Unbiased roughness measurements: Subtracting out SEM effects, part 2

The measurement of roughness of small lithographic patterns is biased by noise in the scanning electron microscopes (SEMs) used to make the measurements. Unbiasing the roughness measurement requires the measurement and subtraction of the image noise based on its unique frequency behavior. Improvement to prior white noise removal is achieved by applying a pink noise model. This pink noise removal technique was applied to roughness measurements made with different electron doses (frames of integration), different operating voltages, and different generations of SEM tools. Effective noise removal to create accurate unbiased estimates of the roughness was achieved over a wider range of SEM tool parameter settings than has been previously achieved. As a result, unbiased roughness measurements can now be used to characterize and improve stochastic variability in semiconductor lithography and patterning.