Identifying optimal photovoltaic technologies for underwater applications

Abstract

Improving solar energy collection in aquatic environments would allow for superior environmental monitoring and remote sensing, but the identification of optimal photovoltaic technologies for such applications is challenging as evaluation requires either field deployment or access to large water tanks. Here, we present a simple bench-top characterization technique that does not require direct access to water and therefore circumvents the need for field testing during initial trials of development. Employing LEDs to simulate underwater solar spectra at various depths, we compare Si and CdTe solar cells, two commercially available technologies, with GaInP cells, a technology with a wide bandgap close to ideal for underwater solar harvesting. We use this method to show that while Si cells outperform both CdTe and GaInP cells under terrestrial AM1.5G solar irradiance, CdTe and GaInP cells outperform Si cells at depths >2 m, with GaInP cells operating with underwater efficiencies approaching 54%.

Keywords: Applied sciences; Engineering; Water resources engineering.

Graphical abstract

Figure 1

LED- and Xe-simulated sunlight (A) AM1.5G solar spectrum from UV (300 nm) to IR (1100 nm). (B) Emission spectra of the 19 LEDs used to simulate both AM1.5G and underwater light, ranging from 425 to 1050 nm. (C) Comparison of irradiance spectra between AM1.5G, a Xe lamp with an AM1.5G filter, and the combined LED spectrum by optimizing each LED to best match the AM1.5G solar spectrum.

Figure 2

Water attenuation and resulting power (A) Accumulated power density from integrating the irradiance spectra shown in (Figure 1C). (B) Attenuation coefficient spectra ($\alpha$) of salt water off the coast of Key West (Jenkins et al., 2014) and pure water (Segelstein, 1981). Dashed lines indicate extensions made to the saltwater spectrum using the pure water spectrum. Inset shows $\alpha$ values in a narrow range from 300 to 800 nm. (C) Underwater irradiance spectra calculated using the Beer-Lambert law and the Key West attenuation spectrum. The AM1.5G spectrum is shown in gray for reference. (D) Resultant input power density obtained from Equation (3).

Figure 3

Simulations of underwater solar spectra (A–H) Spectral match between individual LEDs (filled, colored LED emission curves) and the total LED-array emission (solid black line) compared with the Beer-Lambert-corrected AM1.5G spectra at depths from 2 to 9 m (colored spectra). The insets display photographs of the resulting LED light, showing the transition from white light at 0 m to blue-green light at 9 m. (I) Accumulative power density curves obtained from integrating the total emission from the LEDs adjusted to best fit each underwater solar spectrum (solid black lines, Equation 2) and the underwater-corrected AM1.5G solar spectra (filled circles, Equation 3). The resulting power densities are listed.

Figure 4

Identifying optimal technologies (A) Normalized EQE spectra of c-Si, CdTe, and RHJ-GaInP cells along with the AM1.5G solar spectrum and the calculated underwater spectra. Bandgap values are shown. (B) Optimum bandgap as a function of depth below sea level, calculated using a detailed-balance model (Röhr et al., 2020), showing the Si, CdTe, and GaInP solar cell bandgaps for reference. The Shockley-Queisser limits assuming 6000 K black-body radiation and AM1.5G illumination at D = 0 m are shown. (C) p_out of the same cells under AM1.5G illumination (D = 0 m) and when simulating sunlight at D = 2–9 m. The gray solid line represents the maximum theoretical power density. (D) Corresponding $\eta$ values, using values for p_in given in Figure 1G. The gray line represents the maximum theorical efficiency. Solar J-V curves are shown in Figures S9A–S9C and tables with device characteristics are shown in Table S3.