The Setting of Linewidth Reference on Photomasks through Physical Process Modeling

摘要

The use of photomask can maximally realize the power of parallel pattern replication in photolithography. Although there exists linewidth standard in the metrology tools, in practice, there may exists some variation in the linewidth definition from different chip makers. Experience indicates that such variation can be as much as 1~2 nm (1X). Since the mask bias can affect photo process window, the determination of which can be very important to the setup of the lithography process. Of course, the optimum setting can be obtained through wafer exposure. However, ever since the introduction of Model Based Optical Proximity Correction (MBOPC), the linewidth uniformity across different patterns is mostly accomplished by OPC model and correction. Therefore, it is important that the mask dimension be as close to the reality as possible to make OPC model more physical, which have better extendibility to patterns that may lie along the edge of the design rules or even slightly outside the design rules. Ever since the use of 193 nm immersion lithography, the mask dimension may be very close to the size of the wavelength, mask 3D scattering effect becomes very significant, which can reduce imaging contrast, increase Mask Error Factor (MEF), and reduce common Depth of Focus (DoF) for all patterns. The setting of the accurate mask dimension becomes more important, even critical. However, we can also utilize this process window sensitivity to the mask dimension to determine the real mask dimension through comparing wafer exposure data and simulation. This will require very accurate physical modeling. In this paper, we will propose this method.