Automated material development for solar cells

Until now, the design of molecular compounds suitable for semiconductor components has been based on trial and error. This method has a number of disadvantages: it is labor-intensive and inefficient. This is because the human ability to recognize patterns in huge, complex data sets is limited. In contrast, the research team used machine learning (ML) models and trained them with experimental and computer simulation data to predict molecular structures and properties for optimal device performance. Around 100 molecules were sufficient for this.

Next came a series of optimizations. In the first round, the algorithm suggested 24 molecules. The team then synthesized and tested these. It turned out that they were already better than the current references. This was verified again in a second optimization round. The most efficient material candidates achieved an efficiency of up to 24%, surpassing the previous reference value of 22%.

Automated sample measurement

"With the new approach, we can not only search systematically, but also narrow down the search area. In addition, the newly discovered chemical compounds can be optimized and tested for the desired material properties," explains Prof. Dirk Guldi, holder of the Chair of Physical Chemistry. High-throughput screening (HTS) plays a central role here. These are automated laboratory systems that prepare, dose and measure a large number of samples in parallel. This method is not only more precise, it also reduces the time required in the research process and minimizes human error. In future, it will be possible to search through material libraries containing millions of molecules in a short space of time and discover candidates that are optimal for the desired function of the components.

"We assume that this technology will develop beyond solar energy into an innovation driver for other industries that want to develop new materials using high-throughput processes," explains Prof. Christoph Brabec, holder of the Chair of Materials Science.