Memristors with Initial Low‐Resistive State for Efficient Neuromorphic Systems

Abstract

Memristive electronic synapses are attractive to construct artificial neural networks (ANNs) for neuromorphic computing systems, owing to their excellent electronic performance, high integration density, and low cost. However, the necessity of initializing their conductance through a forming process requires additional peripheral hardware and complex programming algorithms. Herein, the first fabrication of memristors that are initially in low‐resistive state (LRS) is reported, which exhibit homogenous initial resistance and switching voltages. When used as electronic synapses in a neuromorphic system to classify images from the CIFAR‐10 dataset (Canadian Institute For Advanced Research), the memristors offer ×1.83 better throughput per area and consume ×0.85 less energy than standard memristors (i.e., with the necessity of forming), which stems from ≈63% better density and ≈17% faster operation. It is demonstrated in the results that tuning the local properties of materials embedded in memristive electronic synapses is an attractive strategy that can lead to an improved neuromorphic performance at the system level.

Holistic Variability Analysis in Resistive Switching Memories Using a Two-Dimensional Variability Coefficient

“…Memristors, − i.e., resistors whose value can be programmed by applying electrical stresses, are one of the most studied electron devices nowadays because of their excellent electrical performance and easy fabrication, which make them attractive for a great variety of applications in the nanoelectronics landscape. Among these applications, we find nonvolatile memories − (TSMC and INTEL incorporate these devices in the 22 nm node), neuromorphic computing, − and hardware cryptography. − Memristors can be fabricated using different types of materials, including metal oxides for resistive memories, phase-change materials, magnetic materials, and ferroelectric materials …”

“…In this approach the set and reset voltages are key, and this fact could justify a variability analysis focused on these parameters, the current practice at present. However, other applications such as neuromorphic computing require an analogue operation of resistive switching devices. − Different conductance levels are used to mimic synaptic plasticity in hardware neural networks. Moreover, in the particular case of spiking neural networks, a type of network that mimics more closely the operation of the biological neural tissue, the resistive switching device operation is purely analogue.…”

Modeling the Variability of Au/Ti/h-BN/Au Memristive Devices

“…Hence, understanding and quantifying the variability is critical. In a memristive device based on CNF formation and disruption, the reset process normally exhibits higher variability than the set, i.e., V reset presents a higher dispersion than V set and the conductance after the reset (G HRS ) presents a higher dispersion than the current after the set (G LRS ) [1], [3], [37]. The reason is that the current limitation employed during the set process homogenizes the size of the CNFs.…”

Thermal Characterization of Conductive Filaments in Unipolar Resistive Memories

“…RRAMs can be used as non-volatile memories [ 2 ]; several companies commercially use these devices (TSMC for its 40 nm [ 3 ], 28 nm [ 4 ], and 22 nm [ 5 ] nodes; INTEL for its 22 nm [ 6 ] node). Another great field of development is linked to neuromorphic engineering [ 2 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 ]. In addition, these emerging devices are interesting for hardware cryptography in the context of the Internet of Things ecosystem [ 17 , 18 , 19 , 20 ].…”