A5075444766- Advanced Memory and Neural Computing
- Neuroscience and Neural Engineering
- Ferroelectric and Negative Capacitance Devices
- Neural dynamics and brain function
- Semiconductor materials and devices
- CCD and CMOS Imaging Sensors
- Neural Networks and Applications
- Electrochemical Analysis and Applications
- Machine Learning and ELM
- Transition Metal Oxide Nanomaterials
- Advancements in Semiconductor Devices and Circuit Design
- Electron and X-Ray Spectroscopy Techniques
- Magnetic properties of thin films
- Stochastic Gradient Optimization Techniques
- Photoreceptor and optogenetics research
- Magnetic Field Sensors Techniques
- Iron oxide chemistry and applications
- Spectroscopy and Quantum Chemical Studies
- Physical Unclonable Functions (PUFs) and Hardware Security
- Analytical Chemistry and Sensors
- Neural Networks and Reservoir Computing
- Quantum and electron transport phenomena
- Advanced Electron Microscopy Techniques and Applications
- Electronic and Structural Properties of Oxides
- Machine Learning in Materials Science
National Institute of Standards and Technology
2017-2025
Physical Measurement Laboratory
2019-2024
National Institute of Standards
2020-2023
Theiss Research
2021
University of Maryland, College Park
2021
Brown University
2021
Providence College
2021
George Washington University
2021
Western Digital (United States)
2021
Material Measurement Laboratory
2020
Using memristive properties common for titanium dioxide thin film devices, we designed a simple write algorithm to tune device conductance at specific bias point 1% relative accuracy (which is roughly equivalent seven-bit precision) within its dynamic range even in the presence of large variations switching behavior. The high precision state nonvolatile and results are likely be sustained nanoscale devices because inherent filamentary nature resistive switching. proposed functionality...
We report a monolithically integrated 3-D metal-oxide memristor crossbar circuit suitable for analog, and in particular, neuromorphic computing applications. The demonstrated is based on Pt/Al <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> O xmlns:xlink="http://www.w3.org/1999/xlink">3</sub> /TiO xmlns:xlink="http://www.w3.org/1999/xlink">2-x</sub> /TiN/Pt memristors consists of stack two passive 10 × crossbars with shared middle...
Abstract Metal-oxide memristors have emerged as promising candidates for hardware implementation of artificial synapses – the key components high-performance, analog neuromorphic networks - due to their excellent scaling prospects. Since some advanced cognitive tasks require spiking networks, which explicitly model individual neural pulses (“spikes”) in biological systems, it is crucial memristive support spike-time-dependent plasticity (STDP). A major challenge STDP that, contrast...
Recent progress in artificial intelligence is largely attributed to the rapid development of machine learning, especially algorithm and neural network models. However, it performance hardware, particular energy efficiency a computing system that sets fundamental limit capability learning. Data-centric requires revolution hardware systems, since traditional digital computers based on transistors von Neumann architecture were not purposely designed for neuromorphic computing. A platform...
Abstract Oxide-based resistive switching devices are promising candidates for new memory and computing technologies. Poor understanding of the defect-based mechanisms that give rise to is a major impediment engineering reliable reproducible devices. Here we identify an unintentional interface layer as origin in Pt/Nb:SrTiO 3 junctions. We clarify microscopic by which controls switching. show appropriate processing can eliminate this contribution. These findings important step towards more
Abstract Silicon (Si) based complementary metal-oxide semiconductor (CMOS) technology has been the driving force of information-technology revolution. However, scaling CMOS as per Moore’s law reached a serious bottleneck. Among emerging technologies memristive devices can be promising for both memory well computing applications. Hybrid CMOS/memristor circuits with CMOL (CMOS + “Molecular”) architecture have proposed to combine extremely high density robustness technology, leading...
The purpose of this work was to demonstrate the feasibility building recurrent artificial neural networks with hybrid complementary metal oxide semiconductor (CMOS)/memristor circuits. To do so, we modeled a Hopfield network implementing an analog-to-digital converter (ADC) up 8 bits precision. Major shortcomings affecting ADC's precision, such as non-ideal behavior CMOS circuitry and specific limitations memristors, were investigated effective solution proposed, capitalizing on in-field...
Abstract Threshold switching devices are of increasing importance for a number applications including solid-state memories and neuromorphic circuits. Their non-linear characteristics thought to be associated with spontaneous (occurring without an apparent external stimulus) current flow constriction but the extent underlying mechanism subject debate. Here we use Scanning Joule Expansion Microscopy demonstrate that, in functional layers thermally activated electrical conductivity,...
We present a computationally inexpensive yet accurate phenomenological model of memristive behavior in titanium dioxide devices by fitting experimental data. By design, the predicts most accurately I-V relation at small non-disturbing electrical stresses, which is often critical range operation for circuit modeling. While choice functions motivated switching and conduction mechanisms particular devices, proposed modeling methodology general enough to be applied different types memory feature...
Artificial neural networks have advanced due to scaling dimensions, but conventional computing struggles with inefficiencies memory bottlenecks. In-memory architectures using memristor devices offer promise face challenges hardware non-idealities. This work proposes layer ensemble averaging—a hardware-oriented fault tolerance scheme for improving inference performance of non-ideal memristive programmed pre-trained solutions. Simulations on an image classification task and experiments a...
Neuromorphic pattern classifiers were implemented, for the first time, using transistor-free integrated crossbar circuits with bilayer metal-oxide memristors. 10×6- and 10×8-crosspoint neuromorphic networks trained in-situ a Manhattan-Rule algorithm to separate set of 3×3 binary images: into 3 classes batch-mode training, 4 stochastic-mode respectively. Simulation much larger, multilayer neural network based on such technology has sown that their fidelity may be par state-of-the-art results...
The paper presents experimental demonstration of 6-bit digital-to-analog (DAC) and 4-bit analog-to-digital conversion (ADC) operations implemented with a hybrid circuit consisting Pt/TiO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2-x</sub> /Pt resistive switching devices (also known as ReRAMs or memristors) Si operational amplifier (op-amp). In particular, binary-weighted implementation is demonstrated for DAC, while ADC Hopfield neural...
Metal oxide resistive switches are increasingly important as possible artificial synapses in next generation neuromorphic networks. Nevertheless, there is still no codified set of tools for studying properties the devices. To this end, we demonstrate electron beam induced current measurements a powerful method to monitor development local switching TiO2 based By comparing beam-energy dependent currents with Monte Carlo simulations energy absorption different device layers, it deconstruct...
The electrical double layer (EDL) governs the operation of multiple electrochemical devices, determines reaction potentials, and conditions ion transport through cellular membranes in living organisms. few existing methods EDL probing have low spatial resolution, usually only providing spatially averaged information. On other hand, traditional Kelvin probe force microscopy (KPFM) is capable mapping potential with nanoscale lateral resolution but cannot be used electrolytes concentrations...
This is a brief review of our recent work on memristor-based spiking neuromorphic networks. We first describe the experimental demonstration several most biology-plausible spike-time-dependent plasticity (STDP) windows in integrated metal-oxide memristors and, for time, observed self-adaptive STDP, which may be crucial neural network applications. then discuss theoretical an analytical, data-verified STDP model was used to simulate operation classifier spatial-temporal patterns, and...
Skyrmions hold great promise for low-energy consumption and stable high-density information storage, stabilization of the skyrmion lattice (SkX) phase at or above room temperature is greatly desired practical use. The topological Hall effect can be used to identify candidate systems temperature, a challenging regime direct observation by Lorentz electron microscopy. Atomically ordered FeGe thin films are grown epitaxially on Ge(111) substrates with $\ensuremath{\sim}4%$ tensile strain....
Defective devices can severely impact the performance of hardware-based neural networks, in particular resistive crossbar arrays. This study introduces a network training approach that reduces influence defective devices, maintaining inference accuracy. The authors demonstrate this on set dies each containing array consisting 20,000 magnetic tunnel junction devices. They also develop generalized using statistics defects and similar all dies. These results translate to manufacturing setting...
The conduction/set processes of resistive switching have been systemically investigated for Cu/a-Si/Si electrochemical memristive devices. Experimental results indicate that the set process was driven by two different mechanisms, depending on programming pulse amplitude: a purely electrical dielectric breakdown and thermally assisted breakdown. For latter process, we observe time decreased exponentially with increase in amplitude, whereas former shows amplitude independence. Through...
Synapses, the most numerous elements of neural networks, are memory devices. Similarly to traditional applications, device density is one essential metrics for large-scale artificial networks. This application, however, imposes a number additional requirements, such as continuous change state, so that novel engineering approaches required. In this paper, we briefly review our recent efforts at addressing these needs. We start by reviewing CrossNet concept, which was conceived address major...
Neural networks based on nanodevices, such as metal oxide memristors, phase change memories, and flash memory cells, have generated considerable interest for their increased energy efficiency density in comparison to graphics processing units (GPUs) central (CPUs). Though immense acceleration of the training process can be achieved by leveraging fact that time complexity does not scale with network size, it is limited space stochastic gradient descent, which grows quadratically. The main...