Scientists Cut Memory Switching Energy by Up to 100x

Physicists at the University of Edinburgh have developed a theoretical method that could dramatically reduce the energy required to switch magnetic memory. Instead of using a powerful magnetic-field pulse to force a bit directly from one state to another, the researchers used optimal control theory to calculate a precisely shaped pulse that guides the magnetization along a much more energy-efficient trajectory. The simulations focused on three atomically thin van der Waals magnets: Fe₃GaTe₂, Fe₃GeTe₂ and CrSBr. Using optimized magnetic fields, the researchers achieved uniform spin reversal in just 1 to 10 picoseconds. The required field amplitudes were more than ten times lower than those used in conventional switching approaches. The calculated energy savings were even more impressive. Conventional field-driven switching required approximately 42.8 to 91.2 nanojoules per operation. With optimal control, that figure fell to around 0.94 to 9.7 nanojoules — up to roughly two orders of magnitude lower. The key idea is surprisingly simple: instead of applying a stronger pulse, calculate the most efficient shape for that pulse. This allows the spins inside the magnetic material to follow a carefully optimized path between the “0” and “1” states instead of fighting directly against the initial magnetization. The researchers also found that further optimization of device size, magnetic anisotropy and damping could theoretically push switching energies into the femtojoule range. In some modeled conditions, that could make the approach competitive with — or even more efficient than — existing STT-MRAM and SOT-MRAM technologies. The same optimal-control principles may also be applicable to electric currents and laser pulses, potentially creating new ways to reduce energy consumption across spintronic and memory devices. This could become increasingly important for artificial intelligence. AI systems constantly move and rewrite enormous amounts of data, making memory access a significant part of total system power consumption. However, this is still a computational study. Researchers have not yet demonstrated a commercial memory cell using these optimized magnetic pulses. The next step will be translating the predicted pulse shapes into real nanoscale devices and determining whether the same energy savings can be achieved experimentally. Could smarter control signals become the key to ultra-low-power memory for future AI systems? Subscribe to Xenon Lab News for more semiconductor, computing and materials-science breakthroughs. #MagneticMemory #MRAM #LowPowerComputing #Spintronics #VanDerWaals #2DMaterials #ArtificialIntelligence #AIHardware #FutureMemory #OptimalControl #Nanotechnology #ComputerMemory #Semiconductors #ScienceNews #TechNews #XenonLabNews