edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 27 NO. 4 18 nanoprobing platform, employing its semi-blind positioning capability to minimize the number of SEM exposures during probe alignment, sample navigation, and probe contacting. In this mode, only six SEM images at the point of interest were acquired during the probe landing procedure, each confined to a 1 μm × 1 μm field of view, with an imaging time of 120 ms per frame. To minimize irradiation effects, the SEM was operated at a low beam current of 20 pA and an accelerating voltage of 100 eV. The resulting electrical characteristics, specifically the threshold voltage, were then directly compared to the AFP baseline. Despite slight device-to-device variations (± 5%), no significant deviation in threshold voltage was detected, indicating that the minimal SEM exposure associated with semi-blind positioning does not induce observable degradation under these operating conditions. To probe the onset and dynamics of degradation, a follow-up experiment involved continuous SEM exposure on the same 1 μm² area under identical beam current conditions but varying beam energies. The evolution of the threshold voltage over time was monitored, normalized to the value obtained after semi-blind contacting (see Fig. 6). During this time, the SMARPROBE was in constant electrical contact with the pFET, and the contact resistance to the probes remained constant throughout. Although not shown in the figure, immediate and severe threshold shifts were observed at 500 eV, suggesting aggressive charge injection and/or defect activation in the gate oxide. At lower energies—200 eV and 100 eV—the degradation kinetics were substantially slower; however, observable threshold voltage shifts still emerged within the first 60 seconds of continuous exposure. The effect was notably more pronounced at 200 eV, where degradation occurred at approximately eight times the rate observed at 100 eV. Interestingly, the degradation profiles exhibited a near-linear evolution with time at both beam energies, indicating that the underlying physical mechanism— presumed to be charge trapping in the BOX or gate oxide layers—does not saturate within the timescale of observation. Moreover, the threshold shifts were not reversible even after several days. While the slower degradation observed at 100 eV is beneficial, it comes at a significant trade-off: Image quality at this energy level is substantially compromised due to reduced signal-to-noise ratio, image distortion, and strong shadowing from the probes. Consequently, the lower imaging fidelity at 100 eV necessitates longer acquisition times or repeated imaging to achieve sufficient contrast for probe placement. If, for instance, probing at 100 eV takes eight times longer than at 200 eV, the cumulative beam dose and thus the degradation risk may equal—or even exceed—that at 200 eV. This leads to an important insight: In practical probing workflows, minimizing the number of SEM images can be more effective for reducing electron beam-induced damage than merely lowering the beam energy. The semi-blind positioning approach reduces the required number of SEM images by approximately two orders of magnitude compared to manual probe landing. The total electron dose at 200 eV in semi-blind mode is estimated to be at least 10 times lower than during manual probing at 100 eV. This estimation aligns with the empirical data: When using semi-blind positioning at 200 eV, no degradation was observed, and the electrical parameters remained consistent with the AFP benchmark. This is especially important for SEM models that perform poorly in the low eV range. Furthermore, in many applications, it is necessary to probe all six transistors of a standard 6T SRAM cell, which significantly increases cumulative beam exposure during manual probing. If conventional, visually guided methods are used, total imaging times frequently exceed several minutes, Fig. 6 FD-SOI pFET degradation during continuous SEM imaging at 100 eV and 200 eV. In comparison, the semi-blind probing using the SMARPROBE allows probing of a complete 6T SRAM cell in less than two seconds of accumulated imaging time.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==