Figure 5
(a) Process flow for training MLIPs on results from RMC refinements, using an active learning approach. (b) Method for extracting a local environment from an RMC supercell and embedding it in a supercell suitable for DFT calculations. The colored atom positions are from the RMC supercell, within the cutoff radius of the high-uncertainty local environment. The remaining gray atoms are from the average positions. |