Entwicklung effizienter ternärer NFA-basierter organischer Photovoltaik durch Machine-Learning Methoden: ENFA


Photovoltaics play a central role in the expansion of renewable energies. Due to its combination of flexibility, transparency, robustness and cost efficiency, organic photovoltaics has the potential to enter areas that are difficult or impossible to serve with established silicon-based technologies, such as building integrated photovoltaics.

Planning experiments in physical and chemical sciences is very often done by varying one parameter at a time to then find an optimum in a multidimensional parameter space. As many process parameters are highly correlated, this approach of navigating the parameter space along orthogonal lines is not a time-efficient strategy to find the optimum parameter combination.

An alternative is a design of experiment approach based on machine learning algorithms such as support vector regression or Bayesian optimization, which is then applied to the optimization of organic solar cells. Furthermore, using Bayesian inference methods combined with the use of surrogate models for numerical device simulations can help to extract material parameters from spectroscopic experiments such as measuring intensity dependent current-voltage curves, impedance spectroscopy or transient photovoltage. Finally, correlating the extracted material parameters obtained from Bayesian inference with process parameters can lead to a reduction in information entropy.

Last Modified: 05.04.2023