Solar energy is becoming more affordable and easier to produce every day. As the market grows, so too is the field of research. Solar scientists have been working with perovskites, a family of mineral and ceramic materials that are a top contender to replace silicon as the next level up in photovoltaic materials. The problem is that they are enormously difficult to manufacture on a large scale. 

A new AI system, though, uses improved machine learning that could make perovskites the next step in solar evolution. 

The solution to perovskites 

Perovskites are layered crystalline compounds whose atoms form a distinctive crystal lattice. Perovskite photovoltaic cells showed enormous promise in labs, where they were made using a spin-coating technique. The material can be used to make panels that are thinner and lighter, have greater efficiency, are less expensive, and are easier to transport and install. However, making a couple of perovskite cells in the lab is one thing. Manufacturing is another. Spin-coating is not a practical method for large-scale manufacturing. 

Particularly, there are too many variables to consider for perovskite at the manufacturing level, such as the composition of the starting materials, the temperature, the humidity, the speed of the processing path, and the methods of curing the material. Perovskite has a lot of different compound compositions, and this makes it hard to consistently work with. 

A new AI system developed by researchers from MIT and Stanford uses machine learning to solve this problem. The new program includes raw data and, unusually, incorporates human experience. This includes the experimenters’ qualitative observations of the test samples and experiments from other researchers. 

A great turnaround to a new generation

The new machine learning in this AI has enabled more accurate manufacturing of perovskite cells and given it an energy conversion efficiency of 18.5 percent, which puts it at a competitive level for the market today. 

Many solar companies are already gearing up to start producing perovskite cells. Many of them are still working out the specifics of how to manufacture them at scale, and now, with the new AI from MIT and Stanford, the process could soon be streamlined to make perovskite the next generation of solar energy. 

Source Study: MIT News — Engineers enlist AI to help scale up advanced solar cell manufacturing | MIT News | Massachusetts Institute of Technology