Computational Investigation of Lithium Metal Anodes for Solid-State Batteries

Abstract

Solid-state batteries that employ a Li-metal anode (LMSSB) are being widely explored within the battery community due to their potential to achieve improved energy densities. A solid-state battery would also convey safety benefits due to the elimination of the volatile and flammable liquid electrolytes commonly used in existing Li-ion systems. However, realizing these benefits is challenging, as it is widely known that LMSSBs suffer from internal short-circuiting due to dendrite formation and inefficiencies during cycling. One hypothesis for suppressing dendrite formation in LMSSB suggests that maintaining interfacial contact between the Li anode and the solid electrolyte is crucial. At moderate discharge rates, relatively slower diffusion within the anode results in roughening and void formation in Li near this interface. The resulting reduction in interfacial contact focuses the Li-ion current during plating to a reduced number of contact points, generating high local current densities that nucleate dendrites. In chapters 3 and 4, a strategy for minimizing void formation in Li anodes is proposed. Using a multi-scale model, it is shown that capacity and current density in LMSSBs can be improved by reducing the grain size and increasing the dislocation density of Li, thereby exploiting fast diffusion within microstructural defects of Li anodes. Diffusion rates along 55 tilt and twist GBs, and two dislocation types (edge and screw) in Li are predicted using molecular dynamics. Using these atomic-scale data, a 1D meso-scale model of Li depletion in the anode during discharge was developed. The model predicts that grain sizes in the range of 0.1-3 μm, or dislocation densities of 10^11-10^12/cm2, yield sufficiently fast self-diffusivity to enable robust LMSSBs. The range of values reflects the range of diffusivities predicted in the dislocation cores and GBs, and approximations intrinsic to the meso-scale model. As the optimal grain sizes and dislocation densities are different from those in common use by several orders of magnitude, strategies for controlling the microstructures of Li metal are discussed. Lastly, by using atomistic data, a plastic deformation map is constructed for Li anodes under stack pressure. The map indicates that when the grain size is large (~150 μm), dislocation-climb dominates the creep deformation. However, in fine-grained Li (~1 μm), grain boundary sliding creep is the dominant mechanism of plastic deformation. Regarding cycling inefficiencies in LMSSB, another hypothesized failure mode relates to the interfacial wettability of the Li anode and the metal current collector during operation of ‘anode-free’ LMSSBs. When Li and a metal substrate have poor work of adhesion, the nucleation barrier of Li on a metal surface (e.g., Cu foil) can be large. This can result in inhomogeneous deposition of Li and overpotentials of that deposition. To understand the origin of low adhesion between the Li and a Cu substrate, in Chapter 5 the properties of Li/Cu and Li/Cu2O interfaces were computed using density functional theory. These calculations indicate that Cu and Cu2O are lithiophilic; thus Li is predicted to wet both materials. However, the strongly exothermic conversion reaction, Cu2O + 2 Li -> Li2O + 2 Cu, instead favors the formation of a Li/Li2O interface. Prior work has shown that the stoichiometric Li/Li2O interface exhibits poor wetting behavior. A mechanism involving the conversion of native copper oxide into Li2O is consistent with the observation of inhomogeneous Li plating on Cu and high overpotentials observed experimentally in anode-free LMSSBs.