For typical single and double-periodic structures that scatterometry is employed to measure, grating pitch has traditionally been treated as an invariant and well-known parameter. Mask writing processes and lithographic exposure tools are generally regarded to be sophisticated enough to eliminate the possibility of a significantly uncontrolled or unknown grating-pitch. Considering the modern demands in precision and accuracy placed on scatterometry, however, there is value in re-examining this assumption. The factors that can affect grating-pitch variation or inaccuracy include mask writing errors, mask flatness, lithographic magnification errors, focus-height errors, and lens aberrations.
In order to quantify the effects of grating pitch assumptions, several model-based investigations have been performed. Libraries of models were constructed with assumed and invariant grating-pitches. For comparison purposes, an identical second set of libraries was generated assuming a slightly different grating pitch. Those sets were matched such that the pitch difference effects could be investigated by noting the parameter dimension differences in the matched models. We report the estimated severity of errors in parameter dimensions as a result of modeling intentionally mismatched grating pitches. Data from multiple structures that correspond to typical mainstream scatterometry applications will be shown. A summary of the successful model and algorithm-based compensation techniques will follow. While our investigation will include structures that correspond to the current 65 nm technology node, we will also discuss the effect of pitch mismatch for future technology nodes.