecture is confirmed for the investigated driven disordered systems.
• The study demonstrates the power of machine learning in verifying complex theoretical concepts in physics.
• The findings suggest that driven 3D random field XY models effectively behave as their lower-dimensional pure counterparts under specific conditions, potentially exhibiting BKT-like transitions.