Fig. 2: Illustrations of the reference and target molecular models, tasks, and labels in the HiDiscover protocol for studying Li-ion transport in the covalent organic framework COF-PEO-3 containing poly(ethylene oxide) (PEO) chains. | Nature Communications

Fig. 2: Illustrations of the reference and target molecular models, tasks, and labels in the HiDiscover protocol for studying Li-ion transport in the covalent organic framework COF-PEO-3 containing poly(ethylene oxide) (PEO) chains.

From: Hierarchical incremental learning deciphers molecular arrangements in multi-component materials

Fig. 2: Illustrations of the reference and target molecular models, tasks, and labels in the HiDiscover protocol for studying Li-ion transport in the covalent organic framework COF-PEO-3 containing poly(ethylene oxide) (PEO) chains.The alternative text for this image may have been generated using AI.

MD models 1–3 correspond to crystalline and amorphous LiClO4 phases and a LiClO4 cluster, respectively. Molecular dynamics (MD) models 4–6 correspond to LiClO4 blended in COF-42 with various weight ratios. Structures with smaller weight ratios capture the arrangements close to the 2D COF framework, while those with higher ratios explore the arrangements in the 2D COF pores. MD model 7 is LiClO4 blended in pure (PEO)3. MD model 8 is the target system. Circles in the chemical structures highlight the regions in the 2D COFs and (PEO)3 that are expected to interact strongly with Li+ (green, purple, blue, and brown circles correspond to tasks ζA, ζB, ζC, and ζD, respectively). Further information is provided in Sections 2–3 of the SI. The bottom part of the figure illustrates the process of model training and applying the trained model in data analysis in the HiDiscover protocol.

Back to article page