Accelerating 32nm BEOL technology development by advanced wafer inspection methodology

Abstract

In the early development stage of 32nm processes, identifying a nd isolating systematic defects is critical to understanding the issues related to design and process interactions. Conventional inspection methodologies using random review sa mpling on large defect popul ations do not provide the information required to take accurate and qu ick corrective action. This paper demonstrates the successful identification and isolation of systematic defects using a novel methodology that combines Design Based Binning (DBB) and inline Defect Organizer (iDO). This new method of integrating design and defect data produced actionable inspection data, resulting in fewer mask revisions and reduced device development time. 2. INTRODUCTION Engineers must balance design margins and process windows, while achieving fast development times. In the early development stage, identifying systematic defects accurately and quickly is critical in order to minimize cost and shorten the development cycle. Bright field inspectors are often used to help identify integration defects, but high sensitivity inspections can produce very large defect populations that contain both random a nd systematic defect types. Limited review sampling on these high defect counts prevents the effective separati on of design or process-related defects from non-relevant defect types. Thus, these conventional inspection methods produce an incomplete picture of issues related to proces s and design interactions, making it difficult, if not impossible, for engineers to take effective correctiv e action. DBB is a new t echnology that classifies defects into groups based on design background [1,2]. In addition, defect critical index (DCI) can be extracted to describe what the impact of defect is [3], while iDO uses design and defect attributes to identify and separate different cla sses of defects. Using these technol ogies, systematic defects can be identified and separated from the overall defect population, resulting in improved yield relevance and faster solution of production integration issues. For example, the defect population can be