Title:Neural network approach to nonequilibrium phase transitions in open quantum systems

Abstract:Recent advances in the neural network (NN) machine learning technique provide a remarkable ability to identify the phases of matter. Here, we use this NN approach and identify the transition point of the absorbing transition arising in the quantum contact process in one dimension (1D-QCP), which is potentially realized in cold Rydberg atom systems. The training datasets for supervised learning consist of snapshots taken far from the transition point in quantum jump Monte Carlo (QJMC) simulations. Using the convolutional NN approach and finite-size scaling analysis, we identify the transition point, denoted as $\omega_c$, of the 1D-QCP. Moreover, by performing extensive QJMC simulations at $\omega_c$, we successfully determine all the critical exponents of the 1D-QCP and conclude that the 1D-QCP belongs to the directed percolation class. By contrast, the mean-field solution of the QCP shows a different type of phase transition, tricritical phase transition.