Analysis Considering Process Deviations For Design Of MEMS Devices

Yield in manufacturing has been a growing problem with the technological development and commercialization of Micro Electro-Mechanical Systems(MEMS).There is an urgent need for design for manufacturing(DFM).In the manufacturing process of MEMS devices,random deviation of geometric dimensions caused by various physical,chemical,equipment factors will be reflected in the deviation of device performance between actual and nominal value ultimately.It could greatly shorten the development cycle of devices and cut down manufacturing cost to forecast the performance and yield of the final device during the design phase.Existing researches on process deviation of MEMS devices are mostly limited to highly simplified system models and idealized independent normal distributions.Besides,the current main research method is Monte Carlo Simulation(MCS)which has high computation load.Study of uncertainty quantification for manufacturing deviation and its impact on both performance and yield can help optimize structure of device and improve yield,which is of great significance to the development of MEMS technology.Thermal effect of material plays a vital role in all devices in MEMS field.Thermal parameters directly affect device performance.Furthermore,as electronic devices are highly integrated at present,failure to effectively dissipate heat can easily lead to reliability problems.This thesis studies in-situ testing structures and methods for thermal conductivity of polysilicon thin film.Calculated thermal conductivity is taken as the research object.The distribution of calculated thermal conductivity of testing structure under uncertainty of geometrical parameters is simulated,utilizing parametric scanning of finite element simulation tool,COMSOL Multiphysics.Simulated results verify the theoretical model.Layout according to the process flow and tape-out of the chip is successfully completed.The statistical distribution of geometric parameters of the structure and calculated values of thermal conductivity are obtained and compared with the results of Finite Element Method(FEM).The testing structure model has been verified as an appropriate and correct object of uncertainty analysis.This paper takes advantage of Principal Component Analysis(PCA)method to obtain decorrelation and dimensionality reduction of geometrical and electrical data and derive new expression to calculate thermal conductivity.Stochastic collocation method which derives from stochastic spectral method solves Polynomial Chaos Expansion(PCE)coefficient with numerical approximation,which makes it have wide range of applications.It also achieves balance between calculation load and precision and is conducive to subsequent optimization.The numerical analysis of the test structure is carried out by using stochastic collocation method combined with sparse grid sampling technique.The distribution of thermal conductivity and yield under the influence of process deviation is predicted and verified by simulated and experimental results.In addition,the key parameter in testing structure,which is the width of heat sink wire is determined based on parameter sensitivity analysis.Design with optimized heat sink wire is presented and proved to have obviously increased yield compared with the original design.The method proposed in this thesis which is also applicable to complex implicit system functions,has ability to predict performance and yield of MEMS devices during the design phase.Researchers can utilize it to optimize structural design and choose the most suitable process line,thus decreasing the cycle and cost of R&D.