Probing additives for green lubricants with the aid of machine learning molecular dynamics: The case of gallate molecules for aqueous solutions

To address the negative impacts of mineral oil-based lubricants on the environment and human health, eco-friendly lubricants based on aqueous solutions are increasingly being explored as an alternative for several tribological applications. The friction reduction properties of these lubricants crucially depends on the additives included in liquid of lower viscosity than oil. Great efforts are presently devoted to the search of molecular compounds able to function in boundary lubrication conditions, where the extreme pressures and low speeds make the liquid film become so thin that a direct contact between metal asperities can take place. Atomistic simulations represent a powerful tool to design lubricant additives as they can open a window on the sliding buried interface and unravel the mechanisms of function of candidate molecular compounds. However, while classical molecular dynamics simulations relying on empirical or reactive force fields poorly describe tribochemical reactions, the expensive cost of ab initio methods poses a big challenge when it comes to the required large system sizes and long time scales. In this study, we show how a machine-learning interaction potential derived from ab initio data can successfully be used to design lubricant additives. Considering the case of gallates at iron interfaces, our simulations highlight two key molecular features necessary for maintaining interfacial lubricity in boundary conditions: a strong anchoring of the molecules to the substrate, a function that for gallates is smartly activated by tribochemical reactions, and the presence of inert hydrocarbon tails that form a cushion that, thanks to Pauli repulsion, chemically isolates the covered substrate from any reactive counter-surface. These findings, of general validity, provides useful insights for designing new, environmental-friendly lubricants.