Wafer Bonding Defect Detection With Coded Ultrasound Array Imaging and Fourier Domain Analysis

Abstract

The continuous drive for miniaturization and 3D integration in the semiconductor industry necessitates uncompromising quality in wafer bonding. Scanning Acoustic Microscopy (SAM) is a benchmark for detecting micron-scale bonding defects, but its point-by-point scanning mechanism imposes a throughput bottleneck, rendering it unsuitable for large-area wafer inspection. This paper introduces a high-throughput inspection method that integrates high-frequency (30 MHz) ultrasound array imaging with coded plane wave transmissions adapted from medical imaging. Our approach employs a Cascaded Dual-Polarity Wave (CDW) excitation scheme to reduce frequency-dependent attenuation, thereby improving the signal-to-noise ratio (SNR) without compromising inspection speed. We also investigate unique spectral fingerprints of defects through Fourier domain analysis, improving detectability in sub-optimal SNR conditions where conventional time-domain signals are obscured. The proposed method was evaluated using k-Wave simulations and validated via experiments on custom-fabricated patterned wafers. Results demonstrated that our array-based approach increased inspection throughput by more than an order of magnitude compared to conventional SAM. Furthermore, Fourier analysis identified distinct defect signatures at specific resonant frequencies. This work underscores the feasibility of a viable ultrasound array imaging method for rapid and noise-robust in-line quality control in semiconductor manufacturing.