Controlled fabrication of carbon nanotube NO2 gas sensor using ...

Sensors and Actuators B 108 (2005) 398–403

Controlled fabrication of carbon nanotube NO2 gas sensor using dielectrophoretic impedance measurement Junya Suehiro∗ , Guangbin Zhou, Hiroshi Imakiire, Weidong Ding, Masanori Hara Department of Electrical and Electronic Systems Engineering, Graduate School of Information Science and Electrical Engineering, Kyushu University, 6-10-1 Hakozaki, Higashi-ku, Fukuoka 812-8581, Japan Received 13 July 2004; received in revised form 24 September 2004; accepted 30 September 2004 Available online 19 December 2004

Abstract The authors have previously demonstrated an electrokinetic fabrication method of a carbon nanotube (CNT) gas sensor using dielectrophoresis. One advantage of the technique was that one could quantify the amount of trapped nanotubes on a real time basis by monitoring electrical impedance of the sensor (dielectrophoretic impedance measurement, DEPIM). In the present study, we extended the DEPIM technique to controllable assembly of the carbon nanotube gas sensor. This realized a production of CNT gas sensors with identical electrical properties such as initial conductance. The gas sensor response to ppm-level nitrogen dioxide (NO2 ) gas was investigated with various values of the initial conductance. It was found that relative conductance change of the CNT gas sensor after NO2 exposure increased almost proportionally with the initial conductance for a constant NO2 concentration. This enabled to define intrinsic sensitivity of CNT sensors by normalization. It was found that a single-wall CNT gas sensor had higher normalized sensitivity than a multi-wall CNT sensor. © 2004 Elsevier B.V. All rights reserved. Keywords: Carbon nanotubes; Gas sensor; Dielectrophoresis; Impedance; Normalized sensitivity; Nitrogen dioxide (NO2 )

1. Introduction Because of the one-dimensional structure, carbon nanotube (CNT) [1] has distinctive properties in mechanical, chemical and electronic aspects. This makes the CNT a promising material for a variety of potential applications. Recently, CNT-based gas sensors [2–10] have received considerable attention because of their outstanding properties such as faster response, higher sensitivity, and lower operating temperature. The authors have demonstrated that the CNT gas sensor could be fabricated by dielectrophoresis [9,10]. Dielectrophoresis (DEP) is the electrokinetic motion of dielectrically polarized materials in non-uniform electric fields and has been successfully applied to manipulation of biological particle such as cells, bacteria and DNA [11,12]. There have been some attempts to manipulate the CNTs using DEP for ∗

Corresponding author. Tel.: +81 92 642 3912; fax: +81 92 642 3964. E-mail address: [email protected] (J. Suehiro).

0925-4005/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2004.09.048

separation, orientation, alignment and positioning of CNTs [13–19]. In our previous studies, it was demonstrated that the DEP fabrication could provide a way to trap and to fix CNTs on the microelectrode and could establish a good electrical connection between CNTs and the external measuring circuit. The DEP-fabricated CNT gas sensor successfully detected ppmlevel NH3 , NO2 , SO2 , HF and decomposition products of SF6 gas generated by electrical discharge [9,10]. One advantage of the ac electrokinetic manipulation technique was that one could quantify the amount of trapped nanotubes on a real time basis by monitoring electrical impedance of the sensor [9,17] (dielectrophoretic impedance measurement, DEPIM). The DEPIM was originally developed by the authors for electrical inspection of bacteria or micro-organisms [20,21]. In general, the CNT-based gas sensing utilizes the electrical conductance change caused by the gas adsorption as the electrical readout. This means that the sensor response can be dependent on the amount of CNTs, which bridge over the

J. Suehiro et al. / Sensors and Actuators B 108 (2005) 398–403

metal electrodes connected to the external measuring instruments. However, the effects of the CNT amount have not been well clarified, probably because there has been no effective way to control and quantify the number of CNTs retained on the sensor electrode. In this study, the gas sensors composed of single-wall CNT (SWCNT) or multi-wall CNT (MWCNT) were fabricated by the DEP manipulation. The amount of trapped CNTs was quantified by DEPIM as the conductance increase during the DEP process and controlled by adjusting the DEP duration. The CNT sensors were tested for detection of ppm-level nitrogen dioxide (NO2 ) at room temperature. It was found that the conductance change of the CNT sensors after NO2 exposure increased almost proportionally with the initial conductance for a constant gas concentration. It was found that the SWCNT sensor had higher normalized sensitivity than the MWCNT sensor.

2. Experimental Details of fabrication method of the CNT gas sensor using positive DEP have been described elsewhere [9]. MWCNTs (purchased from Nilaco Corp., Japan) had 20 nm average diameter, 5–20 ␮m length and 95% purity. SWCNTs (purchased from Sigma–Aldrich Co., USA) had 1 nm average diameter (15 nm as a bundle), 1–4 ␮m length and 50% purity. The CNTs were suspended in ethanol (1.0 ␮g/ml final concentration) and ultra-sonicated for 60 min. An interdigitated microelectrode of thin chrome film was patterned on a glass substrate by photolithography technique. The electrode finger had a castle-wall pattern and had 5 mm length and 5 ␮m the minimum clearance. The 20 electrode fingers formed 19 castellated gaps. The castle-wall electrodes were surrounded by a silicon rubber spacer to form a sealed chamber (15 ␮l capacity) in which the CNT suspension was stored. The CNT suspension was continuously fed into the microelectrode chamber from a reservoir by a pump at a flow rate of 0.5 ml/min. The DEP trapping of CNTs to the microelectrode was performed with ac voltage of 100 kHz frequency and 10 V amplitude (peak to peak value). During the DEP process, the electrode impedance was continuously monitored using a lock-in amplifier (model 7280, PerkinElmer Instruments, USA) with a sampling interval of 2 s. After a desired time period, the DEP process was stopped and ethanol was gently dried out at room temperature to prepare the microelectrode retaining CNTs as a gas sensor. In the gas detection experiments, NO2 gas, which has been successfully detected by CNT gas sensors [2–5], was employed as a sample gas. NO2 gas concentration was controlled by a mass flow controller employing N2 as a carrier gas. The same microelectrode chamber as used in the DEP fabrication process was employed also for the gas sensing experiments. The chamber was firstly filled with pure N2 gas and then NO2 was introduced. The flow rate of N2 and NO2 was kept constant at 0.2 l/min. The sensor impedance was continuously measured

399

at room temperature and 100 kHz frequency using the same system employed during the DEP sensor fabrication.

3. Results 3.1. Dielectrophoretic manipulation of the CNTs Scanning electron microscope (SEM) images of MWCNTs and SWCNTs trapped on the microelectrode are shown in Fig. 1. In a few minutes after the DEP process was started, CNTs were trapped by positive DEP around the electrode corners where the electric field became higher (Fig. 1a) [9,17]. The trapped CNTs seemed to be aligned along the electric field line and bridged over the electrode gap. The number of trapped CNTs increased with elapsed time. After 3 h, the captured CNTs covered a wider area of the gap space (Fig. 1b). As shown in Fig. 1c, the SEM image of the SWCNTs was not so clear because of the smaller diameter and more impurities such as micro-sized carbon particles. However, a precise observation revealed that SWCNTs were also trapped in the electrode gap. The SEM images also revealed that DEP trapped CNTs could be firmly immobilized on the microelectrode even after the ethanol evaporation. Fig. 2 shows the temporal increase of the electrode conductance measured during the DEP trapping process under an identical condition (CNT concentration, electrode potential, etc.). As one might expect from Fig. 1, the conductance increase might be attributed to the DEP-trapped CNTs bridging the electrode gap [5,9,17]. The conductance was much more increased as the density of CNT trapped in the electrode gap became higher. For MWCNT, the conductance increase was faster than SWCNT. When the conductance increase became a desired value GT , the DEP process was stopped and ethanol was gently dried out. As summarized in Fig. 3, the conductance G0 measured after ethanol evaporation was higher than GT measured in ethanol. This suggests that electrical contact between the metal electrode and the CNT is poor in liquid [9,19]. Since G0 increased almost proportionally with GT , the conductance G0 could be controlled as desired by the real time monitoring of GT during the DEP process. 3.2. NO2 detection using carbon nanotube gas sensors Fig. 4 shows the conductance response of the CNT gas sensors to 1 ppm NO2 gas. In order to examine effects of the CNTs amount retained on the sensor electrode, two sensors were fabricated and tested for MWCNT and SWCNT, respectively. One sensor was fabricated so that the conductance increase G0 measured after the CNT trap and ethanol evaporation (called as ‘initial conductance’ below) became about 10 ␮S (Fig. 4a), while the other one about 100 ␮S (Fig. 4b). That is, the latter retained more CNTs in the electrode gap than the former. When the sensor was exposed to NO2 gas, the conductance increased for all CNT sensors. The conductance increase of CNTs by NO2 exposure suggested

400


Fig. 1. SEM images of CNTs trapped onto a microelectrode by positive DEP: (a) MWCNTs after 10 min of the DEP process; (b) MWCNTs after 3 h of the DEP process; (c) SWCNTs after 3 h of the DEP process. The insets are magnified images near the electrode corners.

that both MWCNT and SWCNT were p-type semiconductors [2–5]. The conductance continuously increased even for one hour and did not saturate especially for the SWCNT sensor. When the NO2 gas was replaced with pure N2 , the conductance returned to the initial value very slowly (it took about 12 h, data are not shown). For the same type CNT, the sensor

conductance response became larger with the sensor initial conductance G0 , that is, with the amount of CNTs. It was also found that the conductance change of the SWCNT sensor was larger than that of the MWCNT sensor. Similar experiments were conducted for various NO2 concentrations and summarized in Fig. 5. Since the conductance increase

Fig. 2. Temporal variation of the electrode conductance GT measured during the DEP trapping process of MWCNTs and SWCNTs under an identical condition.

Fig. 3. Relationship between the conductance measured before (GT ) and after (G0 ) ethanol evaporation.


401

Fig. 6. Relationship between the initial conductance G0 and the sensor response G to 1 ppm NO2 gas.

Fig. 4. Conductance response of MWCNT and SWCNT gas sensors to NO2 gas (1 ppm) measured at room temperature: measured by CNT sensors with the initial conductance G0 about (a) 10 ␮S; (b) 100 ␮S.

did not saturate as shown in Fig. 4, the CNT sensors were calibrated using the conductance response after 9 min exposure to NO2 gas (defined as G below). The calibration time (9 min) was determined so that the conductance increase could be high enough to be measured even for lower values of NO2 concentration and the initial conductance G0 . The sensor response G was almost proportional to NO2 concentration

below 1 ppm and then gradually saturated at higher concentration. The largest G was achieved by the SWCNT sensor with the initial conductance G0 about 100 ␮S. Below 1 ppm, G was almost proportional to NO2 concentration with a coefficient of 200 ␮S/ppm. Since the conductance could be measured with 0.1 ␮S accuracy, it was estimated the SWCNT gas sensor potentially had ppb sensitivity to NO2 gas at room temperature.

4. Discussion As depicted in Figs. 4 and 5, the CNT sensor response depended not only on the structure (SWCNT or MWCNT) but also on the initial conductance G0 , which seemed to increase with the number of DEP-trapped CNTs. The initial conductance dependency of the sensor response G was further investigated for 1 ppm NO2 and summarized in Fig. 6. The CNT sensor response increased almost proportionally with the initial conductance. This implies that the sensor response normalized by the initial conductance, G/G0 , can be a measure for intrinsic sensitivity of the CNT gas sensor [3]. The normalized CNT sensor response may be explained using a simple equivalent circuit model depicted in Fig. 7. The DEP-trapped CNTs bridging the electrodes are modeled as the parallel sum of individual CNT (Fig. 7a). The sensor initial conductance G0 before the gas adsorption can be expressed as G0 = Ng0

Fig. 5. Conductance response of MWCNT and SWCNT gas sensors G measured after 9 min exposure to NO2 gas with various concentration: measured by CNT sensors with the initial conductance G0 about (a) 10 ␮S; (b) 100 ␮S.

(1)

where N is the total number of CNTs and g0 is the average conductance of one CNT. It was implicitly assumed that all CNTs had the same electrical property and that impurities other than the CNTs did not contribute to the initial conductance G0 . After NO2 adsorption, the conductance of individual CNT may be increased by g on average because the positive hole density increases due to electron transfer from p-type CNTs to oxidative NO2 molecules [2–5]. As a result,

402


Fig. 8. Normalized response of CNT gas sensors G/G0 measured for various NO2 concentration.

Fig. 7. An equivalent circuit model of a CNT gas sensor: (a) before NO2 adsorption; (b) after NO2 adsorption.

the sensor conductance is increased by G, which is given by G = N g

(2)

From Eqs. (1) and (2), one can readily obtain G g = G0 g0

(3)

The normalized sensor response G/G0 is equal to relative conductance change of single CNT caused by NO2 adsorption, g/g0 , which should be constant depending only on the intrinsic CNT properties but not on the number of CNTs. Thus, Eq. (3) predicts a linear relationship between G and G0 shown in Fig. 6. In the authors’ previous study, the MWCNT sensor response to NH3 gas was investigated for various values of G0 and we had the impression that G hardly depended on the initial conductance G0 [9]. This seems to be inconsistent with the present results for NO2 sensing. This may be due to the lower normalized sensitivity of the MWCNT sensor for NH3 than for NO2 (typically lower by one order of magnitude), which may make the initial conductance dependency obscure. As shown in Fig. 8, the normalized sensitivity G/G0 increased with NO2 concentration. The normalized sensitivity of the SWCNT sensor was two to six times higher than that of the MWCNT sensor. As mentioned above, the NO2 sensing by CNTs has been attributed to the electronic interaction between the chemisorbed NO2 molecules and semiconducting CNTs [2–5]. It is known that the electronic transport property of a CNT is strongly dependent on the diameter and the helicity [22]. In general, MWCNTs show a conducting (metallic) behavior at room temperature, while SWCNTs behave as semiconducting materials. However, it has been pointed

out that MWCNTs could contain some semiconducting tubes among predominant metallic ones [7,23]. It has been also reported that the semiconducting MWCNTs actually could be utilized as a gas sensor [7–10]. The higher normalized sensitivity of the SWCNT sensor may be attributed to higher abundance of the semiconducting tube, which is responsible for the sensor response. One advantage of the DEP fabrication of a CNT gas sensor is that the number of trapped CNTs can be easily controlled by the DEP force or by the DEP trapping duration (Fig. 1). During the DEP process, the number of trapped CNTs can be quantified by impedance spectroscopy on a real time basis (Fig. 2). Although the normalized sensor response is not dependent on the number of CNTs, the absolute response G increases with G0 or the number of trapped CNTs. This can realize higher signal-to-noise ratio in the gas sensing and may improve the effective sensitivity. 5. Conclusions A CNT-based NO2 sensor was successfully fabricated by positive DEP on a microelectrode array. The CNT sensor could detect NO2 gas at ppm-level and potentially at ppblevel at room temperature. The CNT sensor response to NO2 gas became larger as the initial conductance or the amount of the DEP-trapped CNT was increased. The normalized sensitivity was defined as the sensor response normalized by the initial conductance in order to evaluate intrinsic sensitivity of the CNT sensor. The normalized sensitivity of the SWCNT sensor was higher than that of the MWCNT sensor probably because SWCNTs contained more semiconducting tubes. The DEP fabrication and impedance monitoring could control the initial conductance of the CNT sensor and the resultant sensor response. Acknowledgments This work was partly supported by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion


of Science (No. 16360207), a fellowship from the Iketani Science and Technology Foundation of Tokyo, Japan (No. 0161009A). and the 21st Century COE Program, ‘Reconstruction of Social Infrastructure Related to Information Science and Electrical Engineering’, from the Ministry of Education, Culture, Sports, Science and Technology, Japan.

References [1] S. Iijima, Helical microtubules of graphitic carbon, Nature 354 (1991) 56–58. [2] J. Kong, N.R. Franklin, C.W. Zhou, M.G. Chapline, S. Peng, K.J. Cho, H.J. Dai, Nanotube molecular wires as chemical sensors, Science 287 (2000) 622–625. [3] P. Qi, O. Vermesh, M. Grecu, A. Javey, Q. Wang, H. Dai, S. Peng, K.J. Cho, Toward large arrays of multiplex functionalized carbon nanotube sensors for highly sensitive and selective molecular detection, Nano Lett. 3 (2003) 347–351. [4] L. Valentini, I. Armentano, J.M. Kenny, C. Cantalini, L. Lozzi, S. Santucci, Sensors for sub-ppm NO2 gas detection based on carbon nanotube thin films, Appl. Phys. Lett. 82 (2003) 961–963. [5] J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, M. Meyyappan, Carbon nanotube sensors for gas and organic vapor detection, Nano Lett. 3 (2003) 929–933. [6] K.G. Ong, K. Zeng, C.A. Grimes, A wireless passive carbon nanotube-based gas sensor, IEEE Sens. J. 2 (2002) 82–88. [7] O.K. Varghese, P.D. Kichambre, D. Gong, K.G. Ong, E.C. Dickey, C.A. Grimes, Gas sensing characteristics of multi-wall carbon nanotubes, Sens. Actuators B 81 (2001) 32–41. [8] S. Chopra, A. Pham, J. Gaillard, A. Parker, A.M. Rao, Carbon nanotube-based resonant-circuit sensor for ammonia, Appl. Phys. Lett. 80 (2002) 4632–4637. [9] J. Suehiro, G. Zhou, M. Hara, Fabrication of a carbon nanotubebased gas sensor using dielectrophoresis and its application for ammonia detection by impedance spectroscopy, J. Phys. D: Appl. Phys. 36 (2003) L109–L114. [10] J. Suehiro, G. Zhou, M. Hara, Detection of partial discharge in SF6 gas using a carbon nanotube-based gas sensor, Sens. Actuators B 105 (2) (2005) 164–169. [11] R. Pethig, Dielectrophoresis: using inhomogeneous AC electrical fields to separate and manipulate cells, Crit. Rev. Biotechnol. 16 (1996) 331–348. [12] M.P. Hughes, AC electrokinetics: applications for nanotechnology, Nanotechnology 11 (2000) 124–132. [13] K. Yamamoto, S. Akita, Y. Nakayama, Orientation of carbon nanotubes using electrophoresis, Jpn. J. Appl. Phys. 35 (1996) L917–L918. [14] K. Bubke, H. Gnewuch, M. Hempstead, J. Hammer, M.L.H. Green, Optical anisotropy of dispersed carbon nanotubes induced by an electric field, Appl. Phys. Lett. 71 (1997) 1906–1908. [15] T. Prasse, J.Y. Cavaille, W. Bauhofer, Electric anisotropy of carbon nanofibre/epoxy resin composites due to electric field induced alignment, Compos. Sci. Tech. 63 (2003) 1835–1841. [16] X.Q. Chen, T. Saito, H. Yamada, K. Matsushige, Aligning singlewall carbon nanotubes with an alternating-current electric field, Appl. Phys. Lett. 78 (2001) 3714–3716. [17] X. Liu, J.L. Spencer, A.B. Kaiser, W.M. Arnold, Electric-field oriented carbon nanotubes in different dielectric solvents, Curr. Appl. Phys. 4 (2004) 125–128.

403

[18] F. Wakaya, T. Nagai, K. Gamo, Position control of carbon nanotube using patterned electrode and electric field, Microelect. Eng. 63 (2002) 27–31. [19] R. Krupke, F. Hennrich, H.B. Weber, D. Beckmann, O. Hampe, S. Malik, M.M. Kappes, H.V. L¨ohneysen, Contacting single bundles of carbon nanotubes with alternating electric fields, Appl. Phys. A 76 (2003) 397–400. [20] J. Suehiro, R. Yatsunami, R. Hamada, M. Hara, Quantitative estimation of biological cell concentration suspended in aqueous medium by using dielectrophoretic impedance measurement method, J. Phys. D: Appl. Phys. 32 (1999) 2814–2820. [21] J. Suehiro, D. Noutomi, M. Shutou, M. Hara, Selective detection of specific bacteria using dielectrophoretic impedance measurement method combined with an antigen–antibody reaction, J. Electrostat. 58 (2003) 229–246. [22] J.-C. Charlier, J.-P. Issi, Electronic structure and quantum transport in carbon nanotubes, Appl. Phys. A 67 (1998) 79–87. [23] K. Kaneto, M. Tsuruta, G. Sakai, W.Y. Cho, Y. Ando, Electrical conductivities of multi-wall carbon nano tubes, Synth. Met. 103 (1999) 2543–2546.

Biographies Junya Suehiro received MS and doctor degrees in electrical engineering from Kyushu University in 1985 and 1991, respectively. He was with the Nippon Steel Corporation from 1985 to 1988. Since 1988, he has been at Kyushu University and now an associate professor. His current research interests are applied electrostatics in Bio MEMS and nanotechnology.

Guangbin Zhou received a BS degree in electrical engineering from Shenyang Institute of Technology, China in 1995. He was with the Shenyang Light Industries Research & Design Institute from 1995 to 1998. He received a MS degree in electrical engineering from the Kyushu University in 2002. Currently, he is a doctor course student in Kyushu University. His current research interests are dielectrophoretic manipulation of micro and nanoscale material including biological cells and carbon nanotubes. Hiroshi Imakiire received a BS degree in electrical engineering and computer science from Kyushu University in 2003. Currently, he is a master course student in Kyushu University. His current research interests are electrokinetic manipulation of carbon nanotubes and the application to the gas sensors. Weidong Ding received his BS and MS degrees in high voltage and insulation technology from Xi’an Jiaotong University, China in 1997 and 2000, respectively. Since 2000, he has been at Xi’an Jiaotong University and now a lecturer. In 2003 he became a research student and currently is a doctor course student in Kyushu University under support of the Okazaki Kaheita International Scholarship Foundation. His current research interests are electrokinetic manipulation of carbon nanotubes and the application to diagnosis of gas-insulated swithgear (GIS). Masanori Hara received MS and doctor degrees in electrical engineering from Kyushu University in 1969 and 1972, respectively. Since 1986, he is a professor of Kyushu University. His current research interests are high voltage engineering, superconductivity engineering and pulsed power engineering.