Concentrated thermoelectric generators convert solar energy to electricity, but historically their conversion efficiency has lagged behind their potential. Now, full system efficiencies of 7.4% are achieved by segmentation of two thermoelectric materials and a spectrally selective surface.

Concentrated solar thermoelectric generators offer an intriguing alternative to wind turbines and photovoltaic modules for the production of electricity from renewable sources1,2. Such thermoelectric generators are solid-state heat engines3, directly converting the heat from sunlight to electrical power by exploiting the large temperature difference that develops across a thermoelectric generator under concentrated solar irradiation. While the solar thermoelectric generator literature dates back to 1888, it has been challenging to realize efficient devices, and a modest efficiency value of 3.4% under 50-fold concentration achieved in the mid-1950s remained a high water mark until this century4. However, a flurry of recent activity in solar thermoelectric generators has been inspired by improvements in thermoelectric materials, including advanced models of system performance, optical design optimization, and occasional bench-top demonstrations with efficiencies of 3–5% (refs 1,2,5,​6,​7). Writing in Nature Energy, Gang Chen, Zhifeng Ren and colleagues at MIT and the University of Houston now report on the design, construction, and testing of a 7.4%-efficient concentrated solar thermoelectric generator utilizing the segmented leg geometry of Bi2Te3 and CoSb3 that operates with a 600 °C hot side temperature8.

The researchers have implemented innovations at both the thermoelectric unicouple and system level. The transition from measurements of individual thermoelectric properties to fabrication of a device is highly nontrivial. Critical challenges in this transition include mitigating thermomechanical stresses arising from temperature gradients and developing non-reactive interconnects and bonds that retain their desired electrical and mechanical properties upon cycling. Chen, Ren and colleagues not only address these challenges using state-of-the-art materials, but do so in an unusually transparent way. Optimized ingots of Bi2Te3 and CoSb3 were fabricated in-house, metallized and bonded into a unicouple (a single pair of n- and p-type legs). A spectrally selective absorber was then bonded onto the unicouple and integrated into a custom optical test-bed equipped with a 200 sun source. A schematic illustration of the device is shown in Fig. 1a. Test measurements under such illumination require care, as the optical flux at high concentration must be well-calibrated and thermal stresses during cycling can destroy the generator.

Figure 1: The concept and case for concentrated thermoelectric generators.

Figure 1

a, Schematic illustration of the generator used by Chen, Ren and colleagues. Concentrated solar thermoelectric generators produce electricity from the large temperature gradient across the active thermoelectric materials. Critical innovations in the present work include a spectrally selective receiver and a segmented Bi2Te3/skutterudite unicouple for maximizing efficiency. SKU, skutterudite. b, With increasing photovoltaics on the electricity grid, time-shifting electricity production to the evening is critical to avoid over-generation (where base load production exceeds net demand) and photovoltaic throttling. Shown here are past and predicted future daily electricity demand profiles for the California ISO grid in the US. The 2011 and 2016 profiles correspond to data for March taken from ref. 9. The 2021 forecast is generated starting from 2015 data in ref. 9 and assuming an increase in total electricity demand at a rate of 1.28% per year (ref. 10), an increase of wind power production at a rate of 62.5 MW per year (using data from ref. 9), an increased peak (13:00–14:00) photovoltaic power production at a rate of 1.46 GW per year, and an increased photovoltaic production at all other hours by a proportional amount (using data from ref. 9). The base load is an estimate derived from ref. 11, assuming that all nuclear (2 GW), hydroelectric (1 GW), and 80% of imports (7 GW) and natural gas (8 GW) are inflexible. As can be seen, the rapid increase in demand as the sun sets poses an additional challenge for the grid, which could be partially met by thermoelectric generators. Panel a reproduced from ref. 8, NPG.