Piezoresistance in silicon has been one of the major sensing principles in micro-electro-mechanical systems (MEMS). The principle is also a useful tool to characterize the induced stress in packages for convetional integrated circuits. The producers of MEMS have contiuously made smaller and more complex systems adapted to new and often harsh environments. This has led to the observation of deviations between the common theory for the piezoresistance in p-type silicon and the measured values. This deviation is linkde to temperature and doping effects. The most common model for the piezoresistance coefficients in silicon gives a dependency for both n-and p-type Silicon (Kanda.1982), while early results b Bir and Pikus (Bir and Pikus, 1974),indicate a deviation from this behavior in p-type silicon. The extent to which the deviations are due to the valid rande for the underlying theory being exceede is addressed in this project. In our study we have adapted the 4 point bending method to apply uniaxial stress to the p-type Si resistors. We minimize experimental uncertainties by using model resistors suited for the measurement. Resistors are made into SIMOX wafers homogeneously doped over a depth where the stress is simply related to geometry. Among different resistors the doping is varied over a large range and we have measured piezo-resistance coefficients in the temperature range 5 to 140 degrees Celsius. Using large beams we avoid the increasing errors associated with diminishing geometrical dimensions and errors associated with calibration of small weights etc. The beams are rotated by an angle of 22.5 degrees off the < 110 > crystal direction. This enables measurement of all three piezocoefficients with three resistors rotated 45 degrees with respect to the others.
展开▼
