Microcracks in Silicon Wafers I: Inline Detection and Implications of Crack Morphology on Wafer Strength

Microcracks in silicon wafers reduce the strength of the wafers and can lead to critical failure within the solar-cell production. Both detection of the microcracks and their impact on fracture strength of the wafers are addressed within this study. To improve the accuracy of the crack detection in photoluminescence (PL) and infrared transmission (IR) images of as-cut wafers, we introduce a pattern recognition approach based on local descriptors and support-vector classification. The learning model requires a set of labeled data generated by an artificial insertion of cracks. Within this evaluation, the algorithm detects 81% of the cracks for PL-images and 98% for IR-images at precision rates above 98% in each case, which outperforms the quality of pure IR-intensity-based crack-detection systems with a hit-rate of 65% at a precision of 59%. The proposed algorithm may be combined with the images of the grain structure to avoid the confusion of cracks and grain boundaries. Moreover, the comprehensive set of wafers allows the impact of crack morphology on wafer strength to be investigated. Despite complex crack morphologies, the theoretically expected dependence between crack length and fracture strength is confirmed. Therefore, sorting criteria are derived to rate the cracks with respect to the expected fracture strength of the wafer based on the measured crack length only.