Search

At advanced node of metal layer, defocus conditions need to be taken into account in OPC in order to guarantee metal layer process window. Process window OPC (PWOPC) must be considered for the model-based OPC models under special conditions within a lithographic process window. In this paper, we studied several OPC simulation issues in the application of PWOPC in metal layer. The test run results showed that preparatory work and parameters setting were very important for PWOPC application. Without preparatory work or unbefitting parameters setting, the OPC simulation may lead to undesired results. The depth of focus (DOF) of several hot spots was examined. The pinch and bridge error are hot spots while only OPC model under nominal conditions is employed. These hot spots can be resolved while PWOPC models are in consideration. In addition of PWOPC precision study, PWOPC runtime is also evaluated. The use of PWOPC extends OPC runtime at least 50%, which is unacceptable in the environment of longer OPC runtime for the advanced tech node and continuing pull-in tapeout schedule. The runtime speeding up method is studied to reduce the total runtime.

In 28nm technology node, developing an enough lithographic process window of VIA layer became a major challenge in order to meet the requirements of the connectivity between metal lines. It is widely used to size up VIA in order to enlarge VIA process window if the space between VIA holes and VIA enclosed by metal is big enough. But the traditional retarget method has its own limits as the retarget rule table cannot deal with every complex environment for each VIA hole. In this paper, we studied several issues in the application of CALIBRE auto-retarget function in VIA model-based OPC correction after the traditional retarget treatment. The parameters of the auto-retarget function are optimized based on simulated results and the SEM data of the corresponding VIA holes are collected for DOF verification. The impact of the auto-retarget function on model-based OPC runtime is also evaluated. It is shown in the simulation result that the DOF of certain SRAM patterns can be enlarged as much as to 30nm after applying the auto-retarget function. The silicon data confirms the simulation prediction and that the DOF of the patterns become enough for lithographic process. Meanwhile, auto-retarget has little impact on model-based OPC runtime.

With minimum half-pitch shrinking to 28nm, the accuracy requirement imposed on OPC model is getting harder to achieve. In order to meet the much more challenging requirement, uncertainty in the modeling should be diminished as much as possible. The major factors of OPC modeling accuracy consist of OPC model calibration flow, data analysis of mask modeling, lithographic process parameters, and photoresist CD. In this paper, the parameters of mask modeling are analyzed and the mask induced errors are calculated, such as mask bias, mask corner rounding, and mask 3D effect. Furthermore, the interactive relations among these mask parameters are also investigated. A 28nm test model was constructed with optimized mask modeling approaches. The rms of the test model is significantly reduced to half of that of the traditional modeling approaches.