Observation of the hexatic phase in a two-dimensional complex plasma using machine learning.

Complex plasmas consist of ionized gas and charged solid microparticles, representing the plasma state of soft matter. We apply machine learning methods to investigate a melting transition in a two-dimensional complex plasma. A convolutional neural network is constructed and trained with the numerical simulation. The hexatic phase is successfully identified and the evolution of topological defects is studied during melting transition in both simulations and experiments.