High-Throughput Battery Characterization

High-throughput characterization of battery materials

Subtle structural property changes and defects have appreciable impacts on sample performance and processing requirements. A detailed characterization of these effects with statistical significance can greatly enhance the efficacy of product and process optimization decision making.

Example: NMC antisite defect analysis across processing conditions

High-resolution X-ray powder diffraction measurements were performed on approximately 80 samples of NMC processed by various groups under different conditions. Sequential Rietveld refinement determined Ni-Li antisite defect density (known to affect the performance of the material in a cathode). Systematic evaluation using a random forest model was carried out to find correlations between process parameters and antisite defect density.

Antisite defect distribution across NMC samples

Antisite defect distribution across samples.

Antisite defect values across process steps

Antisite defect values obtained across different process steps.

Antisite defect values across different coatings

Antisite defect values obtained across different coatings.

Feature importance from random forest model

Which feature (processing parameter) has the greatest effect on antisite defects? The top 20 most important features from the random forest model: higher score equals higher importance. The results indicate that the processes of the surface activate coating are most important to the resulting antisite defect levels present. The top two contributors are calcination temperature (most important) and the coating amount (second most important).

Key findings

The processes of the surface-activated coating are most important to the resulting antisite defect levels present.

The top two contributors are calcination temperature (most important) and the coating amount (second most important).

Random forest feature importance analysis identifies which processing parameters have the greatest effect on antisite defects.

Benefits of our method

Statistical significance at scale

Process ~80+ samples in a single campaign to achieve statistically significant structural characterization across processing conditions.
Correlate structure with process

Systematic evaluation using machine learning models finds correlations between process parameters and structural defect properties.
Actionable optimization

Feature importance analysis identifies which processing parameters have the greatest effect—directly guiding process optimization.

Acknowledgements

This work was performed in partnership with various groups in the FESTBATT 2 Cluster with special thanks to Ramon Zimmermanns and Anna-Lena Hansen.

Contributions

HR-XRPD measurements were performed by Momentum Transfer
surface activated treatment: KIT, Group of Joachim Binder, experiments done by Yiran Guo & Valeriu Mereacre
storage/ vacuum drying: JLU Giessen, Group of Jürgen Janek, experiments done by Burak Aktekin & Johannes Schubert
spray drying: Fraunhofer IKTS, experiments done by Jean-Philippe Beaupain
microjet reactor: Uni Saarland, Group of Guido Falk, experiments done by Dina Klippert
chemical coating: FZ Jülich, IMD-2, experiments done by Christoph Roitzheim
heat treatment: collection of reference samples across different groups