Ping Zhang,[1] Hoo-Jeong Lee2 and John C. Bravman3
Mechanical Tests of Free-Standing Aluminum Micro Beams for MEMS Application
REFERENCE: Zhang, P., Lee, H. J., and Bravman, J. C., “Mechanical Tests of Free-Standing Aluminum Micro Beams for MEMS Application,” Mechanical Properties of Structural Films, ASTM STP 1413, C. Muhlstein and S. B. Brown, Eds., American Society for Testing and Materials, West Conshohocken, PA, 2001.
ABSTRACT: We studied mechanical properties as well as alloy effects of free-standing Aluminum micro beams (50mm´500mm´2mm) in a piezo-actuator-driven test apparatus with a load resolution of ±0.2mN and a displacement resolution of ±10nm. Pure Al and Al-2%Ti micro beams were fabricated using micromachining techniques. In tensile tests, we found the yield strength to be about 120MPa for the pure Al beams, and approximately 75% higher for the alloyed beams. We examined the results with respect to those of bulk materials and thin films adhered to substrates. In stress relaxation tests, we observed a load-drop of 56% over 10min for the pure Al beams. We attributed this to grain boundary sliding and the nature of a free-standing thin film, i.e. the absence of a substrate. For the alloyed beams, the load-drop was only 16%. We believed the difference was due to Al3Ti precipitates formed at grain boundaries, which hindered dislocation movements. We used TEM to reveal the microstructural features of the micro beams.
KEYWORDS: Micro Electro Mechanical Systems (MEMS), micromachining, thin film, substrate, piezo-actuator, yield strength, alloy, stress relaxation, grain boundary.
Introduction
In Micro Electro Mechanical Systems (MEMS) devices, many of the active components exist in the form of a free-standing thin film. Such components are constantly in motion under various actuation or stimulation [1]. Understanding the mechanical properties of a free-standing thin film is therefore important for the design of MEMS devices, as well as for predicting their mechanical performances.
The mechanical behavior of a free-standing thin film is expected to be different from that of bulk material or a conventional thin film on substrate. In a free-standing thin film, the grain size is typically very small and the absence of a substrate leads to both its top and bottom surfaces being unconstrained. These microstructural characteristics may result in unique mechanical properties of a free-standing thin film. Although thin films on substrates have been extensively studied [2,3], knowledge on free-standing thin films was not available until the advent of micromachining techniques and is yet to be further explored [4]. The first part of our study examined the mechanical behavior of free-standing Aluminum (Al) micro beams and compared the results with those reported for bulk Al and Al thin films on substrates.
Many MEMS structures employ alloyed, rather than pure, metals for the purpose of improved mechanical or electrical properties as required by certain device requirements. In the second part of our study, we investigated the mechanical properties of Titanium (Ti) alloyed Al thin films. The alloyed free-standing micro beams contain 2% of Ti in atomic percentage. The Al-2%Ti beams were of the same dimension as the pure Al beams and were fabricated in a similar way. By comparing the mechanical test results on the alloyed samples with those of the pure Al samples, we were able to demonstrate the advantages of using Ti as an alloying material for MEMS devices.
The free-standing micro beams (50mm´500mm´2mm) of pure Al and Al-2%Ti were fabricated using micromachining techniques at Stanford Nanofabrication Facility (SNF). For the mechanical tests, we used a dedicated micromechanical testing system. The heart of the system is a custom-designed piezo-actuator-driven test rig with a load resolution of ±0.2mN and a displacement resolution of ±10nm. We carried out microtensile tests as well as stress relaxation tests on both the pure Al and Al-2%Ti samples. Transmission Electron Microscopy (TEM) was used to reveal the microstructural features of the micro beam samples.
Sample Fabrication
The specimens were fabricated using micromachining techniques within a cleanroom facility. For both the pure Al and Al-2%Ti micro beams, the fabrication process made use of two masks. One mask is for the front-side, which defines the beams; the backside mask defines the silicon windows to be etched away to release the beams. There are 34 dice per wafer, with each die (Fig. 1) measuring 8mm´21mm. Each die contains one micro beam (Fig. 2), which measures 50mm wide, 500mm long (gauge section) and 2mm thick. For the ease of die removal upon completion of the process, a groove of about 500mm wide is designed to surround each die. The groove is formed during the silicon backside etch, which leaves the die attached to the remainder of the wafer by only two small silicon support bridges at two corners of the die.
FIG. 1. Die schematic of a micro beam specimen.
FIG. 2. SEM micrograph of a free-standing
Al micro beam specimen (prior to testing).
To fabricate the pure Al micro beams, we used (100)-oriented 4-inch silicon wafers. First, 1mm of silicon nitride was deposited by Low Pressure Chemical Vapor Deposition (LPCVD) on both sides of the wafer. Using Photo Resist (PR) to protect the front-side nitride, the backside nitride was removed by a dry etch; the front-side nitride was used later as an etch-stop layer during the silicon bulk etch to release the micro beams from the backside. Next, 2mm of pure Al was sputter deposited onto the front-side of the wafer, and then patterned with the front-side mask to define the Al micro beams. Afterwards, a thick layer of PR was put down onto the backside of the wafer and patterned with the backside mask. This patterned PR acted as etch mask for the removal of silicon from the backside of the wafer through window regions defined by the backside mask. Silicon was then dry-etched through wafer thickness (about 500mm) until the front-side LPCVD nitride was reached. Lastly, the nitride was removed by dry etch from the backside, and the Al micro beams were released.
For the fabrication of the Al-2%Ti micro beams, the procedure is identical except for the step of metal deposition. Instead of depositing pure Al, a 7-ply 0.5mm-Al/10nm-Ti multi-layer was deposited on the front-side of the wafer followed by subsequent annealing at 550 °C for one hour in a nitrogen atmosphere. The multi-layer was formed by 3 layers of 10nm-Ti sandwiched alternately between 4 layers of 0.5mm-Al, giving a total thickness of about 2mm. This metal layer was then patterned and further processed. TEM studies on the cross-section of the Al/Ti multi-layer (Fig. 3) before and after the annealing indicate that Ti had reacted with Al to form Al3Ti precipitates along the Al grain boundaries during the heat treatment.
FIG. 3. TEM cross-sectional view of (a) alternate layers
of Ti and Al in the as-deposited Al/Ti multi-layer, (b) Al3Ti
precipitates formed at Al grain boundaries after annealing.
(arrows are pointed to the locations of Al3Ti precipitates)
Experimental Setup and Procedure
Micromechanical Testing System
The custom-designed micromechanical testing system (Fig. 4) comprises a piezoelectric actuator with position sensor, a load cell with temperature sensor, a height-adjustable alignment stage, and sample grips that are attached to the piezo-actuator and load cell. The control electronics include a waveform generator, an amplifier, a closed-loop piezo controller, a signal conditioner and an A/D board located in a control PC. Data acquisition is achieved by LabVIEW software. An optical microscope is mounted directly overlooking the top of the sample for the purpose of height alignment as well as test monitoring [5]. During testing, the system is kept in a thermally insulating box to eliminate possible effects that could be caused by temperature fluctuations, such as load cell drift or change in thermal expansion coefficients of the components of the system, which could lead to inaccurate measurement of stress or strain.
FIG. 4. Top view of the main components of the
custom-designed micromechanical testing system.
The load cell has a maximum measurement range of 44.5N and a load resolution of ±0.2mN. The piezo-actuator has a maximum measurement range of 60mm and a displacement resolution of ±10nm. The load to stress conversion is given by
(1)
where is the cross-sectional area of the beam. The displacement to strain conversion is given by
(2)
While we are only interested in the displacement of the gauge section of the specimen, other parts of the system along the stress direction also contribute to the total measured displacement. Strain calibration is done by taking into account the stiffness of the load cell, root section of the specimen, and other compliant parts of the system in addition to the stiffness of the gauge section of the specimen [5].
Experimental Procedure
Prior to testing, the electronics are energized and left to thermally equilibrate. The load cell is pre-calibrated [5]. The sample stage is adjusted under the optical microscope to eliminate the height difference between the two grips.
The test die is removed from the wafer by gently breaking the two silicon support bridges attached to the die corners, and is then placed on the grips with the beam side facing downward. It is then clamped in place by four screws, two on each side, with care taken to ensure even clamping throughout the process (by monitoring the load response on the LabVIEW screen). Upon securing the die onto the grips, the silicon supports on the two long sides of the die are cut with a hand-held rotary diamond saw. The Al micro beam is now free-standing between the two grips, where one end is attached to the load cell and the other to the piezo. Finally, the system is enclosed in the thermally insulating box and allowed to reach thermal equilibrium before actual testing (The temperature is maintained ±0.5 °C).
Tensile tests are performed by allowing the piezo control voltage to increase (i.e. to elongate the beam) monotonically, while the corresponding load (or stress) response of the beam and hence the stress vs. strain relation is recorded by LabVIEW. For a stress relaxation test, the beam is quickly loaded to a certain strain first and then held at this constant strain while the resulting load (or stress) change is recorded.
Results and Discussion
Tensile Test
From the stress vs. strain curve (Fig. 5) of the pure Al free-standing micro beams, we find Young’s modulus from the initial slope of the elastic region. This slope is found to be about 60GPa, close to the value of 62GPa for bulk Al [6]. The yield strength at 0.2% strain for the pure Al free-standing micro beams is about 120MPa. It is much higher than that of bulk Al, which is about 10-20MPa [6]; yet somewhat lower than that of a conventional Al thin film on substrate, which has been reported to range from 200MPa to 400MPa depending on its thickness [7].
FIG. 5. Stress vs. Strain curves of pure Al and Al-2%Ti micro beams.
The significantly higher yield strength of our Al free-standing thin film as compared to bulk Al agrees with classical theory. It is known that the tensile strength of a material is proportional to the inverse square root of its grain size [8]. While bulk materials commonly have grains in the millimeter range, thin films typically have grains in the range of microns or tens of microns, roughly equal to their film thickness. Our pure Al free-standing thin films have grains of about 2mm (Fig. 6). The much smaller grain size results in much larger yield strength.
FIG. 6. TEM bright filed image of a grain in a pure Al
micro beam, showing dislocations after 0.4% strain.
Conventional thin films on substrates typically have higher yield strengths than their bulk counterparts not only because of their small grain sizes, but also the presence of substrates. It has been shown that the stress required to move dislocations in a thin film adhered to substrate is greater than that for bulk material, due to the fact that misfit dislocations are pinned at the film-substrate interface, which make the dislocation motion difficult [9]. The reason for the slightly lower yield strength of our Al free-standing thin film as compared to a conventional Al thin film on substrate, is, we believe, due simply to the absence of a substrate. For the case of a free-standing thin film, since both of its surfaces are free of constraint (because there is no film-substrate interface), dislocations can glide under lower applied loads than those in a conventional thin film adhered to substrate. Consequently, this will result in a lower yield strength for a free-standing thin film. This has been proved by our tensile test result.
From the stress vs. strain curve (Fig. 5) of the Al-2%Ti micro beams, we find a Young’s modulus of about 60GPa, similar to that of the pure Al micro beams. This is expected as the Ti concentration is reasonably low in these samples, and thus the modulus of the alloy is dominated by the modulus of the pure Al. The yield strength of the Al-2%Ti micro beams is found to be 210MPa, about 75% higher than that of pure Al. This significant yield strength increase is due to Al3Ti precipitates that have formed at grain boundaries. Al3Ti has a much larger modulus than that of Al [10]. It is known that hard precipitates (in this case, Al3Ti) can block the motion of dislocations in a soft matrix (in this case, Al), and hence increase the amount of stress required to move these dislocations [11]. In effect, this will result in a higher yield strength in the Al-2%Ti micro beams than of those made from pure Al.
Our TEM observation also showed a consistent result. The TEM samples were prepared from pure Al and Al-2%Ti beams after 0.4% strain of deformation. The bright field image of the pure Al sample (Fig. 6) under a two-beam condition () reveals many dislocations in the grain that are running rather straight. In the alloyed sample (Fig. 7), however, we see a different dislocation pattern, that is, dislocations in the alloyed beams appear to zigzag or swirl around Al3Ti precipitates, indicating that their motion was indeed blocked by the presence of the precipitates.