Uncertainty in roughness measurements: putting error bars on line-edge roughness

Abstract

Measurement of line-edge or linewidth roughness involves uncertainty, like all measurements, and an estimate of that uncertainty should be reported whenever a roughness measurement is reported. However, roughness measurement uncertainty estimates are complicated by the correlations along the length of the rough feature. As a result, roughness measurements are often not accompanied by uncertainty estimates or error bars on graphs. Here, both theoretical considerations and simulations of random rough features will be used to derive a simple formula to estimate the uncertainty of a roughness measurement using the standard parameters describing that roughness: standard deviation, correlation length, and roughness exponent. Additionally, a more accurate formula to estimate the systematic bias in roughness standard deviation is provided.