Inspection and classification of redistribution layer based on principal component analysis of high dimensional ultrasound signal

Abstract

The inspection of redistribution layers (RDL) in modern semiconductor packaging technology requires high-resolution equipment to detect porosity, interlayer separation, and depth information. In particular, owing to the micro/nanometer dimensions, it is difficult to analyze internal characteristics due to overlapping of surface signals using general pulse-echo ultrasound inspection. However, scanning acoustic microscopy (SAM) demonstrates high performance in detecting depth information in multi-layer structures with micro-thickness, offering quantitative analysis in a short time. This study proposes a technique to analyze the Cu-line of the multi-redistribution layer through a principal component analysis (PCA) based algorithm mainly used in dimension reduction methods. The PCA dimension reduction algorithm based on time series data can perform hierarchical analysis between thicknesses smaller than wavelengths compared to conventional pulse-echo inspection. The developed post-processing algorithm shows excellent performance in independent internal layer analysis, indicating that high-frequency focused ultrasound transducer is an optimal choice for visualizing the quality of the interface.