Disclaimer: machine translated by DeepL which may contain errors.
The Rigakubu News
The Rigakubu News Sep. 2026
Frontiers of Science for Undergraduates >
Non-volatile Switching Devices Enabling Ultra-High-Speed, Low-Power Computing
Tsai Hanshen (Project Assistant Professor, Department of Physics)
Satoshi Nakatsuji (Professor, Department of Physics)
The rapid development of generative AI and large-scale machine learning in recent years has created a growing demand for computing technologies that operate at high speeds with low power consumption.
However, switching devices used in current CPUs and GPUs consume increasingly more power as processing speeds increase, making it difficult to achieve ultrafast operation on sub-nanosecond timescales while maintaining low power consumption.
In this study, we demonstrated that the magnetic material Mn₃Sn (manganese tin), which has attracted attention as a promising candidate for addressing this challenge, can have its magnetic state (binary information) switched using ultrashort current pulses as short as 40 picoseconds (one trillionth of a second).
Furthermore, we demonstrated that the device operates with more than an order of magnitude lower power consumption than conventional switching devices based on ferromagnetic materials.
Mn₃Sn is a type of magnetic material known as an “antiferromagnet.” Electrons possess a degree of freedom called spin, which is one of the origins of magnetism in matter. Magnetic materials are classified into several categories based on the spin arrangement within the atoms that make up the material.In ferromagnets—commonly known as magnets—spins align in the same direction to produce a strong magnetization; since the direction of this magnetization can be used to store information, ferromagnets are widely used in storage media such as hard disk drives. In antiferromagnets, on the other hand, spins point in opposite directions and cancel each other out, resulting in a total magnetization that is nearly zero.However, since the spin response speed is approximately three orders of magnitude faster than that of ferromagnets, enabling ultra-high-speed operation, antiferromagnets are attracting attention as next-generation high-speed memory materials. On the other hand, because they exhibit almost no magnetization, it has typically been difficult to read out their spin states electrically.
Mn₃Sn is known to exhibit a large magnetic response rivaling that of ferromagnets despite being an antiferromagnet, due to its unique electronic band structure with topological properties, and is therefore expected to serve as a next-generation high-speed, low-power memory material.However, although high-speed operation in the picosecond range was theoretically expected, previous experiments showed switching speeds limited to the nanosecond range, and performance surpassing that of existing ferromagnets had not been demonstrated.
In this study, we fabricated a thin-film structure combining Mn₃Sn and tantalum (Ta) and evaluated its ultra-high-speed switching characteristics in the picosecond range.By injecting the spin angular momentum generated by a current pulse applied to Ta into Mn₃Sn, we were able to switch the magnetic state of Mn₃Sn known as the magnetic octupole.As a result, we succeeded in non-volatile switching—in which information is retained even after the power is turned off—using a 40-picosecond current pulse, and demonstrated that the device operates at a power density less than one-tenth that of previously reported switching devices.This demonstrates the ability to achieve both ultra-high-speed operation and low power consumption, indicating the potential for this technology to serve as a next-generation memory alternative to ferromagnetic memory.
We also confirmed that switching is possible using photocurrent pulses generated from light in the wavelength band used for communications. These results are expected to lead to applications in new information processing technologies that convert information transmitted via optical fiber into electrical signals and write them directly to memory.
This research is also featured in the following video.
Video Title: Exploring Everything from Elementary Particles to the Universe with the Essence of Quantum Computing
https://youtu.be/UXdPHyjMD-s?t=1005
Schematic diagram of non-volatile quantum switching combining a photoelectric conversion device and an antiferromagnetic switching device