Measurements of compositional profiles using analytical transmission electron microscopy (ATEM) on a spatial scale below the resolution of the electron microprobe (EMPA) has the potential to open new horizons in the study of natural as well as experimental samples. Such short zoning profiles in natural minerals can provide information about short lived or low temperature thermal events and may help in the determination of coexisting mineral compositions even when minerals are sharply zoned. For experimental studies of diffusion, ATEM offers the unique opportunity to simultaneously characterize the diffusion zone chemically as well as structurally. We have improved upon our earlier work on chemical analysis using ATEM (see Annual Report 1995) and begun to structurally characterize the interface region of synthetic diffusion couples. We have already shown that for cases where comparison is possible, compositional profiles in Fe-Mg olivines may be obtained with comparable precision using EMPA and ATEM provided the thin film criterion is fulfilled for the TEM samples.
However, the restriction of analyses to regions where absorption and fluorescence is negligible is a handicap for many practical purposes. Firstly, it is not clear a priori at which points in the sample the thin-film assumption is justified. Secondly, thinning during sample preparation is inhomogeneous and results in uneven sample thickness. Thirdly, even if the sample itself is properly thinned it may be necessary to tilt it within the sample chamber to avoid artefacts such as electron channelling which arise when the sample is in a strong diffracting orientation. Such tilting can increase the effective thickness of the sample through which X-rays travel. This was found to be a particularly important problem for the pre-oriented single crystals of olivines used to prepare our diffusion couples. Therefore, it is worthwhile to develop methods which allow the thin-film criterion to be relaxed such that analyses can be obtained from samples with a range of thicknesses, within some limits.
To achieve this objective we have adapted the approach of parameterless correction of van Capellen and measured the so called zero-thickness k-factors for the elements of interest in our study of olivines. For proper absorption correction to the raw data during a chemical analysis, in addition to the k-factors, it is necessary to know the effective distance that the emergent X-ray beam travels through the sample. This distance depends on a number of factors such as the true sample thickness and sample-detector geometry. An estimate of this distance can be obtained from the absorption of the oxygen-Ka line in stoichiometric oxides and silicates. We have found that using such calculated effective thicknesses and the zero thickness k-factors obtained by us, it is possible to obtain high quality chemical analyses from samples with a variety of orientations and thicknesses. Most notably, we are now able to measure diffusion profiles with a precision comparable to EMPA data in regions that do not fulfil the thin-film criterion.
Preliminary structural analyses of the interface region in synthetic olivine-olivine (Fo-Fo82) diffusion couples reveal that the concentration of dislocations at the interface depends on the extent of initial orientational mismatch between the lattice of the two crystals. Further characterization of the interface region using weak beam and high resolution techniques show that the appearance and defect structure of interfaces in diffusion couples annealed at low temperatures or for short times at high temperatures are quite distinctive.
There appears to be a high concentration of Frank-type dislocation loops with a  Burger's vector in these samples. Detailed analyses are being carried out and the possible influence of these structures on diffusion processes is being further investigated.