A domestic research team has identified for the first time in the world that 'spin loss'—previously cited as a cause of energy efficiency reduction—can actually promote magnetization transition in ferromagnetic materials, thereby increasing energy efficiency.
The research team led by Dr. Dongsoo Han from the Korea Institute of Science and Technology (KIST) announced on the 17th that they collaborated with Professor Jeongil Hong from the Daegu Gyeongbuk Institute of Science and Technology (DGIST) and Professor Kyunghwan Kim from Yonsei University to publish these results in the prestigious scientific journal Nature Communications.
Spintronics is a next-generation semiconductor technology that utilizes the magnetic property known as 'spin' rather than the charge of electrons. With low power consumption and non-volatile characteristics that retain information even when the power is off, it is gaining attention in key fields such as ultra-low power memory, neuromorphic chips, and artificial intelligence (AI) semiconductors.
It changes the magnetic direction inside the ferromagnetic material to store information or perform computations, which requires a strong spin current injection, resulting in a significant loss of spin.
The industry has focused on material design and process improvements to reduce this loss until now, but there were fundamental limitations.
The research team discovered that spin loss actually has the 'counteracting effect' of assisting magnetization transition. It was experimentally proven that the greater the loss, the less power is needed to change the magnetic direction, which is a paradoxical phenomenon.
This finding overturns the conventional wisdom that the loss must unconditionally be reduced, showing that it in fact plays a role as an energy source that induces spontaneous magnetization transition.
The study confirmed that the power required for device operation could be reduced to as low as one-third of the existing level, while energy efficiency improved by more than three times.
The reason this achievement is garnering attention is that it is compatible with existing semiconductor processes. The research team explained that utilizing this technology would enhance mass production capabilities and favor miniaturization and high integration. This implies it could be applied not only in AI semiconductors and ultra-low power memory but also in various next-generation computational structures such as probabilistic computing devices and neuromorphic computing.
Particularly, as the power consumption of AI servers and data centers has sharply increased recently, energy efficiency has emerged as a key issue in the industry, raising expectations for the commercialization potential of the research results.
Dr. Dongsoo Han noted, 'The spintronics field has focused solely on reducing losses until now, but this research has presented a new paradigm where losses can be utilized instead.' He added, 'We plan to actively pursue the development of ultra-small and ultra-low power semiconductor devices essential for the AI era.'