Machine Learning‐Accelerated Design of High‐Efficient Lithium Adsorbents for Salt Lake Brines

The urgent demand for lithium resources has been advancing the development of high-efficient lithium extraction from brines. Aluminum-based lithium adsorbent, as the only one used in salt lake industry, is limited to further improve the extraction efficiency due to the low adsorption capacity. Here, we develop a high-throughput screening framework via interpretable machine learning (ML) to rapidly determine high-performance modification strategies for the aluminum-based lithium adsorbent, avoiding huge workload from traditional trial-and-error doping experiments in view of multiple doping schemes and the unique trade-off between performance and structural stability. Relying on the recommended modifications, a series of doped adsorbents are prepared and verified the structure stability and cyclic adsorption performance, which match well with the uncovered correlation between dopant features and adsorption capacity. Experimental validations confirm the screened doped one exhibits an increased stable adsorption capacity by near 40% in various types of brine. These results indicate that ML-accelerated approach can significantly promote the upgrading of lithium resource adsorption industry in salt lakes.