polyChainStructures
收藏资源简介:
polyChainStructures数据集由佐治亚理工学院材料科学与工程学院计算科学与工程系创建,包含3855个DFT优化后的无限聚合物链结构,用于训练和评估polyGen模型。该模型旨在从最少的输入信息,如重复单元的化学组成,生成真实的聚合物结构,以加快聚合物设计过程。数据集为polyGen提供了训练数据,使其能够生成多种聚合物链构象,尽管在处理原子数较多的重复单元时性能有所下降。该数据集在聚合物科学领域具有开创性意义,首次证明了在考虑聚合物内在结构灵活性的情况下预测真实原子级聚合物构象的可能性。
The polyChainStructures dataset was developed by the School of Computational Science and Engineering, College of Materials Science and Engineering, Georgia Institute of Technology. It contains 3855 DFT-optimized infinite polymer chain structures, which are curated for training and evaluating the polyGen model. The polyGen model is designed to generate realistic polymer structures from minimal input information, such as the chemical composition of the repeating unit, with the goal of accelerating the polymer design process. This dataset provides training data for polyGen, allowing the model to produce a wide range of polymer chain conformations, albeit its performance degrades when processing repeating units with a high number of atoms. The polyChainStructures dataset holds pioneering significance in polymer science, as it is the first work to validate the feasibility of predicting realistic atomistic polymer conformations while accounting for the inherent structural flexibility of polymers.

- 1polyGen: A Learning Framework for Atomic-level Polymer Structure Generation佐治亚理工学院材料科学与工程学院计算科学与工程系 · 2025年



