Optimized operation scheme of flash-memory-based neural network online training with ultra-high endurance

With the rapid development of machine learning,the demand for high-efficient computing becomes more and more urgent.To break the bottleneck of the traditional Von Neumann architecture,computing-in-memory(CIM)has attracted increasing attention in recent years.In this work,to provide a feasible CIM solution for the large-scale neural networks(NN)requiring continuous weight updating in online training,a flash-based computing-in-memory with high endurance(10^(9) cycles)and ultrafast programming speed is investigated.On the one hand,the proposed programming scheme of channel hot electron injection(CHEI)and hot hole injection(HHI)demonstrate high linearity,symmetric potentiation,and a depression process,which help to improve the training speed and accuracy.On the other hand,the low-damage programming scheme and memory window(MW)optimizations can suppress cell degradation effectively with improved computing accuracy.Even after 109 cycles,the leakage current(I_(off))of cells remains sub-10pA,ensuring the large-scale computing ability of memory.Further characterizations are done on read disturb to demonstrate its robust reliabilities.By processing CIFAR-10 tasks,it is evident that~90%accuracy can be achieved after 109 cycles in both ResNet50 and VGG16 NN.Our results suggest that flash-based CIM has great potential to overcome the limitations of traditional Von Neumann architectures and enable high-performance NN online training,which pave the way for further development of artificial intelligence(AI)accelerators.

Study on Simulation and Profile Prediction of Atomic Layer Deposition

The Atomic Layer Deposition process(ALD)is widely used in FinFET,3D-NAND and other important technologies because of its self-limiting signature and low growth temperature.In recent years,the development of computer enables chances for ALD process simulation in order to improve the process R&D efficiency.In this paper,steady state theory and vacuum pump theory are implemented to develop the growth rate algorithm of atomic layer deposition.The dynamic evolution of the deposition profile is realized based on cellular automata method,and fits the relationship between temperature and growth rate in HfO2 deposition.The model accuracy and simulation results are verified with high reliability.Based on the simulation results of this model,the influence of different substrate size and environmental dose on growth rate of pore structure is studied and analyzed.In the case of deep hole,high depth-to-width ratio hole,or when the gas entry time is below saturation,the growth rate decreases at the pore bottom.Meanwhile,the simulation considering the angle-of-inclination of the hole’s tapered sidewall indicates that the greater the angle,the better the distribution of flux.

Flash-based in-memory computing for stochastic computing in image edge detection

The“memory wall”of traditional von Neumann computing systems severely restricts the efficiency of data-intensive task execution,while in-memory computing(IMC)architecture is a promising approach to breaking the bottleneck.Although variations and instability in ultra-scaled memory cells seriously degrade the calculation accuracy in IMC architectures,stochastic computing(SC)can compensate for these shortcomings due to its low sensitivity to cell disturbances.Furthermore,massive parallel computing can be processed to improve the speed and efficiency of the system.In this paper,by designing logic functions in NOR flash arrays,SC in IMC for the image edge detection is realized,demonstrating ultra-low computational complexity and power consumption(25.5 fJ/pixel at 2-bit sequence length).More impressively,the noise immunity is 6 times higher than that of the traditional binary method,showing good tolerances to cell variation and reliability degradation when implementing massive parallel computation in the array.

Memory materials and devices:From concept to application

Memory cells have always been an important element of information technology.With emerging technologies like big data and cloud computing,the scale and complexity of data storage has reached an unprecedented peak with a much higher requirement for memory technology.As is well known,better data storage is mostly achieved by miniaturization.However,as the size of the memory device is reduced,a series of problems,such as drain gate-induced leakage,greatly hinder the performance of memory units.To meet the increasing demands of information technology,novel and high-performance memory is urgently needed.Fortunately,emerging memory technologies are expected to improve memory performance and drive the information revolution.This review will focus on the progress of several emerging memory technologies,including two-dimensional material-based memories,resistance random access memory(RRAM),magnetic random access memory(MRAM),and phasechange random access memory(PCRAM).Advantages,mechanisms,and applications of these diverse memory technologies will be discussed in this review.

AlO_x prepared by atomic layer deposition for high efficiency-type crystalline silicon solar cell

The influence of atomic layer deposition parameters on the negative charge density in AlOx film is investigated by the corona-charge measurement. Results show that the charge density can reach up to -1.56×10^12 cm%-2 when the thickness of the film is 2.4 nm. The influence of charge density on cell conversion efficiency is further simulated using solar cell analyzing software （PC1D）. With AlOx passivating the rear surface of the silicon, the cell efficiency of 20.66% can be obtained.

Modulation of Microstructure and Charge Transport in Polymer Monolayer Transistors by Solution Aging

A few monolayers of organic semiconductors adjacent to the dielectric layer are of vital importance in organic field-effect transistors due to their dominant role in charge transport.In this report,the 2-nm-thick polymer monolayers based on poly(3-hexylthiophene)with different molecular weights(M_(n))were fabricated using dip-coating technique.During the monolayer(solid state)formation from the solution,a disorder-to-order transition of polymer conformation is observed through UV-vis absorption measurement.Meanwhile,high Mn polymer monolayer generates higher crystalline fibrillar microstructure than the low Mn one due to the strongerπ–πintermolecular packing between polymers.More importantly,the solution aging procedure is utilized to further improve the morphology of polymer monolayers.It is obvious that after aging for 6 d,both fiber dimension and density as well as conjugation length are significantly increased under the same processing conditions in comparison to the fresh solution,and consequently the field-effect mobilities are remarkably enhanced by 2—4 times.Note that the maximum mobility of 0.027 cm2·V^(–1)·s^(–1)is among the highest reported values for poly(3-hexylthiophene)monolayer transistors.These results demonstrate a simple but powerful strategy for boosting the device performance of polymer monolayer transistors.

Redox Memristors with Volatile Threshold Switching Behavior for Neuromorphic Computing

The spiking neural network(SNN),closely inspired by the human brain,is one of the most powerful platforms to enable highly efficient,low cost,and robust neuromorphic computations in hardware using traditional or emerging electron devices within an integrated system.In the hardware implementation,the building of artificial spiking neurons is fundamental for constructing the whole system.However,with the slowing down of Moore’s Law,the traditional complementary metal-oxide-semiconductor(CMOS)technology is gradually fading and is unable to meet the growing needs of neuromorphic computing.Besides,the existing artificial neuron circuits are complex owing to the limited bio-plausibility of CMOS devices.Memristors with volatile threshold switching(TS)behaviors and rich dynamics are promising candidates to emulate the biological spiking neurons beyond the CMOS technology and build high-efficient neuromorphic systems.Herein,the state-of-the-art about the fundamental knowledge of SNNs is reviewed.Moreover,we review the implementation of TS memristor-based neurons and their systems,and point out the challenges that should be further considered from devices to circuits in the system demonstrations.We hope that this review could provide clues and be helpful for the future development of neuromorphic computing with memristors.

XOR spin logic operated by unipolar current based on field-free spin-orbit torque switching induced by a lateral interface

Spin logics have emerged as a promising avenue for the development of logic-in-memory architectures.In particular,the realization of XOR spin logic gates using a single spin-orbit torque device shows great potential for low-power stateful logic circuits in the next generation.In this study,we successfully obtained the XOR logic gate by utilizing a spin-orbit torque device with a lateral interface,which was created by local ion implantation in the Ta/Pt/Co/Ta Hall device exhibiting perpendicular magnetic anisotropy.The angle of the lateral interface is set at 45°relative to the current direction,leading to the competition between symmetry breaking and current-driven Néel-type domain wall motion.Consequently,the field-free magnetic switching reversed is realized by the same sign of current amplitude at this interface.Based on this field-free magnetic switching behavior,we successfully proposed an XOR logic gate that could be implemented using only a single spin-orbit torque Hall device.This study provides a potentially viable approach toward efficient spin logics and in-memory computing architectures.

Microscopic mechanism of imprint in hafnium oxide-based ferroelectrics

Hafnia-based ferroelectrics have greatly revived the field of ferroelectric memory(FeRAM),but certain reliability issues must be satisfactorily resolved before they can be widely applied in commercial memories.In particular,the imprint phenomenon severely jeopardizes the read-out reliability in hafnia-based ferroelectric capacitors,but its origin remains unclear,which hinders the development of its recovery schemes.In this work,we have systematically investigated the imprint mechanism in TiN/Hf_(0.5)Zr_(0.5)O_(2)(HZO)/TiN ferroelectric capacitors using experiments and first-principles calculations.It is shown that carrier injection-induced charged oxygen vacancies are at the heart of imprint in HZO,where other mechanisms such as domain pinning and dead layer are less important.An imprint model based on electron de-trapping from oxygen vacancy sites has been proposed that can satisfactorily explain several experimental facts such as the strong asymmetric imprint,leakage current variation,and so forth.Based on this model,an effective imprint recovery method has been proposed,which utilizes unipolar rather than bipolar voltage inputs.The remarkable recovery performances demonstrate the prospect of improved device reliability in hafnia-based FeRAM devices.

Flexible Hf_(0.5)Zr_(0.5)O_(2)ferroelectric thin films on polyimide with improved ferroelectricity and high flexibility

Flexible memory devices are promising for information storage and data processing applications in portable,wearable,and smart electronics operating under curved conditions.In this work,we realized high-performance flexible ferroelectric capacitors based on Hf_(0.5)Zr_(0.5)O_(2)(HZO)thin film by depositing a buffer layer of Al_(2)O_(3)on polyimide(PI)substrates using atomic layer deposition(ALD).The flexible ferroelectric HZO films exhibit high remnant polarization(Pr)of 21μC/cm^(2).Furthermore,deterioration of polarization,retention,and endurance performance was not observed even at a bending radius of 2 mm after 5,000 bending cycles.This work marks a critical step in the development of high-performance flexible HfO_(2)-based ferroelectric memories for next-generation wearable electronic devices.