Thermal Annealing-Driven Modulation of Charge Trapping and Synaptic Plasticity in a Sol-Gel AlOx-Based Floating Gate Transistor

Abstract

The development of high-performance synaptic devices and organic floating gate memory is the most important innovation for electronics technology. Such devices possess huge potential in revolutionizing the storage and processing performance of data in applications that exhibit low power consumption, high areal density, and flexible substrate compatibility. However, achieving such a performance requires an effective charge trapping medium that can efficiently capture and retain charge carriers, essential for storing and processing information. There have been several candidates proposed for a potential charge trapping layer. In our analysis, we made use of a simple and cost-effective, solution-processed sol？gel AlOx as a charge trapping layer. We studied the properties of sol？gel AlOx thin films before and after annealing at different temperatures (pristine, 100, 200, 300, 400, and 500 °C) using X-ray diffraction, atomic force microscopy, and X-ray photoelectron spectroscopy. As the annealing temperature rises, it becomes apparent that the AlOx thin film produced via sol？gel undergoes the decomposition of organic residues and nitrate groups along with the transformation of aluminum hydroxide into aluminum oxide. At low temperatures, the organic floating gate device exhibited a wider hysteresis window (ΔVth), which becomes negligible at high temperatures. This implies that the hysteresis window is affected by the presence of hydroxyl groups. Also, the investigation was done to enhance the device ability to simulate synaptic behavior by using a solution-processed sol？gel AlOx-based floating gate transistor. The channel conductance of a floating gate transistor is stored in synaptic weight, which is modulated by the applied positive and negative electrical pulse stimuli and annealing temperature of the sol？gel AlOx thin-film layer. The key properties of long-term potentiation and long-term depression characteristics such as dynamic range (DR) and nonlinearity (NL), which have a significant impact on the memory, adaptive learning, and decision-making ability of synaptic devices, were studied. The device subjected to annealing temperatures exceeding 200 °C exhibited favorable NL and DR at VG = ±20 V, compared to those annealed at other temperatures, in response to variations in the pulse width. Correspondingly, the devices that were annealed at 200 °C achieved the highest accuracy of ∼93.60% in the MNIST (Modified National Institute of Standards and Technology) deep neural network simulation at a pulse width of 200 ms, surpassing all other annealing conditions. These results underscore the role of the annealing temperature in optimizing device performance, particularly in fundamental aspects of synaptic behavior NL and the DR. This advancement paves the way for more efficient, flexible, and dense electronic devices.