Stochastic Spiking Behavior in Neuromorphic Networks Enables True Random Number Generation

Promotion of Probabilistic Bit Generation in Mott Devices by Embedded Metal Nanoparticles

Considerable attention has been drawn to the use of volatile two-terminal devices relying on the Mott transition for the stochastic generation of probabilistic bits (p-bits) in emerging probabilistic computing. To improve randomness and endurance of bit streams provided by these devices, delicate control of the transient evolution of switchable domains is required to enhance stochastic p-bit generation. Herein, it is demonstrated that the randomness of p-bit streams generated via the consecutive pulse inputs of pump-probe protocols can be increased by the deliberate incorporation of metal nanoparticles (NPs), which influence the transient dynamics of the nanoscale metallic phase in VO2 Mott switches. Among the vertically stacked Pt-NP-containing VO2 threshold switches, those with higher Pt NP density show a considerably wider range of p-bit operation (e.g., up to ≈300% increase in ΔVprobe upon going from (Pt NP/VO2)0 to (Pt NP/VO2)11) and can therefore be operated under the conditions of high speed (400 kbit s-1), low power consumption (14 nJ per bit), and high stability (>105 200 bits) for p-bit generation. Thus, the study presents a novel strategy that exploits nanoscale phase control to maximize the generation of nondeterministic information sources for energy-efficient probabilistic computing hardware.

Efficient Spiking Neural Networks with Biologically Similar Lithium-Ion Memristor Neurons

Benefiting from the brain-inspired event-driven feature and asynchronous sparse coding approach, spiking neural networks (SNNs) are becoming a potentially energy-efficient replacement for conventional artificial neural networks. However, neuromorphic devices used to construct SNNs persistently result in considerable energy consumption owing to the absence of sufficient biological parallels. Drawing inspiration from the transport nature of Na+ and K+ in synapses, here, a Li-based memristor (LixAlOy) was proposed to emulate the biological synapse, leveraging the similarity of Li as a homologous main group element to Na and K. The Li-based memristor exhibits ∼8 ns ultrafast operating speed, 1.91 and 0.72 linearity conductance modulation, and reproducible switching behavior, enabled by lithium vacancies forming a conductive filament mechanism. Moreover, these memristors are capable of simulating fundamental behaviors of a biological synapse, including long-term potentiation and long-term depression behaviors. Most importantly, a threshold-tunable leaky integrate-and-fire (TT-LIF) neuron is built using LixAlOy memristors, successfully integrating synaptic signals from both temporal and spatial levels and achieving an optimal threshold of SNNs. A computationally efficient TT-LIF-based SNN algorithm is also implemented for image recognition schemes, featuring a high recognition rate of 90.1% and an ultralow firing rate of 0.335%, which is 4 times lower than those of other memristor-based SNNs. Our studies reveal the ion dynamics mechanism of the LixAlOy memristor and confirm its potential in rapid switching and the construction of SNNs.

Spike rate dependent synaptic characteristics in lamellar, multilayered alpha-MoO3 based two-terminal devices - efficient way to control the synaptic amplification

Brain-inspired computing systems require a rich variety of neuromorphic devices using multi-functional materials operating at room temperature. Artificial synapses which can be operated using optical and electrical stimuli are in high demand. In this regard, layered materials have attracted a lot of attention due to their tunable energy gap and exotic properties. In the current study, we report the growth of layered MoO3 using the chemical vapor deposition (CVD) technique. MoO3 has an energy gap of 3.22 eV and grows with a large aspect ratio, as seen through optical and scanning electron microscopy. We used transmission electron microscopy (TEM) and X-ray photoelectron spectroscopy for complete characterisation. The two-terminal devices using platinum (Pt/MoO3/Pt) exhibit superior memory with the high-resistance state (HRS) and low-resistance state (LRS) differing by a large resistance (∼MΩ). The devices also show excellent synaptic characteristics. Both optical and electrical pulses can be utilised to stimulate the synapse. Consistent learning (potentiation) and forgetting (depression) curves are measured. Transition from long term depression to long term potentiation can be achieved using the spike frequency dependent pulsing scheme. We have found that the amplification of postsynaptic current can be tuned using such frequency dependent spikes. This will help us to design neuromorphic devices with the required synaptic amplification.

Computation via Neuron-like Spiking in Percolating Networks of Nanoparticles

The biological brain is a highly efficient computational system in which information processing is performed via electrical spikes. Neuromorphic computing systems that work on similar principles could support the development of the next generation of artificial intelligence and, in particular, enable low-power edge computing. Percolating networks of nanoparticles (PNNs) have previously been shown to exhibit critical spiking behavior, with promise for highly efficient natural computation. Here we employ a rate coding scheme to show that PNNs can perform Boolean operations and image classification. Near perfect accuracy is achieved in both tasks by manipulating the spiking activity using certain control voltages. We demonstrate that the key to successful computation is that nanoscale tunnel gaps within the percolating networks transform input data through a powerful modulus-like nonlinearity. These results provide a basis for implementation of further computational schemes that exploit the brain-like criticality of these networks.

Realizing avalanche criticality in neuromorphic networks on a 2D hBN platform

Networks and systems which exhibit brain-like behavior can analyze information from intrinsically noisy and unstructured data with very low power consumption. Such characteristics arise due to the critical nature and complex interconnectivity of the brain and its neuronal network. We demonstrate a system comprising of multilayer hexagonal boron nitride (hBN) films contacted with silver (Ag), which can uniquely host two different self-assembled networks, which are self-organized at criticality (SOC). This system shows bipolar resistive switching between the high resistance state (HRS) and the low resistance state (LRS). In the HRS, Ag clusters (nodes) intercalate in the van der Waals gaps of hBN forming a network of tunnel junctions, whereas the LRS contains a network of Ag filaments. The temporal avalanche dynamics in both these states exhibit power-law scaling, long-range temporal correlation, and SOC. These networks can be tuned from one to another with voltage as a control parameter. For the first time, two different neural networks are realized in a single CMOS compatible, 2D material platform.

Criticality and Neuromorphic Sensing in a Single Memristor

Resistive random access memory (RRAM) is an important technology for both data storage and neuromorphic computation, where the dynamics of nanoscale conductive filaments lies at the core of the technology. Here, we analyze the current noise of various silicon-based memristors that involves the creation of a percolation path at the intermediate phase of filament growth. Remarkably, we find that these atomic switching events follow scale-free avalanche dynamics with exponents satisfying the criteria for criticality. We further prove that the switching dynamics are universal and show little dependence on device sizes or material features. Utilizing criticality in memristors, we simulate the functionality of hair cells in auditory sensory systems by observing the frequency selectivity of input stimuli with tunable characteristic frequency. We further demonstrate a single-memristor-based sensing primitive for representation of input stimuli that exceeds the theoretical limits dictated by the Nyquist-Shannon theorem.

Dynamical stochastic simulation of complex electrical behavior in neuromorphic networks of metallic nanojunctions

Nanostructured Au films fabricated by the assembling of nanoparticles produced in the gas phase have shown properties suitable for neuromorphic computing applications: they are characterized by a non-linear and non-local electrical behavior, featuring switches of the electric resistance whose activation is typically triggered by an applied voltage over a certain threshold. These systems can be considered as complex networks of metallic nanojunctions where thermal effects at the nanoscale cause the continuous rearrangement of regions with low and high electrical resistance. In order to gain a deeper understanding of the electrical properties of this nano granular system, we developed a model based on a large three dimensional regular resistor network with non-linear conduction mechanisms and stochastic updates of conductances. Remarkably, by increasing enough the number of nodes in the network, the features experimentally observed in the electrical conduction properties of nanostructured gold films are qualitatively reproduced in the dynamical behavior of the system. In the activated non-linear conduction regime, our model reproduces also the growing trend, as a function of the subsystem size, of quantities like Mutual and Integrated Information, which have been extracted from the experimental resistance series data via an information theoretic analysis. This indicates that nanostructured Au films (and our model) possess a certain degree of activated interconnection among different areas which, in principle, could be exploited for neuromorphic computing applications.