edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 1 40 ISTFA 2023 FOCUSED ION BEAM (FIB) USER GROUP Chair/Co-Chairs: Valerie Brogden, Steve Herschbein, Michael Wong, and Edward Principe vbrogden@uoregon.edu, steven.herschbein@gmail.com, mike.wong@thermofisher.com, eprincipe@synchres.com Steve Herschbein presented on the status of FIB-related input to the EDFAS Failure Analysis Technology Roadmap gap analysis council, led by Keith Serrels, NXP Semiconductor. Recently published by EDFAS, the FA Technology Roadmap endeavors to capture gaps across all failure analysis areas and examine potential alignment vectors with the CHIPS for America Act. This act allocates more than $50 billion dollars over five years toward restoring U.S. leadership in semiconductor manufacturing and addresses the sobering fact that none of the most advanced logic and memory chips are manufactured at commercial scale in the United States. Herschbein highlighted that failure analysis is not broken out explicitly in the CHIPS Act, but is enveloped into metrology. The Gap Analysis Council is exploring efforts with the Department of Commerce to call out specific opportunities directed at failure analysis. The council is working to partner with companies to bring forward proposed projects, in order to help fund them with need demonstrate through the gap analysis. One avenue of interaction with failure analysis technologies that Herschbein mentioned specifically is the National Semiconductor Technology Center (NSTC), a publicprivate consortium in cooperation with the Department of Commerce who is partnering with the Department of Defense, the Department of Energy, and the National Science Foundation. The NSTC is a critical part of CHIPS that will conduct research and prototyping of advanced semiconductor technology, support workforce-training programs, and maintain an investment fund to help startup companies commercialize new technologies. Focused ion beam, being a broad distributed technology, spans an array of the gap analysis council areas. Noteworthy FIB technology categories include precision ion sectioning and shaping as applied to: i) atom probe sample preparation; ii) isolation of FinFET structures; iii) TEM/STEM analysis of sub 10 nm structures with minimal amorphous damage and; iv) repetitive “slice and view,” requiring higher fidelity to define device structures consisting of hundreds of nanometer scale slices, which is driving demand for quantitative slice thickness and more precise beam control for FIB-based 3D reconstruction. Circuit edit continues to be a monumental challenge and FIB User Group presenters and co-chairs. increasingly requires power connections on the backside, concurrent with signal collection from the frontside of an active device under investigation. Large area selective delayering highlights the opportunity for novel ion sources using alternate ion species, improved gas chemistries, as well as improving the acquisition rates by up to 10X using compressed sensing technologies. Following Herschbein’s gap analysis review, Or Haimson of Annapurna Labs presented on “Computer Vision for Physical Failure Analysis (PFA)”. Haimson’s presentation demonstrated through real-world PFA examples how artificial intelligence (AI) tools and code generators, such as ChatGPT, Amazon Code Whisperer, and Amazon Lookout for Vision are emerging tools to aid the PFA engineer who is not an expert code developer. (Co-chairs’ note: Other code generator tools include OpenAI Codex, GitHub Copilot, ChatGPT, Bing, and Tabnine). Haimson tested Amazon Lookout for Vision on SRAM anomaly detection. The training data set consisted of approximately 20 “normal” defined images to create a model to identify abnormalities. Labels can be defined (supervised training) to augment the model with descriptors such as dark PVC, bright PVC, missing contact, and over etch. Applications include allowing machine vision modules to pre-process FIB-based 3D reconstruction data slices directly to identify the 3D regions of the anomaly or defect. Improvements in end-pointing and feature detection in circuit edit was highlighted as another potential use of machine vision
RkJQdWJsaXNoZXIy MTYyMzk3NQ==