Machine learning-guided discovery of high-efficiency electrolyte additives for aqueous magnesium-air batteries

Abstract

Besides alloying, electrolyte additives have emerged as an effective strategy to overcome parasitic anodic hydrogen evolution reactions, and the formation of detrimental deposit layers at Mg-based anodes, thus improving the discharge behavior of Mg-air batteries. However, discovering suitable electrolyte additives through experimental testing is time-consuming and labor-intensive, given their high number of potential candidates. Our recently developed machine learning-based adaptive design was used iteratively in this work. Based on this, electrolyte additive 2,3-dihydroxynaphthalene was discovered, which achieved in a lab-made (Mg-0.2Ca)-air battery a cell voltage of 1.82 V and anodic utilization efficiency of 83%, yielding a specific energy of 3.37k Wh kg−1. This represents the highest recorded value among all Mg-air batteries reported to date. The results highlight the high potential of machine learning-guided discovery of high-efficiency electrolyte additives to further push the cutting-edge development of high-energy-density Mg-air batteries.