Scholarlink

<p>Nanoimprint lithography (NIL) is a high-resolution nanofabrication technique capable of producing nanoscale patterns with high resolution. However, the main strategy for enhancing replication quality relies heavily on experimental trial-and-error methods, resulting in significant material waste and posing environmental concerns. In this study, back propagation neural network (BPNN) was employed to predict imprinting result. A three-layer BPNN model was established using imprinting parameters including initial resist thickness, imprint temperature, imprint duration time and average pressure. The minimum mean square error achieved was 0.00070, and the overall correlation coefficient of the model reached 0.95143, with an average relative percentage error of only 4.3667%. These results indicate that the model exhibits good predictive performance. By applying this model, it is possible to predict the replication accuracy of thermal nanoimprinting patterns. The established "data + artificial intelligence" research model effectively improves the research efficiency and reliability of NIL.</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

A prediction model for the replication quality of nanoimprinting patterns based on BP neural network

  • 期刊论文

  • 发表日期：

    2025年3月8日

查看全文
  • Hong-Wen Sun,
  • Yan-Chun Huang,
  • Tian-Hua Tang,
  • Jie Shen,
  • Li-Jun Gu,
  • Jing-Sheng Wang

摘要

Nanoimprint lithography (NIL) is a high-resolution nanofabrication technique capable of producing nanoscale patterns with high resolution. However, the main strategy for enhancing replication quality relies heavily on experimental trial-and-error methods, resulting in significant material waste and posing environmental concerns. In this study, back propagation neural network (BPNN) was employed to predict imprinting result. A three-layer BPNN model was established using imprinting parameters including initial resist thickness, imprint temperature, imprint duration time and average pressure. The minimum mean square error achieved was 0.00070, and the overall correlation coefficient of the model reached 0.95143, with an average relative percentage error of only 4.3667%. These results indicate that the model exhibits good predictive performance. By applying this model, it is possible to predict the replication accuracy of thermal nanoimprinting patterns. The established "data + artificial intelligence" research model effectively improves the research efficiency and reliability of NIL.