this type of measurement," says NIST's John Villarrubia, "besides
the real roughness there is also a false roughness
caused by measurement noise. Our method includes
a correction to remove bias or systematic error from
Random noise, by definition, causes the measured
value to be sometimes higher, sometimes lower than
the true value, and can be minimized by simply averaging
an adequate number of measurements. Systematic error,
however, is consistently above or consistently below
the true value due to some quirk of the measurement
The noise in nanoscale scanning electron microscope
(SEM) images consistently adds extra roughness, says
Villarrubia. The NIST/SEMATECH method involves taking
two or more images at exactly the same location on
a circuit feature, comparing the values, and subtracting
the false roughness caused by measurement noise.
SEM manufacturers should be able to incorporate the
new method into their proprietary software for automated
linewidth roughness measurements.
* J.S. Villarrubia and B.D. Bunday, Unbiased Estimation
of Linewidth Roughness, Proceedings of SPIE 5752
(2005) pp. 480-488.
Contact: Gail Porter
National Institute of Standards and Technology