GOALI: "Run-to-Run Feedback Control of Semiconductor Manufacturing Processes"

9510625 Krogh该项目的目标是开发使用半导体制造工艺的在线测量来调整工艺运行点的方法，以补偿工艺变量中系统漂移的来源。半导体制造过程通常使用统计过程控制(SPC)技术进行监控，只有在检测到失控条件时才在操作设置点进行调整。这种方法不能有效地补偿系统参数变化，如工艺操作环境、材料特性或设备条件中的漂移。因此，当消除系统差异太昂贵或在物理上不可能时，需要用于从运行到运行调整工艺工作点的方法，以提高半导体制造工艺的质量和产量。该项目将结合为系统识别、参数估计和反馈控制开发的技术，以及统计和物理建模方法，对商业光伏电池制造设施中的关键过程进行有效的运行控制。将开发参数化模型来表征每个过程的输入-输出关系，并将应用所谓的灰箱参数估计、卡尔曼滤波和约束优化来根据以前运行的过程的后处理测量来更新每个过程的过程设置点。这项工作的主要重点将是开发将来自物理模型和经验数据的过程知识纳入在线计算的方法，以更有效地补偿过程变化。需要探索的关键权衡是模型的复杂性与在线参数估计和控制计算的效率和有效性。除了这些迫在眉睫的技术目标外，这个由NSF GOALI计划赞助的为期一年的项目将使目前在半导体工艺控制领域工作的首席研究员获得在商业半导体制造设施与工业科学家和工程师合作的宝贵经验。这项研究将在Ebara Solar，Inc.进行，并得到部分支持。模型将使用Ebara的工艺数据获得，并将在实际生产设备上实施运行到运行反馈控制技术，以评估其有效性。这一经验将对未来复杂制造过程基于模型控制的理论和应用研究方向的定义做出重大贡献。***

9510625 Krogh The objective of this project is to develop methods for using in- line measurements from semiconductor manufacturing processes to adjust process operating points from run to run to compensate for sources of systematic drift in the process variables. Semiconductor manufacturing processes are typically monitored using statistical process control (SPC) techniques, and adjustments are made in operating setpoints only when out-of- control conditions are detected. This approach cannot effectively compensate for systematic parameter variations such as drifts in the process operating environment, material properties, or equipment conditions. Consequently, methods for adjusting the process operating point from run to run are needed to improve the quality and yield of semiconductor manufacturing processes when it is too costly or physically impossible to eliminate systematic variations. This project will combine techniques developed for system identification, parameter estimation, and feedback control, with methods for statistical and physical modeling to perform effective run-to-run control for key processes in a commercial photovoltaic cell manufacturing facility. Parameterized models will be developed to characterized the input-output relations for each process, and so-called gray-box parameter estimation, Kalman filtering, and constrained optimization will be applied to update the process setpoints for each process run based on post-process measurements from previous runs. A primary emphasis of this work will be to develop methods for incorporating process knowledge from physical models and empirical data into the on-line computations to more effectively compensate for the process variations. The key trade-off to be explored is the complexity of the models versus the efficiency and effectiveness of the on- line parameter estimation and control computations. In addition to these immediate technical objectives, this one- year project sponsored under the NSF GOALI program will enable the principal investigator, who is currently working in the area of semiconductor process control, to gain valuable experience working with industrial scientists and engineers in a commercial semiconductor manufacturing facility. The research will be conducted at - and supported in part by - EBARA Solar, Inc. Models will be obtained using process data from EBARA, and the run-to-run feedback control techniques will be implemented on actual production equipment to evaluate their effectiveness. This experience will contribute significantly to the definition of future theoretical and applied research directions in model- based control of complex manufacturing processes. ***