There are many variables that can easily be changed to increase the performance of solar cells. These include material type, thickness and arrangement. It has been tedious to make small adjustments to each parameter when developing new solar cells. Although computational simulators make it possible to test these changes without actually building each variation, the process is still slow.

 

Researchers at MIT and Google Brain now have a system that allows you to not only evaluate one design but also to give information about the changes that will make it better. This system could significantly increase the speed at which new configurations are discovered.

 

The new system is called a differentiable sun cell simulator. It was described in a paper published in the journal Computer Physics Communications today by MIT junior Sean Mann and Giuseppe Romano, research scientist at MIT’s Institute for Soldier Nanotechnologies. There are also four other authors at MIT and Google Brain.

 

Romano explains that traditional solar cell simulators take details about a solar cell’s configuration and predict their output efficiency. This is the percentage of energy converted from sunlight to electricity. The new simulator predicts the efficiency and shows how each input parameter affects that output. He says, “It shows you exactly what happens to efficiency if this layer is made thicker or what happens if the property of the material changes.”

 

He said that while he didn’t find a new device, the tool he created will allow others to quickly discover other high-performance devices.

 

Mann claims that traditional methods rely on a random search for possible variations. Mann’s tool allows him to “follow a trajectory of changes because the simulator tells us which direction we want to go with our device.” This makes it much quicker because you don’t have to explore every possible option, but instead can follow one path that will lead directly to better performance.

 

Advanced solar cells are often composed of layers interlaced with conductor materials to carry electric charges from one to the next. This computational tool shows how changing the thicknesses between these layers can affect the output of the device. This is crucial because thickness is important. Mann explains that there is a strong interplay among light propagation, the thickness and absorption of each layer.

 

You can also evaluate the amount of doping (the introduction atoms from another element) each layer receives. Also, the dielectric constant for insulating layers or the bandgap can be used to measure the energy levels of photons that can be captured in different materials.

 

Romano states that the simulator is available now as an open-source tool and can be used immediately to guide research in this area. Romano says that the simulator is available and can be used by industry professionals. Researchers would combine this device’s computations and an optimization algorithm or machine learning system to quickly assess a variety of possible changes, and then narrow down the most promising.

 

The simulator currently works with a single-dimensional solar cell model. It will soon be expanded to accommodate two- and three-dimensional configurations. Romano states that even the 1D version of the solar cell “can cover most cells currently in production.” Mann says that while some variations cannot be directly simulated by the tool, Mann says that it is possible to simulate tandem solar cells made from different materials by simulating each individual cell.

Romano states that the simulator is “end to end”, meaning it calculates efficiency and light absorption. Romano says that a future direction would be to combine our simulator with differentiable light propagation simulators. This will allow for greater accuracy.

 

Romano said that this open-source code means that the community can contribute to it once it’s available. That’s why we’re so excited.” Romano says that even though this research group is small, anyone can contribute to the code to improve it and create new capabilities.