Research On Inverse Lithography Technology And Mask Complexity

As the IC process enters the UDSM era, lithography technology has become one of the obstructions of IC technology. As Optical Proximity Effects becomes more and more serious in the process of IC manufacturing, semiconductor manufacturing community adopts many kinds of Resolution Enhancement Technologies to deal with it. As a key technology of traditional Resolution Enhancement Technology, Optical Proximity Correction (OPC) is broadly used in current IC manufacturing. Now the critical dimension becomes even smaller and the leading manufacturers have been working at 22nm node. It becomes more and more difficult to apply OPC in these complex masks, while Inverse Lithography Technology (ILT) becomes more promising for its unique merits. At the same time, the masks created by ILT are so complex that they are very difficult to be manufactured in actual production, which limits the development of ILT.In this paper, ILT and penalty terms are introduced in detail, while mask complexity is studied profoundly. Based on the Local Wavelet Penalty introduced in a former paper, a new mask complexity penalty called Global Wavelet Penalty in the paper is introduced. Unlike Local Wavelet Penalty which reduces the high frequency components of mask on the fixed direction, Global Wavelet Penalty evaluates the high frequency components of mask more profoundly by applying four distinctive Haar wavelet transforms and choosing the optimal direction on which the highest frequency components of the mask will be removed. In this way, the high frequency components of mask can be reduced more effectively, so can be the mask complexity. Based on Global Wavelet Penalty, a new gradient-based ILT algorithm is proposed for mask complexity reduction. Finally according to characteristics of Flash array masks, adjustments are made on the ILT algorithms to speed up the running time, while the accuracy of the lithography results keeps on the same level. Experiments with three typical Flash patterns prove that compared with the original algorithm, the proposed algorithm can reduce mask complexity more effectively.