Proceedings Article | 14 September 2001
KEYWORDS: Error analysis, Semiconducting wafers, Metrology, Critical dimension metrology, Scanners, Statistical analysis, Lithography, Reticles, Fiber optic illuminators, Feature extraction
CD Uniformity (CDU), a.k.a. CD control, linewidth variation, etc., is a lithographic tool parameter that tool vendors and chip manufacturers alike have an interest in improving. A difficult task has been deciding where to concentrate improvement effort. Expending resource to reduce, for instance, lens aberrations may not pay off in finished product if the culprit in CD error is, for example, pupil-fill uniformity. In recent years, formal analysis tools, including error budgets, have been used to analyze CDU error. A good analysis tells us where the errors originate, their relative size, and the expected effect of the different improvements available. Statistical analysis of CDU data is the necessary first step in separating tool systematic errors, process systematic errors, and random errors. This has been discussed by other authors and is easily performed using standard techniques, typically on a sample of features exposed on a wafer at best focus and best exposure. The next step, assignment of responsibility for error to tool subsystems or to process parameters, must employ additional data obtained in other tests, such as CD-through-focus exposures or illumination uniformity measurements. This paper describes an error-budget analysis of CDU data from a Nikon scanner and provides examples not only of the error breakdown but also of observable improvement steps taken during the project. This includes data measured both with CD-SEM metrology and with Nikon's OCD technique. Recommendations for specification limits on tool subsystems as well as for metrology methods are presented.
