▲ Kim Hyun-jin’s team and his lab students took a commemorative photo. (Photo from DKU Portal)

   Semiconductor technology is pivotal for electronics and plays a major role in electronic engineering. Despite the considerable progress made in advancing semiconductor technology, there is an ongoing need for continual improvements. The Embedded & Parallel Systems Laboratory (EPSL), an AI research lab in the Dept. of Electronic Electric Engineering at Dankook University (DKU), contributed significantly to the development of semiconductor technology and as a result, on June 27, the paper, ’Wafer Failure Pattern Detection Using Image Matching Algorithm,’ written by Professor Kim Hyun-jin and students Oh Su-min, and Kwon Young-wook of EPSL, earned the top 2 spot in the Best Paper Award at the 24th academic conference of the Institute of Semiconductor Test of Korea (ISTK).

   ISTK is a prominent entity dedicated to the comprehensive research, development, and supply of overall semiconductor tests and applications. The ISTK’s 24th academic conference was a large-scale semiconductor test conference supported by domestic semiconductor companies such as Samsung Electronics, SK Hynix, and LG Electronics. The paper ‘Wafer Failure Pattern Detection using Image Matching Algorithms’ introduces a paradigm shift in addressing cases of wafer failure. Unlike conventional methods requiring the input of all potential failure scenarios, this approach creates efficiency by offering a streamlined method to mitigate time and cost expenditures associated with wafer failures. Semiconductors have front and back-end processes. The paper deals with the improvements to the backend process, which takes place after the wafer is made. A wafer is a thin slice of a semiconductor for an integrated circuit designed with microdevices. In this study, AI is used to detect failures in semiconductors during the backend process. The paper asserts AI technology was used to match the key points of a normal wafer with those of erroneous ones, leveraging a secondary AI algorithm. This AI algorithm compares and analyzes the key points of the pattern, such as brightness, scale, and angle. This method of testing proved more efficient than ever before. Following this paper, Professor Kim Hyun-jin emphasized the growing importance of wafer error detection due to the continual reduction in semiconductor size and the simultaneous increase in density. Additionally, they plan to apply for a patent for their innovative error detection approach as well as concentrate on a follow-up study about the automatic design of electronic circuits.

▲ Professor Kim Hyun-Jin Who Received the ISTK Best Paper Award (Photo from DKU Portal)

   To gain insights into this latest research achievement, the Dankook Herald (DKH) conducted an exclusive interview with Professor Kim Hyun-jin, and student researchers Kwon Young-wook, and Oh Su-min in their laboratory on August 11. During the interview, the professor wanted to give students more opportunities to speak, and they enthusiastically responded to our queries. The core of this research paper was to develop an algorithm that could quickly verify wafer soundness using an 'image-matching' algorithm embedded in a deep learning algorithm. This approach is a significant improvement over conventional manual scrutiny, where defects were examined by human visual inspection during the wafer production and verification stages, which are critical for semiconductor-integrated circuits.

   In addition to being selected as the best paper, they are also planning to register their design for a patent with the support of SK Hynix. When we asked why they selected this topic, Kwon Young-wook and Oh Su-min said they were studying machine learning when the professor suggested the idea of ‘image-matching.’ However, this was a difficult project because it was a new concept, and there was no data available on the topic. The conversation naturally gravitated towards their forthcoming study. Professor Kim said, "In semiconductor design, most people think of circuit creation, but there is an important field in producing software that helps design or identify problems, which is known as ‘design automation’. Therefore, based on this research paper, we intend to further develop it into the field of design automation and submit it to international academic societies.”

   The DKH asked about their strategies for dealing with the difficulties of research. Professor Kim emphasized the importance of clearing your head. While a complete immersion in research and studying is crucial, taking a break is also necessary to foster creativity. Additionally, he recommended looking at what others have done. Analyzing how other researchers solved similar problems can serve as a great point of reference. Students Kwon Young-wook and Oh Su-min added that organizing existing data is crucial to avoiding future difficulties while preparing a research paper. If information is hard to find, your research will suffer. During the interview, Professor Kim provided opportunities for students Kwon and Oh to express their viewpoints. It was clear he had a genuine concern for their experiences. During the research process, Professor Kim always emphasized three key principles. Firstly, choosing a field for paper writing that holds significance is important. Fields already extensively researched or no longer require further investigation by industries might result in weaker academic outcomes, so they should be avoided. Secondly, he advocates exploring many different possible research ideas rather than fixating on a singular subject. Incorporating insight from new fields and technologies is advisable for broadening your perspective. For instance, machine learning, which was not widely used decades ago, is gaining attention now. Thus, we should broaden our horizons. Lastly, he draws a distinction between short-term and long-term pursuits. Although long-term research is innovative, it may not yield results for a decade or more. Thus, rather than focus only on long-term research, it is beneficial to simultaneously consider short-term studies aimed at improving existing knowledge based on the findings of others. Considering this advice, students Kwon and Oh felt that their experience in this academic conference would enable them to conduct more detailed research in the future. They expressed their gratitude to Professor Kim. Also, they expressed a hope that their research might contribute to advancements in semiconductor testing technology. Furthermore, Professor Kim Hyun-jin had a message for Dankookians (Students of DKU). “We are in an era where computers can do everything, like get information through books or get answers with formulas. Thus, we must cultivate skills that extend beyond what computers can provide.”

   While EPSL was researching AI, Professor Kim Hyun-jin, and students Oh Su-min and Kwon Young-wook combined image matching with an AI algorithm to test for wafer production errors. Utilizing image-matching for error detection represents a novel approach, and it was accomplished with limited existing references. Despite the scarcity of related data, Oh Su-min and Kwon Young-wook pursued their studies independently with some assistance from their professor. Their ability to solve problems represents how well they operate in teams, and their close relationship was evident during the interview. Universities are responsible for providing good quality lectures as well as fostering learning beyond the classroom. The DKH is proud of the collective accomplishments of these Dankookians and anticipates the further advancement of semiconductor technology because of their efforts.