December 1 - 6, 2024
Boston, Massachusetts
Symposium Supporters
2024 MRS Fall Meeting & Exhibit
EL05.05.04

Enhanced Long-Term Synaptic Plasticity of Self-Rectifying Artificial Synapse Device with Pt/Ga2O3/NbOx/ZnO/Pt Structure

When and Where

Dec 3, 2024
8:00pm - 10:00pm
Hynes, Level 1, Hall A

Presenter(s)

Co-Author(s)

Daejae Seo1,Peter Chung1,Tae-Sik Yoon1

Ulsan National Institute of Science and Technology1

Abstract

Daejae Seo1,Peter Chung1,Tae-Sik Yoon1

Ulsan National Institute of Science and Technology1
Neuromorphic computing, which mimics neuron-synapse structure of human brain, has been explored extensively to overcome the limitation of Von-Neumann system. Neuromorphic systems require synapse devices that enable analog or multi-state weight update and stable long-term plasticity for energy-efficient training and inference operations. For the synapse device, memristor device which shows analog or multi-state resistance change has been investigated to mimic the biological synaptic weight update. Although the weight update behaviors have been demonstrated in a variety of memristors, it is still challenging to achieve long-term plasticity with good retention of weight states. In addition, self-rectifying characteristics of memristor synapse devices are highly beneficial to construct high-density crossbar array of synapse device because sneak path current problem can be effectively alleviated even without an additional selector device.[1]

In this study, high self-rectification ratio and long-term synaptic plasticity with good retention was demonstrated with Pt/Ga2O3/NbOx/ZnO/Pt device. In this device, ZnO, a well-known n-type semiconductor, was employed to achieve large analog conductance change with respect to the redistribution of oxygen vacancies. Ga2O3 had a role to present self-rectifying characteristics owing to its ultrawide-bandgap of about 4.9eV.[2] NbOx was used as a good oxygen ion reservoir due to its high oxygen binding energy. Ga2O3 and ZnO were deposited by RF magnetron sputtering in Ar condition, with the thickness of 10nm and 15nm, respectively. NbOx was deposited by reactive sputtering using Nb target in Ar and O2 condition, with the thickness of 18nm. Both top and bottom electrodes were deposited by E-beam evaporator. The device had circular shape with diameter of 100mm.

The current-voltage (I-V) sweep measurement showed that the conductance increased when the positive bias was applied and decreased upon applying negative bias. The current at +3V was about 104 times larger than the current level at -3V, exhibiting high self-rectifying characteristic of the device. In addition, pulse measurement showed analog conductance change with dynamic range of about 10 by applying 30 times of +3V pulses with a width of 0.64ms. Analog conductance modulation originated from the oxygen ion migration between ZnO and NbOx. The increased conductance retained stably to have a long-term synaptic plasticity, remaining 10 times higher current level compared to the initial state after 30 min. This good retention properties came from the captured oxygen ions within NbOx layer that has strong binding energy to oxygen. The performance of the device was evaluated with pattern recognition accuracy using CrossSim:Training.

The comparative study with two reference devices, i.e., Pt/NbOx/ZnO/Pt and Pt/Ga2O3/ZnO/Pt, revealed that Pt/NbOx/ZnO/Pt device had about 1000 times smaller self-rectification ratio, indicating that Ga2O3 layer was responsible for the self-rectifying property. Also, from much poorer retention of Pt/Ga2O3/ZnO/Pt with current which decayed to the initial current level in short time, it was confirmed that NbOx layer was responsible for the long-term retention property as oxygen ion reservoir.

References
[1] Shi, L., Zheng, G., Tian, B., Dkhil, B., & Duan, C. (2020). Research progress on solutions to the sneak path issue in memristor crossbar arrays. Nanoscale Advances, 2(5), 1811-1827.
[2] Saikumar, A. K., Nehate, S. D., & Sundaram, K. B. (2019). RF sputtered films of Ga2O3. ECS journal of solid state science and technology, 8(7), Q3064.

Symposium Organizers

Paschalis Gkoupidenis, Max Planck Institute
Francesca Santoro, Forschungszentrum Jülich/RWTH Aachen University
Ioulia Tzouvadaki, Ghent University
Yoeri van de Burgt, Technische Universiteit Eindhoven

Session Chairs

Ioulia Tzouvadaki
Yoeri van de Burgt

In this Session