"Advanced Control Strategies for Plasma Enhanced Chemical Vapor Deposition"

9312406 Greve这个项目的目标是展示使用新的传感器技术和多变量控制策略来提高集成电路（ic）制造的关键过程的性能。在目前的实践中，集成电路制造的单元过程由宏观变量（如温度、压力和气体流速）的独立调节器控制。这些变量的设定值由离线研究确定，并在处理给定批次晶圆期间不进行调整。虽然设定值调整可以使用统计过程控制（SPC）等技术从批到批进行，但使用实时反馈来实时调整过程设定值直到最近才成为可能，因为缺乏与实际过程性能变量更直接相关的过程变量的现场测量。所研究的特殊工艺是等离子体增强化学气相沉积（PECVD），它具有许多集成电路单元工艺共同的特征。在处理过程中，将使用微分泵浦四极杆质谱仪对直接入射到晶圆平台上的物种进行取样。利用物理模型和实验表征过程的响应曲线和物种浓度相对于调节变量的动态响应，实施多变量控制策略，实现物种浓度的直接调节。然后将根据这些测量的过程变量开发工艺配方，以更直接地影响通过几个性能变量测量的过程结果，包括击穿电压、平带位移、厚度和折射率。将实施和比较两种通用的多变量控制方法（鲁棒多变量调节和模型预测控制），从而对替代多变量控制方法的实际优势和局限性产生新的见解。PECVD工艺的新模型也将在本研究过程中确定，从而根据测量的工艺变量定义的新配方改进工艺性能。许多这些结果将有助于其他关键的IC工艺，如等离子体蚀刻。＊＊＊

9312406 Greve The objective of this project is to demonstrate the use of new sensor technology and multivariable control strategies to improve the performance of critical processes for the manufacture of integrated circuits (ICs). In current practice, unit processes for IC manufacturing are controlled by independent regulators of macroscopic variables such as temperature, pressure, and gas flow rates. Setpoint values for these variables are determined by off-line studies and applied without adjustment during the processing of a given batch of wafers. Although setpoint tuning may be performed from batch-to-batch using techniques such as statistical process control (SPC), the use of real-time feedback to adjust the process setpoints in real time has not been possible until recently because of absence of in situ measurements of process variables more directly related to the actual process performance variables. The particular process to be studied is plasma enhanced chemical vapor deposition (PECVD) which exhibits many features common to several IC unit processes. A differentially-pumped quadrupole mass spectrometer will be used to sample species directly incident on the wafer platform during the processing. Using physical models and experiments to characterize the response curves for the process and the dynamic responses of the species concentrations with respect to the regulated variables, multivariable control strategies will be implemented to enable the direct regulation of the species concentrations. Process recipes will then be developed in terms of these measured process variables to more directly influence the process outcome as measured by several performance variables including breakdown voltage, flat band shift, thickness, and index of refraction. Two general approaches to multivariable control (robust multivariable regulation and model predictive control) will be implemented and compared leading to new insights into the practical advantages and l imitations of alternative multivariable control methodologies. New models of the PECVD process will also be identified in the course of this research leading to improvements in the process performance based on new recipes defined in terms of the measured process variables. Many of these results will be useful for other critical IC processes such as plasma etching. ***