Process control and base line monitoring using optical 'On the Fly' and SEM classification as implemented in advanced DRAM manufacturing

Abstract

As technology advances and device complexity increases, an efficient combination of automatic Optical OTF (On The Fly) and SEM classification is required to ensure reliable and tight in-line process control. In order to monitor a wide variety of critical defects, patterned wafer inspection is performed at higher sensitivity. As a result in some process steps, such as Deep Trench, the overall defect level increases dramatically making the monitoring of critical defects difficult, since they represent a small percentage of the total count. The following article demonstrates that by integrating automatic classification on Optical inspection and SEM review, and setting up class density SPC on fab YMS it is possible to achieve efficient process control. The methodology that was developed at Infineon Dresden using Applied Materials' Compass inspection tool and SEMVision review system is presented together with the results obtained in the project. The implementation of OTF and SEM ADC at Infineon Deep Trench layer used high accuracy & purity classification strategy leading to efficient process control using base line class density monitoring. An alternative strategy based on lower accuracy/purity results allows class excursion control only. The two classification engines are complimentary to one another. OMNIVIEW technology provides the Compass with unique OTF classification capabilities. OTF is performed during the inspection and enables real time comprehensive feedback on the process by isolating and reporting critical defects. Class density monitoring on these defect types is done either with Compass or with the SEMVision, which offers higher resolution ADC. Smart sampling is performed after OTF and the pre-selected defect types are Classified with the SEMVision ADC. As a result of combined OTF and SEM ADC strategy, review capacity is utilized efficiently, and fulfills the lot-sampling requirement for optimal process control.