Machine learning-based prediction of microparticle dynamics in externally driven strongly-coupled dusty plasmas

Abstract We implemented unsupervised machine learning to study strongly coupled dusty plasmas with an emphasis on the dynamics of small plasma crystals rotated as a result of irradiation with a pulsed electron beam. By applying a combination of clustering algorithms such as K-means, DBSCAN, and Hierarchical clustering, along with statistical analysis, we classify the trajectories of microparticles in dust crystals levitated in the sheath of a radio-frequency plasma. This approach enables the identification of dynamic patterns and facilitates predictions based on initial conditions and system parameters. We trained the clustering algorithms to predict dust trajectories, identifying the most likely ones based on the hierarchical relationships between them and the Silhouette and Davies–Bouldin scores. Our results demonstrate the power of combining data science with experimental plasma physics to address challenges in manipulating levitated dust crystals. This predictive model serves as a versatile tool for advancing research in dusty plasma systems and enhances our ability to control dust dynamics in strongly-coupled dusty plasmas.