Neuromorphic Engineering: From Biological to Spike-Based Hardware Nervous Systems

Robust Piezoelectricity with Spontaneous Polarization in Monolayer Tellurene and Multilayer Tellurium Film at Room Temperature for Reliable Memory

Robust neuromorphic computing in the Big Data era calls for long-term stable crossbar-array memory cells; however, the elemental segregation in the switch unit and memory unit that inevitably occurs upon cycling breaks the compositional and structural stability, making the whole memory cell a failure. Searching for a novel material without segregation that can be used for both switch and memory units is the major concern to fabricate robust and reliable nonvolatile cross-array memory cells. Tellurium (Te) is found recently to be the only peculiar material without segregation for switching, but the memory function has not been demonstrated yet. Herein, apparent piezoelectricity is experimentally confirmed with spontaneous polarization behaviors in elementary 2D Te, even in monolayer tellurene (0.4 nm), due to the highly oriented polarization of the molecular structure and the non-centrosymmetric lattice structure. A large memory window of 7000, a low working voltage of 2 V, and high on switching current up to 36.6 µA µm^-1 are achieved in the as-fabricated Te-based memory device, revealing the great promise of Te for both switching and memory units in one cell without segregation. The piezoelectric Te with spontaneous polarization provides a platform to build robust, reliable, and high-density logic-in-memory chips in neuromorphic computing.

Field-Created Coordinate Cation Bridges Enable Conductance Modulation and Artificial Synapse within Metal Nanoparticles

When metal nanoparticles are functionalized with charged ligands, the movement of counterions and conduction electrons is coupled, which enables us to develop electronic devices, including diodes, transistors, and logic gates, but dynamically modulating the conductivity of a synaptic device within these materials has proved challenging. Here we show that an artificial synapse can be created from thin films of functionalized metal nanoparticles using an active silver electrode. The electric-field-injected Ag⁺ coordinates with carboxyl ligands that sets up a conduction bridge to increase the nanoparticle conductivity by reducing the electron tunneling/hopping energy barriers. The dynamic modulation of conductivity allows us to implement several important synaptic functions such as potentiation/depression, paired-pulse facilitation, learning behaviors including short-term to long-term memory transition, self-learning, and massed leaning vs spaced learning. Finally, based on the nonvolatile characteristics, the metal nanoparticle synapse is used to build a single-layer hardware spiking neural network (SNN) for pattern recognition.

Synaptic Segmented Transistor with Improved Linearity by Schottky Junctions and Accelerated Speed by Double-Layered Nitride

Neuromorphic devices have been extensively studied to overcome the limitations of a von Neumann system for artificial intelligence. A synaptic device is one of the most important components in the hardware integration for a neuromorphic system because a number of synaptic devices can be connected to a neuron with compactness as high as possible. Therefore, synaptic devices using silicon-based memory, which are advantageous for a high packing density and mass production due to matured fabrication technologies, have attracted considerable attention. In this study, a segmented transistor devoted to an artificial synapse is proposed for the first time to improve the linearity of the potentiation and depression (P/D). It is a complementary metal oxide semiconductor (CMOS)-compatible device that harnesses both non-ohmic Schottky junctions of the source and drain for improved weight linearity and double-layered nitride for enhanced speed. It shows three distinct and unique segments in drain current-gate voltage transfer characteristics induced by Schottky junctions. In addition, the different stoichiometries of Si_xN_y for a double-layered nitride is utilized as a charge trap layer for boosting the operation speed. This work can bring the industry potentially one step closer to realizing the mass production of hardware-based synaptic devices in the future.

Polarity-Differentiated Dielectric Materials in Monolayer Graphene Charge-Regulated Field-Effect Transistors for an Artificial Reflex Arc and Pain-Modulation System of the Spinal Cord

The nervous system is a vital part of organisms to survive and it endows them with remarkable abilities, such as perception, recognition, regulation, learning, and decision-making, by intertwining myriad neurons. To realize such outstanding efficacies and functions, many artificial devices and systems have been investigated to emulate the operating principles of the nervous system. Here, an artificial reflex arc (ARA) and artificial pain modulation system (APMS) are proposed to imitate the unconscious behaviors of the spinal cord. Gd_x O_y - and Al_x O_y -based charge-regulated field-effect transistors (CRFETs) with a monolayer graphene channel are fabricated and adopted as inhibitory and excitatory synapses, respectively, under the same pulse signals to mimic the biological reflex arc through a connection with a poly(vinylidene fluoride-co-trifluoroethylene)-based actuator. Additionally, a memristor is integrated with a CRFET as the interneuron to regulate the Dirac point by controlling the voltage drop on the graphene channel, analogous to the descending pain-inhibition system in the spinal cord, to prevent excessive pain perception. The proposed ARA and APMS provide a significant step forward to realizing the functions of the nervous system, giving promising potential for developing future intelligent alarm systems, neuroprosthetics, and neurorobotics.

Perovskite-Based Memristor with 50-Fold Switchable Photosensitivity for In-Sensor Computing Neural Network

In-sensor computing can simultaneously output image information and recognition results through in-situ visual signal processing, which can greatly improve the efficiency of machine vision. However, in-sensor computing is challenging due to the requirement to controllably adjust the sensor’s photosensitivity. Herein, it is demonstrated a ternary cationic halide Cs_0.05FA_0.81MA_0.14 Pb(I_0.85Br_0.15)₃ (CsFAMA) perovskite, whose External quantum efficiency (EQE) value is above 80% in the entire visible region (400–750 nm), and peak responsibility value at 750 nm reaches 0.45 A/W. In addition, the device can achieve a 50-fold enhancement of the photoresponsibility under the same illumination by adjusting the internal ion migration and readout voltage. A proof-of-concept visually enhanced neural network system is demonstrated through the switchable photosensitivity of the perovskite sensor array, which can simultaneously optimize imaging and recognition results and improve object recognition accuracy by 17% in low-light environments.

From Theoretical Network to Bedside: Translational Application of Brain-Inspired Computing in Clinical Medicine

Advances in the brain-inspired computing space are growing at a rapid rate, and many of these emerging strategies are in the field of neuromorphic control, robotics, and sensor development, just to name a few. These innovations are disruptive in their own right and have numerous, multi-dimensional medical applications within precision medicine, telematics, device development, and informed clinical decision making. For this discussion, I will define brain-inspired computing in the scope of simulating the architecture of the brain and discuss the realization of integrating hardware and other technologies with the applications of medicine, along with the considerations for the regulatory pathway for approval and evaluating the risk/consequences of failure modes. This perspective is a call for continued discussion of the development of a pathway for translating these technologies into medical treatment and diagnostic strategies. The aim is to align with global regulatory bodies and ensure that regulation does not limit the capacity of these emerging innovations while ensuring patient safety and clinical efficacy. It is my perspective that it is and will continue to be critical that these technologies are correctly perceived and understood in the lens of multiple disciplines in order to reach their full potential for medical applications.

A Fully Solution-Printed Photosynaptic Transistor Array with Ultralow Energy Consumption for Artificial-Vision Neural Networks

Photosynaptic organic field-effect transistors (OFETs) represent a viable pathway to develop bionic optoelectronics. However, the high operating voltage and current of traditional photosynaptic OFETs lead to huge energy consumption greater than that of the real biological synapses, hindering their further development in new-generation visual prosthetics and artificial perception systems. Here, a fully solution-printed photosynaptic OFET (FSP-OFET) with substantial energy consumption reduction is reported, where a source Schottky barrier is introduced to regulate charge-carrier injection, and which operates with a fundamentally different mechanism from traditional devices. The FSP-OFET not only significantly lowers the working voltage and current but also provides extraordinary neuromorphic light-perception capabilities. Consequently, the FSP-OFET successfully emulates visual nervous responses to external light stimuli with ultralow energy consumption of 0.07-34 fJ per spike in short-term plasticity and 0.41-19.87 fJ per spike in long-term plasticity, both approaching the energy efficiency of biological synapses (1-100 fJ). Moreover, an artificial optic-neural network made from an 8 × 8 FSP-OFET array on a flexible substrate shows excellent image recognition and reinforcement abilities at a low energy cost. The designed FSP-OFET offers an opportunity to realize photonic neuromorphic functionality with extremely low energy consumption dissipation.

CMOS-compatible compute-in-memory accelerators based on integrated ferroelectric synaptic arrays for convolution neural networks

Convolutional neural networks (CNNs) have gained much attention because they can provide superior complex image recognition through convolution operations. Convolution processes require repeated multiplication and accumulation operations, which are difficult tasks for conventional computing systems. Compute-in-memory (CIM) that uses parallel data processing is an ideal device structure for convolution operations. CIM based on two-terminal synaptic devices with a crossbar structure has been developed, but unwanted leakage current paths and the high-power consumption remain as the challenges. Here, we demonstrate integrated ferroelectric thin-film transistor (FeTFT) synaptic arrays that can provide efficient parallel programming and data processing for CNNs by the selective and accurate control of polarization in the ferroelectric layer. In addition, three-terminal FeTFTs can act as both nonvolatile memory and access device, which tackle issues from two-terminal devices. An integrated FeTFT synaptic array with parallel programming capabilities can perform convolution operations to extract image features with a high-recognition accuracy.

Exploiting Non-idealities of Resistive Switching Memories for Efficient Machine Learning

Novel computing architectures based on resistive switching memories (also known as memristors or RRAMs) have been shown to be promising approaches for tackling the energy inefficiency of deep learning and spiking neural networks. However, resistive switch technology is immature and suffers from numerous imperfections, which are often considered limitations on implementations of artificial neural networks. Nevertheless, a reasonable amount of variability can be harnessed to implement efficient probabilistic or approximate computing. This approach turns out to improve robustness, decrease overfitting and reduce energy consumption for specific applications, such as Bayesian and spiking neural networks. Thus, certain non-idealities could become opportunities if we adapt machine learning methods to the intrinsic characteristics of resistive switching memories. In this short review, we introduce some key considerations for circuit design and the most common non-idealities. We illustrate the possible benefits of stochasticity and compression with examples of well-established software methods. We then present an overview of recent neural network implementations that exploit the imperfections of resistive switching memory, and discuss the potential and limitations of these approaches.

Tailoring conductive inverse opal films with anisotropic elliptical porous patterns for nerve cell orientation

Background

The nervous system is critical to the operation of various organs and systems, while novel methods with designable neural induction remain to exploit.

Results

Here, we present a conductive inverse opal film with anisotropic elliptical porous patterns for nerve orientation induction. The films are fabricated based on polystyrene (PS) inverse opal scaffolds with periodical elliptical porous structure and poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) mixed polyacrylamide (PAAm) polymers fillers. It is demonstrated that the anisotropic elliptical surface topography allows the nerve cells to be induced into orientation connected with the stretching direction. Because of the anisotropic features of the film which can be stretched into different directions, nerve cells can be induced to grow in one or two directions, forming a neural network and promoting the connection of nerve cells. It is worth mentioning that the PEDOT:PSS-doped PAAm hydrogels endow the film with conductive properties, which makes the composite films be a suitable candidate for neurites growth and differentiation.

Conclusions

All these features of the conductive and anisotropic inverse opal films imply their great prospects in biomedical applications.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-022-01340-w.

Covalent Coupling of Porphyrins with Monolayer Graphene for Low-Voltage Synaptic Transistors

Synaptic devices emulating biological synapses are a key building component of artificial neural networks. Porphyrins and graphene, as two kinds of emerging electronic materials, have attracted extensive attention in the research of photoelectric devices due to their excellent structural and functional properties. Herein, we present a photonic synaptic transistor based on porphyrin-graphene covalent hybrids utilizing 5,10,15,20-tetrakis (4-aminophenyl)-21H,23H-porphine and monolayer graphene linked through the diazo addition reaction. The photonic synaptic device successfully simulates several essential biological functions, and the synaptic plasticity can be regulated by adjusting the parameters of light spikes and gate voltages of the device. Moreover, learning and memory behaviors under different wavelengths are studied to imitate the learning efficiency of humans in diverse emotional states. It is worth noting that all the synaptic functions can be realized at a low operating voltage of -10 mV, which is much lower than that required by most reported photonic synaptic devices. These results indicate that covalent coupling products of porphyrins with graphene have broad prospects in the construction of synaptic transistors and may arouse new research advances in neuromorphic devices with ultralow operating voltage and low energy consumption.

Retina-Inspired Self-Powered Artificial Optoelectronic Synapses with Selective Detection in Organic Asymmetric Heterojunctions

The retina, the most crucial unit of the human visual perception system, combines sensing with wavelength selectivity and signal preprocessing. Incorporating energy conversion into these superior neurobiological features to generate core visual signals directly from incoming light under various conditions is essential for artificial optoelectronic synapses to emulate biological processing in the real retina. Herein, self-powered optoelectronic synapses that can selectively detect and preprocess the ultraviolet (UV) light are presented, which benefit from high-quality organic asymmetric heterojunctions with ultrathin molecular semiconducting crystalline films, intrinsic heterogeneous interfaces, and typical photovoltaic properties. These devices exhibit diverse synaptic behaviors, such as excitatory postsynaptic current, paired-pulse facilitation, and high-pass filtering characteristics, which successfully reproduce the unique connectivity among sensory neurons. These zero-power optical-sensing synaptic operations further facilitate a demonstration of image sharpening. Additionally, the charge transfer at the heterojunction interface can be modulated by tuning the gate voltage to achieve multispectral sensing ranging from the UV to near-infrared region. Therefore, this work sheds new light on more advanced retinomorphic visual systems in the post-Moore era.

Artificial Adaptive and Maladaptive Sensory Receptors Based on a Surface-Dominated Diffusive Memristor

A biological receptor serves as sensory transduction from an external stimulus to an electrical signal. It allows humans to better match the environment by filtering out repetitive innocuous information and recognize potentially damaging stimuli through key features, including adaptive and maladaptive behaviors. Herein, for the first time, the authors develop substantial artificial receptors involving both adaptive and maladaptive behaviors using diffusive memristor. Metal-oxide nanorods (NR) as a switching matrix enable the electromigration of an active metal along the surface of the NRs under electrical stimulation, resulting in unique surface-dominated switching dynamics with the advantage of fast Ag migration and fine controllability of the conductive filament. To experimentally demonstrate its potential application, a thermoreceptor system is constructed using memristive artificial receptors. The proposed surface-dominated diffusive memristor allows the direct emulation of the biological receptors, which represents an advance in the bioinspired technology adopted in creating artificial intelligence systems.

An Ultra-Low Power Threshold Voltage Variable Artificial Retina Neuron

An artificial retina neuron is proposed and implemented by CMOS technology. It can be used as an image sensor in the Artificial Intelligence (AI) field with the benefit of ultra-low power consumption. The artificial neuron can generate signals in spike shape with pre-designed frequencies under different light intensities. The power consumption is reduced by removing the film capacitor. The comparator is adopted to improve the stability of the circuit, and the power consumption of the comparator is optimized. The power consumption of the proposed CMOS neuron circuit is suppressed. The ultra-low-power artificial neuron with variable threshold shows a frequency range of 0.8–80 kHz when the input current is varied from 1 pA to 150 pA. The minimum DC power is 35 pW when the input current is 5 pA. The minimum energy of the neuron is 3 fJ. The proposed ultra-low-power artificial retina neuron has wide potential applications in the field of AI.

A Smarter Pavlovian Dog with Optically Modulated Associative Learning in an Organic Ferroelectric Neuromem

Associative learning is a critical learning principle uniting discrete ideas and percepts to improve individuals' adaptability. However, enabling high tunability of the association processes as in biological counterparts and thus integration of multiple signals from the environment, ideally in a single device, is challenging. Here, we fabricate an organic ferroelectric neuromem capable of monadically implementing optically modulated associative learning. This approach couples the photogating effect at the interface with ferroelectric polarization switching, enabling highly tunable optical modulation of charge carriers. Our device acts as a smarter Pavlovian dog exhibiting adjustable associative learning with the training cycles tuned from thirteen to two. In particular, we obtain a large output difference (>103), which is very similar to the all-or-nothing biological sensory/motor neuron spiking with decrementless conduction. As proof-of-concept demonstrations, photoferroelectric coupling-based applications in cryptography and logic gates are achieved in a single device, indicating compatibility with biological and digital data processing.

Multi-Terminal Memristive Devices Enabling Tunable Synaptic Plasticity in Neuromorphic Hardware: A Mini-Review

Neuromorphic computing based on spiking neural networks has the potential to significantly improve on-line learning capabilities and energy efficiency of artificial intelligence, specially for edge computing. Recent progress in computational neuroscience have demonstrated the importance of heterosynaptic plasticity for network activity regulation and memorization. Implementing heterosynaptic plasticity in hardware is thus highly desirable, but important materials and engineering challenges remain, calling for breakthroughs in neuromorphic devices. In this mini-review, we propose an overview of the latest advances in multi-terminal memristive devices on silicon with tunable synaptic plasticity, enabling heterosynaptic plasticity in hardware. The scalability and compatibility of the devices with industrial complementary metal oxide semiconductor (CMOS) technologies are discussed.

Ionic Sieving Through One-Atom-Thick 2D Material Enables Analog Nonvolatile Memory for Neuromorphic Computing

The first report on ion transport through atomic sieves of atomically thin 2D material is provided to solve critical limitations of electrochemical random-access memory (ECRAM) devices. Conventional ECRAMs have random and localized ion migration paths; as a result, the analog switching efficiency is inadequate to perform in-memory logic operations. Herein ion transport path scaled down to the one-atom-thick (≈0.33 nm) hexagonal boron nitride (hBN), and the ionic transport area is confined to a small pore (≈0.3 nm² ) at the single-hexagonal ring. One-atom-thick hBN has ion-permeable pores at the center of each hexagonal ring due to weakened electron cloud and highly polarized B-N bond. The experimental evidence indicates that the activation energy barrier for H⁺ ion transport through single-layer hBN is ≈0.51 eV. Benefiting from the controlled ionic sieving through single-layer hBN, the ECRAMs exhibit superior nonvolatile analog switching with good memory retention and high endurance. The proposed approach enables atomically thin 2D material as an ion transport layer to regulate the switching of various ECRAM devices for artificial synaptic electronics.

Dual-Mode Learning of Ambipolar Synaptic Phototransistor Based on 2D Perovskite/Organic Heterojunction for Flexible Color Recognizable Visual System

Artificial intelligence vision systems (AIVSs) with information sensing, processing, and storage functions are increasingly gaining attention in the science and technology community. Although synapse phototransistor (SPT) is one of the essential components in AIVSs, solution-processed large-area photonic synapses that can detect and recognize multi-wavelength light are highly desirable. One of the major challenges in this area is the inability of the available materials to distinguish colors from the visible light to the near-infrared (NIR) light for single carrier (hole-only or electron-only) SPTs owing to lack of cognitive elements. Herein, 2D perovskite/organic heterojunction (PEA₂ SnI₄ /Y6) ambipolar SPTs (POASPTs) are developed via solution process. The POASPTs can display dual-mode learning process, which can convert light signals into postsynaptic currents with excitement/inhibition modes (hole-transporting region) or inhibition/excitement (electron-transporting region). The POASPTs exhibit high responsivity to visible light (10⁴ A W^-1 ) and NIR light (200 A W^-1 ), and effectively perform learning and memory simultaneously. The flexible POASPT arrays can successfully recognize the images of different colors of light. This study reveals that the fabricated POASPTs have great potentials in the development of large-area, high-efficiency, and low-cost AIVSs.

Low-power consumption light-stimulated synaptic transistors based on natural carotene and organic semiconductors

Developing synaptic devices with environment-friendly materials is a promising research direction. Here, light-stimulated synaptic transistors based on natural carotene and organic semiconductors were developed. Several important functions similar to biological synapses were realized and an ultra-low power consumption of 3.4 × 10^-18 J was achieved.

2D oriented covalent organic frameworks for alcohol-sensory synapses

Resistive random access memories (RRAMs) based on the electrochemical metallization mechanism (ECM) have potential applications in high-density data storage and efficient neuromorphic computing. However, the high variability of ECM devices still hinders their application in artificial intelligence owing to the random formation of conductive filaments (CFs). Here, we demonstrate 2D covalent organic framework (COF) RRAM with electroforming-free resistive switching behavior, low spatial/temporal variations, and excellent retention capability up to 10⁵ s. The one-dimensional channels of the oriented COF-5 film can not only confine the shape of filaments but also modulate the transition direction of Ag ions. Moreover, alcohol vapors could activate the device to achieve gas-mediated multilevel resistive switching since COF materials can absorb small molecules through host guest interactions to vary the conductivity. An alcohol gas recognition system constructed by integrating the COF RRAM as a sensor and filter part with the k-nearest neighbors (KNN) algorithm as a classifier was demonstrated with a recognition accuracy of 87.2%. Furthermore, the effect of alcohol inhibition stimulation in the human nervous system is successfully emulated by the COF RRAM.