|
1 |
Based on the data used in the industry and domain expertise, we aim to drive industrial innovations by enhancing the design-analysis-manufacturing process using data science and machine learning technology together with our deep understanding of the underlying physics. In particular, we focus on the highly valuable field of semiconductor technology innovation. Image processing technology based on deep learning streamlines measurement and inspection processes. Data- and physics-based lithography computational simulation studies are conducted to innovate lithography processes widely used in various semiconductor fabrications. Additionally, data- and physics-based fatigue life prediction models are devised, and simulation techniques are advanced.
Data-driven Metrology and Inspection
Metrology and inspection (MI) is a crucial task in manufacturing industries. For example, in semiconductor manufacturing, measuring the critical dimension of lithography patterns and identifying unintended pattern defects are important to ensure the quality of semiconductor products. However, conventional algorithm-based MI procedures are often inaccurate or too slow as the feature size to be characterized is getting smaller and smaller. To resolve this problem, we are developing novel MI procedures based on deep neural networks. Our major targets include lithography patterns in semiconductor manufacturing and self-assembled DNA nanomaterials.
Publications
- *Jaehoon Kim, *Jaekyung Lim, Jinho Lee, Tae-Yeon Kim, Yunhyoung Nam, Kihyun Kim, and Do-Nyun Kim. "Hotspot prediction: SEM image generation with potential lithography hotspots." IEEE Transactions on Semiconductor Manufacturing (2024).
- Tae-Yeon Kim, Sunjae Park, Chang-Kue Lim, Jae-Hyuk Choi, Chang-Hyun An, Lak-Hyun Song, and Do-Nyun Kim, "Deep learning-based detection of defects in wafer buffer zone during semiconductor packaging process." Multiscale Science and Engineering (2024).
- *Yunhyoung Nam, *Sungho Joo, Nohong Kwak, Kihyun Kim, and Do-Nyun Kim. "Precise pattern alignment for die-to-database inspection based on the generative adversarial network." IEEE Transactions on Semiconductor Manufacturing (2022).
- Young-Joo Kim, Jaekyung Lim, and Do-Nyun Kim. "Accelerating AFM characterization via deep-learning-based image super-resolution." Small (2022).
- Jaehoon Kim, Yunhyoung Nam, Min-Cheol Kang, Kihyun Kim, Jisuk Hong, Sooryong Lee, and Do-Nyun Kim. "Adversarial defect detection in semiconductor manufacturing process." IEEE Transactions on Semiconductor Manufacturing (2021).
Data-driven and Physics-informed Computational Lithography
Lithography, a technology that fabricates circuit patterns on silicon wafers, is the core process in the semiconductor manufacturing. While various computational methods to predict the lithography patterns formed on a wafer are available, they usually suffer from extremely high computational cost as complex physical and chemical interactions need to be simulated at the molecular level. To overcome this limitation, we focus on developing a novel method for computational lithography by integrating data-driven machine learning approaches with physical models in order to achieve high computational efficiency and prediction accuracy. Major applications include the mask optimization with OPC (optical proximity correction) and EPC (etch proximity correction), virtual printability check of lithography patterns, and hotspot (defect) prediction and correction.
Data-driven and Physics-informed Estimation of Fatigue life
Estimating the life of a system is crucial in many industrial applications. However, abnormal or defect data indicating the potential onset of structural failure is scarce, making it difficult to monitor and evaluate the structural health. Here, we aim to develop a data-driven and physics-informed method for estimating fatigue life by incorporating the small data into the Bayesian networks model. For example, it would be useful to predict the speed and the path of crack growth from a detected crack point minimizing the use of computationally expensive physical simulation.
Publications
|
