|
The current abilities for active feedback loops to correct for various parameters challenge metrology groups to
provide exact input data for these correction cycles. One of the most important feedback loops is the one that
deals with the improvement of the CD (critical dimension) uniformity of structures. Here, several processes
rely on exact metrology data to tackle systematic effects that either have to be overcome by finding better
process conditions or compensated actively, for instance, by tuning the writer data.
Right now most of these processes tackle long range effects on the order of millimetres and do not vary a lot
on the micrometer scale. On the other hand, CD measurements are usually performed with instruments that
measure single points with dimensions of a couple of micrometers (such as the conventional CD-SEM). Thus
noise from the micrometer scale is introduced in the global mapping of the uniformity.
Recently, numerical methods, such as the exponentially weighted penalty approach called TPS (thin plate
splines) have been developed that separate between the true signatures on the millimetre scale from the noise
of the micrometer measurements. In this paper, we will take one step further by showing that the acquired
statistically stable CD signature of a CD-SEM measurement matches the CD data measured by a
scatterometer. Furthermore, we will show that the residual of the CD data of the scatterometer measurement
compared to the found TPS fit has a noise level of about 0.1 nm (3σ), which essentially equals the short-term
reproducibility of the tool. This is of high importance since both methods do essentially the same - they
average out micrometer noise with the only difference being that TPS does it theoretically and a scatterometer
does experimentally. Thus, we have the extremely fortunate situation in which theory and experiment give the
same results. Hence, two separate conclusions can be drawn: the scatterometer measures indeed stable
macroscopic CD signatures and TPS is indeed the right method to extract these signatures from any given CD
data.
|
