Predicting the thermodynamics of formulations via machine learning-enhanced theoretical tools

Characterizing the structure and stability of a formulation is a key challenge in the consumer products industry. Especially when considering a shift to more ‘sustainable’ products, the ability to easily predict how the incorporation of new component(s) affects soft materials’ structure and properties would be a key enabler of sustainable product development. Polymer reference interaction site model (PRISM) theory is a powerful computational technique to efficiently predict the structure and thermodynamics of liquid-like polymer and soft matter systems. However, the use of PRISM theory can pose practical challenges due to issues with accuracy and numerical stability for some classes of systems. In this talk, I describe our efforts to incorporate machine learning (ML)-based approaches to improve the applicability of PRISM theory to realistic multicomponent systems. Our preliminary results show that the ML-enhanced PRISM theory performs favorably in comparison to widely-used theoretical approaches. We then applied the ML closure to model the results of small-angle neutron scattering experiments on polymer solutions, again showing superior performance compared to traditional approaches. I will close with a discussion of how such combined theoretical-experimental tools could be applied to the development of the next generation of sustainable consumer products.