Simultaneous Dielectrophoretic Separation and Assembly of Single

Nanodevices, College of Chemistry and Molecular Engineering, and School of Physics, Peking University,. Beijing 100871, P. R. China. By using the spec...
0 downloads 0 Views 423KB Size
Anal. Chem. 2006, 78, 8069-8075

Simultaneous Dielectrophoretic Separation and Assembly of Single-Walled Carbon Nanotubes on Multigap Nanoelectrodes and Their Thermal Sensing Properties Zhuo Chen,† Zhongyun Wu,† Lianming Tong,† Huapu Pan,‡ and Zhongfan Liu*,†

Center for Nanoscale Science and Technology (CNST), Beijing National Laboratory for Molecular Sciences (BNLMS), State Key Laboratory for Structural Chemistry of Unstable and Stable Species, Key Laboratory for the Physics and Chemistry of Nanodevices, College of Chemistry and Molecular Engineering, and School of Physics, Peking University, Beijing 100871, P. R. China

By using the specifically designed multigap nanoelectrodes, we demonstrated an effective approach for the simultaneous dielectrophoretic separation and assembly of metallic and semiconducting single-walled carbon nanotubes (SWNTs). An approximate metallic-semiconducting-metallic multiarray structure was created by an inward-propagative sequential assembly of SWNTs under ac electric field. Such kinds of SWNT multiarray structures exhibited ultra-low-power consumption and excellent thermal sensing performances with the sensitivity being dependent on the number of gaps: the more gaps, the higher sensitivity. The effective separation of metallic and semiconducting tubes in different gaps is believed to be responsible for the improved sensitivity to temperature. Single-walled carbon nanotubes (SWNTs) have demonstrated great potentials in fabricating various sensing, nanoelectronic, and nanoelectromechanical devices because of their unique physical and chemical properties.1-6 The principal challenges of such applications are the effective separation of metallic and semiconducting nanotubes and the controlled assembly into desired architectures. Current synthesis technology always produces a mixture of SWNTs having ∼33% metallic and ∼67% semiconducting nanotubes with different chiralities.7 Separation of specific nanotubes from the bulk mixture is critical for making a full use of the excellent performances of SWNTs and indispensable for some applications such as fabricating nanoelectronic logic devices * To whom correspondence should be addressed. Tel & Fax: 00-86-10-62757157. E-mail: [email protected]. † College of Chemistry & Molecular Engineering. ‡ School of Physics. (1) Tans, S. J.; Verschueren, A. R. M.; Dekker, C. Nature 1998, 393, 49-52. (2) Rueckes, T. T.; Kim, K.; Joselevich, E.; Tseng, G. Y.; Cheung, C.; Lieber, C. M. Science 2000, 289, 94-97. (3) Heinze, S.; Tersoff, J.; Martel, R.; Derycke, V.; Appenzeller, J.; Avouris, Ph. Phys. Rev. Lett. 2002, 89, Article 106801. (4) Javey, A.; Guo, J.; Wang, Q.; Lundstrom, M.; Dai, H. J. Nature 2003, 424, 654-657. (5) Lee, K.; Wu, Z.; Chen, Z.; Ren, F.; Pearton, S. J.; Rinzler, A. G. Nano Lett. 2004, 4, 911-914. (6) Fennimore, A. M.; Yuzvinsky, T. D.; Han, W. Q.; Fuhrer, M. S.; Cumings, J.; Zettl, A. Nature 2003, 424, 408-410. (7) Saito, R.; Dresselhaus, M. S.; Dresselhaus, G. Physical Properties of Carbon Nanotubes; Imperial College Press: London, 1998. 10.1021/ac0614487 CCC: $33.50 Published on Web 10/31/2006

© 2006 American Chemical Society

and interconnects.1-4 Development of practical assembly technology has been one of the hot topics of SWNT research, and a number of approaches have been used, including chemical assembly via surface chemistry,8,9 electrophoretic10,11 and dielectrophoretic12-17 assembly, surface-directed assembly,18,19 and DNAguided assembly.20 Among these assembly techniques, the dielectrophoretic approach is particularly promising because of its additional capability of separating metallic and semiconducting nanotubes.12 In this work, we demonstrate the use of dielectrophoretic force for simultaneous separation and assembly of SWNTs on the specially designed multigap nanoelectrodes. Electrode geometry design is the key for modulating the electric field distribution for this purpose. On the basis of theoretical analysis of dielectrophoretic force and electric field dependence of different carbon nanotubes, a multigap nanoelectrode structure was fabricated by using focused ion beam (FIB) lithography, which has been proven to be effective for the simultaneous separation and assembly of mixed SWNTs. Thus-obtained SWNT multiarrays take a metallic-semiconducting-metallic nanotube alignment, which showed excellent thermal sensing performances as compared with the reported one-gap carbon nanotube thermal sensors.21 (8) Liu, Z. F.; Shen, Z. Y.; Zhu, T.; Hou, S. F.; Ying, L. Z.; Shi, Z. J.; Gu, Z. N. Langmuir 2000, 16, 3569-3573. (9) Wu, B.; Zhang, J.; Wei, Z.; Cai, S. M.; Liu, Z. F. J. Phys. Chem. B 2001, 105, 5075-5078. (10) Chen, Z.; Yang, Y. L.; Wu, Z. Y.; Luo, G.; Xie, L. M.; Liu, Z. F.; Ma, S. J.; Guo, W. L. J. Phys. Chem. B 2005, 109, 5473-5477. (11) Kamat, P. V.; Thomas, K. G.; Barazzouk, S.; Girishkumar, G.; Vinodgopal, K.; Meisel, D. J. Am. Chem. Soc. 2004, 126, 10757-10762. (12) Krupke, R.; Hennrich, F.; Lohneysen, H. V.; Kappes, M. M. Science 2003, 301, 344-347. (13) Chen, X. Q.; Saito, T.; Yamada, H. Appl. Phys. Lett. 2001, 78, 3714-3716. (14) Nagahara, L. A.; Amlani, I.; Lewenstein, J.; Tsui, R. K. Appl. Phys. Lett. 2002, 80, 3826-3828. (15) Diehl, M. R.; Yaliraki, S. N.; Beckman, R. A.; Barahona, M.; Heath, J. R. Angew. Chem., Int. Ed. 2002, 41, 353-356. (16) Tang, J.; Gao, B.; Geng, H. Z.; Velev, O. D.; Qin, L. C.; Zhou, O. Adv. Mater. 2003, 15, 1352-1355. (17) Chen, Z.; Yang, Y. L.; Chen, F.; Qing, Q.; Wu, Z. Y.; Liu, Z. F. J. Phys. Chem. B 2005, 109, 11420-11423. (18) Rao, S. G.; Huang, L.; Setyawan, W.; Hong, S. Nature 2003, 425, 36-37. (19) Ko, H.; Peleshanko, S.; Tsukruk, V. V. J. Phys. Chem. B 2004, 108, 43854396. (20) Keren, K.; Berman, R. S.; Buchstab, E.; Sivan, U.; Braun, E. Science 2003, 302, 1380-1382.

Analytical Chemistry, Vol. 78, No. 23, December 1, 2006 8069

EXPERIMENTAL SECTION The SWNTs were synthesized by CVD method and were about 500 nm to 2 µm long. After several hours of ultrasonication, the SWNTs were dispersed in dimethylformamide (DMF) as small bundles (∼2-7 nm in diameter). The concentration of the SWNT solution was ∼2.7 mg/L. The multigap electrodes were fabricated by FIB lithography in a DB235 focused ion beam etching and depositing system (FEI Co.). Before dielectrophoresis experiments, the electrodes were treated in SC-1 solution of RCA standard cleaning procedure at 75 °C and washed with ultrapure water. The electrodes were then immersed in SWNT solution and a 5-MHz ac electric field was applied for deposition of SWNTs. Finally, the electrodes were rinsed thoroughly with DMF and dried with N2 gas. Micro-Raman experiments were done with a Renishaw 2000 confocal Raman microscope (Renishaw plc) excited with a 20-mW Ar ion laser (514.5-nm excitation wavelength; Spectra Physics). The thermal sensing experiments were conducted in a UHV chamber (Omicron) equipped with a Keithley 4200 electrical measurement unit. A six-electrode setup in the UHV chamber was utilized for interconnection between the SWNT sensor and Keithley measurement unit. The measuring junction of the thermocouple was attached on the UHV sample clip, and the cold junction was at the intermediate reference point of 25 °C by immersion into a water bath circulator (Digital Uni Ace UA100, Eyela, Japan). The base pressure for all the measurements was less than 5 × 10-10 mbar, which was monitored by a MaxiGauge (Pfeiffer Vacuum). RESULTS AND DISCUSSION Dielectrophoresis (DEP) is a phenomenon where neutral and polarizable particles undergo mechanical motion in a liquid dielectric medium inside a nonuniform electric field.22 Depending on the relative polarizability of the particle to the medium, the dielectrophoretic force could push the particle toward regions of high or low electric field, termed positive and negative dielectrophoresis, respectively. For a quasi-one-dimensional (1D) and highly polarizable SWNT with its long axis aligned with the electric field, the time-averaged dielectrophoretic force can be expressed by eq 1,23 where Erms is the root-mean-square value of 2

2

ω2(cm - m ) + (σcσm - σm ) 2 π FDEP ) r2lm ∇E rms 2 2 2  ω2 + σ m

(1)

m

the electric field, ω is the angular frequency of the applied electric field,  and σ are the permittivity and conductivity, and the indices c and m refer to the carbon nanotube and the medium, respectively. The carbon nanotube here has a radius of r and a length of l, respectively. Equation 1 is only valid when the nanotubes are small in comparison to variations in the electric field. For an accurate determination of the dielectrophoretic force, a Maxwell stress tensor should be used.23 Equation 1 indicates that the magnitude of dielectrophoretic force acting on a SWNT is dependent on the applied voltage, more exactly on both the (21) Fung, K. M.; Wong, T. S.; Chan, R. H. M.; Li, W. J. IEEE Trans. Nanotechnol. 2004, 3, 395-402. (22) Pohl, H. A. Dielectrophoresis; Cambridge University Press: Cambridge, U.K., 1978. (23) Boote, J. J.; Evans, S. D. Nanotechnology 2005, 16, 1500-1505.

8070

Analytical Chemistry, Vol. 78, No. 23, December 1, 2006

strength and the gradient of the electric field, the angular frequency, and the mismatch between the dielectric permittivities and electrical conductivities of the nanotube and the medium. Metallic and semiconducting nanotubes have remarkably different permittivities and conductivities. As compared with the small permittivity value of semiconducting nanotubes, a metallic nanotube is predicted to have infinite permittivity though the finite length and the presence of defects leads to a finite permittivity value.12 For the conductivity in a diffusive regime, one may consider a conductivity level of 105 S‚m-1 for a semiconducting tube and a level of 108 S‚m-1 for a metallic tube.24 It is evident that the metallic and semiconducting tubes would experience greatly different dielectrophoretic forces under the same experimental conditions. This is actually the physical origin behind the dielectrophoretic separation of metallic and semiconducting SWNTs.12,24 On the other hand, a self-assembly of both metallic and semiconducting SWNTs would be observed in-between the electrode gap with large enough positive dielectrophoretic force to overcome the opposite electrothermal and viscous forces, as reported by the previous studies.12-17 In order to realize the simultaneous separation and assembly of SWNTs, the experimental configuration needs to satisfy the following: (1) positive and large enough dielectrophoretic force for both metallic and semiconducting tubes; (2) rational design of the electrode architecture for sequential assembly of the metallic and semiconducting tubes from the SWNT mixture controlled by the magnitude of dielectrophoretic force. With the above theoretical consideration, a multigap nanoelectrode was designed as illustrated in Figure 1a. Panels b-e in Figure 1 show the simulation result of the electric field distribution of a five-gap nanoelectrode, where the gap width was 800 nm, the gap height from the underlying SiO2 substrate was 150 nm, and the voltage across the outmost electrodes was 4 V. Obviously, the field distribution is highly inhomogeneous at such a nanostructured electrode. The field strength along the electrode surface rapidly decreases from the outmost to the middle gaps by an order of ∼2. Edge effect is also remarkable, as indicated by the sharp peaks at the edge locations, which is very important for the dielectrophoresis-driven assembly of SWNTs on the electrode edges. The high electric field gradient at the edges will cause a larger DEP force, which facilitates the assembling of SWNTs from solution. The selection of experimental ac frequency, applied voltage and solvent medium is the critical point of this study. The resultant dielectrophoretic force should have an appropriate value that enables the predominant assembly of metallic nanotubes at the outmost gaps at the beginning. Assembly would not occur at other gaps because of the much weaker field strength. This assembling process would be autoterminated because of the rapid decrease of electric field at the outmost gaps followed by filling in the metallic SWNTs. A similar phenomenon would occur for the second outmost gaps, the third, and so on. Such kind of sequential assembling process is expected to allow the simultaneous separation of metallic and semiconducting nanotubes. The optimized ac frequency, applied voltage, and solvent medium in this work were 5 MHz, 4 V, and DMF, respectively. Figure 2 shows the scanning electron microscope (SEM) image of the SWNT multiarray structure self-assembled on a five(24) Dimaki, M.; Boggild, P. Nanotechnology 2004, 15, 1095-1102.

Figure 1. (a) Multigap nanoelectrode structure for the purpose of simultaneous separation and assembly of SWNTs; (b) electric field distribution of a five-gap nanoelectrode simulated using FEMLAB program, where the gap width was 800 nm, the gap height from the underlying SiO2 substrate was 150 nm, and the voltage across the outmost electrodes was 4 V; (c) zoom-up of (b); (d) zoom-up of (c); (e) zoom-up of (d).

Figure 2. SEM image of the aligned SWNT multiarray structure self-assembled on a five-gap nanoelectrode under ac electric field of 5 MHz and of 4-V peak to peak voltage for 30-min assembly in DMF. The gaps have a width of 800 nm, a height of 150 nm, and a length of 3.6 µm, respectively. The bottom images are the zoom-up of the top. Scale bar is 5 µm.

gap nanoelectrode under an ac electric field of 5 MHz and of 4-V peak to peak voltage for 30-min assembly in DMF. The multigap Au nanoelectrode was fabricated on SiO2/Si substrate by FIB lithography, with a gap size of 800 nm in width, 150 nm in height, and 3.6 µm in length. Obviously all the gaps are filled with aligned SWNTs with their long axes parallel to the electric field. The density of aligned SWNTs was found to show a gradual decrease from the outmost to the middle gaps. This is attributed to the total decrease of electric field in magnitude because of the voltage drop on the filled outside gaps. When the number of gaps was increased from five to nine, no SWNTs were assembled on the

middle gaps (see Supporting Information, Figure S1a). The density of aligned SWNTs was also strongly affected by the gap width. When the gap width was reduced from 800 to 200 nm, with the number, length, and depth of the gap being the same as in Figure 2, only a few SWNTs were deposited on the five gaps for a 30min assembly in DMF (see Supporting Information, Figure S1b). This gap width dependence would be related to the relative dimension of gap width and tube length. The SWNTs used in this work were about 500 nm to 2 µm in length, much longer than the gap width. In the process of dielectrophoretic deposition, such long nanotubes were easily bridging the gaps, leading to the rapid decrease of gap resistance and therefore the decrease of electric field. As a result, the self-assembling process would be autoterminated more rapidly than the wide gap case, which resulted in the decrease of packing density of SWNTs. The motion of SWNTs under an ac electric field may consist of two fundamental steps. The first step is the alignment along the electric field direction via electrostatic interaction between the induced dipole moment and the electric field. Such a field-induced orientation process is very fast, having a time level of nanoseconds. The second step is the directional movement of SWNTs toward the high field gradient region driven by positive dielectrophoretic force. The velocity of SWNTs in a dielectrophoresis-driven process is determined by a lot of factors, including the field strength, the working frequency, the electrode geometry, the size and chirality of carbon nanotube, the viscosity of solvent medium, and the temperature.23-24 This process is relatively slow and may have a time level of seconds.25 Figure 3 illustrates the time dependence Analytical Chemistry, Vol. 78, No. 23, December 1, 2006

8071

of the packing density of SWNTs on a five-gap nanoelectrode having a gap size of 800 nm in width, 150 nm in height, and 4.5 µm in length under ac electric field. With decreasing the deposition time in DMF, the density of aligned SWNTs on the gaps becomes gradually decreased. For 5-min deposition, no SWNTs were found on the middle gap, while for 20-s deposition, the SWNTs were almost only assembled on the outmost gaps. This observation strongly suggests a sequential assembly mechanism of SWNTs on the multigap nanoelectrode under an ac electric field (see Figure 3d) as expected from the above theoretical analysis. The SWNTs subject to larger dielectrophoretic force first fill in the outmost gaps having the highest electric field. Together with the filling-up process, the electric field at the outmost gaps drops down rapidly and the ac voltage is then mainly applied to the rest gaps, which leads to the assembly of SWNTs on the second outmost gaps. Such an assembling process

propagates inward until all the gaps are filled with SWNTs with large enough electric field and long enough deposition time. In the above sequential assembling process, it is reasonable to expect a separation effect of mixed SWNTs considering the greatly different dielectrophoretic forces experienced by metallic and semiconducting tubes. The metallic SWNTs are subject to the largest dielectrophoretic force and therefore would be preferentially deposited on the outmost gaps. Together with the decrease of metallic tubes in DMF solution, the semiconducting tubes would be enriched. As a result, the density of aligned semiconducting SWNTs on the multigap nanoelectrode would become gradually increased in the inward-propagative assembling process. Proof of such a sequential assembly-induced separation of SWNTs is given by a Raman spectroscopy study. It has been shown that there are characteristic differences between the G bands of Raman spectra for metallic and semiconducting tubes.12,26 The graphite-like in-plane mode G bands of SWNTs show two dominant features between 1500 and 1600 cm-1: a low-frequency component ωG- and a high-frequency component ωG+. In metallic tubes, both ωG- and ωG+ are of equal intensity but ωG- is much broader, exhibiting an asymmetric Breit-Wigner-Fano line shape. While in semiconducting tubes, ωG+ is stronger in intensity than ωG- and both components show a Lorenzian line shape.8,20 Therefore the relative intensity of ωG+ and ωG- peaks, IG+/IGcan be approximately used to indicate the relative amount of metallic and semiconducting SWNTs in the assembly, the larger the value, the more the semiconducting tubes. Figure 4 shows the typical G bands Raman spectra obtained from a five-gap sample by averaging 10 single spectra measured on different spots in each gap. The relative intensity value, IG+/IG-, for each gap was calculated after two-peak fitting of the G bands. The results obtained are 2.3 and 2.2 for the outmost gap1 and gap1′, 2.6 and 3.2 for the second outmost gap2 and gap2′, and 3.4 for the middle gap3, respectively. Apparently, the semiconducting SWNTs are gradually enriched from the outmost to the middle gaps, consistent with the theoretical prediction. A similar conclusion has been deduced from all the multigap samples investigated. This experimental observation demonstrates the feasibility of achieving simultaneous separation and assembly of SWNTs by using a multigap nanoelectrode architecture. With carefully selected experimental conditions and electrode design, it is expected that metallic and semiconducting SWNTs are completely separated into different gaps, forming a metallic-semiconducting-metallic assembly structure. The actual experiments are complicated by the existence of length and diameter distribution of SWNTs, the defects, the formation of bundles, which may consist of both metallic and semiconducting tubes, and the sophisticated dependence of dielectrophoretic force on the working frequency. Carbon nanotube has low mass density, small size, enormous flexibility, and high response speed to the thermal variation,21 which make it an excellent material for temperature measurements compared with other types of sensors. Multi-walled carbon nanotubes’ (MWNTs) assembly has been demonstrated as novel thermal sensors with a power consumption of microwatts level.21 The thermal sensing performance of the present SWNT multiarrays dielectrophoretically assembled on the multigap nanoelec-

(25) Krupke, R.; Hennrich, F.; Weber, H. B.; Beckmann, D.; Hampe, O.; Malik, S.; Kappes, M. M.; Lohneysen, H. V. Appl. Phys. A 2003, 76, 397-400.

(26) Dresselhaus, M. S.; Dresselhaus, G.; Jorio, A.; Filho, A. G. S.; Saito, R. Carbon 2002, 40, 2043-2061.

Figure 3. (a-c) SEM images of the SWNT multiarray structures self-assembled on five-gap nanoelectrodes having a gap size of 800 nm in width, 150 nm in height, and 4.5 µm in length. The assembly time in DMF was (a) 30 min, (b) 5 min and (c) 20 s, respectively. The other experimental parameters were the same as in Figure 2. (d) Schematic illustration of the sequential assembly mechanism of SWNTs on a multigap nanoelectrode.

8072

Analytical Chemistry, Vol. 78, No. 23, December 1, 2006

Figure 4. Raman spectra of SWNTs at G band region obtained from a five-gap sample shown in Figure 3a by averaging 10 single spectra measured on different spots in each gap. The inset shows the SEM image of the sample with gap numbering for clarity.

trodes was also investigated for comparison. The outside metallic SWNT arrays could act as electrodes while the middle enriched semiconducting arrays would be as the major sensing element to detect the thermal variation. The unique in-series multiarrays structure and the metallic-semiconducting-metallic configuration are expected to greatly improve the sensitivity of thermal sensors. The thermal sensing experiments were conducted in a UHV chamber. The temperature was monitored by a thermocouple attached on the UHV sample clip. Both the resistance change of the SWNT multiarrays and the voltage of the thermocouple were measured against the sample temperature by a Keithley 4200 electrical measurement system. Before measurement, it was necessary to conduct a thermal annealing treatment to stabilize the initial resistance value. This annealing treatment led to a remarkable decrease of the resistance of the SWNT multiarrays, possibly originating from the improvement of the SWNTs-Au electrode contacts. All the thermal sensing experiments were performed after the resistance of SWNT multiarrays became unchanged by annealing. Figure 5a shows the thermal sensing performance of a five-gap SWNT multiarray upon heating and cooling cycling. The solid curve represents the change of the thermocouple voltage, which indicates the temperature alteration (right-hand y-axis), and the open-circled curve shows the change of current flow in the SWNT multiarrays (left-hand y-axis, the applied voltage was 0.5 V) in the heating and cooling process. Upon heating, the current flow in the SWNT multiarrays showed an immediate increase, while upon cooling, it showed a decrease consistently. Repeating experiments showed that such SWNT multiarrays have a very stable and sensitive response to the temperature change, demonstrating their capability as thermal sensing devices.

The sensitivity of a thermal sensor can be quantified by the temperature coefficient of resistance (TCR), a resistance change factor per degree of temperature change. For pure metals, the TCR has a positive value, suggesting that the resistance increases with increasing temperature, while for semiconductors, the TCR has a negative value, in which the resistance shows a decrease with increasing temperature. The TCR value (RT) of the SWNT multiarrays thermal sensor can be expressed as

RT ) (1/RT)(dRT/dT) × 100%

(2)

where RT is the resistance of the thermal sensor at temperature T. Replotting Figure 5a by a simple calculation, we got the resistance-temperature dependence as shown in Figure 5b. The TCR value was then estimated to be -0.26%/K at 300 K for the five-gap multiarrays thermal sensor. The negative TCR value indicates the semiconducting characteristics of the SWNT multiarrays structure as expected. For comparison, we investigated the thermal sensitivities of SWNT multiarrays sensors with different gaps. Figure 5b also showed the results on three-gap and one-gap samples, respectively. All the experimental conditions and the electrode geometry in Figure 5b were kept the same except the number of gaps. Obviously, the resistance showed a remarkable increase together with the increase of gap number. The sudden rapid increase of resistance in the five-gap sample may also suggest the effective separation of metallic and semiconducting tubes in the middle gaps. The TCR value was estimated to be -0.14%/K for the three-gap sample and -0.11%/K for the one-gap sample at 300 K, respectively. The TCR value for the one-gap SWNT array is on the same level as that reported on the MWNT array.21 It is obvious that the increase of gap numbers Analytical Chemistry, Vol. 78, No. 23, December 1, 2006

8073

Figure 5. Thermal sensing properties of the SWNT multiarray sensors. (a) Changes of the current flow (solid curve) and thermocouple voltage (open-circled curve) in the five-gap sample upon heating and cooling cycling. (b) showed the resistance-temperature dependences of five-, three-, and one-gap thermal sensors, respectively. The gap size and experimental conditions for all the samples were the same as in Figure 3a, and the applied voltage for current measurement was 0.5 V. (Note: different scaling was made in (b) for clarity.)

resulted in a remarkable improvement of the thermal sensitivity of the multigap SWNT array sensors. This can be easily understood by supposing that the multigap nanoelectrode architecture allows the effective separation of metallic and semiconducting tubes under ac electric field because the semiconducting tubes have a negative TCR value and are more sensitive to the temperature change than metallic tubes.27 Further power consumption evaluation has also been carried out. Similar to the TCR measurement results, together with the increase of gap numbers, the consumption of power is decreased to 15.2 × 10-6 (1-gap), 5.95 × 10-6 (3-gap), and 0.21 × 10-6 W (5-gap) at 300 K. The results obtained here clearly demonstrate the advantage of SWNT multiarray structures as a thermal sensing element over the one-

gap carbon nanotube sensors. Moreover, in comparison, the operation power of conventional microelectromechanical system polysilicon sensors is in the order of milliwatts,21,28 which suggested that SWNT array could be used as a resistive element for ultra-low-power consumption devices.

(27) Ebbesen, T. W.; Lezec, H. J.; Hiura, H.; Bennett, J. W.; Ghaemi, H. F.; Thio, T. Nature 1996, 382, 54-56.

(28) Liu, C.; Huang, J. B.; Zhu, Z.; Jiang, F.; Tung, S.; Tai, Y. C.; Ho, C. M. J. Microelectromech. Syst. 1999, 8, 90-99.

8074

Analytical Chemistry, Vol. 78, No. 23, December 1, 2006

CONCLUSIONS In summary, we have demonstrated an effective approach for the simultaneous dielectrophoretic separation and assembly of metallic and semiconducting SWNTs by using the specifically designed multigap nanoelectrodes. Based on the greatly different positive dielectrophoretic forces exerted on the metallic and semiconducting nanotubes, we have been able to perform an

inward-propagative sequential assembly of SWNTs under ac electric field, which generated an approximate metallic-semiconducting-metallic multiarray structure. Such kinds of SWNT multiarray structures exhibited ultra-low-power consumption and excellent thermal sensing performances with the sensitivity being dependent on the number of gaps: the more gaps, the higher sensitivity. The effective separation of metallic and semiconducting tubes in different gaps is believed to be responsible for the improved sensitivity to temperature. The present approach to fabricating SWNT multiarrays may also be extended to other nanotube and nanowire systems. ACKNOWLEDGMENT This work was supported by the National Natural Science Foundation of China (NSFC, 60301001, 90301006, 50521201) and

the Ministry of Science and Technology (MOST, 2001CB6105). SUPPORTING INFORMATION AVAILABLE Effects of gap number and gap width in the multigap nanoelectrodes on the assembling results of SWNTs, Figure S1; controllable diversity of aligned SWNTs in the gaps, Figure S2. These materials are available free of charge via the Internet at http://pubs.acs.org.

Received for review August 4, 2006. Accepted September 25, 2006. AC0614487

Analytical Chemistry, Vol. 78, No. 23, December 1, 2006

8075