卷积神经网络在新型非富勒烯受体分子生成与性能预测上的应用  

Application of convolutional neural networks in the generation and property prediction of novel non-fullerene acceptor molecules

作　　者：杨新玉 彭师平 赵仪[1] YANG Xinyu;PENG Shiping;ZHAO Yi(State Key Laboratory of Physical Chemistry of Solid Surfaces,Fujian Provincial Key Laboratory of Theoretical and Computational Chemistry,College of Chemistry and Chemical Engineering,Xiamen University,Xiamen 361005,China)

机构地区：[1]厦门大学化学化工学院,固体表面物理化学国家重点实验室,福建省理论与计算化学重点实验室,福建厦门361005

出　　处：《厦门大学学报（自然科学版）》2022年第5期777-785,共9页Journal of Xiamen University：Natural Science

基　　金：国家自然科学基金(22033006,21833006,21773191)。

摘　　要：近年来,有机太阳能电池中非富勒烯小分子受体因其拓展了吸收光谱的范围、能够调节激子解离能量和具有灵活的给体-受体形貌等优点使得器件效率越来越接近产业化的目标.本研究借助本课题组之前构建的分子生成和性质预测的卷积神经网络模型,来生成和筛选出具有高效解离激子的前线轨道能量的新型非富勒烯小分子受体.首先生成模型经数据库充分训练并利用小数据集进行微调后生成200多个接近目标轨道能量(最高占据分子轨道(the highest occupied molecular orbital,HOMO)和最低未占据分子轨道(the lowest unoccupied molecular orbital,LUMO)的能量分别为-5.60和-3.60 eV)的分子,然后利用预测模型进一步筛选并预测分子碎片对轨道能量的贡献,接着将这些分子与数据库中具有相近前线轨道能量的分子共同聚类挑选出具有不同化学空间的10个新型受体分子,最后通过从头算验证了轨道能级、分子碎片对轨道贡献预测的准确性,并给出分子光吸收的振子强度.进一步利用生成和预测模型提供了具有另一组轨道能量(HOMO和LUMO能量分别为-5.10和-3.10 eV)的10个非富勒烯小分子,其性质预测与从头算结果一致,证明了生成和预测模型的鲁棒性和结果的可靠性.本研究预测的分子也提供了设计具有高性能非富勒烯受体分子骨架的思路.In recent years,the efficiency of non-fullerene small molecule acceptors in organic solar cells is getting closer and closer to the goal of industrialization because of its expanded absorption spectrum,the ability to adjust the dissociation energy of the exciton,and the flexible donor-acceptor morphology.In this study,we used the convolution neural network model proposed by our group for molecular generation and property prediction to generate and screen new non-fullerene small molecular acceptors with given frontier orbital energies for efficient exciton dissociation.After the molecular generation model was fully trained in the database and fine-tuned with a small data set,more than two hundred molecules close to the targeted orbital energy(the highest occupied molecular orbital(HOMO)and the lowest unoccupied molecular orbital(LUMO)energies of-5.60 and-3.60 eV,respectively)were generated.Then the prediction model was used to further screen the required molecules and predict the contribution of molecular fragments to the orbital energies.After that,these molecules and molecules with similar frontier orbital energies in the database were clustered together and ten new acceptor molecules with different chemical spaces were selected.Finally,the accuracy of the prediction of orbital energies and the contribution of molecular fragments to the orbitals were verified by ab initio calculations,and the oscillator intensity of the molecular light absorption was given.We further used the generation and prediction model to provide ten non-fullerene small molecules with another set of orbital energies(HOMO and LUMO energies of-5.10 and-3.10 eV,respectively).The predicted properties were consistent with the ab initio results,which proves the robustness of the generation and prediction model and the reliability of the results.The molecules predicted in this study also provide the design scheme of potential molecular scaffold with high-performance non-fullerene acceptors.

关 键 词：卷积神经网络 非富勒烯受体 前线分子轨道能量 

分 类 号：O641[理学—物理化学]