Unraveling the reaction mechanisms of the ZDDP anti-wear additive by machine learning molecular dynamics

Wear is a major challenge in mechanical systems, particularly in engines where it compromises efficiency and component lifetime. For over 80 years, zinc dialkyldithiophosphate (ZDDP) has been the industry-standard anti-wear additive, but increasing environmental regulations now demand suitable replacements - a goal that requires a precise understanding of its protective mechanism. While experiments have established the composition of ZDDP-derived tribofilms, the atomistic pathways driving their growth remain unclear due to the difficulty of simulating highly reactive tribochemical environments at realistic time and length scales with the required chemical accuracy.

Leveraging recent advances in machine-learning potentials (MLPs), we develop the first MLP specifically tailored for ZDDP and apply it in large-scale molecular dynamics simulations of tribofilm growth at iron interfaces of varying reactivity, including oxidized surfaces and nano-asperities. We show that linkage-isomers formation by S-to-O substitution accelerates the development of the characteristic polyphosphate network, and identify distinct decomposition mechanisms enabled by C–O bond cleavage to form hydrocarbon by-products, a key step to obtain carbon-free tribofilms.

These simulations provide the first atomistically-resolved, dynamic picture of ZDDP tribofilm formation, bringing novel insights to several open questions, and paving the way for future simulations that can guide the design of sustainable next-generation anti-wear additives.