Computational study of electron-transfers and singlet oxygen in aprotic metal-O2 batteries

Aprotic metal-oxygen batteries (MOBs), based on the electroreduction of molecular oxygen at a porous cathode, have attracted a vast interest in research, owing to their potential upgrade in terms of energy density and costs over present lithium-ion batteries. Despite their highly promising features, aprotic MOBs based on alkali and alkaline-earth metals still suffer severe limitations in their practical applicability. One of the main unresolved issues, especially with Li-O2 batteries, is represented by the high degree of parasitic reactivity. Singlet oxygen (1O2) is today held responsible for a major contribution to such reactivity, and the disproportionation of the superoxide anion is considered as one of the most likely source of 1O2 in the cell environment. Experimental evidences for electrolyte degradation and evolution of 1O2 have been reported, but the fundamental chemical mechanisms underlying these phenomena are still poorly understood. A valid strategy for contrasting the arise of side-reactions and materials degradation is to use redox mediators (RMs), which allow to recharge the battery with greatly reduced overpotentials. Understanding the con- nection of RM-assisted charging with the production 1O2 is likely to play a key role in the design of fully reversible and efficient practical MOBs in the future. In this thesis, quantum chemical computational methods were used to investigate reactive processes of electron-transfer involving reduced oxygen species in aprotic MOBs. The possibility of reactive pathways leading to the release of 1O2 was addressed in particular. The aim of the thesis was to apply theoretical methods to the modeling of reactive systems, in order to unravel part of the mechanisms which underpin the parasitic chemistry of MOBs. Despite their apparent simplicity, the reaction governing the chemistry of the cells involve a complex interplay of radical species and electronic excited states. For this reason, our approach was to use mainly ab-initio correlated multiconfigurational methods for a high-level description of potential energy surfaces and reaction energies. Owing to the computational costs of the methods, such an approach necessarily entails the resort to simplified models, including the exclusive use of implicit solvent and the neglect of solid phases and interfacial effects.