Comprehensive Framework for the Design of High-Performing and Sustainable Polyamides

Developing high-performance materials with tailored properties is a key challenge that requires innovative molecular design methods. This study presents a machine learning-based approach, integrating Monte Carlo tree search and a recurrent neural network for the molecular design of polyamide systems, and we evaluate their properties with molecular dynamics simulations. By benchmarking against the performance of polyimides, we guide our algorithm to design sustainable polyamides with superior properties. A key consideration in our design approach is the incorporation of biodegradability as a critical design factor, in line with the growing demand for sustainable materials. We use an objective function targeting an optimal profile of low octanol–water partition coefficient (log P), high thermal stability (TSef), and substantial thermal conductivity (k). With 20 independent runs, which resulted in an evaluation of 133,655 monomers for design compatibility and generation of 5255 polyamide systems, our approach successfully designed promising 157 polyamide monomers, satisfying the criteria of log P < 0, TSef ≥ 2, and k ≥ 0.5 W/m K. These selections were validated by additional molecular dynamics (MD) calculations, revealing our objective function leads to sustainable and high-performance materials. Furthermore, we performed a detailed analysis of the structure–property relationship. Our algorithm successfully demonstrates the effectiveness of this integrated computational approach in guiding the design of novel polyamides.