February_EDFA_Digital

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 21 NO. 1 8 this analysis are shown in Fig. 5a. In this result, RCD places several big cells (e.g., full-adder and flip-flops) along with Open-Metal1 in the defect Pareto chart as the top root causes.While the cell root causes indicate there is a cell-related issue in the population, the information is not very precise. Next, the same failing die are diagnosed with cell-aware diagnosis and analyzed. Results are shown in Fig. 5b. Now, the defect Pareto chart includes the exact target root cause, i.e., polysilicon shorts. In addition, the fake top root cause of Open-Metal1 has been eliminated. This highlights the tremendous value of cell-aware diagnosis with RCD in resolving physical root causes in the front-end layers inside the cells. With such precise root cause distribution, die selection for FA can be more precisely targeted, resulting in a significant reduction in overall yield analysis cycle time and cost. Similar injection experiments were conducted for five industrial designs and for all front-end cell internal root causes in each design, and cell-aware diagnosis plus RCD achieves an average accuracy of 70%. After verifying the technology through controlled injec- tion experiments, it was then tested on real silicon data. To start, this testing was performed on two existing sets of material for which FA had previously been completed to determine the top defect mechanisms. The objective was achieved, which was to confirm that cell-aware diag- nosis with RCD builds a defect Pareto chart with similar conclusions. In each case, cell-aware diagnosis with RCD matched with FA. Cell-aware diagnosiswith RCDwas then run on a set of new material, on which no FA had been conducted. After creating the defect Pareto chart from diagnosis reports, the die for FA were selected. For one such experiment (Case 3), results are shown in Fig. 6. For FA, the number of die picked for each root-cause in the Pareto chart are shown as a red numeral at the top of each bar. The even- tual match with the FA result is also indicated to the right of each one for which die were picked. Notice in Fig. 6 that no die were picked for FA for CELL_OPEN Layer-A root cause. This was because this mechanism was already known from historical data on other silicon results fromthe same process. This highlights another advantage of this methodology in preventing useless FA. Instead, FA can be focused on finding new issues. FA found defects matching the predicted root causes in six of eight die, a 75% hit rate. Similar work was conducted on other sets of failing die as well, where all data was collected over several wafers from different lots. The overall match rate with FA on 35 die was 77%. Finally, the accuracy of cell-aware diagnosis on these die selected for FAwas 94% (excluding die for which no defect was found by FA). CONCLUSION The rapid adoption of FinFETs threatens the ability of manufacturers to ramp up and sustain semiconductor quality and yield. Because defects are observed at the transistor level, the industry needs new ways to detect and diagnose cell-internal defects. A breakthrough tech- nique in the field of scan diagnosis and machine learning finds defects inside of FinFETs, diagnoses the root cause, and improves FA success rates and efficiency. This tech- nique—cell-aware diagnosis with root cause deconvolu- tion—improves yield ramp and semiconductor quality. ACKNOWLEDGMENTS The authors would like to thank the following col- leagues for significant contributions to this work: Huaxing Tang, Matt Knowles, Wu-Tung Cheng, and Gaurav Veda fromMentor, a Siemens Business; and Kannan Sekar and Neerja Bawaskar from GlobalFoundries. REFERENCES 1. C.H. Gim: “A Novel Bitmap Analysis Technique - Test Sensitivity Intensity Bitmap,” IEEE Proc. Int. Symp. Phys. Fail. Anal. Integr. Circuits (IPFA), 2002. 2. T.Bartenstein,D.Heaberlin,L.Huisman,andD.Sliwinski:“Diagnosing Combinational Logic Designs using the Single Location At-A-Time (SLAT) Paradigm,” in IEEE Proc. Intl. Test Conf. (ITC), 2001, p. 287-296. 3. M. Abramovici, M.A. Breuer, and A.D. Friedman: Digital Systems Testing and Testable Design, IEEE Press, 1990. Fig. 6 Cell-aware diagnosis with RCD results, number of die picked for FA, and FA results for Case 3.