Artificial neural network-driven diffraction imaging for nanoscale optical critical dimension metrology in semiconductor microstructures
Journal
The International Journal of Advanced Manufacturing Technology
ISSN
02683768
Date Issued
2026
Author(s)
Abstract
This study presents an optical critical dimension (CD) measurement technique that integrates artificial neural networks (ANNs) with coherent optical scatterometry (COS) and back focal plane (BFP) imaging for three-dimensional (3D) advanced semiconductor packaging. The proposed method enables simultaneous measurement of line width, line spacing, and height in redistribution layer (RDL) structures by analyzing diffraction signatures from periodic structures. Compared with atomic force microscopy (AFM) and scanning electron microscopy (SEM), which typically require several minutes per measurement, the proposed approach achieves measurement times of less than 3 ms, corresponding to a two-order-of-magnitude improvement in throughput. In comparison to white-light interferometry and confocal microscopy, the method offers approximately a fivefold improvement in lateral resolution, accompanied by significantly higher measurement speed. Experimental validation against AFM demonstrates measurement biases below 2% for line width and spacing and below 1.2% for height. The simple optical configuration also results in lower implementation costs. These characteristics make the proposed method suitable for inline process control in RDL fabrication and 3D IC manufacturing. Current limitations include a restricted depth measurement range and reliance on finite-difference time-domain (FDTD) simulations, which will be addressed in future work.
Subjects
Artificial intelligence
Coherence optical scatterometry
Optical critical dimension
Re-distribution layer
Publisher
Springer Science and Business Media Deutschland GmbH
Type
journal article