A device capable of implementing ultra-low-power, high-performance computing by mimicking the nerves of the human brain has been developed. This device can operate the compression method by using the change in electrical resistance of a material that changes from metal to insulator.
The Korea Basic Science Institute (KBSI) announced on the 28th that Senior Researcher Hong Woong-gi and Korea Chemical Technology Research Institute postdoctoral researcher Jang Hoon-soo have made a nano device with a parallel structure similar to the brain’s signal system and developed a system multi-parallel signal. -memory device switching resistance.
A computer system that mimics the human brain is called ‘neuromorphic computing’. Cranial nerves are synapses that connect neurons in parallel, and provide only meaningful stimuli to neurons, ignoring other signals to reduce energy consumption. Neuromorphic computing is being developed to process information with high efficiency and low power, just like cranial nerves. It is expected to be used mainly in areas with large data scales such as the Internet of Things (IoT) and artificial intelligence (AI).
The research team used vanadium dioxide, a mixture of vanadium and oxygen, as a material that can act as a synapse. Vanadium dioxide, an oxide semiconductor, has the characteristic of a phase transition (a phenomenon that changes to another phase depending on external conditions) from a metal phase to an insulator phase depending on stimuli such as temperature or pressure. In particular, when vanadium dioxide is on a substrate such as silicon or glass, the crystal structure changes according to the interaction with the substrate.
The research team created a passageway for electrons to move by manufacturing several vanadium dioxides in a parallel structure containing 600 nm (nanometer, one billionth of a meter) unit devices. The vanadium dioxide structure realizes a synaptic connection structure that changes from metal to insulator with an external stimulus, just as a spike signal from a neuron goes to another neuron through a synapse.
The phase change process of the vanadium dioxide material can be confirmed by the spectral signal according to the crystal structure change and the electrical signal using a compound Raman microscope. In addition, the structural analysis of multiple resistance changes was measured by computer simulation modeling. The research team explains that the newly developed switchable memory device can express and store different states of resistance rather than binary computing. It is expected to store and process at least twice as much data as binary computing.
Principal Investigator Hong Woong-ki said, “While various new materials are being investigated for next-generation neuromorphic semiconductors, vanadium dioxide is being studied as a next-generation material that can realize neuromorphic computing.” We will study volatile management and the possibility of applying machine learning.”
This research was supported by the KBSI research tools development project and the National Science and Technology Research Council’s Advanced Convergence Research Project. The results of the research were published on the 14th of this month in the online edition of ‘ACS Applied Materials and Interfaces’, an international journal published by the American Chemical Society (ACS).
[참고 자료]
ACS Applied Materials and Interfaces, DOI: https://doi.org/10.1021/acsami.2c21367
