Towards artificial general intelligence with hybrid Tianjic chip architecture.

There are two general approaches to developing artificial general intelligence (AGI)1: computer-science-oriented and neuroscience-oriented. Because of the fundamental differences in their formulations and coding schemes, these two approaches rely on distinct and incompatible platforms2-8, retarding the development of AGI. A general platform that could support the prevailing computer-science-based artificial neural networks as well as neuroscience-inspired models and algorithms is highly desirable. Here we present the Tianjic chip, which integrates the two approaches to provide a hybrid, synergistic platform. The Tianjic chip adopts a many-core architecture, reconfigurable building blocks and a streamlined dataflow with hybrid coding schemes, and can not only accommodate computer-science-based machine-learning algorithms, but also easily implement brain-inspired circuits and several coding schemes. Using just one chip, we demonstrate the simultaneous processing of versatile algorithms and models in an unmanned bicycle system, realizing real-time object detection, tracking, voice control, obstacle avoidance and balance control. Our study is expected to stimulate AGI development by paving the way to more generalized hardware platforms.

Reservoir computing with a random memristor crossbar array.

Physical implementations of reservoir computing (RC) based on the emerging memristors have become promising candidates of unconventional computing paradigms. Traditionally, sequential approaches by time-multiplexing volatile memristors have been prevalent because of their low hardware overhead. However, they suffer from the problem of speed degradation and fall short of capturing the spatial relationship between the time-domain inputs. Here, we explore a new avenue for RC using memristor crossbar arrays with device-to-device variations, which serve as physical random weight matrices of the reservoir layers, enabling faster computation thanks to the parallelism of matrix-vector multiplication as an intensive operation in RC. To achieve this new RC architecture, ultralow-current, self-selective memristors are fabricated and integrated without the need of transistors, showing greater potential of high scalability and three-dimensional integrability compared to the previous realizations. The information processing ability of our RC system is demonstrated in asks of recognizing digit images and waveforms. This work indicates that the 'nonidealities' of the emerging memristor devices and circuits are a useful source of inspiration for new computing paradigms.

A complementary resistive switching neuron.

The complementary resistive switching (CRS) memristor has originally been proposed for use as the storage element or artificial synapse in large-scale crossbar array with the capability of solving the sneak path problem, but its usage has mainly been hampered by the inherent destructiveness of the read operation (switching '1' state to 'ON' or '0' state). Taking a different perspective on this 'undesired' property, we here report on the inherent behavioral similarity between the CRS memristor and a leaky integrate-and-fire (LIF) neuron which is another basic neural computing element, in addition to synapse. In particular, the mechanism behind the undesired read destructiveness for storage element and artificial synapse can be exploited to naturally realize the LIF and the ensuing spontaneous repolarization processes, followed by a refractory period. By means of this biological similarity, we demonstrate a Pt/Ta2O5-x/TaOy/Ta CRS memristor that can exhibit these neuronal behaviors and perform various fundamental neuronal operations, including additive/subtractive operations and coincidence detection. These results suggest that the CRS neuron, with its bio-interpretability, is a useful addition to the family of memristive neurons.

Semiconducting few-layer PdSe2 and Pd2Se3: native point defects and contacts with native metallic Pd17Se15.

PdSe2 is a unique layered two-dimensional (2D) material with pentagonal structural motif and anisotropic properties. In addition, its strong interlayer interaction leads to new 2D form of the exfoliated monolayer, that is, Pd2Se3. Despite the increasing interest in these emerging 2D materials, the landscape of the native point defects, as a fundamental materials property, has not been revealed. In this work, we systematically investigate different types of defects in mono- and bi-layer PdSe2 and monolayer Pd2Se3. In contrast to the common expectation, Se vacancy is not the readily formed defect. Instead, Se-excess defects, such as SePd antisite and Se interstitial, are more likely to form over a majority of the allowed range of the atomic chemical potentials. Se-deficiency defect, Pd interstitial, is able to form under the Se-poor condition in bilayer PdSe2. The defect-mediated interlayer fusion model in the formation of monolayer Pd2Se3 from bilayer PdSe2 is reformulated. These dominant defects are found to stay in the neutral charge state, partly explaining the ambipolar behavior of the PdSe2 transistors. Finally, the stacked and lateral contacts between these few-layer semiconductors and the native Pd17Se15 metal are also studied. All these interfaces show p-type contact properties. This work reveals the important materials properties of few-layer PdSe2 and Pd2Se3 for the better development of new functional devices.

Nanoporous Al-Ni-Co-Ir-Mo High-Entropy Alloy for Record-High Water Splitting Activity in Acidic Environments.

Ir-based binary and ternary alloys are effective catalysts for the electrochemical oxygen evolution reaction (OER) in acidic solutions. Nevertheless, decreasing the Ir content to less than 50 at% while maintaining or even enhancing the overall electrocatalytic activity and durability remains a grand challenge. Herein, by dealloying predesigned Al-based precursor alloys, it is possible to controllably incorporate Ir with another four metal elements into one single nanostructured phase with merely ≈20 at% Ir. The obtained nanoporous quinary alloys, i.e., nanoporous high-entropy alloys (np-HEAs) provide infinite possibilities for tuning alloy's electronic properties and maximizing catalytic activities owing to the endless element combinations. Particularly, a record-high OER activity is found for a quinary AlNiCoIrMo np-HEA. Forming HEAs also greatly enhances the structural and catalytic durability regardless of the alloy compositions. With the advantages of low Ir loading and high activity, these np-HEA catalysts are very promising and suitable for activity tailoring/maximization.

Highly Compact Artificial Memristive Neuron with Low Energy Consumption.

Neuromorphic systems aim to implement large-scale artificial neural network on hardware to ultimately realize human-level intelligence. The recent development of nonsilicon nanodevices has opened the huge potential of full memristive neural networks (FMNN), consisting of memristive neurons and synapses, for neuromorphic applications. Unlike the widely reported memristive synapses, the development of artificial neurons on memristive devices has less progress. Sophisticated neural dynamics is the major obstacle behind the lagging. Here a rich dynamics-driven artificial neuron is demonstrated, which successfully emulates partial essential neural features of neural processing, including leaky integration, automatic threshold-driven fire, and self-recovery, in a unified manner. The realization of bioplausible artificial neurons on a single device with ultralow power consumption paves the way for constructing energy-efficient large-scale FMNN and may boost the development of neuromorphic systems with high density, low power, and fast speed.

LiNbO3 dynamic memristors for reservoir computing.

Information in conventional digital computing platforms is encoded in the steady states of transistors and processed in a quasi-static way. Memristors are a class of emerging devices that naturally embody dynamics through their internal electrophyiscal processes, enabling nonconventional computing paradigms with enhanced capability and energy efficiency, such as reservoir computing. Here, we report on a dynamic memristor based on LiNbO3. The device has nonlinear I-V characteristics and exhibits short-term memory, suitable for application in reservoir computing. By time multiplexing, a single device can serve as a reservoir with rich dynamics which used to require a large number of interconnected nodes. The collective states of five memristors after the application of trains of pulses to the respective memristors are unique for each combination of pulse patterns, which is suitable for sequence data classification, as demonstrated in a 5 × 4 digit image recognition task. This work broadens the spectrum of memristive materials for neuromorphic computing.

Generative complex networks within a dynamic memristor with intrinsic variability.

Artificial neural networks (ANNs) have gained considerable momentum in the past decade. Although at first the main task of the ANN paradigm was to tune the connection weights in fixed-architecture networks, there has recently been growing interest in evolving network architectures toward the goal of creating artificial general intelligence. Lagging behind this trend, current ANN hardware struggles for a balance between flexibility and efficiency but cannot achieve both. Here, we report on a novel approach for the on-demand generation of complex networks within a single memristor where multiple virtual nodes are created by time multiplexing and the non-trivial topological features, such as small-worldness, are generated by exploiting device dynamics with intrinsic cycle-to-cycle variability. When used for reservoir computing, memristive complex networks can achieve a noticeable increase in memory capacity a and respectable performance boost compared to conventional reservoirs trivially implemented as fully connected networks. This work expands the functionality of memristors for ANN computing.

Twist-angle-controlled neutral exciton annihilation in WS2 homostructures.

Exciton-exciton annihilation (EEA), as typical nonradiative recombination, plays an unpopular role in semiconductors. The nonradiative process significantly reduces the quantum yield of photoluminescence, which substantially inhibits the maximum efficiency of optoelectronic devices. Recently, laser irradiation, introducing defects and applying strain have become effective means to restrain EEA in two-dimensional (2D) transition metal dichalcogenides (TMDCs). However, these methods destroy the atomic structure of 2D materials and limit their practical applications. Fortunately, twisted structures are expected to validly suppress EEA through excellent interface quality. Here, we develop a non-destructive way to control EEA in WS2 homostructures by changing the interlayer twist angle, and systematically study the effect of interlayer twist angle on EEA, using fluorescence lifetime imaging measurement (FLIM) technology. Due to the large moiré potential at a small interlayer twist angle, the diffusion of excitons is hindered, and the EEA rate decreases from 1.01 × 10-1 cm2 s-1 in a 9° twisted WS2 homostructure to 4.26 × 10-2 cm2 s-1 in a 1° twisted WS2 homostructure. The results reveal the important role of the interlayer twist angle and EEA interaction in high photoluminescence quantum yield optoelectronic devices based on TMDC homostructures.

Grain-Boundary Engineering of Monolayer MoS2 for Energy-Efficient Lateral Synaptic Devices.

Synaptic devices based on 2D-layered materials have emerged as high-efficiency electronic synapses and neurons for neuromorphic computing. Lateral 2D synaptic devices have the advantages of multiple functionalities by responding to diverse stimuli, but they consume large amounts of energy, far more than the human brain. Moreover, current lateral devices employ several mechanisms based on conductive filaments and grain boundaries (GBs), but their formation is random and difficult to control, also hindering their practical applications. Here, four-terminal, lateral synaptic devices with artificially engineered GBs are reported, which are made from monolayer MoS2 . With lithography-free, direct-laser-writing-controlled MoS2 /MoS2- x Oδ GBs, such synaptic devices exhibit short-term and long-term plasticity characteristics that are responsive to electric and light stimulation simultaneously. This enables detailed simulations of biological learning and cognitive processes as well as image perception and processing. In particular, the device exhibits low energy consumption, similar to that of the human brain and much lower than those of other lateral 2D synaptic devices. This work provides an effective way to fabricate lateral synaptic devices for practical application development and sheds light on controllable electrical state switching for neuromorphic computing.

A new opportunity for the emerging tellurium semiconductor: making resistive switching devices.

The development of the resistive switching cross-point array as the next-generation platform for high-density storage, in-memory computing and neuromorphic computing heavily relies on the improvement of the two component devices, volatile selector and nonvolatile memory, which have distinct operating current requirements. The perennial current-volatility dilemma that has been widely faced in various device implementations remains a major bottleneck. Here, we show that the device based on electrochemically active, low-thermal conductivity and low-melting temperature semiconducting tellurium filament can solve this dilemma, being able to function as either selector or memory in respective desired current ranges. Furthermore, we demonstrate one-selector-one-resistor behavior in a tandem of two identical Te-based devices, indicating the potential of Te-based device as a universal array building block. These nonconventional phenomena can be understood from a combination of unique electrical-thermal properties in Te. Preliminary device optimization efforts also indicate large and unique design space for Te-based resistive switching devices.

A New Opportunity for 2D van der Waals Heterostructures: Making Steep-Slope Transistors.

The use of a foreign metallic cold source (CS) has recently been proposed as a promising approach toward the steep-slope field-effect-transistor (FET). In addition to the selection of source material with desired density of states-energy relation (D(E)), engineering the source:channel interface for gate-tunable channel-barrier is crucial to CS-FETs. However, conventional metal:semiconductor (MS) interfaces generally suffer from strong Fermi-level pinning due to the inevitable chemical disorder and defect-induced gap states, precluding the gate tunability of the barriers. By comprehensive materials and device modeling at the atomic scale, it is reported that 2D van der Waals (vdW) MS interfaces, with their atomic sharpness and cleanness, can be considered as general ingredients for CS-FETs. As test cases, InSe-based n-type FETs are studied. It is found that graphene can be spontaneously p-type doped along with slightly opened bandgap around the Dirac-point by interfacing with InSe, resulting in superexponentially decaying hot carrier density with increasing n-type channel-barrier. Moreover, the D(E) relations suggest that 2D transition-metal dichalcogenides and 2D transition-metal carbides are a rich library of CS materials. Graphene, Cd3 C2 , T-VTe2 , H-VTe2 , and H-TaTe2 CSs lead to subthreshold swing below 60 mV dec-1 . This work broadens the application potentials of 2D vdW MS heterostructures and serves as a springboard for more studies on low-power electronics based on 2D materials.

Perovskite hetero-anionic-sublattice interfaces for optoelectronics and nonconventional electronics.

The perovskite structure provides a versatile framework for functional materials and their high-quality heteroepitaxial interfaces. Perovskite halides (PH) have attracted intense interest for their application in optoelectronics. Oxides are another major class of perovskites that are widely used in fuel cells, nonconventional electronics and electrochemistry. Interfacing different perovskite oxides (POs) has led to a multitude of fascinating discoveries. By introducing anionic degree of freedom, we expect that perovskite hetero-anionic-sublattice interfaces can provide a new platform for emergent phenomena that may or may not have homo-oxygen-sublattice interface analogues. In this work, we investigate the interfaces between the all-inorganic PH CsPbBr3, the emerging double perovskite halide (dPH) Cs2TiBr6 and various common POs. Based on the band alignment properties, these POs are considered to be suitable carrier transport materials (CTMs) for CsPbBr3 and Cs2TiBr6 in either light-harvesting or light-emitting devices. In addition, these perovskite hetero-anionic-sublattice interfaces are found to be defect- and dangling bond-free due to compatible crystal lattices, making POs potentially outperform conventional binary transition-metal-oxide and organic CTMs. Besides optoelectronics, the potential of perovskite hetero-anionic-sublattice interfaces for nonconventional electronics is also explored. As examples, two-dimensionally confined electron-hole systems are predicted at the asymmetric interfaces in both Cs2TiBr6:LaAlO3 and CsPbBr3:LaAlO3 superlattice structures. This finding, along with the optically active properties of PHs, may spark novel applications of light-electron interaction in perovskite systems. This work presents new opportunities for perovskite heteroepitaxial interfaces.

MOF-Derived 2D/3D Hierarchical N-Doped Graphene as Support for Advanced Pt Utilization in Ethanol Fuel Cell.

Development of bifunctional catalysts with low platinum (Pt) content for the ethanol oxidation reaction (EOR) and the oxygen reduction reaction (ORR) is highly desirable, yet challenging. Herein, we present structural engineering of a series of two-dimensional/three-dimensional (2D/3D) hierarchical N-doped graphene-supported nanosized Pt3Co alloys and Co clusters (PtCo@N-GNSs) via a hydrolysis-pyrolysis route. For the ORR, the optimal PtCo@N-GNS exhibits a high mass activity of 3.01 A mgPt-1, which is comparable to the best Pt-based catalyst obtained through sophisticated synthesis. It also possesses excellent stability with minor decay after 50 000 cyclic voltammograms (CV) cycles in acidic medium. For the EOR, PtCo@N-GNS achieves the highest mass-specific and area-specific activities of 1.96 A mgPt-1 and 5.75 mA cm-2, respectively, among all of the reported EOR catalysts to date. The unique 2D/3D hierarchy, high Pt utilization, and valid encapsulation of nanosized Pt3Co/Co synergistically contribute to the robust ORR and EOR activities of the present PtCo@N-GNS. A direct ethanol fuel cell based on PtCo@N-GNS delivers a high open-circuit potential of 0.9 V, a stable power density of 10.5 mW cm-2, and an excellent rate performance, implying the feasibility of the bifunctional PtCo@N-GNS. This work offers a new strategy for designing an ultralow Pt loading yet highly active and durable catalyst for ethanol fuel cell application.

Ultrahigh drive current and large selectivity in GeS selector.

Selector devices are indispensable components of large-scale nonvolatile memory and neuromorphic array systems. Besides the conventional silicon transistor, two-terminal ovonic threshold switching device with much higher scalability is currently the most industrially favored selector technology. However, current ovonic threshold switching devices rely heavily on intricate control of material stoichiometry and generally suffer from toxic and complex dopants. Here, we report on a selector with a large drive current density of 34 MA cm-2 and a ~106 high nonlinearity, realized in an environment-friendly and earth-abundant sulfide binary semiconductor, GeS. Both experiments and first-principles calculations reveal Ge pyramid-dominated network and high density of near-valence band trap states in amorphous GeS. The high-drive current capacity is associated with the strong Ge-S covalency and the high nonlinearity could arise from the synergy of the mid-gap traps assisted electronic transition and local Ge-Ge chain growth as well as locally enhanced bond alignment under high electric field.

Materials Selection and Mechanism of Non-linear Conduction in Chalcogenide Selector Devices.

The electronic structure and conduction mechanism of chalcogenide-based Ovonic threshold switches (OTS) used as selectors in cross-point memory arrays is derived from density functional calculations and quasi-Fermi level models. The switching mechanism in OTS is primarily electronic. This uses a specific electronic structure, with a wide tail of localized states below the conduction band edge. In amorphous GeSe2-x the conduction band consists of Ge-Se σ*states with a low effective mass, and with a broad tail of localized Ge-Ge σ* states below this band edge. This leads to the OTS behavior. At high fields the electron quasi-EF moves up through these tail states, lowering the conductivity activation energy, and giving the non-linear switching process. The 4:2 coordinated GeSe2-x based alloys are the most favorable OTS material because they have the correct network connectivity to give a high electron mobility and lack of crystallization, a favorable band structure to produce the non-linear conduction, an optimum band gap, and with nitrogen or carbon alloying, a sufficiently low off-current.

Self-Doping Memristors with Equivalently Synaptic Ion Dynamics for Neuromorphic Computing.

The accumulation and extrusion of Ca2+ ions in the pre- and post-synaptic terminals play crucial roles in initiating short- and long-term plasticity (STP and LTP) in biological synapses, respectively. Mimicking these synaptic behaviors by electronic devices represents a vital step toward realization of neuromorphic computing. However, the majority of reported synaptic devices usually focus on the emulation of qualitatively synaptic behaviors; devices that can truly emulate the physical behavior of the synaptic Ca2+ ion dynamics in STP and LTP are rarely reported. In this work, Ag/Ag:Ta2O5/Pt self-doping memristors were developed to equivalently emulate the Ca2+ ion dynamics of biological synapses. With conductive filaments from double sources, these memristors produced unique double-switching behavior under voltage sweeps and demonstrated several essential synaptic behaviors under pulse stimuli, including STP, LTP, STP to LTP transition, and spike-rate-dependent plasticity. Experimental results and nanoparticle dynamic simulations both showed that Ag atoms from double sources could mimic Ca2+ dynamics in the pre- and post-synaptic terminals under stimuli. A perceptron network with an STP to LTP transition layer based on the self-doping memristors was also introduced and evaluated; simulations showed that this network could solve noisy figure recognition tasks efficiently. All of these results indicate that the self-doping memristors are promising components for future hardware creation of neuromorphic systems and emulate the characteristics of the brain.

Enhanced Sulfur Transformation by Multifunctional FeS2/FeS/S Composites for High-Volumetric Capacity Cathodes in Lithium-Sulfur Batteries.

Lithium-sulfur batteries are currently being explored as promising advanced energy storage systems due to the high theoretical specific capacity of sulfur. However, achieving a scalable synthesis for the sulfur electrode material whilst maintaining a high volumetric energy density remains a serious challenge. Here, a continuous ball-milling route is devised for synthesizing multifunctional FeS2/FeS/S composites for use as high tap density electrodes. These composites demonstrate a maximum reversible capacity of 1044.7 mAh g-1 and a peak volumetric capacity of 2131.1 Ah L-1 after 30 cycles. The binding direction is also considered here for the first time between dissolved lithium polysulfides (LiPSs) and host materials (FeS2 and FeS in this work) as determined by density functional theory calculations. It is concluded that if only one lithium atom of the polysulfide bonds with the sulfur atoms of FeS2 or FeS, then any chemical interaction between these species is weak or negligible. In addition, FeS2 is shown to have a strong catalytic effect on the reduction reactions of LiPSs. This work demonstrates the limitations of a strategy based on chemical interactions to improve cycling stability and offers new insights into the development of high tap density and high-performance sulfur-based electrodes.

Metal and Nonmetal Codoped 3D Nanoporous Graphene for Efficient Bifunctional Electrocatalysis and Rechargeable Zn-Air Batteries.

Developing bifunctional electrocatalysts with high activities and long durability for the oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) is crucial toward the practical implementation of rechargeable metal-air batteries. Here, a 3D nanoporous graphene (np-graphene) doped with both N and Ni single atoms/clusters is reported. The predoping of N by chemical vapor deposition (CVD) dramatically increases the Ni doping amount and stability. The resulting N and Ni codoped np-graphene has excellent electrocatalytic activities for both the ORR and the OER in alkaline aqueous solutions. The synergetic effects of N and Ni dopants are revealed by density functional theory calculations. The free-standing Ni,N codoped 3D np-graphene shows great potential as an economical catalyst/electrode for metal-air batteries.

Truly Concomitant and Independently Expressed Short- and Long-Term Plasticity in a Bi2 O2 Se-Based Three-Terminal Memristor.

Concomitance of diverse synaptic plasticity across different timescales produces complex cognitive processes. To achieve comparable cognitive complexity in memristive neuromorphic systems, devices that are capable of emulating short-term (STP) and long-term plasticity (LTP) concomitantly are essential. In existing memristors, however, STP and LTP can only be induced selectively because of the inability to be decoupled using different loci and mechanisms. In this work, the first demonstration of truly concomitant STP and LTP is reported in a three-terminal memristor that uses independent physical phenomena to represent each form of plasticity. The emerging layered material Bi2 O2 Se is used for memristors for the first time, opening up the prospects for ultrathin, high-speed, and low-power neuromorphic devices. The concerted action of STP and LTP allows full-range modulation of the transient synaptic efficacy, from depression to facilitation, by stimulus frequency or intensity, providing a versatile device platform for neuromorphic function implementation. A heuristic recurrent neural circuitry model is developed to simulate the intricate "sleep-wake cycle autoregulation" process, in which the concomitance of STP and LTP is posited as a key factor in enabling this neural homeostasis. This work sheds new light on the development of generic memristor platforms for highly dynamic neuromorphic computing.