Unveiling the Phase Diagram and Reaction Paths of the Active Model B with the Deep Minimum Action Method

Area of Science:

• Statistical Physics
• Active Matter Systems
• Phase Transitions

Background:

• Nonequilibrium phase transitions are challenging due to complex dynamics and unknown steady-state distributions.
• Time-reversal symmetry is often broken in these systems.

Purpose of the Study:

• To compute the phase diagram of active Model B.
• To unveil unconventional reaction paths and nucleation mechanisms in dimensions 1, 2, and 3.
• To analyze the system's switching between homogeneous and inhomogeneous phases.

Main Methods:

• Deep neural network implementation of the geometric minimum action method (gMAM).
• Computation of the phase diagram for active Model B.

Main Results:

• Mean time to escape phase separation is extensive with system size L, but non-monotonic in 1D.
• Mean time to escape the homogeneous state is finite.
• Active terms enhance homogeneous phase stability, potentially destroying phase separation in finite systems.

Conclusions:

• The study provides insights into the dynamics of active matter systems, particularly concerning finite-size effects.
• Results are relevant for systems with a limited number of constituents (around 10^7).