Breakthrough in Spintronics: Turning Spin Loss into Energy for Ultra-Low-Power AI Chips

Scientists turn spin loss into energy, unlocking ultra-low-power AI chips

Dr. Dong-Soo Han's research team at the Korea Institute of Science and Technology (KIST) Semiconductor Technology Research Center, in collaboration with the research teams of Prof. Jung-Il Hong at DGIST and Prof. Kyung-Hwan Kim at Yonsei University, has developed a device principle that can utilize "spin loss," which was previously thought of as a simple loss, as a new power source for magnetic control. Spintronics is a technology that utilizes the "spin" property of electrons to store and control information, and it is being recognized as a key foundation for next-generation information processing technologies such as ultra-low-power memory, neuromorphic chips, and computational devices for stochastic computation, as it consumes less power and is more non-volatile than conventional semiconductors. This research is significant because it presents a new approach that can significantly improve the efficiency of these spintronics devices. A team of researchers has identified a new physical phenomenon that allows magnetic materials to spontaneously switch their internal magnetization direction without external stimuli. Magnetic materials are key to the next generation of information processing devices that store information or perform computations by changing the direction of their internal magnetization. For example, if the magnetization direction is upward, it is recognized as '1', and if it is downward, it is recognized as '0', and data can be stored or computed. Traditionally, to reverse the direction of magnetization, a large current is applied to force the spin of electrons into the magnet. However, this process results in spin loss, where some of the spin does not reach the magnet and is dissipated, which has been considered a major source of power waste and poor efficiency. Researchers have focused on material design and process improvements to reduce spin loss. But now, the team has found that spin loss actually has the opposite effect, altering magnetization. This means that spin loss induces a spontaneous magnetization switch within the magnetic material, just as the balloon moves as a reaction to the wind being taken out of it. In their experiments, the team demonstrated the paradox that the greater the spin loss, the less power is required to switch magnetization. As a result, the energy efficiency is up to three times higher than conventional methods, and it can be realized without special materials or complex device structures, making it highly practical and industrially scalable. In addition, the technology adopts a simple device structure that is compatible with existing semiconductor processes, making it highly feasible for mass production, and it is also advantageous for miniaturization and high integration. This enables applications in various fields such as AI semiconductors, ultra-low power memory, neuromorphic computing, and probability-based computing devices. In particular, the development of high-efficiency computing devices for AI and edge computing is expected to be in full swing. "Until now, the field of spintronics has focused only on reducing spin losses, but we have presented a new direction by using the losses as energy to induce magnetization switching," said Dr. Dong-Soo Han, a senior researcher at KIST. "We plan to actively develop ultra-small and low-power AI semiconductor devices, as they can serve as the basis for ultra-low-power computing technologies that are essential in the AI era." This research was supported by the Ministry of Science and ICT (Minister Bae Kyung-hoon) through the KIST Institutional Program, the Global TOP Research and Development Project (GTL24041-000), and the Basic Research Project of the National Research Foundation of Korea (2020R1A2C2005932). The results of this research were published in the latest issue of the international journal Nature Communications (IF 15.7, JCR field 7%).

"Turning Spin Loss into Energy", Developing a Key Technology for Ultra-Low Power Next-Generation Information Devices | Newswise

Newswise -- Dr. Dong-Soo Han's research team at the Korea Institute of Science and Technology (KIST) Semiconductor Technology Research Center, in collaboration with the research teams of Prof. Jung-Il Hong at DGIST and Prof. Kyung-Hwan Kim at Yonsei University, has developed a device principle that can utilize "spin loss," which was previously thought of as a simple loss, as a new power source for magnetic control. Spintronics is a technology that utilizes the "spin" property of electrons to store and control information, and it is being recognized as a key foundation for next-generation information processing technologies such as ultra-low-power memory, neuromorphic chips, and computational devices for stochastic computation, as it consumes less power and is more non-volatile than conventional semiconductors. This research is significant because it presents a new approach that can significantly improve the efficiency of these spintronics devices. A team of researchers has identified a new physical phenomenon that allows magnetic materials to spontaneously switch their internal magnetization direction without external stimuli. Magnetic materials are key to the next generation of information processing devices that store information or perform computations by changing the direction of their internal magnetization. For example, if the magnetization direction is upward, it is recognized as '1', and if it is downward, it is recognized as '0', and data can be stored or computed. Traditionally, to reverse the direction of magnetization, a large current is applied to force the spin of electrons into the magnet. However, this process results in spin loss, where some of the spin does not reach the magnet and is dissipated, which has been considered a major source of power waste and poor efficiency. Researchers have focused on material design and process improvements to reduce spin loss. But now, the team has found that spin loss actually has the opposite effect, altering magnetization. This means that spin loss induces a spontaneous magnetization switch within the magnetic material, just as the balloon moves as a reaction to the wind being taken out of it. In their experiments, the team demonstrated the paradox that the greater the spin loss, the less power is required to switch magnetization. As a result, the energy efficiency is up to three times higher than conventional methods, and it can be realized without special materials or complex device structures, making it highly practical and industrially scalable. In addition, the technology adopts a simple device structure that is compatible with existing semiconductor processes, making it highly feasible for mass production, and it is also advantageous for miniaturization and high integration. This enables applications in various fields such as AI semiconductors, ultra-low power memory, neuromorphic computing, and probability-based computing devices. In particular, the development of high-efficiency computing devices for AI and edge computing is expected to be in full swing. "Until now, the field of spintronics has focused only on reducing spin losses, but we have presented a new direction by using the losses as energy to induce magnetization switching," said Dr. Dong-Soo Han, a senior researcher at KIST. "We plan to actively develop ultra-small and low-power AI semiconductor devices, as they can serve as the basis for ultra-low-power computing technologies that are essential in the AI era." ### KIST was established in 1966 as the first government-funded research institute in Korea. KIST now strives to solve national and social challenges and secure growth engines through leading and innovative research. For more information, please visit KIST's website at https://www.kist.re.kr/eng/index.do This research was supported by the Ministry of Science and ICT (Minister Bae Kyung-hoon) through the KIST Institutional Program, the Global TOP Research and Development Project (GTL24041-000), and the Basic Research Project of the National Research Foundation of Korea (2020R1A2C2005932). The results of this research were published in the latest issue of the international journal Nature Communications (IF 15.7, JCR field 7%).