STEM optical sectioning for imaging screw dislocation core structures

Abstract

The introduction of spherical‐aberration correctors in STEM has allowed an improvement in spatial resolution up to the sub‐angstrom scale also accompanied by a reduction of the depth of focus (due to the increase in probe convergence angles), which in a modern instrument is just a few nanometers, thus often less than the sample thickness. This can be exploited to extract information along the beam direction by focusing the electron probe at specific depths within the sample. This technique has already been used to observe the depth‐dependence of the strain field due to the Eshelby twist associated with dislocations containing a screw component in thin STEM samples. The measurement of the magnitude of the displacement confirmed the screw Burgers vector for dislocations in GaN [1] and allowed the identification of a new dissociation reaction associated with mixed [c+a] dislocations [2]. The optical sectioning approach has also been applied to the direct observation of the c‐component of the dissociation reaction of a mixed [c+a] dislocation in GaN by imaging a dislocation lying transverse to the electron beam [3]. Here we show how optical sectioning in high‐angle annular dark‐field (HAADF) STEM imaging conditions can be used to image the core structure of screw dislocations at atomic resolution. In particular, we evaluate using simulations whether the edge and screw displacements associated with the delocalization of ½[111] screw dislocations in body‐centered cubic (BCC) metals [4] can be detected. In Figure 1 we show that the helicoidal displacements around a screw dislocation can be imaged with the dislocation lying transverse to the electron beam by optically sectioning the plane containing the dislocation. In order to reveal how the edge components of the dislocation contribute to the observed contrast we have created two atomistic models, one using the anisotropic linear‐elastic displacements around the dislocation, which is not capable of modelling the core delocalisation, and the other using the core structure relaxed using the Bond Order Potential for W which does predict a delocalised core. Figures 2 a) and b) show the respective HAADF simulated images. Figure 2 c) is the RGB image made from the (101) component of the Fourier Transform (FT) (shown in Figure 2 d)) of both images. It is possible to observe that the shifts in this Fourier component occur along two distinct lines lying parallel to [111]. The superposition of both filtered images shows that there is a discrepancy on both sides of the core between both models. It is therefore apparent that the delocalisation of the core can in principle be detected using electron‐optical sectioning [5,6].