ABSTRACT
In this paper, the instability mechanism of resistive random access memory (RRAM) was investigated, and a technique was developed to stabilize the distribution of high resistance states (HRS) and better concentrate the set voltage. Due to the accumulation of oxygen, an interface-type switching characteristic was observed on the I-V curves beneath the filament-type switching behavior. In this work, the interface-type switching characteristic is used to fit the natural distribution of HRS as an analysis of the instability mechanism. According to the results, the HRS distribution is attributed to the accumulation of excess oxygen ions left from the lower oxygen content and oxygen vacancy recombination during the reset process. The proposed solution with simple plasma treatment, can create an excess oxygen reservoir by changing the surface topography of the electrode to store the surplus oxygen ions from the reset process, eliminating the oxygen accumulation effect and further improving the device stability.
ABSTRACT
Although good performance has been reported in shallow neural networks, the application of memristor synapses towards realistic deep neural networks has met more stringent requirements on the synapse properties, particularly the high precision and linearity of the synaptic analog weight tuning. In this study, a LiAlOX memristor synapse was fabricated and optimized to address these demands. By delicately tuning the initial conductance states, 120-level continuously adjustable conductance states were obtained and the nonlinearity factor was substantially reduced from 8.96 to 0.83. The significant enhancements were attributed to the reduced Schottky barrier height (SBH) between the filament tip and the electrode, which was estimated from the measured I-V curves. Furthermore, a deep neural network for realistic action recognition task was constructed, and the recognition accuracy was found to be increased from 15.1% to 91.4% on the Weizmann video dataset by adopting the above-described device optimization method.
ABSTRACT
In this study, the impact of moisture on the electrical characteristics of an amorphous In-Ga-Zn-O thin-film transistor (a-IGZO TFT) was investigated. In commercial applications of such TFTs, high stability and quality performance in humid environments are essential. During TFT operation under ambient moisture, the electrolysis of water molecules occurs via the tip electric field effect. Hydrogen diffuses from the etch-stop layer or back-channel into the main channel under a negative electric field. The hydrogen atoms act as shallow donors (which causes the carrier concentration in the channel to rise), causing the threshold voltage (VTH) to shift in the negative direction. Hydrogen diffusion from the overlap of the source/drain and gate electrodes to the channel center caused by the tip electric field induces a significant barrier lowering and VTH shifts in a short-channel device. However, under negative bias stress (NBS) in ambient moisture, the negative VTH shift is more obvious in short- than in long-channel devices, indicating suppressed hydrogen diffusion in long-channel devices. This is attributed to the electrolysis of water by the tip electric field at the source, drain, and gate electrodes, which causes hydrogen to diffuse to the center of the channel. Here, a novel physical model of the capacitance-voltage (C-V) electrical property changes under ambient moisture is proposed, based on the early appearance of abnormalities in the C-V measurements. The electrolysis of water caused by the tip electric field and electrical abnormalities caused by hydrogen diffusion into the a-IGZO active layer are explained by this model. A secondary-ion mass spectrometry analysis shows that hydrogen content in the channel generally increases under NBS in ambient moisture. The degradation behavior due to moisture in a-IGZO is clarified. Thus, inhibiting the tip electric field may benefit future flexible-display and gas-sensing applications.
