**2.2 Etch rate anisotropy**

The photos of the hemispherical specimen before and after etching at 61 °C in TMAH + Triton are taken as one of the examples and presented in Fig. 4. The contour maps made by the 3D Anisotropic-Etching Simulator (FabMeister-ES) are shown in Fig. 5. Due to the crystallographic symmetry of the hemispherical specimen, only one quarter part with averaged etch rates among equivalent orientations is presented. The range of the planes

This chapter starts by providing the completely etch rate anisotropy in surfactant-modified wet etching in section 2; the mechanism behind of the change of etching characters when compared with pure etchants will be analyzed in section 3; several applications for the fabrication of new structures by using this advanced anisotropic wet etching will be

Different etchants can give different etching properties. As one of the most important properties, the etch rate anisotropy clearly manifests the etching behavior in a concentrated solution. In this chapter, the characterization of the etch rate anisotropy is studied by using hemispherical silicon specimens, with and without surfactant in TMAH solutions. Especially, surfactant-modified etching process is analyzed in detail because of its benefit in

The hemispherical specimen enables us to obtain the etch rate for all range of crystallographic orientations under the same etching conditions simultaneously, because all the orientations are placed on its surface. The P-type single-crystal hemispherical specimen with a diameter of 44 mm (resistivity: 6–12 Ω-cm) is used in the evaluation. The ingot of hemispheres is provided by Sumitomo Sitix Corporation. In order to produce hemispheres, it is mechanically ground, lapped and polished into mirrored surfaces with a sphericity of less than 10 µm, latitude from 0 to 90° and surface roughness of 0.005–0.007 µm in the arithmetical average by Okamoto Kogakukosakusho Corporation. The etch rate at each orientation is calculated by measuring the shape change before and after etching. The optimized etching depth should be in the range of 100–150 µm in order to avoid interference between neighboring orientations while maintaining the resolution of the geometry measurement. The shape is measured using a 3D profile machine UPMC550-CARAT (Carl Zeiss Co.) with an accuracy of less than 1.0 µm. Fig. 3 shows the locations of crystallographic orientations on the silicon hemispherical sample and a schematic view of the surface profile measurement. Area 'A' corresponds to the measurable area of the hemispherical silicon sample. The place outside the measurement zone and the bottom area are protected by a thermally grown oxide layer. The surface profile is probed every 2° of latitude ranging from 0° to 70°, and every 2° of longitude ranging from 0° to 360°. Supplements of de-ionized (DI) water into an etching bath every 2 h control the tolerance of etching temperature within 1

TMAH (Toyo Gosei Co. Ltd) and Triton X-100 (Amersham Biosciences) are used as the main etchant and surfactant, respectively. The Triton solution is used to prepare the surfactantadded TMAH solution. The fresh etchant is employed in every subsequent experiment.

The photos of the hemispherical specimen before and after etching at 61 °C in TMAH + Triton are taken as one of the examples and presented in Fig. 4. The contour maps made by the 3D Anisotropic-Etching Simulator (FabMeister-ES) are shown in Fig. 5. Due to the crystallographic symmetry of the hemispherical specimen, only one quarter part with averaged etch rates among equivalent orientations is presented. The range of the planes

presented in section 4.

MEMS applications.

**2.1 Experimental details** 

**2.2 Etch rate anisotropy** 

**2. Characterization of the etch rate anisotropy** 

°C. The total numbers of probe points to be measured are 6480.

affected by surfactant adding is clearly visible. The etch rates of exact and vicinal {100} planes are almost unaffected when the surfactant is added, while the etch rates of exact and vicinal {110} planes are reduced significantly.

Fig. 3. Location of different crystallographic orientations on one silicon hemispherical sample and schematic view of surface profile measurement. (Reproduced with permission from IOP)

Fig. 4. Photos of the hemisphere specimen: (a) before etching and (b) after etching (etching temperature = 61 °C).

The effect of etching temperature on the etch rates is shown in Fig. 6. The etching anisotropy is influenced by temperature. The orientation of the highest etch rate is shifted toward {350} with the increase in temperature. Contrary to 71 and 81 °C, no single plane prominently appears as the highest etching rate plane at 61 °C. The etch rates of the orientations between {100} and {110} largely depend on the temperature; however, those in between {110} and {111} exhibit less dependence. It can obviously be concluded from these results that the etching anisotropy depends upon the etching temperature. This will result in different etched profiles due to the difference in etching temperatures even if the

Advanced Surfactant-Modified Wet Anisotropic Etching 137

{100} 0.133 1.000 0.308 1.000 0.468 1.000 {130} 0.138 1.038 0.462 1.500 0.717 1.532 {120} 0.102 0.767 0.437 1.419 0.709 1.515 {230} 0.093 0.699 0.272 0.883 0.600 1.282 {110} 0.032 0.241 0.055 0.179 0.036 0.075 {551} 0.034 0.256 0.055 0.179 0.098 0.209 {331} 0.034 0.256 0.064 0.208 0.126 0.269 {221} 0.033 0.248 0.062 0.201 0.152 0.325 {111} 0.019 0.143 0.033 0.107 0.035 0.077

Temperature-dependent etching anisotropy is verified by fabricating 3D circular shape cavities on silicon {100} wafers. Fig. 7 shows the SEM images of circular cavities (etch depth = 100 µm) formed in Triton-added 25 wt% TMAH at 61 and 81 °C using circular shape mask opening, respectively. SEM images are taken after removal of the oxide masking layer. In the case of TMAH + Triton, the {110} and its vicinal planes emerge along <100> directions as these planes exhibit significantly low etch rates. The cavity etched at 81 °C provides better roundness than that of etched at 61 °C. This is because of improved anisotropy ({mnl}/{100}) at high temperature (Table 2). When the temperature is increased from 61 to 81 °C, the etch rate ratios of {110}/{100} and {111}/{100} are changed from 0.241 and 0.143 to 0.077 and

Fig. 7. Circle-like 3D microstructures in Si {100} wafers using one-step etching in 25 wt %

The Arrhenius-type dependence of etching rates for six planes {110}, {100}, {111}, {331}, {711}, and {221} is plotted in Fig. 8. The activation energy of TMAH + Triton (0.1-0.2 eV) is lower than that of pure TMAH (or KOH) solution (0.5-0.7 eV). The extreme resemblance of activation energy for {331} and {221}, which lie between {110} and {111}, may be associated with the almost same etch rates as described earlier. In particular, {110} plane has very low

TMAH + 0.1 vol% Triton X-100 at: (a) 61 °C and (b) 81 °C.

Table 2. Orientation- and temperature-dependent etch rates and etch rate ratios of

61ºC 71ºC 81ºC

(µm/min) {mnl}/{100} Etch rate

(µm/min) {mnl}/{100}

Orientatio n

0.075, respectively.

activation energy.

Etch rate

{mnl}/{100} in TMAH + Triton solutions.

(µm/min) {mnl}/{100} Etch rate

same mask is used. Examples of measured etch rates for nine different orientations are listed in Table 2. The etch rate ratios of {mnl}/{100}, where m, n, and l are integers, increase with temperature for the planes lying between {100} and {110} (excluding {100} and {110}). On the other hand, these ratios for the planes lying between {110} and {111} are almost the same for 61 °C; however, for other temperatures (71 and 81 °C), the values vary.

Fig. 5. Contour maps of etch rate in pure and surfactant-added TMAH solutions

Fig. 6. Effect of temperature on the etch rate anisotropy in 25 wt % TMAH + 0.1 vol% Triton X-100.

same mask is used. Examples of measured etch rates for nine different orientations are listed in Table 2. The etch rate ratios of {mnl}/{100}, where m, n, and l are integers, increase with temperature for the planes lying between {100} and {110} (excluding {100} and {110}). On the other hand, these ratios for the planes lying between {110} and {111} are almost the same for 61 °C; however, for other temperatures (71 and 81 °C), the values

Fig. 5. Contour maps of etch rate in pure and surfactant-added TMAH solutions

Fig. 6. Effect of temperature on the etch rate anisotropy in 25 wt % TMAH + 0.1 vol% Triton

vary.

X-100.


Table 2. Orientation- and temperature-dependent etch rates and etch rate ratios of {mnl}/{100} in TMAH + Triton solutions.

Temperature-dependent etching anisotropy is verified by fabricating 3D circular shape cavities on silicon {100} wafers. Fig. 7 shows the SEM images of circular cavities (etch depth = 100 µm) formed in Triton-added 25 wt% TMAH at 61 and 81 °C using circular shape mask opening, respectively. SEM images are taken after removal of the oxide masking layer. In the case of TMAH + Triton, the {110} and its vicinal planes emerge along <100> directions as these planes exhibit significantly low etch rates. The cavity etched at 81 °C provides better roundness than that of etched at 61 °C. This is because of improved anisotropy ({mnl}/{100}) at high temperature (Table 2). When the temperature is increased from 61 to 81 °C, the etch rate ratios of {110}/{100} and {111}/{100} are changed from 0.241 and 0.143 to 0.077 and 0.075, respectively.

Fig. 7. Circle-like 3D microstructures in Si {100} wafers using one-step etching in 25 wt % TMAH + 0.1 vol% Triton X-100 at: (a) 61 °C and (b) 81 °C.

The Arrhenius-type dependence of etching rates for six planes {110}, {100}, {111}, {331}, {711}, and {221} is plotted in Fig. 8. The activation energy of TMAH + Triton (0.1-0.2 eV) is lower than that of pure TMAH (or KOH) solution (0.5-0.7 eV). The extreme resemblance of activation energy for {331} and {221}, which lie between {110} and {111}, may be associated with the almost same etch rates as described earlier. In particular, {110} plane has very low activation energy.

Advanced Surfactant-Modified Wet Anisotropic Etching 139

the type of solution. The samples are now stored in DI water. A surfactant bath consisting of

The density of adsorbed surfactant molecules in the layer increases as the surfactant concentration is increased, reaching a maximum value (adsorption saturation) at a concentration that is similar in magnitude to the critical micelle concentration (CMC), typically between 0.01 vol% and 0.1 vol% (i.e. 100–1000 ppm). Increasing the surfactant concentration further results in a larger number of micelles in solution but does not affect the adsorption density at the interface. The Triton concentration value of 1 vol% in DI water is a simple choice in order to study the effect of the pre-adsorbed layer as it ensures that

The samples are dipped in the surfactant bath at room temperature and 60 ºC for different times. Prior to immersion in the Triton bath, the samples are dipped in 5% hydrofluoric (HF) acid solution for 1 min and then thoroughly rinsed in DI water. This step is attempted to make the surface hydrophobic. After the surfactant bath, the samples are gently dipped in DI water for several times to remove the most weakly adsorbed surfactant molecules from the surface. These dippings are carried out in still water in a Teflon container. The number of dippings in DI water after the surfactant bath may affect the thickness of the adsorbed layer. Therefore, the effect of the number of gentle dippings is also studied. Moreover, the effect of ultrasonic agitation during the surfactant bath is also investigated. After several dippings in DI water, the samples are dried and used for surfactant layer thickness measurements by ellipsometry. Bare silicon samples are used as reference surfaces. For the ellipsometry measurements of the obtained Triton films we used a standard geometry where the sample is placed horizontally and visible light is reflected at a grazing angle and received by the detector. The spectral analysis is performed using a commercial

Although the thickness of the Triton layer (h) increases with the Surfactant Bath Time (SBT) and the thickness can be reduced by performing one or several dippings (ND), we have observed that two Triton layers of equal thickness that have been prepared in different ways, namely, (i) by only the Triton bath and, (ii) by the Triton bath followed by several dippings, have different deviations for the measured thickness in the ellipsometry measurements. When the thickness is determined by focusing the light beam on different regions of a sample without rinsing, the measurements exhibit large fluctuations (±6 Å), indicating an uneven surfactant distribution (large salience). However, the scattering range of the thickness for the samples that have been dipped is less than ±2 Å, thus indicating a more homogeneous packing of the surfactant molecules. For improved repeatability and control, we conclude already at this stage that the Triton layer should be prepared by

In order to evaluate how the thickness of the Triton layer decreases with the number of dippings, we consider a Triton layer that is initially saturated, meaning that the silicon surface has been exposed to the Triton solution for a sufficiently long time (24 h), thus ensuring that the thickness has reached its maximum value. Fig. 9(a) shows the thickness of surfactant layer as a function of the number of dippings. Here, ND = 0 means that the sample is directly dried by air. The plot shows that, for both {100} and {110} oriented wafers, a layer of finite, non-zero thickness is measured, indicating that the surfactants are adsorbed on the silicon surface. Generally, {110} samples show a thicker Triton layer, indicating a larger ability to attract surfactant molecules. Below ND = 3, the difference between Si {110} and Si {100} at the same number of dippings is small, and the thickness is high for both

following the second procedure (Triton bath followed by several dippings).

enough surfactant is adsorbed on the surface according to the CMC argument.

1 vol% Triton X-100 in DI water is prepared.

spectroscopic ellipsometry analyzer (MARY-102).

Fig. 8. The Arrhenius plot of etching rates for different planes.

#### **3. Adsorption of surfactant molecules and its effect on etching**

The change in the etching behavior of TMAH after the incorporation of the surfactant indicates that the reaction mechanism at the surface is affected by the surfactant molecules, which may block selectively some of the active surface sites, particularly those that appear on Si {110}. Triton X-100 has the following molecular structure:

One end of the molecule is hydrophilic, and the other is hydrophobic.

#### **3.1 Adsorption of surfactant molecules on silicon surfaces**

Ellipsometry is utilized as a surface-sensitive technique in order to determine the thickness of thin films adsorbed on different surfaces at the subnanometer. Although the non-ionic surfactants do not take part in the etching reaction, their molecules are known to adsorb on the silicon surface by FT-IR observation. In this section, we report detailed ellipsometry measurements in order to determine and characterize the preferential adsorption of the nonionic surfactant Triton X-100 on {110} and {100} surfaces before etching. The study focuses on the dependence of the thickness of the surfactant layer on various conditions such as the surfactant bath time (SBT), its temperature (T) and the use of ultrasonic agitation (UA), etc. For experiments, we use three-inch diameter, p-type single crystalline {100} and {110} silicon wafers of 5-10 Ω-cm resistivity. These orientations are selected because of two reasons: (i) most widely used and commonly available, (ii) surfactant effects are quite different for Si {100} and Si {110}. Firstly, the wafers are diced into 13×13 mm2 small samples. Thereafter, the samples are properly cleaned in chemical solutions followed by thorough rinse in deionized (DI) water. In this study all the containers are either glass or Teflon, depending on

Fig. 8. The Arrhenius plot of etching rates for different planes.

on Si {110}. Triton X-100 has the following molecular structure:

One end of the molecule is hydrophilic, and the other is hydrophobic.

**3.1 Adsorption of surfactant molecules on silicon surfaces** 

**3. Adsorption of surfactant molecules and its effect on etching** 

The change in the etching behavior of TMAH after the incorporation of the surfactant indicates that the reaction mechanism at the surface is affected by the surfactant molecules, which may block selectively some of the active surface sites, particularly those that appear

Ellipsometry is utilized as a surface-sensitive technique in order to determine the thickness of thin films adsorbed on different surfaces at the subnanometer. Although the non-ionic surfactants do not take part in the etching reaction, their molecules are known to adsorb on the silicon surface by FT-IR observation. In this section, we report detailed ellipsometry measurements in order to determine and characterize the preferential adsorption of the nonionic surfactant Triton X-100 on {110} and {100} surfaces before etching. The study focuses on the dependence of the thickness of the surfactant layer on various conditions such as the surfactant bath time (SBT), its temperature (T) and the use of ultrasonic agitation (UA), etc. For experiments, we use three-inch diameter, p-type single crystalline {100} and {110} silicon wafers of 5-10 Ω-cm resistivity. These orientations are selected because of two reasons: (i) most widely used and commonly available, (ii) surfactant effects are quite different for Si {100} and Si {110}. Firstly, the wafers are diced into 13×13 mm2 small samples. Thereafter, the samples are properly cleaned in chemical solutions followed by thorough rinse in deionized (DI) water. In this study all the containers are either glass or Teflon, depending on the type of solution. The samples are now stored in DI water. A surfactant bath consisting of 1 vol% Triton X-100 in DI water is prepared.

The density of adsorbed surfactant molecules in the layer increases as the surfactant concentration is increased, reaching a maximum value (adsorption saturation) at a concentration that is similar in magnitude to the critical micelle concentration (CMC), typically between 0.01 vol% and 0.1 vol% (i.e. 100–1000 ppm). Increasing the surfactant concentration further results in a larger number of micelles in solution but does not affect the adsorption density at the interface. The Triton concentration value of 1 vol% in DI water is a simple choice in order to study the effect of the pre-adsorbed layer as it ensures that enough surfactant is adsorbed on the surface according to the CMC argument.

The samples are dipped in the surfactant bath at room temperature and 60 ºC for different times. Prior to immersion in the Triton bath, the samples are dipped in 5% hydrofluoric (HF) acid solution for 1 min and then thoroughly rinsed in DI water. This step is attempted to make the surface hydrophobic. After the surfactant bath, the samples are gently dipped in DI water for several times to remove the most weakly adsorbed surfactant molecules from the surface. These dippings are carried out in still water in a Teflon container. The number of dippings in DI water after the surfactant bath may affect the thickness of the adsorbed layer. Therefore, the effect of the number of gentle dippings is also studied. Moreover, the effect of ultrasonic agitation during the surfactant bath is also investigated. After several dippings in DI water, the samples are dried and used for surfactant layer thickness measurements by ellipsometry. Bare silicon samples are used as reference surfaces. For the ellipsometry measurements of the obtained Triton films we used a standard geometry where the sample is placed horizontally and visible light is reflected at a grazing angle and received by the detector. The spectral analysis is performed using a commercial spectroscopic ellipsometry analyzer (MARY-102).

Although the thickness of the Triton layer (h) increases with the Surfactant Bath Time (SBT) and the thickness can be reduced by performing one or several dippings (ND), we have observed that two Triton layers of equal thickness that have been prepared in different ways, namely, (i) by only the Triton bath and, (ii) by the Triton bath followed by several dippings, have different deviations for the measured thickness in the ellipsometry measurements. When the thickness is determined by focusing the light beam on different regions of a sample without rinsing, the measurements exhibit large fluctuations (±6 Å), indicating an uneven surfactant distribution (large salience). However, the scattering range of the thickness for the samples that have been dipped is less than ±2 Å, thus indicating a more homogeneous packing of the surfactant molecules. For improved repeatability and control, we conclude already at this stage that the Triton layer should be prepared by following the second procedure (Triton bath followed by several dippings).

In order to evaluate how the thickness of the Triton layer decreases with the number of dippings, we consider a Triton layer that is initially saturated, meaning that the silicon surface has been exposed to the Triton solution for a sufficiently long time (24 h), thus ensuring that the thickness has reached its maximum value. Fig. 9(a) shows the thickness of surfactant layer as a function of the number of dippings. Here, ND = 0 means that the sample is directly dried by air. The plot shows that, for both {100} and {110} oriented wafers, a layer of finite, non-zero thickness is measured, indicating that the surfactants are adsorbed on the silicon surface. Generally, {110} samples show a thicker Triton layer, indicating a larger ability to attract surfactant molecules. Below ND = 3, the difference between Si {110} and Si {100} at the same number of dippings is small, and the thickness is high for both

Advanced Surfactant-Modified Wet Anisotropic Etching 141

However, for {100} the thickness is only slightly larger than without agitation, indicating little change of the surface properties. The error bars are included in figure 3.6 to stress the

Although our study shows that a larger adsorption of Triton molecules can be obtained by using ultrasonic agitation, especially on {110}, the oscillating force results in large fluctuations in the surfactant thickness. An inhomogeneously adsorbed layer is considered a

Fig. 9. Ellipsometry study: (a) thickness of the surfactant layer obtained in 1 vol% Triton as a function of the number of dipping in DI water at Room temperature; (b) surfactant layer thickness as a function of Surfactant Bath Time (SBT); (c) Arrhenius plot of the saturation thickness of surfactant; (d) comparison of the thickness of surfactant attached on the silicon surface with ultrasonic agitation at room temperature (Reproduced with permission from

In order to observe the effect of the adsorbed surfactant layer on the etched silicon surface morphology and etch rate, 10 wt% TMAH etchant is used at two different temperatures (room temperature and 60 ºC, for both the Triton bath and the etching in TMAH) with ND = 3. The etch rates and etched surface morphology of {110} and {100} with pre-adsorbed Triton

layer of different thicknesses are shown in Fig. 10 (etch depth ~ 30 ± 3 µm).

**3.2 Effect of the adsorbed surfactant layer on etching** 

disadvantage in terms of repeatability and surface roughness control.

larger variations as compared to no agitation.

Elsevier).

orientations. For these particular measurements, the thickness changes significantly between ND = 2 and ND = 3, although it does not change much between ND = 1 and ND = 2. Above ND = 3, the thickness remains finite and constant for {110}, suggesting that the surfactant layers are strongly adsorbed –at least they cannot be easily removed by rinsing– while the thickness is vanishingly small for {100}, indicating a weaker adsorption as the surfactant layer can be easily removed. Thus, we conclude that ND = 3 is the optimal number of dippings and use it for all subsequent experiments. For ND = 3, the typical deviation in the layer thickness is roughly ±1 Å.

The surfactant layer thickness as a function of the Surfactant Bath Time (SBT) is presented in Fig. 9(b) (only ND = 3 is concerned) for two different surfactant bath temperatures (Room Temperature and 60 ºC). The adsorption of the surfactant on both Si {100} and Si {110} are saturation processes, characterized by a saturation time (τsat) and a saturation thickness (hsat). For non-ionic surfactants, the adsorption kinetics involve an initial fast depletion of the surfactant molecules immediately bordering the interface, followed by the diffusion of surfactant molecules from the bulk etchant to the interface and, finally, a rearrangement of the adsorbed surfactant molecules into a final packing structure. Although the actual shape of the monotonic increase of the thickness before saturation should contain valuable information about the manner how the surfactant is packed into a layer, such a study is out of the scope of the present work. From the figure, it is apparent that for the nearly same temperature {100} develops a thinner surfactant layer than {110}. The saturation thickness will become larger for smaller ND values. This observation is good in conformity with the less temperature-dependence of non-ionic surfactant adsorption.

Fig. 9(c) shows an Arrhenius plot of the saturation thickness against inverse temperature. The Arrhenius equation (h α e-Ea/KT) gives the dependence of adsorption on absolute temperature and activation energy. We find that the apparent activation energy is Ea ~ 0.15 eV for {110} while Ea ~ 0.01 eV for {100}, showing that there is clearly bigger barrier to be overcome in the adsorption process of surfactant molecules on the {110} surface. The larger activation energy for {110} suggests that chemisorption may play a role on these surfaces, whereas the small activation energy for {100} indicates that only physisorption is involved for these surfaces.

From the Arrhenius plot, it is observed that adsorption is dominantly thermal for {110}, while nearly athermal for {100}. The increase in adsorption density with temperature is mostly due to the reduction in the size of the hydration shells surrounding the surfactant molecules, especially around the hydrophilic polyethylene oxide chains. The adsorption process is a trade-off between (i) the energy reduction obtained through adsorption by increasing the number of contacts between the hydrophobic alkyl chains and the surface, and (ii) the entropy increase obtained by remaining dissolved in a more disordered state. The hydration shells are smaller at higher temperatures, resulting in less water becoming ordered per adsorbed molecule, thus allowing a larger number of adsorbed surfactant molecules at higher temperature.

Agitation is a key method that can significantly affect the wet etching quality, including the etch rate and surface morphology. The etching properties of etched surfaces with ultrasonic agitation are satisfactory and superior to no agitation. Fig. 9(d) shows the surfactant layer thickness obtained using 110 W ultrasonic cleaner (VS-D100) during residence in the surfactant bath at room temperature. After adding ultrasonic agitation, the surfactant layer thickness for {110} increases with respect to no agitation (cf. figure 3.4), indicating that appropriate forced convection can improve the adsorption of surfactant molecules.

orientations. For these particular measurements, the thickness changes significantly between ND = 2 and ND = 3, although it does not change much between ND = 1 and ND = 2. Above ND = 3, the thickness remains finite and constant for {110}, suggesting that the surfactant layers are strongly adsorbed –at least they cannot be easily removed by rinsing– while the thickness is vanishingly small for {100}, indicating a weaker adsorption as the surfactant layer can be easily removed. Thus, we conclude that ND = 3 is the optimal number of dippings and use it for all subsequent experiments. For ND = 3, the typical deviation in the

The surfactant layer thickness as a function of the Surfactant Bath Time (SBT) is presented in Fig. 9(b) (only ND = 3 is concerned) for two different surfactant bath temperatures (Room Temperature and 60 ºC). The adsorption of the surfactant on both Si {100} and Si {110} are saturation processes, characterized by a saturation time (τsat) and a saturation thickness (hsat). For non-ionic surfactants, the adsorption kinetics involve an initial fast depletion of the surfactant molecules immediately bordering the interface, followed by the diffusion of surfactant molecules from the bulk etchant to the interface and, finally, a rearrangement of the adsorbed surfactant molecules into a final packing structure. Although the actual shape of the monotonic increase of the thickness before saturation should contain valuable information about the manner how the surfactant is packed into a layer, such a study is out of the scope of the present work. From the figure, it is apparent that for the nearly same temperature {100} develops a thinner surfactant layer than {110}. The saturation thickness will become larger for smaller ND values. This observation is good in conformity with the

Fig. 9(c) shows an Arrhenius plot of the saturation thickness against inverse temperature. The Arrhenius equation (h α e-Ea/KT) gives the dependence of adsorption on absolute temperature and activation energy. We find that the apparent activation energy is Ea ~ 0.15 eV for {110} while Ea ~ 0.01 eV for {100}, showing that there is clearly bigger barrier to be overcome in the adsorption process of surfactant molecules on the {110} surface. The larger activation energy for {110} suggests that chemisorption may play a role on these surfaces, whereas the small activation energy for {100} indicates that only physisorption is involved

From the Arrhenius plot, it is observed that adsorption is dominantly thermal for {110}, while nearly athermal for {100}. The increase in adsorption density with temperature is mostly due to the reduction in the size of the hydration shells surrounding the surfactant molecules, especially around the hydrophilic polyethylene oxide chains. The adsorption process is a trade-off between (i) the energy reduction obtained through adsorption by increasing the number of contacts between the hydrophobic alkyl chains and the surface, and (ii) the entropy increase obtained by remaining dissolved in a more disordered state. The hydration shells are smaller at higher temperatures, resulting in less water becoming ordered per adsorbed molecule, thus allowing a larger number of adsorbed surfactant

Agitation is a key method that can significantly affect the wet etching quality, including the etch rate and surface morphology. The etching properties of etched surfaces with ultrasonic agitation are satisfactory and superior to no agitation. Fig. 9(d) shows the surfactant layer thickness obtained using 110 W ultrasonic cleaner (VS-D100) during residence in the surfactant bath at room temperature. After adding ultrasonic agitation, the surfactant layer thickness for {110} increases with respect to no agitation (cf. figure 3.4), indicating that appropriate forced convection can improve the adsorption of surfactant molecules.

less temperature-dependence of non-ionic surfactant adsorption.

layer thickness is roughly ±1 Å.

for these surfaces.

molecules at higher temperature.

However, for {100} the thickness is only slightly larger than without agitation, indicating little change of the surface properties. The error bars are included in figure 3.6 to stress the larger variations as compared to no agitation.

Although our study shows that a larger adsorption of Triton molecules can be obtained by using ultrasonic agitation, especially on {110}, the oscillating force results in large fluctuations in the surfactant thickness. An inhomogeneously adsorbed layer is considered a disadvantage in terms of repeatability and surface roughness control.

Fig. 9. Ellipsometry study: (a) thickness of the surfactant layer obtained in 1 vol% Triton as a function of the number of dipping in DI water at Room temperature; (b) surfactant layer thickness as a function of Surfactant Bath Time (SBT); (c) Arrhenius plot of the saturation thickness of surfactant; (d) comparison of the thickness of surfactant attached on the silicon surface with ultrasonic agitation at room temperature (Reproduced with permission from Elsevier).

#### **3.2 Effect of the adsorbed surfactant layer on etching**

In order to observe the effect of the adsorbed surfactant layer on the etched silicon surface morphology and etch rate, 10 wt% TMAH etchant is used at two different temperatures (room temperature and 60 ºC, for both the Triton bath and the etching in TMAH) with ND = 3. The etch rates and etched surface morphology of {110} and {100} with pre-adsorbed Triton layer of different thicknesses are shown in Fig. 10 (etch depth ~ 30 ± 3 µm).

Advanced Surfactant-Modified Wet Anisotropic Etching 143

In a similar manner, the thickness of the surfactant layer above the silicon surface has an effect on the etch rate in TMAH solutions after the surfactant pretreatment. For {110}, there is a significant reduction in the etch rate for thin surfactant layers. However, for {100} the reduction in the etch rate is only moderate. While {110} has a higher etch rate than {100} in pure TMAH in Fig. 10(b), the pre-treated samples show an opposite behavior, with {100} faster than {110}. This behavior is similar as for directly etching in solutions of Triton added TMAH, indicating that the dissolved surfactant is adsorbed on the surface during etching, as recently shown in FT-IR experiment. Compared with room temperature, there is a less sudden reduction in the etch rate of Si {110} wafers at 60 ºC. In the case of {100}, a moderate etch rate reduction is observed at 60 ºC (as for room temperature). The saturated layer thickness for {110} (≈ 37 Å) produces a very smooth silicon surface (Ra ≈ 20 nm or less).

Significantly different etching behaviour of TMAH + Triton from that of traditionally used anisotropic etchants is very useful for MEMS applications in order to extend the range of 3D structures fabricated by wet etching because the surfactant is adsorbed at the silicon-etchant interface as a thin layer to act as a filter moderating the etching behaviour. In this chapter, we present three applications using surfactant-modified etchants and point out its great

The corner compensation method is the most widely used method for fabricating the sharp edge convex corners. The design and dimensions of the compensation structures require the knowledge of the undercutting ratio and its dependence on the etchant. If the design of MEMS structures does not include any rounded concave and/or sharp convex corners but a smooth etched surface is necessarily required, then the high concentration (20–25 wt% TMAH) should be selected for anisotropic etching. If the structures comprise rounded concave and sharp edge convex corners, the pure TMAH cannot be used due to severe

Fig. 11. Conformal mesa shape fabricated in surfactant-modified solutions.

**4. Applications on MEMS** 

**4.1 Conformal structures** 

undercutting.

potential on advanced MEMS structures.

Fig. 10. Effect of pre-adsorbed surfactant layer (1 vol% Triton in DI water) on the surface roughness (Ra) and etch rate for Si {110} and Si {100} in 10 wt% TMAH: (a) room temperature; and (b) 60 ºC.

For room temperature, the dramatic transformation in the surface morphology correlates directly with a strong reduction in the measured surface roughness. For {110}, it is found that typical zigzag structures emerge in pure TMAH while short pre-treatment in Triton (producing a layer thickness of 12 Å) drastically smoothens the surface, in spite of the reduced thickness of the surfactant layer. The saturated layer thickness for {110} (≈ 16 Å) produces a very smooth silicon surface (Ra ≈ 20 nm). For {100}, the surfactant pre-treatment provides some improvement in the morphology, even when the initial surface is already very smooth. Similar experiments carried out at 60 ºC are shown in Fig. 10(b) (etch depth ~ 35 ± 3 µm). Although {110} shows similar surface morphologies at 60 ºC and RT, the surface morphology of {100} is very different, characterized by the formation of pyramidal hillocks. Nevertheless, the roughness of both {110} and {100} is improved by the use of the surfactant layer.

In a similar manner, the thickness of the surfactant layer above the silicon surface has an effect on the etch rate in TMAH solutions after the surfactant pretreatment. For {110}, there is a significant reduction in the etch rate for thin surfactant layers. However, for {100} the reduction in the etch rate is only moderate. While {110} has a higher etch rate than {100} in pure TMAH in Fig. 10(b), the pre-treated samples show an opposite behavior, with {100} faster than {110}. This behavior is similar as for directly etching in solutions of Triton added TMAH, indicating that the dissolved surfactant is adsorbed on the surface during etching, as recently shown in FT-IR experiment. Compared with room temperature, there is a less sudden reduction in the etch rate of Si {110} wafers at 60 ºC. In the case of {100}, a moderate etch rate reduction is observed at 60 ºC (as for room temperature). The saturated layer thickness for {110} (≈ 37 Å) produces a very smooth silicon surface (Ra ≈ 20 nm or less).
