This paper presents a detection method for characterizing the bonded interface of O2 plasma activated silicon wafer direct bonding. The images, obtained by infrared imaging system, were analyzed by the software based on spatial domain and morphology methods. The spatial domain processing methods, including median filtering and Laplace operator, were applied to achieve de-noising and contrast enhancement. With optimized parameters of sharpening operator patterns, disk size, binarization threshold, morphological parameter A and B, the void contours were clear and convenient for segmentation, and the bonding rate was accurately calculated. Furthermore, the void characteristics with different sizes and distributions were also analyzed, and the detailed statistics of the void's number and size are given. Moreover, the orthogonal experiment was designed and analyzed, indicating that O2 flow has the greatest influence on the bonding rate in comparison with activated time and power. With the optimized process parameters of activated power of 150 W, O2 flow of 100 sccm and time of 120 s, the testing results show that the bonding rate can reach 94.51% and the bonding strength is 12.32 MPa.
Keywords: infrared image; orthogonal experiment; silicon direct bonding; spatial domain and morphology; void characteristics.