A team of researchers led by Prof. Christoph Brabec, spokesperson of the FAU Solar Profile Center and Chair of Materials Science at FAU has developed an innovative, closed-loop workflow to rapidly identify high-performance materials for perovskite solar cells (PSC). This groundbreaking method, published in Science, integrates computational modeling, autonomous synthesis platforms, and quantum-theoretical calculations to predict and test optimal material combinations in an automated manner.
The study addressed a long-standing question in photovoltaic research: “What is the most efficient way to discover new materials with optimal properties for photovoltaic devices?” Over the course of more than a year, a multidisciplinary team of 21 scientists from chemistry, materials science, computer science, and electrical engineering collaborated to develop this advanced approach.
Traditionally, researchers have relied on either fully digital methods—virtually synthesizing molecules to predict properties—or labor-intensive experimental approaches that populate large data libraries. However, both methods are inherently limited by trial-and-error processes and human capacity to analyze vast, complex datasets. “Our hybrid approach leverages machine learning (ML) models trained on both experimental data and computer simulations,” explains Prof. Brabec. “With just around 100 molecules, we were able to train the algorithm to predict molecular structures and properties for optimal device performance.” In the initial round, the algorithm suggested 24 molecular candidates, which were synthesized and tested. Remarkably, these outperformed existing reference materials. In the second optimization cycle, the materials demonstrated power conversion efficiencies of up to 24%, surpassing the current reference efficiency of 22%.
Central to the workflow is high-throughput screening (HTS), an automated laboratory system that prepares, doses, and measures large numbers of samples in parallel. HTS not only enhances precision but also significantly reduces time and minimizes human error. This enables researchers to rapidly explore material libraries containing millions of molecules, identifying candidates with properties tailored for specific photovoltaic applications. “With this approach, we can systematically narrow the search space, optimize new chemical compounds, and directly test their material properties,” adds Prof. Dirk Guldi, Chair of Physical Chemistry at FAU and co-spokesperson of the FAU Solar Profile Center.
Beyond accelerating material discovery, the workflow provides deeper insights into structure-property relationships. While chemists have long predicted molecular properties based on isolated structures, accurately forecasting the performance of a molecule within a solar cell remains a challenge due to the complexity of interacting parameters. “Our algorithm allows us to ask, ‘Which input parameters are most relevant for solar cell performance?’” says Prof. Brabec. “This marks a breakthrough in understanding how molecular structure translates to device efficiency.”
The implications of this technology extend beyond photovoltaics. By enabling systematic high-throughput material discovery, the platform could become a driver of innovation across industries, from renewable energy to advanced materials for electronics and beyond. “This new approach not only accelerates the search for optimal materials but also opens doors to a deeper understanding of their behavior within devices,” concludes Prof. Brabec. “We believe it has the potential to transform materials research across many sectors.”
A team of researchers led by Prof. Christoph Brabec, spokesperson of the FAU Solar Profile Center and Chair of Materials Science at FAU has developed an innovative, closed-loop workflow to rapidly identify high-performance materials for perovskite solar cells (PSC). This groundbreaking method, published in Science, integrates computational modeling, autonomous synthesis platforms, and quantum-theoretical calculations to predict and test optimal material combinations in an automated manner.
The study addressed a long-standing question in photovoltaic research: “What is the most efficient way to discover new materials with optimal properties for photovoltaic devices?” Over the course of more than a year, a multidisciplinary team of 21 scientists from chemistry, materials science, computer science, and electrical engineering collaborated to develop this advanced approach.
Traditionally, researchers have relied on either fully digital methods—virtually synthesizing molecules to predict properties—or labor-intensive experimental approaches that populate large data libraries. However, both methods are inherently limited by trial-and-error processes and human capacity to analyze vast, complex datasets. “Our hybrid approach leverages machine learning (ML) models trained on both experimental data and computer simulations,” explains Prof. Brabec. “With just around 100 molecules, we were able to train the algorithm to predict molecular structures and properties for optimal device performance.” In the initial round, the algorithm suggested 24 molecular candidates, which were synthesized and tested. Remarkably, these outperformed existing reference materials. In the second optimization cycle, the materials demonstrated power conversion efficiencies of up to 24%, surpassing the current reference efficiency of 22%.
Central to the workflow is high-throughput screening (HTS), an automated laboratory system that prepares, doses, and measures large numbers of samples in parallel. HTS not only enhances precision but also significantly reduces time and minimizes human error. This enables researchers to rapidly explore material libraries containing millions of molecules, identifying candidates with properties tailored for specific photovoltaic applications. “With this approach, we can systematically narrow the search space, optimize new chemical compounds, and directly test their material properties,” adds Prof. Dirk Guldi, Chair of Physical Chemistry at FAU and co-spokesperson of the FAU Solar Profile Center.
Beyond accelerating material discovery, the workflow provides deeper insights into structure-property relationships. While chemists have long predicted molecular properties based on isolated structures, accurately forecasting the performance of a molecule within a solar cell remains a challenge due to the complexity of interacting parameters. “Our algorithm allows us to ask, ‘Which input parameters are most relevant for solar cell performance?’” says Prof. Brabec. “This marks a breakthrough in understanding how molecular structure translates to device efficiency.”
The implications of this technology extend beyond photovoltaics. By enabling systematic high-throughput material discovery, the platform could become a driver of innovation across industries, from renewable energy to advanced materials for electronics and beyond. “This new approach not only accelerates the search for optimal materials but also opens doors to a deeper understanding of their behavior within devices,” concludes Prof. Brabec. “We believe it has the potential to transform materials research across many sectors.”