2023-08-012024-05-13https://scholars.lib.ntu.edu.tw/handle/123456789/651383奈米複合薄膜為功能性新興材料，由表面修飾之奈米粒子與高分子基質所組成。在適當的設計下，這些奈米粒子除了可作為固體填料來增加薄膜的機械強度外，奈米粒子本身與其表面嫁接的短鏈長高分子亦可形成特殊的自組裝結構，與高分子基質交互形成多元的多孔薄膜型態，使分子得以有效滲透，而在氣體分離與水過濾等領域有應用性。在過去十年間，雖然奈米複合薄膜的研究論文在Web of Science資料庫中有上萬篇，其中針對薄膜型態與分子擴散係數的研究卻僅有一百多篇，而如此不對稱的研究比例凸顯了此課題的挑戰性與待解決性。由於薄膜的形成過程與最後的微結構受到多種跨尺度作用力的影響，因此本計畫將以嚴謹的統計力學理論結合多尺度模擬，預測不同奈米粒子與高分子組成下的薄膜形態與分子於薄膜之擴散係數。 Nanocomposite membranes (NMs) consisting of oligomer-grafted nanoparticles and a polymer matrix show versatile usages and design strategies. These grafted nanoparticles form the “fillers” to induce desired morphology and mechanical strength of NMs. Moreover, the functionalized nanoparticles can be tactically designed to improve the overall permeability of the membrane in the applications of gas separation and drinking water filtration. As the molecular transport properties depend on the porous microstructure and physicochemical interactions between the separating molecules, functionalizing ligands, and the pore surfaces, a successful membrane design would require careful considerations of suitable membrane morphology with plausible descriptions for intermolecular interactions at the target processing and operating conditions. In the past 10 years, the Web of Science keyword “nanocomposite membrane” search returns 10675 publications. However, including “diffusion and morphology” in the search returns only 126 publications, most of which only perform an experimental investigation on a specific recipe of membranes without systematically assessing the connections between molecular transport properties and mesoscale morphology. This project will utilize statistical mechanical principles and employ multiscale modeling to predict the microstructures of NMs at various compositions and contents of nanoparticles and polymers, and further correlate membrane morphology with the resulting molecular diffusivity.功能性奈米粒子;高分子物理;統計力學;多尺度模擬;物理導向神經網絡;Functionalized nanoparticles; polymer physics; statistical mechanics; multiscale modeling; physics-informed neural networks