Ferroelectric Gate-All-Around Transistors for 3D-Integrated Electronics and Neuromorphic Vision.
basic_science · Level V
Where this comes from
- Record sourced from PubMed, PMID 42265948.
- Also identified by DOI 10.1021/acsnano.6c02136.
- No licence information is recorded for this record.
- Because redistribution is not established, this page shows the abstract only. Follow the links below for the full text.
Abstract
Edge computing demands electronics that simultaneously deliver high performance, ultralow power consumption, and dense integration, yet still remains challenging in growing development of artificial intelligence and Internet of Things. Here, we propose an all-two-dimensional low-power and compact transistor with a ferroelectric gate-all-around (Fe-GAA) architecture for energy-efficient 3D-integrated electronics and neuromorphic edge computing. The Fe-GAA field-effect transistors break the Boltzmann limit of traditional CMOS devices, achieving sub-60 mV dec<sup>-1</sup> switching (down to 25.3 mV dec<sup>-1</sup>), a high on/off ratio (10<sup>8</sup>), and field-effect mobility (310 cm<sup>2</sup> V<sup>-1</sup> s<sup>-1</sup>). Capitalizing on ferroelectric polarization dynamics in the surrounding CuInP<sub>2</sub>S<sub>6</sub> gate, we further demonstrate its application prospects as compact edge computing devices for binary logic operations and spiking neurons on a unified hardware platform. Monolithic 3D-integrated logic circuits (inverter and NOR gate) reduce footprint by 50% versus planar CMOS, and the artificial neuron emulates leaky integrate-and-fire (LIF) behavior without additional capacitors and external reset circuitry in a conventional neuron, significantly reducing the energy consumption and hardware footprint. Implemented in a spiking neural network (SNN) for gesture recognition, this system attains 92.71% accuracy on the DVS128Gesture data set. This work establishes a paradigm for multifunctional edge intelligence processors that transcend traditional power-area trade-offs.