November_EDFA_Digital

edfas.org 5 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 4 electron-transparent (typically ≤ 200 nm-thick) samples, which invariably means microelectronic components must be subjected to some degree of deprocessing prior to imaging. Consequently, in situ TEM observation of an off-the-shelf component’s normal function is rarely feasible. In some cases, analogue devices may be fabri- cated deliberately for TEM imaging without sacrificing function. [6] While these devices may not be suitable for real-world applications, device physics observed while imaging them can be extended to components of the same topology. The much more common approach to TEMsamplepreparation involves cross-sectioning, usually with a focused ion beam (FIB), and studying a subset of the component. The focused beam of ions, typically Ga + , can mill away material for excavation and thinning, and also deposit metal via irradiation of organometallic gas for mounting and making electrical contact. Focused ion beam processing results in damage and contamination that can impair, if not destroy, the electronic structure responsible for a component’s function. However, even if function is compromised during preparation, STEM EBIC characterization of an isolated active device region can still reveal details about electronic structure that cannot be revealed through standard electrical testing, which averages over all electrically active regions. MAPPING ELECTRONIC TRANSPORT WITH SEEBIC In the STEM EBIC images shown here, current is measured with a transimpedance (current input, voltage output) amplifier. The voltage signals from the amplifier andother STEMdetectors are synchronized, pixel-by-pixel, with beam position to produce STEM EBIC and STEM images simultaneously. As a simple example of STEM EBIC on device fabricated specifically for TEM imaging, Fig. 1 shows opposing 25 nm-thick Pt electrodes (with a 5 nmTi adhesion layer) patterned on a 20 nm-thick silicon nitridemembrane in a Si chip. [4] The chip ismounted in an electrical feedthrough holder that allows for connection to the sample while inside the TEM. In the Fig. 1 annular dark-field (ADF) STEM image, the two electrodes have very similar (bright) contrast, indicating that beam elec- trons scatter more effectively from the electrodes than the membrane. The signal from two transimpedance amplifiers, each independently connected to one of the electrodes, is digitized alongside the ADF signal, simultaneously gener- ating all three Fig. 1 images. There are no inherent electric fields in this device and beam absorption is negligible for high beam energies and thin samples, [1] so here SEEBIC is the dominant EBIC mode. In each image the electrode to which the EBIC amplifier is connected appears bright, as the amplifier is collecting positive (hole) current pro- duced by emission of secondary electrons (SEs) from that electrode. The opposing electrode in each image appears slightly dark (negative, electron current) as some SEs ejected from this electrode are recaptured by the EBIC electrode. [4] Themembrane between the electrodes is gray (near zero current), as very fewSEs are emitted from insu- lators due to charging. However, a very subtle transition from bright to dark is visible on the membrane along the middle of each EBIC image, most obviously in the region of closest electrode approach. This transition indicates the “conductance watershed” on the membrane, or the boundary between regions of the insulator on which the few beam-generated holes are better connected to one Fig. 1 Annular dark field (ADF) and EBIC images, acquired with 80 keV beam electrons, of two Pt electrodes patterned on a silicon nitride membrane. For the EBIC images, current in a given image is measured on the electrode indicated by the opaque current meter symbol: the middle image shows EBIC from the left electrode, as indicated by the blue current meter symbol, and the right image shows EBIC fromthe right electrode (red currentmeter). Figure adapted fromHubbard, et al. [4]