New Theory Could Cut Magnetic Memory's Energy Use by Orders of Magnitude

Artificial intelligence and other data-heavy technologies are pushing global computing and data-storage demand to record levels, straining electricity supplies, and researchers are seeking ways to make computing more energy efficient. Physicists at the University of Edinburgh have developed a theoretical framework to cut the energy needed to store and change digital bits, the 0s and 1s that make up all digital information, in future magnetic memory technologies.

Magnetic memory works by switching a material's magnetic state, which is how data is written and changed. Instead of conventional methods for designing this switching process, the Edinburgh team applied Optimal Control Theory, a mathematical approach for finding the most efficient way to reach a goal. Using it, they designed ultrafast magnetic-field pulses that flip magnetic states using as little energy as possible, with calculations that account for real-world experimental limits.

Computer simulations suggest the method could cut switching energy by several orders of magnitude, roughly 1,000 to 100,000 times less, compared with leading memory technologies in use or development today, including DRAM (dynamic random-access memory), STT-MRAM and emerging SOT-MRAM devices. The predicted energy use would also move future magnetic memory much closer to the Landauer limit, the fundamental thermodynamic minimum energy needed to process one bit of information.

The findings, published in the journal Advanced Materials, are theoretical, based on simulations rather than a working device. The framework also gives practical guidance, including device designs and delivery methods, to help researchers test the idea experimentally. Dr. Elton Santos of Edinburgh's Institute for Condensed Matter Physics and Complex Systems, who led the work, said every digital operation carries an energy cost, and that carefully designing how a magnetic field changes over time lets magnetization switch far more efficiently than conventional methods.

Santos added that the same framework, though developed first for magnetic-field pulses, could also be adapted to electrical currents and ultrafast laser pulses, two other approaches explored for future data storage.