Assessment of the critical dimension prediction accuracy for the lumped parameter model for resist lithography

Abstract

Lithography modelling is widely used to predict the critical dimensions (CD) of patterned features after lithographic processing. A lot of full and simplified resist models are available. Previous works on full resist models have shown that the numerous model parameters are very difficult to set and often have a poor range of validity outside the dataset that have been used to generate them [Proc. SPIE 3678 (1999) 877; Proc. SPIE 4404 (2001) 99]. Simplified resist models are an alternative solution, easier to set, and they often provide a good simulation accuracy [Proc. SPIE 4691 (2002) 1266; Proc. SPIE 5040 (2003) 1536]. Among simplified models, lumped parameter model (LPM) is widely used for CD predictions. In this paper, we study the CD prediction accuracy of the LPM, using a comparison between experimental and simulated data. A systematic method is applied for LPM parameters extraction (contrast γ and effective thickness D eff). This assessment shows that a single parameter set giving reasonable accuracy is not found. Moreover, the critical analysis of the model parameters shows that these LPM parameters have a poor physical meaning. We also point out that there is a fundamental disagreement between the LPM theory and experiments.