GlobalFoundries and Mentor create a breakthrough for scan diagnosis with machine learning

Improve yield and failure analysis by detecting, refining, clarifying, and resolving defects inside standard cells.

Today’s advanced technology nodes and the use complex finFETs and multi-patterning can make test and failure diagnosis more difficult than ever. For example, more defects and systematic yield issues now occur in the front-end layers that are inside library cells, complicating the detection and diagnosis of these defects. For yield engineers tasked with finding the root cause of yield loss through failure analysis (FA), the high cost and low success rate of FA on finFETs is becoming a significant burden.

GlobalFoundries set out to solve the problem of advanced node yield improvement by partnering with Mentor, a Siemens business to perform some experiments with advanced scan diagnosis software. The experiments used cell-aware diagnosis, a new and effective way to detect defects inside standard cells. Cell-aware volume diagnosis can report suspects inside the cells because it uses fault models derived from analog simulation. It also uses a failure data collection and diagnosis flow identical to that of traditional diagnosis.

But, semiconductor companies need both to diagnosis defects inside of finFETs and also to clarify the results, to cut through the ambiguity or noise in the diagnosis report. No one benefits from chasing down fake defect suspects.

To cut through diagnosis ambiguity, Mentor developed an unsupervised machine learning algorithm, called root-cause deconvolution (RCD), which estimates the defect distribution from volume diagnosis results in the presence of noise.

GlobalFoundries and Mentor have published the experiments, which you can read here.

The results of the experiments show that Mentor’s cell-aware diagnosis with RCD is very effective at increasing the resolution of the diagnosis by reducing the number of suspects in cell-internal defect data. With more precise root cause distribution, GlobalFoundries found that die selection for FA is much more precise, and overall yield analysis cycle time and cost shrinks accordingly. Take a look at the work GlobalFoundries and Mentor have done to shows that cell-aware diagnosis with RCD improves yield ramp and semiconductor quality.

Leave a Reply

This article first appeared on the Siemens Digital Industries Software blog at https://blogs.sw.siemens.com/tessent/2019/03/24/globalfoundries-and-mentor-create-a-breakthrough-for-scan-diagnosis-with-machine-learning-2/