We study phase transitions in a lattice of square-arranged driven-dissipative polariton condensates with nearest-neighbour coupling. Simulating the polarization (spin) dynamics of the polariton lattice, we observe regions of qualitatively different steady-state behaviour which can be identified in time-integrated measurements. The transition between these regions resemble phase transitions ubiquitous in statistical physics, but have inherently non-equilibrium nature and cannot be classified in the conventional way. To overcome this challenge, we use machine learning methods to determine the boundaries separating the regions. We use unsupervised data mining techniques to sketch the regions of phase transition. We then apply learning by confusion, a neural network-based method for learning labels in the dataset, and extract the polaritonic phase diagram. Our work takes a step towards AI-enabled studies of polaritonic systems.
Download