Scientists use machine learning to identify high-performance solar materials

Now, thanks to a study combining the power of computation, data science and experimental methods, researchers at the Argonne Laboratory of the US Department of Energy (DOE) and the University of Cambridge in England have come up with a new approach to design at the device identifying promising materials for dye solar cells (DSSC). DSSCs can be manufactured with scalable and inexpensive techniques, enabling them to achieve competitive performance / price ratios.

The team, led by Argonne materials scientist Jacqueline Cole, who also heads the Cavendish Laboratory molecular engineering group at the University of Cambridge, used the Theta supercomputer from the Argonne Leadership Computing Facility (ALCF). ) to identify five high-cost technologies for dyeing materials from a pool of nearly 10,000 candidates for device manufacturing and testing. The ALCF is a user facility of the DOE Office of Science.

"This study is particularly interesting because we were able to demonstrate the complete cycle of discovery of data-driven materials, since the use of advanced computer methods to identify materials with optimal properties up to the synthesis of these lab materials and their test in real photovoltaic devices, "Cole said.

As part of an ALCF Data Science Program project, Cole collaborated with Argonne computer scientists to create an automated workflow that combines simulation, data mining and learning techniques. automatic to simultaneously analyze thousands of chemical compounds. The process began with a sorting effort among hundreds of thousands of scientific journals to collect data on chemicals and absorption for a wide variety of organic dye candidates.

"The advantage of this process is that it removes the old manual maintenance of databases, which requires work for several years, and reduces it to a few months or even days," he said. said Cole.

The computational work involved the use of increasingly finer screening techniques to generate pairs of potential dyes that could work together to absorb light across the solar spectrum. "It's almost impossible to find a dye that really works for all wavelengths," Cole said. "This is particularly true with organic molecules because they have narrower optical absorption bands, yet we really wanted to focus only on organic molecules because they are much more environmentally friendly."

To reduce the initial batch of 10,000 potential color candidates to some of the most promising opportunities, we again relied on the ALCF computing resources to carry out a multi-step approach. First and foremost, Cole and his colleagues used data mining tools to eliminate all organometallic molecules, which typically absorb less light than organic dyes at a given wavelength, as well than organic molecules too small to absorb visible light.

Even after this first pass, the researchers still had about 3,000 dye candidates to consider. To further refine the selection, scientists selected dyes containing carboxylic acid components that could be used as chemical "glues" or anchors to fix dyes to titanium dioxide carriers. Next, the researchers used Theta to perform electronic structure calculations on the remaining candidates to determine the molecular dipole moment – or degree of polarity – of each dye.

"We really want these molecules to be polar enough so that their electron charge is high in the molecule," said Cole. "This allows the electron excited by the light to travel the length of the dye, to pass through the chemical glue and to penetrate the titanium dioxide semiconductor to start the electrical circuit."

After having reduced the search to about 300 dyes, the researchers used their computer configuration to examine their optical absorption spectra to generate a batch of about 30 days that could be subjected to an experimental verification. Before synthesizing the dyes, however, Cole and his colleagues performed computationally intensive Theta Thickness Functional Density Calculations (DFT) to evaluate their likely performance in an experimental setting.

The final step of the study was to experimentally validate a collection of the five most promising candidate dyes from these forecasts, which required global collaboration. As each of the different dyes had been synthesized in different labs around the world for another purpose, Cole contacted the original dye developers, who each returned a new dye sample to his team.

"It was really a great team effort to involve so many people from around the world in this research," Cole said.

By examining dyestuffs experimentally at the Argonne Center for Nanoscale Materials, another DOEOffice of Science user facility, and at Cambridge University and Rutherford Appleton Laboratory, Cole and colleagues found that some of between them, once integrated power conversion efficiency roughly equal to that of the standard industrial organometallic dye.

"This result was particularly encouraging because we had simplified our lives by limiting ourselves to organic molecules for environmental reasons, yet we found that these organic dyes were as good as some of the best-known organometallics," said Cole.

An article based on the study, "The Approach from Design to Device Provides Panchromatic Co-sensitized Solar Cells", appeared in the cover story of the February 1 issue of Advanced energy materials.

Explore further:
The discovery of solar cells opens a new window to supply the cities of tomorrow

More information:
Christopher B. Cooper et al., Dye Sensitive Solar Cells: The design-device approach allows for panchromatic co-sensitized solar cells (Adv Energy Energy 5/2019), Advanced energy materials (2019). DOI: 10.1002 / aenm.201970014