edfas.org 5 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 25 NO . 1 provide insight into device physics but may not represent the truebehavior of anoff-the-shelf component, and these devices are therefore more appropriate to basic research purposes. This articlewill focus onFIBsamplepreparation, as it is a standardized process, to the point of automation, and therefore presents themost direct route to routine in situ TEM characterization within existing workflows. FIB-prepared TEM samples are typically mounted to TEM lift-out grids, or semicircular conducting discs with protruding posts on which lamellae can be mounted. Samples mounted on these grids may be used for biasing experiments when loaded into a biasing sample holder equipped with a moveable probe. With the grid serving as ground, bias can be applied via the probe to several thinned devices on a single grid. However, making this mechanical connection in situ produces irregular strain and electric fields that may result in unrealistic device function, and the single probe only allows for a single measurement or applied stimulus at a time. Amuchmore common style of biasing holder includes a chip carrier on which Si-based substrates can be loaded and through which multiple electrical connections canbemade simultaneously. TheseMEMS-based biasing holders are compatible with an array of chips that can be equipped with heaters, platforms for mounting cross sections, as well as more complicated micro-fabricated devices. There are a variety of commercially available and custom-made substrates formounting crosssection samples for in situ biasing. The substrate shown inFig. 1 consists of aSi chiponwhicha thin membrane has been etched via anisotropic KOH etching.[1,2] Large pads are patterned tomatch the mounting/connectionmechanismof the holder, and thinner leads protruding from these pads extenddown to the thinmembrane at the edge of the chip. Here a trench has been plasma-etched in the membrane over which a cross section can be suspended. The standard approach for FIB-preparation of bias-enabled TEM samples, shown in Fig. 2, begins with the milling and extraction of a (relatively thick) cross section from the macroscopic component of interest. The section is then mounted to the Si-based biasing chip with electrical leads deposited, via beam-bombardment of metal-organic gas injected near the sample, to connect to the patterned chip electrodes. Portions of the device must then be milled to electrically isolate device layers. For example, the photodiode in Fig. 2 is milled such that one electrode connects to the substrate and the other to the device’s top electrode. Without this rudimentary circuit edit, the electrode and substrate would be shorted, and electronhole pairs produced in the active device region would recombine. For samples where a bias is to be applied, milling can maintain the fidelity of nominally insulated device regions. Following the selective milling step, the suspended region of the device must be thinned to electron transparency, as in standard FIB sample production. The final step is a low-energy, low-current cleaning of the sample surface. This cleaning reduces surface damage, implantation, and heating. The final surface cleaning is perhaps the most critical step in production of viable TEM biasing samples that behave similar to their parent devices. Optimizing this cleaning procedure is therefore the chief area of improvement necessary for routine implementation of in situ biasing techniques in the TEM. Milling and metal deposition can spray material several micrometers from Fig. 2 FIB preparation of a Si photodiode. (a) Photograph of the photodiode before FIB milling. (b) SEM image of a (relatively thick) cross section FIB-milled at the edge of photodiode’s top electrode, in the region indicated by the small red box on the left side of (a). (c) Cross sectionmounted to the Si-based chip shown in Fig. 1. Metal electrodes are written to either side of the lamella via metal-organic FIB deposition. Sections of the lamella are milled away to electrically isolate regions of the device: the left electrode is connected to the substrate and the right electrode to the top electrode. A section in themiddle of the device is thinned to ~100 nm. (d) Image of the final device, with the region imaged in Fig. 3 indicated by the yellow box.
RkJQdWJsaXNoZXIy MTMyMzg5NA==