The fabrication process of microdevices is presented in the previous section. In order to better understand the process, there is a requirement to study the properties of the material. For example, a thin film in a surface micromachined device has to satisfy a large number of rigorous chemical, mechanical as well as electrical properties. The study can enable us to know whether or not the materials to be used are chemically, mechanically and electrically suitable. For instance, in a particular fabrication, the features such as (i) good adhesion but low in residual stress and pinhole density and (ii) good mechanical strength and chemical resistance, may be desirable. Microdevices are not manufactured for load bearing purpose, hence mechanical strength is sometimes less important. Material properties depend mainly on the deposition process and the growth conditions. Some properties depend on post-deposition thermal processing, such as annealing and surface orientation. The properties of materials change significantly from its bulk state to thin film or microstructure state (e.g., cantilever). For example, thin films display smaller grain size than bulk ones. An important reason for these differences comes from the fact that the thin films exhibit a higher surface-to-volume ratio than large chunks of material. This signifies that the thin film is strongly influenced by its surface properties (Madou 1997). With regard to physical and mechanical characteristics, thin films are also not intended for load-bearing applications. Understanding of mechanical properties is essential for improving the reliability and lifetime of thin films as well as microstructures (Vinci and Braveman 1995). The most influential considerations for surface micromachining are the understanding of stress and stickiness properties. However, others are equally important. Some of the important properties of thin film are described in the following subsections. We preferred to focus on thin films, as the manufacturing of these structures is especially crucial.
Adhesion is defined as the capability of bonding between two surfaces of the same or different materials. It depends on a number of factors including atomic structure, surface roughness and thermal condition. It is one of the most important properties of silicon micromachining materials. There are some other mechanical forces (axial and longitudinal) involved, but the effect of adhesion is more crucial in micromachining. A device may get damaged if its film is lifted from the ground substrate by a repetitive external mechanical force. Adhesion of various films and substrates is equally important as compared to the overall performance and reliability of microdevices. There are various tests done for this. Important adhesion tests include the scotch tape test, abrasion method, scratching, ultrasonic and ultracentrifuge tests, bending, and pulling.
Adhesion can be significantly improved by cleaning the substrate and using a rough bonding surface (Campbell 1970). The latter improves mechanical interlocking. Another technique to improve adhesion is by increasing the adsorption energy of the deposit and/or increasing the number of nucleation sites in the early phase of the film growth. The value of the sticking energy between the film and substrate ranges from 10-20 kcal/mole for physical absorption and chemical absorption, respectively, and the weakest form of adhesion involves van der Waals forces. van der Waals forces are weak physical forces that hold two molecules or two different parts of the same molecule together.
Understanding stress in thin films is the most important consideration in micro structure fabrication. Stress is developed during crystallisation and deposition. Development of stress is risky for the longevity of the component and can cause malfunctioning during normal operations. Stress on a thin film can be analysed qualitatively as well as quantitatively.
Qualitative descriptions of thin film are reflected through cracks, de-lamination and voids. Most the thin films show a state of residual stress. This is mainly due to mismatch in the coefficient of thermal expansion and non-uniform plastic deformation. Other main factors, which affect the thin films, are lattice mismatch, interstitial impurities and growth processes. The stress causing factors can be intrinsic and/or extrinsic. Intrinsic stresses, called growth stresses, are developed during film nucleation. Extrinsic stresses are introduced by external factors such as temperature gradient and packaging. Thermal stress is the most common type of extrinsic stress. It arises either in a structure (with non-homogeneous coefficients) subjected to a uniform temperature change or in a homogeneous material exposed to a thermal gradient (Krulevitch 1994). Intrinsic stresses in thin films are always larger than extrinsic stresses. They are developed due to nonuniform deposition of thin films. An example of a film deposition method is the CVD process. In a typical CVD process the atoms deposited at the beginning of the process occupy lower energy configurations than the latter ones. If the deposition rate is too high or low, atomic surface mobility gets disturbed thereby developing an intrinsic stress. Additionally, extrinsic stresses occur when parts of a material undergo a volume change during a phase transformation. Moreover, misfit stresses arise in epitaxial film due to lattice mismatches between the film and substrate. Interstitial and substitution impurities impose intrinsic residual stress. Local expansion or contraction is associated with point defects. Intrinsic stress in a thin film does not create delaminating unless the film is thick enough. High stress can result in the buckling or cracking of films.
The quantity of thin film stress can be calculated from the basic stress and strain formula. The sum of the stresses is equal to the external applied stress (cextn), an unintended external thermal stress (ffh), and the intrinsic stress components (cint). The total stress in a thin film is given by:
The three non-vanishing stress components are functions of x and y directions of the Cartesian co-ordinate plane. There is no stress in the direction normal to the substrate. In case the principal axes of the substrate coincide with the x, y axes of the plane, the shear stress xxy vanishes (Chou and Pagano 1967). With constant stress through the film thickness, the stress components can be written as:
Eq.6.2 can be reduced to the following strain-stress relationships (Seidel 1990):
In the isotropic case e = sx = sy, so that ax = ay = a, or:
Where, E is Young's modulus and u is the Poisson's ratio of the film. The stress described in Eq.6.4 is called the biaxial modulus. Testing of uniaxial stress of thin films is difficult. The biaxial modulus, rather than Young's uniaxial modulus, is quite useful. Plane stress cannot be avoided. Employing appropriate design techniques can reduce the plane stress. Elwenspoek and Wiegerink have reported that a buckled microbridge clamped at both the ends has a critical length Lcr = 2x^1 EI / sA , where, I, A and s are the moment of inertia, cross sectional area and strain of the beam, respectively. If the microbridge is designed with the shortest length L<Lcr, then the effect of thin film stress can be reduced (Elwenspoek and Wiegerink 2000). Tensile stress can render the surface concave and compressive stress renders the surface convex.
Thin film stress can bend its base substrate by a comparable amount. The most common method for measuring the stress in a microstructure (e.g., thin film) is based on the substrate bending principle. The deformation of a thin substrate due to stress can be measured by observing the displacement at the centre of a circular disk or by using a thin cantilever beam. The radius of the curvature of the beam deflection is measured in order to calibrate the stress.
Recently, more sophisticated stress measurement systems are available and they are based upon analytical tools such as X-ray (Wong 1978), image processing, acoustics measurement, Raman spectroscopy (Nishioka et al. 1985), infrared spectroscopy (Marco et al. 1991), and electron-diffraction techniques. The relationship between the measured force, displacement and differential deformation must be modeled accurately with suitable assumptions. The deflections of suspended and pressurised micromachined membranes can be also measured by a mechanical probe (Jaccodine and Schlegel 1966) like a laser (Bromley et al. 1983), or a microscope (Allen et al. 1987). In the following section, the disk and cantilever methods of stress-measuring techniques are described.
When the thin film under stress deforms in its vertical direction the deformation does not introduce stress in the device, as it is considered to be normal movement. Thus, only stresses in the x and y directions need to be measured. A change in the radius of the curvature of the substrate is determined. Some optical or capacitive gauges can measure such deflections. The disk method is based on a measurement of the deflection at the centre of the disk before and after the application of force or load. Any change in the wafer shape can also reflect stress in the deposited film. It is relatively straightforward to calculate the stress by measuring the radius of the curvature. Hoffman (Hoffman 1976) has developed the relationship between the radius of curvature and thin film stress of the substrate as follows:
Where, R represents the measured radius of the curvature of the bent substrate, E/ (1-u) is the biaxial modulus of the substrate, T is the thickness of the substrate, and t is the thickness of the film. The following assumptions are made:
• The film thickness is uniform
• The disc substrate must be thin
• The film has transverse isotropic elastic properties with respect to normal direction
• The thickness of the component is much less than the substrate thickness (For most films on a silicon substrate it is assumed that t < T, t/T measures ~10_3)
• The measurement is carried out at a uniform temperature
• Mechanical isolation between the substrate and the component exists
• Stress is biaxial and homogeneous over the entire substrate
• Film stress is constant through the thickness
Thin film residual stress can be also measured by a cantilever spiral. There are various cantilever spirals available in the literature but Fan's (Fan et al. 1990) cantilever spiral is famous in stress measurements. The spiral is anchored to the inside spring upwards and can make flexible rotations. This arrangement determines the positive strain. The arrangement in which the spiral is anchored to the outside can measure a negative strain. The positive and the negative strains produce mirror symmetry on the spiral. The strain gradient can be calculated from spiral structures by measuring the following:
• The amount of lateral contraction
• The change in height
• The amount of rotation
Stiction is defined as the amount of force needed to start to move an element. There is a great deal of literature on the theoretical analysis of stiction. However, its effect on microstructures is described briefly (Lober and Howe 1988; Guckel et al. 1987; Alley et al. 1988).
The deposition of a sacrificial layer simplifies the development of movable microstructures. An important limitation of the microstructures is that they tend to deflect because of stress. It could remain attached to the substrate or isolation layer during the rinsing and drying process. This phenomenon is called stiction and it is mostly related to the effect of bonding and residual contamination. Recently, extensive study has been made to overcome stiction in microstructures (Mastrangelo et al. 1993a, 1993b). If the sacrificial layer is removed with buffered oxide enchants and rinsed in de-ionised water for long time and finally dried under an infrared lamp, stiction can be considerably prevented. One can note that as the water dries, the surface tension of the rinse water pulls the microstructure toward the substrate and a combination of forces such as van der Waals forces and bonding keeps the structure firmly attached. It is difficult to release stiction without a mechanical force, which is large enough to damage the structure. Stiction remains as a reliability issue due to contact with adjacent surfaces even after release. Stiction-free passivation can survive during packaging (Howe 1995). Making the structure stiff enough can solve the sticking problem. According to Legtenberg, if the microstructure device is designed with the critical length in Eq.6.7 then stiction can be avoided (Elwenspoek and Weigerink 2000).
Where d and h are dimensional parameters, y is the surface tension of the rinsing liquid and 9t and 02 are the contact angles of the liquid with the substrate. The microstructure will collapse beyond this critical length. An alternative is to increase the roughness of the surface in the microstructure because the end sticking is just a wafer bonding process (see below). Hence, the device should be designed such that its critical length is grater than Lcs. Another difficulty in microstructure fabrication is dimensional uncertainties, which expresses greater concern on reliability. Dimensional uncertainty exists in relatively large dimensional structures. For example, a large resonator developed using any lithography technique of quality factors of up to 100,000, the resonator frequency may vary up to 0.02 Hz over long-term (Gad-el-Hak 2002).
Was this article helpful?
Have you ever been envious of people who seem to have no end of clever ideas, who are able to think quickly in any situation, or who seem to have flawless memories? Could it be that they're just born smarter or quicker than the rest of us? Or are there some secrets that they might know that we don't?