Optimised Workflows to Calculate Solubility and Diffusion Parameters for Polymeric Systems

Material ageing during storage can lead to changes in the structure and functionality of various polymers. By quantifying how molecules and contaminants diffuse into and interact with a material, we can predict how its properties will change over time. In the development of our large-scale baseline system models, we often need to include a range of materials and parameterise these materials accordingly.

Often, we are considering processes such as reaction of materials and onward diffusion of species along with respective solubilities. We approach this very much from a classical perspective by building a representative material cell, geometry optimising and then using molecular dynamics (MD) to study diffusion of species and Monte Carlo (MC) to study solubility of species.

There are numerous challenges in this approach:

• Building representative cells for a material – usually a polymer i.e. chain length / number of monomers / number of atoms overall.
• Optimising the cell ahead of any simulation – going from an initial packed cell to a relaxed simulation cell ready for onward simulation in MD and MC.
• Running suitable numbers of cells / repeats to suitably sample the material / property of interest.
• Choosing an appropriate forcefield – the methods currently applied rely heavily on the accuracy of the force field. Additionally, transferability across different systems can be an issue, requiring empirical adjustments or the development of a new forcefield for each system investigated.

Therefore, the challenge is to explore alternative approaches to address the challenges outlined above. This could include implementation of machine learning to augment or develop suitable forcefields within MD and MC simulations to predict solubility and diffusion coefficients; data-driven models to investigate the potential of statistical learning methods or a hybrid combination of elements from established methods.