edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 4 30 Fig. 1 ADF STEM, STEM EBIC, and a diagram of a Si/SiO2/Pt capacitor. EBIC is measured independently from the top electrode (TE) and from the Si substrate. carriers can still be separated to produce an EBIC, rather than the microscope resolution. In photovoltaic devices, for example, the diffusion length can exceed 100 nm,[2] resulting in STEM EBIC with similar resolution to SEM EBIC (albeit alongside much higher resolution STEM images). The limited return on resolution, at the cost of much more difficult sample preparation, is perhaps the main reason EBIC has been rarely implemented in STEM. But, as we will demonstrate here, higher resolution imaging with this EBIC mode is possible in STEM for devices in which carrier mobility is limited (often by design) in one or more directions. The higher electron energies and thinner samples in TEM minimize multiple-scattering processes such as backscattered electrons (BSEs), which also generate a large number of secondary electrons (SEs), and beam absorption. EBAC is therefore not typically observed in STEM, but resistance contrast imaging can be performed in the TEM via a more recently demonstrated EBIC mode called SE emission EBIC (SEEBIC).[8,9] SEEBIC, which has been demonstrated at atomic resolution,[10,11] measures the positive current generated by emission of secondary electrons and does not require the presence of an electric field. With a single connection to the sample, SEEBIC provides SE contrast in which every SE that leaves the sample is measured (rather than just the SEs that reach a remote detector); SEEBIC provides SE detection with efficiency approaching unity. Such quantitative SE measurements enable mapping signals that influence SE emission, including work function[8] and temperature,[12] and the geometry-agnostic SE detection shows promise for resolving three-dimensional structure.[13] With multiple connections to a sample, the path by which SEEBIC current reaches ground (perhaps via one or multiple EBIC amplifiers) will depend on the resistance to each connection, generating resistance contrast.[9,14] Effectively, the beam (via SE ejection) deposits positive test charges at each pixel, and SEEBIC provides a map of which charges reach a given EBIC amplifier.[8,9] This is currently the only method for obtaining conductivity contrast at high resolution in the TEM. This has been used to observe nanoscale conductivity changes in biased devices purpose-built for functioning in the TEM.[14] But given the small magnitude of the signal (typically ~1000 times smaller than EHP EBIC), SEEBIC is also much more sensitive to current noise and leakage paths, making similar experiments on cross- section samples all the more difficult. While STEM EBIC is generally capable of higher resolution imaging than SEM EBIC, this comes at the cost of a much more limited field of view, especially given that components must be imaged in cross section. For failure analysis, SEM and STEM EBIC may therefore be particularly powerful when used in concert: SEM EBIC can be used to locate a region of interest for cross-section extraction and further high-resolution analysis via STEM EBIC, alongside other TEM-based techniques, to more precisely determine the nature of a defect. STRATEGIES FOR LOW-LEAKAGE CROSS-SECTION SAMPLES The requirements and challenges involved in preparing samples for STEM EBIC or in situ biasing have been discussed elsewhere in detail.[2,15] To briefly summarize, a TEM sample must be thin, requiring extraction of a cross-sectional sample for studying microelectronic components. The standard tool for preparing such samples, the focused ion beam (FIB), can cause damage and contamination that degrade the electronic structure (to which EBIC is highly sensitive) of a sample. The main sources of contamination are from the beam deposition of metal contacts, which can leave conducting residue many micrometers from their intended pattern, and from implantation of Ga from the ion beam. A recently reported technique addresses the former issue by use of weld-free contacting of lamellae to the electrodes of the supporting substrate.[15,16] With this method, cross sections are fully thinned and cleaned while mounted to the extraction needle, and then are mechanically pressed against the substrate electrodes while the needle weld is severed. Van der Waals forces hold the lamella in contact with the electrodes without ELECTRONICALLY VIABLE TEM SAMPLES WITH PFIB AND STEM EBIC (continued from page 27)
RkJQdWJsaXNoZXIy MTYyMzk3NQ==