Reconciling the Role of Pore Structure in Capacitive Deionization: From Molecular Insights to Real-Scale Machine Learning Modeling
Farzin Saffarimiandoab, Shengyin Tang, Yulong Yang, Mohammad‐Hossein Sarrafzadeh, Erçan E. Kuruoğlu, Xihui Zhang
Abstract
High Resolution Image Download MS PowerPoint Slide The rational design of the porous structure of capacitive deionization (CDI) electrodes necessitates a comprehensive analysis approach encompassing insights across different scales. However, translating the idealized molecular-level insights on pore structures into practical CDI performance remains a significant challenge. Here, we bridge this gap by transferring molecular dynamics (MD) simulation insights into machine learning (ML) models trained on experimental data to unravel the role of key pore structure features of micropore crystallinity and size governing the charge compensation mechanisms and salt adsorption capacity (SAC). MD simulations reveal that subnanometer micropores and pore surface curvatures induce higher SAC, though ion transport into amorphous pores exhibits a picosecond-scale delay compared to the more structured slit-pores. Building upon these molecular insights, we incorporated micropore size and the I D / I G ratio into ML models, which confirmed the subnanometer micropore (≤9 Å) role in achieving higher SAC. Additionally, while SAC initially is suppressed by higher I D / I G due to the induced resistance, it progressively significantly boosts equilibrium SAC by the end of deionization, aligning with MD results. This study demonstrates how the presented synergistic multiscale approach can effectively explore new electrode features and translate pore-scale mechanisms into electrode-scale performance, providing a reliable framework for quantitative CDI structure–performance analysis.