Complementary memtransistors for neuromorphic computing: How, what and why

Memtransistors in which the source-drain channel conductance can be nonvolatilely manipulated through the gate signals have emerged as promising components for implementing neuromorphic computing.On the other side,it is known that the complementary metal-oxide-semiconductor(CMOS)field effect transistors have played the fundamental role in the modern integrated circuit technology.Therefore,will complementary memtransistors(CMT)also play such a role in the future neuromorphic circuits and chips?In this review,various types of materials and physical mechanisms for constructing CMT(how)are inspected with their merits and need-to-address challenges discussed.Then the unique properties(what)and poten-tial applications of CMT in different learning algorithms/scenarios of spiking neural networks(why)are reviewed,including super-vised rule,reinforcement one,dynamic vision with in-sensor computing,etc.Through exploiting the complementary structure-related novel functions,significant reduction of hardware consuming,enhancement of energy/efficiency ratio and other advan-tages have been gained,illustrating the alluring prospect of design technology co-optimization(DTCO)of CMT towards neuro-morphic computing.

Reconfigurable and polarization-dependent optical filtering for transflective full-color generation utilizing low-loss phase-change materials

All-dielectric metasurface, which features low optical absorptance and high resolution, is becoming a promising candidate for full-color generation. However, the optical response of current metamaterials is fixed and lacks active tuning. In this work, we demonstrate a reconfigurable and polarization-dependent active color generation technique by incorporating low-loss phase change materials(PCMs) and CaF_2 all-dielectric substrate. Based on the strong Mie resonance effect and low optical absorption structure, a transflective, full-color with high color purity and gamut value is achieved. The spectrum can be dynamically manipulated by changing either the polarization of incident light or the PCM state. High transmittance and reflectance can be simultaneously achieved by using low-loss PCMs and substrate. The novel active metasurfaces can bring new inspiration in the areas of optical encryption, anti-counterfeiting, and display technologies.

Multiply accumulate operations in memristor crossbar arrays for analog computing

Memristors are now becoming a prominent candidate to serve as the building blocks of non-von Neumann inmemory computing architectures.By mapping analog numerical matrices into memristor crossbar arrays,efficient multiply accumulate operations can be performed in a massively parallel fashion using the physics mechanisms of Ohm’s law and Kirchhoff’s law.In this brief review,we present the recent progress in two niche applications:neural network accelerators and numerical computing units,mainly focusing on the advances in hardware demonstrations.The former one is regarded as soft computing since it can tolerant some degree of the device and array imperfections.The acceleration of multiple layer perceptrons,convolutional neural networks,generative adversarial networks,and long short-term memory neural networks are described.The latter one is hard computing because the solving of numerical problems requires high-precision devices.Several breakthroughs in memristive equation solvers with improved computation accuracies are highlighted.Besides,other nonvolatile devices with the capability of analog computing are also briefly introduced.Finally,we conclude the review with discussions on the challenges and opportunities for future research toward realizing memristive analog computing machines.

Volatile threshold switching memristor:An emerging enabler in the AIoT era

With rapid advancement and deep integration of artificial intelligence and the internet-of-things,artificial intelligence of things has emerged as a promising technology changing people’s daily life.Massive growth of data generated from the devices challenges the AIoT systems from information collection,storage,processing and communication.In the review,we introduce volatile threshold switching memristors,which can be roughly classified into three types:metallic conductive filament-based TS devices,amorphous chalcogenide-based ovonic threshold switching devices,and metal-insulator transition based TS devices.They play important roles in high-density storage,energy efficient computing and hardware security for AIoT systems.Firstly,a brief introduction is exhibited to describe the categories(materials and characteristics)of volatile TS devices.And then,switching mechanisms of the three types of TS devices are discussed and systematically summarized.After that,attention is focused on the applications in 3D cross-point memory technology with high storage-density,efficient neuromorphic computing,hardware security(true random number generators and physical unclonable functions),and others(steep subthreshold slope transistor,logic devices,etc.).Finally,the major challenges and future outlook of volatile threshold switching memristors are presented.

An artificial synapse by superlattice-like phase-change material for low-power brain-inspired computing

Phase-change material(PCM)is generating widespread interest as a new candidate for artificial synapses in bioinspired computer systems.However,the amorphization process of PCM devices tends to be abrupt,unlike continuous synaptic depression.The relatively large power consumption and poor analog behavior of PCM devices greatly limit their applications.Here,we fabricate a GeTe/Sb2Te3 superlattice-like PCM device which allows a progressive RESET process.Our devices feature low-power consumption operation and potential high-density integration,which can effectively simulate biological synaptic characteristics.The programming energy can be further reduced by properly selecting the resistance range and operating method.The fabricated devices are implemented in both artificial neural networks(ANN)and convolutional neural network(CNN)simulations,demonstrating high accuracy in brain-like pattern recognition.

A family of flexible two-dimensional semiconductors:MgMX2Y6(M=Ti/Zr/Hf;X=Si/Ge;Y=S/Se/Te)

Inspired by the recently predicted 2D MX_(2)Y_(6)(M=metal element;X=Si/Ge/Sn;Y=S/Se/Te),we explore the possible applications of alkaline earth metal(using magnesium as example)in this family based on the idea of element replacement and valence electron balance.Herein,we report a new family of 2D quaternary compounds,namely MgMX_(2)Y_(6)(M=Ti/Zr/Hf;X=Si/Ge;Y=S/Se/Te)monolayers,with superior kinetic,thermodynamic and mechanical stability.In addition,our results indicate that MgMX_(2)Y_(6)monolayers are all indirect band gap semiconductors with band gap values ranging from 0.870 to 2.500 eV.Moreover,the band edges and optical properties of 2D MgMX_(2)Y_(6)are suitable for constructing multifunctional optoelectronic devices.Furthermore,for comparison,the mechanical,electronic and optical properties of In_(2)X_(2)Y_(6)monolayers have been discussed in detail.The success of introducing Mg into the 2D MX_(2)Y_(6)family indicates that more potential materials,such as Caand Sr-based 2D MX_(2)Y_(6)monolayers,may be discovered in the future.Therefore,this work not only broadens the existing family of 2D semiconductors,but it also provides beneficial results for the future.

Forward stagewise regression with multilevel memristor for sparse coding

Sparse coding is a prevalent method for image inpainting and feature extraction,which can repair corrupted images or improve data processing efficiency,and has numerous applications in computer vision and signal processing.Recently,sev-eral memristor-based in-memory computing systems have been proposed to enhance the efficiency of sparse coding remark-ably.However,the variations and low precision of the devices will deteriorate the dictionary,causing inevitable degradation in the accuracy and reliability of the application.In this work,a digital-analog hybrid memristive sparse coding system is pro-posed utilizing a multilevel Pt/Al_(2)O_(3)/AlO_(x)/W memristor,which employs the forward stagewise regression algorithm:The approxi-mate cosine distance calculation is conducted in the analog part to speed up the computation,followed by high-precision coeffi-cient updates performed in the digital portion.We determine that four states of the aforementioned memristor are sufficient for the processing of natural images.Furthermore,through dynamic adjustment of the mapping ratio,the precision require-ment for the digit-to-analog converters can be reduced to 4 bits.Compared to the previous system,our system achieves higher image reconstruction quality of the 38 dB peak-signal-to-noise ratio.Moreover,in the context of image inpainting,images containing 50%missing pixels can be restored with a reconstruction error of 0.0424 root-mean-squared error.

Ultrasound:A new strategy for artificial synapses modulation

Due to its non-invasive nature,ultrasound has been widely used for neuromodulation in biological systems,where its application influences the synaptic weights and the process of neurotransmitter delivery.However,such modulation has not been emulated in physical devices.Memristors are ideal electrical components for artificial synapses,but up till now they are hardly reported to respond to ultrasound signals.Here we design and fabricate a HfOx-based memristor on 64Y-X LiNbO_(3) single crystal substrate,and successfully realize artificial synapses modulation by shear-horizontal surface acoustic wave(SH-SAW).It is a prominent short-term resistance modulation,where ultrasound has been shown to cause resistance drop for various resistance states,which could fully recover after the ultrasound is shut off.The physical mechanism illustrates that ultrasound induced polarization potential in the HfOx dielectric layer acts on the Schottky barrier,leading to the resistance drop.The emulation of neuron firing frequency modulation through ultrasound signals is demonstrated.Moreover,the joint application of ultrasound and electric voltage yields fruitful functionalities,such as the enhancement of resistance window and synaptic plasticity through ultrasound application.All these promising results provide a new strategy for artificial synapses modulation,and also further advance neuromorphic devices toward system applications.

650 ps SET speed in Ge_(2)Sb_(2)Te_(5)phase change memory induced by TiO_(2) dielectric crystal plane

Crystallization speed of phase change material is one of the main obstaclesfor the application of phase change memory(PCM)as storage classmemory in computing systems,which requires the combination ofnonvolatility with ultra-fast operation speed in nanoseconds.Here,wepropose a novel approach to speed up crystallization process of the onlycommercial phase change chalcogenide Ge_(2)Sb_(2)Te_(5)(GST).By employingTiO_(2)as the dielectric layer in phase change device,operation speed of650 ps has been achieved,which is the fastest among existing representativePCM,and is comparable to the programing speed of commercialdynamic random access memory(DRAM).Because of its octahedralatomic configuration,TiO_(2)can provide nucleation interfaces for GST,thus facilitating the crystal growth at the determinate interface area.Ti–O–Ti–O four-fold rings on the(110)plane of tetragonal TiO_(2)is critical forthe fast-atomic rearrangement in the amorphous matrix of GST thatenables ultra-fast operation speed.The significant improvement of operationspeed in PCM through incorporating standard dielectric materialTiO_(2)in DRAM paves the way for the application of phase change memoryin high performance cache-type data storage.

Deep machine learning unravels the structural origin of mid-gap states in chalcogenide glass for high-density memory integration

The recent development of three-dimensional semiconductor integration technology demands a key component-the ovonic threshold switching(OTS)selector to suppress the current leakage in the high-density memory chips.Yet,the unsatisfactory performance of existing OTS materials becomes the bottleneck of the industrial advancement.The sluggish development of OTS materials,which are usually made from chalcogenide glass,should be largely attributed to the insufficient understanding of the electronic structure in these materials,despite of intensive research in the past decade.Due to the heavy first-principles computation on disordered systems,a universal theory to explain the origin of mid-gap states(MGS),which are the key feature leading to the OTS behavior,is still lacking.To avoid the formidable computational tasks,we adopt machine learning method to understand and predict MGS in typical OTS materials.We build hundreds of chalcogenide glass models and collect major structural features from both short-range order(SRO)and medium-range order(MRO)of the amorphous cells.After training the artificial neural network using these features,the accuracy has reached~95%when it recognizes MGS in new glass.By analyzing the synaptic weights of the input structural features,we discover that the bonding and coordination environments from SRO and particularly MRO are closely related to MGS.The trained model could be used in many other OTS chalcogenides after minor modification.The intelligent machine learning allows us to understand the OTS mechanism from vast amount of structural data without heavy computational tasks,providing a new strategy to design functional amorphous materials from first principles.

Toward memristive in-memory computing:principles and applications

With the rapid growth of computer science and big data,the traditional von Neumann architecture suffers the aggravating data communication costs due to the separated structure of the processing units and memories.Memristive in-memory computing paradigm is considered as a prominent candidate to address these issues,and plentiful applications have been demonstrated and verified.These applications can be broadly categorized into two major types:soft computing that can tolerant uncertain and imprecise results,and hard computing that emphasizes explicit and precise numerical results for each task,leading to different requirements on the computational accuracies and the corresponding hardware solutions.In this review,we conduct a thorough survey of the recent advances of memristive in-memory computing applications,both on the soft computing type that focuses on artificial neural networks and other machine learning algorithms,and the hard computing type that includes scientific computing and digital image processing.At the end of the review,we discuss the remaining challenges and future opportunities of memristive in-memory computing in the incoming Artificial Intelligence of Things era.

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.

Improved multilevel storage capacity in Ge_(2)Sb_(2)Te_(5)-based phase-change memory using a high-aspect-ratio lateral structure

Further improvement of storage density is a key challenge for the application of phase-change memory(PCM)in storage-class memory.However,for PCM,storage density improvements include feature size scaling down and multilevel cell(MLC)operation,potentially causing thermal crosstalk issues and phase separation issues,respectively.To address these challenges,we propose a high-aspect-ratio(25:1)lateral nanowire(NW)PCM device with conventional chalcogenide Ge_(2)Sb_(2)Te_(5)(GST-225)to realize stable MLC operations,i.e.,low intra-and inter-cell variability and low resistance drift(coefficient=0.009).The improved MLC performance is attributed to the high aspect ratio,which enables precise control of the amorphous region because of sidewall confinement,as confirmed by transmission electron microscopy analysis.In summary,the NW devices provide guidance for the design of future high-aspect-ratio threedimensional PCM devices with MLC capability.

Self-selective memristor-enabled in-memory search for highly efficient data mining

Similarity search,that is,finding similar items in massive data,is a fundamental computing problem in many fields such as data mining and information retrieval.However,for large-scale and high-dimension data,it suffers from high computational complexity,requiring tremendous computation resources.Here,based on the low-power self-selective memristors,for the first time,we propose an in-memory search(IMS)system with two innovative designs.First,by exploiting the natural distribution law of the devices resistance,a hardware locality sensitive hashing encoder has been designed to transform the realvalued vectors into more efficient binary codes.Second,a compact memristive ternary content addressable memory is developed to calculate the Hamming distances between the binary codes in parallel.Our IMS system demonstrated a 168energy efficiency improvement over all-transistors counterparts in clustering and classification tasks,while achieving a software-comparable accuracy,thus providing a low-complexity and low-power solution for in-memory data mining applications.

Tailoring the oxygen concentration in Ge-Sb-O alloys to enable femtojoule-level phase-change memory operations

Chalcogenide phase-change materials(PCMs),in particular,the flagship Ge2Sb2Te5(GST),are leading candidates for advanced memory applications.Yet,GST in conventional devices suffer from high power consumption,because the RESET operation requires melting of the crystalline GST phase.Recently,we have developed a conductive-bridge scheme for low-power phase-change application utilizing a self-decomposed Ge-Sb-O(GSO)alloy.In this work,we present thorough structural and electrical characterizations of GSO thin films by tailoring the concentration of oxygen in the phase-separating GSO system.We elucidate a two-step process in the as-deposited amorphous film upon the introduction of oxygen:with increasing oxygen doping level,germanium oxides form first,followed by antimony oxides.To enable the conductive-bridge switching mode for femtojoule-level RESET energy,the oxygen content should be sufficiently low to keep the antimony-rich domains easily crystallized under external electrical stimulus.Our work serves as a useful example to exploit alloy decomposition that develops heterogeneous PCMs,minimizing the active switching volume for low-power electronics.

A special issue on Optical Storage

The use of optics for data storage could trace back to 40 years ago when the researchers explored the possibilities of optical technology to record and playback audio signals. Optical storage devices have great advantages in low cost, high reliability and good compatibility, which make them suitable for data backup, digital publications, audio ＆ visual products, and so on. To meet with the increasing demand for information storage in the era of big data, optical storage technologies are actively involved in continuous improvement of ultra-high densities and data transfer rates. Several promising technologies have grained growing interest, such as holographic data storage （HDS）, multi-layer optical storage, multi-dimensional optical storage and super-resolution near-field structure （super-RENS） optical storage, they are with great expectations because they utilize the volume of optical materials to store information and show great potentials in capacity, density and speed.