Impact of pixel size on unbiased roughness measurements of lines and spaces

Abstract

Background Measuring and subtracting SEM noise from a biased measurement of roughness leads to an unbiased roughness measurement. Biased roughness values are strongly influenced by the pixel size of the image due to both sampling effects and sample damage or charging.

Aim It is important to understand the impact of pixel size on unbiased roughness measurement to establish a range of pixel sizes that can produce measurements that deserve the label “unbiased.”

Approach The impact of pixel size on the measurement of unbiased roughness is investigated using downsampling to explore sampling effects and examining both after develop and after etch images to explore sample damage and charging.

Results Adequate measurement results are obtained when the Y pixel size is less than half of the roughness correlation length (less than a quarter of the correlation length is preferred) and X pixel size is small enough to ensure that the PSD signal to noise ratio is above 2. To reduce the impact of sample damage or charging, especially on resist patterns, the number of frames of averaging can be lowered, but only so long as the linescan signal-to-noise ratio is kept above 2.

Conclusions Smaller pixel sizes produce less systematic error in what is commonly referred to as unbiased roughness measurements but can have the downside of more sample damage. Using the guidelines provided in this paper can help to ensure accuracy in unbiased roughness measurements.