EEL Spectrum Imaging of Extended Defects in Diamond using UHV Enfina in a dedicated STEM *U. Bangert, *A. J. Harvey, **R. Jones, **C. J. Fall, ***A. Papworth * Department of Physics, UMIST, Manchester M60 1QD, UK ** School of Physics, University of Exeter, Exeter EX4 4QL *** Department of Materials Science, University of Liverpool, Liverpool L69 3GH, UK We employ electron energy loss spectrum imaging using an Enfina imaging system adapted to UHV conditions in a cold FEG STEM (VG 601). The samples in this study are quasi-single crystalline CVD grown diamond films on iridium buffers [1]. Data analysis was carried out with the Digital Micrograph software from Gatan, which allows spectrum manipulation such as deconvolution or Gaussian fitting to be applied to an entire spectrum map. We are interested in the low loss region (2- 10 eV) and, to be able to separate the zero loss peak from the low loss signal, we use an energy dispersion of 0.02 eV/channel. Backgrounds were subtracted from each spectrum using a power law fitting to the tail of the zero loss peak. Spectrum maps (of the order of 100 nm in width, comprising of up to 1000 spectra) were taken of dislocations and stacking faults, and they always included surrounding un-dislocated material for comparison. In order to exclude effects of local thickness variations and elastic scattering to the low loss signal, each spectrum was self-normalised. This was achieved via two methods by which contrast maps were extracted from the spectrum maps. These are based on (1) the ratio of integrated counts in different energy windows, and (2) intensity ratios of Gaussian peaks fitted to each spectrum. In both cases the evaluation is based on the knowledge of the spectrum shape, and the energy region in which changes are to be expected from extended defects. Calculated spectra have been obtained from first principle methods [2]. All dislocation types in diamond with pure cores are predicted to introduce changes to the density of states in the lower conduction band. This is demonstrated in Fig. 1 for a shuffle dislocation: a change of the intensity in the dislocation spectrum occurs in the energy region between 6 and 8 eV, when compared to the bulk spectrum. Based on this knowledge, in method (1) two energy windows were defined, the first was used as a probe for extra states, the second was the reference window normally set between 8 and 10 eV, a region where the spectra are less affected by the dislocations (see Fig.1). The values of the integrated counts in both windows were divided by each other for each spectrum, and the value of this ratio was then displayed as grey value map. Fig. 3 depicts additional joint density of states at a partial dislocation bounding a stacking fault. As the probing energy window is stepped towards higher energies, the intensity in the intensity ratio maps changes. At around 6 eV, the dislocation core states show up most pronounced, in accordance with the calculations. In method (2) the experimental spectra were deconvolved into a number of Gaussian functions as shown in Fig. 2, and the ratio of the intensity of the Gaussians with the lower energy to that of the bulk Gaussian were displayed as intensity map. The results are essentially the same as obtained from method (1), however, elastic scattering contrast is further reduced and the spatial resolution of the defect states improved. Fig. 4 depicts energy states at the un-split end of a dislocation with a node, which was commonly observed in CVD diamond. The form of the contrast and the details of the dislocation spectrum (not shown here) suggest the involvement of sp 2 -bonding at the dislocation core as predicted for a shuffle-type dislocation [2]. Fig. 4 (left hand panel) shows the dislocation Microsc Microanal 10(Suppl 2), 2004 Copyright 2004 Microscopy Society of America DOI: 10.1017/S143127604882618 886