Data-science-based reconstruction of 3-D membrane pore structure using a single 2-D micrograph

Abstract

Conventional 2-D scanning electron microscopy (SEM) is commonly used to rapidly and qualitatively evaluate membrane pore structure. Quantitative 2-D analyses of pore sizes can be extracted from SEM, but without information about 3-D spatial arrangement and connectivity, which are crucial to the understanding of membrane pore structure. Meanwhile, experimental 3-D reconstruction via tomography is complex, expensive, and not easily accessible. Here, we employ data science tools to demonstrate a proof-of-principle reconstruction of the 3-D structure of a membrane using a single 2-D image pulled from a 3-D tomographic data set. The reconstructed and experimental 3-D structures were then directly compared, with important properties such as mean pore radius, mean throat radius, coordination number and tortuosity differing by less than 15%. The developed algorithm could dramatically improve the ability of the membrane community to characterize membranes, accelerating the design and synthesis of membranes with desired structural and transport properties.