Uncertainty qualification of surface diffusion barriers on lithium metal surface through multiscale simulations

The barrier for lithium atom self-diffusion on the metal surface is a good descriptor for the dendrite growth with metal anodes. Herein, using density functional theory (DFT) with nugget elastic band (NEB) and molecular dynamics (MD) with machine learning potential, we have calculated the diffusion barriers of lithium on top of various low index surfaces based on five different hopping mechanisms, and qualified the uncertainty of these barriers. It is discovered that the (111) and (211) surfaces have much higher across step diffusion barriers (>0.5 eV) than the (100) surface (similar to 0.2 eV), which partially explains why these surfaces could grow much faster even though the surface energies are similar. The MD simulations further reveal a two-step mechanism for the across step diffusion on (100) surface. The diffusion barriers of foreign atoms on the lithium surface have further been calculated using DFT and AIMD simulations, which show significant reductions in the across step diffusion barriers for barium (Ba) and potassium (K). It can also be concluded that while the barriers obtained from DFTNEB calculations may reveal the relative tendencies for surface diffusions, dynamic simulations are essential to further determine the actual mechanisms of surface diffusion.