Scholarlink

<p>This article investigates the performance of a source pocket diverse dielectric double gate tunnel FET (SP-DD-DG-TFET) as Leaky-Integrated-Fire (LIF) neuron model. Efficient learning and recognition tasks are enabled by a neurobiology-inspired spiking neural network (SNN). For this requirement a power and speed efficient electronic neuron is essential to achieve a large-scale network akin to biological neuron. The working principle of this device is band-to-band tunneling (BTBT), which enables low off-state leakage current (I<sub>OFF</sub>), high firing or spiking frequency (f<sub>o</sub>), and reduced energy consumption per spike, akin to LIF neurons. In this proposed article, LIF neuron consumes 0.4 aJ per spike, which is lower compared to DG-TFET, CMOS, DG-JLFET, PD-SOI-Fin FET, SOI JLFET, Bulk Fin-FET, SOI CMOS, and conventional MOSFET, respectively. This SP-DD-DG-TFET device, based on a LIF neuron, exhibits a firing frequency of 0.173 THz, highlighting its potential for extremely fast computation in neuromorphic circuits. Due to its high spiking frequency, low energy consumption per spike, and CMOS compatibility for neuromorphic computation, the proposed SP-DD-DG-TFET-based LIF neuron device is well-suited for large-scale hardware implementation of SNNs.</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

Investigating Source-Pocket TFETs for High-Speed, Energy-Efficient LIF Neural Devices: A Performance Estimation Study

  • 期刊论文

  • 发表日期：

    2026年1月16日

查看全文
  • Rajeev Ranjan,
  • Nivedita,
  • Priyanka,
  • Dheeraj Kumar Sinha,
  • Sanjay Kumar

摘要

This article investigates the performance of a source pocket diverse dielectric double gate tunnel FET (SP-DD-DG-TFET) as Leaky-Integrated-Fire (LIF) neuron model. Efficient learning and recognition tasks are enabled by a neurobiology-inspired spiking neural network (SNN). For this requirement a power and speed efficient electronic neuron is essential to achieve a large-scale network akin to biological neuron. The working principle of this device is band-to-band tunneling (BTBT), which enables low off-state leakage current (IOFF), high firing or spiking frequency (fo), and reduced energy consumption per spike, akin to LIF neurons. In this proposed article, LIF neuron consumes 0.4 aJ per spike, which is lower compared to DG-TFET, CMOS, DG-JLFET, PD-SOI-Fin FET, SOI JLFET, Bulk Fin-FET, SOI CMOS, and conventional MOSFET, respectively. This SP-DD-DG-TFET device, based on a LIF neuron, exhibits a firing frequency of 0.173 THz, highlighting its potential for extremely fast computation in neuromorphic circuits. Due to its high spiking frequency, low energy consumption per spike, and CMOS compatibility for neuromorphic computation, the proposed SP-DD-DG-TFET-based LIF neuron device is well-suited for large-scale hardware implementation of SNNs.