Simulation And Deformation Analyses Of CMUT Diaphragm

Capacitive Micromachined Ultrasonic Transducer (CMUT) is an ultrasonic transducerbased on Micro Electro Mechanical System (MEMS) technology. The diaphragm of CMUTcould generate ultrasonic by vibrating or generate vibration by detecting ultrasonic, so thework of CMUT can been seen as movement of diaphragm. This paper models and simulatesthe3D models of diaphragm under two conditions, diaphragm generates ultrasonic byproducing vibration with electric force and generates vibration by response to ultrasonicpressure, by the element Solid186in software ANSYS12.0. In the meantime, in order toincrease the efficiency of solution and analysis, the2D models under the two conditions havealso been created. It can be found by contrasting the2D and3D models under the sameconditions that their deformations are very similar, so the3D models can be replaced by2Dmodels to increase their efficiency when the solution needed is not extremely accurate.The capacitance of CMUT restricts the lower limit of transducer bandwidth, and itsbandwidth decreases as capacitance increases. This paper calculates the deformation of squareand circular diaphragm with Finite Element Analysis (FEA), and then assumes that the topelectrode attached to diaphragm and the bottom electrode attached to ground plane comprisesa large number of parallel plate capacitor, so its capacitance can be calculated. This papercalculates the deformation of diaphragm in theoretical functions with three steps and gets itscapacitance by numerical integration. Firstly, its center deformation is got by FEA method,and then its curvature functions are solved by fitting method, at last, the deformation ofdiaphragm is the result of product of the two. In addition, this paper provides various methodsand expressions for solving curvature functions. Through the comparison and analysis ofcapacitance calculated by new models, existing models and FEA, it is obvious that the newmodels can be applied in larger range compared with existing models, and more accuracy formost of the size within effective range, and new models’ relative error are at around1%.However, there are still some singularities causing large error for new models due to thefitting accuracy and the fitting expressions, so the algorithm still have room for improvement.