Lab magnetic-wave chip could make computers 1,000 times faster

In a University of Michigan lab, doctoral student Avery-Ryan Ansbro prepares iron gallium, an alloy, with a technique called pulsed laser deposition. The thin film he makes is one part of M3BOLT, a chip unveiled by researchers at the University of Michigan (U-M) and the University of Wisconsin–Madison. The team believes it could make computers 100 to 1,000 times faster.

The problem starts with the transistor. A processor holds billions of them, switching on and off to handle data, and manufacturers spent decades shrinking them to gain speed and save energy. But the smaller they get, the less efficient they become: electrons pass through them even when they are switched off, and that wastes energy. The team estimates that conventional transistors could reach their theoretical energy-efficiency limits by 2037.

M3BOLT skips the switch. Its designers pair a ferroelectric crystal, lead magnesium niobate-lead titanate (PMN-PT), with a thin layer of magnetic iron-gallium alloy. When an electric field is applied, the crystal changes its internal electrical orientation, which alters the magnetic properties of the iron-gallium layer. That produces magnons, waves that travel through the material with very little resistance and so lose less energy as heat. Those waves could reach terahertz frequencies, against the gigahertz of today's consumer processors, which is why the team says computers could process information hundreds of times faster.

The device could also store data. Researchers would change the magnetic orientation of the iron-gallium layer, so that two magnetic states could serve as 1 and 0 instead of the two voltage levels of current transistors. That orientation stays the same after the power is switched off, so the same material could process information and keep it without a continuous power supply. These remain projections: the 100-to-1,000 figures are the team's estimates. The University of Wisconsin–Madison group will use computer simulations to study performance and identify possible material defects, while U-M continues to develop the materials, devices and circuits needed to make the technology work in real computers.

For now, the team concentrates on low-power devices, Wei Lu, an engineering professor at U-M, said. A press release from the team argues that autonomous or mobile devices could carry better computers without huge power supplies. The longer ambition is larger: “imagine that someday we can build supercomputers with them,” Lu said. “You might not need a nuclear power plant to run a large data center.” The team gives no date for any of this.