Applications of Empirical Potentials and Pathway Combinations to Extending Perovskite Simulations
Abstract
Perovskite oxides, in particular yttrium doped barium zirconium oxides (YBZ), are
promising electrolyte materials for solid oxide fuel cells (SOFCs). Proton conduction
through electrolyte materials is the rate-limiting step in electricity generation by
SOFCs. Proton conductivity can be altered by grain boundaries, frequently occurring
defects in perovskites. The most challenging obstacle in studying grain boundaries
is the scale problem. The experimental data for grain diameters range between 10
and 20 nm while previous proton conduction simulation sizes have been under 2 nm.
This research aims to extend the proton conduction simulation sizes, laying out the
foundation for the investigation of proton conduction across grain boundaries. Ionic relaxations on ab initio potential surfaces generated by Vienna Ab initio
Simulation Package (VASP) are limited by system sizes and computational resources.
In contrast, reactive force field, ReaxFF, allows the investigation of systems with
thousands of atoms in a reasonable amount of time. ReaxFF was used to optimize
YBZ backbone structures and calculate proton binding site energies. The structures
optimized by ReaxFF were more distorted and did not exhibit the symmetry predicted
by VASP. Proton binding site energies calculated by ReaxFF varied on a wider range
and equivalent binding sites did not have the same energies as they did in VASP.
However, ReaxFFs prediction of the most stable Glazer distortion agreed with VASPs
result. ReaxFF relaxes structures 10,000 times faster than VASP. This impressive improvement in speed may compensate for the lower accuracy when the thermal
energy is high enough to obscure ne details of potential energy wells. More studies
need to be done to an a good balance between efficiency and accuracy of ReaxFF.
Dynamic programming and graph theory were utilized in comparing proton conduction
pathways in two different simulation box sizes. The weighted fractions of
rotations, intraoctahedral, and interoctahedral transfers were different in the large
and small systems. However, the average limiting barriers over N-step periodic pathways
were the same in the two systems. Pathways of the large system could be made
by stitching together pathways from the small system. The ensemble of periodic
pathways in the large system was exactly the ensemble of the stitched pathways.
These pathway combinations thus enable the theoretical characterization of periodic
pathways in systems through the experimental grain sizes. Extensions of this approach
could allow connecting paths through large grains with paths through grain
boundaries. Reactive force eld and pathway combinations are useful tools in simulating
long range conduction pathways through large perovskite systems with grain
boundary defects.
Description
Keywords
perovskite, proton conduction, dynamic programming, density functional theory