Researchers have developed a mathematical approach that could make future magnetic memory dramatically more energy-efficient, potentially reducing the power required to write digital information by several orders of magnitude.
The research, led by scientists at the University of Edinburgh, focuses on a fundamental operation behind computing: switching the magnetic state used to represent a digital bit.
Rather than introducing a new memory material, the researchers devised a more efficient way of controlling how magnetic states are switched. Their simulations suggest the technique could achieve ultrafast switching while using far less energy than conventional methods.
The study was published in Advanced Materials.
Why AIโs Memory Energy Use Matters
Artificial intelligence depends on enormous amounts of data being repeatedly stored, transferred, retrieved and rewritten.
Large language models, image generators, recommendation systems and scientific simulations all require extensive interaction between processors and memory.
Each memory operation consumes energy.
Individually, these energy costs can be extremely small, but when multiplied across billions of devices and large data centres, they contribute significantly to the electricity demands of modern computing.
Improving the efficiency of memory technology could therefore help reduce the energy footprint associated with increasingly data-intensive technologies.
How Magnetic Memory Stores Information
Magnetic memory represents information using the orientation of tiny magnetic regions.
One magnetic orientation can represent a digital 0, while the opposite orientation represents a 1.
Writing information requires switching the magnetization from one stable direction to another.
The challenge is to perform this switching rapidly and reliably while consuming as little energy as possible.
Traditional magnetic-field pulses can be inefficient, particularly as memory devices become smaller.
They may require relatively large amounts of energy and can become difficult to focus on extremely small magnetic regions.
These limitations have encouraged the development of alternatives such as spin-transfer torque magnetic random-access memory (STT-MRAM) and spin-orbit torque magnetic random-access memory (SOT-MRAM).
A Smarter Way to Flip a Magnetic Bit
The Edinburgh researchers revisited magnetic-field switching using a mathematical technique known as optimal control theory.
Optimal control theory is designed to determine the most efficient sequence of actions needed to move a system from one state to another.
Instead of applying a magnetic pulse with a simple predetermined shape, the new framework calculates how the magnetic field should change over time.
The aim is to guide the magnetization naturally toward its new state while minimizing wasted energy.
Rather than forcing the magnetic system to reverse direction, the method takes advantage of its inherent dynamics and identifies a more efficient pathway through the switching process.
Testing the Method on Ultrathin Magnets
The researchers tested their approach using computer simulations involving three ultrathin van der Waals magnetic materials:
FeโGaTeโ, FeโGeTeโ and CrSBr.
Van der Waals materials consist of extremely thin atomic layers held together by relatively weak forces.
Their magnetic properties at very small scales have made them promising candidates for future compact electronic and spintronic devices.
The simulations showed that carefully optimized magnetic pulses could reverse the magnetization of these materials extraordinarily quickly.
Switching in Trillionths of a Second
According to the simulations, magnetization could be reversed in approximately 1 to 10 picoseconds.
A picosecond is one trillionth of a second.
The optimized magnetic fields were also more than 10 times weaker than those typically used in standard switching approaches.
Under the conditions modelled by the researchers, the energy needed for switching fell as low as approximately 0.94 nanojoules, compared with values reaching about 91.2 nanojoules for conventional magnetic-field pulses.
The findings demonstrate how reshaping the magnetic pulse rather than simply increasing its strength could significantly improve efficiency.
Toward Femtosecond-Scale Energy Efficiency
The modelling suggests that even larger improvements could be possible by optimizing physical properties such as magnetic damping and anisotropy.
Magnetic damping influences how quickly a disturbed magnetic system settles into a stable configuration, while anisotropy determines which directions the magnetization naturally prefers.
By carefully tuning these properties, the researchers predict that switching energies could potentially fall into the femtojoule range.
A femtojoule is one quadrillionth of a joule.
At such energy levels, optimized magnetic-field switching could potentially compete with or outperform established technologies based on spin-transfer and spin-orbit torques.
Moving Closer to Computingโs Fundamental Energy Limit
The researchers compared their predicted energy requirements with several leading memory technologies, including DRAM, STT-MRAM and emerging SOT-MRAM devices.
Their calculations suggest that optimized switching could reduce energy use by several orders of magnitude under appropriate conditions.
The predicted efficiencies also bring magnetic memory closer to the Landauer limit.
The Landauer limit describes the minimum amount of energy that must theoretically be dissipated when information is irreversibly erased.
Approaching this limit is an important goal in efforts to develop computing systems that perform information processing with minimal energy loss.
A Potential Route to More Energy-Efficient AI
The results are particularly relevant as artificial intelligence systems increase the amount of information being processed worldwide.
AI energy consumption is not determined solely by computation. Moving and repeatedly writing information in memory also contributes to overall energy demand.
Reducing the energy associated with these operations could therefore improve the efficiency of future AI hardware and data-intensive computing systems.
However, the results are currently based on computer simulations and theoretical modelling.
The researchers have not yet demonstrated the predicted energy reductions in a commercial memory device.
The study instead provides mathematical designs and possible device concepts that can now be investigated experimentally.
The Mathematics Could Extend Beyond Magnetic Fields
The researchers say the significance of the approach may extend beyond magnetic-field pulses.
The same optimal-control framework could potentially be adapted to optimize electrical currents and ultrafast laser pulses, both of which are being investigated for future magnetic memory and spintronics technologies.
This flexibility could allow future devices to use combinations of magnetic fields, electrical currents and light to manipulate magnetic states more efficiently.
Dr. Elton Santos, from the University of Edinburghโs Institute for Condensed Matter Physics and Complex Systems, led the research.
The team argues that carefully engineering how the forces used to switch magnetization change over time could provide a more efficient alternative to simply increasing their magnitude.
What Comes Next
The study establishes a theoretical pathway toward faster and lower-energy magnetic switching, but experimental validation will be essential.
Researchers will need to determine whether the precisely shaped pulses predicted by the mathematical framework can be produced reliably in real devices and whether the efficiency gains remain when other engineering constraints are considered.
If those challenges can be overcome, the technique could contribute to future generations of energy-efficient memory, spintronic devices and computing hardware.
As artificial intelligence and other data-intensive technologies continue to expand, improvements at the level of individual memory operations could ultimately translate into substantial reductions in energy use across large computing systems.
Journal reference
Mohammad H. Badarneh, PeiYu Cai and Elton J. G. Santos. โOptimal Control Drives Ultrafast and Energy-Efficient Magnetization Switching in Van der Waals Magnets.โ Advanced Materials, 14 July 2026.