Theoretical models for tip/surface interaction and for the atomic force microscopic imaging of ionic surfaces were considered. It was pointed out that the study of surface point defects faced several problems, not least of which was how to locate them. That is, with a bulk defect concentration of 10-5 sites and a scanning area of some 30 x 30 sites, a search was statistically unfavorable. Also, only those defects which were present at the same surface site during several scans (a period of some ms) could be detected. These also had to have high barriers to diffusion, or the measurements had to be performed at low temperatures. Many defects were active in the adsorption of molecules, so that only high concentrations of stable defects could be observed under ultra-high vacuum. Although cubic insulators could be easily cleaved, they tended to charge under ultra-high vacuum and adsorbed molecules at defect sites in air. Overall, it was logical to study only surface defects which were created  in situ. The surfaces of epitaxially grown films were usually smoother than those of cleaved samples. Electronic defects at these surfaces could be introduced by using the tunnelling current of scanning tunnelling microscopy. Unfortunately suitable materials were rarely exposed to atomic force microscopy and scanning tunnelling microscopy at the same time. Finally, it was difficult to interpret images which contained ostensible defects. Such defects could be imaged only against the background of a perfect lattice. Part of detection problem was to prove that this was not a defect in the image of the perfect lattice due to some artefact. It was therefore necessary to use pattern recognition methods to compare theoretically predicted and experimentally obtained images.

A.Shluger, A.Livshits, H.Rohl: Materials Science Forum, 1997, 239-241, 651-6