Scholarlink

<p>AlGaN/GaN High Electron Mobility Transistors (HEMTs) are essential for high-power and high-frequency electronics; nonetheless, challenges, such as short-channel effects, inadequate gate control, and leakage currents, persist, hindering their scalability. This work presents a novel ferroelectric-spacer (FE-spacer) AlGaN/GaN HEMT architecture, which incorporates a ferroelectric Hf₀.₅Zr₀.₅O₂ layer positioned between the gate and the AlGaN barrier. The proposed design aims to improve electrostatic control and enhance Two-Dimensional Electron Gas (2DEG) confinement while maintaining the integrity of the gate stack. This device undergoes analysis through calibrated Sentaurus TCAD simulations and a compact artificial neural network (C-ANN) that has been trained on over 4000 datasets for predictive modeling. The proposed FE-spacer HEMT demonstrates a 35% increase in drain current (I<sub>on</sub> = 7.12 × 10⁻<sup>4</sup> A), a 45% enhancement in transconductance (g<sub>m</sub> = 2.47 mS/mm), and a 30% rise in cutoff frequency (f<sub>T</sub> = 21.5&#xa0;GHz) relative to conventional HEMTs, along with a further 20% improvement over non-ferroelectric HfO₂ spacer devices. The threshold voltage (V<sub>th</sub>) is determined from the linear and saturation regions, yielding values of 0.311&#xa0;V and 0.275&#xa0;V, respectively. The Subthreshold Slope (SS) attains a value of 68.57&#xa0;mV/dec in the linear region, nearing the theoretical limit of 60&#xa0;mV/dec at 300&#xa0;K, whereas DIBL exhibits a reduction of 18%. The C-ANN model attains a <i>R</i><sup>2</sup> value exceeding 0.995, effectively predicting various performance parameters while requiring minimal computational resources. The results indicate that the incorporation of a ferroelectric spacer offers a scalable and fabrication-compatible method to improve both DC and RF characteristics of GaN-based HEMTs. Additionally, machine learning facilitates efficient and precise design optimization for advanced high-frequency, low-power applications.</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

Performance enhancement of AlGaN/GaN HEMTs using ferroelectric spacer integration and neural network modeling

  • 期刊论文

  • 发表日期：

    2026年3月18日

查看全文
  • K. Vinothkumar,
  • R. Mohan Raj

摘要

AlGaN/GaN High Electron Mobility Transistors (HEMTs) are essential for high-power and high-frequency electronics; nonetheless, challenges, such as short-channel effects, inadequate gate control, and leakage currents, persist, hindering their scalability. This work presents a novel ferroelectric-spacer (FE-spacer) AlGaN/GaN HEMT architecture, which incorporates a ferroelectric Hf₀.₅Zr₀.₅O₂ layer positioned between the gate and the AlGaN barrier. The proposed design aims to improve electrostatic control and enhance Two-Dimensional Electron Gas (2DEG) confinement while maintaining the integrity of the gate stack. This device undergoes analysis through calibrated Sentaurus TCAD simulations and a compact artificial neural network (C-ANN) that has been trained on over 4000 datasets for predictive modeling. The proposed FE-spacer HEMT demonstrates a 35% increase in drain current (Ion = 7.12 × 10⁻4 A), a 45% enhancement in transconductance (gm = 2.47 mS/mm), and a 30% rise in cutoff frequency (fT = 21.5 GHz) relative to conventional HEMTs, along with a further 20% improvement over non-ferroelectric HfO₂ spacer devices. The threshold voltage (Vth) is determined from the linear and saturation regions, yielding values of 0.311 V and 0.275 V, respectively. The Subthreshold Slope (SS) attains a value of 68.57 mV/dec in the linear region, nearing the theoretical limit of 60 mV/dec at 300 K, whereas DIBL exhibits a reduction of 18%. The C-ANN model attains a R2 value exceeding 0.995, effectively predicting various performance parameters while requiring minimal computational resources. The results indicate that the incorporation of a ferroelectric spacer offers a scalable and fabrication-compatible method to improve both DC and RF characteristics of GaN-based HEMTs. Additionally, machine learning facilitates efficient and precise design optimization for advanced high-frequency, low-power applications.