Machine learning may make perovskite solar panels competitive with silicon-based photovoltaics, new research by the Singapore-MIT Alliance for Science and Technology suggests.
Perovskites are a family of crystalline solids whose structure makes them promising for numerous optoelectronic and photonic applications. However, the development of these technologies is currently limited by the difficulties associated with transferring small-scale lab techniques to commercial manufacturing environments. Adapting any given manufacturing process for these larger scales requires the simultaneous optimisation of "a dozen or so" variables, which is "impossible" under normal conditions.
Now, a team from the Singapore-MIT Alliance for Science and Technology has developed a machine learning platform which uses not only conventional experimental variables, but also human observation, to determine the ideal manufacturing process for a given application. The code developed by the team has already been made available on GitHub for anyone - solar cell manufacturers included - to download.
This innovative fusion of machine learning and human expertise may help to bring perovskite solar cells out of the laboratory and into the marketplace.