skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Fatigue failure in thin-film polysilicon is due to subcriticalcracking within the oxide layer

Journal Article · · Applied Physics Letters
DOI:https://doi.org/10.1063/1.1856689· OSTI ID:877321

It has been established that microelectromechanical systems (MEMS) created from polycrystalline silicon thin-films are subject to cyclic fatigue. Prior work by the authors has suggested that although bulk silicon is not susceptible to fatigue failure in ambient air, fatigue in micron-scale silicon is a result of a ''reaction-layer'' process, whereby high stresses induce a thickening of the post-release oxide at stress concentrations such as notches, which subsequently undergoes moisture-assisted cracking. However, there exists some controversy regarding the post-release oxide thickness of the samples used in the prior study. In this Letter, we present data from devices from a more recent fabrication run that confirm our prior observations. Additionally, new data from tests in high vacuum show that these devices do not fatigue when oxidation and moisture are suppressed. Each of these observations lends credence to the '''reaction-layer'' mechanism. Recent advances in the design of microelectromechanical systems (MEMS) have increased the demand for more reliable microscale structures. Although silicon is an effective and widely used structural material at the microscale, it is very brittle. Consequently, reliability is a limiting factor for commercial and defense applications. Since the surface to volume ratio of these structural films is very large, classical models for failure modes in bulk materials cannot always be applied. For example, whereas bulk silicon is immune to cyclic fatigue failure thin micron-scale structural films of silicon appear to be highly susceptible. It is clear that at these size scales, surface effects may become dominant in controlling mechanical properties. The main reliability issues for MEMS are stiction, fatigue and wear. Fatigue is important in cases where devices are subjected to a large number of loading cycles with amplitudes below their (single-cycle) fracture stress, which may arise due to vibrations intentionally induced in the structure (i.e. a resonator) or those which arise from the service environment. While the reliability of MEMS has received extensive attention, the physical mechanisms responsible for these failure modes have yet to be conclusively determined. This is particularly true for fatigue, where the mechanisms have been subject to intense debate. Recently we have proposed that the fatigue of micron-scale polysilicon is associated with stress-induced surface oxide thickening and moisture-assisted subcritical cracking in the amorphous SiO{sub 2} oxide layer (''reaction-layer'' fatigue). The mechanism of oxide thickening is as yet unknown, but is likely related to some form of stress-assisted diffusion. Allameh et al. suggest a complementary mechanism involving stress-assisted oxide thickening, caused by dissolution of the surface oxide which forms deep grooves that are sites for crack initiation. Kahn et al. have criticized these mechanisms and proposed that, instead, fatigue is caused by subcritical cracking due to contacting surface asperities in the compressive part of the cycle. To the authors' knowledge, there is no direct experimental observation of such asperity contact. Also, their model cannot explain why micron-scale silicon, and not bulk silicon, is susceptible to fatigue. Moreover, Kahn et al. do not acknowledge the role of stress-induced oxide thickening, which has been observed directly using TEM and indirectly using atomic-force microscope measurements by several investigators, and have questioned whether the materials utilized by Muhlstein et al. and Allameh et al. were representative due to the relatively thick oxide scales. Accordingly, the goal of the present research is to seek a definitive understanding of the physical mechanisms responsible for fatigue in polysilicon structural thin-films. Our approach is to combine on-chip testing methods with electron microscopy by fatiguing thin-film samples and observing them, in an unthinned condition, using high-voltage transmission electron microscopy (HVTEM). Two principal results are found from this work: (1) fatigue tests conducted in ambient air on polysilicon samples from a more recent fabrication run confirm that the fatigue behavior and oxide-layer thicknesses observed in earlier experiments were not an artifact of that particular fabrication run, and more importantly, (2) fatigue tests run in high vacuum reveal absolutely no evidence of premature fatigue failures. We believe that both of these results add further confirmation to the ''reaction-layer'' mechanism for fatigue of micron-scale polysilicon.

Research Organization:
Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
Sponsoring Organization:
USDOE Director. Office of Science. Office of Basic EnergySciences
DOE Contract Number:
DE-AC02-05CH11231
OSTI ID:
877321
Report Number(s):
LBNL-56771; APPLAB; R&D Project: 511906; BnR: KC0201030; TRN: US200607%%375
Journal Information:
Applied Physics Letters, Vol. 86, Issue 4; Related Information: Journal Publication Date: 01/24/2005; ISSN 0003-6951
Country of Publication:
United States
Language:
English