Machine-learning-potential simulations show that local Li content sets Mn oxidation states at LiMn2O4 surfaces and that the O-H stretch band exposes a dual-site acid-attack vulnerability
Using atomistic simulations with an ab initio-quality machine learning potential, this work investigates aqueous LiMn2O4 (LMO) interfaces across varying lithiation states and shows that local Li content determines surface Mn oxidation states and governs interfacial acid-base chemistry, identifies the O-H stretching band as a sensitive spectroscopic probe of the surface electronic structure whose oxidation-state-dependent shifts reveal that the mixed-valence spinel surface hosts coexisting Lewis-acidic centers and neighboring oxygen sites susceptible to electrophilic attack, extending the conventional acid-centric view of Mn dissolution toward a dual-site mechanism.