Power/Performance Bits: Nov. 26

Magnets hold the key to reducing power consumption and getting heat out of chips. Together, perhaps, we will have a whole new breed of circuits…


Many people are predicting that power will be the issue that brings Moore’s Law to an end. Power creates heat and that heat can be destructive to chips, so there are two paths forward – the first is to reduce heat and the second is to get it off chip. It seems as if magnets may be the common key to both approaches.

Magnetic Transistors
New work by researchers at UC Berkeley soon could transform the building blocks of modern electronics by making nanomagnetic switches, a viable replacement for the conventional transistors found in all computers.

As current passes through a strip of tantalum, electrons with opposite spins separate. Researchers used the resulting polarization to create a nanomagnetic switch that could one day replace computer transistors. (Image by Debanjan Bhowmik, UC Berkeley)

Semiconductor-based transistors have been consuming greater chunks of power at increasingly hotter temperatures as processing speeds grow. For more than a decade, researchers have been pursuing magnets as an alternative to transistors because they require far less energy when switching. However, until now, the power needed to generate the magnetic field to orient the magnets so they can easily clock on and off has negated much of the energy savings that would have been gained by moving away from transistors.
UC Berkeley researchers overcame this limitation by exploiting the special properties of the rare, heavy metal tantalum.
The researchers created a so-called Spin Hall effect by using nanomagnets placed on top of tantalum wire and then sending a current through the metal. Electrons in the current will randomly spin in either a clockwise or counterclockwise direction. When the current is sent through tantalum’s atomic core, the metal’s physical properties naturally sort the electrons to opposing sides based on their direction of spin. This creates the polarization researchers exploited to switch magnets in a logic circuit without the need for a magnetic field.

“This is a breakthrough in the push for low-powered computing,” said study principal investigator Sayeef Salahuddin, UC Berkeley assistant professor of electrical engineering and computer sciences. “The power consumption we are seeing is up to 10,000 times lower than state-of-the-art schemes for nanomagnetic computing. Our experiments are the proof of concept that magnets could one day be a realistic replacement for transistors.”

Getting Heat Out Of Chips
Researchers at MIT and in Australia may have found a low power way to increase cooling by improving a heat-transfer mechanism.

The system utilizes a slurry of tiny particles of magnetite, a form of iron oxide. The magnetite nanofluid flowed through tubes and was manipulated by magnets placed on the outside of the tubes.

According to Lin-Wen Hu, associate director of MIT’s Nuclear Reactor Laboratory, the magnets “attract the particles closer to the heated surface” of the tube, greatly enhancing the transfer of heat from the fluid, through the walls of the tube, and into the outside air. Without the magnets in place, the fluid behaves just like water, with no change in its cooling properties. But with the magnets, the heat transfer coefficient is higher, she says — in the best case, about 300% better than with plain water.

“We were very surprised” by the magnitude of the improvement, Hu says.

Conventional methods to increase heat transfer in cooling systems employ features such as fins and grooves on the surfaces of the pipes, increasing their surface area. That provides some improvement in heat transfer but not nearly as much as the magnetic particles. Also, fabrication of these features can be expensive.

Hu explains that the explanation for the improvement in the new system is that the magnetic field tends to cause the particles to clump together — possibly forming a chainlike structure on the side of the tube closest to the magnet, disrupting the flow there, and increasing the local temperature gradient.

While the idea has been suggested before, it had never been proved in action, Hu says. “This is the first work we know of that demonstrates this experimentally,” she says.

There could be numerous applications for such a system, Jacopo Buongiorno, associate professor of nuclear science and engineering at MIT, says: “You can think of other systems that require not necessarily systemwide cooling, but localized cooling.” For example, microchips and other electronic systems may have areas that are subject to strong heating. New devices such as “lab on a chip” microsystems could also benefit from such selective cooling, he says.

Leave a Reply

(Note: This name will be displayed publicly)