Scholarlink

<p>Magnetron sputtering, a widely used physical vapor deposition method, is a plasma-matter interaction process. The magnetron sputtering application includes semiconductor fabrication, optics, and surface coating. The key roles of sputtering in the semiconductor industry are forming interconnects, barrier layers, and electrode contacts in solar cells, integrated circuits, and other microelectronic devices. Despite wide applications and impact, rare mathematical analyses exist for predicting the deposition rate derived from first principles. A supervised machine learning approach uses process parameters such as power, target-substrate distance, and target material to determine the deposition rate. The work explains the process parameters and their impact on the deposition rate. Seven regression machine learning models are briefly discussed, with the relevance for model sputtering deposition rate, which are evaluated using parameters such as Mean Squared Error, Mean Absolute Error, and Coefficient of determination (<i>R</i><sup><i>2</i></sup>). The average performing models are tree-based regression models with <i>R</i><sup><i>2</i></sup> above 0.9 and minimal error. Random Forest and XGBoost are the top-performing models, with <i>R</i><sup><i>2</i></sup> of 0.96 and 0.97, respectively. Predicting the sputter deposition rate using optimized machine learning is a novel approach to reduce experimental time and expenditure.</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

Predicting magnetron sputtering deposition rate through process parameters using supervised machine learning

  • 期刊论文

  • 发表日期：

    2026年2月19日

查看全文
  • Sri Vishnu Jami,
  • Sakti Prasanna Muduli,
  • Paresh Kale

摘要

Magnetron sputtering, a widely used physical vapor deposition method, is a plasma-matter interaction process. The magnetron sputtering application includes semiconductor fabrication, optics, and surface coating. The key roles of sputtering in the semiconductor industry are forming interconnects, barrier layers, and electrode contacts in solar cells, integrated circuits, and other microelectronic devices. Despite wide applications and impact, rare mathematical analyses exist for predicting the deposition rate derived from first principles. A supervised machine learning approach uses process parameters such as power, target-substrate distance, and target material to determine the deposition rate. The work explains the process parameters and their impact on the deposition rate. Seven regression machine learning models are briefly discussed, with the relevance for model sputtering deposition rate, which are evaluated using parameters such as Mean Squared Error, Mean Absolute Error, and Coefficient of determination (R2). The average performing models are tree-based regression models with R2 above 0.9 and minimal error. Random Forest and XGBoost are the top-performing models, with R2 of 0.96 and 0.97, respectively. Predicting the sputter deposition rate using optimized machine learning is a novel approach to reduce experimental time and expenditure.