Study On The Processing Variations Of MEMS Designs

MEMS has been widely used in the fields of automobiles,smart phones,industry,aerospace,etc.With the development of Internet of Things and artificial intelligence,MEMS is showing exciting prospects.MEMS manufacturing processes are complex.The process deviations in geometry and material properties can cause great difference from the original design expectations to manufacturing results.The manufacturing of MEMS devices always repeats the "trial and error" cycle,which increases costs and the development cycle.For example,process deviations in high-performance MEMS devices,the performance would fail to meet design requirements,while for the low-cost consumer level MEMS devices,process deviations will result in low yields.Considering the effects of bias during the MEMS design phase,the costs and products development cycle can be reduced.However,current MEMS CAD tools can not satisfy these features.In this thesis,the deterministic description(in which the process deviation is determined)models and the uncertainty stochastic analysis(in which the process deviation is randomly distributed)models are established for the influence of process deviations on the MEMS device performance in their manufacturing process.The common structures such as MEMS beams and comb resonators are taken as examples to explain the influence of process variations on the device performance.The main contents of this thesis are as follows:1.Considering the most influential process effects in the DRIE process:the footing effect,the trapezium effect and the ripple effect,respectively,the resonant behavior models of the MEMS beam structures under a single and various process deviations are established.The deviations do exist in the resonant behavior models of the MEMS beam structures.The results of the verified models have been compared with Monte Carlo analysis and the finite element simulations to verify the correctness of the models,given process tolerance rates.2.Considering the process deviations in the MEMS processing as random variables,based on the series of Taylor expansion methods,the resonance behavior models of the beam structures under a single random variable have been proposed.The model of this paper can be applied when the parameters are restricted within small fluctuations(<10%).Compared with large scale of Monte Carlo sample analysis,the correctness of the models has been verified.3.The process deviations of GaAs MMIC beam structures and silicon beam structures have been analyzed throughly.When the multiple input parameters and processing disturbances are different,based on the spectral analysis methods,the structural behavior models of MEMS devices can be proposed.Comparing the results of the models with random Galerkin analysis,stochastic collocation analysis as well as Monte Carlo sampling analysis,the rationality of the stochastic models has been verified.Besides,the device design parameters are no longer limited to the latitude of the analysis methods.Thus,time and space expense-saving and device performance improvement can be achieved.As a result,appropriate guidance for device designers can be proposed to enhance the efficiency of performance predictions as well as design compensations.This thesis has summarized the problems of MEMS device manufactured under the existence of process deviations.The corresponding analysis methods can be provided for the case of insufficient processing data,insufficient device behaviors,or discrete input parameters,avoiding repeated huge sample analysis or repeated experimental tests.At the same time,a software has been designed to integrate the device models under process deviations,which realizes one-button reading of the relations between the process deviations and device performance predictions.It is convenient for device designers to take the influence of process disturbances into consideration.All these results have high reference value for further MEMS designs and the manufacturing.