As the technology keeps scaling down and IC design becomes more and more complex, failure analysis becomes much more challenging, especially for static fault isolation. For semiconductor foundry FA, it will become even more challenging due to lack of enough information. Static fault isolation is the major global fault isolation methodology in foundry FA and it is difficult to access and trigger the failing signal detected by scan and BIST test, which is widely applied in modern IC design. Because, in most of the time, the normal two pin bias (Vdd and Vss) can only get the comparable IV result between bad unit and the reference unit for function related fail. There are two possibilities from reverse engineering perspective. Firstly, the defect location may not be accessed by the DC bias. Secondly, even if the defect can be accessed, but the defect induced current or voltage change is too small to be differentiated from the overall signal. So it will be concealed in the overall current. However, it is still possible for us to do global fault isolation for the second situation. In this paper, a unit with Iddoff failure was analyzed. Although, no significant IV difference was observed between failed and reference units, a distinct Photon Emission (EMMI) spot was successfully observed in the failed unit. Layout analysis and process analysis on this EMMI spot further confirmed the reality of the emission spot.

As the rapid developments of semiconductor manufacturing technologies, the CD of the device keep shrinking. The IC devices have a smaller feature sizes and higher densities, and thus there are many challenges come up in terms of the failure analysis and localized device characterization. Besides the challenge of smaller feature size, there is another challenge as well. Some of the traditional FA method can no longer be employed on advanced technologies, such as 28nm and beyond. Quickly and successfully isolating the failed location and obtaining electrical signature of the defect has become more of a challenge, especially for the device level analysis and characterization. AFP nanoprobing system provides some solutions to advanced nodes fault isolation through its AFM imaging mode of CAFM.

In this study, a 65nm product level low yield case has been investigated and its failure mechanism was identified. Root cause analysis was discussed and concluded. The product has been hit with ATPG failure with a unique wafer map signature - a butterfly pattern. Tools commonality and timeframe analysis show that the highly suspected process is the Metal1 Cu seed PVD step. To understand the failure mechanism and its root cause, product level FA was needed. However due to its functional failure property, the conventional EFA is not applicable in this case. Instead GDS study was performed to isolate the failure sites. Subsequently physical FA analysis was carried out at the identified sites to reveal its failure mechanism. Metal1 void was observed on the sidewall of the metal1. Meanwhile, a very interesting phenomenon was observed. If die was selected on the left part of the butterfly pattern, the void would be on the right side sidewall of the metal. If the die was selected on the right part, the void would be on the left side sidewall of the metal1. All of the voids were towards wafer center. After in-depth study of the PVD process, we suspect the pass die could also have void. These voids must be also towards wafer center. Subsequent PFA on good unit confirmed our suspect. The more detailed mechanism of the void formation was discussed and evidences supporting our analysis are to be presented in the paper. Nevertheless, the butterfly pattern is still a question in our mind. After in-depth analysis, we found the voids formation was associated with Metal1 orientation. Because of the horizontal orientation of Metal1, if the void happens it should locate in the end of the metal line in the butterfly area. While the majority of Via1/contact are stand on the line end, so the open Via1/contact failure will happen. For the die out of the butterfly area, the majority of the void locates in the sidewall of the metal line center. The majority Via1/contact are not stand in the center of the metal line center, of no Via1/contact open happen. But it is still has reliability concern. Much more detailed and in-depth mechanism is investigated in the paper. Moreover, improvement is also touched on. Systematic problem solving method is employed in this case. It is good reference for same kinds of failure analysis.