Semiconductors are the backbone of modern electronics, embedded in everything from smartphones to automobiles and medical devices. Even minor chip failures may have significant consequences to health, safety and operations. It is therefore essential that the semicon industry analyses the failures, understands their root cause and translates the resulting knowledge back into improvements in chip design and manufacturing. However, increased chip complexity greatly expands the variety of root causes that needs to be analyzed in exploding volumes of data. MuSICA will create a novel integrated spectroscopic and data analysis platform that will be much more sensitive and discerning for detecting failures in complex semiconductors, designed to recommend improvements to their design and manufacture.
Multidisciplinary approach
MuSICA will integrate the solution to both the physical and data-related scientific challenges in semiconductor Quality Assurance. The project will focus on new spectroscopic instruments to detect how light interacts with matter and how this may reveal semiconductor failures using new Hyperspectral Imaging Data Science. This joint step will allow (1) detection of failure-associated patterns beyond the existing, limited theoretical models on the studied materials and (2) a more sensitive detection of novel failures. An academic collaboration between physics instrumentation development (for time-resolved spectroscopy) and materials-informed Data Science is rare if not unprecedented and will lead to a new paradigm in interdisciplinary science.
Public-private collaboration
The consortium will stimulate new collaborations at the verge of different disciplines within Radboud University and consolidate ongoing collaborations with NXP, NT-MDT and HAI. NXP Semiconductors and NT-MDT started their collaboration with Alexey Kimel in 2020 within the ETN MSCA project. HAI has been collaborating with Jeroen Jansen in the ongoing Engineering Business Intelligence project funded by TKI Energy & Industry, in which they co-develop a data streaming infrastructure combined with process predictions from newly developed AI, for different process industries. The consortium will focus on improving the instrumentation based on the digital outcomes, to improve the sensitivity, detectability and predictability of chip failure.
Increased chip reliability and life-time
An increase in the success rate of the failure analysis process used by chip manufacturers will improve the understanding of the root causes of chip failures, identifying critical production steps and chip features. Knowing what went wrong, why it happened and how to prevent reoccurrence will provide key input for both the production process and the design to avoid future failures of key electronic components in our everyday life. Improving failure analysis in the semiconductor industry will thereby – in the longer run – address several societal challenges, such as the reduction of waste and critical raw materials, safer and more reliable consumer electronics and transport, more resilient infrastructure and healthcare technologies, thereby mitigating risks in national security and economic growth.
