Study On Electro-resistance Of Two Materials Based On BiVO₄ And BiOI

The resistance random access memory?RRAM?based on oxides show exhilarating advantages for next generation of nonvolatile memories,such as simple structure,low power consumption,fast access speed,three dimensional integration,compatible with microelectronics technology,however,the dispersion of resistive switching parameters limits its applications in memory fields.The memristor is a two-terminal device which“remembers”the flow of charges through it with time dependent memristance feature.It is the most proper electronic component to simulate the human brain processes such as learning,memory and realize the“brain”artificial cognition.Synaptic bionic research is in the booming development,but the lack of high quality memristor material systems and large-scale manufacturing technology compatible with the microelectronics technology,becomes a bottleneck,restricting its rapid progress.Neuromorphic computing is a new computational model that simulates neurobiological processes by constructing a computational architecture that is similar to that of an animal's brain.It can improve the computer system's perception and autonomous learning ability,can cope with the current serious energy consumption problems,and is expected to subvert the existing digital technology.Neuromorphic computing structure emulates a human brain that concurrently performs perception,learning,and memory.These functions are performed by enormous numbers of neurons and synapses.In particular,the synapses perform learning and memory functions by modulating the strength of connection between neurons;this process is called synaptic plasticity.Therefore,emulation of a synapse is an important step to achieve an efficient artificial neuromorphic system.Recently,devices with single artificial synapse based on complementary metal oxide semiconductor?CMOS?analogue circuits with several transistors and capacitors were fabricated,but complex integrated circuits were required with high power consumption.Therefore,the need to study new materials,structures and devices.Using memristors to model key features of biological synapses is considered the most promising solution.Memristors meet the requirement to simulate synaptic behavior using a single device.Most of the artificial synapses using RRAM only work in the gradual‘reset'?resistance changes from low to high?region resulting in depression-only ability.Multiple potentiation states are greatly required for neuromorphic devices to mimic biological synapses with both“learning”and“forgetting”functions.The depression-only issue is still an obstacle for synaptic applications.Based on the above,this paper studies the memristors based on two novel oxide films of iodine bismuth oxide and bismuth vanadate,and tests the performance of these two novel material memristors in simulating biological synapse behavior.The memristor prepared in this paper overcomes the shortcomings of unidirectional regulation in the reset process and achieves bidirectional regulation of device conductance during set and reset process.The bismuth vanadate has the advantages of low price and simple synthesis method.The iodine bismuth oxide has good adsorption properties,and both materials have good photocatalytic properties and are environmentally friendly materials.The main research contents of this article are as follows:The Ti top electrode was grown on the BiVO₄ sample by magnetron sputtering,and a memristor with Ti/BiVO₄/FTO structure was fabricated.The memristor has been studied to simulate several important synaptic behaviors including nonlinear transmission characteristics,peak time dependent plasticity?STDP?,pair pulse facilitation?PPF?,short term plasticity?STP?to long term plasticity?LTP?transition under different stimuli,and short term plasticity at different temperatures.Electrical properties including the characteristics of the retention property,endurance property,the cumulative probability,the switching time,the results show that the prepared memristor Ti/BiVO₄/FTO structure can well simulate the synapse behavior,can promote the realization of large-scale neuromorphic systems.The Ti/Bi VO₄/FTO structure,TiN/BiVO₄/FTO structure,and Pd/BiVO₄/FTO structure resistive memory were fabricated by using magnetron sputtering technique to grow Ti top electrode,TiN top electrode and Pd top electrode on BiVO₄ samples.The conductance quantization behaviors of three different top-electrode structure resistive memory devices were investigated.It was found that the devices exhibited conductance quantized behavior when these three materials were used as top electrodes.Through statistics the quantum conductance appearing in the I/V curve,it is found that the quantum conductance values are mostly integer and half-integer multiples of G₀,and the state of the device is most stable when Ti is used as the top electrode.The resistive memory of the Ti/BiVO₄/FTO structure was tested for retention,fatigue resistance,multi-value memory characteristics,and the device's conductivity mechanism was analyzed.The effective switching time of the device is more than 100 times and after 1.4×10⁴ s,there is no obvious change in the high and low resistance state,and the stepped conductance in the RESET process is maintained and the resistance is maintained for more than 10⁴ s,indicating that the device has good multi-valued storage capabilities.Variable temperature test resistive memory,fitting the device's I/V curve and combining existing research results,considers the device to be a jumping conductance mechanism.Based on BiOI materials,we designed and fabricated Ti/BiOI/FTO structure memristor devices and obtained stable operation devices.The memristor has good resistance adjustability in SET and RESET process,retention property,endurance property.We explored the similarities between memristors and synapses to obtain intelligent devices with autonomous learning capabilities.A variety of important synaptic learning and memory functions have been realized in memristive devices,including peak time dependent plasticity,pair pulse facilitation,and empirical learning behavior.The realization of important synaptic functions and the establishment of a dynamic model would promote more accurate modeling of the synapse for artificial neural network.