Viewpoint Paper Surface oxide effects on failure of polysilicon MEMS after cyclic and monotonic loading H. Kahn, a, * A. Avishai, a R. Ballarini b and A.H. Heuer a a Department of Materials Science and Engineering, Case Western Reserve University, 10900 Euclid Ave, Cleveland, OH 44106, United States b Department of Civil Engineering, University of Minnesota, Minneapolis, MN 55455, United States Received 16 November 2007; accepted 14 December 2007 Available online 14 January 2008 Abstract—Polycrystalline silicon (polysilicon) microelectromechanical systems (MEMS) devices subjected to constant tensile stres- ses do not display delayed fracture in humid ambients unless they also contain thick (>45 nm) surface oxide layers, which are then susceptible to moisture-assisted stress corrosion. Polysilicon MEMS devices with typical (3 nm thick) native oxides do not show any thickening of the surface oxide layer after 3  10 7 fatigue cycles, excluding stress corrosion of the surface oxide as a cause of fatigue failure. Possible origins of polysilicon fatigue are discussed. Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Microelectromechanical systems (MEMS); Fatigue fracture; High-resolution electron microscopy (HREM) The existence of fatigue in polycrystalline silicon (polysilicon) microelectromechanical systems (MEMS) devices has been well established, but its origin has been a matter of debate. As described in two recent reviews [1,2], the two main explanations involve mechanical and environmental effects. Supporting the mechanical damage mechanism are experimental results that show: (i) a strong dependence of fatigue strength on the load ratio R (the ratio of the lowest stress to highest stress in the fatigue cycle), but no dependence on the testing environment using either air or low vacuum (10 Pa) [3]; (ii) no dependence on the frequency of the cyclic load – in other words, a dependence on the number of stress cycles but not on time [4]; (iii) an actual increase in the strength of the devices after certain cyclic stress conditions [5,6]; (iv) no extension of sharp pre-cracks held under constant high tension in humid air environ- ments [3]; and (v) observed polysilicon fatigue in vac- uum (though at much longer times than for air tests) [6]. Also, recent results for bulk single crystal silicon demonstrate decreased strength after cyclic compressive loading but not after constant loading, which is attrib- uted to mechanical damage accumulation [7]. The main drawback to the mechanical explanation for polysilicon fatigue is that no specific physical mech- anisms have been suggested that can explain all the ob- served fatigue behavior. Recent molecular dynamics simulations of nanocrystalline silicon have demon- strated that plasticity can occur in grain boundary re- gions [8]. Incorporating this effect into a finite element model of polysilicon led to a proposed strengthening model [5] in which cyclic loads generate residual com- pressive stresses at the notch root, which in turn result in an apparent increase in tensile strength upon subse- quent loading. The molecular dynamics simulations [8] also suggest that if the applied stresses are high enough, microcracks and voids could develop that would de- crease strength. This has not yet been incorporated into a finite element model with applied cyclic stresses. How- ever, this analysis cannot explain fatigue observed for single crystal silicon devices [9]. The environmental explanation for polysilicon fati- gue describes a degradation process involving subcritical cracking within the thickening silicon dioxide layer on the surface of the devices. This cyclic stress-induced thickening of this oxide reaction layer leads to continued crack growth until a critical length is reached (‘‘reaction layer fatigue” [10]). Stress corrosion cracking of bulk sil- icon dioxide due to humidity is well documented. The strongest evidence in support of this mechanism is the 1359-6462/$ - see front matter Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.scriptamat.2007.12.025 * Corresponding author. Tel.: +1 216 368 6384; e-mail: kahn@case.edu Available online at www.sciencedirect.com Scripta Materialia 59 (2008) 912–915 www.elsevier.com/locate/scriptamat