Scholarlink

<p>Big data intensifies challenges in power efficiency and communication bandwidth. Neuromorphic computing based on emerging devices with intrinsic information processing capabilities enables energy-efficient, high-speed artificial neural networks, playing a crucial role in addressing limitations of traditional von Neumann architectures and attracting widespread attention. As a Mott material, VO<sub>2</sub> is pivotal for next-gen neuromorphic devices, with its near-room-temperature metal-insulator transition and resistance-switching properties. Here, we demonstrate that a fully light-modulated artificial synapse fabricated from single-crystalline VO<sub>2</sub> film could exhibit precisely tunable temporal dynamics and support biologically relevant synaptic plasticity, mimicking key adaptive behaviors of biological synapses. Based on this synapse, a three-layer artificial neural network architecture was implemented, where VO<sub>2</sub> photo-synapses functioned as image pre-processors to compress redundant visual information. Specifically, a 4×4 pixel image was condensed into a 4×1 vector, reducing computational load while maintaining high performance. This study highlights VO<sub>2</sub> photo-synapses’ potential for advanced neuromorphic computing and intelligent vision systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Photo-synapses based on single-crystalline VO2 films for in-sensor information processing

  • 期刊论文

  • 发表日期：

    2026年5月9日

查看全文
  • Xinjie Hou,
  • Xintian Wang,
  • Lei Yin,
  • Xunguo Gong,
  • Ximeng Peng,
  • Yao Wen,
  • Yuchen Cai,
  • Ruiqing Cheng,
  • Jun He

摘要

Big data intensifies challenges in power efficiency and communication bandwidth. Neuromorphic computing based on emerging devices with intrinsic information processing capabilities enables energy-efficient, high-speed artificial neural networks, playing a crucial role in addressing limitations of traditional von Neumann architectures and attracting widespread attention. As a Mott material, VO2 is pivotal for next-gen neuromorphic devices, with its near-room-temperature metal-insulator transition and resistance-switching properties. Here, we demonstrate that a fully light-modulated artificial synapse fabricated from single-crystalline VO2 film could exhibit precisely tunable temporal dynamics and support biologically relevant synaptic plasticity, mimicking key adaptive behaviors of biological synapses. Based on this synapse, a three-layer artificial neural network architecture was implemented, where VO2 photo-synapses functioned as image pre-processors to compress redundant visual information. Specifically, a 4×4 pixel image was condensed into a 4×1 vector, reducing computational load while maintaining high performance. This study highlights VO2 photo-synapses’ potential for advanced neuromorphic computing and intelligent vision systems.