Foundry Division, Samsung Electronics Co. Ltd., Youngin 17113, Korea.
Nanoscale Horiz. 2021 Feb 1;6(2):139-147. doi: 10.1039/d0nh00559b. Epub 2020 Dec 24.
Recently, various efforts have been made to implement synaptic characteristics with a ferroelectric field-effect transistor (FeFET), but in-depth physical analyses have not been reported thus far. Here, we investigated the effects by (i) the formation temperature of the ferroelectric material, poly(vinylidene fluoride-trifluoroethylene) P(VDF-TrFE) and (ii) the nature of the contact metals (Ti, Cr, Pd) of the FeFET on the operating performance of a FeFET-based artificial synapse in terms of various synaptic performance indices. Excellent ferroelectric properties were induced by maximizing the size and coverage ratio of the β-phase domains by annealing the P(VDF-TrFE) film at 140 °C. A metal that forms a relatively high barrier improved the dynamic range and nonlinearity by suppressing the contribution of the tunneling current to the post-synaptic current. Subsequently, we studied the influence of the synaptic characteristics on the training and recognition tasks by using two MNIST datasets (fashion and handwritten digits) and the multi-layer perceptron concept of neural networks.
最近，人们做出了各种努力，试图利用铁电场效应晶体管（FeFET）实现突触特性，但迄今为止尚未进行深入的物理分析。在这里，我们研究了铁电材料聚（偏二氟乙烯-三氟乙烯）（P（VDF-TrFE）的形成温度以及 FeFET 的接触金属（Ti、Cr、Pd）的性质对基于 FeFET 的人工突触的工作性能的影响，包括各种突触性能指标。通过在 140°C 下退火 P(VDF-TrFE) 薄膜，使β相畴的尺寸和覆盖率最大化，从而产生了优异的铁电性能。形成较高势垒的金属通过抑制隧道电流对后突触电流的贡献，提高了动态范围和非线性度。随后，我们使用两个 MNIST 数据集（时尚和手写数字）和神经网络的多层感知器概念研究了突触特性对训练和识别任务的影响。
