Left, a conventional spintronic device structure. Right, the new method proposed in the study. Credit: Korea Institute of Science and Technology (KIST)
Scientists have found a way to improve the efficiency of spintronic devices which are a key foundation of next-generation computing such as ultra-low-power memory and neuromorphic chips.
The newly-discovered physical phenomenon allows magnetic materials to spontaneously change the direction of their internal magnetisation by harnessing the loss of electron ‘spin’ as a natural source of energy.
In traditional electronics, computers perform calculations and transmit and store information by sending pulses of electricity. The presence of a pulse is recognised as a ‘1’ and its absence as a ‘0’.
Magnetic hard drives work a little differently. Tiny regions on a flat magnetic disk record 1s and 0s through the material’s possible magnetic orientations: up and down.
Spintronics combines both approaches by exploiting the charge of an electron, which is always negative, and its intrinsic magnetic property known as ‘spin’.
The information (0 or 1) is written into the spin as a particular magnetic orientation – up or down – and is carried by the electron.
This requires less current, and therefore power, than conventional charge-based electronics.
In a conventional spintronic device, a large electric current is run through the outside of a magnet. This generates and drives spin into the magnet to switch its magnetic orientation. But some of the spin naturally dissipates and this reduces the efficiency of the process.
“Until now, the field of spintronics has focused only on reducing spin losses, but we have presented a new direction by using the losses as energy to induce magnetisation switching,” says Dr Dong-Soo Han of the Korea Institute of Science and Technology (KIST) Semiconductor Technology Research Center.
Han is senior author of the study presenting the findings in Nature Communications.
In the new method, current flows directly into the magnetic material instead. This causes spin to escape in one direction only, which acts on the magnetic material causing it to switch its magnetic orientation.
The team demonstrated that the greater this spin loss was, the less power was required to switch the magnetisation. As a result, the energy efficiency is up to 3 times higher than the conventional method.
“The innovative perspective, where dissipation can be beneficial to achieve efficient spin torques, provides a guideline for designing a new class of low-power spintronic devices, and expands our understanding of intricate spin transport phenomena in magnetic multilayers,” write the authors of the study.
According to Han, the team now plans to actively develop ultra-small and low-power AI semiconductor devices. “They can serve as the basis for ultra-low-power computing technologies that are essential in the AI era,” he says.
