Modelling of physically relevant features in photoluminescence images
Photoluminescence (PL) imaging is a promising measuring technique to rate the quality of multi-crystalline silicon wafers. Defect structures which form during crystallization can be observed in the PL-images. In recent approaches, image features are measured and correlated to global solar cell parameters. So far, an underlying physical effect has not been connected to the large variety of observed PL-image patterns. Next to global solar cell parameters, also spatially resolved solar cell parameters like the dark saturation current density can be measured to rate wafer quality. It is shown that these spatially resolved measurements strongly correlate to the global open circuit voltage. Based on the observation that special patterns within the as-cut PL-images correlate with regions of high and low dark saturation current, a method for the modelling of physically relevant image features in PL-images is introduced in this work. The modelling framework is based on the transfer of spatially resolved quality data of the dark saturation current onto prototypical patterns within the PL-images of as-cut wafers by means of pattern recognition techniques. We evaluate our approach by predicting the images of the local dark-saturation current of the finished cells based on PL-images of as-cut wafers.