Elucidating the Full Potential of OPV Materials Utilizing a High-Throughput Robot-Based Platform and Machine Learning

• Published in: Joule Vol. 5; no. 2; pp. 495 - 506
• Main Authors: Du, Xiaoyan, Lüer, Larry, Heumueller, Thomas, Wagner, Jerrit, Berger, Christian, Osterrieder, Tobias, Wortmann, Jonas, Langner, Stefan, Vongsaysy, Uyxing, Bertrand, Melanie, Li, Ning, Stubhan, Tobias, Hauch, Jens, Brabec, Christoph J.
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 17.02.2021
• Subjects: automation; efficiency; high-throughput methods; machine learning; organic photovoltaics; photostability; processing optimization; efficiency; processing optimization; automation; organic photovoltaics; high-throughput methods; machine learning; photostability
• ISSN: 2542-4351; 2542-4351
• DOI: 10.1016/j.joule.2020.12.013

Evaluating the potential of organic photovoltaic (OPV) materials and devices for industrial production is a multidimensional optimization process with an incredibly large parameter space. Here, we demonstrate automated OPV material and device characterization in terms of efficiency and photostability. Gaussian process regression (GPR) prediction based on optical absorption features guided the optimization process with promising prediction accuracy for PV parameters and burn-in losses. With ∼100 process conditions, screening for efficiency and photostability can be finished within 70 h. The highest power conversion efficiency (PCE) of 14% was achieved by fully automated device fabrication in air with a model material system PM6:Y6. Improving molecular ordering has been identified as the most promising motif for further efficiency optimization. Thin active layers combined with medium thermal annealing temperature are favorable to simultaneously improve efficiency and suppress burn-in losses. The platform and protocol may be expanded to any solution-processed organic semiconductor and interface materials. [Display omitted] •Automated robot-based organic solar cell fabrication and characterization•A high-quality dataset enables machine-learning-based predictions•Reliable screening of more than 100 processing variations within 70 h•A platform and methodology for broader research areas in solution-processed devices Evaluating the potential of OPV materials and devices for industrial viability is a multi-dimensional, large parameter space exploration. Manual experimentation is extremely limited in throughput and reproducibility. Automated platforms for fabricating and characterizing functional devices have the potential to accelerate experimentation speed with precise process control. Here, we demonstrate a multi-target evaluation of OPV materials at the full-device level with an automated platform called “AMANDA Line One.” Over 100 processing variations are automatically screened, which allows a reliable evaluation in terms of efficiency and photostability. The high-quality dataset enables a promising prediction of performance with Gaussian process regression. The platform and methodology can be generalized to broad research areas including, but not limited to, other solution-processed PV technologies, light emitting diode, photodetectors, and transistors. A self-developed platform called AMANDA Line One is demonstrated for fabricating and characterizing organic solar cells automatically in a high-throughput manner. Over 100 processing variations are screened in terms of efficiency and photostability within 70 h. Combined with machine learning methods, a multi-target evaluation of OPV materials at the full-device level is achieved based on the high-quality dataset, which allows a promising prediction accuracy for both efficiency and photostability. The platform could be generalized for broader research areas.