Research Briefing

Cellular neural networks enable real-time, parallel computation for tasks such as high-speed image processing and solving partial differential equations. We found that low-power memristors are well suited for synaptic weights in such networks, even at extreme temperatures, allowing a single system-on-a-chip to be reconfigured for different tasks.

A heat-shrinking method for fabricating conformal electronics is realized by patterning semi-liquid metal circuits onto thermoplastic substrates that are then heated to induce shrinkage around a three-dimensional target object. The method can be applied to diverse targets of different shapes and sizes, with the shape-adaptive electronics showing good electrical stability.

A direct bonding–debonding method has been developed to fabricate stacked two-dimensional semiconductors at the wafer scale with engineered layer numbers and interlayer twist angles. The as-produced structures feature pristine surfaces and interfaces, and wafer-scale uniformity — all of which are essential for application in next-generation electronic devices.

Complex architectures of conductive and dielectric domains have been created through the thermal drawing of liquid-metal-embedded elastomers. The stretchable electronic fibres from this scalable process can be easily integrated in textiles, including for sport or health-monitoring applications.

Conventional analogue compute-in-memory suffers from limited accuracy and reliability. Now, a spintronic digital compute-in-memory macro integrates in-bitcell multiplication and digitization, and flexible-precision accumulation, to deliver software-equivalent artificial intelligence accuracy. The macro achieves computation latencies of 7.4–29.6 ns and energy efficiencies of 7.02–112.3 tera-operations per second per watt.

The insertion of thin cobalt layers increases the thermal stability of β-tungsten. Using this composite as the spin current source in a 64-kb spin–orbit torque magnetic random-access memory array enables back-end-of-line process compatibility and efficient magnetization switching — achieving 1 ns switching, data retention of more than 10 years and 146% tunnelling magnetoresistance.

A skin-like organic field-effect transistor developed using semiconducting polymer nanofibres and a medical-grade elastomer demonstrates high biocompatibility and stable electrical performance under mechanical strain, providing a viable route towards robust and safe implantable bioelectronics.

An ultrathin, ultrasoft and ultrastretchable magnetoelastic on-eyelid sensor is developed to track eye movements. Decoding the eye blinking signals collected from this sensor, a one-dimensional convolutional neural network combined with an unsupervised clustering model classifies the level of cognitive fatigue with an accuracy of 96.4%.

Although neuromorphic hardware is key to mimicking brain-like information processing, existing systems have much higher energy consumption than biological systems. A molybdenum disulfide-based neuron module has been developed that emulates the intrinsic plasticity of low-power biological neural systems and exhibits visual adaptation similar to the human visual system.

Chronic pain is a widespread global health challenge. A wireless ultrasonic implant is developed that can deliver programmable spinal electrical stimulation for pain relief. This system is integrated with a machine-learning-based detection module to classify the pain severity and automatically apply the appropriate level of stimulation.

In scaling quantum computers to millions of qubits, the large mass of metallic connecting cables would create unacceptable heat loads on cryogenic cooling systems. Instead, a highly efficient wireless interconnect approach — using complementary metal–oxide–semiconductor (CMOS) terahertz transceivers with backscatter communication schemes — could provide a high-capacity, low-heat interconnect solution for future cryogenic electronic hardware.

Quantum-dot light-emitting diodes have been achieved with electroluminescent response speeds faster than expected for their organic–inorganic hybrid structure. This breakthrough is enabled by an excitation-memory effect during pulsed operations, whereby the device ‘remembers’ past input signals to emit light faster on subsequent excitation, bypassing delays caused by slow charge transport.

The particle-like behaviour of magnetic skyrmions is exploited to perform a fundamental neuromorphic computing operation: the weighted summation of synaptic signals. The use of skyrmions enables a compact and energy-efficient implementation that could pave the way for neuromorphic spintronic components that mimic the efficiency of biological systems.

A hybrid optoelectronic synthesizer is developed that combines simplified optical frequency division with direct digital synthesis to generate tunable, low-phase-noise microwaves across the X-band. This approach also achieves high frequency stability while reducing the size, weight and power demands, paving the way for chip-scale photonic microwave sources.

The multiple layers in electrochromic displays complicate the device structure and fabrication. Now, a transparent conductive polymer is shown to be both an effective conductor and ion-storage layer, enabling the fabrication of flexible, all-polymer electrochromic displays with a simplified device architecture.

A wearable single-transducer echomyography system has been developed that can be used to track and identify different breathing patterns by detecting diaphragm activity with sub-millimetre resolution. Moreover, hand gestures can be recognized from the single-channel radiofrequency ultrasound signals detected from muscles in the forearm, using a customized deep learning algorithm.

Monitoring driver alertness could improve road safety, but the use of biosensors in moving vehicles is challenging. Now, a metamaterial biosensor has been designed that can be directly embroidered onto seat belts, enabling the contactless measurement of drivers’ vital signs. This biosensor shows reliable and consistent performance, even in challenging kinetic environments.

Radiofrequency tuning elements made of quantum paraelectric materials are demonstrated at temperatures close to absolute zero — a temperature regime in which conventional electronic tuning components do not work. This advance greatly improves the read-out sensitivity of quantum circuits that require operation at such low temperatures.

Designing a rectifier for harvesting low ambient radiofrequency energy and converting it into useful d.c. power is challenging. Now, a spin-rectifier-based rectenna and a spin-rectifier array with on-chip coplanar waveguides are designed for harvesting ambient radiofrequency signals with good sensitivity and efficiency.

Graphene plasmon polaritons are expected to enable rapid data transfer and processing; however, these plasmons are difficult to access. Terahertz electronics now facilitate the efficient generation, manipulation and on-chip detection of wave packets lasting as little as 1.2 ps. This advance could lead to the development of nanoscale terahertz circuits.