Study On Lifetime Model Of Silicon Microstructure Based On Cyclic Stress Enhanced Oxidation Fatigue

Silicon microstructure is the key structural component of Micro Electromechanical System(MEMS).The fatigue properties of the silicon microstructure plays a significant role in the reliability and safety of MEMS application.Thus,it is important to develop an accurate and effective fatigue lifetime model for silicon microstructure.Currently,there are mainly two types of reported methods for lifetime model of silicon microstructure,namely data-driven statistical modeling based on experimental data analysis and physical modeling based on fatigue failure mechanism of micro-damage.For the data driven methods,the main challenges include acquisition of experimental data,incorporation of the influence of environmental factors and etc.,which limits the application of the model.On the other hand,the physical model can be used to predict life effectively.However,there are still debates on the fatigue failure mechanism of silicon microstructures.Therefore,we aim to analyze the fatigue failure mechanism of silicon microstructures from the perspective of cyclic stress enhanced oxidation,and develop a fatigue lifetime model.The main contents are listed as follows:(1)We propose the approach as well as its supporting theory for developing the fatigue lifetime model of silicon microstructure.With a comprehensive analysis of the reported fatigue lifetime model of silicon microstructure,we propose an approach to develop a lifetime model from the view of fatigue failure mechanism.The mechanism of cyclic stress enhanced oxidation fatigue is adapted as the supporting principle.The detailed explanation of the mechanism is given from the view of chemical kinetics.(2)We explore the basic requirements of oxide thickening and give the functional forms.Based on the mechanism of cyclic stress enhanced oxidation,basic requirements of oxide thickening are explored.And functional forms of oxide layer thickening rate model are derived by gas-solid reaction kinetics,in which the parameters in the model is simplified according to the actual situation.(3)An approach for the parameters estimation and model selection is proposed.Given the initial conditions and unified input &output and evaluation criteria,the initial values of the parameters are optimitzed by genetic algorithm.The results of the two models are compared and discussed based on the experimental data.