Machine-learned collision operators from Particle-in-Cell simulations

Computational plasma physics has seen significant advances in the ability to model nonlinear plasma dynamics from first principles. However, capturing the complex multi-scale dynamics of plasmas, especially in non-equilibrium and strongly coupled regimes, remains challenging. One of such challenges is the development of collisional operators that capture the macroscopic dynamics in regimes where existing analytical theory is expected to fail.

In this talk, we will highlight how differentiable simulators, coupled with machine learning algorithms and self-consistent Particle in Cell (PIC) data, can be used to efficiently learn collision operators [1]. As a test case, we will focus on extracting the collision operator for an electromagnetic PIC code and compare the retrieved operator against existing theory [2,3]. We will conclude by illustrating how, using the same simulation tools, one can shed light on the nature of collisions in strongly coupled regimes and, in the future, elucidate how collisional operators are modified in strongly coupled, relativistic, and far-from-thermodynamic-equilibrium conditions.

[1] D. Carvalho et al, in preparation (2025)

[2] A. B. Langdon. Proc. 4th Conference on Numerical Simulation of Plasmas (1970), pp. 467–495

[3] H. Okuda and C. K. Birdsall, Phys. Fluids 13.8 (1970), pp. 2123–2134