|
Photomask technology has attained feature sizes of about 50nm and below. Whereas the main feature size is still above
70-80nm at 20nm technology node recently reported e.g. by Toppan Printing Company as developed, assist features for
this node are in the range of 50-60nm.
One of the critical aspects of this technology development is the cleaning process. Processes are supposed to clean off
contamination and particles down to a defect size of about 40nm and at the same time prevent damage to assist features
in the same size range. Due to obvious trade offs between cleaning power and Feature Damage Probability (FPD), this
task becomes tricky. Improvement of cleaning processes by raising the power of megasonic (MS) cleaning, or adjusting
the speed and size of droplets for spray cleaning occurs at the expense of increased feature damage. Prolongation of
physical cleaning steps does not necessarily leads to improvement of the cleaning as shown previously. Susceptibility
to feature damage occurs predicatively according to dimension and orientation. This allows us to extrapolate a Feature
Damage Limit (FDL) which approximates the smallest feature size for which a process has an acceptable probability of
success.
In a practical sense, the most advantageous approach seems to be to adjust the cleaning power to the maximum allowed
by the FDP and then optimize to the lowest process time necessary to reach expected cleaning efficiency.
Since there are several alternative physical cleaning principles, we have to pick the best one for a given application.
At this point we have to raise the question of how to compare the cleaning efficiency of processes. The goal of this
work is to provide a method for evaluation and comparison of cleaning efficiency between physical cleaning processes
and demonstrate the method on an example. We will focus on comparing two physical cleaning processes 1MHz
megasonic and binary spray process.
|
