Unsupervised machine learning for detection of phase transitions in off-lattice systems. II. Applications

Area of Science:

• Physics
• Statistical Mechanics
• Computational Science

Background:

• Phase transitions are fundamental phenomena in physical systems.
• Analyzing particle configurations is crucial for understanding system behavior.
• Detecting transitions in off-lattice systems, especially out of equilibrium, presents challenges.

Purpose of the Study:

• To demonstrate the utility of principal component analysis (PCA) for identifying phase transitions.
• To apply PCA to diverse off-lattice systems, both equilibrium and nonequilibrium.
• To showcase PCA's capability in analyzing complex particle configuration data.

Main Methods:

• Application of principal component analysis (PCA) to particle configuration data.
• Analysis of nonequilibrium systems, including the random organization (RandOrg) model.
• Study of equilibrium phase transitions in hard ellipse systems.
• Investigation of demixing transitions in binary mixtures like the Widom-Rowlinson mixture.

Main Results:

• PCA successfully identified phase transitions in all tested systems.
• The method distinguished between quiescent and steady-state behaviors in the RandOrg model.
• PCA revealed orientationally and positionally driven transitions in hard ellipses.
• The demixing transition in the Widom-Rowlinson mixture was effectively detected.

Conclusions:

• Principal component analysis is a versatile tool for detecting phase transitions in various off-lattice systems.
• PCA provides insights into both equilibrium and nonequilibrium phenomena.
• This approach offers a robust method for analyzing complex particle configurations.