Volume 6 Issue 11, November 2023

A two-dimensional metallic alloy that has a composition-tunable work function can be used to engineer metal–semiconductor contacts.

An effective gate voltage doping method can be used to create single-gate molybdenum ditelluride field-effect transistors that can be reconfigured between rectification, memory, logic and neuromorphic functions.

Dual-gate heterojunction transistors that are based on monolayer molybdenum disulfide and carbon nanotubes can provide tunable Gaussian and sigmoid functions for support vector machine computing.

Event-driven, in-sensor computing can be performed by individual vision sensors composed of two parallelly connected photodiodes, enabling vision recognition of dynamic motion.

Ring oscillator circuits that operate at gigahertz frequencies and are based on monolayer molybdenum disulfide can be created with the help of a design–technology co-optimization approach.

A silicon photonics modulator design approach is proposed, in which the inductive networks and termination resistors are designed in conjunction with the optical phase shifter. A complementary metal–oxide–semiconductor (CMOS) silicon photonics transmitter developed with this approach achieved 112 gigabaud transmission with an energy efficiency better than 1 pJ per bit.

The Perspective explores the future design of lifelong learning artificial intelligence (AI) accelerators that are intended for deployment in untethered environments, identifying key desirable capabilities for such edge AI accelerators and guidance on metrics to evaluate them.

This Review provides a full-spectrum classification of computing-in-memory technologies by identifying the degree of memory cells participating in the computation as inputs and/or output, creating a platform for comparing the advantages and disadvantages of each of the different technologies.

A ternary metallic alloy VS_2xSe_2(1–x) that has a tunable work function can be grown using chemical vapour deposition and used as contacts for two-dimensional semiconductors.

An effective-gate-voltage-programmed graded-doping method can be used to reconfigure a single-gate molybdenum ditelluride device to different states, including a polarity-switchable diode, memory, Boolean logic and artificial synapse.

Dual-gated van der Waals heterojunction transistors can provide Gaussian, sigmoid and mixed-kernel functions for use in low-power machine learning classification operations.

A spiking neural network that is based on event-driven vision sensors can be created using two parallel photodiodes of opposite polarities that output programmable spike signal trains in response to changes in light intensity.

Five-stage ring oscillators that operate at frequencies of up to 2.65 GHz can be created using monolayer molybdenum disulfide field-effect transistors developed with a design-technology co-optimization process.

A model that predicts the force behaviour for solid/liquid-dielectric multilayer stacks independent of actuator design, and solely based on the material properties, can be used to develop actuators that provide a steady force output under constant-voltage operation.

A low-power radio-frequency multiplexing cryo-electronics system, which is based on complementary metal–oxide–semiconductor technology, can operate below 15 mK and provide the control and interfacing of superconducting qubits with minimal cross-coupling.

Switching-current-based low-power transmitters with a high throughput can be created using an approach in which silicon-photonics-based Mach–Zehnder modulators and complementary metal–oxide–semiconductor electrical drivers are co-designed.