Ce Song, 1 Xigao Jian 
Dalian University of Technology, Dalian, 116024, China 

EXTENDED ABSTRACT: Polymers have become pivotal and indispensable components for numerous domains spanning from daily life to high-technology. However, traditional intuition-driven development through trial-and-error approaches, demanding high costs and long research cycles, is impractical to efficiently pinpoint optimal candidates in the nearly infinite chemical space of polymers. To accelerate the discovery and deployment of advanced materials, the Materials Genome Initiative (MGI) advocates the fourth paradigm of data-driven materials research and development (R&D), proposing the strategic target of reducing both R&D cycle time and cost by half. This work proposes a new data-driven approach for rational design of high-performance polymers, following the instruction of MGI. To implement this approach for developing high-performance poly(aryl ether)s (PAEs), an organized database containing more than 2800 PAEs was constructed and a new presentation and featurization framework was designed and employed to characterize molecular compositions of PAEs. By means of the database and the presentation and featurization framework, machine learning models with high predictive accuracies were constructed and favorable and unfavorable genes for high-performance PAEs were unraveled. Thereby, molecular design strategies and mechanistic perspectives for developing high-performance PAEs were presented, and a series of new PAEs whose glass transition temperature can reach 368'C were proposed. The exemplified application in developing high-performance PAEs suggests that the proposed approach is supposed to boost the development of other genres of polymers. This work has been selected as the front cover of Journal of Materials Chemistry A.
Keywords: Materials Genome Initiative; poly(aryl ether)s; machine learning; polymer genome
REFERENCES
[l] C. Song, H. Gu, et al. Journal of Materials Chemistry A, 11(32), (2023) 16985-16994.