Very high-cycle fatigue failure in micron-scale polycrystalline silicon films: Effects of environment and surface oxide thickness

Abstract

Fatigue failure in micron-scale polycrystalline silicon structural films, a phenomenon that is not observed in bulk silicon, can severely impact the durability and reliability of microelectromechanical system devices. Despite several studies on the very high-cycle fatigue behavior of these films (up to 1012cycles), there is still an on-going debate on the precise mechanisms involved. We show here that for devices fabricated in the multiuser microelectromechanical system process (MUMPs) foundry and Sandia Ultra-planar, Multi-level MEMS Technology (SUMMiT V™) process and tested under equi-tension/compression loading at ∼40kHz in different environments, stress-lifetime data exhibit similar trends in fatigue behavior in ambient room air, shorter lifetimes in higher relative humidity environments, and no fatigue failure at all in high vacuum. The transmission electron microscopy of the surface oxides in the test samples shows a four- to sixfold thickening of the surface oxide at stress concentrations after fatigue failure, but no thickening after overload fracture in air or after fatigue cycling in vacuo. We find that such oxide thickening and premature fatigue failure (in air) occur in devices with initial oxide thicknesses of ∼4nm (SUMMiT V™) as well as in devices with much thicker initial oxides ∼20nm (MUMPs). Such results are interpreted and explained by a reaction-layer fatigue mechanism. Specifically, moisture-assisted subcritical cracking within a cyclic stress-assisted thickened oxide layer occurs until the crack reaches a critical size to cause catastrophic failure of the entire device. The entirety of the evidence presented here strongly indicates that the reaction-layer fatigue mechanism is the governing mechanism for fatigue failure in micron-scale polycrystalline silicon thin films.

Mechanisms for Fatigue of Micron‐Scale Silicon Structural Films

“…[11] Another recent study, by Alsem et al [48,49] , confirmed, using HVTEM, the presence of this cyclic stress-assisted oxidation on resonator specimens similar to the ones used by Muhlstein et al (i.e., MUMPs fabricated 2 µm thick polysilicon [44,45] ). Enhanced oxide thicknesses were imaged at the locations of maximum stresses after fatigue cycling, but not in samples that had been monotonically loaded and failed by overload fracture; such reaction layers were also not found after cycling in high vacuum (2 × 10 -5 Pa) ( Figure 14).…”

“…First and foremost, the absence of thin-film silicon fatigue failures in high vacuum, and the observed influence of humidity on the fatigue life are strong indications of a significant environmental contribution to cracking ( Figure 5) [19,20,30,31,48,49] . However, observations that the number of cycles to failure is frequency-independent are clear indications of a true fatigue contribution too ( Figure 18) [26,27,55] .…”

“…Based on a wide spectrum of studies (listed in Tables 1 and 2) involving a range of different testing methods (described in the Appendix), stress-lifetime data (Figures 1, 10) for thin-film silicon show relatively consistent trends, specifically that stress amplitudes as low as half the (single-cycle) fracture stress can cause delayed fatigue failure, typically after 10 11 cycles or more [11,24,31,33,48] . Such fatigue failure has been reported for various modes of cyclic loading, specifically for fully reversed cyclic loading (R = -1) [4,[11][12][13][14][15]18,19,24,[30][31][32][33][34][35][41][42][43]48,49,[52][53][54]59] and tensile loading with a positive mean stress (R ≥ 0) [13,[21][22][23][25][26][27][28][29][33][34][35]40,[55]…”

“…Similar to the effects of coating silicon films with monolayer barrier coatings [11,42] samples subjected to applied stresses up to 3.45 GPa survived lifetimes of more than 10 10 cycles without failure ( Figure 15) [48] . They also showed that samples tested in high relative humidity (>95% RH) air failed after fewer numbers of cycles than corresponding samples tested in ambient air (~35% RH); indeed, in the moist environment, failures were seen at ~10 9 cycles at applied stresses as low as 2.4 GPa (Figure 15) [49] .…”

“…One puzzling aspect of the these studies was the thick post-release oxide layers found on the MUMPs [44,45] fabricated fatigue samples studied by Muhlstein, Ritchie, Alsem and coworkers [11,41,42,48,49] ; whereas native oxides on the order of a few nanometers thick are expected for polysilicon films, the initial oxides on the MUMPs samples were 20 nm or more.…”