Mimicking the brain functions of learning, forgetting and explicit/implicit memories with SrTiO3-based memristive devices

Programmable Retention Characteristics in MoS2-Based Atomristors for Neuromorphic and Reservoir Computing Systems

In this study, we investigate the coexistence of short- and long-term memory effects owing to the programmable retention characteristics of a two-dimensional Au/MoS2/Au atomristor device and determine the impact of these effects on synaptic properties. This device is constructed using bilayer MoS2 in a crossbar structure. The presence of both short- and long-term memory characteristics is proposed by using a filament model within the bilayer transition-metal dichalcogenide. Short- and long-term properties are validated based on programmable multilevel retention tests. Moreover, we confirm various synaptic characteristics of the device, demonstrating its potential use as a synaptic device in a neuromorphic system. Excitatory postsynaptic current, paired-pulse facilitation, spike-rate-dependent plasticity, and spike-number-dependent plasticity synaptic applications are implemented by operating the device at a low-conductance level. Furthermore, long-term potentiation and depression exhibit symmetrical properties at high-conductance levels. Synaptic learning and forgetting characteristics are emulated using programmable retention properties and composite synaptic plasticity. The learning process of artificial neural networks is used to achieve high pattern recognition accuracy, thereby demonstrating the suitability of the use of the device in a neuromorphic system. Finally, the device is used as a physical reservoir with time-dependent inputs to realize reservoir computing by using short-term memory properties. Our study reveals that the proposed device can be applied in artificial intelligence-based computing applications by utilizing its programmable retention properties.

Quantum Topological Neuristors for Advanced Neuromorphic Intelligent Systems

Neuromorphic artificial intelligence systems are the future of ultrahigh performance computing clusters to overcome complex scientific and economical challenges. Despite their importance, the advancement in quantum neuromorphic systems is slow without specific device design. To elucidate biomimicking mammalian brain synapses, a new class of quantum topological neuristors (QTN) with ultralow energy consumption (pJ) and higher switching speed (µs) is introduced. Bioinspired neural network characteristics of QTNs are the effects of edge state transport and tunable energy gap in the quantum topological insulator (QTI) materials. With augmented device and QTI material design, top notch neuromorphic behavior with effective learning-relearning-forgetting stages is demonstrated. Critically, to emulate the real-time neuromorphic efficiency, training of the QTNs is demonstrated with simple hand gesture game by interfacing them with artificial neural networks to perform decision-making operations. Strategically, the QTNs prove the possession of incomparable potential to realize next-gen neuromorphic computing for the development of intelligent machines and humanoids.

Optically modulated ionic conductivity in a hydrogel for emulating synaptic functions

Ion-conductive hydrogels, with ions as signal carriers, have become promising candidates to construct functional ionotronics for sensing, actuating, and robotics engineering. However, rational modulation of ionic migration to mimic biological information processing, including learning and memory, remains challenging to be realized in hydrogel materials. Here, we develop a hybrid hydrogel with optically modulated ionic conductivity to emulate the functions of a biological synapse. Through a responsive supramolecular approach, optical stimuli can trigger the release of mobile ions for tuning the conductivity of the hydrogel, which is analogous to the modulation of synaptic plasticity. As a proof of concept, this hydrogel can be used as an information processing unit to perceive different optical stimuli and regulate the grasping motion of a robotic hand, performing logical motion feedback with "learning-experience" function. Our ionic hydrogel provides a valuable strategy toward developing bioinspired ionotronic systems and pushes forward the functional applications of hydrogel materials.

DFT-1/2 and shell DFT-1/2 methods: electronic structure calculation for semiconductors at LDA complexity

It is known that the Kohn-Sham eigenvalues do not characterize experimental excitation energies directly, and the band gap of a semiconductor is typically underestimated by local density approximation (LDA) of density functional theory (DFT). An embarrassing situation is that one usually uses LDA+Ufor strongly correlated materials with rectified band gaps, but for non-strongly-correlated semiconductors one has to resort to expensive methods like hybrid functionals orGW. In spite of the state-of-the-art meta-generalized gradient approximation functionals like TB-mBJ and SCAN, methods with LDA-level complexity to rectify the semiconductor band gaps are in high demand. DFT-1/2 stands as a feasible approach and has been more widely used in recent years. In this work we give a detailed derivation of the Slater half occupation technique, and review the assumptions made by DFT-1/2 in semiconductor band structure calculations. In particular, the self-energy potential approach is verified through mathematical derivations. The aims, features and principles of shell DFT-1/2 for covalent semiconductors are also accounted for in great detail. Other developments of DFT-1/2 including conduction band correction, DFT+A-1/2, empirical formula for the self-energy potential cutoff radius, etc, are further reviewed. The relations of DFT-1/2 to hybrid functional, sX-LDA,GW, self-interaction correction, scissor's operator as well as DFT+Uare explained. Applications, issues and limitations of DFT-1/2 are comprehensively included in this review.

Molecular-Layer-Defined Asymmetric Schottky Contacts in Organic Planar Diodes for Self-Powered Optoelectronic Synapses

Optoelectronic synapses have been utilized as neuromorphic vision sensors for image preprocessing in artificial visual systems. Self-powered optoelectronic synapses, which can directly convert optical power into electrical power, are promising for practical applications. The Schottky junction tends to be a promising candidate as the energy source for electrical operations. However, fully utilizing the potential of Schottky barriers is still challenging. Herein, organic self-powered optoelectronic synapses with planar diode architecture are fabricated, which can simultaneously sense and process ultraviolet (UV) signals. The photovoltaic operations are facilitated by the built-in potential originating from the molecular-layer-defined asymmetric Schottky contacts. Diverse synaptic behaviors under UV light stimulation without external power supplies are facilitated by the interfacial carrier-capturing layer, which emulates the membranes of synapses. Furthermore, retina-inspired image preprocessing functions are demonstrated on the basis of synaptic plasticity. Therefore, our devices provide the potential for the development of power-efficient and advanced artificial visual systems.

Learning to Approximate Functions Using Nb-Doped SrTiO3 Memristors

Memristors have attracted interest as neuromorphic computation elements because they show promise in enabling efficient hardware implementations of artificial neurons and synapses. We performed measurements on interface-type memristors to validate their use in neuromorphic hardware. Specifically, we utilized Nb-doped SrTiO3 memristors as synapses in a simulated neural network by arranging them into differential synaptic pairs, with the weight of the connection given by the difference in normalized conductance values between the two paired memristors. This network learned to represent functions through a training process based on a novel supervised learning algorithm, during which discrete voltage pulses were applied to one of the two memristors in each pair. To simulate the fact that both the initial state of the physical memristive devices and the impact of each voltage pulse are unknown we injected noise into the simulation. Nevertheless, discrete updates based on local knowledge were shown to result in robust learning performance. Using this class of memristive devices as the synaptic weight element in a spiking neural network yields, to our knowledge, one of the first models of this kind, capable of learning to be a universal function approximator, and strongly suggests the suitability of these memristors for usage in future computing platforms.

The Electronic Properties of Extended Defects in SrTiO3—A Case Study of a Real Bicrystal Boundary

This study investigates the impact of extended defects such as dislocations on the electronic properties of SrTiO3 by using a 36.8° bicrystal as a model system. In order to evaluate the hypothesis that dislocations can serve as preferential reduction sites, which has been proposed in the literature on the basis of ab initio simulations, as well as on experiments employing local-conductivity atomic force microscopy (LC-AFM), detailed investigations of the bicrystal boundary are conducted. In addition to LC-AFM, fluorescence lifetime imaging microscopy (FLIM) is applied herein as a complementary method for mapping the local electronic properties on the microscale. Both techniques confirm that the electronic structure and electronic transport in dislocation-rich regions significantly differ from those of undistorted SrTiO3. Upon thermal reduction, a further confinement of conductivity to the bicrystal boundary region was found, indicating that extended defects can indeed be regarded as the origin of filament formation. This leads to the evolution of inhomogeneous properties of defective SrTiO3 on the nano- and microscales.

Flexible Neuromorphic Electronics for Computing, Soft Robotics, and Neuroprosthetics

Flexible neuromorphic electronics that emulate biological neuronal systems constitute a promising candidate for next-generation wearable computing, soft robotics, and neuroprosthetics. For realization, with the achievement of simple synaptic behaviors in a single device, the construction of artificial synapses with various functions of sensing and responding and integrated systems to mimic complicated computing, sensing, and responding in biological systems is a prerequisite. Artificial synapses that have learning ability can perceive and react to events in the real world; these abilities expand the neuromorphic applications toward health monitoring and cybernetic devices in the future Internet of Things. To demonstrate the flexible neuromorphic systems successfully, it is essential to develop artificial synapses and nerves replicating the functionalities of the biological counterparts and satisfying the requirements for constructing the elements and the integrated systems such as flexibility, low power consumption, high-density integration, and biocompatibility. Here, the progress of flexible neuromorphic electronics is addressed, from basic backgrounds including synaptic characteristics, device structures, and mechanisms of artificial synapses and nerves, to applications for computing, soft robotics, and neuroprosthetics. Finally, future research directions toward wearable artificial neuromorphic systems are suggested for this emerging area.

Gate Tuning of Synaptic Functions Based on Oxygen Vacancy Distribution Control in Four-Terminal TiO2-x Memristive Devices

Recent developments in artificial intelligence technology has facilitated advances in neuromorphic computing. Electrical elements mimicking the role of synapses are crucial building blocks for neuromorphic computers. Although various types of two-terminal memristive devices have emerged in the mainstream of synaptic devices, a hetero-synaptic artificial synapse, i.e., one with modulatable plasticity induced by multiple connections of synapses, is intriguing. Here, a synaptic device with tunable synapse plasticity is presented that is based on a simple four-terminal rutile TiO_2−x single-crystal memristor. In this device, the oxygen vacancy distribution in TiO_2−x and the associated bulk carrier conduction can be used to control the resistance of the device. There are two diagonally arranged pairs of electrodes with distinct functions: one for the read/write operation, the other for the gating operation. This arrangement enables precise control of the oxygen vacancy distribution. Microscopic analysis of the Ti valence states in the device reveals the origin of resistance switching phenomena to be an electrically driven redistribution of oxygen vacancies with no changes in crystal structure. Tuning protocols for the write and the gate voltage applications enable high precision control of resistance, or synaptic plasticity, paving the way for the manipulation of learning efficiency through neuromorphic devices.

Quasi-Hodgkin-Huxley Neurons with Leaky Integrate-and-Fire Functions Physically Realized with Memristive Devices

Artificial neurons with functions such as leaky integrate-and-fire (LIF) and spike output are essential for brain-inspired computation with high efficiency. However, previously implemented artificial neurons, e.g., Hodgkin-Huxley (HH) neurons, integrate-and-fire (IF) neurons, and LIF neurons, only achieve partial functionality of a biological neuron. In this work, quasi-HH neurons with leaky integrate-and-fire functions are physically demonstrated with a volatile memristive device, W/WO₃ /poly(3,4-ethylenedioxythiophene): polystyrene sulfonate/Pt. The resistive switching behavior of the device can be attributed to the migration of protons, unlike the migration of oxygen ions normally involved in oxide-based memristors. With multifunctions similar to their biological counterparts, quasi-HH neurons are advantageous over the reported HH and LIF neurons, demonstrating their potential for neuromorphic computing applications.

Oxygen vacancy migration/diffusion induced synaptic plasticity in a single titanate nanobelt

Neuromorphic computational systems that emulate biological synapses in the human brain are fundamental in the development of artificial intelligence protocols beyond the standard von Neumann architecture. Such systems require new types of building blocks, such as memristors that access a quasi-continuous and wide range of conductive states, which is still an obstacle for the realization of high-efficiency and large-capacity learning in neuromorphoric simulation. Here, we introduce hydrogen and sodium titanate nanobelts, the intermediate products of hydrothermal synthesis of TiO2 nanobelts, to emulate the synaptic behavior. Devices incorporating a single titanate nanobelt demonstrate robust and reliable synaptic functions, including excitatory postsynaptic current, paired pulse facilitation, short term plasticity, potentiation and depression, as well as learning-forgetting behavior. In particular, the gradual modulation of conductive states in the single nanobelt device can be achieved by a large number of identical pulses. The mechanism for synaptic functionality of the titanate nanobelt device is attributed to the competition between an electric field driven migration of oxygen vacancies and a thermally induced spontaneous diffusion. These results provide insight into the potential use of titanate nanobelts in synaptic applications requiring continuously addressable states coupled with high processing efficiency.

Silicon synaptic transistor for hardware-based spiking neural network and neuromorphic system

Brain-inspired neuromorphic systems have attracted much attention as new computing paradigms for power-efficient computation. Here, we report a silicon synaptic transistor with two electrically independent gates to realize a hardware-based neural network system without any switching components. The spike-timing dependent plasticity characteristics of the synaptic devices are measured and analyzed. With the help of the device model based on the measured data, the pattern recognition capability of the hardware-based spiking neural network systems is demonstrated using the modified national institute of standards and technology handwritten dataset. By comparing systems with and without inhibitory synapse part, it is confirmed that the inhibitory synapse part is an essential element in obtaining effective and high pattern classification capability.