Solid-state devices that convert heat to energy have a reputation for being inefficient. New materials may eventually change the equation.
Leland Teschler • Executive Editor
Estimates are that 61% of the energy consumed in the U.S. is lost to heat. No wonder, then, that there is interest in finding ways to harvest some of these losses and convert waste heat into something useful One of the methods increasingly under investigation is in thermoelectric materials, materials that generate electricity from a heat differential.
The classic example of a thermoelectric generator is a Peltier module. Though they are usually employed as solid-state cooling devices, Peltier modules also can work as generators. Here, one side of the device is heated to a temperature greater than the other side. Because of the Seebeck effect, a difference in voltage will build up between the two sides.
But several issues limit the scenarios in which Peltier thermogenerators can make sense. One is that the typical efficiency of TEGs is only around 5–8%. Modern devices use highly doped semiconductors made from bismuth telluride (Bi2Te3), lead telluride (PbTe), calcium manganese oxide (Ca2Mn3O8), or combinations of them depending on temperature. Several of these materials can be pricey. Finally, it generally takes a high temperature for a Peltier module to generate much electricity. Alloys based on bismuth and antimony, tellurium or selenium are considered low-temperature thermoelectrics but like to see temperatures above 300°F. Thermoelectrics based on lead alloys handle temperatures up to about 1,000°F, and silicon-germanium thermoelectrics are for temperatures up to about 1,800°F. Consequently, Peltier thermogenerators tend to find use only for low-power remote applications.
However, there is great interest in devising thermoelectric devices that can work at lower temperatures and which convert heat to electricity more efficiently. Research is progressing in two main areas: materials that generate electricity at lower temperatures, and device structures that convert the infrared radiation given off by hot bodies into electrical current.
To understand the direction of this work, it pays to know the figure of merit used for thermoelectric materials, often given as ZT=S2σT/κ. Here, S is the Seebeck coefficient, σ is electrical conductivity, T is working temperature, and κ is thermal conductivity. Researchers at Beihang University in China say its tough to improve the ZT because making making improvements in one of the parameters tends to make one or more of the others go in the wrong direction. Consequently, many of the strategies for improving ZT so far only work in a narrow range of temperatures.
One factor limiting thermoelectric performance is the bandgap, i.e. the discrete energies of the thermoelectric material’s electrons. The bandgap is given by Es=2eSmaxT where e is unit charge, Smax is the maximum Seebeck coefficient, and T is the temperature corresponding Smax. The Seebeck coefficient basically measures the voltage produced with a temperature gradient (S=ΔV/ΔT).
To get a thermoelectric material that works over a span of several hundred degrees, the usual approach is to use several materials that all have narrow bandgaps or one material with a wide bandgap. There are practical problems with material mismatches in thermoelectrics that employ several narrow-bandgap materials, so the more typical approach currently is to use wide bandgap materials such as tin selenide (SnSe) whose bandgap energy is about 0.86 eV. Beihang researchers report seeing a thermoelectric effect in SnSe that spans the 80 to 980ºF range.
However, materials with wide bandgaps also have another problem that can limit their utility as thermoelectrics: They tend to have low carrier densities, i.e. too few charge carriers available to support significant electrical current flow. The approach used to solve the problem is to set up the orientation of the SnSe crystalline material in a layered fashion that makes more carriers available.
Researchers at Beihang University say they’ve used this approach to uncover several promising thermoelectric materials, including BiCuSeO, BiSbSe3, K2Bi8Se13, and Sb2Si2Te6. But they warn that it can be challenging to turn material having a high ZT value into a commercial device, particularly one that can work at high temperatures. One issue: The resistivity of the material used for electrical contact can grow over time, particularly in the presence of high temperatures.
Harvesting thermal light
Objects at a given temperature radiate heat according to their surface temperature. The sun, for example, has a surface temperature of 6,050°C. Photovoltaic cells convert this radiant energy to electricity.
Of course, most terrestrial sources are a lot cooler than the sun. From Wien’s law, as the temperature of a black-body source drops, the wavelength at its peak power rises such that source temperatures between 100 and 400°C have a spectrum in the thermal infrared range (7 to 12 μm wavelengths). Estimates are that more than 95% of waste heat generated in the U.S. is below 400°C (752°F).
Problem is, ordinary photovoltaic cells don’t efficiently convert this kind of light to electricity. A photovoltaic cell is basically a p-n diode where collected photons create what is basically a reverse current for the diode. But the ability of a photovoltaic cell to generate electricity depends on the band gap of its material; the photovoltaic effect doesn’t occur if the energy of the absorbed light is lower than the bandgap energy of the (typically) silicon photodiode. Silicon at room temperature has a band gap energy of 1.12 eV and a cutoff wavelength of 1.1 μm.
To create a photodiode able to better detect mid-range IR light wavelengths, one approach is to rectify IR using a special type of ultra-fast diode structure called a tunnel-junction diode. Rather than create charge carriers from photons, as with ordinary photovoltaic cells, it rectifies light waves via tunneling in a manner analogous to the way high-speed diodes rectify radio waves.
A tunnel diode is characterized by heavy doping to a point where the Fermi level of the diode P-type material lies below the valence band, and the Fermi level of its N-type material lies above the conduction band. The quantum mechanics of this setup are complicated, but the point of this structure is that it creates a current flow via quantum tunneling through the P-N junction. (It also has a region in its IV characteristics where it displays negative resistance: When the voltage increases the current through the tunnel-junction diode decreases.)
Most of the heat-to-energy devices created thus far work best at temperatures above 1,000°C. Creating devices that work well at temperatures below this level is proving to be tough. One reason is that there are fewer photons to work with than at higher temperature extremes.
Nevertheless, there are promising developments taking place in lower-temperature thermovoltaic devices. One device created recently at Sandia National Labs is called a bipolar MOS tunnel-junction diode. The device uses an optical grating to couple light into a small (3-4 nm) area SiO2 barrier which results in a concentrated electromagnetic field that drives photon-assisted tunneling of electrons from a doped-P type silicon to the N-type silicon part.
The Sandia device uses what’s called photon-assisted tunneling where a photon gets absorbed in an occupied state near the Fermi level of the metal gate, followed by field-enhanced tunneling into an unoccupied state of the silicon. The result is a small direct photocurrent. A similar time-reversed process happens in the semiconductor that causes a back flow current into the metal gate. The overall direct current is due to the difference between these two currents, which arises from the difference in effective mass for electrons in the metal and in the semiconductor.
Sandia researchers devised a special circuit structure to make use of the photon-assisted tunneling effect. They use an interdigitated bipolar P-N junction array under the tunneling gate electrode which acts as a charge pump moving electrons from the P-type region to the N-type well.
The amount of power this experimental device generates is small. Researchers say they saw a peak power density of 27 μW/cm2 for heat sources of 250 and 400°C and 61 μW/cm2 for 350°C. The open-circuit voltages produced are in the range of a few millivolts. Researchers also say they can adjust the temperature at which the device is most efficient by changing the thicknesses of the semiconducting and metal layers involved.
The energy conversion efficiency of this setup is modest. Researchers say it is 0.4%, but there are ways of improving it using slightly different gate dielectrics, and heat collection designs. Also, the experimental devices were fabricated on a CMOS platform which may make it possible to eventually scale up for mass production. DW