UC Berkeley

Anthropogenic climate change represents one of the greatest present threats to human society and global ecology. Rapid and dramatic reductions in greenhouse gas emissions are essential to avoid catastrophic global warming and environmental collapse. Although renewable energy from solar and wind resources is cheap and widely available, these resources are intermittent, and energy storage is needed to reliably supply renewable energy to the electrical grid and power vehicles. In particular, electrochemical energy storage in batteries can be highly efficient and energy-dense, showing tremendous promise to decarbonize the energy and transportation industries.

Metal-ion batteries like lithium-ion batteries (LIBs) are the current state of the art for commercial and research energy storage technologies. To achieve high energy density, metal-ion batteries typically operate at voltages outside of the electrochemical stability window of their electrolytes. As a result, electrolyte components react electrochemically, reducing and/or oxidizing to trigger a complex reaction cascade. In some batteries, such as LIBs with graphitic negative electrodes and electrolytes with certain cyclic and linear carbonate solvents, electrolyte degradation is managed by the formation of a passivation film known as an interphase. Once an interphase forms, these batteries are extremely stable over many years and thousands of charge-discharge cycles. To realize the goal of next-generation energy storage technologies with higher energy densities, including lithium-ion batteries with lithium metal or silicon electrodes as well as multivalent batteries (e.g. magnesium-ion batteries or MIBs), electrolytes must be designed either to avoid decomposition altogether or to react in a controlled manner such that they form stable interphase films.

Herein, I discuss efforts to predict the behavior of reactive electrochemical systems such as battery electrolytes using computational simulations and data science. I begin (Chapter 1) with an introduction to battery electrochemistry, focusing primarily on the role of electrolytes and interphase formation, as well as the inherent challenges of interphase engineering. In Chapter 2, I review the literature on analysis of interphase films using both experimental and theoretical approaches. As I will show, there are few techniques available that can provide mechanistic explanations for electrochemical reactivity accounting for the complex interactions between electrodes, electrolyte species, impurities, and decomposition intermediates and products.

Chapter 3 details a traditional study of electrolyte decomposition using a first-principles quantum chemical method: density functional theory (DFT). Specifically, I use DFT to explain the chemical and thermal instability of lithium hexafluorophosphate (LiPF6) salt in terms of elementary reaction mechanisms. This study highlights the limitations of chemical intuition in understanding reaction cascades and the need for new methods to broadly explore diverse (electro)chemical interactions.

In Chapter 4, I describe such a methodology that combines high-throughput DFT, chemical reaction networks (CRN), and stochastic simulations to predict reaction outcomes and pathways in complex systems with minimal prior knowledge. As a proof of concept (Chapter 5), I apply this approach to reductive electrolyte decomposition and interphase formation in LIBs with ethylene carbonate as a solvent. I recover most previously reported interphase products and predict several new, previously unreported, but chemically plausible species. From this starting point, I apply a kinetic Monte Carlo (kMC) algorithm to construct a model of LIB interphase formation and evolution (Chapter 6). This model successfully captures the expected bilayer structure of the interphase without relying on adjustable parameters or fitting to experiment.

Finally, in Chapter 7, I apply the previously described CRN methods to a new system where significantly less information is known: electrolyte decomposition and gas evolution in MIBs. I show how CRNs can help to interpret experimental spectra (in this case differential electrochemical mass spectroscopy or DEMS) and explain not only why certain species form in abundance but also why other species cannot form due to kinetic limitations.

I conclude (Chapter 8) by highlighting opportunities to apply computational modeling --- in particular using CRNs --- to understand complex (electro)chemical systems. The developments laid out here represent a significant step forward towards the rational design of reactive processes not only in the world of batteries and energy storage but also in electrochemical synthesis, pollution management, and much more. I also point out remaining challenges to studies of reactivity --- both experimental and computational --- and suggest possible avenues for future research to alleviate them.