Matching crystal structures atom-to-atom: applications in material science

Abstract

Materials simulations and modeling paired with high performance computing resources have been successful in accelerating the discovery of new materials for renewable energy production and energy efficiency. However, relatively few of these materials have been metastable since predictive theories of materials realizability and lifetime are still limited. In that context, we develop a new approach to establish the best correspondence, or match, between different crystalline structures, a crucial step in estimating the lifetime of metastable materials. The new approach circumvents problems related to the choice of the unit cell, by mapping large finite portions of the crystals together. To obtain the best match, a cost function is minimized by simultaneously updating the mapping and the alignment between the two sets of atoms using the Khun-Munkres algorithm and a gradient descent respectively. The periodicity in the mapping is retrieved and given as a final output. We apply the aforementioned methodology to the martensitic transformation in steel and show how the resulting minimal distance pathway, with no energy minimization, is consistent with experimental observations. We also show that, compared to other transformation mechanisms it yields lower energy martensite when coexistence with austenite is considered. Moreover, we demonstrate how this new approach for matching crystal structures can be adapted to matching termination planes at heterointerfaces. We use the algorithm to create optimal interface structures between platinum and Gallium Oxide (beta-Ga2O3), an interface that is relevant for high-power electronics. Using density functional theory, we compute the local density of states and the Schottky barrier height (SBH) at the interface. We find that the presence of decomposed water on the (-201) beta-Ga2O3 surface has a strong impact on the SBH which emphasizes the importance of sample preparation on device performance. Finally, we present a potential approach for computing the nucleation energy in solid-solid transformation, an important quantity that is typically difficult to calculate.