Page 434 -
P. 434

Single-Crystal Silicon Carbide MEMS: Fabrication, Characterization, and Reliability        7-23


             If this reaction proceeds to the right, the limiting reactant would be silicon, since it is fully consumed
             by  platinum. This  proposition  is  made  more  likely, given  that  the  heat  of formation  of Pt Si
                                                                                                          2
                                                                              1
                                 1
             ( ∆H  10kcalmol ) is less than that of TaSi ( ∆H  28kcalmol ) [Andrews and Phillips, 1975].
                   f                                     2      f
               It is important to examine the oxygen concentrations in Figures 7.15(a) and 7.15(b). Both parts of the
             figure show very little migration of oxygen into the metallization. This finding implies a much more sta-
             ble contact structure than the Ti/TiN/Pt approach [Okojie et al., 1999], which was affected by significant
             oxidation of the metals in the contact. Various groups have extensively studied the oxidation kinetics of
             the silicides. Murarka (1988) concluded that the oxidation mechanisms for most silicides essentially have
             the same heats of formation as those of normal oxides. Lie (1984) and Razouk (1982) confirmed that the
             oxidation kinetics of tantalum disilicide has a parabolic rate. This result implies that it would take an
             appreciable length of time for oxygen to diffuse through the entire contact.




             7.5 Sensor Characteristics


               Generally, the design of devices that sense physical phenomena and provide electrical readouts calls for
             the interpolation of two or more kinds of mathematical relationships. In the design of high-temperature
             pressure sensors, the Equations that describe the physical phenomena (i.e., pressure and diaphragm deflec-
             tion) are interpolated with the electrical Equations that express the resulting output voltage.
               The Equations that model deflecting diaphragms are classified into two main categories. One category
             models  maximum  diaphragm  deflections  that  are  less  than  the  diaphragm  thickness  (linear  case),
             whereas the other supports diaphragms with maximum deflections greater than the diaphragm thickness
             (nonlinear case). If the maximum deflection of the membrane is less than its thickness, as occurs in appli-
             cations  of short-range  pressure  measurement, there  is  generally  a  reasonable  degree  of linearity  of
             diaphragm deflection in response to applied pressure. For larger pressure measurements where deflection
             is equal to or greater than the thickness of the diaphragm, the deflection of the diaphragm in response to
             applied pressure is no longer linear [Timoshenko and Woinowsky-Krieger, 1959]. For the device to be
             used continuously over a long period of time, the membrane must be capable of repeatedly deflecting
             under applied pressure with precision and little hysteresis. To achieve this aim, the membrane must retain
             its elastic property after it is subjected to maximum applied pressures. To that effect, there is a need to
             choose materials with an appreciable linear region on the Stress-Strain curve. For the diaphragm to retain
             its elastic integrity, the stress induced by pressure must not exceed the yield or fracture point. In essence,
             the maximum operating stress should be at a point below the yield and fracture stress limit. If the oper-
             ating stress reaches the elastic or fracture limit, there is a very strong likelihood that the diaphragm would
             lose its elasticity, become permanently deformed, and possibly fracture.
               Aresistor on a circular diaphragm arranged tangentially, with current flowing parallel to the resistor,
             will experience a longitudinal stress induced by the tangential strain component. It will also experience a
             transverse (radial) strain component (strain perpendicular to resistor length), which usually inserts a
             negative piezoresistance coefficient. On the other hand, as depicted in Figure 7.3(b), the radially oriented
             resistor, with current flowing parallel along its length, will be dominated by the stress induced by the lon-
             gitudinal strain component. The transverse effect is introduced via the tangential stress, with its corre-
             sponding negative piezoresistance coefficient. These findings have been extensively verified and used in
             silicon. The output of the sensor is strongly affected by the orientation of the resistors. Therefore, the
             resistor geometry and orientation should be such that only one strain component exists and the other is
             suppressed.
               When the maximum deflection, w, of a clamped circular plate is less than its thickness, the equation
             that describes it is expressed as [Timoshenko and Woinowsky-Krieger, 1959]:



                                                               Pa 4
                                                      w   0.89                                         (7.20)
                                                               64D



             © 2006 by Taylor & Francis Group, LLC
   429   430   431   432   433   434   435   436   437   438   439