Stephen Harris1,Harsh Jagad2,Chunmei Ban3,Brian Sheldon2,Yue Qi2
Lawrence Berkeley National Laboratory1,Brown University2,University of Colorado Boulder3
Stephen Harris1,Harsh Jagad2,Chunmei Ban3,Brian Sheldon2,Yue Qi2
Lawrence Berkeley National Laboratory1,Brown University2,University of Colorado Boulder3
The use of lithium or sodium metal anodes together with highly ion-conductive solid electrolytes (SEs) could provide batteries with a step improvement in volumetric and gravimetric energy densities. Unfortunately, these SEs face significant technical challenges, in large part because Li and Na dendrites can penetrate through SEs, leading to short circuits. The ability of such a soft material (Li or Na metal) to penetrate through ceramic is surprising from the point of view of models widely used in the Li-battery field. We introduce a concept, new to the battery field, for preventing penetration of lithium dendrites through SEs by putting the SE surfaces into a state of residual compressive stress. Residual stress is the stress that remains in a material after all external forces have been removed. As such, it is a material property, having nothing to do with stack pressure. For a sufficiently high residual compressive stress, which can be on the order of several GPa or higher, cracks have difficulty forming, and cracks that do form are forced to close, inhibiting dendrite penetration. This approach is widely used in industry to toughen ceramics and glasses (e.g., Gorilla Glass); and making SEs more fracture-resistant will make even very thin SE films possible to handle in factory environments. However, the technique will not be useful for SEs if the Li-ion transport rate through a SE is substantially reduced when the SE is under residual compressive stress. Our molecular dynamics calculations for Li-ion transport through a common SE demonstrate that the introduction of even very high residual compressive stresses (up to 10 GPa) has only a modest effect on Li-ion transport kinetics. Combined with our recent experimental and theoretical results, we will argue that this approach is both general and commercially viable.