Thermoelastic anisotropy of lithium niobate from molecular dynamics with a machine-learned interatomic potential

Our latest paper is now freely accessible for the next 50 days from this link: https://authors.elsevier.com/a/1nUCG3In-v8hqF 

We developed a machine-learned interatomic potential (MLIP), and used it to compute the temperature-dependent elastic behavior of bulk lithium niobate from molecular dynamics. The MLIP was validated against structural, elastic, and phonon benchmarks. 

Five of the six independent elastic constants soften monotonically with temperature, whereas C₄₄​ is essentially temperature-independent. The coupling constant C₁₄​ shows the strongest relative decrease, falling by 28% between 100 and 700 K. 

The diagonal stiffnesses are 6 to 9% below room-temperature experiments. This offset is attributed primarily to the exchange–correlation functional, with smaller contributions from congruent-to-stoichiometric composition differences and nuclear quantum effects.

Our complete workflow is available here: https://github.com/nuwan-d/MLIP_LiNbO3_VASP_LAMMPS

This research was supported by DARPA.