Thermophotovoltaic efficiency of 40

Abstract

Thermophotovoltaics (TPVs) convert predominantly infrared wavelength light to electricity via the photovoltaic effect, and can enable approaches to energy storage^1,2 and conversion^3-9 that use higher temperature heat sources than the turbines that are ubiquitous in electricity production today. Since the first demonstration of 29% efficient TPVs (Fig. 1a) using an integrated back surface reflector and a tungsten emitter at 2,000 °C (ref. ¹⁰), TPV fabrication and performance have improved^11,12. However, despite predictions that TPV efficiencies can exceed 50% (refs. ^11,13,14), the demonstrated efficiencies are still only as high as 32%, albeit at much lower temperatures below 1,300 °C (refs. ^13-15). Here we report the fabrication and measurement of TPV cells with efficiencies of more than 40% and experimentally demonstrate the efficiency of high-bandgap tandem TPV cells. The TPV cells are two-junction devices comprising III-V materials with bandgaps between 1.0 and 1.4 eV that are optimized for emitter temperatures of 1,900-2,400 °C. The cells exploit the concept of band-edge spectral filtering to obtain high efficiency, using highly reflective back surface reflectors to reject unusable sub-bandgap radiation back to the emitter. A 1.4/1.2 eV device reached a maximum efficiency of (41.1 ± 1)% operating at a power density of 2.39 W cm^-2 and an emitter temperature of 2,400 °C. A 1.2/1.0 eV device reached a maximum efficiency of (39.3 ± 1)% operating at a power density of 1.8 W cm^-2 and an emitter temperature of 2,127 °C. These cells can be integrated into a TPV system for thermal energy grid storage to enable dispatchable renewable energy. This creates a pathway for thermal energy grid storage to reach sufficiently high efficiency and sufficiently low cost to enable decarbonization of the electricity grid.

Fig. 1. Tandem TPVs.

a, History of some TPV efficiencies with different cell materials: Ge^, (dark grey), Si (yellow), GaSb (light grey), InGaAs^,,– (dark blue), InGaAsSb (light blue) and GaAs (orange). The black line shows the average thermal efficiency of power generation in the United States using a steam turbine (coal and nuclear)^,. Before the year 2000, turbine efficiencies shown also include natural gas. b, Energy that is incident on the TPVs ($P_{\mathrm{inc}}$) can be converted to electricity ($P_{\mathrm{out}}$), reflected back to the emitter ($P_{\mathrm{ref}}$) or thermalized because of inefficiencies in the cell and back reflector ($Q_{\mathrm{c}}$). c, d, The 1.2/1.0 eV (c) and 1.4/1.2 eV (d) tandems that were fabricated and characterized in this work, and a representative spectrum shape at the average emitter temperature (2,150 °C blackbody) indicating the spectral bands that can be converted to electricity by the top and bottom junction of a TPV cell. A gold mirror on the back of the cell reflects approximately 93% of the below bandgap photons, allowing this energy to be recycled. TJ represents the tunnel junction.

Fig. 2. TPV characterization.

a, Reflectance of the 1.4/1.2 eV and 1.2/1.0 eV tandems. The 2,150 °C blackbody spectrum is shown for reference, which is the average emitter temperature in the TEGS application. b, Internal quantum efficiency (IQE) of the 1.4/1.2 eV and 1.2/1.0 eV tandems. The EQE is shown in Extended Data Fig. 3. c, d, Current density–voltage curves measured in the efficiency setup at varying emitter temperatures for the 1.4/1.2 eV (c) and 1.2/1.0 eV (d) tandems.

Fig. 3. TPV efficiency.

a, TPV efficiency measured at different emitter temperatures ranging from approximately 1,900 °C to 2,400 °C. The error bars indicate the uncertainty of the efficiency measurement, which is discussed in Methods. The dashed lines show the model predictions and the shaded regions show the uncertainty in the model predictions (see Methods). b, Predicted efficiency of the 1.4/1.2 eV and 1.2/1.0 eV tandems as the weighted sub-bandgap reflectance ($R_{\mathrm{sub}}$) is extrapolated assuming a W emitter with AR = 1 and VF = 1 and a 25 °C cell temperature (Extended Data Fig. 5). The solid lines show the average efficiency within the TEGS operating temperature range of 1,900 °C to 2,400 °C. The shaded bands show the maximum and minimum efficiencies within the temperature range. The dots show the present value of $R_{\mathrm{sub}}$ based on the measured reflectance in Fig. 2a weighted by the W AR = 1, VF = 1 spectrum.

Extended Data Fig. 1. TPV applications.

a) Conceptual illustration of TEGS, which takes in electricity, converts it to heat via Joule heating, stores the heat in insulated graphite blocks, and then uses TPV for conversion of heat to electricity. A unit cell of the power block is also shown. B) Sankey diagram showing the energy flows in the TEGS system at scale and different efficiency metrics. c) The relationship between TPV sub-system efficiency and power block size or volume to surface area ratio, Φ, assuming the system is a cube. d) Conceptual illustration of a combustion-based electricity generation system using TPV. The system consists of an all-ceramic recuperator, similar to a printed circuit heat exchanger, with the end comprising of a combustion chamber. Air is preheated by exhaust and then combined with fuel for combustion near the end facing the TPV. The hot exhaust then delivers heat to the ceramic which radiates it to the TPV. e) Sankey diagram showing energy flows in a combustion-based TPV system at scale.

Extended Data Fig. 2. Tandem device structures.

Device structures of the 1.4/1.2-eV and the 1.2/1.0-eV tandems.

Extended Data Fig. 3. External Quantum Efficiency.

The external quantum efficiency (EQE) of the two cells. The blue curve shows the 2150 °C blackbody spectrum for reference.

Extended Data Fig. 4. Emitter spectrum measurements and model.

The emitter spectrum was measured at different emitter temperatures spanning the test temperature range. A model (Methods) was fit to the measurement and used to extend the spectra measurements to longer wavelengths. The spectral radiance goes to zero > ~4500 nm due to the presence of the quartz envelope around the bulb, as quartz is absorbing beyond this wavelength.

Extended Data Fig. 5. Comparison of spectra shapes.

a) A comparison between the spectrum shape at an intermediate test temperature (2130 °C) The red curve shows the modeled spectrum which agrees well with the measurement (see Extended Data Fig. 4). The gray curve shows comparison to a blackbody spectrum shape at the same emitter temperature. The blue curve shows comparison to the spectrum described by the literature emission of tungsten with AR=1, VF=1. All curves are normalized by their peak to show the comparison in spectra shapes. The spectrum shape under which the cells were characterized (red curve) is similar to that of a blackbody (gray curve), particularly above bandgap. Comparison of modeled TPV efficiency under the spectrum in this work with emitters which could be incorporated into a TPV system in which the $\mathit{AR}$ and $\mathit{VF}$ allow for the reflected light to be recycled. Shown is a tungsten (W) emitter with $\mathit{AR}=1$ and $\mathit{VF}=1$ as well as a blackbody emitter (cavity) with $\mathit{VF}=1$. An example of systems which could have this geometry is shown in Extended Data Fig. 1. The W emitter results in a higher efficiency because the selective emissivity properties of W suppress some of the below-bandgap energy. Additionally, the W emitter causes the peak in efficiency to shift to lower temperature because the emissivity of W weights the spectrum towards shorter wavelengths. The blackbody emitter results in a lower efficiency because the high irradiance causes a larger penalty of series resistance loss due to the high current density. The comparison shows that the efficiency measured under the lightbulb spectrum in this work provides an appropriate and relevant characterization for TPV efficiency in a real TPV sub-system. In all cases, the cell temperature is 25 °C.

Extended Data Fig. 6. Cell temperature and parasitic heat flows.

a) Cell temperature vs emitter temperature. The cell temperature increases with emitter temperature due to the heat flux sensor which undesirably impedes heat flow. b) Schematic (not to scale) showing parasitic heat flows in the experiment. c) Calculated parasitic heat flows for the 1.4/1.2-eV device. A positive value would act to increase the measured heat flow and reduce the measured efficiency, while a negative value would have the opposite effect. d) Comparison of the efficiency measurement (solid circles) to the measurement with the addition of the modeled parasitic heat flows (open circles) for both tandems.

Extended Data Fig. 7. Experimental setup.

a) Schematic of the concentrator setup showing the relative placement of the ellipsoidal and compound parabolic reflectors, water-cooled aperture, TPV cell, HFS, and heat sink. b) Image of the concentrator setup. c) Schematic of the heat and electricity flows through the measurement device. Electric power is extracted by two copper clips which interface with the cell bus bars on the top surface of the cell and are thermally and electrically insulated from the heat sink. d) Image of the cell on the heatsink with electrical leads. The aperture was removed for clarity.

Extended Data Fig. 8. J_sc, V_oc, and FF.

Modeled vs measured a) $J_{\mathit{sc}}$, b) $V_{\mathit{oc}}$, and c) $\mathit{FF}$. Good agreement can be seen between the measurement and model predictions. For each device, the $\mathit{FF}$ measurement and model exhibit the same trend and the minimum in $\mathit{FF}$ for 1.2/1.0-eV agrees well between the model and measurement which suggests good calibration of the emitter temperature.

Extended Data Fig. 9. Electric power modeling.

a) Measurements of $V_{\mathit{oc}}$ and $\mathit{FF}$ vs $J_{\mathit{sc}}$ for the 1.2/1.0-eV device under the high irradiance flash simulator over a wide range of irradiances, but fixed spectrum and fixed cell temperature at 25 °C. A model was fit to the data using the three fitting parameters to determine the cell characteristics. The measurement over a wide irradiance range is critical to extract the $R_{\mathit{series}}$ parameter under the high-irradiance conditions of interest. b) Low irradiance measurements of $V_{\mathit{oc}}$ and $\mathit{FF}$ under a continuous 1 sun simulator in which the spectral content could be varied to produce photocurrent ratios of the two junctions corresponding to different emitter temperatures. Cell temperature was fixed at 25 °C. The model was determined using the cell characteristic parameters which were extracted from fitting to the data over a wide range of irradiances. The good agreement suggests that the model can be used to predict $V_{\mathit{oc}}$, $J_{\mathit{sc}}$, $\mathit{FF}$ over a wide range of conditions (irradiance and spectra). c) Modeled cell performance parameters under the measured spectra showing a comparison between results for a 25 °C cell temperature and the measured cell temperature.