ACS Applied Materials & Interfaces · 2024 · 10 citations · 52 references
EngineeringMaterials EngineeringMaterials ScienceElectronic MaterialsSemiconductor DevicesSemiconductor MaterialsMemristive MaterialsMetal-semiconductor ContactsOxide SemiconductorsSemiconductor Device MaterialsSemiconductor InterfacesSemiconductor PhysicsAnalog Switching BehaviorEpitaxial InterfacesMetal-semiconductor InterfacesOxygen VacanciesOxygen Vacancies In Metal OxidesReciprocal Space MappingSchottky Barrier CharacteristicsSemiconductor Heterojunctions
Neuromorphic computing, inspired by the brain’s architecture, promises to surpass the limitations of von Neumann computing. In this paradigm, synaptic devices play a crucial role, with resistive switching memory (memristors) emerging as promising candidates due to their low power consumption and scalability advantages. This study focuses on the development of metal/oxide-semiconductor heterojunctions, which offer several technological advantages and have broad potential for applications in artificial neural synapses. However, constructing high-quality epitaxial interfaces between metal and oxide semiconductors and designing modifiable contact barriers are challenging. Herein, we construct high-quality epitaxial metal/semiconductor interfaces based on the metallicity of the perovskite phase SrFeO 3−δ (PV-SFO) and a small Schottky barrier in contact with Nb-doped SrTiO 3 (NSTO). X-ray diffraction patterns, reciprocal space mapping results, and cross-sectional transmission electron microscopy images reveal that the prepared PV-SFO film exhibits a perfect single-crystal structure and an excellent epitaxial interface with the NSTO (111) substrate. The corresponding memristor exhibits analog-type resistive-variable characteristics with an ON/OFF ratio of ∼1000, stable data retention after 10,000 s, and no noticeable fluctuation in resistance after 10,000 pulse cycles. Electron energy loss spectroscopy, first-principles calculations, and electrical measurements reveal that compensating or restoring oxygen vacancies at the NSTO surface decreases or increases the contact barrier between PV-SFO and NSTO, respectively, thereby gradually regulating the resistance value. Furthermore, high-quality epitaxial PV-SFO/NSTO devices achieve up to 98.21% recognition accuracy for handwriting recognition tasks using LeNet-5-based network structures and 92.21% accuracy for color images using visual geometry group (VGG) network structures. This work contributes to the advancement of interface-type memristors and provides valuable insights into enhancing synaptic functionality in neuromorphic computing systems.
52
Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing
Zhongrui Wang, Saumil Joshi, Sergey Savel’ev et al. · Nature Materials · 2016 · 2.3K citations · Full text
Fully hardware-implemented memristor convolutional neural network
Peng Yao, Huaqiang Wu, Bin Gao et al. · Nature · 2020 · 2.1K citations
Engineering, Computer Engineering, Electrical Engineering +12
Short-term plasticity and long-term potentiation mimicked in single inorganic synapses
Takeo Ohno, Tsuyoshi Hasegawa, Tohru Tsuruoka et al. · Nature Materials · 2011 · 1.8K citations