Silicon Wafer Dry Etching Notching Effect Benefit -prevent -prevention

Where, the tensors Eijkl and Sjkl are replaced by Em,n and Sm,n, respectively. This can be obtained by substituting the indices ij = m and kl = n. In a cubic structure it can be considered that x, y and z directions are replaced by 11, 22, 33 and the corresponding planes xy ^12 and 21, xz ^13 and 31, yz ^ 23 and 32 are in symmetry. With this representation the stress and strain components can be represented as:

^x = E11^x + E12^y + E13^z + E14/xy + E15^xz + E16^yz y y y (5.8)

Sx = S11CTx + S12ay + S13^z + S14rxy + S15Txz + S16Tyz

Each of these equations has six elastic constants. The tensor matrices Em,n and Sm,n are composed of 36 coefficients. A material without symmetry has 21 independent coefficients. Hence, in an isotropic crystal with symmetry the stress and strain components can be expressed by a maximum of two the elastic coefficients, mainly E and u. In the case of a cubic crystal, the three elastic moduli are Eu, E12 and E44. Furthermore, an additional relationship among them is as follows:

Where, a<1 represents the anisotropic coefficient. Hence, the system can always be reduced to its independent stiffness constants. The stiffness constants for the silicon substrate <100> and <110> are functions of the plane but they are independent in the case of Si<111>. A plate made up of Si<111> can be considered as having isotropic elastic properties. Most of the elastic properties, such as the Young's modulus and Poisson ratio remain constant for both lightly and highly doped silicon. Due to the size of atomic radius, huge boron doping in many situations creates shrinkage on the lattice when compared to pure silicon, generating residual stress. A cantilever beam etched from such silicon would be expected to bend out of the plane of the wafer.

All metals deform beyond their elasticity limit when the linear relation between stress and strain does not sufficiently hold. Once metal reaches this stage, it cannot return to its original position. The point P where yielding starts (Fig. 5.4(b)), is known as the proportional limit or yielding point. At some specified strain offset, usually 0.002, a straight line is drawn parallel to the elastic portion of the stressstrain curve. The stress corresponding to the intersection of the line and the practical stress-strain curve as it bends over in the plastic region is defined as the yield strength ay. The yield strength of a material gives its resistance to plastic deformation. The tensile strength of a material is the maximum stress of the stress-strain curve and it corresponds to the stress that a structure can withstand by the applied load. Silicon substrates do not show any plastic deformation up to 800°C. Silicon sensors are inherently insensitive to fatigue failure even at high cyclic loads. They can be used in excess of 100 million cycles without any failures unless sticking and residual stresses come into the picture. This ability is due to the fact that the crystal has no energy absorbing or heat generating mechanism because of inter-granular movement of dislocations in its atoms at atmospheric temperatures. However, single crystal silicon is a brittle material, and shows yield catastrophically, when stressed beyond the yield limit, rather than deform plastically as metals do. SiO2 and Si3N4 exhibit linear-elastic behavior at low strain and high modulus. Plastic deformation is considered as a dislocation generated in the grain boundaries due to stress. Macroscopic deformation forms inter-grain shifts within the structural material. Single crystal silicon has no grain boundaries, for which there is no plastic deformation. The deformation occurs due to two reasons i) the migration of the defects originally present in the lattice, and ii) the defects are generated at the surface. If the defects are nominal, the materials can be considered as perfect elastic material. Perfect elasticity is taken to mean that i) proportionality between stress and strain exists and ii) there is an absence of irreversibility or mechanical hysteresis. Hence, silicon has the advantage over metal as far as micromachining at relatively high temperatures is concerned.

5.5.4 Thermal Properties of Silicon

Single crystal silicon has a high temperature dependency on conductivity as compared to metals but a low thermal coefficient of expansion. The above property makes the silicon not only a better structural material but also helps microfabrication processed such as wet etching. In some processes, the devices are made by joining two wafers such as Si and Pyrex glass together as both wafers have nearly same thermal coefficient of expansion. Any mismatch will introduce thermal stress. This degrades the device performance. It is difficult to fabricate stress-free composite structures over a wide range of temperatures. Silicon on insulator (SOI) wafers can sustain high temperatures. Highly doped silicon bears a linear relationship with the temperature coefficient, resistance and sensitivity over a wide range of temperatures. A thermally isolated structure imposes difficulties in fabrication. Glass or quartz with lower thermal conductivity provides advantages.

5.6 Wet Etching Process

Chemical solutions are primarily used to etch the silicon substrate on the plane of crystallisation. Hence, it is known as wet etching. Wet etching of silicon is used mainly for cleaning, shaping, polishing and characterizing structural and compositional features (Uhlir 1956). Wet etching can provide a higher degree of selectivity than dry etching techniques. In many cases wet etching is faster and an etch rate up to 6 ^m/min cam be achieved. The etching process passes through the following steps: i) reactant transportation to the surface, ii) surface reaction and iii) reaction product transportation (away from the surface). Stirring can increase the diffusion-limited etching. If (ii) is considered as the rate-determining step then etching is reaction-limited and it depends strongly on temperature, material and composition of the solution. Reaction-rate controlled processes are more active than diffusion-limited processes because of their high activation energy. The later is relatively insensitive to temperature variations. In practice, the former is preferred as it is easy to reproduce a temperature setting than a stirring. The etching apparatus should have a good temperature controller and a reliable stirring facility (Kiminsky 1985)). Wet etching can be classified into two categories depending on the direction of etch. Isotropic etching is independent of the plane of the crystal, whereas anisotropic etching is dependent on the plane of orientation.

5.6.1 Isotropic Etchants

In isotropic etching materials are removed uniformly from all directions and it is independent of the plane of orientation of the crystal lattice. Isotropic etching is used for polishing, cleaning and unidirectional etching of the materials. The etchants are a mixture of acidic solutions such as HF, HNO3, and CH3COOH. In single crystal silicon the etchant leads to round isotropic features. They can be used at room temperature or slightly above, but below 50°C. The mixtures have been introduced as silicon etchants by many scientists (Robbins and Schwartz 1960; Uhlir 1956; Hallas 1971; Turner 1958; Kern and Deckert 1978). Later it was discovered that some alkaline chemicals can be used to etch anisotropically. Anisotropic etchants can etch away the crystalline silicon at different rates and different directions. The pH value stays above 12. High temperatures can be maintained for slower etchants (>50°C). Isotropic etchants show diffusion limitation, while anisotropic etchants are reaction rate limited.

5.6.2 Reaction Phenomena

In acid media, the etching starts as a hole injected into the silicon covalence band. Holes are injected using an oxidant, electrical field or photons. Nitric acid and H2O2 in the HNA solution act as oxidants. The holes attack the covalent bond in silicon and oxidise the material. Consider the following reaction produced by HNO3 together with water and HNO2:

In the above reaction, the holes are generated by an autocatalytic process. HNO2 re-enters, sustaining further reaction with HNO3 so as to produce more holes. The reaction is continued so as to produce more holes. This type of reaction takes time to introduce the oxidation process, and continues until a steady-state concentration of HNO2 is reached. The phenomenon is observed at a low concentration of HNO3 (Tuck 1975). After the holes are injected, OH- groups attach to the oxidised silicon to form SiO2 and eventually hydrogen is released.

Subsequently, hydrofluoric acid (HF) dissolves the SiO2 and forms the water-soluble compound H2SiF6. The overall reaction of HNA system with silicon is given by:

The above reaction model considers only the holes. The actual reaction takes place with both holes and electrons. In the isotropic model the rate-determining step in acidic etching involves hole injection into the covalence band, while in alkaline anisotropic etching it involves hole injection into the conduction band. In the first case a hole injected into the covalence band is significantly greater than the latter. The difference between the two etching processes is their reactivity.

5.6.3 Isotropic Etch Curves

Isotropic etching was introduced in late 1950s and was well characterised in the early 1960s. Schwartz and Robbins have published a series of literature in this regard (Robbins and Schwartz 1960). Isotropic etch curves are presented in two forms, one is called iso-etch curves and other one is known as the Arrhenius plot. The former characterises the HNA system for various weight percentages of the constituents as shown in Fig. 5.5 (a).

The concentration of acid plays a major role in the reaction process. In this example, concentrated acids of 49 wt% HF and 70 wt% HNO3 (in both the cases water acts as a diluent) are shown in Fig. 5.5(a) and in Fig. 5.5(b), respectively. The curves are plotted in the sides of a triangle and the axes show the increment of concentrations from low to high. The characteristics of the HNA system can be summarised from Fig. 5.5 as follows:

• The etching is isotropic

• At high HF and low HNO3 concentration, the iso-etch curves describe lines of constant HNO3 concentrations

• HNO3 concentration controls the etch rate

• Etching at those concentrations takes more time to initiate and unstable in silicon growth, hence SiO2 forms slowly

• The etch rate is limited by amount of oxidisation as the process depends on the plane of orientation, and is affected by dopant concentration and catalyst

• At low HF and high HNO3 concentration, iso-etch curves are lines parallel to the acid axis

• The etch rate is controlled by the ability of HF to remove the SiO2

• In the region of maximum etch rate both the agents play important roles

• The addition of acetic acid diluents instead of water does not reduce the oxidising power of the nitric acid until large amount of diluents is added

• The rate contours remain parallel to constant nitric acid over a considerable amount of added acetic acid diluents

These plots can be used for the study of temperature effects on the HNA system. In the region around the HF vertex the surface reaction rate controls the quality of silicon surfaces and sharpness of the corners and edges (Fig. 5.6). The topology of the silicon surfaces depend strongly on the composition of the etching solution. Around the maximum etch rates the surfaces appear quite flat with rounded edges. Slow etching solutions lead to rough surfaces (Robbins and Schwartz 1960). The inner section marked '1' has a smooth surface and subsequent sections marked '2' and '3' have sharpened edges, and rough surfaces, respectively. The etch rate is highly dependent on temperature variation. The reaction rate is increased as the temperature increases. In the low temperatures, etching can be preferential and the activation energy is developed with oxidation reactions. At higher temperatures, etching creates smooth surfaces.

Iso Etch Curves
80 60 40 20 H20 *--
Iso Etch Curves

80 60 40 20

( HCJHjOJ

Fig. 5.5. (a) & (b) Isotropic etch curves (with permission from Dr. J-B. Lee, UTD)

Iso Etch Curves

Fig. 5.6. (a) & (b) Topology or Arrhenius plots: (a) Water as diluent (b) Acetic acid as diluent

Fig. 5.6. (a) & (b) Topology or Arrhenius plots: (a) Water as diluent (b) Acetic acid as diluent

5.6.4 Masking

Isotropic etchants are very fast and do not differentiate silicon planes, all planes are etched equally. Hence, masking is essential. An etch rate of up to 50 ^m/min can be obtained with 66% HNO3 and 34% HF (Kern and Deckert 1978). Silicon oxide is used as masking material with an etch rate of 300 to 800 A/min in the above HF:HNO3.. One can choose a thick layer of SiO2 for shallow etching. A mask of non-etching Au or Si3N4 is needed for deeper etching. Diluted HF etches Si at a higher rate because the reactions in aqueous solutions proceed by oxidation of Si by OH- groups (Hu et al. 1986). A typically buffered HF (BHF) solution has been reported to etch Si at radiochemically measured rates of between 0.23 and

0.45 A/min, depending on the type of doping, and the dopant concentration (Hoffmeister 1969). The doping dependence of the etch rate provides another means of patterning a Si surface. A summary of the dopant dependence (DD) of silicon isotropic etchants is given below.

5.6.5 DD Etchants

The dopant dependence of isotropic etchants is essentially a charge-transfer mechanism. Mostly, the etch rate depends on the dopant type and its concentration. A typical etch rate with an HNA system (1:3:8) for n- or p-type dopant concentration above 1018/cm3 substrate is 1 to 3 ^m/min. Reduction of the etch rate by 150 times can be obtained in n- or p-type regions with a dopant concentration of 1017/cm3 or less (Murarka and Retajczyk 1983). Isotropic preferential etching is a kind of DD etching process. A variety of additives can be mixed into the HNA system (mainly oxidants). They can reduce and modify the etch rate, surface finishes or isotropic characteristics. This effect makes the system selective or preferential. Additives that change the viscosity of the solution can modify the etch rate. One can note that the additive only changes the diffusion coefficient of the reactants (Tuck 1975). An isotropic etchant can be transformed to anisotropic etchant using preferential additives and appropriate masking. Discussion of these effects is not within the scope of this chapter but the basic concept of the anisotropic etching is presented below.

5.7 Anisotropic Etching

Anisotropic etching presents the opposite behavior of isotropic etching. Anisotropic etchants remove the materials based on the crystal plane and do not etch uniformly in all directions. This method is used for the manufacturing of a desired shape and geometry. Anisotropic wet etching techniques were started at the Bell Laboratories in the 1960s. The process is summarised in Table 6.1.

Table 6.1. Anisotropic etching process steps

Process

Duration

Temp. (oC)

Oxidation

Variable

900 - 1200

Spinning (5000 to 6000 rpm)

25-30 sec.

Room temp.

Exposure

20-30 sec.

Room temp.

Develop

60 sec.

Room temp.

Stripping of oxide and resist

500-600 sec.

Room temp.

RCA1(NH3(25%)+ H2O + H2O2:1: 5:1)

600 sec.

Boiling

RCA2 (HCl + H2O + H2O2:1: 6: 1)

600 sec.

Boiling

HF-dip (2% HF)

10 sec.

Room temp

Anisotropic etch

Several minutes

70-100

Anisotropic etching solves the problem of lateral control. The laterally masked geometry of the planar surface can be etched and a vertical profile can be easily made. Etch-stop techniques are also available for anisotropic etchants. Although it solves the problem of lateral etching, the process is not problem-free. The process is slow even in the fast etching direction of the plane (100), with an etch rate below 1 ^m/min, and also consumes more time. Anisotropic etchants are temperature sensitive and insensitive to agitation agents.

5.7.1 Anisotropic Etchants

Alkaline solutions such as KOH, NaOH, LiOH, CsOH, NH4OH, and ammonium hydroxides, with the addition of alcohol are primarily used as anisotropic etchants. Alkaline organics such as ethylenediamine, choline (trimethyl-2-hydroxyethyl ammonium hydroxide) or hydrazine with additives such as pyrocathechol and pyrazine are also used. The best fit of the etch rate can be expressed using the empirical formula:

Where, k is a constant. The rate is dopant insensitive over several orders of magnitude, but at certain concentration it shows bias dependence (Allongue et al. 1993). Etching of silicon can occur without external voltage. Alcohols such as propanol and isopropanol butanol slow the attack whereas pyrocathechol speeds up the etch rate. Some additives such as pyrazine and quinone are used as catalysts (Linde and Austin 1992). The etch rate increases with temperature. So etching at higher temperatures can give the best results. However, for all practical purposes, temperatures of 80 to 85°C should be avoided because the solvent starts evaporating at this temperature. The topology plot (also called Arrhenius plot) presented in Fig. 5.6 shows the effect of temperature on silicon etching (Lee et al. 1995). The effect of temperature on etch rate is high, compared to the effect of plane orientation. The slope differs for each plane. While choosing an anisotropic etchant, a variety of issues are considered. Some of them are toxicity, etch rate, etch-stop, compatibility, etch selectivity over other materials, mask material and thickness and topology of the bottom surface.

The principal anisotropic etchants are KOH and ethylene-diamine/ pyrocatechol + water (EDP). More recently, quaternary ammonium hydroxide solution such as tetraethyl ammonium hydroxide (TEAH) has become popular.

5.7.2 Masking for Anisotropic Etchants

SiO2 cannot be used as a masking material due to its long exposure to KOH etchants. Etch rate of SiO2 is a function of KOH concentration. The etch rate of thermally grown SiO2 in KOH-H2O varies and depends on the quality of the oxide, the etching container and the age of the etching solution. Thermal oxides create a strong compressive force in the oxide layer because one silicon atom takes nearly twice as much space as in single crystalline silicon. Oxides can be removed by pinholes and the etch rate in KOH remains higher than that of thermal oxides. Low pressure chemical vapor deposited (LPCVD) oxides can be used as mask materials of good quality as compared to thermal oxides. LPCVD nitride can serve as a better mask than less dense plasma deposited nitride. Nitride can easily be patterned with photoresist and etched in CF4/O2-based plasma. Nitride deposition does not pose any problem. In many applications it is necessary to protect the backside of the wafer from isotropic or anisotropic etchants. This can be done either by mechanical or chemical protection methods. In the former method, the wafer is held in a Teflon holder and in the latter case wax or other organic coatings spin onto the back side of the wafer. Two wafers can be glued back-to-back for faster processing.

Anisotropic etchants can leave the surface rough. This type of roughness can be found during microbridge etching. Macroscopic roughness is called notching or pillowing. Notching increases linearly with etch depth but decreases with higher concentrations of KOH. However, increases in temperature can decrease the roughness of the surface. The macroscopic smoothness can be degraded into microscopic roughness.

5.8 Etching Control: The Etch-stop

It is important to control and stop etching on the wafer when the desired shape and depth of the structure has been reached. An effective etch-stop can produce high resolutions and high yields. Practically, the wafer is not uniform and therefore, the devices cannot be etched uniformly. These effects are impossible to avoid in any microdevice. Advanced techniques in controlling, monitoring, stabilising and visualising the etch rates have been reported. The etch-stop process can be handled through etchant compositions, etchant agents, N2 sparing, loading effects, etchant temperatures and diffusion effects. The depth of etch depends on temperature, etchant concentration, and stirring control. A stirring can lead to a smooth surface and uniform etching. Light can also affect the etch rate. Sample preparation as well as surface pre-processing can affect the etch rate. The pre-treatment and surface cleaning with 5% HF can remove the native oxides. Normally, a constant etch time method is used for the etch-stop but in the case of thin membranes it can be stopped with sufficient precision. Etch-stop can be prepared using an etch rate dependant dopant. Electrochemical etch-stop can, however, improve the accuracy. It is also easy to handle.

5.8.1 Boron Diffusion Etch-stop

The boron diffusion method is a widely used etch-stop technique. It is based on the fact that anisotropic etchants do not attack boron-doped (p+) Si layers. It can create a stress-free, dislocation-free and slow etching layer. This effect was first discovered by Greenwood. The presence of a p-n junction causes the effect (Greenwood 1969). The boron diffusion method is explained in Fig. 5.7 for micronozzle device fabrication:

Fig. 5.7. Boron diffusion stop

Fig. 5.7. Boron diffusion stop

The wafer is first deposited with SiO2 followed by lithography and mask patterning. A boron p+ profile is diffused from the back of the wafer prior to the anisotropic etching. The etching process is automatically stopped at the junction. Boron atoms are smaller than silicon, hence a highly doped and freely suspended membrane or diaphragm can be stretched to create tensile stress. The microstructures can become flat and do not buckle. Boron doping causes lattice constant to decrease while germanium doping causes it to increase. A membrane doped with both boron and germanium can be etched much slower than that of undoped silicon. One disadvantage is that high concentration boron doping can not be used for bipolar type CMOS ICs.

The method can only be applied to microstructures where electronics circuits are not needed. Another limitation of this process is that it can only be used for a fixed number and angles of (111) planes. The method is not limited to anisotropic etching but can also be used for isotropic etch-stop.

5.8.2 Electrochemical Etch-stop

For microdevices such as piezoresistive pressure sensors, the doping concentration should be kept as low as 1019/cm3 as piezoresistive coefficients may drop beyond this value. Moreover, boron doping affects the quality of the crystal and induces tensile stress and prevents the incorporation of integrated electronics. Therefore, a boron diffusion etch-stop cannot be used for all purposes. An alternative method is called an electrochemical technique that can be used as isotropic etch-stop. A lightly doped p-n junction is used as cathode and a counter electrode as anode in the etchant KOH.

A bias voltage is applied between the wafer and anode. The technique was proposed by Waggener (Waggener et al. 1967). The p-n junction forms as a large diode. A wafer is mounted on a sapphire plate, with an acid-resistant wax and immersed in the solution. An ohmic contact to the n-type epitaxial layer is connected to one pole of the voltage source and the other pole is connected via an ammeter to the solution as shown in the Fig. 5.8. In this process a p-type substrate is etched and stopped at he p-n junction.

Electrochemical Etch Paddle

Fig. 5.8. Electrochemical etch-stop

Similarly, an n-type substrate can be etched and stopped at the n-p junction. Electrochemical etch-stop is a highly controlled method and it replaces many high temperature aggressive chemical etching processes. The method utilises a much milder solution with less complexity as far as preparation and transformation of photoresist mask is concerned (Kern and Deckert 1978). In the case of acidic etching, an electrical power supply can be employed to drive the chemical reaction by supplying holes in the silicon surface. The dissolution rate of silicon is related to the current density. Increasing the acceptor concentration can further increase the rate.

The acidic electrochemical technique in micromachining is primarily used for polishing the surfaces. Since the etching rate increases with current density, high spots on the surface are more rapidly etched and very smooth surfaces are produced. Electrochemical etch-stops with anisotropic etchants such as KOH and EDP have been performed (Waggener et al. 1967). In electrochemical anisotropic etching the p-n junction is made by the epitaxial growth of an n-type layer on a p-type substrate.

5.8.3 Thin Films and SOI Etch-stop

Another etch-stop technique is based on the principle of the silicon-on-insulator (SOI) fabrication method. Anisotropic etchant does not react with a number of materials. Hence, this idea can be employed to establish an etch-stop. An example is a Si3N4-based diaphragm fabrication. Silicon nitride is a strong and hard chemical and it can form stress in the film. This stress can be transformed from tensile to compressive for thin films. Another example of this category is a SiO2 layer. A layer of SiO2 between two layers of silicon can form an excellent etch-stop because many etchants of silicon do not react with SiO2. SOI is a simple technique compared to the electrochemical etch-stop method.

5.9 Problems with Etching in Bulk Micromachining

Despite its simplicity in controlling the etching in bulk micromachining, wet etching is not a flexible and reliable process (Gad-el-Hak 2002). A dry etching process is an alternative choice. Unavoidable problems associated with this process are as follows:

• Extensive real estate (RE) consumption

• Difficulties in etching at convex corners

• Difficult in preparing the mask with high precision

• Etch rate is very sensitive to both agitation and temperature making it difficult to control both lateral and vertical geometries of the structure

5.9.1 RE Consumption

Bulk micromachining is involved with extensive real estate consumption because the process removes a comparable amount of materials from the bulk substrate. Fig. 5.9 illustrates the fabrication of two membranes created by etching through a <100> wafer from the backside until an etch-stop is reached. A large amount of silicon has been wasted to create this structure. The consumption of silicon can not be stopped but it can be reduced best by practice and careful design as well as appropriate etching techniques. This is defined as real estate gain by reducing the consumption of wafer materials. Two different real estate gain techniques are:

• RE gain by etching from the front

• RE gain by using silicon bonded wafers

Fig. 5.9. Wastage of material

RE can be gained by removing fewer amounts of material and the use of thinner bulk instead of thick wafers. This approach is useful for wafers with a thickness more than 200 ^m. An alternative solution is to etch the wafer from the front. Anisotropic etchants can remove an amount of materials depending on the orientation of the wafer with respect to the mask, and etch until a pyramidal pit is formed (Fig. 5.9). These produce an angle of 54.7° to the surface of the silicon wafer. The etching stops when the pyramid pit is completed. Hence, there is no further wastage of the wafer.

Anisotropic Etching Silicon
Fig. 5.10. Real estate gain by silicon diffusion bonding (SDB)

Another process to gain real estate is use of thinner wafers and join them together using the silicon direct bonding (SDB) technique. SDB is a process to bond two silicon wafers by applying pressure and annealing at a certain temperature. An example is given in Fig. 5.10. A groove (Fig. 5.10(a)) can be made using two parts such as (b) and (c). First, a thin silicon wafer <100> is taken and then it is etched to form a pyramid pit as shown in Fig. 6.10(b). In the second stage it is bonded or fused with another thin wafer to produce another desired structure (Fig. 5.10(c)). Finally, the bonded wafer is etched from the backside of the surface in order to get the final desired shape as shown in Fig. 5.10(d). The RE gain by using silicon fusion-bonded wafers can reduce the consumption of wafers, waste of etchants and time. This process saves RE and makes the design and etching simple.

5.9.2 Corner Compensation

The corners of the structure do not etch uniformly especially in anisotropic etching. In microdevices it is necessary to preserve the shape while etching. In practice, difficulty arises due to the fact that some planes etch faster than others. This results in a loss of the desired structure. The effect can be reduced through a technique called corner compensation. Convex corners can be made by adding extra structures and then removing them during etching. The effect of corner etching can lead to either underetching or undercutting of the substrate.

In general, convex corner structures and non-crystal planes are undercut during wet anisotropic etching. These characteristics have been utilised to fabricate freely suspended structures. Conversely, these effects need to be prevented in various applications. Vertical posts and vertical walls have been employed to protect convex corners and non-crystal planes from undercutting. Experimental results have shown that cavities with arbitrary shapes can be fabricated using wet anisotropic etching. In the case of etching rectangular corners, deformation occurs at the edges due to undercutting. This type of effect is unwanted where symmetry and perfect 90° corners are essential. The undercutting effect is dependent on etch time and related to the etch depth. The effect can be reduced by using saturated KOH solutions with isopropanol. This may cause a formation of hillocks and piles on the surface. Peeters has claimed that these hillocks and piles are due to carbonate contamination of the etchants and he has suggested that the etchant ingredients should be stored under an inert nitrogen atmosphere and the etching process should be carried out under an inert atmosphere (Peeters 1994). Another way to reduce the undercut is at the expense of a reduced anisotropy ratio. A mask is added to the corners in the layout. The thickness of the mask and the shape depend on the type of etchants and the shape of the corner. In the case of square corner compensation, the square of SiO2 in the mask is enhanced by adding an extra SiO2 square to the corner.

5.10 Conclusions

Microelectromechanical systems (MEMS) are fabricated by using three methods which include: i) bulk micromachining, ii) surface micromachining, and iii) wafer bonding. The most important and common sub-process within the above methods is etching. This chapter characterised the etching process in bulk micromachining. In particular, the chapter discussed the following topics.

• Silicon crystallography

• Silicon as the material

• The importance of properties of material with respect to etching process

• Thermal properties

• The difference between the isotropic and anisotropic etching and etchants

• Various types of etch-stops and their techniques

• The problems encountered in the bulk micromachining process

5.11 References

Allen DM (1986) The principles and practice of photochemical machining and photoetching. Inst of Physics Pub Inc. pp 1-18 Allongue P, Kieling VC, Gerischer H (1993) Etching of Si in NaOH solutions, Part I and II,

J. Electrochem. Soc. 40:1009-1018 and 40:1018-1026 Ammar ES, Rodgers TJ (1980) UMOS transistors on (110) silicon, IEEE Trans. Electron

Devices ED-27:907-914 Bassous E, Taub HH, Kuhn L (1977) Ink jet printing nozzle arrays etched in silicon, Appl.

Phys. Lett. 31:135-137 Boivin LP (1974) Thin film laser-to fiber coupler, Appl. Opt. 13:391-395 Editorial (2004) Analog Devices combines micromachining with BICMOS, Semiconductor

International, available at http://www.reed-electronics.com Editorial (1974) Transducers, pressure and temperature catalog. White paper, National

Semiconductor, Sunnyvale, USA Editorial (1977) Thermal character print head. Technical Report, Texas Instruments, Austin

Gad-el-Hak M (2002) The MEMS handbook. CRC press, pp 200-211

Greenwood JC (1969) Ethylene Diamine-Cathechol-Water mixture shows preferential etching of p-n junction, J. Electrochem. Soc. 116:1325-1326 Greenwood JC (1984) Etched silicon vibrating sensor, J. Phys. E, Sci. Instrum. 17:650-652 Hallas CE (1971) Electropolishing silicon, Solid State Technology 14:30-32 Harris TW (1976) Chemical Milling. Clarendon Press, Oxford

Hoffmeister W (1969) Determination of the etch rate of silicon in buffered HF using a 31 Si tracer method, Int. J. Appl. Radiation and Isotopes 2:139 Hu JZ, Merkle LD, Menoni CS, Spain IL (1986) Crystal data for high-pressure phases of silicon, Phys. Rev. B 34:4679-4684 Kaminsky G (1985) Micromachining of silicon mechanical structures, J. Vac. Sci.

Technol., B3:1015-1024 Kern W, Deckert CA (1978) Chemical etching in thin film processes. Vossen JL, Kern W

(eds) Academic Press, Orlando Lee JB, Chen Z, Allen MG, Rohatgi A, Arya R (1995) A miniaturised high-voltage solar cell array as an electrostatic MEMS power supply, J. Microelectromech. Syst. 4:102108

Linde H, Austin L (1992) Wet silicon etching with aqueous amine gallates, J. Electrochem.

Soc. 139:1170-1174 Madou MJ (1997) Fundamentals of microfabrication, CRC Press, New York Madou MJ, Morrison SR (1989) Chemical sensing with solid state devices. Academic Press, New York

Michalicek AM (2004) Introduction to MEMS. http://mems.colorado.edu/c1.res.ppt /ppt/g.tutorial

Murarka SP, Retajczyk TFJ (1983) Effect of phosphorous doping on stress in silicon and polycrystalline silicon, J. Appl. Phys. 54:2069-2072 O'Neill P (1980) A monolithic thermal converter, Hewlett-Packard J. 31:12-13 Palik ED Faust JW, Gray HF, Green RF (1982) Study of the Etch-Stop Mechanism in

Silicon, J. Electrochem. Soc. 129:2051-2059 Peeters E (1994) Process development for 3D silicon microstructures, with application to mechanical sensor design, Ph.D. thesis, Catholic University of Louvain Petersen KE (1982) Silicon as a mechanical material, Proc. IEEE, 70:420-457 Pfann WG (1961) Improvement of semiconducting devices by elastic strain, Solid State Electron. 3:261-267

Robbins H, Schwartz B (1960) Chemical etching of silicon-II: The system HF, HNO3,

H2O, and HC2C3O2, J. Electrochem. Soc. 107:108-111 Rodgers TJ Hiltpold WR, Frederick B, Barnes JJ, Jenne FB, Trotter JD (1977) VMOS

memory technology, IEEE J. Solid-State Circuits SC12:515-523 Schwartz B, Robbins H (1976) Chemical etching of silicon-IV, Etching Technology, J.

Electrochem. Soc. 123:1903-1909 Smith CS (1954) Piezoresistance effect in germanium and silicon, Phys. Rev., 94:42-49 Tuck B (1975) The chemical polishing of semiconductors, J. Mater. Sci. 10:321-339 Tuft ON, Chapman PW, Long D (1962) Silicon diffused-element piezoresistive diaphragms, J. Appl. Phys. 33:3322 Turner DR, (1958) Electropolishing silicon in hydrofluoric acid solutions, J. Electrochem. Soc. 105:402-408

Uhlir A (1956) Electrolytic shaping of germanium and silicon, Bell Syst. Tech. J. 35:333347

Waggener HA, Krageness RC, Tyler AL (1967) Two-way etch, Electronics 40:274

Was this article helpful?

0 0

Post a comment