Publication date: 10th April 2024
Magnesium (Mg)-batteries, which form a subset of multivalent (MV) electrochemical system, are a promising alternative to state-of-the-art lithium ion batteries (LIBs) due to their high theoretical volumetric energy density, natural abundance of Mg, safety, and low cost. However, the need for a high-voltage intercalation cathode with high-enough Mg-ion mobility remain a bottleneck for the realization of practical Mg-batteries. The poor mobility of Mg2+ in ionic solids is often attributed to stronger local electrostatic interactions that cause larger fluctuations in the underlying potential energy surface, resulting in higher migration barriers (Em). Thus, if the potential energy surface is flattened, via suitably modifying the underlying structure and/or the local coordination environment of Mg, the Em can be lowered and Mg motion can be made facile. Thus, in this work, we use a combination of density functional theory (DFT), machine learned interatomic potentials (MLIPs), and molecular dynamics (MD) to evaluate the Mg intercalation thermodynamics and Mg-transport properties of amorphous oxides as potential Mg/MV cathodes. Specifically, we generated a DFT-calculated dataset based on vanadium oxides and trained a moment tensor potential (i.e., an MLIP) to model large supercells mimicking an amorphous framework over longer time scales. We use pair distribution functions to verify the creation of an amorphous structure and active learning to ensure that our MLIP model has trained successfully. Subsequently, we find that the amorphous structure can indeed enhance Mg motion compared to the crystalline framework, with a loss of ~0.2 V in average Mg intercalation voltage over the entire Mg composition range. Our work demonstrates that computational frameworks can be used for unearthing novel amorphous Mg cathode materials, which can eventually result in the construction of a practical MV battery.