Research On The Visual/Tactile In-Sensor Computing Of Lead-Free Inorganic Perovskite Memristors

Currently,memristors with both storage and computing functions are the core devices of non-von Neumann architecture in the future,and the in-sensor computing based on memristors has become the focus of attention in academic and industrial fields.In addition,the development of intelligence and robotics has put forward higher requirements for visual/tactile perception,and the superiority of memristors makes it possible to build intelligent sensing systems with complex environment perception,intelligent human-computer interaction and precise control.In recent years,perovskite materials have achieved a breakthrough for the memristors in the applications of new material,device structure design and resistive switching mechanism due to their excellent optoelectronic properties.However,perovskite memristors still suffer from complex resistive switching mechanism and low integration of in-sensor computing.This work focuses on the fabrication of lead-free inorganic perovskite memristors,resistive switching mechanism and realization of visual/tactile in-sensor computing:（1）The optical Reset phenomenon is discovered in memristors based on lead-free inorganic Cs₂AgBiBr₆ perovksite.A model involving the formation of Br Frenkel defects is established and the mechanism that UV light illumination will induce the annihilation of the Br defects and cause the fracture of conductive filaments is revealed.The memristor exhibits an average SET voltage of 1 V,a high ON/OFF ratio of 10³ and a long data retention of 10³s.The one-diode/one-resistor structure is utilized to suppress the crosstalk current in the cross-bar structure,and the reading resistance in high resistance state is increased from 4.5kΩto 4.1 MΩ.（2）A visual in-sensor computing system is demonstrated based on Cs₂AgBiBr₆memristors,realizing reinforcement learning and the integrated functions of image sensing and memory for a long-term retention time（>6000 s）.Besides,an all-optically controlled visual in-sensor computing system is demonstrated by integrating the Cs₂AgBiBr₆memristors with the perovskite solar cells.（3）A memristor based on lead-free inorganic AgBiI₄ perosvkite is demonstrated for the first time.The mechanism of resistive switching is proposed involving conducting filaments formed by Ag cations.The memristors exhibit an ultralow operating voltage（0.16V）,high ON/OFF ratio（10⁴）,long data retention（10⁴ s）and reversible resistive switching by pulse voltage operation（700）.The prepared flexible memory displays high ON/OFF ratio（10²）and good stability under repeated bending tests（1000 cycles）.The above work paves the way for realizing low-power consumption flexible nonvolatile memristors.（4）A tactile in-sensor computing system based on AgBiI₄ perovskite is demonstrated,which realizes the capability of contrast enhancement and noise reduction under pressure stimulations with various intensities.The synaptic device realizes several functions of biological synapses including long-term potentiation,long-term depression,paired-pulse facilitation and spike-timing-dependent plasticity.After information preprocessing with tactile in-sensor computing system,the recognition accuracy reaches up to almost 100%with only 500 training epochs by using the MNIST handwritten data set.In this work,the in-sensor computing devices are designed and fabricated based on lead-free inorganic perovskite,which are applied to neuromorphic reinforcement learning,synaptic function simulation and in-sensor computing integration successfully,providing a potential solution to solving the problems of sensor data increase,slow processing speed and high energy consumption in the intelligent era.