Scholarlink

<p>A study on the synaptic characteristics dependent on programming-pulse properties, such as the programming pulse width (PPW), of amorphous In-Ga-Zn-O thin-film transistors (a-IGZO TFTs) is presented. To validate this, the weight-update characteristics with respect to PPWs (e.g., 1.5&#xa0;s, 3&#xa0;s, and 6&#xa0;s) are monitored. Here, the memory functionality of a-IGZO TFTs can be achieved with defects (e.g., trap states) in the gate insulator deposited at a low temperature process, leading to the electron trapping or de-trapping phenomena. Based on this function, when the programming pulses with same duty cycle and total time are applied to the gate terminal, the dynamic ratio for PPW = 3&#xa0;s is found to be 18.5, which is larger compared to other PPWs. This is because the energy per programming-pulse for PPW = 1.5&#xa0;s is relatively low whereas the recovery during the read process for PPW = 6&#xa0;s is faster, degrading the weight-update characteristics. In order to fairly compare this recovery effect for each PPW, another pulsed experiment with different duty cycles is conducted maintaining the same number of programming pulses. As a result of the experiment, when the PPW is increased while keeping the same recovery condition, it is found to be a larger dynamic ratio due to a higher energy per programming pulse. With results of this device level, the analog AI accelerator simulation is also performed monitoring the recognition accuracy for three PPW cases.</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

Influence of programming-pulse properties on weight-update characteristics of charge-trapping IGZO synaptic transistors

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月26日

查看全文
  • Younggeon Ko,
  • Jaeyeong Ryu,
  • Jeongseok Pi,
  • Danyoung Cha,
  • Sungsik Lee

摘要

A study on the synaptic characteristics dependent on programming-pulse properties, such as the programming pulse width (PPW), of amorphous In-Ga-Zn-O thin-film transistors (a-IGZO TFTs) is presented. To validate this, the weight-update characteristics with respect to PPWs (e.g., 1.5 s, 3 s, and 6 s) are monitored. Here, the memory functionality of a-IGZO TFTs can be achieved with defects (e.g., trap states) in the gate insulator deposited at a low temperature process, leading to the electron trapping or de-trapping phenomena. Based on this function, when the programming pulses with same duty cycle and total time are applied to the gate terminal, the dynamic ratio for PPW = 3 s is found to be 18.5, which is larger compared to other PPWs. This is because the energy per programming-pulse for PPW = 1.5 s is relatively low whereas the recovery during the read process for PPW = 6 s is faster, degrading the weight-update characteristics. In order to fairly compare this recovery effect for each PPW, another pulsed experiment with different duty cycles is conducted maintaining the same number of programming pulses. As a result of the experiment, when the PPW is increased while keeping the same recovery condition, it is found to be a larger dynamic ratio due to a higher energy per programming pulse. With results of this device level, the analog AI accelerator simulation is also performed monitoring the recognition accuracy for three PPW cases.