A5013038382- Advanced Memory and Neural Computing
- Ferroelectric and Negative Capacitance Devices
- Semiconductor materials and devices
- Metallurgy and Material Forming
- Neuroscience and Neural Engineering
- Metal Alloys Wear and Properties
- Advancements in Semiconductor Devices and Circuit Design
- Simulation and Modeling Applications
- Hydraulic and Pneumatic Systems
- Textile materials and evaluations
- Geomechanics and Mining Engineering
- Neural dynamics and brain function
- Microstructure and Mechanical Properties of Steels
- Photoreceptor and optogenetics research
- Industrial Technology and Control Systems
- Nanowire Synthesis and Applications
- Neural Networks and Reservoir Computing
- Machine Learning and ELM
- Powder Metallurgy Techniques and Materials
- Advanced Sensor and Control Systems
- Vibration Control and Rheological Fluids
- Smart Materials for Construction
- Mining and Gasification Technologies
- Thermal properties of materials
- CCD and CMOS Imaging Sensors
Tsinghua University
2004-2024
Suzhou University of Science and Technology
2024
Institute of Microelectronics
2018-2021
Xi'an Jiaotong University
2008-2020
Shaanxi University of Technology
2012
Shanxi University
2005-2012
Ansteel (China)
2011-2012
Sorbonne Université
2012
Institut des NanoSciences de Paris
2012
Centre National de la Recherche Scientifique
2012
Three-Dimensional NAND flash technology is one of the most competitive integrated solutions for high-volume massive data storage. So far, there are few investigations on how to use 3-D in-memory computing in neural network accelerator. In this brief, we propose using vertical channel array architecture implement vector-matrix multiplication (VMM) with first time. Based array-level SPICE simulation, bias condition including selector layer and unselected layers optimized achieve high...
Memristor with topotactic phase transition demonstrates controllable analog switching and implements neural network pruning.
A physically unclonable function (PUF) is a creditable and lightweight solution to the mistrust in billions of Internet Things devices. Because this remarkable importance, PUF need be immune multifarious attack means. Making concealable considered an effective countermeasure but it not feasible for existing designs. The bottleneck finding reproducible randomness source that supports repeatable concealment accurate recovery data. In work, we experimentally demonstrate at chip level with...
The parallelism and analog computing features of neuromorphic systems bring great challenges in developing a compact model resistive random access memory (RRAM). In this article, we develop physics-based for RRAM devices crossbar array. Nonideal effects device, such as variability, I-V nonlinearity, programming nonlinearity asymmetry, tuning voltage sensitivity, are modeled verified with the statistical data measured from Modeling parallel-vector-matrix-multiplication weight update process...
The physicochemical properties of steel slag were investigated using SEM and IR, it was found that free calcium oxide magnesium in produce hydroxide when contact with water, leading to volume expansion. Thus, the expansion rate itself first investigated, more obvious seven days after water immersion. Then, cement dosages 5% 6% cement-stabilized gravel changes between intrinsic link further explored study bonding effect can be partially inhibited due caused by slag, so seen increasing dosage...
Computing-in-memory (CIM) with analog resistive random access memory (RRAM) has recently shown great potential in building energy-efficient hardware for artificial intelligence (AI). However, the relaxation effect of RRAM featuring post-programming conductance drift become a key performance-limiting factor. In this work, comprehensive study is presented from analysis its causes to strategy device optimization as well impact on CIM applications. An application-oriented quantitative indicator...
The understanding of the retention and endurance degradation behavior different levels filamentary analog RRAM is critical for development neuromorphic computing. This paper investigates conductance distribution during tests. low states high change from normal at beginning to asymmetric skewed with baking time increasing. But intermediate remain distribution. A model proposed predict evolution levels. It also found that lifetime depends on ratio position window. longer attributed switching...
Mott memristors have been considered as a promising candidate to implement artificial neurons for neuromorphic computing thanks their low-power consumption and superior scalability. However, the large variability poor reliability hinder large-scale applications. The complex working mechanism associated with thermoelectric coupling in correlated oxides such niobium oxide (NbOx) has led lack of physics-based model guide device optimizations. In this work, we present microscopic NbOx-based...
Thermoreflectance experiments are sensitive to the thermal properties of thin layers deposited on substrates (conductivity and diffusivity). However, retrieving these from experimental data remains a difficult issue. The case conducting layer an insulating substrate was studied previously. We present here mathematical analysis thermoreflectance response in opposite case: substrate. show theoretically that conductivity diffusivity can be determined independently thanks comparison with method...
A physics-based compact model of reliability degradation in analog resistive random access memory (RRAM) is developed. The captures the stochastic behaviors retention, bit yield, and endurance during switching. verified with statistical data measured from RRAM arrays. Based on this model, a device-to-system simulation framework for computation-in-memory (CIM) system This silicon-verified versatile simulator that supports both inference training, fully considers device nonideal effects...
Analog RRAM is considered as a promising emerging device technology for the future storage and neuromorphic computing. Different from long-term retention degradation, usually overlooked relaxation effect shows more significant impact on computing applications, which manifested in low energy efficiency of data mapping high accuracy loss application functions. In this work, we have statistically studied analog arrays. The random conductance fluctuation behaviors due to captured quantified with...
A new device structure based on Ge–Si core–shell nanowire is proposed. Owing to the fact that Si shell stripped at source/drain end, density and electrostatic potential of holes gradually distributed in region structure, which results a higher gradient quasi-Fermi channel underneath gate. We observe current ratio for on-off state improved output characteristics via TCAD simulation. Besides, proposed we can adjust on-state by changing length same gate length. This provides feasible approach...
Analog RRAM is considered as a promising emerging device for the future computation-in-memory system. However, relaxation effect shows significant impact on system performance. It causes high accuracy loss inference application. In this work, we have statistically studied of analog RRAM. Based statistical measurement results, compact model established. The simulation results can match experiment well different conductance levels at time. Furthermore, device-level to system-level framework...
The relaxation effect in analog resistive random access memory (RRAM) poses a significant challenge the implementation of neuromorphic systems, as it leads to loss accuracy computing. However, due inherent interdependence various fluctuations and underlying mechanisms, is still challenging. In this study, we have developed high-quality adaptive signal analysis method by analyzing read current during process. This enables identification all conductivity demarcation points categorizes them...
In this work, a metal gate with high effective work function (Weff) that is suitable for junctionless field-effect transistors (JLFETs) has been fabricated. Weff modulated by inserting an Al interfacial layer different thicknesses between the HfO2 dielectric and Pt metal. Transmission electron microscopy together capacitance–voltage (C–V) measurement used to investigate underlying mechanism of phenomenon. It suggested oxygen scavenging from leads formation interface dipoles, which...
Developing analog RRAM model for studying the effects of nonideal characteristic on neural network is necessary. In this work, we developed an compact considering temperature coefficient, I-V nonlinearity, variability, and programing which well consistent with data measured from array. Then evaluated accuracy MNIST task using based multilayer perceptron. Especially, coefficient computing studied.
In this paper, a novel route was developed to fabricate new flexible force sensor using carbon composites that consist of micro particles as conductive matrix, silicone rubber insulating matrix and elastomer fillers elastic matrix. The can not only show gradual change in electrical resistivity with applied quasi-static pressure, but also measure the changes compression stress relaxation soft substrates under it. Furthermore, it is thin (1.2 mm) enough be adhered on measured substrates, so...
With the rich internal ion dynamics, memristor-based neuromorphic computing emerges as a non-von Neumann paradigm to mimic biological neural networks and achieve high energy efficiency. However, implement large-scale memristive networks, reliability issue of devices, including artificial synapse, dendrite, soma, should be properly addressed. In this paper, recent works investigating physical mechanisms optimizations device are presented. particular, relaxation effect <tex...