Nonequilibrium molecular dynamics for accelerated computation of ion–ion correlated conductivity beyond Nernst–Einstein limitation

Abstract

Condensed matters with high ionic conductivities are crucial in various solid devices such as solid-state batteries. The conduction is characterized by the cooperative ionic motion associated with the high carrier density. However, the high cost of computing correlated ionic conductivities has forced almost all ab initio molecular dynamics (MD) to rely on the Nernst–Einstein dilute-solution approximation, which ignores the cross-correlation effect. Here we develop a chemical color-diffusion nonequilibrium MD (CCD-NEMD) method, which enables to calculate the correlated conductivities with fewer sampling steps than the conventional MD. This CCD-NEMD is demonstrated to well evaluate the conductivities in the representative solid electrolyte bulk Li10GeP2S12 and Li7La3Zr2O12. We also applied CCD-NEMD to the grain boundary of Li7La3Zr2O12 and demonstrated its applicability for calculating interfacial local conductivities, which is essential for investigating grain boundaries and composite electrolytes. CCD-NEMD can provide further accurate understanding of ionics with ionic correlations and promote developing solid devices.

How Concerted Are Ionic Hops in Inorganic Solid-State Electrolytes?

“…Figure shows the level of correlation estimated for the tracer ion diffusion coefficient, D x , and atomistic descriptors described above considering all the SSE families examined in this study (for this analysis, we considered the tracer diffusion coefficient instead of the full ion diffusion coefficient − because of its ubiquity in computational studies). Such correlations were obtained by following the same data-analysis approach that was introduced in our previous work, which essentially involves the computation of Spearman correlation coefficients and p -values for the assessment of statistical significance.…”

“…In the dilute-solution limit, the interactions between mobile ions are regarded as negligible; hence, the full ionic diffusion coefficient reduces to the tracer diffusion coefficient − (Computational Methods), and its dependence on temperature can be expressed as , where a is a hopping distance, E a is the activation energy barrier for ionic migration, and ν 0 is the hopping frequency.…”

“…Finally, in order to quantify the influence of neglecting many-ion correlation on the calculation of the activation energy barrier for ionic migration, E a , we estimated this quantity for Li 7 La 3 Zr 2 O 12 (LLZO) and Li 10 GeS 2 P 12 (LGSP), considering both the tracer and full ionic diffusion coefficients (Computational Methods). − We selected these two materials because the first lies inside the coincidence interval defined for the ion hopping frequency, while the second is outside. For LLZO, it was found that when disregarding many-ion correlations E a amounted to 0.16 eV, whereas it decreased to 0.14 eV when accounting for them.…”

“…The tracer diffusion coefficient, D , then was obtained based on the Einstein relation In practice, we considered 0 < t ≤ 50 ps and estimated D by performing linear fits to the averaged MSD­( t ) obtained over the last 25 ps. When taking into account many-ion correlations, the full diffusion coefficient was estimated by considering additional i – j particle positions crossed terms in eq …”

“…How does the neglection of many-ion correlation affect the estimation of key atomistic quantities like the ion hopping frequency? Answering these questions is not only relevant from a fundamental point of view but also necessary to justify the broad adoption of formulas obtained in the dilute-solution limit (e.g., the Nerst–Einstein relation for the ionic conductivity), which assumes mobile ions to be fully uncorrelated. − …”