Discovery of polymers with high bulk modulus and low thermal conductivity using a hybrid generative pipeline

HX Yue and SY Wang and XP Sha and QY Wang, CHEMICAL ENGINEERING JOURNAL, 518, 164763 (2025).

DOI: 10.1016/j.cej.2025.164763

Polymers with high mechanical rigidity and low thermal conductivity are critical for applications in aerospace, energy systems, and deep-sea exploration, yet their design remains a formidable challenge due to the inherent trade-off between bulk modulus and thermal conductivity. Here, we introduce a hybrid genetic algorithm-reinforcement learning (GA-RL) framework integrated with cheminformatic validation to systematically discover polymers that defy this trade-off. By combining evolutionary exploration of monomer sequences with goal-directed optimization, our method identified chemically valid and synthesizable candidates by SMiPoly, while prioritizing high bulk modulus (K) and low thermal conductivity (T-c ). The framework identifies 10 novel polymers with T-c as low as 0.075 Wm(-1)K-1 and K exceeding 4.59 GPa, validated by molecular dynamics simulations, with T-c predictions achieving <5 % error for a representative candidate and K predictions exhibiting approximately 10 % deviation. This work demonstrates that machine learning-guided monomer sequencing is promising to resolve long- standing property conflicts, paving the way for computationally driven design of multifunctional polymers for extreme environments.

Return to Publications page