Phase-Field Model of SEI Formation in Li-Metal Batteries

The formation of solid electrolyte interphase (SEI) is critical to the overall performance of lithium batteries. Here we have developed a rather comprehensive phase-field model for predicting the formation and growth of SEI, taking into account of multiple electrochemical/chemical reaction pathways leading to multiple SEI reaction products, diffusion of ionic and molecular species, and electron tunneling. All the SEI products are treated stoichiometric compounds without the typical approximation of parabolic composition dependence in almost all existing phase-field models involving stoichiometric compounds, allowing the direct connection between the phase-field model and atomistic calculations without further approximation. As a model system, we consider a 1-D prototypical half-cell with Li metal anode, liquid ethylene carbonate (EC) electrolyte with 1 molar (M) LiPF 6 , and SEI products Li 2 BDC and Li 2 CO 3 under the open-circuit condition. The reduction potentials and activation barriers of the reactions are determined using atomistic simulations while the diffusivities of Li + and EC in the bulk electrolyte are obtained by performing molecular dynamics simulations. The phase-field simulation is capable of predicting the evolution behavior of SEI products within a wide range of time scales from nanoseconds to a few seconds. It reveals that: (i) a porous Li 2 BDC layer forms at the anode/electrolyte interface and grows to a stable thickness of several tens of nanometers within ~20 μs; (ii) a porous Li 2 CO 3 layer forms between anode and Li 2 BDC and grows to ~10 nm by partially consuming the existing Li 2 BDC layer within ~33 μs; and (iii) Li 2 CO 3 forms a dense layer by filling the pores within ~20 ms. The stage (i) and (ii) are dominated by Li + diffusion kinetics whereas stage (iii) is controlled by the reaction kinetics. The general computational framework can be extended and applied to the simultaneous evolution of Li-metal deposition and stripping and the SEI evolution.