Effect of the Gate Dielectric Layer of Flexible InGaZnO Synaptic Thin-Film Transistors on Learning Behavior

In this work, flexible InGaZnO (IGZO) synaptic thin-film transistors (TFTs) with different gate dielectric layers are fabricated and analyzed to investigate the effect of the gate insulator of flexible IGZO synaptic TFTs in terms of weight window and retention characteristics. The gradual weight modulation of these devices comes from the migration of hydrogens in the Al2O3 layer deposited by low-temperature atomic layer deposition and can be controlled by gate bias. In addition, the learning behaviors with identical and incremental pulse schemes are verified for a linear weight modulation, and its effect in pattern recognition accuracy is studied considering device variation and retention properties in a 784 x 10 fully connected neural network with handwritten digit images.

Automated summary

Flexible IGZO synaptic TFTs with different gate dielectric layers were fabricated and analyzed to investigate the effect of the gate insulator on weight window and retention characteristics. The weight modulation can be controlled by gate bias, and its impact on pattern recognition accuracy in a neural network was studied.