Abstract
Surfactants are key ingredients in foaming and cleansing products acrossvarious industries such as personal and home care, industrial cleaning, andmore, with the critical micelle concentration (CMC) being of major interest.Predictive models for CMC of pure surfactants have been developed based onrecent ML methods, however, in practice surfactant mixtures are typically useddue to to performance, environmental, and cost reasons. This requiresaccounting for synergistic/antagonistic interactions between surfactants;however, predictive ML models for a wide spectrum of mixtures are missing sofar. Herein, we develop a graph neural network (GNN) framework for surfactantmixtures to predict the temperature-dependent CMC. We collect data for 108surfactant binary mixtures, to which we add data for pure species from ourprevious work [Brozos et al. (2024), J. Chem. Theory Comput.]. We then developand train GNNs and evaluate their accuracy across different prediction testscenarios for binary mixtures relevant to practical applications. The final GNNmodels demonstrate very high predictive performance when interpolating betweendifferent mixture compositions and for new binary mixtures with known species.Extrapolation to binary surfactant mixtures where either one or both surfactantspecies are not seen before, yields accurate results for the majority ofsurfactant systems. We further find superior accuracy of the GNN over asemi-empirical model based on activity coefficients, which has been widely usedto date. We then explore if GNN models trained solely on binary mixture andpure species data can also accurately predict the CMCs of ternary mixtures.Finally, we experimentally measure the CMC of 4 commercial surfactants thatcontain up to four species and industrial relevant mixtures and find a verygood agreement between measured and predicted CMC values.