Automated probe technique for robust and high-throughput electrical characterization and measurements

• Main Author: Chenardy, Frederick Hubert
• Other Authors: Kedar Hippalgaonkar
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Materials::Material testing and characterization; Engineering::Materials::Microelectronics and semiconductor materials
• Online Access: https://hdl.handle.net/10356/147697
• Institution: Nanyang Technological University

In materials research, high-throughput experiments and characterization techniques have been instrumental in driving new materials development and design. In recent years, the use of advanced data analytics and application of state-of-the-art machine learning and artificial intelligence models to materials science have resulted in optimization and automation in both fundamental and applied research. The combination of statistical computing and algorithms has proved to be effective in not only eliminating inefficient and repetitive ways, but also providing useful insights through prediction and high-speed computation. In this research, we seek to explore and implement computer vision to automate an electrical probe technique to enhance the robustness and throughput of electrical characterization and measurements, such as conductivity measurement using the four-point probe method and the field-effect transistor (FET) transfer measurement. As a subset of artificial intelligence (AI), computer vision is a highly powerful tool which gives a computer system the ability to understand, process and recognize features within digital images or videos. Throughout the project, various image processing and object detection techniques, such as image segmentation, contour detection and analysis and the advanced Haar cascade classifier approach, are studied and utilized to build a robust detection model to identify and recognize small-sized thin-film device, from which useful data is extracted. This report introduces the implementation strategies, such as the hardware and software set-up, image processing and object detection frameworks used, image data preparation method as well as user-friendly parameter tuning to optimize the algorithms. Moreover, the report also discusses the detection results obtained from the model and evaluates the effectiveness of the algorithms in performing precise detections and optimizing the electrical probe technique for robust and high-throughput electrical characterization and measurements.