Linking Microscopic Structures to Ionic Currents in Solid Electrolytes via Kinetic Monte Carlo and Regression Analysis

The relationship between dopant configuration and ionic current is critical for designing solid-state devices composed of ionic conductors, and this relationship becomes more important when the device is fabricated to a smaller size in device miniaturization. However, this relationship is still ambiguous due to limitations in experimental measurements. In this study, we focused on yttria-stabilized zirconia (YSZ) as a representative ionic conductor and theoretically investigated the relationship between the dopant (yttrium) configuration and ionic current density in 8 mol % YSZ by a data-driven approach. Using a large data set of over 2000 YSZ systems with different yttrium dopant configurations, drift current caused by the oxygen ions diffusing under a bias voltage was calculated using mesoscale kinetic Monte Carlo simulation. To characterize the dopant configurations, three types of physically motivated structural descriptors were defined as histograms of the 6-site local configurations related to oxygen migration and compared with respect to ionic conductivity prediction accuracy. This prediction was conducted using a linear regression model trained on the data set of dopant configurations converted to structural descriptors and calculated ionic conductivities. As a result, we succeeded in identifying the most informative structural descriptors and linking microscopic dopant configurations and macroscopic ionic transport features through multiple computational and data-analytic tools. Furthermore, we found that features related to the waveform of ionic current density, such as local fluctuations and fast Fourier transform (FFT) coefficients, can also be predicted with high accuracy. Thus, our approach allows reconstruction of the ionic current density from the presented structural descriptors. Although YSZ was selected in this study, the present methods can be applied to other ionic conductors in a similar way.