Dicing Lane Quality Quantification & Wafer Assessment Using Image Thresholding Techniques

• Published in: 2024 IEEE 10th Electronics System-Integration Technology Conference (ESTC) pp. 1 - 6
• Main Authors: Leech, Damien Jon, Hiro, Akito, Schoofs, Geert, Altintas, Bensu Tunca, Phommahaxay, Alain, Kennes, Koen, Beyer, Gerald, Beyne, Eric
• Format: Conference Proceeding
• Language: English
• Published: IEEE 11-09-2024
• Subjects: blade dicing; chipping; dicing; Image edge detection; image thresholding; Mechanical variables measurement; Microscopy; Optical imaging; Optical microscopy; Optical variables measurement; Particle measurements; plasma dicing; Plasmas; Semiconductor device measurement; Systematics; wafer assessment
• ISSN: 2687-9727
• DOI: 10.1109/ESTC60143.2024.10712068

The quality of a dicing lane is a qualitative measurement, with many characterization methods available including mechanical and optical profilometry, however the lowest barrier to entry is to simply inspect the dicing lane in an optical microscope and make a spot assessment of the chipping, dicing width and particulate presence. This is useful to the trained eye, however it remains difficult to accurately compare and contrast dicing lanes with differing processing parameters. As such we produce a series of Python scripts that seek to assess the quality of a dicing lane, irrespective of the process or materials, in a manner similar to the 'kerf checks' done in situ on many vendor tools, using the OpenCV library to effectively isolate the dicing lane in an image, locate the edges and assess the number of particles surrounding the lane. Kerf checks are common method of assessing dicing quality during the process, however here the tests are done following completion of the whole process, so they allow for a more complete assessment of the wafer and dicing quality and can be repeated following other process steps. In addition, these methods can be extended to other dicing features such as grooved lanes and are intended to be agnostic of the exact processing parameters used. As a simple example, we use these scripts to quantify the differences between blade dicing and plasma dicing. We then expand the process, using automated image acquisition tools, that allow for a more systematic assessment of our diced wafers and the ability to store data on how the dicing quality varies across a wafer. We periodically take images of dicing lanes from centre to edge, assess the dicing quality and track the variation across the wafer to highlight areas of issue or overall inconsistencies.