Thermophotovoltaic (TPV) Power Conversion

The objective of several research programs at PSU over the last five years has been to develop a portable, energy-dense power source that could replace batteries or fuel cells in various environments. These specifically included powering cardiac assist devices, funded by NIH, and, most recently, general portable power requirements, funded by DOE.  For each of these programs, Dr. Robert Backes has been the Principal or Senior Investigator at PSU, while I collaborated.

The basic premise of the intended technology is to convert heat from various possible combustibles (compressed gases, liquid fuels) directly into electricity, using specially adapted solid state photovoltaic cells as the heat-to-electricity conversion medium.  Such thermophotovoltaic (TPV) cells are solar cells tailored to operate with blackbody radiators much cooler than the sun. A crude way to describe this is a lantern coupled to photocells.

A common approach to TPV conversion is illustrated (very schematically) in Figure 1.  This omits various ways to recirculate exhaust heat or to preheat fuel to reach higher combustion temperatures, which are needed for sufficiently intense emission in the part of the spectrum needed for efficient operation of the photocell.

One key to making such devices efficient in energy conversion is management of the optical path between radiator emitter and photovoltaic conversion device. In the approach shown, a blackbody emitter illuminates a filter(s), which transmits energy above the TPV cell’s conversion threshold (bandgap).  The rest are reflected back to the emitter to help maintain emitter temperature.  Energy passed by the filter is then absorbed and converted to electricity.

One potential simplification might be to eliminate the filter by incorporating its band-pass functions directly into the emitter. Such selective emitters can be produced by doping with certain rare earth elements or by micro-texturing the surface. Unfortunately, these techniques are either unstable at high temperatures or limited in selectivity. We concluded that both a filter and selective emitter were needed to achieve the required efficiency.

Figure 1.  General configuration of thermophotovoltaic energy converters.

The fraction of energy absorbed by the photocell and converted to electricity can theoretically be quite large, more than 50%, when the emission spectrum is tailored to match the cell’s optimum conversion band, the region just above its band-gap energy.  Photons outside this band can be recycled (reflected) to the emitter to help maintain its operating temperature.

Overall system efficiency depends critically on radiator temperature and emission characteristics, as well as on cell coating, temperature, and bandgap while minimizing parasitic absorption losses.  However, even with remaining losses, a well-designed thermophotovoltaic system should yield an overall fuel-to-electricity conversion efficiency much higher than (for example) the one-to-two percent typical of a storage battery.

Silicon was the photocell material most widely used in early experimental TPV systems. Although readily available, its relatively high bandgap dictated emitter temperatures in the 2300 oC range for significant efficiency to be possible.1,2
Emitter stability has generally restricted heat sources of experimental systems to about 1400 oC.  This in turn requires photocells with a bandgap less than half that of silicon.

Cells exploiting binary or ternary material systems offer bandgaps in the necessary range, and the technology of these low bandgap cells is relatively far advanced.  Indium-gallium-arsenide (InGaAs), gallium-antimonide (GaSb), and mercury-cadmium-telluride (HgCdTe) converters having bandgaps in the 0.50 to 0.75 eV range (1.6 to 2.5 microns) are all available from commercial suppliers.  The key to improved TPV systems is therefore not the cell, but the radiator, which must provide photon energies closely matching the cell’s absorption spectrum, and do so reliably and stably for long periods of unattended operation.

Recent TPV configurations include emitters coated with ceramic materials, such as the rare earth oxides, radiating line or narrow band spectra.3-5  The drawback to the rare earths, however, is their mobility at high temperatures; they diffuse and sublime and subsequently condense on filters, reflectors, photocells, or other components, greatly reducing their effectiveness and limiting practical lifetimes of systems based on them.  Also, rare earths contribute emission lines well outside the desired band.  To be really effective, they will still require the addition of filters.

Ion Optics and PSU have worked together to overcome the disadvantages of rare earth ceramic emitters by microscopically texturing a refractory metal emitter supported by a ceramic burner element.  Texturing produces tiny features that act as antennae selectively tuned to the wavelength required to match photocell bandgap energy.  Outside the desired energy band, the emitter is reflective; inside the band it is highly emissive.

To operate efficiently with combustion sources, the emitter must have both a high melting point (above 1,500 oC) and a low vapor pressure, allowing it to function for prolonged periods without damaging the rest of the system.  While a textured metal emitter will not produce a perfect spectrum by itself, modeling its emission in combination with the transmission characteristics of a single optical grid filter7,8 shows the emitter-filter pair produces much greater efficiency than any rare earth emitter configuration.

TPV cells currently have an unexpectedly high price compared to silicon ($2,500 per 5 cm2 of GaSb, JX Crystal, Inc., more for other types), and this led to further modifications of the original designs to include a study of concentrating chamber designs for the model TPV system to minimize cell area, particularly parabolic and elliptical chambers and concentrators.  An important consideration for these concentrators is uniform illumination. Otherwise, one part of a cell or array effectively works against another. 

Ray tracing and Monte Carlo performance models were developed for each chamber type.  From these models it was concluded that an elliptical chamber could focus on a linear 5 cm2 array of cells up to 91% of the emitted energy, as opposed to 41% for a parabolic chamber, and 30% for the originally proposed cylindrical chamber.9

The key problem for this type of TPV system is matching the emission spectrum of the radiator to the operating waveband of the photovoltaic cell.  For a radiator which produces an ideal blackbody spectrum, a large fraction of the energy emitted will be in the long wavelength tail where it cannot contribute useful energy to the system.  This can be corrected by adjusting emitter emissivity such that it is high at short wavelengths and low in the “tail,” thus combining the function of the radiator and that of a reflective micromesh filter in a single component.

Ion beam texturing was first described in 1942, before it could be imaged by electron microscopy, inferred from angular changes in reflectivity of glow-discharge cathodes.11  Since then, many researchers have studied ion texturing processes and the physical mechanisms involved have become relatively clear.12  Energetic (keV) ions striking a solid surface collide with target atoms and transfer enough kinetic energy to break atomic bonds in the outer surface layers, sputtering away surface atoms.  Efficiency of the sputtering process depends mostly on kinematic factors, specifically mass and energy of projectile ions and mass of target atoms.  Typically, heavier projectile ions sputter more efficiently, and lighter target atoms are removed more easily than heavy atoms.  However, target atoms which form strong chemical bonds to their nearest neighbor atoms are more difficult to remove.  For each projectile ion/target atom combination, there is an optimum sputtering energy related to the relative speed of the ion and the strength of chemical bonding in the solid.

The several constituents of alloyed targets materials usually sputter at different rates, as do impurities on the surface or in the body of the target.  If sputtering removes an impurity more slowly than it does the host, impurity islands shield the underlying target material from ion impact, producing impurity-capped “pillars” on the resulting surface.  This has been observed in many classes of materials: metals, ceramics, and polymers.13  Artificially depositing “seed” impurities on the surface, instead of relying upon naturally-occurring differential erosion, enhances this texture.  If impurities are not randomly deposited but deliberately placed in regular patterns by, say, a lithographic process, the texturing effect can be intensified by superposing it over the larger structure.14

It has been reported that, for natural texturing, surface temperature impacts feature size, apparently through diffusion of impurities away from islands;15 residual gasses in the vacuum chamber during ion bombardment also appear to affect feature size. Control of these parameters, together with those described above, now enables deliberate control of feature size and density.

In previous research, Ion Optics has carried out an extensive matrix of experiments on a number of materials (copper, titanium, stainless steel), varying texturing parameters in a search of a well-defined process producing an emissivity profile optimized for GaSb photocells.  The ideal wavelength-dependent emissivity would track GaSb‘s efficiency curve, requiring a reflector for wavelengths longer than the (1.65 micron) bandgap, where photons lack sufficient energy to participate in the conversion process, and a highly emissive (absorbing) surface for shorter wavelengths.


1. Paul Scherrer Institut,
2. SBIR Proposal, “Photonic Bandgap Surface Structures for Mechanically Rugged Thermophotovoltaic Emitters”, submitted by IOI, Feb 28, 2000.
4. Stem, Chemicals for Research, Cat # 18, 1999-2001.
5. Alfa Aesar; Inorganics, Organics, Metals and Materials for Research; 2001-02 Catalog.
6. Paul Scherrer Institut,
7. B. Bitnar, et al, “Critical Components for TPV Applications”, PSI Annual Report, 1999
8. E. Yablonovitch and T.J. Gmitter, Phys. Rev. Lett., 63, 1950 (1989).
9. E. Yablonovitch and T.J. Gmitter, Phys. Rev. Lett., 67, 3380 (1991).
10. J.Y. Decker, etal, J. Vac. Sci. Technol., B 15(6), p 1949 (1997).
11. E-mail communications from IOI to Pittsburg State University on Sept 28, 2000.
12. M. U. Pralle, Internal Report, IOI, TECHNICAL MEMORANDUM – ATP PROJECT, Jan 3, 2001
13. C. J. Crowley and N.A. Elkouh, “Plasma-Sprayed Selective Emitters for Thermophotovoltaics”, Creare, Inc. for Glenn Research Center, Date (?)
14. J. Ferber, et al, “Microstructured Tungsten Surfaces as Selective Emitters in Thermophotovoltaic Systems”, Frauhofer Institute for Solar Energy Systems ISE, Proceeding of the 16th European Photovoltaic Solar Energy Conference and Exhibition, Glasgow 2000