UC Berkeley

The demand for battery energy storage is rapidly growing, necessitating the development of novel battery chemistries that are affordable, utilize earth-abundant metals, and can deliver high energy density with long cycling stability. Promising avenues of research to achieve these goals include the development of solid-state electrolytes and novel cathode materials. The discovery and optimization of novel candidate phases can be accelerated through first principles calculations to rationalize the atomic structure, phase diagrams, and other relevant properties, without empirical parameters. In our work, we investigate the phase stability and electronic structure of sulfide superionic conductor electrolytes and Mn-rich rock-salt oxide cathodes from first-principles calculations. Although the chemical nature of these materials is distinct, we are able to harness similar statistical thermodynamics approaches to model the phase stability. Specifically, these sulfides within the Li-P-S (LPS) composition space and Mn-rich oxides (Li-Mn-Ti-O, or LMTO) are crystalline ionic materials exhibiting configurational disorder, which we capture by using lattice model methods, namely the cluster expansion (CE) coupled with Monte Carlo (MC) sampling. We find that accounting for the vibrational free energy contributions within the harmonic phonon approximation is also essential for accurate predictions of the LPS phases.The calculations of configurational and vibrational contributions to the free energy require accurate electronic structure methods to precisely evaluate the possible energy states. We harness density-functional theory (DFT) to calculate these quantities and train surrogate CE models. However, the LPS conductors and LMTO cathodes exhibit physical complexities that are non-trivial to treat within semi-local DFT exchange-correlation functionals. Specifically, the LPS conductors contain relatively long-range dispersion-type interactions within the sulfur (S) sublattice, which are not accurately captured within the generalized gradient approximation (GGA), but can be reasonably described within more recently developed meta-GGAs such as r2SCAN. The LMTO phases exhibit correlated and localized 3d electronic states, Jahn-Teller (JT) distortions, and antiferromagnetism of Mn3+, leading to the failures of GGA and meta-GGA approaches to predict the correct ground states, a deficiency primarily attributed to the self-interaction error (SIE) that is inherent to these functionals. Instead, DFT methods that more precisely correct the SIE, such as hybrid-GGA and GGA with non-empirical Hubbard corrections, are needed to achieve reasonable ground state prediction of Mn3+-rich compositions such as LiMnO2. We unveil the subtle interplay between the atomic and orbital ordering, magnetism, and degree of electronic localization/hybridization that determines the phase stability of LiMnO2.The thorough benchmarking and application of DFT methods enable us to calculate the relevant phase diagrams of LPS superionic conductors and LMTO rock-salt cathodes with reasonable accuracy, recovering and rationalizing experimentally observed trends. Within the LPS phase diagram, we predict the experimentally known polymorphic phase transitions among the Li3PS4 and Li7PS6 phases, the metastability of Li7P3S11 at all temperatures, and quantify the free energy differences between phases. We also clarify the details of the Li sublattices within specific LPS superionic conductors, which we find to be crucial for deriving physically accurate CE lattice models.In the LMTO system, we predict the order-disorder transition temperatures (Tdisord) as a function of composition to rationalize the thermodynamic accessibility of the disordered rocksalt (DRX) phase, a promising cathode material. Importantly, we predict the LMTO phase diagram to be eutectoid-like, in which the Tdisord decreases as off-stoichiometry is introduced to the end-point compositions, which is confirmed from X-ray diffraction (XRD) experiments. Our LMTO phase diagram reveals that DRX is stable over a significant composition range at relatively low temperatures of 800 − 900 C, significantly lower than the conventional synthesis temperatures (> 1000 C), offering a route to optimize particle morphology and rate capability. We also elucidate the importance of Ti doping towards stabilizing DRX at synthetically accessible temperatures.