Answers are grounded in this paper and cite their sources.
We demonstrate die-level field-of-view (FoV) metrology for µLED (micro-light-emitting diode) patterns assisted by a proposed learning-based defect inspection framework. In this study, high-resolution, large-FoV images were obtained by reflection-mode Fourier ptychographic microscopy (FPM) providing sub-1 µm pitch resolution imaging over a 1×1mm 2 FoV. Patterns with varying pitches and diverse geometries were fabricated, incorporating both natural and artificial defects to emulate lithography and pattern transfer processes. Defect types were systematically defined to facilitate supervised and unsupervised learning defect detection methods. Deep learning models, including a YOLO-based detector and a self-supervised segmentation framework, achieved high accuracy and robustness in identifying diverse defect types. The proposed approach offers a scalable and efficient metrology solution for inline defect detection and process optimization for next-generation µLED and advanced display manufacturing.
No takes yet. Share an insight, caveat, or question.
Mahmoudpour et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: