Reticle side wall clamping
Optical lithography has been the primary manufacturing technology for the integrated circuit or microchip for the past decades. In a chip, up to 40 layers of different materials are stacked to create billions of transistors. To create a layer, the critical step is to illuminate the pattern on a photomask or reticle onto a silicon wafer. The image on this mask is exposed on the wafer up to hundreds of times with a demagnification of a factor four. The current state of the art exposure tools are step-and-scan systems, use Deep Ultra-Violet light with a wavelength of 193 nm and a numerical aperture of 1.35. Most chip manufacturers have already navigated beyond the lithographic printing limits of these systems at around 40 nm by turning to double patterning techniques, where two exposure steps are used to produce a single layer of a chip. It is essential to keep production costs down by e.g. increasing wafer throughput. This can be acquired by increasing scanning speed and acceleration of the positioning stages. Meanwhile, the allowed alignment error of subsequent layers, called overlay error budget is decreasing with an accelerated rate. Therefore, contributors to the overlay error budget due to reticle slip, reticle clamping should decrease as well. This thesis addresses both the elimination of reticle slip as well as the reduction of non-correctable reticle deformation, while meeting clamping stiffness requirements.
The reticle clamping concept investigated in this thesis allows for scanning accelerations of 400 m/s2. Reticle slip is eliminated entirely, by avoiding acceleration force transfer by means of friction. Instead of the conventional way of clamping the reticle on the lower area only, reticle side wall clamping was investigated. This reticle clamping concept was designed and realized, employing struts to kinematically constrain the reticle, minimizing the non-correctable reticle deformation. The struts clamping on the reticle side walls cope with the reticle side wall skewness and squareness tolerances by means of an aerostatic spherical joint at the head of the strut, using a ball diameter of 7 mm. During alignment toward the side walls, compressed air with a pressure of 3 bar is used to lift the ball out of the socket. The aerostatic spherical joint performance was proven with an air film height of about 5 micrometer at an air film preload of 4.5 N, with an air film stiffness of 340 N/mm. In all struts, consistent performance of the spherical joint was achieved. The strut is made of Tungsten Carbide for high stiffness and includes elastic hinges to allow for differences in thermal expansion between the reticle and the chuck. Three contact pads are present on the flat side of the ball section of the spherical joint. A strut has a vacuum cup of 1 mm height, installed onto this ball section. This allows for a vacuum preload of the reticle contact of about 4 N. The compressed air required for the spherical joint and vacuum for the vacuum cup are provided from the distal end of the strut. The strut is modular and fits in a hole of no larger than 14.6 mm.
The clamping concept was validated in a test setup where the reticle was clamped in six degrees of freedom. The stiffness of the reticle clamp was validated in experiments and was shown to be 1.5*107 N/m in x-direction and 3.1*107 N/m in y-direction. Furthermore, the reticle clamping induced out-of-plane deformation was measured in the same test setup. A Twyman-Green type laser interferometer was used to measure the entire pattern area. This measurement was used to calculate the clamping induced in-plane deformation of the pattern. A reticle clamping induced in-plane error of 0.02 nm was the result, clearly showing the high potential of the reticle side wall clamping concept.