edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 1 42 micromachining in the 3-5 nm range in and around active transistors when using a gallium ion source. Neon micromachining with a non-conductive particle is still an active area of research to mitigate the contrast issue and Zeiss is also continuing research in the circuit edit domain using the gas field ion source. A final question posed by a FIB user group attendee sought to understand the reasoning behind the anecdotal evidence that lower voltage (e.g., 12 kV) milling with a plasma FIB mitigated formation of trapped bubbles, causing porosity in the deposition layer. Lukáš Hladík from Tescan indicated they use a protocol of 10-15 kV to avoid formation of xenon-induced pores during platinum or tungsten deposition. Rick Livengood further commented on the well-documented observations of subsurface trapping of lighter element helium and neon ions, which coalesce into nano and micro bubbles via mobility driven by diffusion. The speculation being that the fundamental process in the xenon plasma FIB is similar. Livengood added that the cesium laser cooled source has the interesting property that it tends to diffuse to the surface and boil off, as opposed to coalescing into micro bubbles. Presenters and co-chairs of the System in Package User Group. ISTFA 2023 SYSTEM IN PACKAGE (SIP) USER GROUP Chair/Co-Chairs: Yan Li, Bryan Tracy, Wentao Qin yan.a.li@samsung.com, btracy@tesla.com, wentao.qin@microchip.com The System in Package (SIP) User Group adopted a new user group panel discussion format for 2023 to encourage more interactions between panel members and the audiences. This year’s topic, “Overcome the Challenges in SIP FA,” was fully discussed with active participations from panel members and audiences. SIP FA gaps identified by the FA technical roadmap councils were illustrated by Yan Li from Samsung, which include: 1) AI applications in SIP FI-FA. 2) Challenges related to big SIP package dimension, sample prep, fault isolation, non-destructive imaging. 3) FI and FA tool robustness. Bryan Tracy from Tesla asked a few insightful questions to stimulate dynamic discussions among panelists and audiences relating to the FA roadmap. Zhiyong Wang at Analog Devices highlighted the importance of test program in SIP FA. Sebastian Brand from the Fraunhofer Institute IMWS, emphasized the additional efforts to ensure reliability and investigate robustness during SIP FI-FA. This requires the availability of nondestructive techniques of different contrast mechanisms to allow for screening and repetitive analyses in SIP development and production. Novel analysis and data evaluation and processing concepts are being developed to increase the certainty and accuracy for defect detection and localization while remaining nondestructive. Florian Felux, Infineon Technologies AG, elaborated a few case studies where the application of AI played a critical role in SIP FI-FA. Jim Colvin of FA Instruments Inc., asked a challenging question to the panel members and audiences: How to find out the defect causing the same failure in five SIP devices, which unfortunately can recover during physical failure analysis? Audiences and panelists got excited about the problem statement and were eager to express opinions of how to establish a FI-FA workflow to systematically find the recoverable defect. Colvin later revealed that the defect was a metal dendrite, which led to an intermittent leakage failure. The discussion was then switched to how to do non-destructive fault isolation and high-resolution 3D x-ray CT of the metal dendrite. Colvin was happy to know that subtle metal dendrite can be nondestructively imaged by high-resolution 3D x-ray CT. To wrap up the discussion, he eventually disclosed his unique physical FA technique, which avoided the failure recovery and preserved the metal dendrite. The audiences started to think about more challenging cases, which were very common in product FA support: How to find out the root cause with very limited failing samples, for instance, a single failing unit. A very detailed background study was recommended, including test and yield data, and previous similar failure data.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==