Evaluation of ferroelectric devices for smart computers
Transistors or “microchips” partly explain why our paper-thin laptops can perform far more complex tasks than their giant, clumsy predecessors of them. To maximize computing power, engineers are trying to make transistors as small as possible and pack billions of transistors into a single computer chip.
However, despite the rapid development of manufacturing techniques, traditional transistors are approaching their physical limits—the nanoscale device cannot shrink any further after a certain point—and that hinders the growth of computing power.
However, as data continues to pour in, the need for computing power continues to grow. New devices, especially new storage and logic device with higher speed and Lower power consumptionis needed to unleash new computing capabilities while breaking down major obstacles to existing computing systems.
Recently, a group of researchers from China pointed to ferroelectric devices as a promising solution and published a paper introducing the emerging devices. ferroelectric material and devices for smart computing. Review published in Intelligent computing.
Ferroelectric materials are quite versatile and are widely used as special-purpose memory in aerospace storage devices. They have special polarizing properties, a magnetic-like property that can be retained even after the external electric field is removed. But when the film thickness falls below 10 nm, most conventional ferroelectric materials lose their polarizing properties at 25°C, and thus are not suitable for large integrated circuit (IC) fabrication processes. nano scale.
New ferroelectric materials with high expansion potential can solve these problems. “The discovery of the polarization effect in a high material, which is a commonly used gate oxide material for nanoscale MOSFETs [metal-oxide-semiconductor field-effect transistors]is a breakthrough for mass production of ferroelectric transistors,” the researchers pointed out.
They reviewed two prominent examples of amorphous oxide-based and polycrystalline Hf-based ferroelectric materials, and briefly described several recently reported new materials and devices. All of them are compatible with the complementary metal oxide semiconductor (CMOS) fabrication process.
For the most advanced ferroelectric devices, the researchers classified them into low-power logic devices, high-performance memory cells, and neural simulators, and detailed each category. . The abstracts address the development of the devices and their ability to break down the “thermal wall”, the “memory wall”, and the von Neumann bottleneck respectively.
Negative capacitor field-effect transistors (NCFETs) are low-power logic devices that can break down a “thermal wall”, hindering the processor’s main frequency boost due to increased power density and effects. heat. “Reducing a chip’s control voltage is a potential method for breaking the ‘thermal wall’, and its feasibility is highly dependent on SS [subthreshold swing] of the transistor,” the researchers explain.
“The ferroelectric NCFEs, together with the voltage amplification effect, can overcome Boltzmann tyranny and achieve SS below 60 mV/dec. They are therefore considered to have one of the design architectures. most promising for ultra-low power applications and could re-ignite the rapid growth of the microchip industry.”
Ferrous capacitor-based random access memory (FeRAM) and ferroelectric field-effect transistor (FeFET-)-based memory, classified as high-performance memory cells, exhibit excellent performance. great in motion. temporary access memory (DRAM) replacement and embedded applications.
Ferroelectric capacitors, unlike conventional DRAM capacitors, can store information via charge Pr, which is non-volatile and possesses a much higher charge-per-area density.
Therefore, replacing the dielectric material of the flash device with a doped HfO2 ferroelectric or amorphous oxide ferroelectric to generate FeFETs is an alternative method for further reduction, the researchers say. energy or latency of these memories”. This will help bridge the large gap in performance or area between the logic device and the memory cell, overcoming the so-called “memory wall”.
Furthermore, the FeFET can be used as a neuroimaging device to break the von Neumann bottleneck. The von Neumann bottleneck refers to delays and power problems caused by inefficient data transfers between the initially decoupled memory module and the logic processor, where computing emulates neural networks—mimicking neural systems to process information—is a possible solution.
In a neural simulation system, artificial neurons and synapses are the most important components, and FeFET is reported to be able to do both. For applications in neurons, FeFET was used as the impulse neural network; for artificial synaptic applications involving spike neural networks (SNNs) and convolutional networks neural network (CNNs), FeFETs are adopted because of their ability to perform storage and processing functions simultaneously.
In addition, ferroelectric tunnel junctions (FTJs) have attracted considerable attention to synaptic device applications due to their compact device structure, non-destructive read scheme, and write/access speed high reading.
In conclusion, the researchers point out that if a balance can be struck between process compatibility and device performance, then NCFET, FeRAM, or FeFET memory and iron synaptic devices power can be integrated into the same chip to build a multifunctional intelligent computer system.
“Based on the advances made in ferroelectric device processing technology, integrating low-power logic, high-performance memory, and neural simulation systems on a single chip seems feasible,” they stressed. with continuous improvement”. “This will help realize the development of high-performance and high-performance intelligent computing systems in the future.”
Genquan Han et al, Ferroelectric devices for intelligent computers, Intelligent computing (2022). DOI: 10.34133/2022/9859508
Powered by Intelligent Computing
quote: Review of ferroelectric devices for smart computers (2022, December 5) accessed December 5, 2022 from https://techxplore.com/news/2022-12-ferroelectric-devices-intelligent .html
This document is the subject for the collection of authors. Other than any fair dealing for private learning or research purposes, no part may be reproduced without written permission. The content provided is for informational purposes only.