November_EDFA_Digital

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 4 6 electrode or the other. The SEEBIC technique provides insight into conductivity (signal intensity) and connectivity (location of bright/dark transition) in this simple sample, even on the nominally insulating membrane. Standard STEM imaging, on the other hand, reveals no details about electronic properties, it can only discern between regions of bare (presumably insulating) membrane and metal (presumably conducting) electrodes. The SEEBIC method reveals even more interesting electronic features in regions with non-trivial resistance distributions. The images in Fig. 2 show a cross section extracted from a multilayer ceramic capacitor (MLCC), a component ubiquitous in many electronic systems, con- sisting of interdigitated Ni electrodes separated by poly- crystalline BaTiO 3 . A FIB was used to extract a thin cross section froma commercially availableMLCC and tomount it toaSi-based liftout chippatternedwithelectrodes. [7] The chip’s electrodes are connected to the cross section’s Ni electrodes on either side of the polycrystalline BaTiO 3 film and electrical connection to the chip is made via a TEM feedthrough holder. The Ni electrodes are visible just on the left and right of the frame in the Fig. 2 ADF image. This standard STEM image shows physical contrast, related to density and crystal structure, and distinguishes individual crystal grains of both materials. In the Fig. 2 EBIC image, much like the middle Fig. 1 EBIC image, current ismeasured on the left electrode, the right is held at ground, and the electrodes are separated by an insulator. The EBIC is bright on the EBIC electrode (left) anddark on the opposing electrode. However, unlike in Fig. 1, the EBIC in Fig. 2 decreases stepwise, moving from left to right, with sharp steps (often within a single 14 nmpixel) at some of the BaTiO 3 grain boundaries. In the Fig. 1 sample, the insulatingmembrane charges positively at the onset of SE emission, limiting further emission, however the FIB-induced damage and contamination on the surface of the Fig. 2 sample helps dissipate charge and allows for stronger SEEBIC signal. Amore in-depth discus- sion of surface conductivity and different EBIC modes is provided by Hubbard et al. [8] The Fig. 2 device can be modeled as a current divider, with two connections to ground (either directly or through the EBIC amplifier) on either side of the resistive BaTiO 3 and the electron beamas a current source. The Fig. 2 EBIC image can therefore be recast as a resistance contrast image (RCI), [3,8] indicating the resistance between each pixel and the right Ni electrode. Near to the right electrode the dark BaTiO 3 contrast is indicative of the low resistance to the adjacent Ni. Near the left electrode the bright con- trast represents the full resistance between electrodes. Steps in contrast on either side of some grain boundaries indicate a large resistance at those boundaries. Inverting the contrast would produce a similar map of resistance relative to the left electrode. Reliability models for these MLCCs determine the mean time to failure (MTTF) for a given voltage to be a function of the number of grains per layer, [9] assuming that eachgrain contributes equally to the dielectric resistance between electrodes in a given region. The Fig. 2 EBIC/RCI image instead shows that resistance in this region is dominated by just a fewgrain boundaries. While no potential was applied between the Ni elec- trodes while EBIC was acquired, the Fig. 2 EBIC image Fig. 2 ADF, EBIC, and EBIC-derived electric field map for a Ni/BaTiO 3 /Ni MLCC acquired with 300 keV beam electrons. The EBIC is measured on the left Ni electrode (just barely in frame) and the right Ni electrode is grounded. The right map shows the negative gradient (vector difference between nearest-neighbor pixels in x and y) of the EBIC map, after binning by 2 in x and y. The arrow indicates field direction and intensity of the green color indicates magnitude. This map represents the electric field that would be present if a potential were applied to the left electrode, with the right grounded (despite no potential being applied during image acquisition). The pixel size in the ADF and EBIC images is 14 nm. Figure adapted from Hubbard, et al. [8]