PROCEEDINGS ARTICLE | March 10, 2010
Proc. SPIE. 7640, Optical Microlithography XXIII
KEYWORDS: Lithography, Optical lithography, Switching, Manufacturing, Image resolution, Computer simulations, Photomasks, SRAF, Critical dimension metrology, Algorithm development
SRAF (Sub Resolution Assist Feature) technique has been widely used for DOF enhancement. Below 40nm design
node, even in the case of using the SRAF technique, the resolution limit is approached due to the use of hyper NA
imaging or low k1 lithography conditions especially for the contact layer. As a result, complex layout patterns or random
patterns like logic data or intermediate pitch patterns become increasingly sensitive to photo-resist pattern fidelity. This
means that the need for more accurate resolution technique is increasing in order to cope with lithographic patterning
fidelity issues in low k1 lithography conditions. To face with these issues, new SRAF technique like model based SRAF
using an interference map or inverse lithography technique has been proposed. But these approaches don't have enough
assurance for accuracy or performance, because the ideal mask generated by these techniques is lost when switching to a
manufacturable mask with Manhattan structures. As a result it might be very hard to put these things into practice and
production flow.
In this paper, we propose the novel method for extremely accurate SRAF placement using an adaptive search algorithm.
In this method, the initial position of SRAF is generated by the traditional SRAF placement such as rule based SRAF,
and it is adjusted by adaptive algorithm using the evaluation of lithography simulation. This method has three advantages
which are preciseness, efficiency and industrial applicability. That is, firstly, the lithography simulation uses actual
computational model considering process window, thus our proposed method can precisely adjust the SRAF positions,
and consequently we can acquire the best SRAF positions. Secondly, because our adaptive algorithm is based on optimal
gradient method, which is very simple algorithm and rectilinear search, the SRAF positions can be adjusted with high
efficiency. Thirdly, our proposed method, which utilizes the traditional SRAF placement, is easy to be utilized in the
established workflow. These advantages make it possible to give the traditional SRAF placement a new breath of life for
low k1.
