edfas.org 5 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 1 generated at each point of the electron beam raster.[8] Multibeam diffraction was reported to improve the orienta- tion mapping of a small-grained sample compared to a map generated using a single beam measurement.[8] Template matching is frequently used to index diffraction patterns, where the patterns for all possible orientations of each phase are pre-calculated and compared to the experimental data by cross-correlation. This means the angular resolution depends on the density of templates simulated and thereby creates a tradeoff in angular resolution versus the time to generate the initial library plus comparing it to experimental results.[9,10] Angular resolutions as small as ≈ 0.3° have been reported using template matching, but is commonly on the order of 1 degree.[11] Improvements to angular resolution have been realized using artificial neural networks (ANNs). ANNs were trained using diffraction patterns simulated using the Bloch wave algorithm, meaning it incorporated dynamical scattering effects.[12] The angular resolution of the ANN-based analysis was reportedly 0.009°, over an order of magnitude improvement compared to previous reports using tem- plate matching. Henry et al.[13] used a combination of precession 4D-STEM and elemental mapping via energy dispersive x-ray spectroscopy (EDX) to determine the RESET failure mechanism in Ge-Sb-Te, a non-volatile phase-change memory (Fig. 1, Panel I). By using the 4D-STEM phase mapping results to segment the EDX maps, they were able to demonstrate that failure occurred due to depletion of the Ge content in the amorphous volume of the cell while the composition of the crystalline regions was unchanged. Another semiconductor application of 4D-STEM was reported by daSilva et al.[14] who investigated the formation of superlattices in PbS nanocrystal assemblies. By precisely measuring the position and orientation of nanocrystals in the transient region of these assemblies, they were able to determine that the transformation and subsequent superlattice orientation is dominated by the correlated alignment of the individual nanocrystals. Another interesting application reported by Londoño-Calderon et al.[15] quantified the orientation variation along the length of semiconducting Te nanowires and therefore was able to measure the overall twist rate and chirality (Fig. 1, Panel II). Finally, the speed and sensitivity of modern detectors has enabled phase and orientation mapping of semiconducting materials prone to electron beam damage. Panova et al.[16,17] used 4D-STEM to characterize conjugated semiconducting polymers and copolymers. By characterizing the response of the materials to the electron dose and staying below the damage threshold for a given material, they were able to map the crystal location and orientation in these technologically important soft materials. Finally, this approach was extended by Wu et al.[18] who employed a modified electron optical approach to improve the signal intensity and lower the required electron dose even further. In doing so, they were able to monitor the evolution of crystal size, distribution, and orientation as a function of heat treatment by serial collection of 4D datasets while remaining below the dose threshold. Beyond using zero order Laue zone (ZOLZ) Bragg reflections, higher order Laue zone (HOLZ) features also provide unique measurement opportunities. Where the ZOLZ includes the origin of the reciprocal lattice and the transmitted beam, HOLZ reflections originate from diffraction vectors that are not perpendicular to the electron beam direction. As a result, the HOLZ reflection contains information about the crystal parallel to the beam direction and, as such, can provide information about the threedimensional ordering of the crystal. Analysis of HOLZ diffraction features were used to map a displacive transformation in a perovskite heterostructure of La0.7Sr0.3MnO3/ LaFeO3/SrTiO3, where cell tilting and La A-site modulation in the LaFeO3 is suppressed at both interfaces. [19] DISORDERED MATERIALS So far, this article has discussed crystalline materials, that is, materials with long range order (translational symmetry) and by doing so, ignored an extensive category of technologically important materials, i.e., disordered or amorphous materials. The lack of long-range order poses unique challenges to characterization. For example, highresolution TEM (HRTEM) images of amorphous materials can depict fringes which may be tempting to interpret as crystal lattice planes and the presence of local ordering. However, these fringes can be an artifact of the band filtering action of the image forming (objective) lens.[20] Additionally, a small local region of order may not be detected above the background when embedded in a thicker disordered matrix. In response to such challenges, several techniques have been developed providing insight into the structure of amorphous materials. While many of these techniques predated modern day 4D-STEM, the large, low-noise datasets provided by modern pixelated STEM detectors are advantageous because many of these techniques are sensitive to noise and also rely on statistical methods which benefit from greater sampling. The pair distribution function (PDF) and its variants (e.g., radial distribution function), describe the probability distribution of interatomic distances.[21] A benefit of using STEM to generate PDFs is the small interaction volume;
RkJQdWJsaXNoZXIy MTYyMzk3NQ==