November_EDFA_Digital

edfas.org 41 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 19 NO. 4 defines the overall region of interest for the montage as well as the size of each field of view. The degree of overlap is another variable as well as the dwell time. Altogether, a user has control over the resolution, signal-to-noise, and hence time required for each montage. Optimizing imaging conditions for speed quality and the application of multiresolution imaging strategies is the subject of another paper. [10] The ion beam conditions may vary between 15 and 30 kV, while the beam current density may range from 0.5 to 2.0 pA/µm 2 . At lower ion beam energy, there is a dual benefit. Moderate ion beam energy yields an optimal material-removal rate in conjunctionwith the gas chemistry. In addition, lower ion beam energy permits a larger field of view and therefore can be applied to yield larger delayering areas. In the case of 15 kV xenon ions, areas as large as 800 µm 2 can be accessed. The distribu- tion of the gas chemistry is also a critical parameter in governing the uniformity that can be achieved over large delayering areas. Multiple gas-injection nozzles and/or gas-concentration schemes may be applied to optimize the gas distribution for the purpose of maximizing large- area delayering uniformity. The entire process of chemical-assisted ion beam delayering coupled to montage imaging in a plasma FIB-SEM lends itself quite well to integrated automated processing. Aspects of automation, as well as opportu- nities in computationally guided microscopy in plasma FIB-SEM delayering, is the subject of a related paper. [10] Figure 5 shows a pair of images acquired at two differ- ent accelerating voltages to highlight different informa- tion. The 5 kV image (Fig. 5a) is a montage of 49 images acquired using a BSE detector, while Fig. 5(b) is a 30 kV BSE imagemontage. The low-voltage secondary electron image is more surface-sensitive, while the 30 kV image has a larger interaction volume, because the BSE signal emanates from a greater depth in the sample, allowing the structure to be discerned past the metal 1 layer. Each image in the montages contains 4096 × 4096 pixels and required approximately 50 s/image to acquire under the imaging conditions used. The imaging conditions selected in this work do not represent an optimal imaging condi- tion to minimize acquisition time. Optimization of both imaging conditions and strategy, including the applica- tion of multiresolution imaging, is the subject of a related paper. [10] Here, the purpose is to validate and demonstrate the overall process and to evaluate a specific delayering gas chemistry applied to this particular device. Delayering exposure times variedbetween 3 and 7min/cycle between imaging, depending on the area exposed. A detailed image pair is shown in Fig. 6, taken from the region highlighted by the yellowbox in Fig. 5(b). Metal vias are distinctly bright in the low-kV image (Fig. 6a), while the underlying structure is visible in the 30 kV image (Fig. 6b). Stitching errors in the montage created by the native instrument software are evident. Improving the correlated stitching functions as well as segmentation and feature extraction is the subject of future work. [10] Figure 7 shows an image pair from a region of interest within the delayering sequence following removal of the contact layer. The gate structures are highlighted at 5 kV, and the underlying M1/M2 structure is observed at 30 kV. The M2 layer is relatively “fuzzy” due to the electron scat- tering at greater depth. As the lower layers are removed from the backside, the near-surface structures seen at 30 kV become progressively sharper. A final image pair (a) (b) Fig. 6 ImagepairtakenfromtheregionofinterestwithinFig. 5(b). Metal vias and doping contrast are emphasized in the 5 kV image (a), while the underlying structure deeper into the device is seen in the 30 kV montage section (b). The fieldof view is 127 µm inboth images.