Presentation Information

[22a-A302-9]Mechanism of leaky-integration in oxide field-effect transistors

〇Hisashi Inoue1, Ai Kitoh1, Isao Inoue1 (1.AIST)

Keywords:

Leaky-integration,Neuromorphic computing,SrTiO3

Neuromorphic computing processes signals by mimicking the behavior of biological neural networks, in which typical timescales of neurons are of order 1-100 ms. For this purpose, we have developed a field-effect transistor made of semiconducting oxides SrTiO3 that can implement the leaky-integrate function of timescales ~100 ms with little power consumption. Here we investigate the mechanism of the leaky integration of this device by numerical simulation and show that the drift-diffusion of the oxygen vacancies is the source of the long timescale. The result is a step toward realizing energy-efficient neuromorphic hardware.