The Berkeley MET5, funded by EUREKA, is the world’s highest-resolution EUV projection lithography tool. With a 0.5-numerical aperture (NA) Schwartzchild objective, the Berkeley MET5 is capable of delivering 8-nm resolution for dense line/space patterns. In order to achieve this resolution, optical aberrations must be accurately characterized and compensated, a task that is complicated by the difficulty in finding a bright, high quality reference wave, and nonlinear effects associated with high incident angles on interferometry targets. The Berkeley MET5 was designed with an in-situ lateral shearing interferometer (LSI) to provide real-time wavefront diagnostics alongside its imaging capabilities.
The geometry of the MET5 makes it a particularly difficult optical system to measure interferometrically. Unlike EUV production tools, the 2-bounce Schwartzchild design is non-telecentric at the image, with an image plane whose normal vector is tilted 1.12 degrees with respect to the optical axis. Shearing interferometers have shown good results measuring EUV wavefronts at low to medium NAs (0.1 - 0.33) with telecentric geometry. However, to accommodate the MET5 geometry, a generalized model of LSI was developed to inform the design and build of a lateral shearing interferometer capable of operating at high-NA and with a tilted image plane. This model predicts non-negligible systematic errors that must be compensated in the analysis.
Specialized pinhole arrays were patterned onto the mask to fill the pupil with spatially filtered light that is incoherently multiplexed from multiple apertures. Due to the relatively large amount of DC flare compared with the signal in the interferograms, illumination profiles were chosen to match the NA of the obscuration so that zero-order light coming through the mask absorber is blocked in the pupil, which results in a finite coherence function width. Because of this, the design of the arrays required balancing the efficiency of the pattern while maintaining enough separation between apertures to accommodate the coherence function width.
Analysis of the interferometric data shows a total RMS wavefront error of 0.6 nm after removal of systematic errors predicted by the LSI model. The bulk of this error lies in astigmatism and coma terms which can be corrected by field position and small adjustments to the alignment of the Schwartzchild optic respectively. The aberration signature of this wavefront is in good agreement with preliminary print data of aberration targets according to aerial image modeling of these features.
The interferometric capability of the Berkeley MET5 is an indispensable part of commissioning the tool, and will allow for the diagnosing and monitoring of tool performance as it begins user operations in the coming months.