A specialized genetic algorithm (GA) is used to search the structural space of samarium-doped ceria (SDC) for the most energetically stable configurations which will predominate in low temperature fuel cells. A systematic investigation of all configurations of 3.2% SDC and a GA investigation of 6.6% SDC are presented for the first time at the DFT+U level of theory. It was found that Sm atoms prefer to occupy the nearest neighbor (NN) position relative to the oxygen vacancy at 3.2%, while at 6.6%, a balance exists between various Sm-vacancy interactions and the vacancies prefer to be separated by ∼6 Å. Also, the migration barriers for oxygen diffusion are calculated amongst the best structures in 3.2% and 6.6% SDC and are found to be comparable to the barriers in Gd-doped ceria at the DFT+U level of theory. While the migration calculations provide insight on the oxygen diffusion mechanism in this material, the favored configurations from our GA enable future research on concentrated SDC and contribute to the atomistic understanding of the influence of dopant-vacancy and vacancy-vacancy interactions on ionic conductivity.

A genetic algorithm ͑GA͒-inspired method to effectively map out low-energy configurations of doped metal oxide materials is presented. Specialized mating and mutation operations that do not alter the identity of the parent metal oxide have been incorporated to efficiently sample the metal dopant and oxygen vacancy sites.The search algorithms have been tested on lanthanide-doped ceria ͑L =Sm,Gd,Lu͒ with various dopant concentrations. Using both classical and first-principles density-functional-theory ͑DFT͒ potentials, we have shown the methodology reproduces the results of recent systematic searches of doped ceria at low concentrations ͑3.2% L 2 O 3 ͒ and identifies low-energy structures of concentrated samarium-doped ceria ͑3.8% and 6.6% L 2 O 3 ͒ which relate to the experimental and theoretical findings published thus far. We introduce a tandem classical/DFT GA algorithm in which an inexpensive classical potential is first used to generate a fit gene pool of structures to enhance the overall efficiency of the computationally demanding DFT-based GA search.

Classical force field simulations and genetic algorithms are used to navigate low-energy configurations of samarium-doped ceria (SDC) at a number of concentrations, up to 20% SDC, such that the experimentally observed peak in ionic conductivity is mapped out in its entirety and fresh insight into samarium's role is reported.

We study the (111) surface of 10.3%, 14.3%, and 18.5% samarium-doped ceria (SDC) using a genetic algorithm (GA) to search for the most energetically stable configurations. In all cases, both Sm ions and oxygen vacancies segregate to the surface, which is similar to experimental findings for 5.3% SDC. 1 Importantly, at the optimal doping level of SDC (∼11%), where conductivity is maximal, defect segregation is limited such that vacancies remain 6 Å apart and pairs of vacancies do not form. At higher concentrations, pairs of vacancies are present, which likely contributes to the observed decrease in ionic conductivity. We also investigate the low-energy bulk structure of SDC from 10.3% to 18.5%, at the DFT+U level of theory, which has not been previously reported. The DFT+U energetics allow us to gain further insight on the fundamental interactions that influence ionic conductivity and defect segregation, as well as to confirm the insight reported from classical simulations. 2 The low-energy configurations found by our GA search enable future studies of SDC at experimentally relevant concentrations and identify the important interactions at the bulk and surface of the fuel cell electrolyte.

No abstract