The surface of metal oxides is of technological importance and is extensively used as a substrate for various electronic and chemical applications. A real surface, however, is not a perfectly well-defined and clean surface, but rather contains a diverse class of atomistic defects. Here, we show the direct determination of the 3D surface atomic structure of SrTiO3 (001) including termination layers and atomistic defects such as vacancies, adatoms, ledges, kinks, and their complex combinations, by using depth sectioning of atomic-resolution annular dark-field scanning transmission electron microscopy (ADF STEM). To overcome the poor depth resolution in STEM, we statistically analyze the column by column depth profiles of ADF STEM images with a Bayesian framework fitting algorithm, and we achieve depth resolution at the entrance surface of ±0.9 Å for 1518 individual atomic columns. The present atomic-resolution 3D electron microscopy at the surface will provide fertile ground especially in surface science.
Keywords: Markov chain Monte Carlo; atomic-resolution STEM; depth sectioning; surface imaging; surface point defects.