A systematic study on the metallophilicity of ordered five-atomic-layer MXenes using high-throughput automated workflow and machine learning

Dendrite growth is an important issue hindering practical applications of various kinds of metal (e.g. Li, Na, K, Mg, Ca, Fe, Zn, and Al) batteries. To tackle this problem, an effective method is to introduce metallophilic materials as hosts for metal anodes. Due to its high electrical conductivity, ease of assembly, and large compositional space, MXene is an excellent choice for anode host materials. However, how do MXenes’ compositions influence their metallophilicity to different kinds of metals, and how do we find MXenes with the strongest metallophilicity to different kinds of metals? To answer these questions, a density functional theory (DFT) based high-throughput automated workflow (HTAW) is designed and a machine learning (ML) study is performed focusing on the metallophilicity of ordered five-atomic-layer MXenes. For the first time, the metallophilicity of ordered five-atomic-layer MXenes from their entire compositional space to a total of eight kinds of metal anodes is studied. The metallophilicity of ∼10 % of the compositional space is investigated by the HTAW and serves as the dataset for ML study, and the remaining ∼90 % is predicted by the trained ML model. Based on the predictions, a ‘cooperation with neighbor’ mechanism governing MXenes’ metallophilicity is summarized, MXenes with the strongest metallophilicity to the eight kinds of metals are discovered, and the already synthesized -O terminated Cr₂TiC₂ is found to be a member of them, which represents a readily available subject for further experimental verifications and practical applications.