Fabrication of Single-Walled Carbon-Nanotube-Based Pressure

NANO LETTERS

Fabrication of Single-Walled Carbon-Nanotube-Based Pressure Sensors

2006 Vol. 6, No. 2 233-237

C. Stampfer,*,† T. Helbling,† D. Obergfell,‡ B. Scho1 berle,† M. K. Tripp,†,§ A. Jungen,† S. Roth,‡ V. M. Bright,§ and C. Hierold† Micro and Nanosystems, ETH Zurich, 8092 Zurich, Switzerland, Max Planck Institute for Solid State Research, 70569 Stuttgart, Germany, and Department of Mechanical Engineering, UniVersity of Colorado, Boulder, Colorado 80309 Received November 3, 2005

ABSTRACT We report on the fabrication and characterization of bulk micromachined pressure sensors based on individual single-walled carbon nanotubes (SWNTs) as the active electromechanical transducer elements. The electromechanical sensor device consists of an individual electrically connected SWNT adsorbed on top of a 100-nm-thick atomic layer deposited (ALD) circular alumina (Al2O3) membrane with a radius in the range of 50−100 µm. A white light interferometer (WLI) was used to measure the deflection of the membrane due to differential pressure, and the mechanical properties of the device were characterized by bulge testing. Finally, we performed the first electromechanical measurements on strained metallic SWNTs adhering to a membrane and found a piezoresistive gauge factor of approximately 210 for metallic SWNTs.

Research in the field of nanomechanics1 and nanoelectromechanical systems2,3 is growing rapidly with considerable potential for future ultrafast high-sensitivity low-power devices.4 At the nanoscale level, a number of new applications may become possible including resonating charge shuttles,5 single-electron spin detection,6 ultrasensitive mass7 and force8 sensing, and carbon nanorelays.9 However, the inevitable development of microelectromechanical systems (MEMS) down to the nanoscale (NEMS) level suffers from a discontinuity in process technologies and a nonoptimal transfer of transducer concepts from microscale to nanoscale systems.10 Among the investigations of alternative transducer concepts,7-9 new nanomaterials for sensing at the nanoscale level are being considered to overcome these so-called scaling problems. Carbon nanotubes, in particular single-walled carbon nanotubes (SWNTs), have outstanding potential as a new material for novel NEMS devices. Carbon nanotubes have been investigated intensively during the past decade11-14 and have attracted great interest because of their exceptional electronic, mechanical, and electromechanical properties. Electronically, SWNTs can behave as either metallic or semiconducting, depending on the chirality of their atomic * Corresponding author. E-mail: [email protected]. † ETH Zurich. ‡ Max Planck Institute for Solid State Research. § University of Colorado. 10.1021/nl052171d CCC: $33.50 Published on Web 01/21/2006

© 2006 American Chemical Society

arrangement. They are highly elastic with a Young’s modulus in the range of 1 TPa.14 For sensor applications, the electromechanical properties of SWNTs are most interesting. Recent experiments15-19 have proven SWNTs potential for use as piezoresistors in a variety of applications. Most recently, Grow et al. studied the electromechanical response of semiconducting and small-gap semiconducting (SGS) SWNTs adhering to a silicon nitride surface. They found gauge factors of down to -400 and up to 850 for semiconducting and SGS SWNTs, respectively. According to Yang et al.20 the sign of the gauge factor depends directly on the wrapping indices (n, m). More precisely, it depends on the value of (n - m) modulus 3. Thus, half of all of the semiconducting SWNTs will have a positive gauge factor, while the other half show a negative gauge factor. However, metallic SWNTs show positive gauge factors. The magnitudes of the reported gauge factors of SWNTs exceed significantly the gauge factor (200) of state-of-the-art doped-silicon strain gauges. In an earlier experiment, Cao et al. stretched metallic, SGS, and semiconducting SWNTs suspended between a surface micromachined cantilever and a platform. Cao et al.16 observed, for all three types, an increase in resistance under mechanical load (on the cantilever). Effective piezoresistive gauge factors between 600 and 1000 are reported. Here we present the fabrication and operation of a new pressure sensor based on carbon nanotubes as piezoresistive strain gauges. These devices show, in contrast to state-of-

Figure 1. (a) Schematic of the carbon-nanotube-based pressure sensor consisting of an ultrathin alumina membrane of radius r0 ) d/2. The single-walled carbon nanotube adhering to the membrane is additionally clamped by two metal electrodes (31 nm Ti/Au). A side gate is patterned to control the electrical response of the different nanotube types. (b and c) SEM images of a final device consisting of an alumina membrane with electrically contacted single-walled carbon nanotubes. The nanotubes are placed in the center of the membrane in order to gain an isotropic strain distribution (r < 0.2r0).

the-art piezoresistive pressure sensors, a potentially higher sensitivity in terms of a larger gauge factor and a higher scalability due to the nanoscaled (SWNT based) strain gauges. With the presented device we report for the first time electromechanical measurements of metallic SWNTs adhering to alumina membranes. Each pressure sensor consists of an ultrathin atomic layer deposited (ALD) alumina (Al2O3) membrane with an integrated electrically contacted SWNT as the strain gauge. The SWNTs adhere to the alumina surface by van der Waals forces and are also clamped in place by two metal electrodes. It is assumed that the SWNTs experience the same stretching as the membrane, similar to the doped-silicon strain gauges used in state-of-the-art MEMS pressure sensors. A schematic of the SWNT-based pressure sensor is shown in Figure 1a. The process flow used for the fabrication of SWNT-based pressure sensors is shown schematically in Figure 2. The 300-µm-thick silicon (Si) samples are cleaned in a 5:1 piranha solution for 15 min to remove all organics, and the native oxide is stripped using HF just before the growth of the Al2O3. The Al2O3 layer is grown on the silicon substrate using the atomic layer deposition (ALD)21 technique. Standard photolithography and lift-off are subsequently used to pattern 32-nm-thick chromium/gold (Cr/Au) alignment markers on the frontside of the samples (Figure 2a). These large markers (100 µm) are used to align the following frontside and backside processing. The membrane openings in the Al2O3 film on the backside are patterned using standard deep ultraviolet (DUV) photolithography and are aligned to the frontside markers by infrared backside alignment. Then the Al2O3 layer is patterned using inductively coupled plasma 234

Figure 2. Process flow to fabricate single-walled-carbon-nanotubebased pressure sensors: (a) 100 nm of Al2O3 is deposited by atomic layer deposition (ALD) on a 300-µm-thick Si sample. Photolithography and lift-off are used to pattern markers and electrodes. (b) The membrane openings are patterned using infrared backside alignment and anisotropically dry etched from the backside. (c) SWNTs are dispersed from a SDS solution onto the Al2O3. (d) AFM image recording followed by PMMA spin coating and e-beam exposure. (e) Metalization and lift-off to electrically connect the SWNTs, and the final dry etch membrane release (f).

reactive ion etching (ICP). The Al2O3 is further used as a hard mask for the subsequent anisotropic etch step to form the membrane cavities. A deep reactive ion etching ICP (DRIE-ICP) bulk micromachining BOSCH22 process is used to form the cavities by anisotropic etching (Figure 2b) until only a few micrometers (∼10 µm) of Si remain. This remaining Si is crucial to mechanically support the Al2O3 thin film during the next processing steps. The following adsorption and contacting of individual SWNTs is described in more detail in ref 8. Small reference alignment markers (2 µm) are patterned by electron-beam (e-beam) lithography and formed by metal (2 nm Cr and 30 nm Au) evaporation and lift-off. These markers are later used to precisely align the metal electrodes to the SWNTs. Next, the alumina surface is functionalized with a DAS, N-[3-(trimethoxysilyl)propyl]ethylene-diamine (97%), self-assembled monolayer to enhance the adsorption of nanotubes. The SWNTs (from an arc-discharge process) have been dispersed in a sodium dodecyl sulfate (SDS) solution and are randomly adsorbed on the Al2O3 thin film (Figure 2c). The reference alignment markers serve to identify the location and orientation of individual SWNTs, which are recorded by atomic force microscope (AFM) imaging in tapping mode. These AFM images are further used to design the e-beam illumination pattern (mask) for the electrode structures that are used to electrically connect the nanotubes. The following e-beam lithography step is used to structure the individual electrode patterns (Figure 2d). After development, a thin Ti adhesion layer (1 nm) and 30 nm Au are evaporated and lifted-off resulting in electrically contacted individual SWNTs (Figure 2e). Here, Ti is used instead of Cr in order to decrease the SWNT-metal contact resistance. The final process step is the release of the ALD alumina membrane from the remaining few micrometers of bulk Si. This process step is very crucial and is optimized to prevent damage to the ultrathin membrane. Thus, an isotropic release by reactive ion etching (RIE) is performed (Figure 2f). Before the final Nano Lett., Vol. 6, No. 2, 2006

(w0) of two membranes with different radii, r0 ) 54 µm (triangles) and r0 ) 101 µm (circles) are plotted against the differential pressure (∆p). The following analytical model for large membrane deflection24 has been used to fit the measured data ∆ p(w0) )

Figure 3. Bulge test of two alumina membranes (with different radii) to extract Young’s modulus and initial in-plane stress. (a) Deflection (w0) vs differential pressure (p) for 100-nm-thick alumina membranes with r0 ) 54 µm and r0 ) 101 µm. (b) Strain vs differential pressure relation extracted from the measurement shown in Figure 4a. The inset shows a typical white light interferometer image of the deflected membrane under load.

release takes place, the frontside is covered with photoresist to prevent the SWNT from incurring damage because of the plasma. Sulfur hexafluoride (SF6) and low power are used to achieve an isotropic release profile. A final cleaning step removes the protective photoresist layer. In Figure 1b and c we show SEM images of a final device. The released Al2O3 circular membrane with the Au electrodes is depicted in Figure 1b. The large gold electrodes are heading to the center area of the membrane (black disk), where the reference alignment markers have been patterned and the individual electrodes are placed to contact the SWNT. An electrically contacted SWNT (white arrows) can be seen clearly in Figure 1c (a close up of Figure 1b). The load (differential pressure) versus deflection relations of t ) 100nm-thick alumina membranes with connected SWNTs are measured by the bulge testing method,23 where a white light interferometer (ZYGO New View 5020) is used to measure the maximum deflection, w0, of the membranes. The inset in Figure 3b shows a typical white light interferometer image. The maximum deflection is in the center of the membrane (point B), and w0 is defined as the height difference between points A, A′, and B. In Figure 3a, the measured deflections Nano Lett., Vol. 6, No. 2, 2006

(7 - ν)Et 3(1 -

ν)r40

w30 +

4tσ0 r20

w0

(1)

where E is the Young’s modulus, ν is Poissons ratio, and σ0 is the initial stress. This model fits to the measured data very well and confirms the deflection behavior of circular membranes (see also the inset in Figure 3b). Moreover, we extracted the Young’s modulus, E, and the initial stress, σ0, of the alumina thin film from the curve fitting. This yields a Young’s modulus E ) 189 ( 25 GPa (E ) 181 ( 20 GPa) and an intrinsic in-plane stress σ0 ) 332 ( 27 MPa (σ0 ) 383 ( 26 MPa) for the membrane with radius of r0 ) 54 µm (r0 ) 101 µm). These results are derived assuming a Poissons ratio ν ) 0.24.26,27 The two different membranes are from two different processing runs, which may explain the discrepancy of the intrinsic in-plane stress. However, the results agree well with the results obtained previously from bulge measurements performed on pure alumina membranes (without electrode structures).25 Thus, we conclude that the 30-nm-thick Ti/Au electrodes (see Figure 1b and the inset of Figure 3b), which are placed on top of the membrane, do not alter the mechanical properties of the Al2O3 membrane significantly. The pressure sensor design based on circular membranes (in contrast to rectangular membranes) has been chosen mainly because of the isotropy of the strain, , in the center area (r/r0 < 0.2) of the membrane. The strain in this area is approximated by  ) 1/2[r0/w0 + w0/r0](θ - sin θ), where θ ) arccos[(r20 - 2w20)/(r20 + 2w20)].28 Because the strain is now independent of the orientation, all randomly adsorbed SWNTs in the inner area of the membrane experience the same strain, . The relation between applied differential pressure and extracted strain for both membranes is shown in Figure 3b. Electromechanical measurements of a metallic nanotube adhering to the alumina membrane are performed at room temperature (T ) 300 K). The results are plotted in Figure 4. I-V measurements for various applied differential pressures (∆p) are shown in Figure 4a. The conductance clearly decreases with increasing pressure. The independent nature of the metallic SWNT (M-SWNT) to the gate voltage, Vg (at Vds ) 20 mV), is shown in the inset of Figure 4a. The side-gate electrodes, which are approximately 1 µm away from the nanotube, exhibited reasonable functionality, as demonstrated on semiconducting nanotubes (see S-SWNT no. 1 measurement in the inset). In Figure 4b the transducer property of the proposed sensor is plotted and shows the resistance as a function of the applied differential pressure (∆p). The plot shows the measured average resistance over the Vds interval from -50 mV to 0 mV. The error bars correspond to the standard deviation of the averaging. For small pressure (up to ∆p ≈ 70 kPa), we find a monotonic 235

not yet explained. However, the clear trend and the significant change in resistance makes the proposed sensing scheme favorable. The piezoresistive gauge factor of the presented metallic SWNT (extracted from Figure 4c) defined as βGF ) (∆R/R0)-1 is 210 ( 8. This value slightly exceeds the value of state-of-the-art doped-silicon strain gauges (βGF ) 200).29 To discuss our results, we compare them to theoretical investigations reported previously.20 The resistance change is plotted as a function of strain (Figure 4c), where we make use of the pressure-strain relation plotted in Figure 3b. Furthermore, we assume that the resistance change due to small strains (up to 0.04%) can be explained in terms of a strain-dependent band-gap opening.15,18,20 Therefore, we linearly fit (dashed line in Figure 4c) the resulting resistance versus strain measurements (we restrict ourselves to the small strain regime for fitting) by R() ) R0 + R ˜ 1. This expression can be physically interpreted by comparing it with the strain dependent resistance of contacted SWNTs in the small strain regime18 (terms of the order 2 and higher are neglected) R() ) Rs +

[ ( )] ( ) (

E ˜ g 1 h 1 + exp 2 2 kBT |t| 8e Rs +

Figure 4. Electromechanical measurements on a metallic singlewalled carbon nanotube (M-SWNT) adhering to the alumina membrane. (a) I-V characteristics of the metallic nanotube under different applied pressures. The inset shows a logarithmic plot of the resistance of semiconducting (S-SWNT) and the investigated M-SWNT as a function of the gate voltage. This plot proves the basic functionality of the implemented side gates. (b) The average resistance of the M-SWNT as a function of the differential pressure, p. (c) The resistance change as a function of axial strain, where a linear curve is fitted (dashed line) to the measured data. The inset shows the drain current, Ids, as a function of the gate voltage, Vg, for different strains. The black arrow highlights the gate dependence of Ids at higher strain.

increase in resistance with increasing pressure, which allows further physical interpretation (see below). Alternatively, for larger pressures we see a nonmonotonic anomaly, which is 236

≈

)

E ˜g h 1 h + 2 2  (2) 2 2 |t| 4e |t| 8e kBT

where Rs is the resistance in series with the junction due to metal-SWNT contacts, |t|2 is the transmission through the ˜ g is the strain-dependent band gap nanotube, and Eg ) E for metallic nanotubes, neglecting the torsion contributions. From the parameter fitting (see the dashed line in Figure 4c), we obtain for the zero strain resistance R0 ) 79.5 ( 0.1 kΩ and for E ˜ g/|t|2 ) 1.23 ( 0.05 eV/%. To estimate the range of band-gap opening per percent strain, we limit the range of the transmission, |t|2. For the lower boundary, we assume completely transparent metal-nanotube contacts (hence, Rs ) 0) leading to a transmission of |t|2 ) 0.081 (thus, E ˜ gmin ) 100 meV/%). This estimation might be correct for Pd contacts;15 however, for Ti/Au contacts we expect a nonzero series resistance, Rs, because of a mismatch in the work function of the metal and the SWNT. Moreover, we observed at a strain of  ) 0.072% a clear gate dependence in the current, Ids (see the black arrow in the inset of Figure 4c). We conclude that at this strain the band gap is exceeding the thermal excitation energy (kBT ≈ 25meV) of the electrons at room temperature. This leads to a lower limit of E ˜ gmin ) 340 meV/% (|t|2 ) 0.28) and a series resistance of Rs ) 55 kΩ. An upper limit for the transmission can be estimated by examining the gate dependence of Ids at lower strains (e.g.,  ) 0.058%, see the white triangles in inset of Figure 4c) where no deviation from the metallic behavior is found. Because up to a strain of 0.058% no gate dependence has been observed, we conclude that the transmission is smaller than E ˜ gmax ) 430 meV/% and |t|2 ) 0.35. In summary, we find that E ˜ g ) dEg/d lies in the range of 340 to 430 meV/%. It is noteworthy that according to Yang et al.20 the maximum band-gap opening per percent strain of a metallic Nano Lett., Vol. 6, No. 2, 2006

SWNT under pure axial strain (no torsion and distortion) is limited by E ˜ gmax ≈ 100 meV/%. This is significantly lower than the range for E ˜ g we found. However, this discrepancy cannot be due to pressure and strain uncertainties in the performed measurements. It seems to be more likely that the nanotube surface interactions lead to an increasing strain, torsion, and possibly collapse of SWNTs, which may effect the band gap and transmission amplitude significantly. This may also explain the anomalies at higher strains and the nonmonotonic behavior, which is shown clearly in Figure 4b and c. Hence, further investigations on metallic and semiconducting SWNTs are needed to study these effects. Variable temperature studies are needed to separate the different contributions for the strain-dependent resistance, R(). In summary, we have designed, fabricated, and tested a pressure sensor device based on SWNTs as active elements for a pressure range of 0-130 kPa. We developed a new fabrication process for making ultrathin alumina membranes supporting electrically connected SWNTs. Through mechanical bulge test measurements, we verified that alumina shows excellent properties for use as an ultrathin pressure membrane. I-V measurements on metallic and semiconducting SWNTs have been carried out and prove the feasibility of the presented device. Electromechanical measurements on metallic SWNTs adhering to the alumina membrane have been performed, and a piezoresistive gauge factor of 210 has been found, which is slightly exceeding state-of-the-art Si-based strain gauges. Acknowledgment. We thank Cari Hermann and Steven George for the atomic layer deposition of Al2O3, and we thank David Juncker, Jochen Ho¨tzel, Otte Homan, Cari Hermann, Steven George, Patric Strasser, and Udo Lang for helpful discussions. Support by the ETH FIRST Lab and financial support by the TH-18/03-1 grant and Swiss National Science Foundation (20021-108059/1) are gratefully acknowledged. M.T. and V.M.B. gratefully acknowledge the CAMPmode Industrial Advisory Board and NSF IGERT: 1530112, DARPA/MTO Grant No.: NBCH1040003, DGE9870665, and the Air Force Office of Scientific Research for their financial support. References (1) Cleland, A. N. Foundations of Nanomechanics: From Solid-State Theory to DeVice Applications; Springer, New York, 2003.

Nano Lett., Vol. 6, No. 2, 2006

(2) Craighead, H. G. Nanoelectromechanical Systems; Science 2000, 290, 1532-1535. (3) Roukes, M. L. Nanoelectromechanical Systems, Technical Digest of the 2000 Solid-State Sensor and Actuator Workshop, Hilton Head Isl., SC, June 4-8, 2000. (4) Schwab, K. C.; Roukes, M. L. Phys. Today 2005, July, 36. (5) Erbe, A.; Weiss, C.; Zwerger, W.; Blick, R. H. Phys. ReV. Lett. 2001, 87, 096106. (6) Rugar, D.; Budakian, R.; Mamin, H. J.; Chui, B. W. Nature 2004, 430, 329. (7) Nishio, M.; Sawaya, S.; Akita, S.; Nakayama, Y. Appl. Phys. Lett. 2005, 86, 133111. (8) Stampfer, C.; Jungen, A.; Hierold, C. IEEE Sen. J., accepted for publication, 2005. (9) Lee, S. W.; Lee, D. S.; Morjan, R. E.; Jhang, S. H.; Sveningsson, M.; Nerushev, O. A.; Park, Y. W.; B Campbell, E. E. Nano Lett. 2004, 4, 2027-2030. (10) Hierold, C. J. Micromech. Microeng. 2004, 14, 1-11. (11) Iijima, S. Nature 1991, 354, 56-58. (12) Carbon Nanotubes: Synthesis, Structure, Properties and Applications; Dresselhaus, M. S., Dresselhaus, G., Avouris, P., Eds.; Springer: Berlin, 2001; Vol. 80. (13) Saito, R.; Dresselhaus, G.; Dresselhaus, M. S. Physical Properties of Carbon Nanotubes; Imperial College Press: London, 2001. (14) Reich, S.; Thomsen, C.; Maultzsche, J. Carbon Nanotubes; WileyVCH: Hoboken, NJ, 2003. (15) Grow, R. J.; Wang, Q.; Cao, J.; Wang, D.; Dai, H. Appl. Phys. Lett. 2005, 86, 093104. (16) Cao, J.; Wang, Q.; Dai, H. Phys. ReV. Lett. 2003, 90, 157601157604. (17) Tombler, T. W.; Zhou, C.; Alexseyev, L.; Kong, J.; Dai, H.; Liu, L.; Jayanthi, C. S.; Tang, M.; Wu, S. Nature 2000, 405, 769-772. (18) Minot, E. D.; Yaish, Y.; Sazonova, V.; Park, J.-Y.; Brink, M.; McEuen, P. L. Phys. ReV. Lett. 2003, 90, 156401-156404. (19) Maiti, A.; Svizhenko, A.; Anantram, M. P. Phys. ReV. Lett. 2002, 88, 1268051-1268054. (20) Yang, L.; Anantram, M. P.; Han, J.; Lu, J. P. Phys. ReV. B 1999, 60, 13874. (21) Groner, M. D.; Fabreguette, F. H.; Elam, J. W.; George, S. M. Chem. Mater. 2004, 16, 639-645. (22) Laermer, F.; Schilp, A. U.S. Patent 5501893. (23) Vinci, R. P.; Vlassak, J. J. Annu. ReV. Mater. Sci. 1996, 26, 431462. (24) Small, M. K.; Nix, W. D. J. Mater. Res. 1992, 7, 1553-1563. (25) Tripp, M. K.; Stampfer, C.; Miller, D.; Helbling, T.; Herrmann, C.; Hierold, C.; Gall, K.; George, S.; Bright, V. Sens. Actuators, A, accepted for publication, 2005. (26) Proost, J.; Saepen, F. J. Appl. Phys. 2002, 91, 204-215. (27) Chechenin, N.; Bootiger, J.; Krog, J. Thin Solid Films 1997, 304, 70-77. (28) Jaccodine, R. J.; Schlegel, W. A. J. Appl. Phys. 1966, 37, 6, 24292434. (29) Kovacs, G. Micromachined Transducers Sourcebook; WCB, McGrawHill: Boston, MA, 1998.

NL052171D

237