Nano-scale Defect Analysis Through K-Means Clustering of CuInSe2 Solar Cells with Ag and K Incorporation

Nano-X-ray fluorescence microscopy is used to correlate elemental inhomogeneity to device performance measured through X-ray beam induced current (XBIC). Unsupervised machine learning techniques can be useful in the processing and correlation of XRF/XBIC data. A case study using CuInSe
₂
solar cells treated with Ag-alloying and KF-post- deposition treatment (PDT) is presented. The implementation of K-means clustering is studied for its ability to identify statistically meaningful results from XRF/XBIC data. The procedure described may be applied to datasets of a similar nature.