Using machine learning to turn particles into probes in dusty plasmas

In laboratory dusty plasmas, micron-sized dust particles are often levitated at the edge of a plasma sheath where the electric field balances gravitational forces. In the traditional study of this environment, intrusive, millimeter-sized Langmuir probes are used to measure plasma sheath properties, and particle interaction forces are calculated by either numerical simulation or mode analysis around their equilibrium positions. Here we simultaneously do both of these things by using machine learning to analyze experimental highly dynamic particle trajectories in 3D, thereby revealing previously inaccessible information about particles and the plasma without external probes. We track the motion 10-30 dust particles in 3D using scanning laser sheet tomography in a laboratory RF dusty plasma. The motion of these particles inherently contains information about the ambient plasma environment, including each particle's ion wake in the plasma sheath. We train machine learning (ML) models using this motion, and importantly, the model treats each particle individually with different sizes. In this presentation, I will detail our ML model and its recent discoveries, including a position-dependent plasma charge density, the position and dependence of the particle charge, particle interactions, and plasma wake structure.