CoSSPIA – Computational Screening of Sustainable Polymers for Industrial Applications

Polymers are ubiquitous in our society thanks to their durability, low manufacturing costs, and ability to be formulated into diverse materials. However, due to environmental concerns, the replacement of fossil-based polymers with biopolymers has emerged as a promising strategy towards a more circular economy.

Since nature only provides us with a limited number of bio-based building blocks, the fine tuning of the polymer properties is expected to be achieved through combining these building blocks into copolymers and polymer blends. Since testing all possible combinations via experiments is too slow, costly and inefficient, the computational route has emerged as an attractive alternative. On the one hand, Molecular Dynamics (MD) simulations of polymers can accurately reproduce the experimental behavior for a series of properties, but are in practice limited to some hundred polymers.

On the other hand, data-driven solutions such as Machine Learning (ML) allow to screen millions of candidate polymers, but their development is hindered by the scarcity of high-quality training data. In this project, the team will combine the two approaches and perform high-throughput (HT) MD simulations to generate an extended and consistent computational dataset of polymer properties that can subsequently be used for the training of predictive ML models.