钻石
人工神经网络
材料科学
领域(数学)
电磁场
微波食品加热
计算机科学
生物系统
计算机模拟
计算物理学
航程(航空)
等离子体
电磁学
核工程
功能(生物学)
基质(水族馆)
化学气相沉积
最优化问题
声学
球(数学)
机械
物理
作者
Jie Shen,Zhiyin Gan,D H Zhang,Yinghao Wang
出处
期刊:Physica Scripta
[IOP Publishing]
日期:2026-05-29
卷期号:101 (25): 255603-255603
标识
DOI:10.1088/1402-4896/ae750a
摘要
Abstract To enhance the electromagnetic field performance of microwave plasma chemical vapor deposition (MPCVD) reactors for diamond growth, a 915 MHz MPCVD system was selected as the research model. Six geometric parameters were chosen as optimization variables, and numerical simulations were conducted to analyze and optimize the reactor’s electromagnetic field distribution. The relationship between reactor dimensions and electromagnetic field characteristics was fitted using a back-propagation neural network. By maximizing a weighted objective function incorporating central field intensity, field uniformity, and growth radius, multi-parameter optimization followed by secondary refinement was performed. The optimized reactor exhibited an increased central field intensity and an enlarged growth region. Furthermore, narrowing the fitting range of the neural network enhanced the prediction accuracy and optimization performance. Multi-physics coupled simulations confirmed that the optimized 915 MHz reactor generated a more stable and uniform plasma ball above the substrate holder, demonstrating its suitability for high-quality diamond growth.
科研通智能强力驱动
Strongly Powered by AbleSci AI