|
It is necessary to apply extreme illumination condition on real device as minimum feature size of the device
shrinks. As k1 decrease, ultra extreme illumination has to be used. However, in case of using this illumination, CD and
process windows dramatically fluctuate as pupil shapes slightly changes. For past several years, Pupil Fit Modeling
(PFM) was developed in order to analyze pupil shape parameters which are independent from each others. The first
object in this work is to distinguish pupil shape of different scanner by separating more parameters. According to pupil
parameter analysis, the major factors of CD or process window difference between two scanner systems obviously
appear. Due to correlation between pupil parameter and scanner knob, pupil parameter analysis would be clearly
identified which scanner knob should be compensated. The second object is to define specification of each parameter by
using analysis of CD budget for each pupil parameters. Using periodic monitoring of pupil parameter which is controlled
by previous specification, scanner system in product lines can be maintained at ideal state. Additionally, OPC model
accuracy enhancement should be obtained by using highly accurate fitted pupil model. Recently, other application of
pupil model is reported for improvement of OPC and model based verification model accuracy. Such as modeling using
average optics and hot spot detection of scanner specific model are easily adopted by using pupil fit model. Therefore,
applications of pupil fit parameter for process model are very useful for improvement of model accuracy.
In our study, the quantity of model accuracy enhancement using PFM is investigated and analyzed. OPC and
hotspot point detection capability results with pupil fit model would be shown. Also, in this paper, trends of CD and
process window for each scanner parameter are evaluated by using pupil fit model. As of results, we were able to find
which pupil parameter has influence in critical layer CD and application of this result resulted in better accuracy in
detecting hotspot for model based verification.
|
