Brain-inspired computing is an emerging field, which intends to extend the capabilities of information technology beyond digital logic. The progress of the field relies on artificial synaptic devices as the building block for brainlike computing systems. Here, we report an electronic synapse based on a ferroelectric tunnel memristor, where its synaptic plasticity learning property can be controlled by nanoscale interface engineering. The effect of the interface engineering on the device performance was studied. Different memristor interfaces lead to an opposite virgin resistance state of the devices. More importantly, nanoscale interface engineering could tune the intrinsic band alignment of the ferroelectric/metal-semiconductor heterostructure over a large range of 1.28 eV, which eventually results in different memristive and spike-timing-dependent plasticity (STDP) properties of the devices. Bidirectional and unidirectional gradual resistance modulation of the devices could therefore be controlled by tuning the band alignment. This study gives useful insights on tuning device functionalities through nanoscale interface engineering. The diverse STDP forms of the memristors with different interfaces may play different specific roles in various spike neural networks.
Memristors, acting as artificial synapses, have promised their prospects in neuromorphic systems that imitate the brain's computing paradigm. However, most studies focused on the understanding of the memristive mechanism and how to optimize the synaptic performance, and the implementations of higher‐order cognitive functions are quite limited. Here the experimental demonstration of a representative network level learning function, i.e., associative learning and extinction, in a compact memristive neuromorphic circuit with only one memristor is reported. The association of the conditioned and unconditioned stimulus is established within a temporal window through the spike‐timing‐dependent plasticity rule, whereas the extinction of the formed memory is due to the synaptic depression. The temporal contiguity consists with biological behaviors and reflects nature's cause and effect rule. An efficient methodology of integrating memristors into large‐scale neuromorphic systems for massively parallel computing, such as pattern recognition, is provided herein.
An electro-photo-sensitive synapse based on a highly reliable InGaZnO thin-film transistor is demonstrated to mimic synaptic functions and pattern-recognition functions.
A high-energy electron scattering study of the electronic structure and elemental composition of O-implanted Ta films used for the fabrication of memristor devices
The memristor is a promising candidate for the next generation non-volatile memory, especially based on HfO2−x, given its compatibility with advanced CMOS technologies. Although various resistive transitions were reported independently, customized binary and multi-level memristors in unified HfO2−x material have not been studied. Here we report Pt/HfO2−x/Ti memristors with double memristive modes, forming-free and low operation voltage, which were tuned by oxidation conditions of HfO2−x films. As O/Hf ratios of HfO2−x films increase, the forming voltages, SET voltages, and Roff/Ron windows increase regularly while their resistive transitions undergo from gradually to sharply in I/V sweep. Two memristors with typical resistive transitions were studied to customize binary and multi-level memristive modes, respectively. For binary mode, high-speed switching with 103 pulses (10 ns) and retention test at 85 °C (>104 s) were achieved. For multi-level mode, the 12-levels stable resistance states were confirmed by ongoing multi-window switching (ranging from 10 ns to 1 μs and completing 10 cycles of each pulse). Our customized binary and multi-level HfO2−x-based memristors show high-speed switching, multi-level storage and excellent stability, which can be separately applied to logic computing and neuromorphic computing, further suitable for in-memory computing chip when deposition atmosphere may be fine-tuned.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.