Approach to critical dimension error budget analysis and specification estimation by the Monte Carlo method

Abstract

A systematic approach to determine the specifications for process latitude in a 80 nm node device is suggested based on critical dimension (CD) error budget analysis using the Monte Carlo method. Using an aerial image simulation with the Gaussian convolution model, CD error budget analysis is performed on line/space (L/S), island, and contact patterns in a 90 nm node device. The portions contributed by mask uniformity are 60%, 66%, and 50% for the L/S, island, and contact patterns, respectively, and they occupy a dominant part in comparison with ones from other factors. The contact pattern occupies 12% for focus variation, which shows the larger portion than the analysis result for L/S and island patterns. The contribution of the 1 nm grid size in the mask layout is approximately 10% of the in-field uniformity. The residual terms which include variations from aberration, inaccurate measurement, postexposure baking, development, line-edge roughness, etc. occupy 21% for L/S and contact patterns, and 16% for island patterns. Based on CD error budget analysis, specifications in a 80 nm node device are calculated. The portion contributed by each process variation in the 80 nm node device is assumed to be approximately equal to those in the 90 nm node device. The specifications for mask uniformity are 6.8 and 5.8 nm for L/S and island patterns, respectively, while they are 2.9 nm for the contact pattern. The small margin for the contact pattern is due to a large mask error enhancement factor. The contact pattern shows a drastic decrease in flare, focus margin, illumination uniformity, and mask transmittance. In the 80 nm node device, the grid size of the mask layout to generate the patterns should be reduced to less than 0.3 nm.