Department of Electronics and Communication Engineering, Vellore Institute of Technology, Tamilnadu, India
Machining is defined as a process of removing material from a workpiece in the form of chips in order to obtain exact shapes and sizes. It is a well-defined process method in "machine design and tooling". It includes several methods, which are usually divided into three main classes such as cutting, abrasive processes, and nontraditional machining. Machining processes design and manufacture macro components and systems. Micromachining processes bear similar meaning but the process handles microsystems. Micromachining has become a new technology, as there has been significant growth in the manufacture of components having dimensions less than a millimeter upon which micro features in the range of 1-200 ^m are built. In particular, micromachining deals with microelectronics, micromechanical systems, micro-opto-electrical systems and micro-optomechanical systems. When these are integrated into one platform, the system is known as a MEM (Microelectromechanical) system, or MEMS. Micromachining is considered to be a process as well as a technology that is utilised to structure wafer materials or thin films in order to fabricate miniature devices such as microsensors, microactuators and passive components. The process is classified under two different process techniques: bulk micromachining and surface micromachining. If structuring is performed on wafer materials the process is called a bulk micromachining process and if it is performed on the thin film the process is called a surface micromachining process.
5.2 Wet Bulk Micromachining (WBM)
Bulk micromachining is a method of fabricating microdevices by etching the substrates and implementing the features in the bulk of the materials such as silicon, germanium, quartz, SiC, GaAs, and InP. The etching process is a special technique to cut the substrate, which can be used to form the desired shape. The process falls under three categories in terms of the state of the etchants used: wet, vapor, and plasma states. When it uses wet etchants it is called wet bulk micromachining (WBM) and the etching depends on the plane of orientation of the substrate. Orientation dependent etching is called anisotropic etching, whereas orientation independent etching is known as isotropic etching.
A vast majority of wet bulk micromachining processes are based on single crystal silicon. Some work has been done on quartz, Ge and GaAs and recently on GaP, InP and SiC. WBM is understood to be a well-established fabrication process in silicon technology. Among other techniques, surface micromachining can be named and it has been popular in recent years. The IC (integrated Circuit) industry has been using WBM for a long time, but further development has been made to adopt it into the MEMS domain. Although the technique is emerging fast, it is unlikely that the popularity of bulk micromachining will be decreased. This chapter deals with the features and properties of the etching process of the bulk micromachining method. Readers can follow the included references for further details. The next chapter deals with the important aspects of the surface micromachining process. Silicon is used as the substrate throughout the chapter.
Wet etching using a mask (e.g., wax) and etchants (e.g., acid) has been found in the fifteenth century for decorating armors (Harris 1976). By the early seventeenth century, the etching process had become an established technique. Initially it was known as chemical milling. Introduction of photosensitive masks were found in 1822 but the accuracy was less then 0.05 mm (Madou and Morrison 1989). Later on, lithography and chemical milling processes were combined to achieve a new level of accuracy. PCB is the best example of an application of lithography-based chemical milling. Silicon based ICs had been fabricated by 1961. Later, many other applications like color television shadow masks, integrated circuit lead frames, light chopper and encoder discs, and decorative objects such as costume jewelry have been found that use photochemical machining (Allen 1986).
Silicon processing using isotropic etching has been available since the early 1950s. Some details of the work are available from a series of papers published by Robbins and Schwartz during 1959 to 1976, on chemical isotropic etching (Robbins and Schwartz 1960; Schwartz and Robbins 1976). They have observed the use of the chemical isotropic etchant HF and HNO3 with or without acetic acid and water for the silicon process. Uhlir's work is based on electrochemical isotropic etching (Uhlir 1956). Isotropic etching was first applied on electrochemical cells or for electro-polishing, but the deposition of metal on the surface created a problem (Hallas 1971). Turner has found that if the critical current density is exceeded, silicon can be electro-polished in aqueous HF solutions without formation of any metal deposition (Turner 1958).
In the mid sixties, Bell Telephone Laboratories started the work on anisotropic silicon etching in the mixtures of KOH, water and alcohol and later in KOH and water. Various scientists also pursued chemical and electrochemical anisotropic etching methods during 1966 to 1977. In the mid-seventies, a new surge of activity in anisotropic etching was associated with the work on V-groove and U-groove transistors (Rodgers et al. 1977; Ammar and Rodgers 1980). Smith discovered piezoresistance using Si and Ge in 1954 (Smith 1954). Pfann et al. proposed a diffusion technique for the fabrication of Si piezoresistive sensors for stress, strain and pressure measurement (Pfann et al. 1961). Tufte et al. made the first thin Si piezoresistive diaphragms for pressure sensors using a combination of a wet isotropic etch, dry etching and oxidation processes (Tufte et al. 1962). National Semiconductor became the first to make stand-alone Si sensor products in 1972. They also broadly described Si pressure transducers in 1974 and completed a silicon pressure transducer catalogue (Editorial 1974).
Other important commercial suppliers of micromachined pressure sensor products were Foxboro/ICT, Endevco, Kulite and Honeywell's Microswitch. Other micromachined structures have been explored during the mid- to late seventies. Texas Instruments (Editorial 1977) produced a thermal print head in 1977. Hewlett Packard (O'Neill 1980) made thermally isolated diode detectors in 1980. Fiber optic alignment structures were developed at Western Electric (Boivin 1974) and IBM produced ink jet nozzle arrays (Bassous et al., 1977).
Many Silicon Valley microsensor companies played vital roles in the development of the market for Si sensor products. Druck Ltd. in the U.K. started exploiting micromachined pressure sensors in the mid-eighties (Greenwood 1984). Petersen explored the mechanical properties of a single crystalline of silicon (Petersen 1982). It is estimated that today there are more than 10,000 scientists pursuing research on the design and development of Si based microsystems. It has become a pressing need to understand the intended applications.
Silicon crystalline inherits a covalent bond and diamond-cubic (DC) structure. The atoms form a tetrahedron with its four covalent bonds as shown in Fig. 5.1. These tetrahedrons make a diamond-cubic structure. The structure can be explained as two interpenetrating face-centered cubic (FCC) lattices, one displaced (1/4, 1/4, 114) with respect to the other, as shown in Fig. 5.2. The axes x, y, z are perpendicular to a plane with the three integers (h, k, l). It simplifies the illustration concerning the crystal orientations.
The lattice parameter is defined as distance between two atoms. It is approximately 5.4309 Â for the silicon structure. The DC lattice is wide-opened, with a packing density of 34%, compared to 74% for a regular FCC lattices. The plane (111) presents the highest packing density. In addition to the DC structure, silicon has many special features like high-pressure crystalline phases and stress-induced meta-stable properties.
Fig. 5.1. Silicon diamond-cubic structure and covalent bonds
Fig. 5.1. Silicon diamond-cubic structure and covalent bonds
Fig. 5.2. Silicon plane orientation
Silicon wafer <100> <110> and <111> orientations are commonly used for micromachining. Wafer <100> can be machined in order to obtain (110) and (111) orientation planes on the wafer. The <110> wafer cleaves much easier than other orientations. The <111> wafers are not normally used because it is difficult to etch anisotropically. Additionally, the wafer entails a laser-assisted etching process. On a <100> silicon wafer, the <110> direction can be made easily. The flatness can be up to 3° precision. Flat areas of the wafer can be used for the determination of the orientation. The flat area also helps to place the slices on the equipment. It is important to understand the geometric relationships between the different planes within the silicon lattice. Only wafers <100> or <110> are considered to be ground planes. The wafer <111> has the highest atom packing density and cannot be etched using anisotropic etchants. Hence, wafers <100> or <110> are bound to form the sidewalls of a (111) plane. The micromachined planes are dependent on the geometry and the orientation of the mask.
Fig. 5.3 shows a silicon wafer <100> together with non-etching (111) planes as sidewalls. During etching, truncated pyramids or V-grooves do not widen. After etching, the (111) planes are joined to their intersection and the (100) bottom plane disappears, creating a pyramidal pit or a V-groove. The slope of the sidewall in a cross section perpendicular to the wafer surface is approximately 54.74°. The alignment of the wafer surface determines the accuracy of this angle. Due to this behavior of silicon, its mechanical properties differ from common metals. Some of the important features are discussed in the next section.
5.5 Silicon as Substrate and Structural Material
Due to its intrinsic mechanical stability, sensing, integrating capabilities and electronic properties, silicon is mostly selected as the prime material for the fabrication of electromechanical microsystems. Many thin films and other microcomponent based devices use silicon as substrates or wafers.
It is observed that the chemical sensor development in industry has moved from silicon in the 1970s and early 1980s to a hybrid thick film on ceramic approach in the late 1980s and early 1990s. However, recently, silicon has been considered to be the best choice in terms of metalisation and machinability as compared to ceramic, glass and plastic. Both ceramic and glass substrates are immune to machining processes. Plastic substrates are susceptible to metalisation. The silicon substrate is relatively costly but it can be substituted with small sizes. Some of the special features like ease of passivation layers, extreme flatness, well-established coating procedures, stability in high temperatures and formation of thin films have made the silicon substrate acceptable. Silicon has greater flexibility in design and manufacturing as compared to other substrates. Although it is much expensive, the initial capital equipment investment is not product specific, as the micromachined products require a change in mask but not in the equipment itself. Another factor that determines the choice for silicon is the final packaging of the device. For instance, a chemical sensor is easy to package on an insulating substrate rather than on a conductive substrate. Silicon has also some disadvantages. It is not worthwhile to use silicon for large devices and low production volumes, as well as for devices where the design of electronics circuits is not important.
In mechanical sensors, an active structural element converts a mechanically external input signal, i.e. force, pressure and acceleration, into an output electrical signal like voltage, current or frequency. The transfer functions can be described in terms of mechanical, electromechanical and electrical conversion. In mechanical conversion, a structural active member of the microdevice is optimally loaded or stressed. An active member can be a high-aspect-ratio suspended beam or membrane. Electromechanical conversion is the transformation of a mechanical quantity into an electrical quantity such as capacitance or resistance. Sometimes, the electrical signal is amplified and further converted to voltage, current or frequency. For conversion into voltage, a Wheatstone bridge may be utilised. A charge amplifier may be provided for amplification.
To optimise the transfer functions, system modeling is necessary. One of the most important features of modeling is the method of determination of independent elasticity constants. It follows that silicon has the requisite property of elasticity constants. In the case of actuators, the same transfer function can be used in reverse way, but the mechanical and electrical properties remain same. Important mechanical properties such as stress, strain and plastic deformation and thermal properties of the silicon materials are described below.
Stress is defined as amount of force acting in a unit area while strain is the elongation in unit length. Another parameter which is used to define the elastic property is known as the Poisson ratio. It is defined as the amount of lateral contraction in unit diameter. Yield, tensile strength, hardness and creep of a material are all related to the elasticity properties. The stress-strain relationship curve is illustrated in Fig. 5.4. It varies from material to material, but the nature of the curve for most materials is similar. The curve follows a linear relationship in certain regions and then it becomes nonlinear. The linear part draws interest as far as the modeling of micromechanical devices is concerned. The stress is proportional to the strain of a material with a constant slope in this region. This is known as Hooke's law and the slope is defined as elastic modulus E or Young's modulus. The region marked P as in Fig. 5.4 (b), is called the elastic deformation. For isotropic media (e.g., amorphous and polycrystalline materials) the applied axial force per unit area or tensile stress aa, and the axial or tensile strain ea, are related by (Seidel et. al. 2990):
Where, sa is a dimensionless ratio of (L2 - L1)/L1, i.e., the ratio of the elongation to its original length. The elastic modulus can be considered as the resistance of the materials to deformation. Materials with high elastic modulus are stiffer. Tensile stress leads to a lateral strain, called Poisson's ratio £1. This is given by the dimensionless ratio of (D2 - D1)/D1= (-AD/D), where D1 is the original diameter and AD is the change in diameter under axial stress. The negative sign means contraction rather than extension. The ratio of lateral strain to axial strain can be written as:
For most materials, v is unchanged within the elastic limit. Small volume change is accompanied by the deformation and v does not exceed 0.5. The value of v changes from 10% to 50% for different materials. For an elastic isotropic medium under triaxial (x,y,z directions) forces, the strain in the x direction sx , can be written as:
Where ay and az are stresses in the y and z directions, respectively. Similarly, sy and sz can be written using other parameters.
There exists another form of strain in the mechanical structure. This is known as the shear strain, y. This phenomenon does not involve volume change but changes such as shape, twisting and warping appear. The corresponding stress is called shear stress. Shear strain can be written as follows:
Where, x is the shear stress and G is the elastic shear modulus or the modulus of rigidity. Three similar equations can also be written in the x, y, and z planes. The shear modulus, G, can be expressed in terms of Young's modulus, E, and the Poisson ratio, v, as follows:
Isotropic materials can be characterised by these two independent elastic constants. However, anisotropic materials require more than two elastic constants and the number increases with decreasing symmetry. For example, a cubic crystal such as FCC features 3 elastic constants, hexagonal crystal characterises 5 constants, and materials without symmetry require up to 21 elastic constants. The stress-strain relationship is more complex in these cases and depends on the spatial orientation of the crystallographic axes. Hooke's law can still be applied and expressed in the following form:
Where, aj and akl are called stress tensors of rank 2 and expressed in N/m2, sij and skl are called strain tensors of rank 2 and are dimensionless, EiJkl is a stiffness coefficient tensor of rank 4, and Sykl = [Ejkl]-1 is a compliance coefficient tensor of rank 4 and expressed in m2/N. The Eq. (5.6) above can be transformed as:
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?