With relevant feature sizes approaching atomic length scales, the transmission electron microscope (TEM) is the most practical tool for inspection of modern silicon transistors. TEM is unmatched in its ability to resolve physical structure and composition at high resolution, but it provides virtually no information about a sample’s electronic properties. TEM may therefore be blind to critical device features for which a tiny physical change has a major impact on local electronic properties, such as the selective introduction of a relatively low concentration of dopant atoms. A direct measure of local electronic properties, as a complement to the physical information provided by standard TEM-based approaches, could potentially accelerate process optimization efforts for current and next-generation components. Scanning TEM electron beam-induced current (STEM EBIC) imaging is a promising candidate for introducing high-resolution electronic metrology within the framework of existing TEM-based workflows. EBIC has been a standard technique for imaging electric field-based contrast from semiconductors in the scanning electron microscope (SEM) since the 1960’s [1], with occasional implementation in STEM since the 1970’s [2]. Recent advances in STEM EBIC have enabled entirely new types of EBIC contrast [3], including conductivity mapping [4], as well as atomic-resolution EBIC [5]. EBIC images are acquired alongside standard STEM detector signals, pixel-by-pixel, and with electronic contrast visible (and often obvious) in the raw images. Despite the effectiveness of STEM EBIC as an electronic complement to TEM’s physical contrast, sample preparation challenges have been a bottleneck for high-throughput STEM EBIC characterization. Locating and making electrical connection to relevant devices in a thin cross section can be prohibitive, especially for advanced-node MOSFETs. Here we present STEM EBIC images of several different advanced-node (5-14 nm) finFET samples, and discuss challenges and approaches to effective sample preparation. In the four examples in Fig. 1, samples were extracted from three different source components, in various states of de-processing, and using different ion beam species and cleaning procedures [6,7]. All cross sections are extracted and mounted to STEM EBIC-compatible sample support chips using mostly standard focused ion beam (FIB)-based lamella preparation procedures. For each, electrical contact is made to the device by welding to the support chip’s two electrodes with metal-organic gas deposition in the electron and ion beams. The source components in Figs. 1A-C were off-the-shelf components, with connection made to metal interconnects and to the substrate, with the latter connection intended to provide a path for EBIC currents without the need to directly connect to specific transistor terminals. The Fig. 1D source was a test structure, fabricated at a major foundry on a multi-project wafer, that was specifically designed to facilitate STEM EBIC sample preparation and acquisition. For this test structure, all source contacts are routed to a large electrode on one side of the lamella, and all drain contacts to a large electrode on the other, and these lamella electrodes are welded to the chip electrodes to make direct contact to the appropriate device terminals. In Fig. 1a, the EBIC contrast is caused by secondary electron emission EBIC (SEEBIC) conductivity contrast [3,4] and shows a clear distinction between electrically active fins (bright) and “dummy” fins (dark) [5]. For the EBIC images in Figs. 1b-d, the contrast is due to the separation of electron-hole pairs in electric fields at dopant interfaces, with the sign of the current depending on the direction of the electric field relative to the point of contact. This mechanism is the standard mode of contrast in SEM EBIC, but the clarity and spatial resolution is substantially improved in these STEM EBIC images compared to SEM EBIC results on similar structures [8,9]. The difference in dominant contrast modes in the Fig. 1A and Fig. 1B-D EBIC images is likely due to the continuity of the conducting path from the metal cap to the device layer, and to differences in surface leakage, but further study is necessary to more completely understand the STEM EBIC sample preparation parameter space. In each case presented here, dopant-related electronic contrast is obvious in the raw images, but entirely absent from the ADF STEM images acquired simultaneously. These results exemplify the significant progress made in STEM EBIC sample preparation and acquisition over just the past few years, and the feasibility of higher-throughput STEM EBIC metrology of advanced-node MOSFETs in the near future. STEM ADF and EBIC images showing dopant-related contrast in 5-12 nm-node finFETs. Samples in (A) and (B) were extracted from a 7 nm-node commercial processor that was decapped but otherwise intact, with electrical connection to the device layer via the substrate (where the displayed EBIC signal was measured) and through the device’s interconnects. The device in (A) was prepared only with Ga+ FIB (Thermo Fisher Scientific Scios FIB-SEM) [5]. The sample in (B) was prepared in the same FIB as (A), but in an inverted orientation and with a post-FIB low-kV Ar+ milling (Fischione Instruments PicoMill®) [6]. (C) Sample is extracted from a 5 nm-node commercial processor that had first been mechanically polished down to the M0 metal layer [7]. Polishing allows metal-side contact to be made to virtually all fins via beam-deposition during subsequent Ga+ FIB extraction (ZEISS Crossbeam 550L FIB-SEM), with the other contact made to the substrate. (D) A sample extracted from a STEM EBIC-optimized test structure fabricated in a commercial foundry on a multi-project wafer. This sample was prepared using a Xe+ PFIB (Thermo Fisher Helios G4 PFIB UXe DualBeam FIB/SEM). The test structure routes all source terminals to one side of the lamella, and drain terminals to the other, where they connect to larger electrodes in the lamella which can be welded directly chip electrodes. All STEM EBIC was measured using a two-terminal STEM EBIC system (NanoElectronic Imaging), but here only one EBIC channel is shown as the other channel has either similar (A) or identical (B-D) contrast but with the opposite sign.
No takes yet. Share an insight, caveat, or question.
Hubbard et al. (2026) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: