Electron tomography of defects

Abstract

Tomography of crystal defects in the electron microscope was rst attempted in 2005 by the author and colleagues. This thesis further develops the technique, using a variety of samples and methods. Use of a more optimised, commercial tomographic reconstruction program on the original GaN weakbeam dark- eld (WBDF) tilt series gave a ner reconstruction with lower background, line width 10-20 nm. Four WBDF tilt series were obtained of a microcrack surrounded by dislocations in a sample of indented silicon, tilt axes parallel to g = 220, 220, 400 and 040. Moir e fringes in the defect impaired alignment and reconstruction. The e ect on reconstruction of moir e fringe motion with tilt was simulated, resulting in an array of rods, not a at plane. Dislocations in a TiAl alloy were reconstructed from WBDF images with no thickness contours, giving an exceptionally clear reconstruction. The e ect of misalignment of the tilt axis with systematic row g(ng) was assessed by simulating tilt series with di raction condition variation across the tilt range of n = 0, 1 and 2. Misalignment changed the inclination of the reconstructed dislocation with the foil surfaces, and elongated the reconstruction in the foil normal direction; this may explain elongation additional to the missing wedge e ect in experiments. Tomography from annular dark- eld (ADF) STEM dislocation images was also attempted. A tilt series was obtained from the GaN sample; the reconstructed dislocations had a core of bright intensity of comparable width to WBDF reconstructions, with a surrounding region of low intensity to 60 nm width. An ADF STEM reconstruction was obtained from the Si sample at the same microcrack as for WBDF; here automatic specimen drift correction in tomography acquisition software succeeded, a signi cant improvement. The microcrack surfaces in Si reconstructed as faint planes and dislocations were recovered as less fragmented lines than from the WBDF reconstruction. ADF STEM tomography was also carried out on the TiAl sample, using an detector inner angle ( in) that included the rst order Bragg spots (in other series in had been 4-6 B). Extinctions occurred which were dependent on tilt; this produced only weak lines in the reconstruction. Bragg scattering in the ADF STEM image was estimated by summing simulated dark- eld dislocation images from all Bragg beams at a zone axis; a double line was produced. It was hypothised that choosing the inner detector angle to omit these rst Bragg peaks may preclude most dynamical image features. Additional thermal di use scattering (TDS) intensity due to dilatation around an edge dislocation was estimated and found to be insigni cant. The Huang scattering cross section was estimated and found to be 9 A, ten times thinner than experimental ADF STEM dislocation images. The remaining intensity may be from changes to TDS from Bloch wave transitions at the dislocation; assessing this as a function of tilt is for further work. On simple assessment, only three possible axial channeling orientations were found over the tilt range for GaN; if this is typical, dechanneling contrast probably does not apply to defect tomography.