Qualitative and Quantitative Analysis of Multi-Pattern Wafer Bin Maps

• Published in: IEEE transactions on semiconductor manufacturing Vol. 33; no. 4; pp. 578 - 586
• Main Authors: Kong, Yuting, Ni, Dong
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.11.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Artificial neural networks; boundary detection; convolutional neural network; Convolutional neural networks; Feature extraction; Monitoring; Multi-pattern wafer map; overlapped pattern segmentation; Pattern analysis; Qualitative analysis; Quantitative analysis; Segmentation; Segments; Semiconductor device manufacture; semiconductor manufacturing
• ISSN: 0894-6507; 1558-2345
• DOI: 10.1109/TSM.2020.3022431

Wafer map analysis is one of the most critical steps for monitoring wafer quality and tracking failures in the semiconductor manufacturing process. Defective dies on wafer bin maps (WBMs) usually cluster into specific spatial patterns, which contain critical information for cause identification and yield improvement. Failures frequently arise due to the high complexity of production processes. Multiple patterns are likely to appear on one WBM. This article proposes a qualitative and quantitative analysis approach for multi-pattern WBMs. The boundary detection method is proposed to locate and segment pattern groups at first. Then, the overlapped pattern unwrapping method is proposed to segment overlapped patterns. With two stages of segmentation, multi-pattern WBMs are separated into multiple single pattern WBMs. A convolutional neural network is introduced to determine the classes of patterns. After pattern segmentation and classification, the segmented single pattern WBMs will be remapped to the original multi-pattern WBMs, and the impacts of patterns are calculated. Our approach is validated with the real-world dataset, and the results demonstrate the proposed method has a good performance of qualitative and quantitative analysis for multi-pattern WBMs.