A perovskite cell tuned for blue-green light hits 34.7% underwater
Serguey Asael Shinder2 min readSeawater absorbs light past about 630 nanometres — red, infrared and everything longer. A conventional silicon panel is built to harvest roughly 400 to 1,100 nm, so most of what it is designed for is gone before the light reaches any useful depth. A team led by Wen-Hua Zhang at Yunnan University built perovskite cells for the light that actually survives: the blue-green 400 to 600 nm band that penetrates five to ten metres. The write-up is at https://spectrum.ieee.org/perovskite-underwater-solar-panels.
The headline result is a reversal. Under standard land conditions the same cells convert 17.08% of incident light — below the 20 to 25% a commodity silicon panel manages. Given the spectrum found ten metres down, they reach 34.71%. Nothing about the cell changed; the light did.
They then took it to water. Cells encapsulated in glass, synthetic rubber and epoxy were fitted to small underwater robots with lithium-ion batteries and deployed near Weizhou Island in the South China Sea, holding a fixed depth to measure performance against distance from the surface. At 2 metres, 115 cm² of cells produced 1,416 mWh over two hours — roughly half a rechargeable AA. At 10 metres the same array still produced 324 mWh in two hours, which Zhang says beat his expectations by a factor of three to six.
What it means
This is a spectrum-matching result, not a solar result, and that is the transferable part. A cell's efficiency is quoted against a standard terrestrial spectrum that almost no deployed cell actually sees — not under water, not under snow, not under a sodium lamp, not behind tinted glass. Design for the light that arrives and the number moves a long way, in both directions. The 17.08% figure in air is the same device failing the test it was not built for.
The application is not grid power and should not be sold as such. 324 mWh over two hours from a hand-sized array is a sensor budget, not a generator. But that is precisely the gap in subsea work: autonomous vehicles, moored sensors, cameras and acoustic modems all end their missions on battery, and a trickle that recharges in daylight changes mission length rather than power ratings.
And the packaging is where this will actually be decided. Glass, rubber and epoxy held for an experiment near an island. Marine deployment is a story of biofouling, pressure cycling and seawater ingress over years, and perovskites are not famous for tolerating moisture. The efficiency number is real; the service life is the thing to ask about next.