Nitrogen-induced ultralow power switching in flexible ZnO-based memristor for artificial synaptic learning

Abstract

An energy-efficient memristive synapse is highly desired for the development of brain-like neurosynaptic chips. In this work, a ZnO-based memristive synapse with ultralow-power consumption was achieved by simple N-doping. The introduction of N atoms, as the acceptor, reduces the carrier concentration and greatly increases the resistance of the ZnO film. The low energy consumption, which is as low as 60 fJ per synaptic event, can be achieved in our device. Essential synaptic learning functions have been demonstrated, including excitatory postsynaptic current, paired-pulse facilitation, and experience-dependent learning behaviors. Furthermore, the device can still exhibit the synaptic performance in the bent state or even after 100 bending cycles. Our memristive synapse is not only promising for energy-efficient neuromorphic computing systems but also suitable for the development of wearable neuromorphic electronics.

Highly Flexible Artificial Synapses from SiO2-Based Conductive Bridge Memristors and Pt Nanoparticles through a Crack Suppression Technique

“…In Table , the switching performance of recently reported RRAM devices is presented. In most of the reported studies, synaptic properties of the switching materials that have been employed toward the demonstration of flexible devices have not been studied. ,,− , In some other published studies in which synaptic properties of the switching material have been studied, transferring of the active material or of the completed device is required, , the mechanical strain of devices is low, − while in others, the preparation of the active material is time-consuming. , These steps complicate device processing and could act as a bottleneck toward the scaling up of the process. Regarding the devices presented here, which are fabricated using wafer scale vacuum-based semiconductor processing, both V SET and V RESET are among the lowest in literature, while the large memory window, even at a high strain (4.16% strain), makes them attractive for potential applications that require multibit properties.…”

Memristor Based on Inorganic and Organic Two-Dimensional Materials: Mechanisms, Performance, and Synaptic Applications

“…In 2018, the RS characteristics of the multiterminal device with MoS 2 were further studied. The multiterminal devices exhibited diverse synaptic plasticity including heterosynaptic plasticity, long-term potentiation (LTP) and depression (LTD) plasticity, and spike-timing dependent plasticity (STDP). , Memristors have the ability to modulate the resistance states by external voltages, which makes the two-terminal device mimic a biological synapse in the neural network of the human brain. , A biological synapse is the most fundamental link in a neural network, transmitting excitatory signals via chemical neurotransmitters from the presynaptic terminal to the postsynaptic terminal. The two terminals of the memristor can be employed to serve as presynaptic and postsynaptic terminals in an artificial synapse.…”

Bio‐Inspired 3D Artificial Neuromorphic Circuits

“…[152][153][154][155][156][157][158][159] Also, the tunable resistive states of memristors can be nonvolatile, which allows the storage of the computed outputs in the circuits without the need of power supply and can significantly suppress the static power. [115,[160][161][162][163] The chemistry-physics mechanism of memristors feature several categories, [115] such as electrochemical reaction, [164][165][166] amorphous-crystalline phase transition, [167] spin dependent tunnel resistance, [168] lattice-polarization dependent resistance. [169] For detailed exploration of corresponding operating mechanism, the readers are encouraged referring to several recent comprehensive reviews.…”