| AVS 54th International Symposium | |
| Manufacturing Science and Technology | Thursday Sessions |
| Session MS-ThA |
| Session: | MEMS Manufacturing |
| Presenter: | L.A. Mosher, University of Maryland |
| Authors: | L.A. Mosher, University of Maryland B.C. Morgan, U.S. Army Research Laboratory R. Ghodssi, University of Maryland |
| Correspondent: | Click to Email |
We report the development of a new double-exposure gray-scale photolithography technique to batch fabricate three-dimensional structures in photoresist with improved vertical resolution. Deep reactive ion etching is used to transfer the patterned photoresist into silicon, enabling a complete three-dimensional microfabrication platform with many applications for MEMS devices, such as lenses and microengines. Gray-scale photolithography utilizes a photomask to spatially control the ultraviolet light intensity incident on a photoresist layer. This control is achieved by diffraction through sub-resolution pixels on the mask using projection photolithography. Projection optics keep only the zeroth order intensity, preventing the higher-order diffractions from reaching the wafer. The local pixel size is correlated to the transmitted intensity and therefore determines the height in photoresist after development. Vertical resolution, determined by the number of available pixels, is limited by the mask vendor. Current mask fabrication techniques allow for tens of pixel sizes, which is insufficient for some MEMS applications. We introduce a double-exposure gray-scale technique, utilizing two gray-scale exposures prior to a single development. In this technique, two pixel sizes are used with two partial exposures, resulting in a third unique photoresist height after development. By using all pixel combinations, we achieved an exponential increase in the number of available levels. Our initial test design utilized only eight pixel sizes, but realized 64 unique height levels in photoresist after development. We created an empirical model to correlate the pixel combination and exposure times with the photoresist height, which facilitates the design of pixel layouts for nearly arbitrary geometries. This model was used to optimize the exposure times to minimize the average and maximum vertical step height. We fabricated a seventeen-pixel test structure based on this model and observed a significant improvement over single-exposure. A decrease in the average vertical step height from 0.19 µm to 0.03 µm was achieved as well as a decrease in the maximum vertical step height from 0.63 µm to 0.24 µm. Detailed modeling parameters and experimental results from double-exposure gray-scale structures will be presented.