Neural-Network-Assisted Boltzmann Approach for Dilute Microswimmer Suspensions

2025/10/02 by Hayano, Haruki, Furukawa, Akira, Kim, Kang
#FOS: Physical sciences #Soft Condensed Matter (cond-mat.soft)

paper · doi:10.48550/arxiv.2510.02559

Abstract

We introduce a neural-network-assisted Boltzmann framework that learns the binary-collision map of microswimmers directly from data and uses it to evaluate collision integrals efficiently. Using a representative model swimmer, the learned map quantitatively predicts translational and rotational diffusivities and enables a linear-stability analysis of isotropy against polar ordering in dilute suspensions. The resulting predictions closely match direct simulations. The present framework is agnostic to active matter models and broadly applicable: once two-body collision data are obtained -- either from simulations or experiments -- the same surrogate can be used to evaluate kinetic transport across dilute conditions where binary collisions dominate. Because the workflow relies only on pre- and post-collision statistics, the present approach provides a general data-driven route linking particle-scale interactions to macroscopic transport and collective behavior in active suspensions.

Citations

• Self-Aligning Polar Active Matter
• Bacterial Hydrodynamics
• Enhanced diffusion of tracer particles in dilute bacterial suspensions

Related

• Fast-Forward Lattice Boltzmann: Learning Kinetic Behaviour with Physics-Informed Neural Operators
• Learning fast, accurate, and stable closures of a kinetic theory of an active fluid
• Solving the QCD effective kinetic theory with neural networks
• Topological Regularization for Force Prediction in Active Particle Suspension with EGNN and Persistent Homology
• Correlations, mean-field limits, and transition to the concentrated regime in motile particle suspensions