Tunnel junction based memristors as artificial synapses

Photonics for Neuromorphic Computing: Fundamentals, Devices, and Opportunities

In the dynamic landscape of Artificial Intelligence (AI), two notable phenomena are becoming predominant: the exponential growth of large AI model sizes and the explosion of massive amount of data. Meanwhile, scientific research such as quantum computing and protein synthesis increasingly demand higher computing capacities. As the Moore's Law approaches its terminus, there is an urgent need for alternative computing paradigms that satisfy this growing computing demand and break through the barrier of the von Neumann model. Neuromorphic computing, inspired by the mechanism and functionality of human brains, uses physical artificial neurons to do computations and is drawing widespread attention. This review studies the expansion of optoelectronic devices on photonic integration platforms that has led to significant growth in photonic computing, where photonic integrated circuits (PICs) have enabled ultrafast artificial neural networks (ANN) with sub-nanosecond latencies, low heat dissipation, and high parallelism. In particular, various technologies and devices employed in neuromorphic photonic AI accelerators, spanning from traditional optics to PCSEL lasers are examined. Lastly, it is recognized that existing neuromorphic technologies encounter obstacles in meeting the peta-level computing speed and energy efficiency threshold, and potential approaches in new devices, fabrication, materials, and integration to drive innovation are also explored. As the current challenges and barriers in cost, scalability, footprint, and computing capacity are resolved one-by-one, photonic neuromorphic systems are bound to co-exist with, if not replace, conventional electronic computers and transform the landscape of AI and scientific computing in the foreseeable future.

Voltage-dependent plasticity of spin-polarized conductance in phenyl-based single-molecule magnetic tunnel junctions

Synaptic strengths between neurons in brain networks are highly adaptive due to synaptic plasticity. Spike-timing-dependent plasticity (STDP) is a form of synaptic plasticity induced by temporal correlations between the firing activity of neurons. The development of experimental techniques in recent years enabled the realization of brain-inspired neuromorphic devices. Particularly, magnetic tunnel junctions (MTJs) provide a suitable means for the implementation of learning processes in molecular junctions. Here, we first considered a two-neuron motif subjected to STDP. By employing theoretical analysis and computer simulations we showed that the dynamics and emergent structure of the motif can be predicted by introducing an effective two-neuron synaptic conductance. Then, we considered a phenyl-based single-molecule MTJ connected to two ferromagnetic (FM) cobalt electrodes and investigated its electrical properties using the non-equilibrium Green’s function (NEGF) formalism. Similar to the two-neuron motif, we introduced an effective spin-polarized conductance in the MTJ. Depending on the polarity, frequency and strength of the bias voltage applied to the MTJ, the system can learn input signals by adaptive changes of the effective conductance. Interestingly, this voltage-dependent plasticity is an intrinsic property of the MTJ where its behavior is reminiscent of the classical temporally asymmetric STDP. Furthermore, the shape of voltage-dependent plasticity in the MTJ is determined by the molecule-electrode coupling strength or the length of the molecule. Our results may be relevant for the development of single-molecule devices that capture the adaptive properties of synapses in the brain.

3D Stackable Synaptic Transistor for 3D Integrated Artificial Neural Networks

Although they have attracted enormous attention in recent years, software-based and two-dimensional hardware-based artificial neural networks (ANNs) may consume a great deal of power. Because there will be numerous data transmissions through a long interconnection for learning, power consumption in the interconnect will be an inevitable problem for low-power computing. Therefore, we suggest and report 3D stackable synaptic transistors for 3D ANNs, which would be the strongest candidate in future computing systems by minimizing power consumption in the interconnection. To overcome the problems of enormous power consumption, it might be necessary to introduce a 3D stackable ANN platform. With this structure, short vertical interconnection can be realized between the top and bottom devices, and the integration density can be significantly increased for integrating numerous neuromorphic devices. In this paper, we suggest and show the feasibility of monolithic 3D integration of synaptic devices using the channel layer transfer method through a wafer bonding technique. Using a low-temperature processible III-V and composite oxide (Al₂O₃/HfO₂/Al₂O₃)-based weight storage layer, we successfully demonstrated synaptic transistors showing good linearity (α_p/α_d = 1.8/0.5), a high transconductance ratio (6300), and very good stability. High learning accuracy of 97% was obtained in the training of 1 million MNIST images based on the device characteristics.

Skyrmions in Magnetic Tunnel Junctions

In this work, we demonstrate that skyrmions can be nucleated in the free layer of a magnetic tunnel junction (MTJ) with Dzyaloshinskii-Moriya interactions (DMIs) by a spin-polarized current with the assistance of stray fields from the pinned layer. The size, stability, and number of created skyrmions can be tuned by either the DMI strength or the stray field distribution. The interaction between the stray field and the DMI effective field is discussed. A device with multilevel tunneling magnetoresistance is proposed, which could pave the ways for skyrmion-MTJ-based multibit storage and artificial neural network computation. Our results may facilitate the efficient nucleation and electrical detection of skyrmions.

Solid-State Synapse Based on Magnetoelectrically Coupled Memristor

Brain-inspired computing architectures attempt to emulate the computations performed in the neurons and the synapses in the human brain. Memristors with continuously tunable resistances are ideal building blocks for artificial synapses. Through investigating the memristor behaviors in a La_0.7Sr_0.3MnO₃/BaTiO₃/La_0.7Sr_0.3MnO₃ multiferroic tunnel junction, it was found that the ferroelectric domain dynamics characteristics are influenced by the relative magnetization alignment of the electrodes, and the interfacial spin polarization is manipulated continuously by ferroelectric domain reversal, enriching our understanding of the magnetoelectric coupling fundamentally. This creates a functionality that not only the resistance of the memristor but also the synaptic plasticity form can be further manipulated, as demonstrated by the spike-timing-dependent plasticity investigations. Density functional theory calculations are carried out to describe the obtained magnetoelectric coupling, which is probably related to the Mn-Ti intermixing at the interfaces. The multiple and controllable plasticity characteristic in a single artificial synapse, to resemble the synaptic morphological alteration property in a biological synapse, will be conducive to the development of artificial intelligence.

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

To implement the complex brain functions of learning, forgetting and memory in a single electronic device is very advantageous for realizing artificial intelligence. As a proof of concept, memristive devices with a simple structure of Ni/Nb-SrTiO₃/Ti were investigated in this work. The functions of learning, forgetting and memory were successfully mimicked using the memristive devices, and the "time-saving" effect of implicit memory was also demonstrated. The physics behind the brain functions is simply the modulation of the Schottky barrier at the Ni/SrTiO₃ interface. The realization of various psychological functions in a single device simplifies the construction of the artificial neural network and facilitates the advent of artificial intelligence.

Trends and Challenges in Neuroengineering: Toward "Intelligent" Neuroprostheses through Brain-"Brain Inspired Systems" Communication

Future technologies aiming at restoring and enhancing organs function will intimately rely on near-physiological and energy-efficient communication between living and artificial biomimetic systems. Interfacing brain-inspired devices with the real brain is at the forefront of such emerging field, with the term "neurobiohybrids" indicating all those systems where such interaction is established. We argue that achieving a "high-level" communication and functional synergy between natural and artificial neuronal networks in vivo, will allow the development of a heterogeneous world of neurobiohybrids, which will include "living robots" but will also embrace "intelligent" neuroprostheses for augmentation of brain function. The societal and economical impact of intelligent neuroprostheses is likely to be potentially strong, as they will offer novel therapeutic perspectives for a number of diseases, and going beyond classical pharmaceutical schemes. However, they will unavoidably raise fundamental ethical questions on the intermingling between man and machine and more specifically, on how deeply it should be allowed that brain processing is affected by implanted "intelligent" artificial systems. Following this perspective, we provide the reader with insights on ongoing developments and trends in the field of neurobiohybrids. We address the topic also from a "community building" perspective, showing through a quantitative bibliographic analysis, how scientists working on the engineering of brain-inspired devices and brain-machine interfaces are increasing their interactions. We foresee that such trend preludes to a formidable technological and scientific revolution in brain-machine communication and to the opening of new avenues for restoring or even augmenting brain function for therapeutic purposes.

Magnetic Tunnel Junction Mimics Stochastic Cortical Spiking Neurons

Brain-inspired computing architectures attempt to mimic the computations performed in the neurons and the synapses in the human brain in order to achieve its efficiency in learning and cognitive tasks. In this work, we demonstrate the mapping of the probabilistic spiking nature of pyramidal neurons in the cortex to the stochastic switching behavior of a Magnetic Tunnel Junction in presence of thermal noise. We present results to illustrate the efficiency of neuromorphic systems based on such probabilistic neurons for pattern recognition tasks in presence of lateral inhibition and homeostasis. Such stochastic MTJ neurons can also potentially provide a direct mapping to the probabilistic computing elements in Belief Networks for performing regenerative tasks.

Unsupervised Learning by Spike Timing Dependent Plasticity in Phase Change Memory (PCM) Synapses

We present a novel one-transistor/one-resistor (1T1R) synapse for neuromorphic networks, based on phase change memory (PCM) technology. The synapse is capable of spike-timing dependent plasticity (STDP), where gradual potentiation relies on set transition, namely crystallization, in the PCM, while depression is achieved via reset or amorphization of a chalcogenide active volume. STDP characteristics are demonstrated by experiments under variable initial conditions and number of pulses. Finally, we support the applicability of the 1T1R synapse for learning and recognition of visual patterns by simulations of fully connected neuromorphic networks with 2 or 3 layers with high recognition efficiency. The proposed scheme provides a feasible low-power solution for on-line unsupervised machine learning in smart reconfigurable sensors.