Simulating Ion Wake Dynamics in Dusty Plasmas with ML-Augmented N-body Models

Charged dust grains immersed in a plasma perturb the local plasma density and, consequently, the electric potential. This is evident in complex plasma laboratory experiments where the ions flowing past the dust grains are deflected, creating an ion wake downstream of the grains. Individual ion wakes are modified by the presence of nearby dust grains. To understand the complex structure of the ion wake fluctuations, we use an N-body model of dust and ions (DRIAD) [1] that explicitly simulates the formation of the ion wake field for multiple interacting dust particles. To understand the macroscopic effects of the wake field fluctuations on the effective dust-dust interaction potential, we employ two methods. In one, we use sparse regression techniques to determine the form of an interaction force from position data of simulated particle pair interactions. In the second, we use an N-body model of the dust and ions and calculate the dust and ion potential for fixed positions of the dust particle pairs. We then use machine learning to determine the position-dependent coefficients for the potential of the dust plus ion wake, which is represented by a Gaussian cloud [2]. We implement these methods in an N-body code modeling dust only, where the effect of the ions is modeled through the learned interaction force. This allows us to model large dust systems and investigate dynamic phenomena.



[1] Matthews, Sanford, Kostadinova, et al., (2020) Phys of Plasmas 27(2), 023703

[2] Vermillion, et al., (2024) Phys of Plasmas 31(7), 073701