No abstract

The implementation of highly anticipated hardware neural networks (HNNs) hinges largely on the successful development of a low-power, high-density, and reliable analog electronic synaptic array. In this study, we demonstrate a two-layer Ta/TaO x /TiO2/Ti cross-point synaptic array that emulates the high-density three-dimensional network architecture of human brains. Excellent uniformity and reproducibility among intralayer and interlayer cells were realized. Moreover, at least 50 analog synaptic weight states could be precisely controlled with minimal drifting during a cycling endurance test of 5000 training pulses at an operating voltage of 3 V. We also propose a new state-independent bipolar-pulse-training scheme to improve the linearity of weight updates. The improved linearity considerably enhances the fault tolerance of HNNs, thus improving the training accuracy.

A two-terminal analog synaptic device that precisely emulates biological synaptic features is expected to be a critical component for future hardware-based neuromorphic computing. Typical synaptic devices based on filamentary resistive switching face severe limitations on the implementation of concurrent inhibitory and excitatory synapses with low conductance and state fluctuation. For overcoming these limitations, we propose a Ta/TaOx/TiO2/Ti device with superior analog synaptic features. A physical simulation based on the homogeneous (nonfilamentary) barrier modulation induced by oxygen ion migration accurately reproduces various DC and AC evolutions of synaptic states, including the spike-timing-dependent plasticity and paired-pulse facilitation. Furthermore, a physics-based compact model for facilitating circuit-level design is proposed on the basis of the general definition of memristor devices. This comprehensive experimental and theoretical study of the promising electronic synapse can facilitate realizing large-scale neuromorphic systems.

No abstract

No abstract

Three-dimensional vertical resistive-switching random access memory (VRRAM) is the most anticipated candidate for fulfilling the strict requirements of the disruptive storage-class memory technology, including low bit cost, fast access time, low-power nonvolatile storage,and excellent endurance. However, an essential self-selecting resistive-switching cell that satisfies these requirements has yet to be developed. In this study, we developed a TaOx/TiO2 double-layer V-RRAM containing numerous highly desired features, including: (1) a self-rectifying ratio of up to 10³ with a sub-μA operating current, (2) little cycle-to-cycle and layer-to-layer variation, (3) a steep vertical sidewall profile for high-density integration, (4) forming-free and self-compliance characteristics for a simple peripheral circuit design, and (5) an extrapolated endurance of over 10¹⁵ cycles at 100 °C. Furthermore, the switching and self-rectifying mechanisms were successfully modeled using oxygen ion migration and homogeneous barrier modulation. We also suggest the new possibility of monolithically integrating working and storage memory by exploiting a unique tradeoff between retention time and endurance.

A neuro-inspired computing paradigm beyond the von Neumann architecture is emerging and it generally takes advantage of massive parallelism and is aimed at complex tasks that involve intelligence and learning. The cross-point array architecture with synaptic devices has been proposed for on-chip implementation of the weighted sum and weight update in the learning algorithms. In this work, forming-free, silicon-process-compatible Ta/TaOx/TiO2/Ti synaptic devices are fabricated, in which >200 levels of conductance states could be continuously tuned by identical programming pulses. In order to demonstrate the advantages of parallelism of the cross-point array architecture, a novel fully parallel write scheme is designed and experimentally demonstrated in a small-scale crossbar array to accelerate the weight update in the training process, at a speed that is independent of the array size. Compared to the conventional row-by-row write scheme, it achieves >30× speed-up and >30× improvement in energy efficiency as projected in a large-scale array. If realistic synaptic device characteristics such as device variations are taken into an array-level simulation, the proposed array architecture is able to achieve ∼95% recognition accuracy of MNIST handwritten digits, which is close to the accuracy achieved by software using the ideal sparse coding algorithm.

No abstract