Improve the IDM recipe creation flow by integrating and benchmarking a new solver

The core business of ASML is to develop lithography tools (scanners) for chip manufacturing. As the feature size decreases on the chips, the need to control the performance of ASML scanners is increased therefore this requires more advanced metrology techniques/tools. These advanced metrology tools help to analyze what is actually printed on the wafer with the scanner. YieldStar (YS) is an optical metrology tool developed by ASML. YieldStar is an angle-resolved scatterometer which allows to measure both Optical Critical Dimension (OCD) and Overlay (OV). Since YS uses visible light and the dimensions of the features of interest are generally a few nm, it is difficult to measure these features on wafers in a direct manner. Therefore, it is needed to model the structures/features and use optical modeling and optimization techniques to reconstruct the right profile of the structure. The modelling of the geometrical and material parameters is performed in YS software. Within the YS software, ASML provides their customer a platform to support metrology recipe creation to reconstruct CD (critical dimensions: all the geometrical parameters/features on a chip) based on the measurements performed on YS tool. This metrology recipe creation platform is called In Device Metrology (IDM) Recipe Creation Flow (RCF).
The current IDM RCF is comprised of several steps, which together with the complexity of the IDM profile contribute to the total recipe creation time of an IDM profile. The recipe creation time for complex 3D profiles is high because some of the steps in the RCF require long computation time and large manual user interaction. The recipe creation time is needed to be reduced in order to bring the recipe creation flow within specification which can be achieved by automating the IDM recipe creation step(s). Therefore, to automate the IDM RCF, a new solver has been designed.
In this work, an improvement in the IDM RCF has been proven by validating the performance of the new solver for different IDM profiles. As a result of this study, it is found that the new solver has fully automated the IDM RCF step(s) and the time to recipe is reduced, while, at the same time it has merged some of the steps together and reduced the number of steps in the IDM RCF. The implementation of the new solver has also increased the application scope of the IDM RCF.