Bipolar resistive switching with unidirectional selector function in nitride/oxide heterostructures

Abstract

In this work, we propose a self-rectifying Ni/SiNx/HfO2/p++Si resistive memory device to alleviate the sneak-path current occurring in crossbar array. The bilayer (Ni/SiNx/HfO2/p++Si) device exhibits a much higher rectification ratio (>104) in the low-resistance state for DC sweep mode and pulse mode than single-layer devices (Ni/SiNx/p++Si and Ni/HfO2/p++Si). The suppressed current of the bilayer device can be explained by the high Schottky barrier of the HfO2 layer under a negative bias. The modified read bias scheme in the crossbar array structure ensures a large number of word line (∼3971 at a read margin of 10%) using the advantage of the high rectification of the bilayer device. The bilayer device with the proposed read bias scheme is promising for high-density memory applications.

Synaptic Characteristics from Homogeneous Resistive Switching in Pt/Al2O3/TiN Stack

“…It is important to implement multilevel conductance by analog switching to provide synaptic functions in a neuromorphic chip. Resistive switching random access memory (RRAM) can act as a synapse in a neuromorphic chip because of its low-power operation [7], fast switching time [8], high-density integration [9], and multi-level cells (MLC) with analogue switching [10][11][12][13][14]. The various resistive switching characteristics are achieved using a dielectric material and metal electrodes [15][16][17][18][19][20][21][22][23][24][25][26][27].…”

Pseudo-Interface Switching of a Two-Terminal TaOx/HfO2 Synaptic Device for Neuromorphic Applications

“…Compared with von Neumann architecture, which has limitations, such as a bottleneck between processor and memory, the human brain can learn and process a complex task with low power consumption and parallel computation [1,2]. In recent years, with the advent of big data and artificial intelligence (AI), a greater amount of data that is more complex can be efficiently processed with lower amounts of energy [3][4][5][6][7]. In response to the demand for new computing architecture, a neuromorphic system has been proposed and self-learning neuromorphic chips have already been developed.…”

Post-Moore Memory Technology: Sneak Path Current (SPC) Phenomena on RRAM Crossbar Array and Solutions

“…The built-in SPC immunity of RRAM with intrinsic nonlinearity for suppressing the sneak path current, namely “selectorless RRAM” [ 31 , 32 , 33 ]. The nonlinearity (NL) is defined as the current at full-read voltage divided by the current at 1/2 read voltage as the V/2 read [ 34 ]. The built-in nonlinear nature can alleviate the sneak current because the on-state of the selected cell can be read at a “high-voltage” region, while the sharp conductance decrease at the “low-voltage region” effectively suppresses sneak current results from the unselected cells [ 35 ].…”