**4. Dry etch process optimization using RIE**

The first batch of wafers was used to optimize the release of the final device. For this purpose whole fabrication process was skipped and only the mask layers which were required for the fabrication of cantilever beams were used. For optimizing the release process Aluminum (Al) metal was used as it was readily available rather than expensive Au metal. Once the release process was carefully optimized, Au layer was used for the final fabrication of RF MEMS switches.

A 2.5μm thick layer of photo-resist (AZ6612) was deposited and then patterned for anchor. A 1.5μm thick layer of evaporated Al was deposited using e-beam evaporator. A layer of 1.0μm of photo-resist was deposited to pattern the cantilever beam. Two approaches were

Plasma Based Dry Release of MEMS Devices 277

the resist from top and sides of the beam but under the beam resist was observed. Now, the power and pressure parameters of the RIE tool were changed from high power to low power and from low pressure to high pressure. This created an isotropic behavior of the plasma instead of anisotropic behavior which was observed in the first setting. Sample was

(a) (b)

The fabrication of the RF MEMS switches is a six mask all metal fabrication process, as shown in figure 8. All processing steps are developed on the basis of standard CMOS processing. A standard one-layer photo-resist was used as a mask during the fabrication process to provide precise pattern definition. However, during release step the photo-resist was also used as a sacrificial layer. The photo-resist (AZ6612) was a positive photo-resist sensitive to ultraviolet (UV) radiation and can be developed with AZ-300 MIF solution. Throughout the fabrication process, alignment was performed with Quintal Q-6000 mask aligner with UV light exposure. In order to achieve good RF performance device, the switch

The fabrication is a six mask all metal process. The process started with the standard wafer cleaning process. DC bias lines and actuation pads were defined by evaporating 0.04µm layer of Cromium (Cr). This layer was then patterned with mask one. The Cr metal evaporation was done using Lesker evaporator operating in 10-6 Torr range. An insulator layer in a series switch served as a mechanical connection as well as electrical isolation between the actuator and the contact. Since the switch was made of metal, the insulator layer also acted as a dielectric layer which was needed to prevent direct contact between the

A 0.75µm thick layer of silicon nitride was deposited as dielectric layer using PECVD and patterned with mask two. The deposition of Si3N4 was performed using VACUTEC-1500 series PECVD equipment. The CPW lines were defined by evaporating/RF sputtering of 0.04/1.0µm thick layer of Cr/Au and patterned with mask three. Cr was used as an adhesion layer between the Au and substrate. The sputtering was performed using Edwards

was fabricated on a low loss alumina substrate with dielectric constant 9.9.

E-306 series sputtering tool which was used for RF sputtering of the Au film.

metal cantilever bridge and the actuation pad.

Fig. 7. SEM of discolored/damaged areas after RIE (a) bias line bridge (b) cantilever beam

**5. Fabrication process** 

exposed to plasma for one hour which resulted in a clean release of the structures.

used to dry release the cantilever beams. First, one wafer was used to directly dry etch the sacrificial layer of photo-resist without using any other process on it. In this approach, it was observed that RIE tool was not able to remove the sacrificial layer material sufficiently from the devices. Some leftover metal residues were also observed which could not be cleaned even after extensive DI water rinsing of the wafer. A prolonged exposure of anisotropic RIE also damaged the cantilever beam structures.

In the second approach, a combination of wet and dry release was used to remove the leftover metal residues after etching while replacing a new layer of photo-resist as a supporting layer. The sample was inspected under microscope followed by SEM and in this case metal residues were not observed on the sample or supporting layer. Wafer was exposed under O2 plasma in RIE chamber for dry etch. The wafer was exposed to high RF power (70W) and high pressure (30Pa). The high RF power generated an intense bombardment of plasma atoms with high pressure.

Devices were inspected under the microscope after the first etching exposure. It was observed that although plasma etched the photo-resist from top and sides of the device a significant amount of resist was observed under the cantilever beam. A second exposure of plasma was given again to the samples. After second exposure of plasma it was observed that resist was still visible under the beam. However, the beam structures were discoloured. It was assumed that some resist was still on the beam which created this discolourization. However, when the samples were observed under SEM, it showed that this discolourization was not due to resist but the beam structures were damaged due to high power plasma particles.

Figure 7(a) shows the SEM image of bridge over bias line with damaged surface. A metal peeling from some parts of the bridge is also visible. One can observe that high power bombardment of plasma atoms has damaged the metal layers on the device. Figure 7(b) shows the cantilever beam structure damaged due to RIE plasma while optimizing the release process.

After a number of iterations, it was revealed that power and pressure were the main factors for the optimization of dry release process. Variation of power and pressure from high to low and vise versa can change the plasma behavior inside the chamber. The voltage bias was also controlled once these parameters were changed. With high RF power and low pressure we achieved a bias of 232±6V which indicated that plasma particles generated inside the chamber strike the surface of the substrate with more power giving anisotropic etching behavior. With low power and high pressure the bias changed to as low as 90±6V which changed the plasma atoms behavior from anisotropic to isotropic giving etch profile below the surface of the cantilever beam also.

Three wafer samples were used to optimize the RIE process using the supporting layer technique. In this case, a careful shifting of power and pressure parameters was done. Once sample was ready with supporting layer of photo-resist underneath the cantilever beam, the sample was exposed to high power and low pressure for one hour. In this case, the plasma particles struck the wafer surface with high power but under low pressure. It did not damage the device surface. The sample was observed under the microscope and a significant amount of photo-resist was observed on the sample as well as under the beam. The sample was exposed to plasma for 30min and then inspected again under microscope. It was observed that much of the resist from top of the beam and sides was removed but a small amount of resist was still visible on the beam and significant amount of resist was observed under the beam. Sample was exposed to another 30min exposure which cleaned

used to dry release the cantilever beams. First, one wafer was used to directly dry etch the sacrificial layer of photo-resist without using any other process on it. In this approach, it was observed that RIE tool was not able to remove the sacrificial layer material sufficiently from the devices. Some leftover metal residues were also observed which could not be cleaned even after extensive DI water rinsing of the wafer. A prolonged exposure of anisotropic RIE

In the second approach, a combination of wet and dry release was used to remove the leftover metal residues after etching while replacing a new layer of photo-resist as a supporting layer. The sample was inspected under microscope followed by SEM and in this case metal residues were not observed on the sample or supporting layer. Wafer was exposed under O2 plasma in RIE chamber for dry etch. The wafer was exposed to high RF power (70W) and high pressure (30Pa). The high RF power generated an intense

Devices were inspected under the microscope after the first etching exposure. It was observed that although plasma etched the photo-resist from top and sides of the device a significant amount of resist was observed under the cantilever beam. A second exposure of plasma was given again to the samples. After second exposure of plasma it was observed that resist was still visible under the beam. However, the beam structures were discoloured. It was assumed that some resist was still on the beam which created this discolourization. However, when the samples were observed under SEM, it showed that this discolourization was not due to resist but the beam structures were damaged due to high power plasma

Figure 7(a) shows the SEM image of bridge over bias line with damaged surface. A metal peeling from some parts of the bridge is also visible. One can observe that high power bombardment of plasma atoms has damaged the metal layers on the device. Figure 7(b) shows the cantilever beam structure damaged due to RIE plasma while optimizing the

After a number of iterations, it was revealed that power and pressure were the main factors for the optimization of dry release process. Variation of power and pressure from high to low and vise versa can change the plasma behavior inside the chamber. The voltage bias was also controlled once these parameters were changed. With high RF power and low pressure we achieved a bias of 232±6V which indicated that plasma particles generated inside the chamber strike the surface of the substrate with more power giving anisotropic etching behavior. With low power and high pressure the bias changed to as low as 90±6V which changed the plasma atoms behavior from anisotropic to isotropic giving etch profile

Three wafer samples were used to optimize the RIE process using the supporting layer technique. In this case, a careful shifting of power and pressure parameters was done. Once sample was ready with supporting layer of photo-resist underneath the cantilever beam, the sample was exposed to high power and low pressure for one hour. In this case, the plasma particles struck the wafer surface with high power but under low pressure. It did not damage the device surface. The sample was observed under the microscope and a significant amount of photo-resist was observed on the sample as well as under the beam. The sample was exposed to plasma for 30min and then inspected again under microscope. It was observed that much of the resist from top of the beam and sides was removed but a small amount of resist was still visible on the beam and significant amount of resist was observed under the beam. Sample was exposed to another 30min exposure which cleaned

also damaged the cantilever beam structures.

bombardment of plasma atoms with high pressure.

below the surface of the cantilever beam also.

particles.

release process.

the resist from top and sides of the beam but under the beam resist was observed. Now, the power and pressure parameters of the RIE tool were changed from high power to low power and from low pressure to high pressure. This created an isotropic behavior of the plasma instead of anisotropic behavior which was observed in the first setting. Sample was exposed to plasma for one hour which resulted in a clean release of the structures.

Fig. 7. SEM of discolored/damaged areas after RIE (a) bias line bridge (b) cantilever beam

#### **5. Fabrication process**

The fabrication of the RF MEMS switches is a six mask all metal fabrication process, as shown in figure 8. All processing steps are developed on the basis of standard CMOS processing. A standard one-layer photo-resist was used as a mask during the fabrication process to provide precise pattern definition. However, during release step the photo-resist was also used as a sacrificial layer. The photo-resist (AZ6612) was a positive photo-resist sensitive to ultraviolet (UV) radiation and can be developed with AZ-300 MIF solution. Throughout the fabrication process, alignment was performed with Quintal Q-6000 mask aligner with UV light exposure. In order to achieve good RF performance device, the switch was fabricated on a low loss alumina substrate with dielectric constant 9.9.

The fabrication is a six mask all metal process. The process started with the standard wafer cleaning process. DC bias lines and actuation pads were defined by evaporating 0.04µm layer of Cromium (Cr). This layer was then patterned with mask one. The Cr metal evaporation was done using Lesker evaporator operating in 10-6 Torr range. An insulator layer in a series switch served as a mechanical connection as well as electrical isolation between the actuator and the contact. Since the switch was made of metal, the insulator layer also acted as a dielectric layer which was needed to prevent direct contact between the metal cantilever bridge and the actuation pad.

A 0.75µm thick layer of silicon nitride was deposited as dielectric layer using PECVD and patterned with mask two. The deposition of Si3N4 was performed using VACUTEC-1500 series PECVD equipment. The CPW lines were defined by evaporating/RF sputtering of 0.04/1.0µm thick layer of Cr/Au and patterned with mask three. Cr was used as an adhesion layer between the Au and substrate. The sputtering was performed using Edwards E-306 series sputtering tool which was used for RF sputtering of the Au film.

Plasma Based Dry Release of MEMS Devices 279

explained in last section, in single dry release process, the problem of left over residues of metal after etching the metal layer was experienced. So a unique dry release process was

Motivation for this unique process was that some left over residues were observed after the single step or traditional RIE process. Secondly, this process was more cost effective as compared to a wet release CPD technique using CO2 dryer. The process not only produced less residual waste but achieved a clean dry release. The steps for dry release process are

First, the sacrificial layer was removed using acetone. This also included the removal of some Au leftover residues on photo-resist from the previous wet etching with mask 6 [figure 9(a)]. After this, sample was dipped again into clean acetone for 30 min for final cleaning. Then the structure was immediately dipped into another resist (AZ5214E), until all the liquid covering the sample was concentrated resist [figure 9(b)] (Forsen et al., 2004 & Orpana & Korhonen, 1991). The resist covered sample was spun at 2500 rpm to achieve uniform layer of resist and then soft backed at 90ºC resulting in a thick layer of photo-resist

It must be noted that the wafer was never allowed to dry during the the process or else structure would be permanently bonded to the substrate. The structure was then dry released by Oxygen plasma using the single process RIE in two steps. In step one the etching was done using high power and low pressure (15sccm O2, 180 W, 8 Pa) giving an anisotropic etch of the photo-resist [figure 9(c)]. In step two low power and high pressure (15sccm O2, 50 W, 40 Pa) was used. This resulted in isotropic etching of the photo-resist thus giving a free

(a) (b)

(c) (d)

Fig. 9. Schematic representation of process steps involved in dry release process of MEMS structures (a) patterned cantilever beam over sacrificial layer of AZ-6612 (b) cantilever beam

dipped in structural layer of AZ-5214E (c) anisotropic etching (d) isotropic etching

developed with a combination of wet and dry release to achieve better results.

fully encapsulating the suspended beam as a supporting layer.

standing structure at the end [figure 9(d)].

**6.1 Dry release model** 

described in figure 9.

Fig. 8. Six mask fabrication process for RF MEMS switches

Then a 2.5µm thick layer of photo-resist (AZ-6632) was deposited as sacrificial layer and patterned for anchor and dimple with mask four and five respectively. While defining the anchor and dimple full dark masks were used to expose only the anchor and dimple areas. This was followed by a 1.5µm thick layer of RF sputtered Au which was patterned with mask six to form the cantilever beam. Finally, the bridge structure was released using a unique dry release process.

#### **6. Dry release process**

During fabrication of RF MEMS switches both dry and wet release methods were applied. The yield of wet release was very low and no working prototype was achieved. Problems related to wet release and stiction have already been discussed in the previous sections. As explained in last section, in single dry release process, the problem of left over residues of metal after etching the metal layer was experienced. So a unique dry release process was developed with a combination of wet and dry release to achieve better results.

#### **6.1 Dry release model**

278 Microelectromechanical Systems and Devices

Fig. 8. Six mask fabrication process for RF MEMS switches

unique dry release process.

**6. Dry release process** 

Then a 2.5µm thick layer of photo-resist (AZ-6632) was deposited as sacrificial layer and patterned for anchor and dimple with mask four and five respectively. While defining the anchor and dimple full dark masks were used to expose only the anchor and dimple areas. This was followed by a 1.5µm thick layer of RF sputtered Au which was patterned with mask six to form the cantilever beam. Finally, the bridge structure was released using a

During fabrication of RF MEMS switches both dry and wet release methods were applied. The yield of wet release was very low and no working prototype was achieved. Problems related to wet release and stiction have already been discussed in the previous sections. As Motivation for this unique process was that some left over residues were observed after the single step or traditional RIE process. Secondly, this process was more cost effective as compared to a wet release CPD technique using CO2 dryer. The process not only produced less residual waste but achieved a clean dry release. The steps for dry release process are described in figure 9.

First, the sacrificial layer was removed using acetone. This also included the removal of some Au leftover residues on photo-resist from the previous wet etching with mask 6 [figure 9(a)]. After this, sample was dipped again into clean acetone for 30 min for final cleaning. Then the structure was immediately dipped into another resist (AZ5214E), until all the liquid covering the sample was concentrated resist [figure 9(b)] (Forsen et al., 2004 & Orpana & Korhonen, 1991). The resist covered sample was spun at 2500 rpm to achieve uniform layer of resist and then soft backed at 90ºC resulting in a thick layer of photo-resist fully encapsulating the suspended beam as a supporting layer.

It must be noted that the wafer was never allowed to dry during the the process or else structure would be permanently bonded to the substrate. The structure was then dry released by Oxygen plasma using the single process RIE in two steps. In step one the etching was done using high power and low pressure (15sccm O2, 180 W, 8 Pa) giving an anisotropic etch of the photo-resist [figure 9(c)]. In step two low power and high pressure (15sccm O2, 50 W, 40 Pa) was used. This resulted in isotropic etching of the photo-resist thus giving a free standing structure at the end [figure 9(d)].

Fig. 9. Schematic representation of process steps involved in dry release process of MEMS structures (a) patterned cantilever beam over sacrificial layer of AZ-6612 (b) cantilever beam dipped in structural layer of AZ-5214E (c) anisotropic etching (d) isotropic etching

Plasma Based Dry Release of MEMS Devices 281

From the SEM images and optical microscopy it was observed that the released beam structure showed higher curling up trend. This was due to residual gradient stress in the film and lead to the increase in the actuation voltage. The stress gradient lead to the lift of beam around 1µm after the release of structure. The measured lift of cantilever front end is 4.3µm after release. Figure 13 shows a DEKTAK profile of the unreleased and released beam tip of RF MEMS switch. In figure 13(a) DEKTAK profile indicates the beam height after patterning mask six which also confirms the gap height distance of 2.5μm. When beam structure was released using RIE plasma technique, the lift of the front tip of the cantilever beams was measured again which confirms the curling up trends of the beam stated above.

(a) (b)

The yield of the released structures on the wafer was measured using visual examination and SEM. No stiction was observed with new release process. However presence of some residues was observed on the outer samples of the substrate. This was due to non uniform plasma distribution during the RIE. The yield of the release process was worked out on full cleaned samples. A yield of more than 70% was achieved with contact resistance of less than 2.7Ω.

Fig. 13. DEKTKK image of beam tip (a) before release (b) after release

**6.2.1 Yield** 

Fig. 12. Released RF MEMS switch cantilever beam (front tip view)

### **6.2 Dry release using RIE**

Figure 10 displays a SEM image of the fabricated switch. The sacrificial layer (AZ-6612) has been removed after two dips in acetone; supporting layer below the structure has been made with another photoresist (AZ5214E). It can be observed that structure has got a clear standing on the supporting layer. There is no indication of left over residues of the Au after acetone cleaning.

Fig. 10. SEM of the RF MEMS switch with supporting layer

Figure 11 shows a SEM image of the structure after anisotropic etching during the first step of single process RIE. The structure rests on supporting layer. Some leftover parts of the chemical waste are also visible. The chemical waste observed during the dry etching was comparatively less than as seen in the wet etching.

Fig. 11. SEM of the cantilever beam structure resting on supporting layer after anisotropic etch

During the isotropic etching step of RIE, plasma moves in all directions and etches the photo-resist layer located below the cantilever beam structure. Figure 12 shows the released RF MEMS cantilever beam structures. The clean standing structure of the MEMS bridge can be observed. The release of structure was clean and results achieved by this process technique were satisfactory.

Figure 10 displays a SEM image of the fabricated switch. The sacrificial layer (AZ-6612) has been removed after two dips in acetone; supporting layer below the structure has been made with another photoresist (AZ5214E). It can be observed that structure has got a clear standing on the supporting layer. There is no indication of left over residues of the Au after

Figure 11 shows a SEM image of the structure after anisotropic etching during the first step of single process RIE. The structure rests on supporting layer. Some leftover parts of the chemical waste are also visible. The chemical waste observed during the dry etching was

Fig. 11. SEM of the cantilever beam structure resting on supporting layer after anisotropic etch During the isotropic etching step of RIE, plasma moves in all directions and etches the photo-resist layer located below the cantilever beam structure. Figure 12 shows the released RF MEMS cantilever beam structures. The clean standing structure of the MEMS bridge can be observed. The release of structure was clean and results achieved by this process

**6.2 Dry release using RIE** 

Fig. 10. SEM of the RF MEMS switch with supporting layer

comparatively less than as seen in the wet etching.

technique were satisfactory.

acetone cleaning.

Fig. 12. Released RF MEMS switch cantilever beam (front tip view)

From the SEM images and optical microscopy it was observed that the released beam structure showed higher curling up trend. This was due to residual gradient stress in the film and lead to the increase in the actuation voltage. The stress gradient lead to the lift of beam around 1µm after the release of structure. The measured lift of cantilever front end is 4.3µm after release. Figure 13 shows a DEKTAK profile of the unreleased and released beam tip of RF MEMS switch. In figure 13(a) DEKTAK profile indicates the beam height after patterning mask six which also confirms the gap height distance of 2.5μm. When beam structure was released using RIE plasma technique, the lift of the front tip of the cantilever beams was measured again which confirms the curling up trends of the beam stated above.

Fig. 13. DEKTKK image of beam tip (a) before release (b) after release

#### **6.2.1 Yield**

The yield of the released structures on the wafer was measured using visual examination and SEM. No stiction was observed with new release process. However presence of some residues was observed on the outer samples of the substrate. This was due to non uniform plasma distribution during the RIE. The yield of the release process was worked out on full cleaned samples. A yield of more than 70% was achieved with contact resistance of less than 2.7Ω.

Plasma Based Dry Release of MEMS Devices 283

Figure 16 shows the beam labeled Design-2, with three supported cantilever bars and an extended cantilever at the front. The dimensions of the three supported cantilever beams are the same as that for Design-1, with the centre 60µm×60µm and the extended cantilever at the

Fig. 15. SEM of fabricated RF MEMS switch

Fig. 16. SEM of fabricated RF MEMS switch

front 40µm×20µm.
