Polymer Structure Prediction from First Principles

Developing a large database of polymers structures and properties, for which suitable polymer structural models are a prerequisite, is critical for polymer informatics. We present a simple strategy, referred to as polymer structure predictor (PSP), for predicting the crystal structural models of linear polymers, given their chain-level atomic connectivity information. The PSP, which was designed specifically for polymers, relies on properly defining and sampling the configuration space. Using this approach, we have successfully recovered eight known crystal structures of six common linear polymers, including polyethylene, polyvinylidene fluoride, poly­(vinyl chloride), poly­(p-phenylene sulfide), polyacrylonitrile, and poly-2,5-benzoxazole, while discovering some new stable structures of three of them, i.e., polyvinylidene fluoride, polyacrylonitrile, and poly­(p-phenylene sulfide). The PSP is very simple, highly scalable, suitable for automatic workflows, and comparable to the best major structure prediction method in terms of efficiency in polymer crystal structure prediction. Although challenges in comprehensively accounting for possible chain-level conformations remain, the PSP will be very useful in efficiently generating polymer data and strengthening the emerging polymer informatics ecosystems.

开发包含高分子结构与性能的大型数据库，而合适的高分子结构模型是该数据库的先决条件，这对于高分子信息学（polymer informatics）而言至关重要。我们提出了一种被称为高分子结构预测器（polymer structure predictor, PSP）的简便策略，可在给定链级原子连接信息的前提下，预测线性高分子的晶体结构模型。该PSP专为高分子设计，依赖于对构象空间（configuration space）的合理定义与采样。通过这一方法，我们成功复现了6种常见线性高分子的8种已知晶体结构，涵盖聚乙烯、聚偏氟乙烯、聚氯乙烯、聚对苯硫醚、聚丙烯腈以及聚2,5-苯并恶唑；同时还为其中3种高分子——即聚偏氟乙烯、聚丙烯腈与聚对苯硫醚——发现了若干新的稳定结构。PSP极为简便、可扩展性极强，适配自动化工作流，且在高分子晶体结构预测的效率方面，可与当前主流的顶尖结构预测方法相媲美。尽管在全面覆盖所有可能的链级构象方面仍存在挑战，但PSP将在高效生成高分子数据、推动新兴的高分子信息学生态系统发展方面发挥重要作用。