Neural Network-Assisted Data Processing Improved Tomography Characterizations of Reverse Osmosis Polyamide Layers

Electron tomography (ET) can providecomprehensive and detailedstructural information on the polyamide selective layer of reverseosmosis membranes. Albeit powerful, the segmentation of low-contrastgrayscale images from transmission electron microscope tomographyremains a challenge. In this study, a neural network-assisted imagesegmentation method was developed to solve this problem. A well-trainedneural network based on the U-Net model was able to automaticallyidentify the boundary between polymer and vacuum, which reached a93.0% compatibility with the baseline within a 0.5 h effort of datasegmentation, significantly improved upon other segmentation methods.The results were compared with those of a previous study and conventionalcharacterization results for validation. In addition, we performedtopography analyses to quantitatively characterize the surface geometryand proposed an entire "sample preparation to 3D structuralmodel" workflow that calibrates the data sets and minimizestedious manual inputs of ET. For the first time, ET is qualified tobe a powerful quantitative characterization technique for separationmembranes and other polymer materials widely used in energy and environmentalapplications.