The accuracy of quantitative image matching for HRTEM applications

The accuracy of atomistic model refinement in quantitative high resolution transmission electron microscopy was analyzed using simulated images of aluminum, calculated from crystalline atomistic models and including a thin coating of amorphous aluminum. The images, which contained a realistic background 'noise' resulting from the amorphous aluminum layers, were processed using two filtering methods, and then compared to images simulated from models without the amorphous layer. The effects of filtering were assessed by quantitatively matching the processed images with those from a perfect structure, and a method to select optimized matching criteria for comparing simulated and experimental images was developed. Errors introduced by filtering noise are converted into accuracy of atomic column determination. (c) 2004 Elsevier Inc. All rights reserved.