Importance of data quality in virtual metrology

The purpose of VM is to enable the manufacturers to conjecture the wafer quality and deduce the causes of defects without performing physical metrology. VM requires a large amount of sensor data retrieved from production tools. However, inappropriateness and instability of the data collection system, which leads to incorrectness, fragment and asynchrony of data collected, may lead to inaccurate conjecture results. Hence, not only precision of the VM conjecture module but also quality of the collected data are essential to ensure accurate and stable VM results for improving production yield. In this work, the importance of data quality to VM is investigated. The data quality mechanism is proposed for data characteristic analysis, data anomaly detection, data cleaning, data normalization, and data reduction. Besides, the equipment of semiconductor chemical vapor deposition (CVD) is adopted as a practical example to illustrate the significance of data quality mechanism, and further verify feasibility and effectiveness of VM.