W1E.002
HIGH ASPECT RATIO MICROELECTROMECHANICAL SYSTEMS: A VERSATILE APPROACH USING CARBON NANOTUBES AS A FRAMEWORK D.N. Hutchison, Q. Aten, B. Turner, N. Morrill, L.L. Howell, B.D. Jensen, R.C. Davis†, and R.R. Vanfleet* Brigham Young University, Provo, Utah, USA ABSTRACT We recently developed a fabrication process for carbon nanotube templated MEMS. The fabrication process involves growing a three dimensional pattern from carbon nanotube forests and filling that forest by chemical vapor infiltration to make a solid structure. This templating process allows us to fabricate extremely high aspect ratio microscale structures from a wide variety of materials. The nanotube structures can be hundreds of microns tall with lateral pattern dimensions down to a few microns. The chemical vapor infiltration has been shown with silicon and silicon nitride but could be extended to many other materials. In this paper, we investigate the microstructure of the filling material and extend the process to the fabrication of comb actuators.
KEYWORDS Vertically aligned carbon nanotubes, chemical vapor infiltration, high aspect ratio, comb actuators.
INTRODUCTION Silicon based microelectromechanical systems (MEMS) have a long and well established history. Silicon processing dominates the literature and applications for MEMS. The use of silicon imposes some constraints on the aspect ratio of silicon-based MEMS due to the cost and quality of deep reactive ion etching (DRIE). Additionally, materials properties are limited to those of silicon. Alternative
materials including metals or ceramics could enable applications of MEMS in chemically reactive or high temperature environments or provide enhanced electrical, mechanical, or optical functionality. Successful MEMS have been limited with many of these materials due the lack of effective processes for deep etching. In particular the difficulty in deep etching of metals has lead to alternate fabrication process like LIGA [1]. LIGA in its original form, however, requires the use of a synchrotron radiation source. To be cost effective, a LIGA mold needs to be formed and used many times making development and prototyping very challenging. LIGA-like processes using SU-8 have been demonstrated using more conventional lithography. These LIGA-like processes achieve smaller aspect ratios and shorter structures than true LIGA. We have developed a novel method to make high aspect ratio microstructures out of a wide variety of materials by growing forests of patterned verticallyaligned carbon nanotubes (VACNTs) as a framework and then filling in the nanotube “forests” with various materials by chemical vapor infiltration [2]. The process for fabricating carbon nanotube framework MEMS (CNTMEMS) is outlined in items a-g of figure 1 and examples are shown in figures 2 and 4. The VACNT structures are grown from a Fe catalyst on a solid surface. During the VACNT growth, multi-walled carbon nanotubes grow vertically upward from the solid surface at controllable rates over 100µm/min. The high aspect ratio three
Figure 1 CNT framework MEMS (a-g) Process diagram. (a-d) a 2 nm Fe catalyst film is patterned by lift-off on a silicon wafer coated with 30 nm of alumina. (e) A forest of VACNT’s is grown by atmospheric CVD from the patterned iron. (f) Chemical vapor infiltration is used to fill in the carbon nanotube framework with silicon or some other material. (g) The underlying release layer is etched to free moving parts.
978-1-4244-4193-8/09/$25.00 ©2009 IEEE
1604
Transducers 2009, Denver, CO, USA, June 21-25, 2009
the shape of the CNT framework but that consists mostly of the filler material. We have filled in the framework with polysilicon and silicon nitride using low pressure chemical vapor deposition (LPCVD); as shown in figures 2(c) and 2(b) respectively. However many other CVDdepositable materials could be used in the infiltration process and are currently being explored.
MICROSTRUCTURE Figure 3 shows SEM and TEM images of the interior regions of CNT-MEMS filled by a 300 nm thick LPCVD process step. For complete filling the vapor infiltration process would ideally be limited by surface reaction
(a)
kinetics rather than mass transport of the gases to and from the reaction surfaces. This allows process gasses to penetrate deeply into the nanotube forest before reacting. The silicon LPCVD process is good at deep filling, however, the quality of deep filling still depends some upon proximity to the edges of the structure for distances over ~ 10 microns. In 3(a), vapor access holes were made in large features to facilitate filling of tall and wide structures. These vapor access holes are visible in crosssection as the vertical features to either side of the center of the image. Notice how these small features, which are about 2-3 microns in diameter, remain vertical and unfilled under standard silicon LDCVD conditions even in the presence of vapor access holes in a tall structure. It should be noted that silicon deposition conditions were not optimized for the infiltration process so more complete filling of these geometries may be possible. In 3(b), an isolated line is shown in cross section with greater than 90% solid filling [2]. Some unfilled voids within the filled structure can be seen in both 3(a) and 3(b). Figure 3(c) shows a TEM dark field image with some of the crystalline grains showing up as bright. The size of these grains is about 175 nm. These are typical of the interior of the
(b)
(c) Figure 2 (a) CNT structure corresponding to fabrication step (1e). (b) Si3N4 filled CNT structure corresponding to fabrication step (1f). (c) Silicon filled and released composite MEMS corresponding to fabrication step (1g).
dimensional structures that result come from the extremely vertical nature of VACNT growth combined with the two dimensional patterning of the Fe catalyst film. We have grown patterned VACNT structures over 1 mm tall with vertical sidewalls and lateral feature sizes down to a few microns. The nanotube forest is a very low density (0.009 g/cc) material, consisting mostly of air but with the ability to form a 3-dimensional structure with high fidelity as shown in figure 2(a). The “as grown” CNT structures are not useful as mechanical materials because they are extremely fragile. However, when the spaces between tubes in the forest regions are replaced with a "filler material" using chemical vapor infiltration, a solid structure results with
1605
(a)
(b)
(c)
(d)
200 nm
200 nm
Figure 3. (a) and (b) are SEM images of the interior of filled CNT-MEMS structures. (c) and (d) are TEM images of the interior filling silicon. (c) is a Dark field image showing the polycrystalline nature of the material and (d) shows the CNTs as lighter thread like patterns.
(a)
(b)
100 μm
100 μm
Figure 4 (a) high aspect ratio silicon coated lines exhibiting undesired bending. (b) examples of stabilizingribs which allow the growth of higher aspect ratio features
structure with similar length and width of grains. On nearby planar surfaces the grains are columnar. The interior deposition on nanotubes perturbs the nucleation and growth of grains in these regions. Figure 3(d) shows a bright field TEM image with two features of interest. The thread like patterns are the nanotubes incorporated into the silicon solid. The nanotubes remain distinct and no reaction or mixing between the carbon and silicon is seen. On the left side of 3(d) are two lighter areas that seem packed with particles. These are the unfilled (by CVD) voids that have been filled with polishing residue left from the TEM sample preparation. The nearly complete dense filling with silicon may result in a composite with mechanical properties close to silicon. Carbon nanotubes do have have a very high strength to weight ratio. A nanotube composite might therefore be expected to have significant strength enhancement. This potential strength enhancement would be best along the length of the nanotubes. Although the nanotubes in these structures are partially aligned vertically, there is lateral intertwining of the tubes that may increase strength in that direction. Previous bending tests did show Young’s modulus and strength values close to that of polysilicon, as might be expected with good filling of the nanotube template.[2] . Strength measurements may show a different story as the intimate connection to the
Figure 5 Force v.s. displacement for fixed-free (cantilever) and fixed-guided beams both with and without stabilizing ribs.
nanotube fibers to the silicon may impede crack propagation The nanotubes are intact after the silicon deposition as seen in figure 3(d). In addition to the structural template, the nanotubes provide a conducting path through the filler material, providing conductivity when the filler materials are insulators or low conductivity materials. This is consistent with resistivity measurements on the Si infiltrated nanotubes that showed much lower resistivity (4 ohm-cm) than that of the deposited silicon that was deposited[2].
ASPECT RATIO Under the optimal growth conditions the CNT structures have very straight and smooth side walls (< 1 μm roughness). The filling process then retains the vertical smooth structure. Figures 2, 4, and 6 show structures with very high aspect ratios. We regularly produce features of a few microns width that are hundreds of microns tall. Fine isolated features tend to bend during the CNT growth process as shown in figure 4(a), but patterns with interlocking features maintain excellent verticality. This bending is detrimental to MEMS mechanisms using long, slender beams. The growth of a high aspect ratio VACNT forests can be stabilized by the addition of “ribs” along the beam’s length as seen in figure 4(b). Figure 5 shows calculations using a finite element model to compare straight and rib-stabilized beams. These results indicate that during MEMS device design, ribstabilized beams can be used in place of straight beams without significantly increasing stiffness. Additionally, the stiffness of rib-stabilized beams can be closely approximated by the well established equations for straight beams.
COMB ACTUATOR Figure 6 shows SEM images of a silicon-filled CNTMEMS comb actuator. The sample is tilted in figure 6(a) so both the overall comb structure can be seen as well as the 106 μm height. Figure 6(b) is a top down view. Figure 7 shows optical images of the off (a) and on (b) states of the actuator, with the on-state corresponding to ~100 V. The high aspect ratios possible with the process are helpful in increasing the force output of the comb actuator (or the sensitivity of a comb variable capacitor) by increasing the thickness of the comb array while allowing small lateral features for high overall capacitance. In addition to the comb actuator shown here, we have also successfully tested thermomechanical actuators [3] and bistable mechanisms [4] fabricated from both silicon-filled and silicon nitride-filled forests.
1606
(a)
(b) Figure 6 (a) an SEM image of a 106 µm tall silicon filled CNT comb actuator shown at a 50˚ tilt. (b) the comb actuator shown in a top down view.
CONCLUSION We are developing CNT-MEMS fabrication processes and related process design constraints for several materials systems including silicon and silicon nitride. To demonstrate the power and flexibility of this approach we have fabricated and tested a variety of CNT-MEMS structures and devices including TEM grids, cantilevers, thermomechanical in-plane microactuators (TIMs), bistable mechanisms and comb drives. This is a potential materials breakthrough for MEMS. Whereas most bulk micro-machining has been limited to silicon, the CNT framework process opens up the possibility of fabricating MEMS with very tall features from a wide range of materials based on the desired (a)
(b)
Figure 7. Comb fingers in the off (a) and on (b) configurations. The triangle structures at the bottom are the moveable side of the comb drive.
materials properties, which could include mechanical, electrical, electrochemical, or optical properties. CNTMEMS offers several potential benefits including: (1) high aspect ratio MEMS structures made in a wide variety of materials, (2) allowing conductive MEMS made from insulating materials like metals and functional oxides, and (3) VACNT growth is much faster than deep reactive ion etching, the current leading technology for creating high aspect ratio structures.
CONTACTS †R.C.D.tel: +1-801-422-3238
[email protected] * R.R.V., tel: +1-801-422-1702;
[email protected]
REFERENCES [1] Malek CK, Saile V, “Applications of LIGA technology to precision manufacturing of high-aspect-ratio microcomponents and -systems: a review”, Microelectronics Journal 35 (2), 131-143 (2004). [2] D. Hutchison, N. Morrill, Q. Aten, B. Turner, B. Jensen, L. Howell, R. Vanfleet, R. Davis, “High Aspect Ratio Patterned Nanocomposites Fabricated by Carbon Nanotube Templating”, In Review [3] C. D. Lott, T. W. McLain, J. Harb, and L. L. Howell, “Modeling the thermal behavior of a surfacemicromachined linear-displacement thermomechanical microactuator,” Sensors and Actuators, A: Physical, vol. 101, pp. 239–250, 2002. [4] B. D. Jensen, M. B. Parkinson, K. Kurabayashi, L. L. Howell, and M. S. Baker, “Design optimization of a fully-compliant bistable micro-mechanism,” in Microelectromechanical Systems (MEMS), 2001 ASME Int. Mechanical Engineering Congress and Exposition, 2001.
1607
