The comparison of the effectiveness of model-based SRAFs and rule-based SRAFs

The SRAFs that are usually in the form of scattering bars (SBs) are inserted to the contact layer to maintain proper local pattern densities, ensuring a good resolution on a full chip scale. There exist 2 practical addition schemes of SBs; namely, the rule-based (RB) SB and model-based (MB) SB. The RB SBs are set up by the inspection of local pattern geometries and their effectiveness heavily relies on engineers' experiences. In contrast, the MB scheme adds the SBs in accordance to the computed contrasts of contact edges. It is fully environment adaptive and can hence save a huge amount of human efforts. Moreover, the MB scheme is capable of taking the process window (PW) conditions into account so that DOF (depth of focus) and EL (energy latitude) can be optimized on a full chip scale due to the added SBs. Since the PW conditions are not explicitly included in the RB scheme, it is expected that the MB SBs should have a better performance than the RB SBs. Using a real chip as a test case, we have found that the MB scheme can better control the PV (process variability) band width though, the overall performances of these 2 SB schemes are comparable. However, optimizing a limited number of adjustable parameters can further improve the performance of MB SBs. Both schemes yield some weak points. The weak points that are resulted from one scheme can be fixed by the other scheme. Hence, these 2 schemes are complementary to each other. This inspires us to incorporate the MB SB into the post-OPC repair flow.