Predicting room-temperature conductivity of Na-ion super ionic conductors with the minimal number of easily-accessible descriptors

Abstract

Given the vast compositional possibilities Na$_nM_mM_{m'}$Si$_{3-p-a}$P$_p$As$_a$O$_{12}$, Na-ion superionic conductors (NASICON) are attractive but complicate for designing materials with enhanced room-temperature Na-ion conductivity $\sigma_{\rm Na,300K}$. We propose an explicit regression model for $\sigma_{\rm Na,300K}$ with easily-accessible descriptors, by exploiting density functional theory molecular dynamics (DFT-MD). Initially, we demonstrate that two primary descriptors, the bottleneck width along Na-ion diffusion paths $d_1$ and the average Na-Na distance $\langle d_{\rm Na-Na} \rangle$, modulate room-temperature Na-ion self-diffusion coefficient $D_{\rm Na,300K}$. Then, we introduce two secondary easily-accessible descriptors: Na-ion content n, which influences $d_1$, $\langle d_{\rm Na-Na} \rangle$, and Na-ion density $\rho_{\rm Na}$; and the average ionic radius $\langle r_M \rangle$ of metal ions, which impacts $d_1$ and $\langle d_{\rm Na-Na} \rangle$. These secondary descriptors enable the development of a regression model for $\sigma_{\rm Na,300K}$ with $n$ and $\langle r_M \rangle$ only. Subsequently, this model identifies a promising yet unexplored stable composition, Na$_{2.75}$Zr$_{1.75}$Nb$_{0.25}$Si$_2$PO$_{12}$, which, upon DFT-MD calculations, indeed exhibits $\sigma_{\rm Na,300K} > 10^{^3}$ S$\cdot$cm$^{-1}$. Furthermore, the adjusted version effectively fits $140$ experimental values with $R^2=0.718$.