Understanding and Designing the Frustration in Super-Ionic Conductors aided by Machine Learning

Frustration is a physical phenomenon in which plenty competing states exsit with similar energy levels. The frustration in super-ionic conductors enables their exceptionally high ionic conductivities. Although atomistic modeling reveals a wide range of specific mechanisms in causing the frustration, a long-standing challenge is that analyzing many disordered configurations of the atomistic systems and their energies using first principles computation is computationally expensive. With the aid of machine learning interatomic potential to provide atomistic energies of individual atoms, we proposed density of atomistic states (DOAS) as quantitive analytics to elucidating, characterizing, and understanding the frustration mechanisms involving a diverse range of locally disordered atomic configurations. Using Li-ion conductors as model systems, the DOAS quantitatively characterizes the onset and degree of disordering, reveals the local configurational disorder and energetics in causing the frustration, and reveal how the frustration of atomistic states enhances ion diffusion. Furthermore, materials design strategies aided by the DOAS are devised and demonstrated for new super-ionic conductors. As demonstrated, the combination of atomistic modeling and machine learning can extend and empower conventional physics analytics for unraveling fundamental mechanisms and for guiding material design.