Scholarlink

Artificial intelligence has advanced with the emergence of deep learning techniques. However, the performances of deep neural networks in Von Neumann architectures are limited by the memory bottleneck. In-memory computing using memristive devices, is a promising solution, with significant improvement in speed, efficiency, and scalability. This work demonstrates the application of RF-sputtered Hafnium Oxide (HfOx) based memristive device for implementing logic gates targeted at In-memory computing applications. A model for Pt/HfOx/Ag based memristor was developed, adopting the Voltage ThrEshold Adaptive Memristor (VTEAM) model, showing close match with experimental I-V data (rms error:0.0078). The Verilog-A model for this memristive device was developed and implemented in Cadence Virtuoso circuit simulator and realized basic logic gates using the Memristor Aided Logic design-MAGIC and analyzed its performances in terms of delay and energy. This work bridges the gap between experimental validation, theoretical modeling, and circuit implementation, illustrating the potential of HfOx-based memristor for high performance, energy efficient computations.

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

Logic Gate Design for In-memory Computing with Sputtered Hafnium Oxide Based Memristor

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rose Mary Kuruvithadam,
  • Antony Sharon,
  • Aldrin Antony,
  • M. K. Jayaraj,
  • S. Nalesh

摘要

Artificial intelligence has advanced with the emergence of deep learning techniques. However, the performances of deep neural networks in Von Neumann architectures are limited by the memory bottleneck. In-memory computing using memristive devices, is a promising solution, with significant improvement in speed, efficiency, and scalability. This work demonstrates the application of RF-sputtered Hafnium Oxide (HfOx) based memristive device for implementing logic gates targeted at In-memory computing applications. A model for Pt/HfOx/Ag based memristor was developed, adopting the Voltage ThrEshold Adaptive Memristor (VTEAM) model, showing close match with experimental I-V data (rms error:0.0078). The Verilog-A model for this memristive device was developed and implemented in Cadence Virtuoso circuit simulator and realized basic logic gates using the Memristor Aided Logic design-MAGIC and analyzed its performances in terms of delay and energy. This work bridges the gap between experimental validation, theoretical modeling, and circuit implementation, illustrating the potential of HfOx-based memristor for high performance, energy efficient computations.