Efficient thin-film stack characterization using parametric sensitivity analysis for spectroscopic ellipsometry in semiconductor device fabrication

• An improved method for multilayered thin-film stack characterization is proposed.
• The screening-type technique based on so-called “elementary effects” was employed.
• The model parameters were ranked according to relative importance for model output.
• The method is tested using two examples of complex thin-film stack characterization.
• The approach can be useful in many practical cases in the microelectronic industry.

During semiconductor device fabrication, control of the layer thicknesses is an important task for in-line metrology since the correct thickness values are essential for proper device performance. At the present time, ellipsometry is widely used for routine process monitoring and process improvement as well as characterization of various materials in the modern nanoelectronic manufacturing. The wide recognition of this technique is based on its non-invasive, non-intrusive and non-destructive nature, high measurement precision, accuracy and speed, and versatility to characterize practically all types of materials used in modern semiconductor industry (dielectrics, semiconductors, metals, polymers, etc.). However, it requires the use of one of the multi-parameter non-linear optimization methods due to its indirect nature. This fact creates a big challenge for analysis of multilayered structures since the number of simultaneously determined model parameters, for instance, thin film thicknesses and model variables related to film optical properties, should be restricted due to parameter cross-correlations. In this paper, we use parametric sensitivity analysis to evaluate the importance of various model parameters and to suggest their optimal search ranges. In this work, the method is applied practically for analysis of a few structures with up to five-layered film stack. It demonstrates an evidence-based improvement in accuracy of multilayered thin-film thickness measurements which suggests that the proposed approach can be useful for industrial applications.