Multi-level, forming and filament free, bulk switching trilayer RRAM for neuromorphic computing at the edge

CMOS-RRAM integration holds great promise for low energy and high throughput neuromorphic computing. However, most RRAM technologies relying on filamentary switching suffer from variations and noise, leading to computational accuracy loss, increased energy consumption, and overhead by expensive program and verify schemes. We developed a filament-free, bulk switching RRAM technology to address these challenges. We systematically engineered a trilayer metal-oxide stack and investigated the switching characteristics of RRAM with varying thicknesses and oxygen vacancy distributions to achieve reliable bulk switching without any filament formation. We demonstrated bulk switching at megaohm regime with high current nonlinearity, up to 100 levels without compliance current. We developed a neuromorphic compute-in-memory platform and showcased edge computing by implementing a spiking neural network for an autonomous navigation/racing task. Our work addresses challenges posed by existing RRAM technologies and paves the way for neuromorphic computing at the edge under strict size, weight, and power constraints.

High-density transparent graphene arrays for predicting cellular calcium activity at depth from surface potential recordings

Optically transparent neural microelectrodes have facilitated simultaneous electrophysiological recordings from the brain surface with the optical imaging and stimulation of neural activity. A remaining challenge is to scale down the electrode dimensions to the single-cell size and increase the density to record neural activity with high spatial resolution across large areas to capture nonlinear neural dynamics. Here we developed transparent graphene microelectrodes with ultrasmall openings and a large, transparent recording area without any gold extensions in the field of view with high-density microelectrode arrays up to 256 channels. We used platinum nanoparticles to overcome the quantum capacitance limit of graphene and to scale down the microelectrode diameter to 20 µm. An interlayer-doped double-layer graphene was introduced to prevent open-circuit failures. We conducted multimodal experiments, combining the recordings of cortical potentials of microelectrode arrays with two-photon calcium imaging of the mouse visual cortex. Our results revealed that visually evoked responses are spatially localized for high-frequency bands, particularly for the multiunit activity band. The multiunit activity power was found to be correlated with cellular calcium activity. Leveraging this, we employed dimensionality reduction techniques and neural networks to demonstrate that single-cell and average calcium activities can be decoded from surface potentials recorded by high-density transparent graphene arrays.

Spatial Interactions in Hydrogenated Perovskite Nickelate Synaptic Networks

A key aspect of how the brain learns and enables decision-making processes is through synaptic interactions. Electrical transmission and communication in a network of synapses are modulated by extracellular fields generated by ionic chemical gradients. Emulating such spatial interactions in synthetic networks can be of potential use for neuromorphic learning and the hardware implementation of artificial intelligence. Here, we demonstrate that in a network of hydrogen-doped perovskite nickelate devices, electric bias across a single junction can tune the coupling strength between the neighboring cells. Electrical transport measurements and spatially resolved diffraction and nanoprobe X-ray and scanning microwave impedance spectroscopic studies suggest that graded proton distribution in the inhomogeneous medium of hydrogen-doped nickelate film enables this behavior. We further demonstrate signal integration through the coupling of various junctions.

Multimodal monitoring of human cortical organoids implanted in mice reveal functional connection with visual cortex

Human cortical organoids, three-dimensional neuronal cultures, are emerging as powerful tools to study brain development and dysfunction. However, whether organoids can functionally connect to a sensory network in vivo has yet to be demonstrated. Here, we combine transparent microelectrode arrays and two-photon imaging for longitudinal, multimodal monitoring of human cortical organoids transplanted into the retrosplenial cortex of adult mice. Two-photon imaging shows vascularization of the transplanted organoid. Visual stimuli evoke electrophysiological responses in the organoid, matching the responses from the surrounding cortex. Increases in multi-unit activity (MUA) and gamma power and phase locking of stimulus-evoked MUA with slow oscillations indicate functional integration between the organoid and the host brain. Immunostaining confirms the presence of human-mouse synapses. Implantation of transparent microelectrodes with organoids serves as a versatile in vivo platform for comprehensive evaluation of the development, maturation, and functional integration of human neuronal networks within the mouse brain.

E-Cannula reveals anatomical diversity in sharp-wave ripples as a driver for the recruitment of distinct hippocampal assemblies

The hippocampus plays a critical role in spatial navigation and episodic memory. However, research on in vivo hippocampal activity dynamics mostly relies on single modalities, such as electrical recordings or optical imaging, with respectively limited spatial and temporal resolution. Here, we develop the E-Cannula, integrating fully transparent graphene microelectrodes with imaging cannula, which enables simultaneous electrical recording and two-photon calcium imaging from the exact same neural populations across an anatomically extended region of the mouse hippocampal CA1 stably across several days. The large-scale multimodal recordings show that sharp wave ripples (SWRs) exhibit spatiotemporal wave patterns along multiple axes in two-dimensional (2D) space with different spatial extents and temporal propagation modes. Notably, distinct SWR wave patterns are associated with the selective recruitment of orthogonal CA1 cell assemblies. These results demonstrate the utility of the E-Cannula as a versatile neurotechnology with the potential for future integration with other optical components.

A Neuromorphic Brain Interface based on RRAM Crossbar Arrays for High Throughput Real-time Spike Sorting

Real-time spike sorting and processing are crucial for closed-loop brain-machine interfaces and neural prosthetics. Recent developments in high-density multi-electrode arrays with hundreds of electrodes have enabled simultaneous recordings of spikes from a large number of neurons. However, the high channel count imposes stringent demands on real-time spike sorting hardware regarding data transmission bandwidth and computation complexity. Thus, it is necessary to develop a specialized real-time hardware that can sort neural spikes on the fly with high throughputs while consuming minimal power. Here, we present a real-time, low latency spike sorting processor that utilizes high-density CuO_x resistive crossbars to implement in-memory spike sorting in a massively parallel manner. We developed a fabrication process which is compatible with CMOS BEOL integration. We extensively characterized switching characteristics and statistical variations of the CuO_x memory devices. In order to implement spike sorting with crossbar arrays, we developed a template matching-based spike sorting algorithm that can be directly mapped onto RRAM crossbars. By using synthetic and in vivo recordings of extracellular spikes, we experimentally demonstrated energy efficient spike sorting with high accuracy. Our neuromorphic interface offers substantial improvements in area (~1000× less area), power (~200× less power), and latency (4.8μs latency for sorting 100 channels) for real-time spike sorting compared to other hardware implementations based on FPGAs and microcontrollers.

Decoding of cortex-wide brain activity from local recordings of neural potentials

Objective. Electrical recordings of neural activity from brain surface have been widely employed in basic neuroscience research and clinical practice for investigations of neural circuit functions, brain-computer interfaces, and treatments for neurological disorders. Traditionally, these surface potentials have been believed to mainly reflect local neural activity. It is not known how informative the locally recorded surface potentials are for the neural activities across multiple cortical regions.Approach. To investigate that, we perform simultaneous local electrical recording and wide-field calcium imaging in awake head-fixed mice. Using a recurrent neural network model, we try to decode the calcium fluorescence activity of multiple cortical regions from local electrical recordings.Main results. The mean activity of different cortical regions could be decoded from locally recorded surface potentials. Also, each frequency band of surface potentials differentially encodes activities from multiple cortical regions so that including all the frequency bands in the decoding model gives the highest decoding performance. Despite the close spacing between recording channels, surface potentials from different channels provide complementary information about the large-scale cortical activity and the decoding performance continues to improve as more channels are included. Finally, we demonstrate the successful decoding of whole dorsal cortex activity at pixel-level using locally recorded surface potentials.Significance. These results show that the locally recorded surface potentials indeed contain rich information of the large-scale neural activities, which could be further demixed to recover the neural activity across individual cortical regions. In the future, our cross-modality inference approach could be adapted to virtually reconstruct cortex-wide brain activity, greatly expanding the spatial reach of surface electrical recordings without increasing invasiveness. Furthermore, it could be used to facilitate imaging neural activity across the whole cortex in freely moving animals, without requirement of head-fixed microscopy configurations.

Multimodal neural recordings with Neuro-FITM uncover diverse patterns of cortical-hippocampal interactions

Many cognitive processes require communication between the neocortex and the hippocampus. However, coordination between large-scale cortical dynamics and hippocampal activity is not well understood, partially due to the difficulty in simultaneously recording from those regions. In the present study, we developed a flexible, insertable and transparent microelectrode array (Neuro-FITM) that enables investigation of cortical-hippocampal coordinations during hippocampal sharp-wave ripples (SWRs). Flexibility and transparency of Neuro-FITM allow simultaneous recordings of local field potentials and neural spiking from the hippocampus during wide-field calcium imaging. These experiments revealed that diverse cortical activity patterns accompanied SWRs and, in most cases, cortical activation preceded hippocampal SWRs. We demonstrated that, during SWRs, different hippocampal neural population activity was associated with distinct cortical activity patterns. These results suggest that hippocampus and large-scale cortical activity interact in a selective and diverse manner during SWRs underlying various cognitive functions. Our technology can be broadly applied to comprehensive investigations of interactions between the cortex and other subcortical structures.

Energy-efficient Mott activation neuron for full-hardware implementation of neural networks

To circumvent the von Neumann bottleneck, substantial progress has been made towards in-memory computing with synaptic devices. However, compact nanodevices implementing non-linear activation functions are required for efficient full-hardware implementation of deep neural networks. Here, we present an energy-efficient and compact Mott activation neuron based on vanadium dioxide and its successful integration with a conductive bridge random access memory (CBRAM) crossbar array in hardware. The Mott activation neuron implements the rectified linear unit function in the analogue domain. The neuron devices consume substantially less energy and occupy two orders of magnitude smaller area than those of analogue complementary metal-oxide semiconductor implementations. The LeNet-5 network with Mott activation neurons achieves 98.38% accuracy on the MNIST dataset, close to the ideal software accuracy. We perform large-scale image edge detection using the Mott activation neurons integrated with a CBRAM crossbar array. Our findings provide a solution towards large-scale, highly parallel and energy-efficient in-memory computing systems for neural networks.

Evaluation of Durability of Transparent Graphene Electrodes Fabricated on Different Flexible Substrates for Chronic In Vivo Experiments

OBJECTIVE

To investigate chronic durability of transparent graphene electrodes fabricated on polyethylene terephthalate (PET) and SU-8 substrates for chronic in vivo studies.

METHODS

We perform systematic accelerated aging tests to understand the chronic reliability and failure modes of transparent graphene microelectrode arrays built on PET and SU-8 substrates. We employ graphene microelectrodes fabricated on PET substrate in chronic in vivo experiments with transgenic mice.

RESULTS

Our results show that graphene microelectrodes fabricated on PET substrate work reliably after 30 days accelerated aging test performed at 87 °C, equivalent to 960 days in vivo lifetime. We demonstrate stable chronic recordings of cortical potentials in multimodal imaging/recording experiments using transparent graphene microelectrodes fabricated on PET substrate. On the other hand, graphene microelectrode arrays built on SU-8 substrate exhibit extensive crack formation across microelectrode sites and wires after one to two weeks, resulting in total failure of recording capability for chronic studies.

CONCLUSION

PET shows superior reliability as a substrate for graphene microelectrode arrays for chronic in vivo experiments.

SIGNIFICANCE

Graphene is a unique neural interface material enabling cross-talk free integration of electrical and optical recording and stimulation techniques in the same experiment. To date, graphene-based microelectrode arrays have been demonstrated in various multi-modal acute experiments involving electrophysiological sensing or stimulation, optical imaging and optogenetics stimulation. Understanding chronic reliability of graphene-based transparent interfaces is very important to expand the use of this technology for long-term behavioral studies with animal models.

Hippocampal-Cortical Memory Trace Transfer and Reactivation Through Cell-Specific Stimulus and Spontaneous Background Noise

The hippocampus plays important roles in memory formation and retrieval through sharp-wave-ripples. Recent studies have shown that certain neuron populations in the prefrontal cortex (PFC) exhibit coordinated reactivations during awake ripple events. These experimental findings suggest that the awake ripple is an important biomarker, through which the hippocampus interacts with the neocortex to assist memory formation and retrieval. However, the computational mechanisms of this ripple based hippocampal-cortical coordination are still not clear due to the lack of unified models that include both the hippocampal and cortical networks. In this work, using a coupled biophysical model of both CA1 and PFC, we investigate possible mechanisms of hippocampal-cortical memory trace transfer and the conditions that assist reactivation of the transferred memory traces in the PFC. To validate our model, we first show that the local field potentials generated in the hippocampus and PFC exhibit ripple range activities that are consistent with the recent experimental studies. Then we demonstrate that during ripples, sequence replays can successfully transfer the information stored in the hippocampus to the PFC recurrent networks. We investigate possible mechanisms of memory retrieval in PFC networks. Our results suggest that the stored memory traces in the PFC network can be retrieved through two different mechanisms, namely the cell-specific input representing external stimuli and non-specific spontaneous background noise representing spontaneous memory recall events. Importantly, in both cases, the memory reactivation quality is robust to network connection loss. Finally, we find out that the quality of sequence reactivations is enhanced by both increased number of SWRs and an optimal background noise intensity, which tunes the excitability of neurons to a proper level. Our study presents a mechanistic explanation for the memory trace transfer from the hippocampus to neocortex through ripple coupling in awake states and reports two different mechanisms by which the stored memory traces can be successfully retrieved.

A Soft-Pruning Method Applied During Training of Spiking Neural Networks for In-memory Computing Applications

Inspired from the computational efficiency of the biological brain, spiking neural networks (SNNs) emulate biological neural networks, neural codes, dynamics, and circuitry. SNNs show great potential for the implementation of unsupervised learning using in-memory computing. Here, we report an algorithmic optimization that improves energy efficiency of online learning with SNNs on emerging non-volatile memory (eNVM) devices. We develop a pruning method for SNNs by exploiting the output firing characteristics of neurons. Our pruning method can be applied during network training, which is different from previous approaches in the literature that employ pruning on already-trained networks. This approach prevents unnecessary updates of network parameters during training. This algorithmic optimization can complement the energy efficiency of eNVM technology, which offers a unique in-memory computing platform for the parallelization of neural network operations. Our SNN maintains ~90% classification accuracy on the MNIST dataset with up to ~75% pruning, significantly reducing the number of weight updates. The SNN and pruning scheme developed in this work can pave the way toward applications of eNVM based neuro-inspired systems for energy efficient online learning in low power applications.

Investigation of Propagating Cortical Waves and Spirals Recorded by High Density Porous Graphene Arrays

Propagating waves along the cortical surface have recently attracted significant attention by the neuroscience community. However, whether these propagating waves imply network connectivity changes for the neural circuits is not known. In this work, we employ a high density porous graphene microelectrode array and perform in vivo experiments with rodents to investigate network connectivity during cortical propagating waves. The spatial-temporal analysis of the cortical recordings reveals various types of propagating waves across the recording area. Network analysis results show that these propagating waves are consistent with the functional connectivity changes in the neural circuits, suggesting that the underlying network states are reflected by the cortical potential propagation patterns.

Ultra-low Impedance Graphene Microelectrodes with High Optical Transparency for Simultaneous Deep 2-photon Imaging in Transgenic Mice

The last decades have witnessed substantial progress in optical technologies revolutionizing our ability to record and manipulate neural activity in genetically modified animal models. Meanwhile, human studies mostly rely on electrophysiological recordings of cortical potentials, which cannot be inferred from optical recordings, leading to a gap between our understanding of dynamics of microscale populations and brain-scale neural activity. By enabling concurrent integration of electrical and optical modalities, transparent graphene microelectrodes can close this gap. However, the high impedance of graphene constitutes a big challenge towards the widespread use of this technology. Here, we experimentally demonstrate that this high impedance of graphene microelectrodes is fundamentally limited by quantum capacitance. We overcome this quantum capacitance limit by creating a parallel conduction path using platinum nanoparticles. We achieve a 100 times reduction in graphene electrode impedance, while maintaining the high optical transparency crucial for deep 2-photon microscopy. Using a transgenic mouse model, we demonstrate simultaneous electrical recording of cortical activity with high fidelity while imaging calcium signals at various cortical depths right beneath the transparent microelectrodes. Multimodal analysis of Ca2+ spikes and cortical surface potentials offers unique opportunities to bridge our understanding of cellular dynamics and brain-scale neural activity.

A Compact Closed-Loop Optogenetics System Based on Artifact-Free Transparent Graphene Electrodes

Electrophysiology is a decades-old technique widely used for monitoring activity of individual neurons and local field potentials. Optogenetics has revolutionized neuroscience studies by offering selective and fast control of targeted neurons and neuron populations. The combination of these two techniques is crucial for causal investigation of neural circuits and understanding their functional connectivity. However, electrical artifacts generated by light stimulation interfere with neural recordings and hinder the development of compact closed-loop systems for precise control of neural activity. Here, we demonstrate that transparent graphene micro-electrodes fabricated on a clear polyethylene terephthalate film eliminate the light-induced artifact problem and allow development of a compact battery-powered closed-loop optogenetics system. We extensively investigate light-induced artifacts for graphene electrodes in comparison to metal control electrodes. We then design optical stimulation module using micro-LED chips coupled to optical fibers to deliver light to intended depth for optogenetic stimulation. For artifact-free integration of graphene micro-electrode recordings with optogenetic stimulation, we design and develop a compact closed-loop system and validate it for different frequencies of interest for neural recordings. This compact closed-loop optogenetics system can be used for various applications involving optogenetic stimulation and electrophysiological recordings.

Spatiotemporal evolution of focal epileptiform activity from surface and laminar field recordings in cat neocortex

New devices that use targeted electrical stimulation to treat refractory localization-related epilepsy have shown great promise, although it is not well known which targets most effectively prevent the initiation and spread of seizures. To better understand how the brain transitions from healthy to seizing on a local scale, we induced focal epileptiform activity in the visual cortex of five anesthetized cats with local application of the GABA_A blocker picrotoxin while simultaneously recording local field potentials on a high-resolution electrocorticography array and laminar depth probes. Epileptiform activity appeared in the form of isolated events, revealing a consistent temporal pattern of ictogenesis across animals with interictal events consistently preceding the appearance of seizures. Based on the number of spikes per event, there was a natural separation between seizures and shorter interictal events. Two distinct spatial regions were seen: an epileptic focus that grew in size as activity progressed, and an inhibitory surround that exhibited a distinct relationship with the focus both on the surface and in the depth of the cortex. Epileptiform activity in the cortical laminae was seen concomitant with activity on the surface. Focus spikes appeared earlier on electrodes deeper in the cortex, suggesting that deep cortical layers may be integral to recruiting healthy tissue into the epileptic network and could be a promising target for interventional devices. Our study may inform more effective therapies to prevent seizure generation and spread in localization-related epilepsies. NEW & NOTEWORTHY We induced local epileptiform activity and recorded continuous, high-resolution local field potentials from the surface and depth of the visual cortex in anesthetized cats. Our results reveal a consistent pattern of ictogenesis, characterize the spatial spread of the epileptic focus and its relationship with the inhibitory surround, and show that focus activity within events appears earliest in deeper cortical layers. These findings have potential implications for the monitoring and treatment of refractory epilepsy.

Computational analysis of network activity and spatial reach of sharp wave-ripples

Network oscillations of different frequencies, durations and amplitudes are hypothesized to coordinate information processing and transfer across brain areas. Among these oscillations, hippocampal sharp wave-ripple complexes (SPW-Rs) are one of the most prominent. SPW-Rs occurring in the hippocampus are suggested to play essential roles in memory consolidation as well as information transfer to the neocortex. To-date, most of the knowledge about SPW-Rs comes from experimental studies averaging responses from neuronal populations monitored by conventional microelectrodes. In this work, we investigate spatiotemporal characteristics of SPW-Rs and how microelectrode size and distance influence SPW-R recordings using a biophysical model of hippocampus. We also explore contributions from neuronal spikes and synaptic potentials to SPW-Rs based on two different types of network activity. Our study suggests that neuronal spikes from pyramidal cells contribute significantly to ripples while high amplitude sharp waves mainly arise from synaptic activity. Our simulations on spatial reach of SPW-Rs show that the amplitudes of sharp waves and ripples exhibit a steep decrease with distance from the network and this effect is more prominent for smaller area electrodes. Furthermore, the amplitude of the signal decreases strongly with increasing electrode surface area as a result of averaging. The relative decrease is more pronounced when the recording electrode is closer to the source of the activity. Through simulations of field potentials across a high-density microelectrode array, we demonstrate the importance of finding the ideal spatial resolution for capturing SPW-Rs with great sensitivity. Our work provides insights on contributions from spikes and synaptic potentials to SPW-Rs and describes the effect of measurement configuration on LFPs to guide experimental studies towards improved SPW-R recordings.

Transformation of Cortex-wide Emergent Properties during Motor Learning

Learning involves a transformation of brain-wide operation dynamics. However, our understanding of learning-related changes in macroscopic dynamics is limited. Here, we monitored cortex-wide activity of the mouse brain using wide-field calcium imaging while the mouse learned a motor task over weeks. Over learning, the sequential activity across cortical modules became temporally more compressed, and its trial-by-trial variability decreased. Moreover, a new flow of activity emerged during learning, originating from premotor cortex (M2), and M2 became predictive of the activity of many other modules. Inactivation experiments showed that M2 is critical for the post-learning dynamics in the cortex-wide activity. Furthermore, two-photon calcium imaging revealed that M2 ensemble activity also showed earlier activity onset and reduced variability with learning, which was accompanied by changes in the activity-movement relationship. These results reveal newly emergent properties of macroscopic cortical dynamics during motor learning and highlight the importance of M2 in controlling learned movements.

Implantable optoelectronic probes for in vivo optogenetics

More than a decade has passed since optics and genetics came together and lead to the emerging technologies of optogenetics. The advent of light-sensitive opsins made it possible to optically trigger the neurons into activation or inhibition by using visible light. The importance of spatiotemporally isolating a segment of a neural network and controlling nervous signaling in a precise manner has driven neuroscience researchers and engineers to invest great efforts in designing high precision in vivo implantable devices. These efforts have focused on delivery of sufficient power to deep brain regions, while monitoring neural activity with high resolution and fidelity. In this review, we report the progress made in the field of hybrid optoelectronic neural interfaces that combine optical stimulation with electrophysiological recordings. Different approaches that incorporate optical or electrical components on implantable devices are discussed in detail. Advantages of various different designs as well as practical and fundamental limitations are summarized to illuminate the future of neurotechnology development.

Graphene neural interfaces for artifact free optogenetics

Light-induced artifacts in metal-based microelectrodes significantly limit simultaneous use of electrophysiology with optogenetics or optical imaging. In this work, we systematically investigate light-induced artifacts in Au and transparent graphene electrodes fabricated in the same batch. We demonstrate that the Au electrodes show light artifacts resembling genuine local field potentials, while graphene electrodes are free of light-induced artifacts. With its other desirable properties including high mechanical strength, good conductivity, transparency, and biocompatibility, graphene electrodes show great promise for combining electrical and optical modalities in the same experiment.

Bioresorbable silicon electronics for transient spatiotemporal mapping of electrical activity from the cerebral cortex

Bioresorbable silicon electronics technology offers unprecedented opportunities to deploy advanced implantable monitoring systems that eliminate risks, cost and discomfort associated with surgical extraction. Applications include postoperative monitoring and transient physiologic recording after percutaneous or minimally invasive placement of vascular, cardiac, orthopaedic, neural or other devices. We present an embodiment of these materials in both passive and actively addressed arrays of bioresorbable silicon electrodes with multiplexing capabilities, which record in vivo electrophysiological signals from the cortical surface and the subgaleal space. The devices detect normal physiologic and epileptiform activity, both in acute and chronic recordings. Comparative studies show sensor performance comparable to standard clinical systems and reduced tissue reactivity relative to conventional clinical electrocorticography (ECoG) electrodes. This technology offers general applicability in neural interfaces, with additional potential utility in treatment of disorders where transient monitoring and modulation of physiologic function, implant integrity and tissue recovery or regeneration are required.

Transparent and flexible low noise graphene electrodes for simultaneous electrophysiology and neuroimaging

Calcium imaging is a versatile experimental approach capable of resolving single neurons with single-cell spatial resolution in the brain. Electrophysiological recordings provide high temporal, but limited spatial resolution, because of the geometrical inaccessibility of the brain. An approach that integrates the advantages of both techniques could provide new insights into functions of neural circuits. Here, we report a transparent, flexible neural electrode technology based on graphene, which enables simultaneous optical imaging and electrophysiological recording. We demonstrate that hippocampal slices can be imaged through transparent graphene electrodes by both confocal and two-photon microscopy without causing any light-induced artefacts in the electrical recordings. Graphene electrodes record high-frequency bursting activity and slow synaptic potentials that are hard to resolve by multicellular calcium imaging. This transparent electrode technology may pave the way for high spatio-temporal resolution electro-optic mapping of the dynamic neuronal activity.

Brain-like associative learning using a nanoscale non-volatile phase change synaptic device array

Recent advances in neuroscience together with nanoscale electronic device technology have resulted in huge interests in realizing brain-like computing hardwares using emerging nanoscale memory devices as synaptic elements. Although there has been experimental work that demonstrated the operation of nanoscale synaptic element at the single device level, network level studies have been limited to simulations. In this work, we demonstrate, using experiments, array level associative learning using phase change synaptic devices connected in a grid like configuration similar to the organization of the biological brain. Implementing Hebbian learning with phase change memory cells, the synaptic grid was able to store presented patterns and recall missing patterns in an associative brain-like fashion. We found that the system is robust to device variations, and large variations in cell resistance states can be accommodated by increasing the number of training epochs. We illustrated the tradeoff between variation tolerance of the network and the overall energy consumption, and found that energy consumption is decreased significantly for lower variation tolerance.

Synaptic electronics: materials, devices and applications

In this paper, the recent progress of synaptic electronics is reviewed. The basics of biological synaptic plasticity and learning are described. The material properties and electrical switching characteristics of a variety of synaptic devices are discussed, with a focus on the use of synaptic devices for neuromorphic or brain-inspired computing. Performance metrics desirable for large-scale implementations of synaptic devices are illustrated. A review of recent work on targeted computing applications with synaptic devices is presented.

Nanoelectronic programmable synapses based on phase change materials for brain-inspired computing

Brain-inspired computing is an emerging field, which aims to extend the capabilities of information technology beyond digital logic. A compact nanoscale device, emulating biological synapses, is needed as the building block for brain-like computational systems. Here, we report a new nanoscale electronic synapse based on technologically mature phase change materials employed in optical data storage and nonvolatile memory applications. We utilize continuous resistance transitions in phase change materials to mimic the analog nature of biological synapses, enabling the implementation of a synaptic learning rule. We demonstrate different forms of spike-timing-dependent plasticity using the same nanoscale synapse with picojoule level energy consumption.

SiGe optoelectronic metal-oxide semiconductor field-effect transistor

We propose a novel semiconductor optoelectronic switch that is a fusion of a Ge optical detector and a Si metal-oxide semiconductor field-effect transistor (MOSFET). The device operation is investigated with simulations and experiments. The switch can be fabricated at the nanoscale with extremely low capacitance. This device operates in telecommunication standard wavelengths, hence providing the surrounding Si circuitry with noise immunity from signaling. The Ge gate absorbs light, and the gate photocurrent is amplified at the drain terminal. Experimental current gain of up to 1000x is demonstrated. The device exhibits increased responsivity (approximately 3.5x) and lower off-state current (approximately 4x) compared with traditional detector schemes.