Silicon Nanowires Turn Heat to Electricity

11 January 2008--Silicon nanowires may lead the way to converting waste heat into electricity, according to research reported yesterday in the journal Nature.

Two separate teams, one at Caltech and the other at the University of California, Berkeley, reported that they could increase silicon's ability to convert heat into electric current by as much as 100 times. If they can use what they've learned to improve silicon even further, or translate their findings to other materials, the discovery could lead to new ways to cool computer chips, build refrigerators, or get more power out of car engines.

Thermoelectric conversion relies on a difference between hot and cold areas in a device. Heat flowing from the hot side to the cold side creates current, which can be captured and used to power a device or stored for subsequent use. Bulk silicon has traditionally been considered a poor material for thermoelectric conversion, because its thermal conductivity is too high; heat travels across it so well that it's difficult to create the necessary temperature differential.

”If you were going to make a high-performance thermoelectric, you would never use silicon, because as a bulk material it's pretty lousy,” says James Heath, a chemist who led the research at Caltech. He was surprised by his own results; he expected some increase in efficiency, but not as much as he got.

Thermoelectric conversion efficiency is measured by a number dubbed ZT. Several factors go into that number, and it can be increased both by lowering the thermal conductivity of a material and by increasing its electrical conductivity. Whereas bulk silicon at room temperature has a ZT of 0.01, the Berkeley team increased that to 0.4, and the Caltech team increased it to 0.6. That puts silicon nanowires about on par with bismuth telluride, the compound from which commercial converters are made despite the fact that it is relatively expensive and challenging to work with. Making thermoelectric devices out of silicon, which is abundant, cheap, and easily handled, could help create a new market for the devices.

Both research teams found that they could decrease silicon's thermal conductivity--and therefore increase the conversion efficiency--by fashioning the material into nanowires with diameters of 10 to 100 nanometers and introducing defects in the silicon that slowed the flow of phonons--the acoustic vibrations in the crystal lattice of a material that carry heat.

”Defects are important here,” says Peidong Yang, a materials scientist at Berkeley. ”They can block the phonon transport from one end to the other end, so the thermal conductivity can be drastically reduced.”

Yang says his group engineered defects into the nanowires at three different length scales. First, by fashioning the bulk silicon into nanowires, they made the material very small compared with the phonons so that the size of the wires themselves affected how the phonons could move. They also made the surface of the wires rough, introducing a set of defects at a smaller scale. Finally, they doped the silicon with boron to introduce defects at an atomic level.

Heath induced a greater drop in thermal conductivity by making his nanowires even smaller than Yang's--only 10 to 20 nm in diameter. Normally, a wire would carry two types of phonons, he explains: one that causes the wire's diameter to expand or contract, and one that causes it to lengthen or shorten. Like a rubber band that gets thinner when stretched, the two work in opposition. But when the nanowires get small enough, the two types merge into a single type of phonon, and that slows down the heat transport even more.

Unfortunately, when Heath made the wires 10 nm wide, which gave him the best results for thermal conductivity, the electrical conductivity crucial to thermoelectric conversion also dropped.

Mildred Dresselhaus, a physicist at MIT who had predicted that using nanowires would lead to better thermoelectric conversion, says she's pleased with Yang's and Heath's research. Their reports ”represent a significant advance in the field,” she says. ”The applications field is now taking off, and interest in the field by the science community has grown a lot in the last two to three years.”

One of the easiest applications would be for recycling waste heat from computer chips into electricity. ”You gain twice,” says Heath. ”Number one, you're getting rid of heat, which is bad in a laptop, and number two, you're gaining efficiency.” He thinks that applications could come with just a couple of years' work.

Even better, he says, would be making refrigerators based on thermoelectric conversion. Instead of taking heat and converting it into electricity, the same system would use electricity to cool the material. Such thermoelectric coolers are already used to chill CCD detectors in infrared cameras and are found in high-tech picnic baskets and in the seats of luxury cars as well. Thinking on a larger scale, Yang would like to see systems that convert the waste heat from car engines or power plants.

Both teams are pressing ahead to see what they can achieve next. The researchers believe a material with a ZT of 3 or 4 would be very appealing commercially. Heath hopes to apply his findings to other materials that might start out with better properties than silicon and be improved further. He's doing work with silicon germanium, for instance, which has much lower thermal conductivity than pure silicon.

Yang, too, is looking at new materials, but he also thinks he'll get improvement using silicon at higher temperatures, like those in, say, a car engine.

Cronin Vining, a consultant on thermoelectrics, says the commercial market for thermoelectric devices is very small at present but could grow with better materials. He says the nanowire work is impressive, but he's not ready to say that thermoelectrics could, for instance, help stem global warming by increasing the efficiency of power plants. ”As they stand, their properties are not really good enough to be useful,” Vining says. ”But this is the very first result on silicon in 60 years that's of any interest at all.”