Statistical methods for linking material composition to recombination losses in optoelectronic devices.

Area of Science:

• Materials Science
• Renewable Energy
• Statistical Analysis

Background:

• Industrial solar cell production generates large datasets from compositional characterization.
• Multicollinearity in high-dimensional compositional data hinders pattern discovery and performance correlation.

Purpose of the Study:

• To develop a statistical framework for analyzing compositional data from thin-film solar cells.
• To correlate compositional gradients with device performance parameters.
• To provide a data-driven approach for optimizing solar cell design.

Main Methods:

• Glow-discharge optical emission spectroscopy (GDOES) for depth-resolved composition analysis of 280 Cu(In,Ga)(S,Se)2 solar cell samples.
• Parameterization of [Ga]/([Ga] + [In]) and [S]/([S] + [Se]) gradients.
• Two-way clustering to group similar samples and features.
• Principal Component Analysis (PCA) for dimensionality reduction.

Main Results:

• A comprehensive map visualizing GDOES data and feature correlations across all samples.
• Identification of relationships between compositional grading and performance metrics like open-circuit voltage deficit.
• Successful grouping of samples and features based on compositional similarity.

Conclusions:

• Statistical analysis of GDOES data provides valuable insights into solar cell composition-performance relationships.
• The developed method enables precise planning for compositional grading optimization.
• Clustering and dimensionality reduction facilitate prediction of performance for new solar cell samples.