Impact of registration error of reticle on total overlay error budget

As the overlay specification decreases drastically, it is necessary to consider how the total overlay is influenced by each contributing factor. In particular, it is expected that the contribution on overlay error budget can be quantitatively analyzed in terms of the correlation among registration errors of the reticle. The reticle contribution of about 25% is assessed by the breakdown of the sources of overlay metrology uncertainty through the double exposure technique (DET) process. A positive correlation of around 0.7 mitigates the reticle contribution by 180%, compared to the uncorrelated case. In both DET and double patterning technique (DPT) processes, it is needed to positively correlate the registration errors among many reticles in order to decrease the reticle contribution. To maintain the gain of 180% due to positive correlation, the correlation coefficient requisite has to be increased it is difficult to achieve highly positive correlation among more than three reticles. From an integration point of view, three reticles used in DET and DPT generate three combinations of reticle sets. When the positive correlation between two reticle sets is determined to be 0.78, the other reticle set will statistically have a smaller correlation coefficient, 0.61 < 0.78. With an overlay specification of 12% of design rule and a reticle contribution of 25%, the specification for registration error was expected to be 7.7 nm (4 X) at the 45 nm node and 5.5 nm (4 X) at the 3 2 nm node in the uncorrelated case between two reticles. The positive correlation of around 0.7-0.78 helps to reduce the reticle specification by around 200% even in the DPT process. (c) 2006 American Vacuum Society.