May_EDFA_Digital

edfas.org 25 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 2 This shows that there is some small contact resistance causing the Idsat to decrease by ~1 order of magnitude for this transistor. When both source and drain contacts lack any SCM signal as in PG cell 2, the Idsat is reduced by ~3 orders of magnitude. In this case, the fact that the transistor is still operational (although, not ideal) when both contacts do not showany SCMsignal implies there is a limit to the sen- sitivity of the SCM technique to contact resistance. When the SCM signal is simplymissing, one cannot conclusively determine if the contact is resistiveor openwithout further analysis by nanoprobing or TEM. In the saturation regime, the Idsat difference is ~1 order of magnitude if one contact has reduced SCM contrast vs. two orders of magnitude if both contacts have no SCM contrast. DISCUSSION The analysis shows good correlation between SCM signal and the device’s electrical behavior. CAFM is often em- ployed to acquire some electrical infor- mation about failing devices, which is less time consuming than performing nanoprobing. Because CAFM does not work on SOI technologies, having SCM as an alternative approach is beneficial for FA flow. Further physical failureanaly- sis on the devices discussed identified a small amount of resistive material was present between the source/drain con- tacts and the fin, which explains the SCM and nanoprobing data. High resolution TEMandelemental analysiswas needed, as the SEM was unable to resolve this material. CONCLUSIONS This case study reviewed 14 nm SOI finFETs in which SCM and nanoprobing methods were utilized to determine the electrical nature of cluster-type fails in the SRAM after standard FIB cross-section imaging methods failed to show any visible defect. The results obtained by SCMand nanoprobingwere compared to show that SCM is a viable technique for screening for resistive fails Fig. 6 Nanoprobing data, saturation regime, for failing PGs in cell 2 and cell 3 (blue and red, respectively). The Idsat is lower in cell 2compared to cell 3. Fig. 7 (a) SCM image (3V AC, 0V DC) and (b) Layout. Failing bits in the cluster are shaded in red. Fig. 8 (a) Nanoprobing data (linear regime) and legend for various nFETs with various contact resistance scenarios. (b) and (c) SCMdata and corresponding SCM line scans for the various nFETs discussed. Source, drain and gate contacts are labelled and correlated to the nanoprobing data in the legend. (a) (b) (a) (b) (c) (continued on page 28)