Professor Gunuk Wang’s group develops a brain-mimic electronic device for low-power accelerated learning inspired by Dopamine-Facilitated Synaptic Activity
Article highlighted on the back cover of Advanced Functional Materials
▲ Gunuk Wang, professor at the KU-KIST Graduate School of Converging Science and Technology (left, corresponding author) and Seonggil Ham, student (right, first author).
Professor Gunuk Wang’s Group in the KU-KIST Graduate School of Converging Science and Technology developed a brain-mimic electronic device that may mimic the mechanism of dopamine, a neurotransmitter, by using an organic/inorganic halide perovskite, a substance which has recently drawn much attention as a next-generation solar cell material.
The secretion of dopamine is known to be a key mechanism in people’s repetition of action by contributing to addiction, pleasure, and motivation. A recent study has shown that the secretion of dopamine accelerates learning or improves adoption in a new environment or in an environment exposed to light. The KU research team mimicked the biological mechanism to develop a novel artificial synapse electronic device that may accelerate learning and increase the energy efficiency by using natural light.
The synapse device is a biomimetic device inspired by the biological functions of synapse, a junction between neurons. The study of a synapse-mimic device is important because the chemical information system of synapse that transmits information from the brain is capable of performing high-level parallel computing while consuming a very little amount of energy.
The research group employed a two-terminal memristor structure that may the electric resistance information in several stages by forming/annihilating with an electric field the iodine vacancy filament of an organic/inorganic halide perovskite. Through this structure, the research group realized the synaptic plasticity to generate and delete memory, which includes the long-term potentiation to store memory and the long-term depression to delete memory by controlling the synaptic weight. In addition, the authors proposed the applicability of neuromorphic technology that provides the functions of information storage, memorization, and learning.
In particular, the authors showed that the high-order operation system, which lowers the activation energy needed for the migration of iodine vacancy by giving a light-assisting effect to the conventional operation method based on electrical stimulus, can reduce the onset threshold of the synaptic plasticity, allowing the acceleration of memorizing and learning. The result of the study also showed that the developed system can decrease the energy consumption in the early learning stage by more than 2,600 times, and give a pattern recognition accuracy over 80% even after a relatively small number of learning epoch.
The study may provide a foundation for the development of future artificial hardware technologies that allow low-power and high-efficient neuromorphic computing-based image recognition and machine learning. Given the important contribution of this research, its results were featured on the back cover of January 30, 2019 issue of Advanced Functional Materials (impact factor: 13.325), one of the most authoritative journals in the field of material science.
The research was supported by the Basic Science Research Program through the National Research Foundation of Korea (Young Researcher Program, NRF-2016R1C1B2007330), the KU-KIST Research Fund, the Korea University Future Research Grant, and Samsung Electronics.
- Title of article: Photonic Organic Halide Perovskite Artificial Synapse Capable of Accelerated Learning at Low Power Inspired by Dopamine-facilitated Synaptic Activity.
- Five authors of the article: Seonggil Ham (KU, first author), Sanghyeon Choi (KU), Haein Cho (KU), Seok-In Na (Chonbuk National University) and Gunuk Wang (KU, corresponding author)
▲ (Figure) Schematic of the organic/inorganic perovskite-based artificial synaptic device and its working principle (Ag/CH3NH3PbI3). The electric field generated in the device by the light-assisted effect helps the migration of iodine vacancy and generates a conductive switching filament.