Detection and analysis of micro-cracks in multi-crystalline silicon wafers during solar cell production

The reduction of wafer thickness requires an improved quality control of the wafer strength, which is significantly influenced by cracks. We introduce a machine learning framework to establish photoluminescence (PL) imaging as an optical inspection technique for the detection of cracks in multi-crystalline silicon wafers. The specially derived algorithm enables reliable crack detection in spite of similar background structures in the PL image from grain boundaries and dislocations. Within an experiment on thin wafers with artificially induced cracks we evaluate our approach by comparing the PL detection results to the findings of an infrared-transmission system and fracto-graphical reference data. Based on the optical detection result, we derive a description of the crack structure. Since wafer strength may change after etching and thermal processes, wafer strength is analyzed during cell production and correlated to the optical detection results.