**4. Sample preparation and microfabrication techniques**

Sample preparation is one of the main challenges in tensile testing of thin film specimens. Thin film materials are usually fabricated using one of the deposition techniques. In order to utilize any of these techniques to fabricate free-standing thin film "dog-bone" specimens, a designated microfabrication process has to be developed. This process depends on the specific requirements defined by the choice of gripping and sample handling method, the film material and deposition technique, and the availability of the specific procedures in any fabrication laboratory.

Ruud et al., 1993 used a relatively simple technique to fabricate free-standing films. They evaporated Cu and Ag films on glass substrate and after patterning the film to a dog-bone shaped specimen, they took the films off the substrate by sliding a razor blade underneath them while submerged in water. For Ni films, glass substrate was first coated with a layer of photoresist and Ni was then sputter deposited on it. The film was then released by etching the resist in acetone. Although both processes developed by them are relatively simple, films are prone to be damaged and wrinkle while releasing.

The concept of using a window frame in the substrate which was originally introduced by Read & Dally, 1993 was among the most popular methods that was used and further developed by other researchers. In this process, double-sided polished (DSP) <100> silicon wafers, were first coated by a thin layer of silicon oxide. Oxide layer on front-side was patterned and etched at specimen locations. Thin film material of interest was then deposited by e-beam evaporation and patterned to dog-bone shape specimens. The oxide layer on both sides was then patterned and etched in HF to form a hard mask for silicon substrate etching. Silicon was then etched in hydrazine to open window frames. Sharpe et al., 2003 utilized this technique to test thin polysilicon films. Figure 1-a shows a silicon carbide specimen that was fabricated by Edwards et al., 2004 using this concept. Emery & Povirk, 2003a, 2003b used the same process to fabricate e-beam evaporated gold. The main issue with this technique is the long Si substrate etching times that may cause the specimen film be attacked during etching process and special care is required in this regard.

Cornella, 1999 improved this concept by using dry etching processes rather than wet etching processes to fabricate specimens with higher film quality and process yield. In their process, Si substrate was first coated on front side with 1μm thick LPCVD silicon nitride to be used as an etch stop. Aluminum was then sputter deposited on front side and patterned to the dog-bone shape. Backside of the substrate was coated with thick photoresist to act as the

Standalone Tensile Testing of Thin Film Materials for MEMS/NEMS Applications 441

Fig. 2. Microfabrication process for electroplated (A), and evaporated (B) dog-bone

deposited via lift-off process (i), PMMA is spun coated and patterned through

patterned to the dog bone shape specimen (ii), a thick Au layer is deposited via

(ii), and the sacrificial layer is then removed in stripper (iii).

**5. Gripping** 

issue.

specimens. (Chasiotis et al., 2007) In brief, for electroplated specimens (A), Ti/Au anchor is

electroplating to realize the final dog bone specimen (iii), PMMA sacrificial layer is etched to release the structure (iv). In the case of evaporated films (B), photoresists is used as the sacrificial layer and patterned via photolithography (i), Au film is evaporated and patterned

Gripping a film that usually has smaller thickness than even the surface roughness of the macro-machined grippers is a tough challenge. Under these circumstances, the film may slip or experience high stresses at the gripping location due to stress concentrations. On the other hand, aligning the two grippers is, in fact, a demanding task. Therefore, many researchers have designed and utilized a variety of gripping techniques to overcome this

Ruud et al., 1993 sandwiched the free-standing thin film specimens between polished aluminum grippers using 5μm thick Cu foils. The concept of window frame in substrate (Read & Dally, 1993; Cornella, 1999) has made gripping much easier and common macromachined grippers can be used to mount thick end grips of the specimen which is basically the thick silicon substrate rather than the thin film. As shown in Figure 3, Greek & Johnson, 1997 and Greek et al., 1997 used a connecting ring as a gripper. They inserted a probe connected to the load-train setup in the ring and loaded the specimen. Buchheit et al., 2003 used the same concept to pull micromachined silicon films. A cylindrical sapphire nanoindenter tip was inserted into the so-called "pull-tab" and utilizing the lateral loading capability of a nano-indenter, samples were loaded in tension. Emry & Povirk, 2003a, 2003b used the same technique for pulling tensile specimens on a substrate window frame. This

photolithography and RIE etching, and then, a thin Ti/Au seed layer is evaporated and

etching mask during substrate etching. Silicon substrate was entirely dry-etched until it reached silicon nitride layer. This layer was then removed in RIE to release the aluminum specimen. The specimen fabricated through this process is shown in Figure 1-b.

Fig. 1. (a) A free-standing silicon carbide specimen on substrate window frame (Sharpe et al., 2003) and (b) SEM image of an Al free-standing film on window frame. (Zhang et al., 2001)

A few researchers, who mainly worked on the mechanical behavior of polysilicon, used factory processes like polyMUMPS to fabricate their specimens. The specimen is then released by etching the oxide sacrificial layer. Although these processes are well developed and are readily available, they are limited to a few thin film materials, most of which are silicon based. When metallic materials are used as the structural layer, traditional siliconbased films are not good choices for sacrificial layer. A common practice in the fabrication of free-standing metallic devices in RF MEMS devices is to use polymers as the sacrificial layer. Chasiotis et al., 2007 used this technique to fabricate tensile specimens of Au films. As shown in Figure 2, they used PMMA and AZ 4110 photoresist as the sacrificial material for electroplated and evaporated Au films, respectively. For electroplated Au films, a molding process was utilized to pattern gold on PMMA and sacrificial layer was then etched to release the film. Evaporated Au films, however, were lithography patterned and the photoresist sacrificial layer was then stripped to release the structure. Although polymeric sacrificial layers are easier to remove and hence result in less attack to the metallic film, they are less applicable when high temperature processes are involved. In fact in high temperatures two problems arise; above the glass transition temperature, polymer layer starts a significant flow which causes deformation and wrinkling in the metallic film; on the other hand, the thermal mismatch between the polymer and the metallic film causes significant stresses on the film that, in high temperatures, may result in creep and permanent deformation. (Stance et al., 2007) Tajik, 2008 optimized this process in order to realize free-standing thin film specimens that are free of wrinkle and warpage.

Fig. 2. Microfabrication process for electroplated (A), and evaporated (B) dog-bone specimens. (Chasiotis et al., 2007) In brief, for electroplated specimens (A), Ti/Au anchor is deposited via lift-off process (i), PMMA is spun coated and patterned through photolithography and RIE etching, and then, a thin Ti/Au seed layer is evaporated and patterned to the dog bone shape specimen (ii), a thick Au layer is deposited via electroplating to realize the final dog bone specimen (iii), PMMA sacrificial layer is etched to release the structure (iv). In the case of evaporated films (B), photoresists is used as the sacrificial layer and patterned via photolithography (i), Au film is evaporated and patterned (ii), and the sacrificial layer is then removed in stripper (iii).

### **5. Gripping**

440 Microelectromechanical Systems and Devices

etching mask during substrate etching. Silicon substrate was entirely dry-etched until it reached silicon nitride layer. This layer was then removed in RIE to release the aluminum

specimen. The specimen fabricated through this process is shown in Figure 1-b.

Fig. 1. (a) A free-standing silicon carbide specimen on substrate window frame (Sharpe et al., 2003) and (b) SEM image of an Al free-standing film on window frame.

**Silicon Substrate** 

(a) (b)

realize free-standing thin film specimens that are free of wrinkle and warpage.

A few researchers, who mainly worked on the mechanical behavior of polysilicon, used factory processes like polyMUMPS to fabricate their specimens. The specimen is then released by etching the oxide sacrificial layer. Although these processes are well developed and are readily available, they are limited to a few thin film materials, most of which are silicon based. When metallic materials are used as the structural layer, traditional siliconbased films are not good choices for sacrificial layer. A common practice in the fabrication of free-standing metallic devices in RF MEMS devices is to use polymers as the sacrificial layer. Chasiotis et al., 2007 used this technique to fabricate tensile specimens of Au films. As shown in Figure 2, they used PMMA and AZ 4110 photoresist as the sacrificial material for electroplated and evaporated Au films, respectively. For electroplated Au films, a molding process was utilized to pattern gold on PMMA and sacrificial layer was then etched to release the film. Evaporated Au films, however, were lithography patterned and the photoresist sacrificial layer was then stripped to release the structure. Although polymeric sacrificial layers are easier to remove and hence result in less attack to the metallic film, they are less applicable when high temperature processes are involved. In fact in high temperatures two problems arise; above the glass transition temperature, polymer layer starts a significant flow which causes deformation and wrinkling in the metallic film; on the other hand, the thermal mismatch between the polymer and the metallic film causes significant stresses on the film that, in high temperatures, may result in creep and permanent deformation. (Stance et al., 2007) Tajik, 2008 optimized this process in order to

**Specimen** 

(Zhang et al., 2001)

Gripping a film that usually has smaller thickness than even the surface roughness of the macro-machined grippers is a tough challenge. Under these circumstances, the film may slip or experience high stresses at the gripping location due to stress concentrations. On the other hand, aligning the two grippers is, in fact, a demanding task. Therefore, many researchers have designed and utilized a variety of gripping techniques to overcome this issue.

Ruud et al., 1993 sandwiched the free-standing thin film specimens between polished aluminum grippers using 5μm thick Cu foils. The concept of window frame in substrate (Read & Dally, 1993; Cornella, 1999) has made gripping much easier and common macromachined grippers can be used to mount thick end grips of the specimen which is basically the thick silicon substrate rather than the thin film. As shown in Figure 3, Greek & Johnson, 1997 and Greek et al., 1997 used a connecting ring as a gripper. They inserted a probe connected to the load-train setup in the ring and loaded the specimen. Buchheit et al., 2003 used the same concept to pull micromachined silicon films. A cylindrical sapphire nanoindenter tip was inserted into the so-called "pull-tab" and utilizing the lateral loading capability of a nano-indenter, samples were loaded in tension. Emry & Povirk, 2003a, 2003b used the same technique for pulling tensile specimens on a substrate window frame. This

Standalone Tensile Testing of Thin Film Materials for MEMS/NEMS Applications 443

electrostatic gripping, although seems straight forward, is only applicable for static low-load (<0.1 N) tests. The utilization of adhesive layer is necessary when conducting time-

Fig. 4. (a) Schematic representation of electrostatic gripping technique (Tsuchiya et al., 1997) and (b) Combination of the electrostatic and UV adhesive gripping. (Chasiotis & Knauss,

> Grip Point Gripper Arm

Course Stage Fine Actuator

Course Actuator

2002)

Piezo Connection

Fig. 5. The double action gripper (Tajik, 2008)

Laser Beam Pass

dependent tests or applying a dynamic load that requires a reliable no-slip gripper.

(a) (b)

methods, however, is only useful when tension-tension loading scenario is used. In cases where loading direction is changed or set to zero, backlash and rigid displacements cannot be avoided.

Fig. 3. Tensile testing specimen with a ring at one end for gripping and loading (Greek & Johnson, 1997)

Tsuchiya et al., 1997, 1998 introduced a novel technique to grip tensile testing specimens using electrostatic force. This technique is schematically shown in Figure 4-a. In this technique, a free-standing specimen with large end grip (puddle) is fabricated and is fixed to the gripper by electrostatic force. Specimen can be easily fixed to and released from the gripper by changing the polarity of the applied voltage. Sharpe & Bagdahn, 2004 argued that although the electrostatic gripping is very useful in static tensile tests, it fails during tension-tension fatigue testing. To overcome this issue, they glued a silicon carbide fiber to the puddle using viscous UV curable adhesive.

Chasiotis & Knauss, 2002 also reported that specimens mounted by electrostatic gripping slip during long-time static loadings and they experienced rigid-body motion of the specimen during their long-time AFM scans for deformation measurement. It was shown that the electrostatic gripping is only reliable when the applied tensile loads are below 0.1 N for their specimen geometry. They improved the technique by combining the electrostatic actuation with UV adhesive to meet their demanding requirements for a no-slip reliable gripper. (Figure 4-b)

In order to avoid slipping of the film at the gripper, Tajik, 2008 used a novel gripping method that could reliably grip the specimen for different modes of loading. In this method, the conventional serrated jaw macro-machined gripper was mounted on a double action arm. This mechanism provides a tight gripping of thin film micro-specimens, though the gripper itself is of macro-scale size. (Figure 5)

In conclusion, the application of substrate frame window concept makes griping much easier in the expense of having a more complicated specimen fabrication process. The

methods, however, is only useful when tension-tension loading scenario is used. In cases where loading direction is changed or set to zero, backlash and rigid displacements cannot

Fig. 3. Tensile testing specimen with a ring at one end for gripping and loading

**50 μm** 

Tsuchiya et al., 1997, 1998 introduced a novel technique to grip tensile testing specimens using electrostatic force. This technique is schematically shown in Figure 4-a. In this technique, a free-standing specimen with large end grip (puddle) is fabricated and is fixed to the gripper by electrostatic force. Specimen can be easily fixed to and released from the gripper by changing the polarity of the applied voltage. Sharpe & Bagdahn, 2004 argued that although the electrostatic gripping is very useful in static tensile tests, it fails during tension-tension fatigue testing. To overcome this issue, they glued a silicon carbide fiber to

Chasiotis & Knauss, 2002 also reported that specimens mounted by electrostatic gripping slip during long-time static loadings and they experienced rigid-body motion of the specimen during their long-time AFM scans for deformation measurement. It was shown that the electrostatic gripping is only reliable when the applied tensile loads are below 0.1 N for their specimen geometry. They improved the technique by combining the electrostatic actuation with UV adhesive to meet their demanding requirements for a no-slip reliable

In order to avoid slipping of the film at the gripper, Tajik, 2008 used a novel gripping method that could reliably grip the specimen for different modes of loading. In this method, the conventional serrated jaw macro-machined gripper was mounted on a double action arm. This mechanism provides a tight gripping of thin film micro-specimens, though the

In conclusion, the application of substrate frame window concept makes griping much easier in the expense of having a more complicated specimen fabrication process. The

be avoided.

(Greek & Johnson, 1997)

gripper. (Figure 4-b)

the puddle using viscous UV curable adhesive.

gripper itself is of macro-scale size. (Figure 5)

electrostatic gripping, although seems straight forward, is only applicable for static low-load (<0.1 N) tests. The utilization of adhesive layer is necessary when conducting timedependent tests or applying a dynamic load that requires a reliable no-slip gripper.

Fig. 4. (a) Schematic representation of electrostatic gripping technique (Tsuchiya et al., 1997) and (b) Combination of the electrostatic and UV adhesive gripping. (Chasiotis & Knauss, 2002)

Fig. 5. The double action gripper (Tajik, 2008)

Standalone Tensile Testing of Thin Film Materials for MEMS/NEMS Applications 445

Read & Dally, 1993 monitored the cross-head displacement and used it as a measure for strain. There are many sources of error involved in this technique. Specimen may slip at the gripper. On the other hand, the gripper itself may have clearances that cause backlash during the changes in load direction. Compliance of the test setup is the other source that deteriorates the accuracy of the method. The cross-head displacement is a combination of all of the deformations in the load train, i.e. the deformations in load-cell, load actuator, the test rig, grippers and albeit in the specimen itself. Therefore, this measurement will not provide

Cornella, 1999 measured the compliance of the test setup by compressing the load actuator to the load cell in the absence of specimen and subtracted this compliance from the actual measurements to find the deformation in the specimen. They reported that 76% of the measured displacement accounts for the actual deformation in the specimen (Zhang et al., 2006). In order to validate the strain relaxation measurements and to show that the drop in the stress level over time is the actual behavior of the specimen itself and not the test setup, they used iridium specimens. Iridium, due to its high melting point, has a very low relaxation at room temperature. Since these tests revealed no relaxation, they argued that their test setup is stiff enough and that the relaxation behavior that they monitored during

Chasiotis et al., 2007 used the cross-head displacement to study the relaxation in gold thin films. They measured the deformation of the load-cell and the apparatus compliance and subtract it from the results. Due to the high compliance of their specimens compared to the setup, 99% of the cross-head displacement was due to the deformation in the specimen. They verified the accuracy of the crosshead displacement method by testing brittle materials

Greek & Johnson, 1997 cancelled out the effect of the compliance of the test setup by testing specimens with identical gages section areas and different gage lengths. Assuming that the compliance of the test setup is constant for any test, they calculated the deformations caused by test setup and subtract it from the test results. This method is only applicable to the cases that the compliance of the specimens are sufficiently different

Emry & Povirk, 2003a also measured the displacement in grippers by monitoring the displacement of two markers using a video camera. They argued that their method has the limitation that the measured strain is not the actual strain in the gage section. They also reported a non-linearity in the stress-strain curve in low loads. Figure 6 shows this nonlinearity which is technically an experimental error. They extrapolated the linear portion of the results to find the zero point of stress-strain curve. Due to this non-linearity in the curve, calculating the yield stress using the 0.2% offset rule was erroneous. Therefore they found the yield point by defining it as the point where the slope of the stress-strain curve drops to

Due to the uncertainties involved in the application of cross-head displacement for strain measurements, a number of techniques have been introduced to measure the strain directly on the gage section. An inexpensive way of measuring strain is to put markers on the specimen's gage section and monitor their displacement using a camera. Allameh et al., 2004 used a video camera and monitored the deformation of two markers milled by Focused Ion Beam (FIB) on LIGA Ni specimens. Markers were 300μm apart and were located by block matching in a series of images captured by a camera during tensile testing. They have not

an accurate measure of stain in the gage section of the sample.

the tensile testing of Al films is the actual material behavior.

with known elastic modulus.

at different gage lengths.

one tenth of the elastic modulus.
