AVS 52nd International Symposium
    MEMS and NEMS Monday Sessions
       Session MN-MoM

Paper MN-MoM5
Quantitative Work-of-Adhesion Values for use as an In-Fab Monitor of Stiction

Monday, October 31, 2005, 9:40 am, Room 207

Session: Processing & Characterization of Materials for MEMS & NEMS
Presenter: N.A. Burnham, Worcester Polytechnic Institute
Authors: E.J. Thoreson, Worcester Polytechnic Institute
J. Martin, Analog Devices, Inc.
N.A. Burnham, Worcester Polytechnic Institute
Correspondent: Click to Email
The Atomic Force Microscope (AFM), a common tool in the fab, can measure the work of adhesion between AFM tips and MEMS surfaces, which could become an eventual predictor of device stiction and failure. The goal of our study was to ensure reliable and quantitative values for the work of adhesion, i.e., the adhesive ("pull-off") force normalized for tip radius. Seventeen tips of four different types were used, with radii from 200 nm to 60µm, covering the range of typical MEMS contacts. The samples were unpatterned amorphous silicon dioxide MEMS die with two types of surface conditions (untreated and treated with a few angstroms of vapor deposited diphenylsiloxane). The cantilever's length, angle of repose, the radius and height of the tip, and the surface roughness all contribute to the measured pull-off force and work of adhesion. A simple correction for the surface roughness resulted in the expected linear dependence of pull-off force on radius, but the magnitudes for the pull-off force and work of adhesion were higher than expected. Normal heat-treated AFM tips have minimal surface roughness and result in magnitudes that are more reliable. The typical relative standard deviation and current relative uncertainty for these data are 20% and 15%, respectively. In this presentation, we derive how the cantilever and tip parameters contribute to the measured work of adhesion, show how the corrected results compare with theory, and list our recommendations for using the AFM as a quantitative in-fab stiction monitor. Work-of-adhesion data can now be meaningfully compared to actual device performance.