SGER: Sequential Bayesian Decision Making for End-Point Detection and Control in Chemical Mechanical Planarization (CMP) Processes

本探索性研究小额资助（SGER）项目的目的是应用（概率）决策理论，寻找化学机械平面化过程中测量到的传感器信号特征与动态状态和材料去除率之间的定量关系，用于终点检测和控制。该方法将使用顺序贝叶斯模型在每个时间步估计过程的状态，从测量的振动和其他传感器信号。将定义（过程状态和决策的）效用函数，并在每个时间步骤评估。决定是停止机器还是继续抛光，将在每个时间步骤中根据最小化预期效用的替代方案进行。该研究将结合pi在化学机械抛光科学，传感器和信号分析以及贝叶斯决策理论方面的互补专业知识。如果成功，探索性的努力将引入一种根植于决策理论的全新方法，用于实时预测和控制半导体制造过程中的质量和生产率。该方法可以推广到其他超精密制造工艺和复杂系统的质量和性能预测和控制中。该传感器联网抛光机可作为学生和行业人员指导和培训的试验台。该研究将与pi的研究生课程相结合。pi将继续让来自代表性不足群体的本科生和研究生参与他们的研究。他们将接触到基础的多学科研究和先进制造工艺的行业实践。研究重点将在pi的网站上公布，并在各种期刊、贸易杂志和会议论文集上发表。

The objective of this Small Grant for Exploratory Research (SGER) project is to apply (probabilistic) decision theory to find a quantitative relationship that connects the measured sensor signal features with the dynamic state and the material removal rate in chemical mechanical planarization process for end-point detection and control. The approach will be to use a sequential Bayesian model to estimate at each time step the state of the process from the measured vibration and other sensor signals. A utility function (of the process state and decisions) will be defined, and evaluated at each time step. A decision on whether to stop the machine or to continue polishing will be made at every time step based on the alternative that minimizes the expected utility. The research will combine the complementary expertise of the PIs in chemical mechanical polishing science, sensors and signal analysis, and Bayesian decision theory.If successful, the exploratory effort will introduce a radically new approach, rooted in decision theory, for real-time prediction and control of quality and productivity in semiconductor manufacturing processes. This methodology can be extended to prediction and control of quality and performance in other ultra-precision manufacturing processes and complex systems. The sensor-networked polishing machine can serve as a test bed for instruction and training of students as well as industry personnel. This research will be integrated with the PIs' graduate courses. The PIs will continue to involve undergraduate and graduate students from underrepresented groups as part of their research. They will be exposed to fundamental multi-disciplinary research and industry practices in advanced manufacturing processes. The research highlights will be made available in the PIs' website and published in various journals, trade magazines, and conference proceedings.