May_EDFA_Digital

edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 2 22 This process has better control for layer-by-layer material removal compared to conventional mechani- cal polishing. Figure 7 shows an SCM image of 14 nm SRAM cells in a cross section that is perpendicular to the fins. The SCM image shows doping contrast, confirming good electrical contact and spatially resolved individual fins within the device. The NFETS, N0 and N1, are two-fin devices while PFET, P0, is a single-fin device (Fig. 7). The strength of the doping signal, however, is limited by the formation of a surface amorphous layer. The formation of an amorphous layer by Ar ion milling is dependent upon beam voltage, time, and angle. The results show that the optimized Ar ion milling parameters (beam power, acceleration voltage, time, and angle) yields a sample surface that meets the requirements necessary for acquiring a strong dopant signal using SCM. Figure 8 shows the cross-section- al topography and dopant/carrier profile of a 45 nm NFET device. The topography image shows the poly- silicon, barrier nitride, and spacer layers. The sample was prepared by the inverted TEM sample-preparation method with a final surface clean- ing using a low-voltage, non-gallium-based beam. SCM analysis did not show any dopant signal after the FIB preparation. The absence of an SCM signal is due to resi- dual surface damage after the final FIB clean-up step. After low-voltage, low-angle, Ar ionmilling, a dopant signal was obtained. The ionmillingwas done by using beamvoltage of 500 eV at 10 degrees (at a rate of 4 nm/min for 3 min). Thedopant signal clearly shows the sourceanddrain (S/D), source-drain extension (SDE), and channel areas. CONCLUSION This work shows that a significant improvement in the SCM doping signal was achieved with samples pre- pared by the inverted TEM sample-preparation method followed by non-gallium-based ion milling. Additionally, our results show that low-voltage (<500 eV), shallow-angle (~10 degrees), nano-milling minimizes surface amorph- ization, resulting in a strong SCM signal representative of local activedopant concentration. The samplepreparedby mechanical polishing and referencemarkingwith FIB-SEM shows proof of concept that SCM and SCS have the capa- bility to resolve dopant-related, root-cause anomalies on location-specific 14 nm FinFET devices. ACKNOWLEDGMENTS The authors thank the GlobalFoundries FAB9 TEM lab team, Erik McCullen, Thom Hartswick, and Gregory Tidman for their support in FIB sample preparation and process optimization. Inaddition, theauthorswould like to give credit to Jochonia Nxumalo (former GlobalFoundries employee) for the design and fabrication of the sample polishing fixture. REFERENCES 1. P. Kaszuba, L. Moszkowicz, and R. Wells: “Scanning Capacitance Microscopy and Spectroscopy for Root Cause Analysis on Location Specific Individual FinFET Devices,” Proc. Int. Symp. Test. Fail. Anal. (ISTFA), 2019. 2. J. Mody, et al.: “3D-carrier Profiling in FinFETs using Scanning Spreading Resistance Microscopy,” 2011, p. 6–1, DOI: 10.1109/ IEDM.2011.6131498. 3. Prime Nano: Imaging Finfet using Scan wave pro 4. D. Burnett, et al.: “Modeling the Impact of the Vertical Doping Profile on FinFET SRAM VT Mismatch,” Proc. IEEE SOI-3D-Subthreshold Microelectronics Technology Unified Conference, 2016, p. 1-3. Fig. 7 Topography and doping profile of SRAMarray 14 nmFinFETs prepared by the inverted TEM sample-preparation technique. Fig. 8 Cross-sectional dopant profile image of a 45nm NFET before and after 3 minutes of Ar ion milling with 500 eV at 10 degrees.