Fig. 1: The identification of silver surface oxide.
From: Machine-learning atomic simulation for heterogeneous catalysis
a The 2000-step SSW-NN trajectory for identifying the structure of silver surface oxide (Ag12O6 phase on p(4 × 4) supercell of Ag(111)). The initial structure (IS) and four representative structures under different stage of optimization during SSW-NN are shown. b The building of Ag12O6 optimal surface phase with the given input of p(4 × 4) supercell and Ag12O6 composition by reinforcement learning based atomistic structure learning algorithm, in which the depth of blue in the figure represents the Q-values (the preference for the location of the next atom) predicted by the CNN, reproduced with permission from ref. 89. Copyright American Institute of Physics, 2019.
