In the 4th Industrial Revolution, artificial intelligence is evaluated as a major technology, receiving public attention, and active research is underway in the education industry. In particular, artificial intelligence was implemented through an algorithm called deep learning and deep neural network hardware. However, as the amount of computation increases and becomes more complex, the Spiking Neural Network, an artificial neural network capable of biological neuron computation, emerges as the problem of the deep neural network, Bon Neumann bottleneck and power consumption, are limited. Therefore, various devices to be used in spiking neural networks called third-generation artificial neural networks are being studied. However, in the case of artificial neurons, most of them use complementary metal oxide semiconductors (CMOS), so there is a disadvantage in terms of device cost.
This dissertation focuses on the validity of ferroelectric field effect transistors to be used in neurons in spiking neural networks, and achieves and optimizes leakage characteristics. The first purpose of this work is to identify the depolarization field and leakage current trapping, which are the causes of the leaky characteristics of ferroelectric field effect transistors(FeFET) and to confirm their influence on the leaky effect. The second purpose is to fabricate neuron devices using leaky ferroelectric field effect transistor(Leaky-FeFET) having similar characteristics to biological neurons and to propose neuron device that includes refractory period features.