Thermal effects in spintronics systematically evaluated for the first time

Spintronics, or devices that use microscopic magnetism in conjunction with electric current, could lead to computing technology that is as fast as classical electronics, but much more energy efficient. While such devices are being developed and studied, an important unresolved question is how the operation of these devices is affected by heat.

A new experimental technique, reported by researchers at the University of Illinois at Urbana-Champaign in the journal APL materialsdirectly measures heating in spintronic devices, allowing direct comparison with other effects. The researchers say this technique can be used to select spintronic materials whose magnetic behavior is little affected by heating, leading to faster devices.

“Spintronic devices rely on the ability to change magnetization using electric currents, but there are two possible explanations for this: electromagnetic interactions with the current, or the temperature increase caused by the current,” says Axel Hoffmann, project leader and professor of materials science and engineering at Illinois. “If you want to optimize the operation of the device, you need to understand the underlying physics. That’s what our approach allowed us to do.”

Unlike electronics, which use electrical signals to store information and perform computations, spintronics exploits a fundamental property of electrons called spin, which results in microscopic magnetic behavior. These devices have the potential to consume much less energy than their electronic counterparts, due to the magnetic nature of their operation. It has even been suggested that spintronics controlled by fast electronics could remain energy-efficient while matching the speed of conventional computers. “It’s like getting the best of both worlds,” Hoffmann says.

The challenge has been to find suitable materials for such devices. Antiferromagnets have attracted attention because of their periodic arrangements of opposite spins and limited sensitivity to neighboring devices. To use these materials for memory and computing, the spin structure must be controlled by an electric current. The currents required to do this are so large that the temperature of the devices increases to the point where thermal effects impact the spin structure in addition to electromagnetic effects.

“There is a debate now about whether the current is directly responsible for the spin changes or whether the resulting heating has the dominant effect,” Hoffmann said. “If it is a current-induced effect, it is very easy to make the effect very fast. If it is a thermally induced effect, then thermal conductance and thermal relaxation are important and can limit how fast you can operate the device. So the exact functionality of the device depends on the physics responsible.”

Efforts to date to clarify the importance of current- and temperature-induced effects have been hampered by the inability to directly measure heating effects in small-scale devices. Myoung-Woo Yoo, a postdoctoral researcher in Hoffmann’s group, demonstrated an experimental method in which thermal effects are inferred from how a device heats substrates with different thermal conductivities.

“We prepared antiferromagnetic samples on silicon dioxide substrates of different thicknesses,” Yoo explained. “The substrate’s ability to conduct heat decreases with thickness, which means that antiferromagnets on thicker samples have higher temperatures when the same electric current is applied. If device heating is important for spin structure changes, then there will be a difference between devices on different substrates.”

The researchers found that heating had a significant effect on the antiferromagnet they studied, Mn₃However, they noted that there are many other antiferromagnets under study for spintronics, and this technique provides a framework for systematically comparing the role of heating to that of electric current effects.

“We now have a well-defined strategy to evaluate the influence of electrical heating in spintronic devices,” Yoo said. “Moreover, it is very easy to implement in very general terms, so it can be applied to any system, including standard electronics. This methodology can be used to optimize the functionality of any type of microscopic device.”