Synthesis Of Low-Dimension Memristive Materials And Their Applications In Neurosynaptic Devices

With the rapid development of the information age,technology is ushering society into a new era characterized by digitalization and intelligence.Against this backdrop,there’s a growing demand for memory technology,especially in processing extensive data and achieving efficient computation.The development of novel storage technologies and information processing techniques represents a significant challenge for both the scientific community and industry.Due to its capabilities for parallel processing,adaptive learning,and low energy consumption,neuromorphic computing is seen as one of the most promising strategies to overcome the limitations of the von Neumann architecture.However,to realize neuromorphic computing,it’s imperative to develop high-performance artificial neuromorphic devices that can simulate biological synaptic behavior.Low-dimensional materials exhibit a host of novel physical and chemical properties,such as ballistic transport,quantum confinement effects,and intrinsic anisotropy,because their size in three dimensions is close to the average free path of electrons.These properties offer new avenues for creating wide-spectrum,high-sensitivity,rapid,and stable neuromorphic synaptic devices.This dissertation addresses several challenges faced by optoelectronic memristors,including low stability,poor repeatability,high energy consumption,slow operational speeds,low integration levels,and unclear working mechanisms.It designs and fabricates novel low-dimensional memristive materials（including two-dimensional materials,silicon-based,and oxide thin films）through doping engineering or defect design to modulate their physical properties.By constructing ion transport channels,it designs wide-spectrum,stable,and low-power neuromorphic synaptic devices.The operating mechanisms of low-dimensional material memristors are thoroughly investigated through characterization and physical modeling.Additionally,by fabricating wafer-level array devices,the study examines device stability and implements arrayed device design,achieving high-stability,high-repeatability array optoelectronic memristors for vision recognition,array stability assessment,and in-memory computing applications.The main research findings include:1.Through the coordinated modulation of conducting filaments,Sn O₂-based memristors achieved high resistance switching ratios（on/off ratio greater than 10⁶）,long retention times（test results exceeding 10⁵,extrapolation suggesting stability for up to 10years）,excellent endurance（greater than 5×10³ cycles）,and various synaptic plasticity functions.Using conductive atomic force microscopy（C-AFM）testing and density functional theory（DFT）calculations,the mechanisms of formation and rupture of conducting filaments by Vo,Ag⁺,and Vo/Ag⁺were verified.Device simulations,incorporating symmetric weight update and multiple conductance states,achieved a convolutional image processing accuracy of≥91.39%.Preliminary results highlight the potential of coordinated control of conducting filament technology in optimizing memristor performance,balancing memory windows,enhancing retention and durability,proving its significant prospects for advanced neuromorphic applications.2.By doping metal Ru,a-Si:Ru thin films with a low bandgap for wide-spectrum sensing were innovatively designed and a device structure based on a Si O₂ Schottky barrier was constructed to modulate photoexcited electron-hole pair transport.This bandgap engineering approach facilitated wide-spectrum recognition and memory in memristors,enabling biomimetic functions such as excitatory postsynaptic current（EPSC）,paired-pulse facilitation（PPF）,short-term plasticity（STP）,long-term plasticity（LTP）,and"learning experiences,"while reducing device switching energy consumption（116.86 f J）.Finally,using a convolutional network design,wide-spectrum image recognition and classification were successfully implemented.After training for five epochs,the system achieved a recognition accuracy of 97%,providing a new approach for artificial vision applications.3.Based on the innovative synthesis of new two-dimensional Ag Bi P₂Se₆ materials,memristors were constructed that achieved precise control of ion migration within van der Waals materials,leading to persistent and stable switching states.These memristors exhibited highly reliable switching characteristics,with minimal cycle-to-cycle variation（C_V of 0.22%for SET and 0.19%for RESET）and low device-to-device variation（C_V of4.87%for SET and 9.53%for RESET）,while also possessing multifunctional synaptic traits.Furthermore,a 16×16 integrated array demonstrated exceptional performance in simulating complex synaptic functions,achieving a recognition accuracy of 94.06%.Detailed in-situ scanning transmission electron microscopy（STEM）analysis further confirmed the close correlation between reliable resistive switching and highly ordered pathways for ion migration.This atomic-level confinement strategy offers a new pathway for designing ideal memristors,bringing fresh possibilities into the field of bio-inspired computing.4.The optoelectronic synaptic functions of the developed two-dimensional Ag Bi P₂Se₆ material-based optoelectronic memristors were further explored,successfully simulating biomimetic synaptic plasticity transformation functions and photoelectronic co-modulation biomimetic functions.Based on detailed in-situ STEM analysis,the stability and optoelectronic modulation mechanisms of optoelectronic memristors were elucidated.Additionally,wafer-level（4-inch）16×16 array devices were fabricated,and corresponding PCBs were designed and produced,integrating the optoelectronic memristor arrays.Within 22 cycles for a single device,the integrated array devices exhibited minimal pulse-to-pulse variation（P_v of 4.1032%）,and within 64 devices,P_vwas only 6.2686%.Moreover,the system implemented hardware multiply-accumulate（MAC）operations,showcasing a narrow error distribution of only 0.03%and an energetic efficiency of 260.21 TOPS/W for MAC operations.This extended research validated the potential of the designed optoelectronic memristors for more efficient storage and processing of massive data,laying a solid foundation for further improving computing system performance and energy efficiency.In summary,this dissertation delves into the preparation,performance modulation,and application research of low-dimensional memristive materials and their neuromorphic synaptic devices.Through studies on material design and device optimization,the dissertation not only provides a solid material foundation for memristor device design but also offers new insights and methodologies for achieving neuromorphic computing.