Superlattices Chill Hot Processors

Tiny heat pumps offer rapid cooling for electronics, fiber optics.

Researchers have developed highly efficient thermal transfer devices that can cool or heat an area thousands of times faster than existing methods. An alloy-based substance can be deposited in microscopic layers on hot spots in electronics or next-generation fiber optic switches to improve their efficiency. The technology also makes possible the creation of tiny, localized heat sources for use in biochemistry, laboratory-on-a-chip systems, and mobile power sources for soldiers.

Excess heat has troubled electrical engineers for decades. Uncontrolled overheating may lead to chain reactions that destroy microchips and other sensitive electronics. Although heat sinks and other types of cooling systems prevent these catastrophes, protection often comes at the cost of added weight, space and bulk. Efficient miniaturized heat sinks save valuable component space and can be placed directly on the appropriate area. Such thermal control systems have applications in fiber optic switches and other technologies that rely on temperature changes to move components or alter physical states within a device.

Heat sink technology is an outgrowth of U.S. government research to develop better thermoelectric materials, which are substances that generate an electrical current when two conductive materials of differing temperatures are in contact with each other. According to Rama Venkatasubramanian, senior program director at the Research Triangle Institute (RTI), Research Triangle Park, North Carolina, studies began in 1993 with funding from the Defense Advanced Research Projects Agency (DARPA) and the Office of Naval Research (ONR), Arlington, Virginia. RTI has developed a material that is 2.4 times more efficient and responds 23,000 times faster than existing thermoelectric substances. Devices made from this material provide a variety of functions such as heating, cooling, precise temperature control and converting heat into electricity. 

The substance combines two alloys, bismuth telluride and antimonic telluride, which are laid down in alternating layers via a thin film deposition process similar to that used in microchip construction. Venkatasubramanian describes the resulting structure as a superlattice that disrupts the flow of phonons—heat or sound vibrations—between the alloy layers. Differences between the two substances cause this blocking effect.

Because thermal vibrations cannot easily move along the lattice, electron flow is actually enhanced. Venkatasubramanian explains that the great advantage of this technology is that it allows electrons—current—to flow without transmitting thermal vibrations in the form of phonons. By passing an electric current through the interface formed by superlattice layers, the interface grows hotter or colder, depending on the direction of the current. Heating a thermoelectric interface also causes electrical current to flow through the device, generating power.

The alloy’s properties and the thin film deposition process create a more efficient cooling process. Current thermoelectric systems use pellet-like elements that are one millimeter thick. These components take roughly two to three seconds to cool electronic devices such as microchips. “Instead of a millimeter, we are doing this in as few as five microns. Now cooling or heating is almost instantaneous—10 microseconds,” Venkatasubramanian says.

Such high-speed cooling can prevent events like thermal runaway, caused when a microchip’s hot spots continue to absorb more heat in a cascade effect until the entire device fails catastrophically. The thin film deposition process allows RTI’s heat sinks to pump more heat per unit area than larger, bulk thermoelectric devices.

Venkatasubramanian notes that while existing bulk systems can pump one to two watts of heat per square centimeter, RTI’s technology has a power density that pumps up to 700 watts in the same volume. “That doesn’t mean we can pump 700 watts of heat within a centimeter squared. What it means is the power density can be as high as that. Suddenly, if we can pack these devices densely within a centimeter squared, we can pump a lot of heat,” he says.

Advanced microprocessors such as Pentium chips generate 50 to 60 watts of heat. This is not distributed uniformly, but scattered across the device in localized hot spots. Temperature fluctuations in these areas can be significant, with future chips expected to experience heat fluctuations ranging between 10 to 100 watts per square centimeter. These potential temperature variations create many opportunities for the application of small heat sinks, Venkatasubramanian says.

Because thermoelectric devices also can generate energy, researchers are investigating their use as portable power sources for individual soldiers. DARPA’s Palm Power program is exploring the potential for these systems to replace or augment batteries in various types of portable communications and sensor equipment.

The U.S. Army also is exploring ways to build thermoelectric cooling systems into soldiers’ uniforms. “If we are successful, we can integrate these lightweight, thin film devices into a cooling jacket. It won’t add much weight or create the vibration and noise of a typical small mechanical refrigeration system,” Venkatasubramanian says.

Power for the coolers would come from batteries or some other power source carried on soldiers’ belts. By achieving an efficient system, less battery power would be needed for the unit. It is also possible that some of the nearly 100 watts of waste heat produced by a soldier could be used to supplement some of the equipment’s power requirements, he speculates.

Another application is for switches used in fiber optic trunk lines. These devices change the direction of laser light by using heaters to move a bubble of liquid. By replacing a switch’s temperature heating control with a single solid-state heating and cooling device, RTI scientists predict that operating speeds may increase by 100 percent. This also may increase efficiency and reduce the cost of regenerating optical signals over long distances. Enhanced cooling permits switches to operate at lower voltages and enables engineers to pack more of them into smaller spaces. “When you enhance this efficiency, you can have repeaters every 100 kilometers instead of every 10 kilometers. Or you can actually switch light from one fiber to another using switches. There are any number of applications,” Venkatasubramanian explains.

Thermoelectric technology also has potential applications in high-temperature superconductors. One of the original goals of RTI’s research was to develop solid-state cryogenic coolers. Current thermoelectric materials used in superconducting usually operate at a temperature of 160 Kelvin or below because their efficiency decreases dramatically at higher temperatures.

“We believe that with the superlattices, assuming we can make a six-stage cooler and work out all the issues, it should be theoretically possible to achieve cascade cooling to cryogenic temperatures of 100 Kelvin or so,” Venkatasubramanian says. Operating in the 100-Kelvin area puts the device in the realm of high-temperature superconductors. Although superconductivity research is a long-term goal, he notes that RTI is not currently working in this area because a module capable of operating at room temperature must be built and successfully tested before any additional work can proceed.

The precise temperature control achieved by RTI’s thermoelectric alloy superlattices creates potential applications in chip-based chemistry and biotechnology. An example of the latter involves using controlled heating and cooling to cause deoxyribonucleic acid (DNA) microarrays to self-assemble. This approach could replace the existing time-consuming process of individually constructing the nodes with robotic tools. Other biotechnology applications include using microscopic thermoelectric devices to create new proteins and enzymes for use as pharmaceuticals by temperature control to manipulate the binding of ribonucleic acid polymerase with DNA.

Despite their advanced characteristics, the bismuth telluride and antimonic telluride alloys have been in use for decades, Venkatasubramanian explains. Thermoelectric technology had been stagnant for nearly 40 years before the government backed RTI’s efforts in the early 1990s. He notes that initial research by the U.S. Navy in the 1950s was promising, and the service invested roughly $60 million—the equivalent of $2 billion in 2002—into thermoelectric research. But the technology did not progress beyond the discovery of the alloys because material processes such as thin film deposition did not exist and the concept of superlattices only emerged in the 1970s and 1980s. This inability to fully exploit the alloy’s potential meant that Navy researchers could only use the materials in bulk quantities as heat sinks, he explains.

Supported by DARPA and ONR funding, the institute has developed several prototype devices to prove the technology works, Venkatasubramanian says. One technology demonstrator built by RTI scientists allows a user to observe a temperature drop or increase when he or she presses a switch or reverses the current. The prototype can generate a cooling of 32 degrees Celsius at room temperature.

Beyond this example, the institute is moving quickly to develop the technology for use in existing devices. Solid-state cooling/heating systems for fiber optics and wafer-scale cooling for microchips are the alloy’s most immediate applications. Venkatasubramanian predicts that RTI may have a product ready within a year.

Because RTI is a nonprofit research organization, it is spinning off a company to make commercial prototypes. This new firm will conduct prototype fabrication runs and develop a manufacturing plan for producing large volumes of thermoelectric devices. He notes that some 35 companies are interested in the technology.

Additional information on thermoelectrics is available on the World Wide Web at www.rti.org