Resistive memory technologies could be used to create intelligent systems that learn locally at the edge. However, current approaches typically use learning algorithms that cannot be reconciled with the intrinsic non-idealities of resistive memory, particularly cycle-to-cycle variability. Here, we report a machine learning scheme that exploits memristor variability to implement Markov chain Monte Carlo

A flexible biosensing system with in-sensor machine-learning functionality can recognize up to 21 hand gestures in real time based on surface electromyography patterns from a forearm.

Hardware for deep neural network (DNN) inference often suffers from insufficient on-chip memory, thus requiring accesses to separate memory-only chips. Such off-chip memory accesses incur considerable costs in terms of energy and execution time. Fitting entire DNNs in on-chip memory is challenging due, in particular, to the physical size of the technology. Here, we report a DNN inference system—termed

Electronics with skin- or tissue-like mechanical properties, including low stiffness and high stretchability, can be used to create intelligent technologies for application in areas such as health monitoring and human–machine interactions. Stretchable transistors that provide signal-processing and computational functions will be central to the development of this technology. Here, we review the development

Wearable devices that monitor muscle activity based on surface electromyography could be of use in the development of hand gesture recognition applications. Such devices typically use machine-learning models, either locally or externally, for gesture classification. However, most devices with local processing cannot offer training and updating of the machine-learning model during use, resulting in

Technology breakthroughs at the 2020 IEEE International Electron Devices Meeting, which this year takes place online.

Two-dimensional semiconductors have a range of electronic and optical properties that can be used in the development of advanced electronic devices. However, unlike conventional silicon semiconductors, simple doping methods to monolithically assemble n- and p-type channels on a single two-dimensional semiconductor are lacking, which makes the fabrication of integrated circuitry challenging. Here we

Metasurfaces, which consist of arrays of subwavelength scatterers, can be used to precisely control incident electromagnetic fields, but are typically static once fabricated. A dynamically programmable array of terahertz meta-elements, in which each element can be individually reconfigured to allow controlled wavefront shaping, could be of value in terahertz applications such as wireless communication

The development of small, energy-efficient artificial intelligence edge devices is limited in conventional computing architectures by the need to transfer data between the processor and memory. Non-volatile compute-in-memory (nvCIM) architectures have the potential to overcome such issues, but the development of high-bit-precision configurations required for dot-product operations remains challenging

Isolated point defects in silicon that emit light at telecom wavelengths could help accelerate the development of quantum information technologies using commercial platforms.

Security is a critical aspect in modern circuit design, but research into hardware security at the device level is rare as it requires modification of existing technology nodes. With the increasing challenges facing the semiconductor industry, interest in out-of-the-box security solutions has grown, even if this implies introducing novel materials such as two-dimensional layered semiconductors. Here

A magnonic directional coupler based on yttrium iron garnet could be used to create integrated magnonic nanocircuits for logic operations.

Spin–orbit torque magnetization switching is an efficient method to control magnetization. In perpendicularly magnetized films, two types of spin–orbit torque are induced by driving a current: a damping-like torque and a field-like torque. The damping-like torque assists magnetization switching, but a large field-like torque pushes the magnetization towards the in-plane direction, resulting in a larger

Antiferromagnets are of potential use in the development of spintronic devices due to their ultrafast dynamics, insensitivity to external magnetic fields and absence of magnetic stray fields. Similar to their ferromagnetic counterparts, antiferromagnets can store information in the orientations of the collective magnetic order vector. However, the readout magnetoresistivity signals in simple antiferromagnetic

The electrical manipulation of magnetization and exchange bias in antiferromagnet/ferromagnet thin films could be of use in the development of the next generation of spintronic devices. Current-controlled magnetization switching can be driven by spin–orbit torques generated in an adjacent heavy-metal layer, but these structures are difficult to integrate with exchange bias switching and tunnelling

Given its potential for integration and scalability, silicon is likely to be a key platform for large-scale quantum technologies. Individual electron-encoded artificial atoms, formed by either impurities or quantum dots, have emerged as a promising solution for silicon-based integrated quantum circuits. However, single qubits featuring an optical interface, which is needed for long-distance exchange

Near-term applications of quantum information processors will rely on optimized circuit implementations to minimize the circuit depth, reducing the negative impact of gate errors in noisy intermediate-scale quantum (NISQ) computers. One approach to minimize the circuit depth is the use of a more expressive gate set. The XY two-qubit gate set can offer reductions in circuit depth for generic circuits

Could halide perovskites be of use in electronic devices beyond solar cells and LEDs?

The number of nodes typically used in sensory networks is growing rapidly, leading to large amounts of redundant data being exchanged between sensory terminals and computing units. To efficiently process such large amounts of data, and decrease power consumption, it is necessary to develop approaches to computing that operate close to or inside sensory networks, and that can reduce the redundant data

An ultrashort pulse of electric current can be used to switch the magnetization of a micrometre-scale magnetic element via spin–orbit torques.

An epicardial bioelectronic patch is an important device for investigating and treating heart diseases. The ideal device should possess cardiac-tissue-like mechanical softness and deformability, and be able to perform spatiotemporal mapping of cardiac conduction characteristics and other physical parameters. However, existing patches constructed from rigid materials with structurally engineered mechanical

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

The development of approaches that can efficiently control the magnetization of magnetic materials is central to the creation of fast and low-power spintronic devices. Spin transfer torque can be used to electrically manipulate magnetic order in devices, but is typically limited to nanosecond timescales. Alternatively, spin–orbit torque can be employed, and switching with current pulses down to ~200 ps

The efficient generation, manipulation and detection of magnetic skyrmions are important for the development of future spintronic devices. One approach is to use electric-current-induced spin torques. Recently, thermally induced skyrmion motion has also been observed, but wider experimental evidence and its capabilities remain limited. Here we report the thermal generation, manipulation and thermoelectric

Ferroelectric field-effect transistors could play a key role in the development of data-centric computing hardware.

A cleaning–healing–cleaning method can effectively eliminate ionic defects at the surface of perovskite films, resulting in reliable and high-performance perovskite transistors.

Inexpensive radio-frequency devices that can meet the ultrahigh-frequency needs of fifth- and sixth-generation wireless telecommunication networks are required. However, combining high performance with cost-effective scalable manufacturing has proved challenging. Here, we report the fabrication of solution-processed zinc oxide Schottky diodes that can operate in microwave and millimetre-wave frequency

Light-emitting diodes based on halide perovskites have recently reached external quantum efficiencies of over 20%. However, the performance of visible perovskite light-emitting diodes has been hindered by non-radiative recombination losses and limited options for charge-transport materials that are compatible with perovskite deposition. Here, we report efficient, green electroluminescence from mixed-dimensional

Two-dimensional materials could play an important role in beyond-CMOS (complementary metal–oxide–semiconductor) electronics, and the development of memristors for information storage and neuromorphic computing using such materials is of particular interest. However, the creation of high-density electronic circuits for complex applications is limited due to low device yield and high device-to-device

Magnons, the quanta of spin waves, could be used to encode information in beyond-Moore computing applications, and magnonic device components, including logic gates, transistors and units for non-Boolean computing, have already been developed. Magnonic directional couplers, which can function as circuit building blocks, have also been explored, but have been impractical because of their millimetre

The discovery of ferroelectricity in oxides that are compatible with modern semiconductor manufacturing processes, such as hafnium oxide, has led to a re-emergence of the ferroelectric field-effect transistor in advanced microelectronics. A ferroelectric field-effect transistor combines a ferroelectric material with a semiconductor in a transistor structure. In doing so, it merges logic and memory

Organometal halide perovskite semiconductors could potentially be used to create field-effect transistors (FETs) with high carrier mobilities. However, the performance of these transistors is currently limited by the migration of ionic surface defects. Here, we show that a surface cleaning and passivation technique, which is based on a sequence of three solution-based steps, can reduce the concentration

Field-effect transistors that use carbon nanotubes as the channel material and an ion gel as the gate exhibit a high tolerance to radiation and can be recovered following radiation damage using a simple annealing process.

With the help of a gate electrode to control the charge state of individual molecules on graphene, information can be moved along a one-dimensional molecular chain, mimicking the behaviour of an electronic shift register.

Resistive switches, which are also known as memristors, are low-power, nanosecond-response devices that are used in a range of memory-centric technologies. Driven by an externally applied potential, the switching mechanism of valence change resistive memories involves the migration, accumulation and rearrangement of oxygen vacancies within a dielectric medium, leading to a change in electrical conductivity

The ability to tune the electronic properties of molecular arrays is an important step in the development of molecule-scale electronic devices. However, control over internal device charge distributions by tuning interactions between molecules has proved challenging. Here, we show that gate-tunable charge patterning can occur in one-dimensional molecular arrays on graphene field-effect transistors

Two-dimensional (2D) semiconductors are attractive for electronic devices with atomically thin channels. However, controlling the electronic properties of the 2D materials by incorporating impurity dopants is inherently difficult due to the limited physical space in the atomically thin lattices. Here we show that a solid-state ionic doping approach can be used to tailor the carrier type in 2D semiconductors

Atomically thin molybdenum disulfide (MoS2) is a promising semiconductor material for integrated flexible electronics due to its excellent mechanical, optical and electronic properties. However, the fabrication of large-scale MoS2-based flexible integrated circuits with high device density and performance remains a challenge. Here, we report the fabrication of transparent MoS2-based transistors and

Negative capacitance field-effect transistors have been proposed as a route to low-power electronics, but a lack of fundamental understanding limits progress.

The ability to estimate the uncertainty of predictions made by a neural network is essential when applying neural networks to tasks such as medical diagnosis and autonomous vehicles. The approach is of particular relevance when deploying the networks on devices with limited hardware resources, but existing competency-aware neural networks largely ignore any resource constraints. Here we examine the

Progress towards low-power electronics based on negative capacitance has been slow. For the field to develop, the gap between fundamental research on ferroelectric materials and the engineering of practical devices needs to be bridged.

A monocentric lens and a sensitive hemispherical imager can be combined to create a miniaturized camera that offers a field of view of 120°, deep depth of field and minimal optical aberration.

Accurately detecting a potential collision and triggering a timely escape response is critical in the field of robotics and autonomous vehicle safety. The lobula giant movement detector (LGMD) neuron in locusts can detect an approaching object and prevent collisions within a swarm of millions of locusts. This single neuronal cell performs nonlinear mathematical operations on visual stimuli to elicit

Electronics devices that operate in outer space and nuclear reactors require radiation-hardened transistors. However, high-energy radiation can damage the channel, gate oxide and substrate of a field-effect transistor (FET), and redesigning all vulnerable parts to make them more resistant to total ionizing dose irradiation has proved challenging. Here, we report a radiation-hardened FET that uses semiconducting

Two-dimensional materials could first find widespread commercial application in analogue electronics, rather than as a replacement for silicon in digital devices.

Spintronic devices exploit the spin, as well as the charge, of electrons and could bring new capabilities to the microelectronics industry. However, in order for spintronic devices to meet the ever-increasing demands of the industry, innovation in terms of materials, processes and circuits are required. Here, we review recent developments in spintronics that could soon have an impact on the microelectronics

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Digital electronics are ubiquitous in the modern world, but analogue electronics also play a crucial role in many devices and applications. Analogue circuits are typically manufactured using silicon as the active material. However, the desire for improved performance, new devices and flexible integration has—as for their digital counterparts—led to research into alternative materials, including the

The range of luminescent materials that can be used in electroluminescent devices is limited due to material processing challenges and band alignment issues. This impedes the development of electroluminescent devices at extreme wavelengths and hinders the use of electroluminescence spectroscopy as an analytical technique. Here, we show that a two-terminal device that uses an array of carbon nanotubes

Spin splitting in graphene is required to develop graphene-based multifunctional spintronic devices with low dissipation and long-distance spin transport. Magnetic proximity effects are a promising route to realize exchange splitting in the material, which is otherwise intrinsically non-spin-polarized. Here, we show that monolayer graphene can be magnetized by coupling to an antiferromagnetic thin

The two-dimensional semiconductor Bi2O2Se can be oxidized to create an atomically thin layer of Bi2SeO5 that can be used as the insulator in scaled field-effect transistors.

A van der Waals ferroelectric tunnel junction with asymmetric metal and graphene contacts exhibits a high resistance ratio between on and off states, and could be of value in the development of low-power computing.

Silicon-based transistors are approaching their physical limits and thus new high-mobility semiconductors are sought to replace silicon in the microelectronics industry. Both bulk materials (such as silicon-germanium and III–V semiconductors) and low-dimensional nanomaterials (such as one-dimensional carbon nanotubes and two-dimensional transition metal dichalcogenides) have been explored, but, unlike

Hydrogen-resist lithography with the tip of a scanning tunnelling microscope can be used to fabricate atomic-scale dopant devices in silicon substrates and could potentially be used to build a dopant-based quantum computer. However, all devices fabricated so far have been based on the n-type dopant precursor phosphine. Here, we show that diborane can be used as a p-type dopant precursor, allowing p-type

Integrated circuits and certain silicon-based quantum devices require the precise positioning of dopant nanostructures, and hydrogen resist lithography can be used to fabricate such structures at the atomic-scale limit. However, there is no single technique capable of measuring the three-dimensional location and electrical characteristics of these dopant nanostructures, as well as the charge dynamics

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Neuromorphic computing might be the answer to AI’s hardware problem.

Neuromorphic engineering aims to create computing hardware that mimics biological nervous systems, and it is expected to play a key role in the next era of hardware development. Carver Mead recounts how it all began.

Neuromorphic engineering attempts to create brain-like computing hardware and has helped reawaken interest in computer chip start-ups. But is the technology ready for mainstream application?

The rapid development of artificial intelligence (AI) demands the rapid development of domain-specific hardware specifically designed for AI applications. Neuro-inspired computing chips integrate a range of features inspired by neurobiological systems and could provide an energy-efficient approach to AI computing workloads. Here, we review the development of neuro-inspired computing chips, including