It is interesting to note that is of the same order of magnitude as the step size x and the primary beam size, i.e. 15 μm, indicating that the entire data reduction and reconstruction process does not significantly degrade the spatial resolution (in this direction). The apparent interface width , however, is roughly 2× higher than . This effect can also be seen in Fig. 6 where the XRD tomograms of rutile and barite as obtained by the FBP and MLEM methods are compared. In the FBP tomograms, sharply defined layers can be observed, but the tomograms as a whole contain numerous artifacts outside the sample area and are noisy. [Some of the algebraic reconstruction techniques (ART and SIRT) were also evaluated in this context but yielded similar results as the FBP method.] The MLEM tomograms, on the other hand, show a more robust reconstruction result, i.e. smoother layers without artifacts are obtained. However, the layer edges are blurred/averaged out compared to the FBP distributions. Since the average thickness is not affected by the use of the MLEM method, we conclude that this reconstruction method it well suited for reconstruction of XRD tomographic datasets of this type. For obtaining the highest crystallographic contrast, however, the FBP (or equivalent) methods are more suitable.