Machine Learning and Correlative Microscopy: How ‘Big Data’ Techniques Can Benefit Thin Film Solar Cell Characterization

Correlative microscopy techniques have improved tremendously in the last 5 years and enabled simultaneous high spatial resolution mapping of a variety of material parameters. As acquisition speeds and resolution increase giving us a greater density of data points and the functionality of the measurements add more dimensions to be analyzed; the handling, management and analysis of data sets becomes more and more complicated. Operando measurements as well as in-situ studies pose a new challenge. Finding correlations in the 3+ dimensional data sets that result from many of these measurements is not straight forward, and the possibility of missing correlations and trends is increasingly concerning. In the following, we evaluate the potential of two machine-learning methods, an unsupervised Gaussian-Mixture-Model and a supervised Naïve Bayes model, to identify two groups of pixels in a micro-X-ray fluorescence map, given varied input data sizes.