The limited operational stability of perovskite solar cells is a drawback that requires a greater knowledge of degradation behaviour to be addressed. While some insights have been gleaned from individual studies based on relatively small datasets, a statistical analysis of stability data from a large number of devices with different characteristics would afford a more comprehensive understanding. Yet, such an analysis requires not only a large dataset but also homogenous data entries — that is, consistency in the measurement conditions and parameter reporting — both of which are currently lacking. Now, Noor Titan Putri Hartono, Antonio Abate and colleagues at the Helmholtz-Zentrum Berlin and the Freie Universität Berlin in Germany collected in a consistent way over 2,000 pieces of operational stability data from solar cells based on various materials and architectures and tested under controlled conditions.
The researchers benchmark the loss in the power conversion efficiency with respect to its maximum value in the first 150 h of the stability experiments. Devices with high efficiency are statistically more likely to show longer lifetimes, suggesting that efforts in increasing the device efficiency go in parallel with stability improvements. Using a self-organizing map machine learning method, Hartono et al. identify four shapes in the degradation curves: slow, medium and fast exponential decays, and initial gain in the efficiency. The majority of the devices with high efficiency show an initial gain in performance while none of them show a fast-exponential decay linked to a catastrophic failure, indicating that the shape of the degradation curve could be used as an early indicator of long-term stability. This kind of high-level analysis could help identify overarching trends and complement more in-depth investigations of the mechanisms and factors underlying degradation.
This is a preview of subscription content, access via your institution