Stress Stability of Poly-SiGe and Various Oxide Films in Humid Environments

This paper presents the stress stability of thin films for MEMS structural and sacrificial layers. The average residual stresses of the thin films were monitored via wafer curvature measurement over a long period of time. Poly-Si, poly-SiGe, poly-Ge and thermally growth SiO2 films are found to be stable in humid environments. Moisture makes LPCVD and TEOS-based PECVD SiO2 films more compressive over time. Multi-layer thin film stress is modeled with the same methodology used to derive the Stoney Equation [1]. INTRODUCTION Reliability specifications typically require MEMS structural layers to have long-term material stability. Poly-SiGe is a promising structural material for modular integration of MEMS with electronics, because of its relatively low deposition temperature (<450°C) [2]. SiO2 films can also be deposited at high rates at low temperatures, and are attractive for use as sacrificial layers because they can be easily etched selectively with respect to poly-SiGe. In past research, results from the analysis of wafer curvature over time appeared to indicate that poly-SiGe films experience a stress drift in humid environments [3]. This created a major challenge for the plausibility of poly-SiGe MEMS technology. The stress drift phenomenon has been investigated further in this research. Results show that the residual stress of poly-SiGe films is, in fact, stable in ambient conditions. The apparent residual stress drift of the poly-SiGe films in Ref. [3] was caused by the unstable low temperature LPCVD oxide on the backside of the wafers. EXPERIMENTAL DETAILS The average residual stresses of various thin films were determined with wafer curvature measurements before and after thin film deposition using a Tencor FLX-2320. Long term average residual stress monitoring was done with various layer stacks as shown in Fig. 1. Poly-Si, polySiGe, and poly-Ge, as well as various oxides, were deposited and removed from single crystal silicon (SCS) wafers under different conditions as summarized in Table I. Initial wafer curvature measurements were taken from the bare Si wafer for the single layer stacks (Fig. 1a & b), and from the oxidized wafer before poly-Si, poly-SiGe or poly-Ge deposition for the bi-layer stacks (Fig. 1c). Fig. 1 Layer stacks for stress monitoring c) Poly-Si, SiGe or Ge on oxide a) Poly-Si, SiGe or Ge on silicon b) Oxide on silicon Si Substrate Si Substrate Si Substrate Poly-Si, poly-SiGe or poly-Ge Various Oxides Table I. Deposition and removal conditions of the various thin films. Film Deposition Method Removal Method Poly-Si (0.6 μm) LPCVD @ 620 °C RIE @ 60 °C Poly-SiGe (0.2-1 μm) LPCVD @ 400-450 °C RIE @ 60 °C Poly-Ge (0.4 μm) LPCVD @ 350 °C RIE @ 60 °C Dry Thermal Oxide (1200 A) Thermally growth @ 1050 °C HF solution @ 21 °C Wet Thermal Oxide (1600 A) Thermally growth @ 1050 °C HF solution @ 21 °C LPCVD Oxide (2 μm) LPCVD @ 450 °C HF solution @ 21 °C PECVD Oxide (0.5 μm) PECVD @ 390 °C Single side deposition RESULTS and DISCUSSIONS Wafers used in Ref. [3] to monitor the stress stability had a poly-SiGe film deposited on top of a 2 μm LPCVD oxide, as shown in Fig. 1c. This layer stack is commonly used in MEMS: the thick oxide serves as a sacrificial layer and the poly-SiGe serves as the structural layer. Under further investigation, the results reported in Ref. [3] have been reproduced in this work, as plotted in Fig. 2. The measured stresses of poly-SiGe and poly-Ge on LPCVD oxide become more tensile over time, but all poly-Si, poly-SiGe and poly-Ge films on thermal oxide or SCS are stable. These results indicate a problem with the LPCVD oxide. It should also be noted that poly-Si films on LPCVD oxide are more stable than poly-SiGe and poly-Ge films on the same oxide. This is because, during the poly-Si deposition, the LPCVD oxide is annealed at 620 °C. Further experimentation was done with the poly-SiGe on LPCVD oxide wafers (Fig. 3). If the backside poly-SiGe films of two similar wafers are removed at different times, the drift profiles and absolute stresses of the wafers are nearly identical, with an offset in the x-axis. When the backside poly-SiGe and LPCVD oxide films are both removed, the stresses of the topside poly-SiGe and LPCVD oxide films become stable. This data suggests that the stress drift reported in Ref. [3] is due solely to the instability of the LPCVD oxide film exposed to the ambient on the backside of the wafer. 0 10 20 30 40 50 60 70 -200 0 200 400 600 80