4.7 Article

Emulating synaptic plasticity and resistive switching characteristics through amorphous Ta2O5 embedded layer for neuromorphic computing

Journal

CERAMICS INTERNATIONAL
Volume 47, Issue 21, Pages 30764-30776

Publisher

ELSEVIER SCI LTD
DOI: 10.1016/j.ceramint.2021.07.257

Keywords

Neuromorphic computing; Biological synapse; Electronic synapse; Memristors; Ta2O5 barrier layer; ZrO2 resistive switching layer

Funding

  1. National Research Foundation of Korea (NRF) - Korean government (MSIP) [2021R1C1C1004422]
  2. National Research Foundation of Korea (NRF) - Ministry of Science, ICT & Future Planning [NRF-2019R1F1A1057243, NRF-2016R1A6A1A03013422]

Ask authors/readers for more resources

By introducing an amorphous Ta2O5 layer into the ZrO2 memristor, significant improvements in resistive switching and synaptic characteristics were achieved, leading to multilevel storage capability and effective mimicry of biological synaptic learning patterns.
Memristors with controllable resistive switching (RS) characteristics gain significant attention for neuromorphic computing applications in high-density memory and artificial synapses. Herein, we present the consequence of an amorphous Ta2O5 embedded layer on RS and synaptic characteristics of ZrO2 memristor. Structural and electronic properties of the memristor without and with-Ta2O5 are exemplified by x-ray diffraction (XRD) and xray photoemission spectroscopy (XPS) measurements. Memristor with-Ta2O5 exhibits excellent performance in RS parameters such as lower forming/SET-voltage, improved uniformity of switching cycling, admirable pulse endurance (10(4)), and long retention time (10(4) s). Moreover, multilevel storage capability is achieved through limiting the current compliance in SET-operation or stop-voltages in RESET-operation. Likewise, diverse synaptic characteristics such as long-term potentiation (LTP), long-term depression (LTD), spike-rate-dependent plasticity (SRDP), paired-pulse facilitation (PPF), and post-tetanic potentiation (PTP), spike-timing-dependent plasticity (STDP) are effectively mimicked, which is regarded as an imperative learning regime of biological synapses. Owing to improved linearity of LTP/LTD behaviors in memristor with-Ta2O5, a solid recognition rate (similar to 85%) is achieved for pattern recognition with the modified national institute of standards and technology database (MNIST) handwritten numbers. Memristor with embedded similar to 2 nm Ta2O5 barrier layer shows significant potential applications in high-performance multilevel data storage and neuromorphic computing systems.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.7
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available