Most solar panels convert less than 20 percent of the energy in the sunlight that falls on them into electricity. A new $2.4 million project funded by the U.S. Advanced Research Projects Agency for Energy aims to greatly increase the amount of sunlight that becomes electricity. Its goal is a conversion efficiency of more than 50 percent, which would more than double the amount of power generated by a solar panel of a given size. This would cut the number of solar panels needed in half and potentially make solar power more competitive with fossil fuels.

In the new research effort, Harry Atwater, a professor of applied physics and materials science at Caltech, plans to use precisely structured materials to sort sunlight into eight to 10 different colors and direct those to solar cells with semiconductors that are matched perfectly to each color. As a result, more of the solar spectrum will be absorbed, and the energy contained in each slice of the spectrum will be converted mostly to electricity, rather than heat.

The general idea of sorting sunlight by color isn’t new. One approach involves growing multiple semiconductor materials in a stack—light moves through the stack until it’s absorbed by a semiconductor that can convert it efficiently. This approach has yielded commercial solar cells with efficiencies of over 43 percent. But the process for making such solar cells is expensive, and the power output of the device is limited by the worst-performing layer.

Others have tried sorting light into various colors using conventional lenses, mirrors, and filters, but the prototypes have been bulky and haven’t reached very high efficiencies, in part because of the imprecision of the optics—it’s proved difficult to direct exactly the right wavelengths of light to each solar cell. It’s also been difficult to split the light up into more than a couple of different colors in one device. 

In the last several years, however, scientists have gotten better at manipulating light at a very small scale, sorting it by color, trapping it, and guiding it from one spot to another using thin layers of material that incorporate tiny features that are often smaller than the wavelength of light.  Atwater plans to draw on these advances to manipulate light precisely and in a compact flat package that might not look that much different than a conventional solar panel. One layer would split light up, sort it by color, and then deliver it to a second layer that contains an array of solar cells matched to each color.

The challenge with this approach is that no one makes these precisely structured materials over the large areas and in the large volumes needed in the solar industry. But Atwater compares the device to a flat screen TV, which is itself a sophisticated device for manipulating light, with its millions of transistors for switching on and off different colored pixels.

“The first ones that came out were many thousands of dollars and had defects. Now you can get one for less than a hundred dollars that’s essentially perfect, and the costs are going down all the time,” he says. “Flat displays are an example of something that’s at the scale of a solar panel, but are incredibly complex optoelectronic circuits. What we’re proposing is primitive by that standard.” 

Atwater says the manufacturing tools needed to make his nanostructured materials are starting to come on the market. They’ll remain expensive, however, as long as production volumes are low. Researchers are also closing in on the ability to make thin wafers of various semiconductors and transfer them to a device like the one he’s envisioning.