| The visual system is essential for biological survival and competition.During visual information handling,the retina can detect light stimuli and preprocess image information at the same time before the brain conducts more complex actions.In recent years,digital vision systems,based on conventional complementary metal-oxide-semiconductor(CMOS)imagers or charge-coupled device(CCD)cameras,have been rapidly developed to achieve the computer vision through extended interfaced digital processing units on serial or coarsely parallel structures.However,these conventional digital artificial vision systems tend to consume a lot of power,and have a large size and high cost for practical applications.Hence,the main challenge for developing an artificial vision system is to recreate the flexibility,sophistication,and adaptability of biological systems with computational efficiency and elegance.Similar to biological systems,these neuromorphic vision systems that integrate image sensing,memory and processing into a same space in the device,and process different types of spatio-temporal computations of a continuous analog brightness signal in real-time,are expected to overcome these disadvantages.For the development of a high-performance neuromorphic vision system,high integrated flexible optoelectronic sensors with ultra-high responsivity,detectivity,and signal-to-noise ratio are necessary to offer enhanced imaging capability under extreme dim light conditions.For the selection of an active sensing material,the all-inorganic perovskite CsPbBr3-QDs have excellent optoelectronic response performance,and CNTs can significantly improve the detection signal-to-noise ratio of the sensor due to the excellent carrier mobility and on/off ratio.Both materials can be fabricated into uniform large-area films with excellent flexibility and stability,and the combination of these two materials provides a new strategy for the design and fabrication of high-performance,high-resolution neuromorphic vision sensors.In this dissertation,we have fabricated a flexible neuromorphic optoelectronic sensor array with 1024 pixels using a combination of semiconducting carbon nanotubes(CNTs)and perovskite quantum dots(CsPbBr3-QDs)as the active materials.In addition,the effects on neuromorphic reinforcement learning are systematically investigated by the ultra-weak optical image pulse(1 ?W/cm2),and an ultrasensitive artificial vision system which possess image sensing,information memory and data preprocessing is realized.The results are summarized as followed:(1)By optimizing the deposition parameters,a fast and uniform method of depositing the separated and purified semiconducting CNT film is obtained.And the CNT thin film transistor with excellent electrical performance is fabricated,which lays a solid foundation for designing and fabricating high integrated and uniform flexible sensor array.(2)The all-inorganic CsPbBr3-QDs is used as the photo-absorption and photo-generated charge trapping layer,and CNT is used as electrical transport layer.The neuromorphic optoelectronic sensor based on this new material system shows an extraordinary sensitivity to light with an ultra-high Gesponsivity of 5.1×107 A/W,an ultra-high external quantum efficiency of 1.6×10 10%,an ultra-high specific detectivity of 2×1016 Jones and an ultra-high signal-to-noise ratio of 106.At the same time,the device also exhibits excellent memory characteristics with the retention time beyond 10,000 s,and optically tunable synaptic plasticity which can mimic synaptic behaviors.(3)The flexible neuromorphic optoelectronic sensor array with 1024 pixels has been fabricated successfully with high uniformity and high yield of 100%.After training the sensor array with optical image pulse signals,the visual image reinforcement learning is successfully realized,which has been also experimentally demonstrated by training a highly integrated physical device with an ultra-weak light pulse of 1 ?W/cm2 at the first time.The neuromorphic optoelectronic sensor simultaneously acts as an artificial photoreceptor and a biological synapse,and,thus,directly responds to optical stimuli and performs light-tunable synaptic plasticity for functional preprocessing.These results provide motivation for the development of artificial neuromorphic visual systems to simulate the flexibility,complexity,and adaptability of biological vision systems. |
PDF Full Text Request