Measurement of thickness of thin coatings on rough substrates via computational analysis of SEM images

This work introduces a computational method to quantify the thickness of thin films deposited on highly rough substrates analyzing top-down Scanning Electron Microscope (SEM) images. The method entails measuring the bright areas of top-down SEM images of the rough surface obtained before and after deposition, allowing for the prediction of film thickness through the ratio of bright area enhancement caused by deposition to the average perimeter of these areas before and after deposition. Validation of this technique was conducted via synthetic SEM images with predefined film thicknesses, incorporating simple and complex substrate morphologies generated through Diffusion-Limited Aggregation (DLA) simulations for added realism. Experimental applications were explored through the analysis of SEM images of plasma-etched polymer (PMMA) surfaces coated with carbyne and of nanorods of TiO 2 coated with NiO, demonstrating the method's efficacy across varying surface roughness and morphologies. This work lays the foundation for future advancements, including the implementation of a neural network trained on synthetic datasets to enhance the measurement accuracy of coating thickness on rough substrates as well as the reconstruction of true surface morphologies prior to metal layer sputtering via SEM image analysis. • Quantified thin film thickness on rough substrates using top-down SEM images, avoiding destructive cross-sectional imaging. • Validated with synthetic SEM images of known thickness and real cases incl. carbyne-coated PMMA and NiO-coated TiO 2 nanorods. • Demonstrated effectiveness across diverse surface morphologies. • Provided a robust alternative to traditional techniques, even for substrates with extreme roughness.