Computational Study on a Variety of Lithium Diffusion Pathways

Lithium metal batteries are promising storage systems with better electrochemical characteristics than ubiquitous Li-ion. Lithium anode is the key to creating high-power batteries and the main challenge at the same time. One of the main flaws of using lithium as anode material is the formation of heterogeneous deposits on lithium surface during charge process. The most common type of such deposits is whiskers. Understanding of whiskers’ formation and growth mechanism is required for solving this problem. Last studies have shown that solid-state processes run the whole show in the whiskers’ formation processes. <br/> In this study we investigated and compared different pathways of lithium diffusion within metal electrode: diffusion of point defects in the metal bulk, self-diffusion through grain boundaries and the influence of Li-Li₂O interface on the point defects. Due to high reactivity of lithium, we weren’t able to provide all these investigations in chemical lab, we had to estimate diffusion processes with computational methods: Molecular Dynamics (MD) and DFT. <br/> We have chosen MD for modeling all-lithium-systems (LAMMPS package) because this method can handle a relatively large system size and time scale. For avoiding inaccuracies, we have compared several force fields firstly. One of them was Machine-learning Interatomic Potential (MLIP), which we had trained ourselves. The results obtained with MLIP have shown DFT-level of accuracy. For instance, we calculated melting parameters and point defect energies obtained with 5 different potentials and compared these data with the published experimental and DFT data. Thus, we revealed the superiority of MLIP over other potentials and the possibility of using it to simulate large systems such a grain boundary with DFT-accuracy. <br/> Further we simulated lithium grain boundary and calculated lithium diffusion coefficient from mean squared displacement slopes. Simulation cell consist of two lithium slabs rotated at a certain angle, which was chosen based on coincidence site lattice. We showed that the boundary is 3-5 atomic layers thick and is amorphous at room temperature. The mobility of grain boundary atoms was found to be extremely high and only 5 times lower than it is in supercooled molten lithium. <br/> For comparison of the diffusion coefficients in the grain boundary and bulk we had to estimate self-diffusion value in solid lithium bulk. We investigated the mobility of point defects (vacancy and self-interstitial atom) in the lithium bulk and discovered an anomalous phenomenon. Temperature dependence of vacancy diffusion coefficient corresponds to Arrhenius law, simultaneously this dependence of self-interstitial atom does not correlate with Arrhenius law and even demonstrates an inverse dependence, i.e. diffusion coefficient decreases with the temperature increasing. We have further demonstrated that this anomaly is due to the interplay between ballistic motion of interstitial defect along &lt;111&gt; direction and its rotation which obeys to Arrhenius law.<br/> There is the third way for lithium diffusion. It would be surface diffusion for any less reactive metal, but for lithium this way is not available due to presence of SEI. It is considered that the closest to lithium layer of SEI consists of lithium oxide, so we simulated Li-Li₂O interface using DFT (VASP) and estimated the behavior of point defects at the heterogeneous interface. We have studied the influence of space-charge at the interface on the defect energetics and mobility along the interface. <br/>Eventually, we have been able to consider all possible mass-transfer pathways within the electrode to build a complete picture of processes during lithium deposition. We have discovered that solid-state diffusion along grain boundaries and Li-SEI interfaces is sufficient to realize the proposed mechanism of whiskers’ formation and growth.