Measurement of GaAs structure factors from the diffraction of parallel and convergent electron nanoprobes

Abstract

Present techniques in quantitative Transmission Electron Microscopy (TEM) frequently rely on the comparison of experimental and simulated data. For example, the evaluation of the chemical composition in ternary semiconductor nanostructures (CELFA, [1]) is based on both the computation and the measurement of the contrast in (200) lattice fringe images. This technique takes advantage of the chemical sensitivity of the 200 reflections in zincblende crystals. High accuracy of the CELFA method requires a precise knowledge of structure factors (SF) to simulate the correct contrast. Conventionally, theoretical structure factors are calculated from atomic scattering amplitudes (ASA) obtained by the isolated atom approximation [2]. However, the redis-tribution of electrons due to the bonding in crystals may significantly affect ASA and SF of low order, such as 002. We recently published modified ASA obtained by density functional theory (DFT) for the 002 reflection in In x Ga 1-x As alloys [3]. Here, we present a new approach to measure SF from diffraction patterns formed by parallel electron nanoprobes. To judge the plausibility of this method, the results will be compared with those we obtained from convergent beam electron diffraction (CBED) and from DFT. The high efficiency in the refinement of e.g. atom positions, Debye parameters and specimen thickness using parallel beam electron diffraction (PBED) has motivated the development of the software package ELSTRU by J. Jansen during the last decade. We use the subroutine GREED to extract the background-corrected Bragg intensities from experimental PBED patterns. The subroutine MSLS [4] is then used to refine orienta-tion, specimen thickness and Debye-Waller factors using up to 15 data sets simulta-neously. The refinement of low-order structure factors is finally performed by the Bloch-wave refinement program Bloch4TEM with the MSLS results as input. The applicability and the precision of this procedure was initially checked with simulated diffraction patterns containing thermal diffuse background. We deduced conditions to recognize physically plausible refinement results for the thicknesses and the Debye-parameters, because the simultaneous refinement of both quantities may be ambigous