A5023902529- Advanced Memory and Neural Computing
- Ferroelectric and Negative Capacitance Devices
- Neuroscience and Neural Engineering
- Transition Metal Oxide Nanomaterials
- Semiconductor materials and devices
- Neural dynamics and brain function
- CCD and CMOS Imaging Sensors
- Magnetic properties of thin films
- ZnO doping and properties
- Machine Learning and ELM
- Neural Networks and Reservoir Computing
- Magnetic Properties and Applications
- Radio Frequency Integrated Circuit Design
- Photoreceptor and optogenetics research
- Electromagnetic wave absorption materials
- Action Observation and Synchronization
- Thin-Film Transistor Technologies
- Advanced Neural Network Applications
- Brain Tumor Detection and Classification
- Metallic Glasses and Amorphous Alloys
- GaN-based semiconductor devices and materials
- Microwave Engineering and Waveguides
- Advanced DC-DC Converters
- Advanced Power Amplifier Design
- Electrochemical Analysis and Applications
Samsung (South Korea)
2019-2025
Pohang University of Science and Technology
2013-2023
Georgia Institute of Technology
2006
IBM Research - Almaden
1999
Dense crossbar arrays of non-volatile memory (NVM) devices represent one possible path for implementing massively-parallel and highly energy-efficient neuromorphic computing systems. We first review recent advances in the application NVM to three paradigms: spiking neural networks (SNNs), deep (DNNs), 'Memcomputing'. In SNNs, synaptic connections are updated by a local learning rule such as spike-timing-dependent-plasticity, computational approach directly inspired biology. For DNNs, can...
We analyze the response of identical pulses on a filamentary resistive memory (RRAM) to implement synapse function in neuromorphic systems. Our findings show that multilevel states conductance are achieved by varying measurement conditions related formation and rupture conductive filament. Furthermore, abrupt set switching behavior RRAM leads an unchanged state, leading degradation accuracy pattern recognition. Thus, we demonstrate linear potentiation (or depression) under using effect...
We propose TiO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</sub> -based resistive switching device for neuromorphic synapse applications. This is capable of 64-levels conductance states because their optimized interface between the metal electrode and TiOx film. To compensate change in power with increasing pulse number, we use fixed voltage current pulses potentiation depression conditions, respectively. By adopting a hybrid scheme,...
We demonstrated a proton-based 3-terminal synapse device which shows symmetric conductance change characteristics. Using the optimized device, we successfully confirmed improved classification accuracy of neural networks for on-chip training.
Abstract This study demonstrates an integrate and fire (I&F) neuron using threshold switching (TS) devices to implement spike‐based neuromorphic system. An I&F can be realized the hysteric voltage switch characteristics of a TS device. To investigate effects various on behavior, neurons are compared three different types device: NbO 2 ‐based insulator‐to‐metal transition (IMT) device, B–Te‐based ovonic Ag/HfO atomic‐switching The results show that off‐state resistance time determine...
In this brief, we demonstrate the multilevel cell (MLC) characteristics of an HfO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> -based resistive memory (RRAM) array as a synaptic element for neuromorphic systems. We utilize various programming schemes to linearly change resistance state with either set voltage/pulse ramping or gate voltage ramping. Our results reveal that MLC relates size conductive filament involved in movement oxygen...
We perform a comparative study of HfO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> and Ta O xmlns:xlink="http://www.w3.org/1999/xlink">5</sub> resistive switching memory (RRAM) devices for their possible application as electronic synapses. By means electrical characterization simulations, we link behavior (digital or analog switching) to the properties evolution conductive filament (CF). More specifically, identify that...
STT-RAM (Spin-Transfer Torque Random Access Memory) is a second-generation magnetic random access memory (MRAM) technology that fast, non-volatile, durable, and scalable to future nodes [1-2]. In this paper, we present the latest advances in in-plane perpendicular development outline STT-RAM's prospects, applications roadmap.
We report on material improvements to non-filamentary RRAM devices based Pr <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.7</sub> Ca xmlns:xlink="http://www.w3.org/1999/xlink">0.3</sub> MnO xmlns:xlink="http://www.w3.org/1999/xlink">3</sub> by introducing an MoOx buffer layer together with a reactive Al electrode, and device measurements designed help gauge the performance of these as bidirectional analog synapses for on-chip acceleration...
We report on a 1-transisor/2-resistor (1T2R) synapse device with improved conductance linearity and ratio under an identical pulse condition for hardware neural networks high pattern-recognition accuracy. Utilizing additional series-connected resistor, the of was significantly owing to reduced initial voltage drop resistive RAM (RRAM) during depression conditions. Moreover, maximize device, we utilized steep subthreshold region MOSFET by parallel connection RRAM transistor. A small change in...
In this letter, we demonstrate a steep slope field-effect transistor (FET) using threshold switching (TS) device. The Ag/TiO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> -based TS device reported in our previous work was implemented series with the drain region of transistor. Since has an abrupt transition between OFF- and ON-states vice versa, 5-mV/decade subthreshold high ON/OFF-current ratio (I...
This paper presents the development of a novel aluminum-filled high dielectric constant composite for embedded passive applications. Aluminum is well known as low-cost and fast self-passivation metal. The forms nanoscale insulating boundary outside metallic spheres, which has dramatic effects on electrical, mechanical, chemical behaviors resulting composites. Influences aluminum particle size filler loading properties composites were studied. Because self-passivated oxide layer fine level...
We have investigated the analogue memory characteristics of an oxide-based resistive-switching device under electrical pulse to mimic biological spike-timing-dependent plasticity synapse characteristics. As a synaptic device, TiN/Pr0.7Ca0.3MnO3-based exhibiting excellent was used control weight by applying various amplitudes and cycles. Furthermore, potentiation depression with same spikes can be achieved negative positive pulses, respectively. By adopting complementary...
We report oxide based analog synpase for neuromorphic system. By optimizing redox reaction at the metal/oxide interface, we can obtain stable synapse characteristics and wafer scale switching uniformity. have confirmed feasibility of hardware system with array device which recognizes electroencephalogram (EEG) signal rat's neural signal.
This letter presents an investigation of analog synapse characteristics a PCMO-based interface switching device with varying electrode materials. In comparison the filamentary having only 1-b storage and variability issues, devices exhibit excellent electrical properties, such as 5-b (32-level) multi-level cell characteristics, wafer-scale uniformity, scalability energy area. To improve data retention device, we propose Mo to increase oxidation barrier height (~0.4 eV) that, in turn,...
We assess the impact of conductance response Non-Volatile Memory (NVM) devices employed as synaptic weight element for on-chip acceleration training large-scale artificial neural networks (ANN). briefly review our previous work towards achieving competitive performance (classification accuracies) such ANN with both Phase-Change (PCM) [1], [2] and non-filamentary ReRAM based on PrCaMnO (PCMO) [3], assessing potential advantages ML over GPU-based hardware in terms speed (up to 25× faster)...
To improve the classification accuracy of an image data set (CIFAR-10) by using analog input voltage, synapse devices with excellent conductance linearity (CL) and multi-level cell (MLC) characteristics are required. We analyze CL MLC TaOx-based filamentary resistive random access memory (RRAM) to implement device in neural network hardware. Our findings show that number oxygen vacancies filament constriction region RRAM directly controls characteristics. By adopting a Ta electrode (instead...
The advancements in 3D-NAND technology have significantly increased the number of vertically stacked cells, which are controlled via word lines (WLs), enabling higher cell density and reducing costs. However, increase vertical layers has also introduced challenges such as power consumption diminished current levels, both compromise reliability memory cells. At same time, demand for high low been growing, driven by expanding needs storage applications big data cloud services. In this study,...
The sensorimotor system is broadly organized somatotopically. However, an action-type organization has also been found: a division based on independent of acting body parts shown for reaching and grasping actions. Does this generalization extend to non-ethological actions? Here, we examined fMRI responses tool-use actions that participants performed with their hands or feet. We additionally tested individuals born without control hand motor imagery when performing foot show the primary...
Abstract Neuromorphic hardware systems emulate the parallel neural networks of human brain, and synaptic weight storage elements are crucial for enabling energy-efficient information processing. They must represent multiple data states be able to updated analogously. In order realize highly controllable devices, replacing high-k gate dielectric in conventional transistor structures with either solid-electrolytes that facilitate bulk ionic motion or ferroelectric oxide allows steady...
We report our progress on material improvement, device design, wafer processing, integration with CMOS, and testing of STT-RAM memory chips at 54 nm node cell sizes 14 28 F <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> (F=54 nm). A dual tunnel barrier MTJ structure was found to have lower more symmetric median spin transfer torque writing switching currents, much tighter parallel antiparallel current distribution. In-plane devices...
A 3D high-density switching device is realized utilizing titanium oxide, which the most optimum material, but not practically demonstrated yet. The 1S1R (one ReRAM with developed device) exhibits memory characteristics a significantly suppressed sneak current, can be used to realize applications.
We report novel nanoscale synapse and neuron devices for ultra-high density neuromorphic system. By adopting a Mo electrode, the redox reaction at Mo/Pr <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.7</sub> Ca xmlns:xlink="http://www.w3.org/1999/xlink">0.3</sub> MnO xmlns:xlink="http://www.w3.org/1999/xlink">3</sub> (PCMO) interface was controlled which in turn significantly improve characteristics such as switching uniformity, disturbance,...
Current overshoot has severe effects on the reliability of resistive random access memory (RRAM). It is well known that current during SET process caused by parasitic capacitance. In this letter, we observed a different type RESET process. The was confirmed to have endurance RRAM. We also demonstrated relation between and intrinsic capacitive elements each state Finally, an optimized pulse shape proposed minimize experimentally verified significantly improve variability in typical RRAM...
Large arrays of the same nonvolatile memories (NVM) being developed for Storage-Class Memory (SCM) - such as Phase Change (PCM) and Resistance RAM (ReRAM) can also be used in non-Von Neumann neuromorphic computational schemes, with device conductance serving synaptic "weight." This allows all-important multiply-accumulate operation within these algorithms to performed efficiently at weight data.