文献类型：期刊文献

英文题名：Machine learning-driven doping optimization: Enhancing structural stability of Mn-based ion sieves for Lithium extraction from brine

作者：Huang, Zhiai[1];Bao, Luri[1];Qu, Xiaorong[2];Wei, Ting[3];Sun, Shu-Ying[1,2]

机构：[1]East China Univ Sci & Technol, Sch Resources & Environm Engn, Shanghai 200237, Peoples R China;[2]Qinghai Transcend Separat & Extract Tech CO Ltd, Langfang, Peoples R China;[3]East China Univ Sci & Technol, Sch Chem & Mol Engn, Shanghai 200237, Peoples R China

年份：2026

卷号：392

外文期刊名：SEPARATION AND PURIFICATION TECHNOLOGY

收录：;EI(收录号：20260720055558);WOS:【SCI-EXPANDED(收录号:WOS:001691572100002)】；

基金：This work was supported by the Key Research and Development and Transformation Program of Haixi Prefecture (2025-YZH04) , and Qing-hai Kunlun Talent Plan.

语种：英文

外文关键词：Lithium extraction; Machine learning; Dopant engineering; Mn dissolution

摘要：Manganese-based adsorbents are a key technology for lithium extraction from salt-lake brines, but they suffer from Mn dissolution. Conventional dopant screening relies on trial-and-error methods, which are both timeconsuming and costly. Here, we develop a machine learning (ML) model to predict the effects of 16 metallic dopants on Li1.6Mn1.6O4 adsorbents. To integrate literature data with our laboratory dataset, we employed the ComBat method to correct for batch effects, thereby enabling robust predictions based on limited data. SHAP analysis reveals that electronegativity (En), dopant loading (ratio), and melting point (m.p.) are the key features governing Mn dissolution. Experiments confirmed that Sr and Nd doping reduced the Mn dissolution rate to 1.27% and 1.03%, respectively, and the modified adsorbent maintained excellent cycling stability after 6 cycles. This data-driven workflow accelerates the screening of dopants and provides actionable guidance for mitigating Mn dissolution issues