Design and simulation of ultra high sensitive piezoresistive MEMS sensor with structured membrane for low pressure applications Piotr Mackowiak, Michael Schiffer, Xin Xu, Ernst Obermeier, Ha-Duong Ngo Technical University of Berlin, Microsensor & Actuator Technology (MAT) TIB 4/2-1, Gustav-Meyer-Allee 25,13355 Berlin, Germany
[email protected], Ph: +49-30-314522, Fax: +49-30-31472603
Abstract In order to increase the sensitivity of a piezoresistive pressure sensors, the membrane needs to be very thin or very large to achieve good results. But there is a trade-off between stability, linearity and sensitivity. The thinner the membrane the more instable is the sensor structure. The original sensor developed for wall pressure measurement has membrane thickness of 4µm [1][2]. Our idea is to use partly-structured thicker membrane to improve the sensor performance. In this paper we show the optimization of the new sensor structure by using DoE (Design of Experiment) and FEA with ANSYS software. In our work with the new membran structure the sensor sensitivity could be increased up to 300% in compare to sensor using traditional unstructured silicon membrane. Introduction The piezoresistive effect is well known and deployed for a wide range of applications such as aeromems, automotive, pharmaceuticals and many more. The sensors normally contain an electrical Wheatstone bridge, which is made up of 4 piezoresisitors that are implanted in the silicon membrane. Well designed MEMS piezoresistive sensors are stable and have a high sensitivity and good linearity. For low pressure ranges (< 10mbar) the sensor dimensions become crucial. The membrane thickness needs to be reduced to gain sensitivity but the sensor looses stability (over pressure) and linearity. To avoid the problem resulting in thinning the membrane our solution is to structure the membrane. The piezoresistors are then located in the peninsula structures so called stress concentration zones as shown in figure 1. Different stress concentration structures have been investigated to optimize the sensor structure for later realization. Design of experiment approach The approach of our solution was to define first simple structures and then using the FEA. As next step an emphasis of the used parameters was performed and the next sensor model is generated using the genetic algorithm. Various parameters, such as area, amplitude and homogeneity of stress concentration zones are involved to evaluate the advantages and disadvantages of the new sensor structure. Totally eight base forms were used - triangle, oval, trapezium, rectangle, rectangle with 20% beveling, Spline1, Spline2 and Spline3. The aim was to concentrate the mechanical stress in the areas where the piezoresistors are located, a homogeneous distribution of stress in this area and the avoidance of stress
peaks, which could lead to local deformation of the membrane. FEA results Figure 1 shows the structured membrane as the result of the DoE process of a developed piezoresistive MEMS sensor. The silicon membrane geometry is 800µm x 800µm and has thicknesses of 5µm at the peninsula structures and 3µm in the structured area. To obtain a uniform membrane thickness, SOI (Silicon-On-Isolator) wafer is needed. The device dimensions are 1600µm x 1600µmx200µm. The peninsula structures can be produced by using DRIE (Deep reactive ion etching). The cavity bellow the membrane can be produced by using standard KOH or DRIE. Details of the optimized Spline3 structure and the pieszoresistors are shown in figure 2.
Fig. 1: Part of the FEM sensor model with structured membrane using Spline3. The piezoresistors are located in stress concentration structures.
Fig. 2: Details of the Spline3 structure and piezoresistors used in this work. 2010 12th Electronics Packaging Technology Conference
Figure 2 shows the mechanical stress distributions as FEM simulation results on the eight peninsula base forms. The red colored areas in the membrane show a high stress and the blue areas a low stress. It can be clearly seen, that the Spline3 has much higher stress concentration in compare to other base forms at the same pressure level.
triangle
oval
trapezium
rectangle
Conclusion and future work It could be shown, that performance of piezoresistive MEMS sensors for low pressure ranges can be improved by using partly-structured silicon membrane. Our next steps will be the design, simulation and manufacturing of MEMS sensors with structured membrane, which has ultra high sensitivity and improved performances for low and very low pressure ranges especially in aero, automotive, pharmaceuticals and medium vacuum applications. Literature 1. “Aeromems Pressure sensor array featuring the trough-wafer vias for high-resulution wall pressure measurements”, A. Berns, January 13-17, Tucson, AZ, 2008, pp. 896–899. 2. „FEM Simulation und Entwurfsoptimierung von auf Si basierenden Drucksensoren“, X. Xu, Diploma thesis, Technical University Berlin, 2005.
Spline 1
rectangle with 20% beveling
Spline 2
Spline 3
Fig. 3: FEA results (1/8 models). Stress distribution across of the sensor membrane for the base forms (triangle, oval, trapezium, rectangle, rectangle with 20% beveling, Spline1, Spline2 and Spline3). Table 1 shows the sensitivities with different sensor membrane geometries at 100Pa pressure level. Table 1. Sensitivity comparison between different membrane geometries at input pressure of 100Pa. The supply voltage is 1V. Membrane geometry 5 µm (unstructured) 3 µm (unstructured) 5 µm structured (Spline3)
Sensitivity 312 µV/Pa 867 µV/Pa 926 µV/Pa
Comparison 100% 36% 278% 100% 297% 106,8%
The sensor with the structured membrane (spline3) shows a much higher first resonance frequency (~78kHz) than the resonance frequency of the developed sensor (~50kHz). 2010 12th Electronics Packaging Technology Conference
