Data-Driven\nStructure Recognition of Scanning Tunneling\nMicroscopy Images in a Case of Iron Carbide

Scanning tunneling microscopy (STM) serves as a critical\ntool for\nhigh-resolution surface imaging, yet deciphering the atomic structures\nfrom STM images on multielement surfaces, such as oxides and carbides,\nremains a challenging task that heavily relies on the expertise and\nintuition of researchers. In this study, we introduce a data-driven\nmethod for rapid structural recognition from STM images. This method\ninvolves extracting structural features, filtering through a structural\ndatabase, and matching with simulated STM images and surface energy\nanalyses, thereby providing researchers with several of the most probable\nstructures. We demonstrate the capabilities of this technique using\nour previously reported iron carbide grown on an Fe(110) crystal.\nBy proposing a candidate structure set and establishing a comprehensive\ndatabase linking STM images to corresponding structures and surface\nenergies, we selected 6 out of more than 10 000 possible surfaces.\nOn the basis of these 6 recommendations, researchers can conveniently\ndetermine the real surface structures. Our work provides an efficient\ntool for the structure recognition of STM images to construct surface\nstructures, potentially serving as a universal auxiliary tool for\nSTM structural analysis.