FATIGUE AND WEAR IN SILICON STRUCTURAL FILMS FOR MICROELECTROMECHANICAL SYSTEMS APPLICATIONS
Last updated 03/05/08 by Daan Hein Alsem ([email protected])
To study wear mechanisms, n+-type polysilicon MEMS side-wall friction test specimens were run in ambient air with loads ~1-6mN. The devices were fabricated using the Sandia National Laboratories SUMMiT V process. Specifically, two electrostatic comb-drive actuators create motion in two orthogonal directions. A DC voltage applied to one of the actuators pulls a beam against a post; sinusoidal AC signals leading to the other, perpendicular, comb drive actuator cause the beam to wear back and forth against the post (Figure 1).
transmission electron microscope images of worn silicon beam surfaces
and wear debris were used to deduce the morphology of the wear
particles and worn surfaces, the latter prepared using focused-ion beam
lift-out and thinning sample preparation techniques. The debris
particles (Figure 2A, B), which varied in diameter from a minimum of
50-100 nm to a maximum of ~500 nm with the larger particles being
agglomeration of the smaller particles, exhibited a relatively equiaxed
morphology, unlike the more flake-like debris that is often encountered
during delamination wear. The smaller particles (50-100 nm) did not
appear to be made up of even smaller particles, as one might expect
from atomic-scale wear of the top silicon dioxide layer. Selected-area
diffraction patterns revealed diffuse rings around the forward
scattered beam indicating that the debris particles were amorphous
(Figure 2B). The worn surface of the beam also showed evidence of long
plow tracks (several mm in length), with a width of 100-400 nm
(Figure 2A). This is more than an order of magnitude larger than the
roughness of the sidewalls (~10 nm) or native oxide on these surfaces
(~3 nm), as measured by atomic force and transmission electron
microscopy respectively, but is of the same order of magnitude as the
wear debris. This suggests that debris formation is associated with
abrasive wear, caused by third-body debris particles.
FIGURE 2: Typical SEM image of worn beam after wearing (A); wear debris (~100-500 nm in size) on the surface of a worn beam after lift-off from the chip. Typical TEM bright-field image of debris (B) with accompanying diffraction pattern (inset) of debris particle agglomerates, show the particles to be amorphous. (C) TEM bright-field image and selected area diffraction (upper right) of surface layer in worn area of beam. The image shows that the top surface layer, which has been worn in the indicated direction, has a different microstructure than the beam. The diffraction pattern shows that this layer is polycrystalline.
These observations lead to three conclusions: (I) most wear occurs by abrasive removal of ~100 nm particles and does not occur by atomic-scale grinding of the oxide layer, or subsequently exposed silicon, (II) because the inherent grain size of the polysilicon in the beam is ~500 nm, the wear debris must be generated by fracture through the grain, and (III) since the debris is not flake-like, but equiaxed, most likely the first debris particles are formed by adhesive, rather than delamination wear.
Quantitative energy dispersive X-ray spectroscopy (EDX) analysis inside a TEM of debris particles revealed a silicon-oxygen atomic ratio of ~34:66 at the edges, but ~50:50 in the middle of the particles. This suggests that although the debris particles are amorphous, only the edge is stoichiometric SiO2, the dominant stochiometry for room temperature oxidation, and not the entire particle consist of SiO2. Accordingly, we can conclude that the particles oxidize heavily when they are worn off and are comprised of an amorphous SiO2 shell on an amorphous Si core.
It was also clear that sections of the worn beam exhibited a surface layer with a microstructure that was different from the rest of the beam. These surface layers varied in their in-plane thickness from ~20 nm to almost 200 nm, and displayed amorphous contrast, whereas the beam itself is populated with ~500 nm sized polycrystalline silicon grains (Figure 2C). Closer investigation revealed that this surface layer is actually nano-crystalline. Because of the sharp rings visible in the diffraction pattern, it can be concluded that there are numerous small crystals present in this relatively small volume. These nano-crystals are much smaller than the overall grain size and are even smaller than the debris particles. Control samples of areas that were not subject to wear confirmed that the microstructure inside the beam looked similar in both cases. However, these control samples showed no evidence of a surface layer with a different microstructure, as was observed in worn areas.
Quantitative EDX results shows that the layer is not fully SiO2 (Si:O 65:35 to 80:20 atomic percent). This is consistent with the fact that the layer is actually nanocrystalline, with Si as the dominant species, although with a significant degree of oxidation. These results suggest that smaller («100 nm) partly-oxidized debris particles become attached to the beam, oxidize and lead to the formation of this surface layer. These particles can originate from the initial surface asperities since the roughness is ~12 nm.
Finally, we investigated the effect of plasticity during these processes. After wear no dislocation pile-ups can be seen left in the film as well as the debris. Since plasticity in silicon below ~500ºC generally does not occur, infrared microscopy studies of the device while operating were performed and only showed a temperature rise of several K. These observations strongly indicate that plasticity does not influence the wear process at these size scales.
Comparing abrasive wear tracks, wear debris, grain size and surface roughness, it is apparent that after an initial adhesive wear mechanism, where ~100 nm particles are removed from the surface by fracture through the grains (~500 nm), an abrasive wear mechanism regime occurs and creates wear plough tracks with widths similar to the size of debris particle agglomerates.