Effective defect detection and classification methodology based on integrated laser scanning inspection and automatic defect classification

Abstract

In-line defect monitoring in VLSI manufacturing is indispensable for yield management in the sub-0.25 /spl mu/m era. One of the most important functions of wafer inspection is to capture process excursions and identify the sources of yield-limiting (killer) defects. Wafer inspection has three stages: (1) defect detection, (2) defect review and classification, and (3) process defectivity trend analysis. This paper presents a new methodology for wafer inspection and defect classification by integrating a production-proven wafer inspection system (KLA-Tencor AIT) with a production-proven ADC (automatic defect classification) system (KLA-Tencor IMPACT ADC). The integrated system takes a cassette of wafers and automatically produces all information needed to analyze defectivity trends by type. Adding the on-board ADC system does not increase inspection system footprint, which is an important consideration in an industry where production floor space is precious. In this paper, we present results of characterization of the combined inspection/ADC system, including case studies from manufacturers. The results include: ADC accuracy and purity compared to manual classification on various process layers; overall time-to-results compared to traditional inspection/classification strategies; and defect sizing based on high resolution defect images of ADC compared with SEM measurement. Key advantages of the combined inspection/ADC system were found to include high classification accuracy and consistency, improved ability to track defectivity trends by defect type, improved overall time to results, and reduction of process excursion costs to IC manufacturers.