MRS Meetings and Events


EL20.09.02 2023 MRS Fall Meeting

Investigation of Connection Strength Modulation Between the Artificial Neuron Devices Based on 2D hBN Threshold Switching RRAM

When and Where

Nov 30, 2023
10:15am - 10:30am

Hynes, Level 3, Room 301



Yooyeon Jo1,Gichang Noh1,Eunpyo Park1,Min Jee Kim1,Yong Woo Sung1,Dong Yeon Woo1,Dae Kyu Lee1,Joon Young Kwak1,2

Korea Institute of Science and Technology1,Korea University of Science and Technology2


Yooyeon Jo1,Gichang Noh1,Eunpyo Park1,Min Jee Kim1,Yong Woo Sung1,Dong Yeon Woo1,Dae Kyu Lee1,Joon Young Kwak1,2

Korea Institute of Science and Technology1,Korea University of Science and Technology2
The brain-inspired neuromorphic computing has attracted many researchers to solve the von Neumann bottleneck problems of conventional computing systems [1-2]. Various emerging materials and structures have been proposed to emulate the performance of biological neurons and synapses successfully. In particular, 2D materials-based memristors have been investigated by many researchers because of their excellent electrical performance, simple structure, and small device scale as promising building blocks of neuromorphic computing [3-5]. Although many artificial synaptic devices based on 2D materials-based memristors have been reported, there are very few reports in artificial neuron research, which is one of the key components in neuromorphic computing systems. Also, the integration of developed artificial neuromorphic devices based on emerging materials is an essential field of research to verify the performance of an artificial neural network (ANN) on the system level and successfully demonstrate large-scale ANN hardware [6-7].<br/>In this study, we fabricated the threshold switching resistive random-access memory (RRAM) based on 2D multi-layer hBN film to mimic the behavior of the biological neuron for neuromorphic computing. The fabricated device was composed of Au/Ti bottom and Au/Ag top electrodes, respectively. The fabricated volatile memristor showed the highly-reliable resistance switching characteristics with 0.22 V of the average set voltage even at 110 fA of compliance current. Additionally, it showed a huge ON/OFF ratio over 3 × 10<sup>6</sup> at 0.1 V of reading voltage. The threshold switching characteristics were well maintained with the different series resistors from 100 kΩ to 10 MΩ. We evaluated the switching speed of the fabricated device via an AC response experiment. The 100 kΩ series resistor was connected to protect the device from a breakdown. As a result, it showed 22.7 and 30.9 µs of ON and OFF time, respectively.<br/>The leaky-integrate-and-fire (LIF) neuron behaviors were demonstrated with the fabricated device and passive components. According to the RC time delay and the input pulse amplitude, the number of output spikes of the artificial neuron could be modulated, resulting in tailoring the artificial neuron properties to the desired performance for various applications. Lastly, we integrated the two artificial neurons to demonstrate the biological neural networks. The two LIF neurons were connected via a synaptic resistor as a biological synapse. Depending on the resistance (connection strength between the neurons), the number of output spikes of the post-synaptic neuron was modulated. We believe that our experimental results pave the way for the development of large-scale neural network hardware in the future.<br/><br/>Acknowledgments: This work was supported by the National Research Foundation of Korea (NRF) (grant no. 2021M3F3A2A01037738) and Korea Institute of Science and Technology (KIST) through 2E32260.<br/><br/>References<br/>[1] G. Cao, Adv. Funct. Mater., 2021 [2] H.-M. Huang, Adv. Intell. Syst., 2020 [3] C. Y. Wang, Adv. Electron. Mater., 2020 [4] H. Bian, Adv. Mater., 2021 [5] S. Chen, Nat. Electron., 2020 [6] J. Woo, IEEE Electron Device Lett., 2019 [7] Q. Duan, Nat. Commun., 2020


2D materials

Symposium Organizers

Gina Adam, George Washington University
Sayani Majumdar, Tampere University
Radu Sporea, University of Surrey
Yiyang Li, University of Michigan

Symposium Support

APL Machine Learning | AIP Publishing

Publishing Alliance

MRS publishes with Springer Nature