The Korea Institute of Materials Science (KIMS, President Lee Jung-hwan), a government-funded research institute under the Ministry of Science and ICT, is expected to accelerate the realization of artificial intelligence semiconductor technology by successfully developing next-generation neuromorphic semiconductor core materials.

The research team of Dr. Kwon Jeong-dae and Dr. Kim Yong-hoon from the Energy and Electronic Materials Laboratory of the Korea Materials Research Institute announced on the 18th that they recently implemented a new concept memtransistor device using a two-dimensional nanomaterial a few nanometers thick in collaboration with the research team of Professor Cho Byeong-jin from Chungbuk National University.

Memtransistor is a compound word of ‘memory’ and ‘transistor.’

The research team succeeded in obtaining a high pattern recognition rate of approximately 94.2% (simulation-based pattern recognition rate of 98%) by reproducibly mimicking the electrical plasticity of neural synapses with more than 1,000 electrical stimulations.

Molybdenum sulfur (MoS2), widely used as a semiconductor material, operates on the principle that defects within a single crystal are moved by an external electric field.

On the other hand, this had the problem of making it difficult to precisely control the concentration or shape of the defect. To solve the problem, the research team chose a method of sequentially layering niobium oxide (Nb2O5) oxide layers and molybdenum sulfur materials.

Through this, we were able to develop an artificial synaptic element with a memtransistor structure that has high electrical reliability under external electric fields.

Additionally, the research team freely controlled the resistance switching characteristics by changing the thickness of the niobium oxide layer, and demonstrated that brain information related to memory and forgetting can be processed even with an extremely low energy of 10 picojoules (pJ).

Currently, artificial intelligence hardware is in the form of GPUs, FPGAs, and ASICs, which consume a lot of power and are expensive, and explosive demand is expected in line with future industrial growth.

The wearable artificial intelligence market is expected to grow at a compound annual growth rate of 29.75% from approximately $11.5 billion in 2018 to nearly $42.4 billion in 2023.

The research team of Dr. Kwon Jeong-dae and Dr. Kim Yong-hoon of the Materials Research Institute said, “If AI semiconductor devices based on the new concept of high-reliability memtransistor structure are utilized, circuit integration and operating energy can be significantly reduced,” adding, “It is expected to be applied to low-power edge computing and wearable AI systems in the future.”

This research was conducted as a major project of the Korea Institute of Materials Science with the support of the Ministry of Science and ICT, and the research results were published as a cover paper in the October 1 issue of Advanced Functional Materials (IF=18.808), a world-renowned academic journal.