Predicting Atomic Force Microscopy Topography from Optical Microscopes Using Deep Learning

Atomic force microscopy (AFM) is routinely used as a metrological tool among diverse scientific and engineering disciplines. A typical AFM, however, is intrinsically limited by low throughput and is inoperable under extreme conditions. Thus, this work attempts to provide an alternative to a conventional optical microscope (OM) by training a deep learning model to predict surface topography from surface OM images. The feasibility of this novel methodology is shown with germanium-on-nothing (GON) samples, which are self-assembled structures that undergo surface and sub-surface morphological transformations upon high-temperature annealing. Their transformed surface topographies are predicted based on the OM-AFM correlation of three different surfaces, bearing an error of about 15% with a 1.72x resolution upscale from OM to AFM. The OM-based approach brings about significant improvement in topography measurement throughput (equivalent to OM acquisition rate, up to 200 frames per second) and area (approximate to 1 mm(2)). Furthermore, this method is operable even under extreme environments when an in situ measurement is impossible. Based on such competence, the model's simultaneous application in further specimen analysis, namely surface morphological classification and simulation of dynamic surfaces' transformation is also demonstrated.