Virtual Integrated Sensors and Data Fusion for Model Based Control In Intelligent Processing of Materials

Significant progress has been made in recent years (through DARPA and industry initiatives) in developing real-time, on-line, model-based control systems to improve the productivity and flexibility of the US semiconductor industry. However, none of these initiatives have addressed the design of sensing systems which are crucial to the operation and efficiency of feedback control systems. The Phase I effort will focus on developing advanced physical models for sensors, i.e., virtual sensors. Models will be developed for sensors based on interferometric methods (such as IR laser interferometry and fourier transform interferometry) and will be tested on an existing virtual reactor environment developed at CFDRC (through DARPA programs). The study will be used to determine (i) combination of sensors required to diagnose the desired quantity non-intrusively, (ii) location of sensors in the virtual reactor, (iii) sensor accuracy and response characteristics, and, (iv) data fusion methodologies to extract the desired information from multiple sensor input data. The performance of the sensing system will be quantified on a typical Rapid Thermal Processing (RTP) system for the fabrication of semiconductor devices. In Phase II, the effort will be extended to a wider range of sensors based on other non-invasive interferometry methods such as in-situ ellipsometer, thin film interferometer, etc. The virtual sensor will be coupled with virtual controllers to demonstrate the complete real-time control system on fabrication reactors. This effort will be coordinated with the DARPA initiative on Virtual Integrated Prototyping (VIP) for thin films which will focus on model development, model reduction and model based control systems for on-line control of fabrication equipment and processes.