Gd-Doped HfO2 Memristor Device, Evaluation Robustness by Image Noise Cancellation and Edge Detection Filter for Neuromorphic Computing

Abstract

This paper estimates the robustness of non-volatile device (NVM) Gadolinium (Gd) doped Hafnium oxide (HfO2) nanoparticles (NPs) based memristor which was constructed using as-formed and annealed nanoparticles in 600 °C and 800 °C temperature, based on their performance in various basic image processing applications (i.e. edge detection filter and noise-canceling filter). A catalytic free glancing angle deposition technique (GLAD) is employed to grow Gd doped HfO2 nanoparticles (NPs) of 8 nm range on the thin film of Silicon oxide SiOx in 30 nm dimension. Annealing process is performed on Gd-doped HfO2 NPs and and the changes were demonstrated in its surface morphology. The elemental composition of the device was analyzed by Energy Dispersive X-ray (EDX). Photoluminescence (PL) analysis revealed that the topography and electrical characteristics of Gd-doped HfO2 alter swiftly after annealing process. A leakage current, interface state density (Dit) factor emphasizes that the device annealed at 600 °C portrayed significant improvement in the non-volatile characteristics in comparison with other devices. Additionally, the endurance of the device annealed at 600 °C was seen to possess more than decades of memory potential. The C-V and hysteresis curve measurement demonstrated maximal charge accumulation relative to other devices. Crossbar array is designed from both as-formed and annealed memristor devices.