Microstructural engineering in interface-type synapse device for enhancing linear and symmetric conductance changes

Abstract

The origins of the nonlinear and asymmetric synaptic characteristics of TiO x -based synapse devices were investigated. Based on the origins, a microstructural electrode was utilized to improve the synaptic characteristics. Under an identical pulse bias, a TiO x -based synapse device exhibited saturated conductance changes, which led to nonlinear and asymmetric synaptic characteristics. The formation of an interfacial layer between the electrode and TiO x layer, which can limit consecutive oxygen migration and chemical reactions, was considered as the main origin of the conductance saturation behavior. To achieve consecutive oxygen migration and chemical reactions, structural engineering was utilized. The resultant microstructural electrode noticeably improved the synaptic characteristics, including the unsaturated, linear, and symmetric conductance changes. These synaptic characteristics resulted in the recognition accuracy significantly increasing from 38% to 90% in a neural network-based pattern recognition simulation.

Hafnium Oxide (HfO2) – A Multifunctional Oxide: A Review on the Prospect and Challenges of Hafnium Oxide in Resistive Switching and Ferroelectric Memories

“…[30,36,[52][53][54]115] Several studies have confirmed the possibility of interfacial switching in RRAM. [42,[55][56][57][58][59] Goux et al reported an ultra-thin 1-nm Al 2 O 3 interface in TiN/HfO 2 /Hf RRAM, which can minimize I SET to sub-500 nA. [30] The higher HRS in these devices originates from the large bandgap of the Al 2 O 3 layer.…”

“…Pt/HfO 2 /TiO x /Pt [42] 2 nm free -2 +0.5 46 nA 454 --10 2 @ 25 Pt/TiO 2-x /TiO 2 /Pt [43] 40 nm +2 +1.5 -1.4 750 µA 50 ---TiN/a-Si/ TiO 2 /TiN [56] -Free -3 +6 1 µA 100 -10 6 10 8 @ 55 W/WO x /HfO 2 /Pd [57] 1 µm Free +1.5 -3 100 µA 100 100 ns 10 9 10 4 @ 85 W/WO x /AlO x /IrO x [59] 4 µm Free +1.5 -1 30 µA 500 -10 5 10 4 @ 85 TiN/TiO x /Mo [58] 50 nm Free +3 -2.5 100 nA 10 ---filamentary switching can experience nanosecond speed, which can be further scaled down by scaling of charging time of crossbar devices. Lee et al achieved a 300-ps switching speed in HfO xbased RRAM.…”

Neural Network Training Acceleration With RRAM-Based Hybrid Synapses

“…In this work, however, we exploit the switching mechanism of the Mo/TiO x -based RRAM, i.e., area-dependent conductance scaling, to implement the gain at the device level. Previously, we reported a microstructural engineered Mo/TiO x RRAM for electronic synapse applications ( Park et al, 2019 ); the study presented some promising synaptic features of the Mo/TiO x RRAM, such as gradual and linear conductance programming. However, the present expanded work adds significantly more explanatory details regarding the areal dependency of the conductance precision, which is utilized to construct a hybrid synapse.…”

“…First, we deposited a 15 nm thick TiO x layer through RF sputtering process by using a ceramic Ti 4 O 7 target at room temperature. Then, 50 nm thick Mo top electrode was deposited by the sputtering system ( Park et al, 2019 ). The device structure and composition of each layer are shown in Figure 3A via transmission electron microscopy (TEM) image and its energy dispersive X-ray spectroscopy (EDS) line profile.…”

“…The device structure and composition of each layer are shown in Figure 3A via transmission electron microscopy (TEM) image and its energy dispersive X-ray spectroscopy (EDS) line profile. The switching mechanism of the RRAM is based on gradual oxygen migration and chemical reactions at the interface between the Mo top electrode and the TiO x layer under an electric field ( Park et al, 2019 ). As shown in Figure 3B , the areal conduction contributes to a gradual increase (or decrease) in the conductance states when a positive (or negative) bias is applied, which are called potentiation and depression, respectively.…”