ELECTRONIC-STRUCTURE CLASSIFICATIONS USING SCANNING-TUNNELING-MICROSCOPY CONDUCTANCE IMAGING

The electronic structure of atomic surfaces is imaged by applying mult ivariate image classification techniques to multibias conductance data measured using scanning tunneling microscopy. Image pixels are groupe d into classes according to shared conductance characteristics. The im age pixels, when color coded by class, produce an image that chemicall y distinguishes surface electronic features over the entire area of a multibias conductance image. Such ''classed'' images reveal surface fe atures not always evident in a topograph. This article describes the e xperimental technique used to record multibias conductance images, how image pixels are grouped in a mathematical, classification space, how a computed grouping algorithm can be employed to group pixels with si milar conductance characteristics in any number of dimensions, and fin ally how the quality of the resulting classed images can be evaluated using a computed, combinatorial analysis of the full dimensional space in which the classification is performed. (C) 1998 American Institute of Physics. [S0021-8979(98)01917-3].