Artificial intelligence and other digital services are creating huge amounts of data that must be stored and processed, and this work uses a lot of electricity. Data centres already use enormous amounts of power, and demand is expected to grow quickly in coming years. Because of this, scientists are looking for ways to make computer memory use far less energy.
Researchers at the University of Edinburgh have now developed a new theoretical approach that could sharply cut the energy needed to store and change digital information in future magnetic memory devices. These devices store bits by flipping the direction of tiny magnetic regions. The team used Optimal Control Theory, a mathematical method for finding the most efficient way to reach a goal, to design ultrafast bursts of magnetic field that flip these magnetic states using as little energy as possible, while remaining realistic enough to build in a laboratory.
Approaching a physical limit
Computer simulations suggested that this method could cut the energy needed to switch a bit by a very large factor compared with current memory technologies such as DRAM (a common type of fast computer memory), STT-MRAM and the newer SOT-MRAM (two types of magnetic memory that use electric current to flip magnetic bits). The predicted energy use comes close to the Landauer limit, a fundamental rule from physics setting the smallest possible amount of energy needed to process a single bit of information, no matter how good the technology becomes.
The methods and results of this study are published in Advanced Materials. The paper also offers practical guidance for building such devices, including suggestions for device designs and ways of delivering the magnetic field pulses, giving researchers a realistic path toward testing the idea in real experiments.
Dr Elton Santos, who led the research, explained that every digital operation carries an energy cost, and that this matters more as AI and data-heavy technologies keep growing. He added that although the theory was first built around magnetic field pulses, the same mathematics could also be applied to electrical currents and even ultrafast laser pulses, among the most advanced tools being explored for future data storage. This suggests the approach could have uses well beyond the systems studied so far.