Revolutionizing Solid‐State NASICON Sodium Batteries: Enhanced Ionic Conductivity Estimation through Multivariate Experimental Parameters Leveraging Machine Learning

• Published in: ChemSusChem Vol. 17; no. 6; pp. e202301284 - n/a
• Main Authors: Zhang, Yuyao, Zhan, Tingjie, Sun, Yang, Lu, Lun, Chen, Baoliang
• Format: Journal Article
• Language: English
• Published: Germany Wiley Subscription Services, Inc 22.03.2024
• Subjects: Algorithms; Chemical synthesis; Electrolytes; Electronegativity; Electronic properties; Ion currents; Ionic Conductivity; Machine Learning; Molten salt electrolytes; NASICON; Neural networks; Parameters; Sodium; Sodium-Ion Batteries; Solid electrolytes; Solid-State Electrolytes; Sodium-Ion Batteries; Ionic Conductivity; Machine Learning; NASICON; Solid-State Electrolytes
• ISSN: 1864-5631; 1864-564X
• DOI: 10.1002/cssc.202301284

Na superionic conductor (NASICON) materials hold promise as solid‐state electrolytes due to their wide electrochemical stability and chemical durability. However, their limited ionic conductivity hinders their integration into sodium‐ion batteries. The conventional approach to electrolyte design struggles with comprehending the intricate interactions of factors impacting conductivity, encompassing synthesis parameters, structural characteristics, and electronic descriptors. Herein, we explored the potential of machine learning in predicting ionic conductivity in NASICON. We compile a database of 211 datasets, covering 160 NASICON materials, and employ facile descriptors, including synthesis parameters, test conditions, molecular and structural attributes, and electronic properties. Random forest (RF) and neural network (NN) models were developed and optimized, with NN performing notably better, particularly with limited data (R2=0.820). Our analysis spotlighted the pivotal role of Na stoichiometric count in ionic conductivity. Furthermore, the NN algorithm highlighted the comparable significance of synthesis parameters to structural factors in determining conductivity. In contrast, the impact of electronegativity on doped elements appears less significant, underscoring the importance of dopant size and quantity. This work underscores the potential of machine learning in advancing NASICON electrolyte design for sodium‐ion batteries, offering insights into conductivity drivers and a more efficient path to optimizing materials. The potential of machine learning to predict ionic conductivity in NASICON was explored using a database of 211 datasets, covering 160 NASICON materials. Facile descriptors, including synthesis parameters, test conditions, molecular and structural attributes, and electronic properties, were employed to optimize and develop the models.