In situ synthesized carbon nanotube networks on a microcantilever for sensitive detection of explosive vapors

Sensors and Actuators B 176 (2013) 141–148

Contents lists available at SciVerse ScienceDirect

Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb

In situ synthesized carbon nanotube networks on a microcantilever for sensitive detection of explosive vapors Wenzhou Ruan a , Yuanchao Li a , Zhimin Tan a , Litian Liu a , Kaili Jiang b , Zheyao Wang a,∗ a b

Institute of Microelectronics, and Tsinghua National Laboratory for Information Science and Technology, Tsinghua University, Beijing 100084, China Department of Physics, and Tsinghua-Foxconn Nanotechnology Research Center, Tsinghua University, Beijing 100084, China

a r t i c l e

i n f o

Article history: Received 1 July 2012 Received in revised form 29 September 2012 Accepted 4 October 2012 Available online 13 October 2012 Keywords: Carbon nanotubes Chemical sensor Explosive Microcantilever

a b s t r a c t This paper reports a chemical sensor that consists of a suspended microcantilever and a carbon nanotube (CNT) network that is in situ synthesized on the microcantilever. The in situ synthesis achieves minimized thermal resistance of the interface between the CNTs and the microcantilever, which together with the low thermal mass and the high thermal conductivity allows the sensor to operate in microcalorimeter mode. By heating the CNT networks with an integrated heater, the explosive vapors adsorbed on the CNT surfaces are ignited to deflagration, which releases extra heat to deflect the microcantilever and thus changes the resistance of an integrated piezoresistor. This thermal-mechanical coupling transduction mechanism bridges the heat changes on the CNT networks and the mechanical deflection of the microcantilever. The large surface to volume ratio of CNTs enables fast adsorption of the chemical sensor to explosive vapors and improved equivalent limit of detection (LOD). An equivalent LOD of 2.4 pg is achieved for TNT detection, and run-to-run repeatability and device-to-device reproducibility are better than 4% and 10%. The preliminary results demonstrate the feasibility of development of sensitive CNT chemical sensors using in situ synthesized CNT networks and thermal-mechanical coupling transduction mechanism. © 2012 Elsevier B.V. All rights reserved.

1. Introduction Detection of explosives, such as 2,4,6-trinitrotoluene (TNT) and 1,3,5-trinitro-1,3,5-triazacyclohexane (RDX), is a rapidly increasing task in forensics, anti-terrorist activities, global de-mining projects, and environment monitoring. Most of these applications, if not all, need potable detection technologies with low LOD, high selectivity, and fast response time. However, it is still a technical challenge to fulfill these requirements due to the extremely low vapor pressures of explosives, complex interferents, and various kinds of explosives. Among the large variety of detection technologies, miniaturized sensors based on micro and nano technologies have attracted considerable research attentions in recent years. In nano regime, carbon nanotubes (CNTs) has promised unprecedented opportunities for developing trace explosive sensors. Due to the high electronic conductivity for electron transfer reactions, CNTs have been used to modify the electrodes of electrochemical sensors to enhance electrochemical signals [1]. For example, multi-wall CNT (MWCNT) and Cu-single walled CNT (SWCNT) have been employed to modify glassy carbon electrodes (GCE), and electrochemical

∗ Corresponding author. Tel.: +86 10 62772748; fax: +86 10 62771130. E-mail address: [email protected] (Z. Wang). 0925-4005/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.snb.2012.10.026

sensors with TNT detection limits down to 0.6 ␮g/l and 1 ppb have been reported for 10 min preconditioning [2,3]. Individual CNTs and CNT networks have also been exploited to develop a large variety of chemical sensors with different configurations and transduction mechanisms such as FETs [4,5], chemiresistor [6], thin-film-transistor (TFT) [7,8], and optical detection [9]. High-performance chemical sensors using individual CNTs have been realized for detection of TNT in aqueous solution with a LOD of 1 fM [22] and nitrotoluene vapors with a LOD of 262 ppb [10], and they are facing the challenges of variation in electrical properties [11], manufacturing difficulties [12], and poor reproducibility [13,14]. Randomly distributed CNT networks, which are able to mitigate the deviations by averaging the electrical heterogeneity of individual CNTs [15,16], have attracted considerable research attention more recently to develop sensors for various volatile organic compounds (VOCs), nerve agents, and explosives [17–20]. To improve selectivity, metal nanoparticles [21] and self-assembled monolayers (SAMs) [22] of ligands, lipid membranes and peptide receptors, and conjugated polymers have been developed as surface coating reagents to functionalize CNT sidewalls to selectively detect TNT [22,23]. In micro scale, microcantilever sensors with SAMs immobilized on the surfaces have been extensively investigated for explosive detection by measuring either the deflection or the resonance

142

W. Ruan et al. / Sensors and Actuators B 176 (2013) 141–148

Fig. 1. Schematic illustration of the sensor configuration and the operation principle. (a) Explosive vapors adsorbed on the CNT sensors. (b) Heating explosive vapors to micro deflagration.

frequency shift induced by adsorption of target chemicals [24]. Microcantilevers operated in bending mode has achieved a LOD of 300 ppt for detection of 2,4-dinitrotoluene (DNT) using a selfassembled layer as receptors [25]. Microcantilevers coated with 4-mercaptobenzoic acid monolayer have been developed for detection of PETN and RDX vapors with LODs in the level of 30 ppt or a few femtograms [26]. To improve selectivity and sensitivity, nanomaterials such as nanoporous framework materials [27] and nanoporous zeolites [28,29] have been employed as surface coaters for detection of explosives or VOCs. A microcantilever with post-transferred MWCNTs have been developed, and by grafting special sensing groups on CNT sidewalls, a LOD of 4.6 ppb has been obtained for TNT detection [30]. Even selective coaters are used, microcantilevers based on chemical receptors normally suffer from interferences such as water molecules that occur at concentrations in the orders of magnitude higher than explosives. Microcalorimeters, which has been thoroughly investigated for measuring the thermal properties of chemical and biological reactions by analyzing the exothermic and endothermic characteristics of thermal reaction [31–33], were first employed for explosive detection in 1999 [34]. Thanks to the capability in measuring the thermal fingerprints of explosives, microcalorimeters are evolving as an important technique for trace explosive detection. Microcalorimeters with LOD down to 6 pg for TNT detection [35–40] and 3 ppm for combustible gases [41] have been achieved. However, the low volatility of most explosives and the small surfaces of the microcalorimeters need long times to adsorb enough explosives for detection, normally around several tens to hundreds seconds [35]. To overcome the shortcomings of microcantilevers using chemical receptors, this paper reports a microcantilever sensor with integrated CNT networks for explosive detection. The CNT networks are the in situ synthesized on the microcantilever surface rather than self-assembled (post transferred). Direct synthesis enables very low interface thermal resistance between CNTs and microcantilevers, and thus allows the microcantilevers to operate in micro-calorimetry mode by measuring the microcantilever deflection induced by the heat released from the deflagration of the TNT molecules adsorbed on the CNT surfaces. Since micro-calorimetry is able to detect the thermal fingerprints of explosives without the need of chemical receptors, the microcantilevers integrated with CNT networks avoid chemical receptors that are inevitable in normal microcantilever sensors for selectivity, and thus it is possible to achieve short recovery time, reusability, and immunity to interferences and sensor poisoning. Compared with literature reported microcalorimeters, the proposed sensor integrates CNT networks to efficiently adsorb TNT vapors for fast adsorption and detection by exploiting the extremely large surface areas of CNTs.

2. Configuration and operation principle The sensor is a suspended bi-material microcantilever that consists of two silicon dioxide layers and a silicon layer sandwiched in between. Two silicon resistors fabricated in the silicon layer act as a heater and a piezoresistor. A CNT film is in situ synthesized on the surface of the microcantilever. When the chemical sensor is exposed to explosives, the vapors are adsorbed onto the CNT surfaces, as shown in Fig. 1(a). By heating the microcantilever to the deflagration temperature of the explosives using the integrated heater, the CNT film is heated simultaneously to almost the same temperature due to the extremely high thermal conductivity. Thus, the vapors adsorbed on the CNTs are ignited to deflagration, which is an inherently exothermic process and generates a heat increment to the CNT film. This extra heat is conducted to the microcantilever and results in bending of the microcantilever as a result of bi-material structure, which can be detected by the integrated piezoresistors, as shown in Fig. 1(b). Using this thermalmechanical coupling transduction mechanism, the explosives can be detected by measuring the microcantilever deflection induced by the micro-deflagration. The transduction mechanism is based on the excellent thermal properties of the microcantilever and the extraordinary physical properties of CNTs. The extremely low thermal mass (around 10 nJ/K [42]) and the low thermal dissipation of the microcantilever allow it to be heated to several hundreds of degrees in milliseconds by the integrated heater to ignite the adsorbed explosive vapors, and allow a distinct deflection to be generated by the infinitesimal heat released from micro-deflagration. The high thermal conductivity of CNTs and the low interface thermal resistance between the CNTs and the microcantilever allow fast heat transfer between the microcantilever and the CNT networks, which is critical to achieve high sensitivity that is proportional to the heating rates. The huge surface to volume ratios of CNTs, as well as the fact that the aromatic explosives tend to be physisorbed onto CNTs through ␲-stacking interaction [43], enhances the adsorption ability to TNT vapors. 3. Sensor fabrication The microcantilever is fabricated from a silicon-on-insulator (SOI) wafer [44]. Integrated heaters and piezoresistors are fabricated on the SOI device layer by ion implantation, which together with a buried oxide layer and a silicon dioxide passivation layer constitutes the mainbody of the microcantilever. The CNTs are locally synthesized on the surface of the microcantilever using a custom-built laser-assisted chemical vapor deposition (LACVD) system [45]. The extremely small heat mass and the minimized thermal dissipation allow the microcantilever to be heated rapidly to high temperatures (600–800 ◦ C) for CNT synthesis by focusing a

W. Ruan et al. / Sensors and Actuators B 176 (2013) 141–148

143

Fig. 2. (a) A microcantilever with dimensions of 70 ␮m (length) × 30 ␮m (width) × 1.6 ␮m (thickness). (b) The SEM photo the synthesized CNTs on the microcantilever. (c) The TEM photo of the multi-walled CNTs.

Fig. 3. (a) A CNT sensor packaged in a chamber. (b) Typical responses of the reference and the sensing device. The inset shows the deflection signal after subtracting the reference from the sensing device.

laser beam on the microcantilever, whereas other areas are out of focus and remain at low temperatures. This ensures local high temperature and self-confined CNT synthesis on the microcantilever, and protects the devices on the substrate from being destroyed. The growth rate of the CNTs is about 3 ␮m/min. Fig. 2(a) shows the as-fabricated microcantilever. The length and the width of the microcantilever are 70 ␮m and 30 ␮m, respectively. The thicknesses of the silicon, the top silicon dioxide, and

(a)

3.276

x 10

the bottom silicon dioxide layers are 200 nm, 1000 nm and 400 nm, respectively, creating an asymmetrical structure across the thickness. Fig. 2(b) and (c) shows the SEM photo of the CNT networks on the microcantilever and the TEM photo of the MWCNTs, respectively. CNTs resemble sparse grass with random distribution and porous structures. When CNTs are longer than 10 ␮m, they tend to be densely-packed and vertically aligned due to Van der Waals forces.

4

(b)

3

3.2755

Mass (m/µg)

Frequency (f/Hz)

Experimental Linear Fitting

2.5

Adsorption Begin

3.275

3.2745

3.274

2 1.5 1 Adsorption Begin

3.2735

0.5

3.273

3.2725

3.5

0

0

100

200

Time (t/min)

300

-0.5 0

100

200

300

Time (t/min)

Fig. 4. (a) Frequency change of a tuning fork vs. the adsorption time of TNT vapors. (b) TNT mass calculated from the frequency changes.

144

W. Ruan et al. / Sensors and Actuators B 176 (2013) 141–148

(b)

50 0 min 5 min 7.5 min 10 min 12.5 min 15 min

Amplitude (mV)

40 30 20 10 0 -10 -20 0

50 0 min 5 min 7.5 min 10 min 12.5 min 15 min

40

Amplitude (mV)

(a)

30 20 10 0 -10

20

40

60

80

100

-20 0

20

Time (ms)

40

60

80

100

Time (ms)

Fig. 5. The changes in the voltages versus the adsorption times of a CNT sensor and of a bare microcantilever. (a) Bare microcantilever without CNTs. (b) CNT sensor (microcantilever with CNTs).

The in situ synthesis has the following advantages. First, it minimizes the interface thermal resistance between the CNTs and the substrate, and thus enables fast heating rates that is critical to obtain low LODs. Second, the in situ synthesized CNT networks use a great number of CNTs to minimize the statistical uncertainty, and are more uniform and reproducible than those transferred from liquid. Third, it avoids the liquid deposition methods for CNT network preparation, which are inapplicable to suspended structures because they tend to stick to the substrate in liquids as a result of the adhesion phenomenon.

microcantilever relative to the reference, as shown in Fig. 3(b). The TNT deflagration is an exothermic process and releases heat to the CNT networks, which causes additional downwards deflection of the sensing microcantilever relative to the reference. Using a 50 ms pulse voltage as the heating power, the explosive deflagrates within 40 ms, and after deflagration the CNT sensor instantly restores to the initial state within 10 ms due to the minute heat mass and the large surface to volume ratio.

4. TNT vapor detection

Fig. 5 shows the measured voltages as a function of the TNT adsorption times at the same vapor concentration. For comparison, microcantilevers without CNTs (bare microcantilever) are also tested for CNT detection. It is clear that the peak voltages increase with the absorption time, indicating an increasing amount of adsorbed TNT. For 5 min adsorption, the peak voltage and the SNR of the bare microcantilever are 11.5 mV and 11.7, respectively, whereas those of the CNT sensor are 24 mV and 26.4, indicating a 2.3 times improvement in the LOD. This improvement is associated with the increased amount of TNT. The LOD of the bare microcantilever for TNT detection, which corresponds to the quantity of TNT to produce a SNR of 3 [46], can be obtained as 24 × 5 × 3/11.7 = 30.8 pg. This result is consistent with the literature reports [35,39]. Similarly, an equivalent LOD of the CNT sensors can be obtained as 30.8/2.3 = 13.4 pg. Comparing the phases of the response curves in Fig. 5(a) and (b), it can be found that, for the same adsorption time the occurrence of deflagration of the CNT sensors shift backwards relative to those of the bare microcantilever, and for the same sensor, either with or without the CNTs, the deflagration also shift backwards with the increase in the adsorption time. The time lags of the CNT sensors indicate that the CNT networks increase the heat mass of the microcantilever, and the dependence of the deflagration points on the adsorption time implies that the adsorbed CNT vapors also slightly increase the heat mass. During heating, the adsorbed TNT molecules experience sublimation before reaching deflagration [39]. Sublimation decreases the quantity of the TNT molecules that have been already adsorbed on the CNTs, reducing the heat released from deflagration. Therefore, a heating rate that is rapid enough to shorten or even avoid sublimation and melting is of great benefit to achieve low LOD. In addition, rapid heating also reduces the heat dissipation of the sensor to the circumstance during heating, further improving the LOD. Fig. 6 shows three measured responses of both a CNT sensor and a pure microcantilever to TNT vapors with a 6 ms rapid heating

4.1. Experimental setup To validate the proposed CNT sensors and the thermalmechanical coupling transduction mechanism, TNT is adopted as a target explosive for detection test. Fig. 3(a) shows the CNT sensor packaged in a test chamber. A voltage-controlled current source is supplied to the heater to heat the CNT sensor. The microcantilever deflagration is measured with the integrated piezoresistor. Two sensors are connected in a Wheatstone bridge and operated in a differential mode. One shielded from surrounding environment serves as the reference and the other exposed to TNT vapor is used as the sensing device. The deflection is obtained by subtracting the reference from the sensing device. The TNT vapors with constant pressure are obtained by heating solid TNT powder at 60 ◦ C. Weight difference method using a tuning fork is employed to measure the mass of the TNT adsorbed on the bare microcantilever. By measuring the resonant frequency shift caused by TNT adsorption, the mass of the TNT molecules adsorbed on the tuning fork can be calculated. Fig. 4 shows the measured frequency changes and the corresponding mass from the tuning fork. It can be seen that the adsorbed mass follows an almost linear relationship with time. The TNT mass on the bare microcantilever is then calculated through dividing the TNT mass on the tuning fork by their surface ratio. It should be noted that the prerequisites for this approach are that the adsorbed TNT molecules are uniformly distributed on solid surfaces [40] and the tuning fork and the microcantilever have similar adsorption ability. Upon these assumptions, the mass of the TNT adsorbed on the microcantilever is calculated as 24 pg/min. During heating, the TNT vapors adsorbed on the CNT surfaces experience sublimation, melting, and deflagration, which inherently involve the characteristic endothermic or exothermic processes and show unique thermal features of TNT [39]. The sublimation and melting of TNT vapors are endothermic processes and causes additional upwards deflection of the sensing

4.2. TNT detection

W. Ruan et al. / Sensors and Actuators B 176 (2013) 141–148

sensor immediately after deflagration. Thus, the proposed CNT sensor has the ability of self-cleaning, and can be used for detection reversibly and immediately after deflagration, without the need of waiting or special treatment.

180

w/o CNT test1

Amplitude (mV)

150

w/o CNT test2

w/o CNT test3

120

145

w/o CNT no TNT 90

w/ CNT test1

5. Interfering compounds

w/ CNT test2 60

w/ CNT test3 w/ CNT no TNT

30 0 -30

0

1

2

3

4

5

6

Time (ms) Fig. 6. The voltage responses of the bare microcantilever and the CNT sensor with 6 ms heating rate and 5 min adsorption.

rate and 5 min adsorption. The average peak voltage and the SNR of the CNT sensor are 146.3 mV and 147.2, whereas those of the bare microcantilever are 20.8 mV and 20.7 mV. By setting the SNR to 3, the LOD of the bare microcantilever is 120 × 3/20.7 = 17.4 pg, which is about three times larger than the literature reported 6 pg [35,39]. Considering that the peak voltage of the CNT sensor is 7.1 times that of the bare microcantilever, the equivalent LOD of the CNT sensor with 6 ms fast heating rate is 17.4/7.1 = 2.4 pg. Although rapid heating significantly improves the LOD, the characteristic thermal features disappear; therefore it should be combined with slow heating rates to differentiate the explosives. Changing the heating rates from slow to fast improves the LOD of the bare microcantilever and the CNT sensor, respectively, from 30.8 pg to 17.4 pg and from 13.4 pg to 2.4 pg, corresponding to 1.8 times and 5.6 times improvements, respectively. This means that fast heating rates are more effective in improving LOD of the CNT sensors than the bare microcantilevers, i.e., the contribution of fast heating rates to the sensors with large surface areas is more pronounced than those with small surface areas. As most TNT molecules are adsorbed on the TNT network surfaces, the sublimation of TNT molecules on the CNT networks are more significant than that on the bare microcantilever, thus increasing the heating rates and shortening the heating duration achieves more remarkable benefits to the CNT sensors than to the bare microcantilevers. The prerequisites of fast heating are the high thermal conductivity of the CNTs and the low interface thermal resistance of the interface between the CNTs and the microcantilever. Direct in situ synthesis of CNTs in high temperatures enables the CNTs to be tightly rooted on the metal catalyst, so good phonon interaction can be achieved and the interface thermal resistance are much lower than that fabricated using post-transfer methods. The deflagration occurs at 1 ms for CNT sensors and about 0.7 ms for pure microcantilevers when heated using fast heating rates, which implies that the influences of the thermal resistance of the interface is insignificant. This is critical to fast heat transfer between the microcantilever and the CNTs. Normally the recovery times to the baseline of adsorption-based CNT chemical sensors are on the order of 10 min to several hours because of the slow desorption rates of the molecules that are adsorbed on CNT surfaces [4,10,47]. Due to the inherent high temperatures of micro-deflagration, the recovery time of the proposed CNT chemical sensor is significantly shortened to the extent that can be neglected. The deflagration burns the TNT vapors adsorbed on the CNT surfaces, and the associated high temperatures accelerate the escape of the impurities and the residues generated during TNT deflagration. Hence the CNT surfaces are clean and ready again for next detection after each deflagration. This is evidenced by the near zero differential voltage between the reference and the CNT

VOCs with high flammability including acetone, ethanol, and toluene were tested to evaluate the immunity to interferential materials. Saturated VOC vapors are injected and completely fill the sensor chamber for 5 min adsorption, and then a slow heating voltage is applied from 50 ms to 90 ms. Fig. 7(a) and (b) shows the responses to acetone, ethanol, and toluene vapors. It can be seen that the CNT sensor remarkably responses to the these VOCs and the peak voltages are comparable to the responses to TNT, as indicated by the two repeated measured curves Acetone 1 and Acetone 2 in Fig. 7(a). It should be noted that the characteristics of the response curves are quite different from those of TNT vapors in the curvatures of the descending and ascending parts. Specifically, the curves increase in a linear-like behavior but restore to room temperature much quickly than the recovery time of TNT deflagration. The extremely fast restoration prompts a conjecture that there is no heat released during acetone heating, and the linear-like responses could be induced by the changes in the thermal conductivity of the acetone air. To validate this conjecture, further experiments were carried out. After acetone adsorption, air is conducted to the test chamber to clean the acetone vapor in the sensor chamber, and then the sensor is heated to high temperature. The measured results are shown by the curve Acetone differential in Fig. 7(a). It can be seen that after removing the acetone from the chamber, the CNT sensors do not show distinct responses to the adsorbed acetone. This implies that the responses to acetone is not caused by acetone adsorption, but could be attributed to acetone induced changes in the thermal conductivity of circumstance. This result is consistent with literature report [41], in which palladium catalyst is needed for detection of combustible gases using microcalorimeters. To verify this mechanism, inert gases helium and argon are tested and the results are shown in Fig. 7(c). It can be seen that, although inert gases are stable during heating, significant responses still occur. The thermal conductivity, the convection coefficient, and the heat capacity of argon are lower than the air, whereas those of helium are just on the opposite. The inert helium is stable during heating, but its good thermal conductivity facilitates heat dissipation of the sensor to environment, causing a remarkable positive response of the sensor. Similarly, the thermal conductivity of argon is worse than the air, so the sensor temperature is higher and the response is on the opposite. Ethanol and toluene vapors are also tested and similar results are obtained. Upon this mechanism, the fast recovery time can be expected because there is no additional heat released for these compounds after heating. Therefore, the adsorption of aforementioned VOCs does not influence the CNT sensors. 6. Repeatability and reproducibility The repeatability is a serious concern for most chemical sensors. A test with adsorption times of 5 min, 7.5 min, 10 min, 12.5 min and 15 min is performed three times for the same sensor, and the peak voltages versus the adsorption times of the three test runs are shown in Fig. 8(a). The relative standard deviations of the three test runs at each adsorption time are calculated to represent the non-repeatability. It can be seen that the relative standard deviations range from 2.2% to 4.0%. This relatively good repeatability is due to the thermal-mechanical based operation mode. Specifically,

146

W. Ruan et al. / Sensors and Actuators B 176 (2013) 141–148

(a)

(b)

20

20 10

Amplitude (mV)

Amplitude (mV)

0 - 20 - 40

- 60

Acetone 1

0 -10

Acetone 2

-80

Ethanol

- 20

Toluene Ethanol differential

- 30

Toluene differential

Acetone differential - 40

-100

0

20

40

60

0

100

80

20

40

Time (ms)

(c)

60

80

100

Time (ms)

2000

He

Amplitude (mV)

1500

Ar

1000 500

0 - 500

- 1000 0

20

40

60

80

100

Time (ms) Fig. 7. The responses of the CNT sensor to interferents. (a) Acetone. (b) Ethanol and toluene. (c) Helium and Argon.

Test run 1 Test run 2 Test run 3 Relative deviation

40

4 35 3 30 2 25 20

4

6

8

50

5

10

12

14

40

Amplitude (mV)

Peak voltage (mV)

45

15

(b)

6

50

Relative deviation (%)

(a)

30

20

1

10

0 16

0

Adsorption time (min)

Device1 Device2 Device3 Device4

5

7.5

10

12.5

15

Time (min)

Fig. 8. (a) The repeatability of three test runs and the relative deviations. (b) The reproducibility of 4 different CNT sensors to TNT vapors at various adsorption times.

it can be attributed to the deflagration associated self-cleaning, which is of great advantages to adsorption-based chemical sensors in achieving fast recovery times and avoiding performance decay induced by contamination. Besides, the high-temperature stability of CNTs also ensures the repeatability and reliability even the deflagration temperature is as high as 250–300 ◦ C. Reproducibility is another concern for practical applications of CNT chemical sensors, and is characterized through testing four different sensors. Specifically, the first sensor is tested at the same temperature with adsorption times of 5 min, 7.5 min, 10 min, 12.5 min, and 15 min, and the same process is repeated

for other three sensors. Then the corresponding responses of all the sensors at each adsorption time are compared, and the largest relative difference of all the adsorption times is used to represent the reproducibility. Fig. 8(b) illustrates the voltage responses of four different CNT sensors to TNT adsorption times to evaluate the device-to-device reproducibility. The relative standard deviations of the four CNT sensors vary within 6–10% at the different adsorption times, which indicates a rather good reproducibility for CNT-based chemical gas sensors. This good device-to-device reproducibility is attributed to the good uniformity of the CNT films in situ synthesized on the microcantilever, which not only greatly

W. Ruan et al. / Sensors and Actuators B 176 (2013) 141–148

improves the adsorption ability of the sensor, but also statistically reduces the scatter ranges that are serious in individual CNTs or sparse CNT networks as a result of the difficulties in precise control over the process to replicate nanotubes with identical parameters. It should be noted that this 6–10% deviation includes all the nonuniformities arising from the CNT networks, the microcantilevers, the heater, and the piezoresistor. 7. Conclusions A CNT chemical sensor that consists of a microcantilever and an in situ synthesized CNT network and is operated in thermalmechanical coupling transduction mechanism has been developed for detection of trace explosive vapors. The CNTs in situ synthesized on the suspended microcantilever act as an efficient adsorbent to accelerate vapor adsorption. The large surface to volume ratio and the excellent thermal properties of CNTs, together with the extremely low thermal mass of the microcantilever, enable the realization of a high performance CNT chemical sensor. The CNT networks improve the adsorption ability as much as 7.1 times, and an equivalent LOD of 2.4 pg has been achieved for TNT detection with 6 ms heating rate. The repeatability and the reproducibility are better than 4% and 10%, respectively. The preliminary results demonstrate that the excellent physical and thermal properties of CNTs allow the realization of the CNT chemical sensors operated in thermal-mechanical coupling transduction mechanism for TNT detection. Acknowledgments This research is supported in part by NSFC under Grant 60871006, 863 Program under Grant 2007AA03Z304, and Tsinghua University Initiative Scientific Research Program 2009THZ01005. References [1] S. Marin, A. Merkoc¸i, Nanomaterials based electrochemical sensing applications for safety and security, Electroanalysis 24 (2012) 459–469. [2] J. Wang, S.B. Hocevar, B. Ogorevc, Carbon nanotube-modified glassy carbon electrode for adsorptive stripping voltammetric detection of ultratrace levels of 2,4,6-trinitrotoluene, Electrochemistry Communications 6 (2004) 176–179. [3] S. Hrapovic, E. Majid, Y. Liu, K. Male, J.H.T. Luong, Metallic nanoparticle-carbon nanotube composites for electrochemical determination of explosive nitroaromatic compounds, Analytical Chemistry 78 (2006) 5504–5512. [4] T. Zhang, S. Mubeen, N.V. Myung, M.A. Deshusses, Recent progress in carbon nanotube-based gas sensors, Nanotechnology 19 (2008) 332001. [5] B. Paolo, L. Pierre, P. Didier, Carbon nanotubes based transistors as gas sensors: state of the art and critical review, Sensors and Actuators B 140 (2009) 304–318. [6] M. Penza, R. Rossi, M. Alvisi, G. Cassano, E. Serra, Functional characterization of carbon nanotube networked films functionalized with tuned loading of Au nanoclusters for gas sensing applications, Sensors and Actuators B 140 (2009) 176–184. [7] P.-C. Chen, S. Sukcharoenchoke, K. Ryu, L.G. de Arco, A. Badmaev, C. Wang, C. Zhou, 2,4,6-Trinitrotoluene (TNT) chemical sensing based on aligned singlewalled carbon nanotubes and ZnO nanowires, Advanced Materials 22 (2010) 1900–1904. [8] W. Xue, T.H. Cui, A thin-film transistor based acetylcholine sensor using selfassembled carbon nanotubes and SiO2 nanoparticles, Sensors and Actuators B 134 (2008) 981–987. [9] M. Consales, S. Campopiano, A. Cutolo, M. Penza, P. Aversa, G. Cassano, M. Giordano, A. Cusano, Carbon nanotubes thin films fiber optic and acoustic VOCs sensors: performances analysis, Sensors and Actuators B 118 (2006) 232–242. [10] J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, M. Meyyappan, Carbon nanotube sensors for gas and organic vapor detection, Nano Letters 3 (2003) 929–933. [11] M. Terrones, Science and technology of the twenty-first century: synthesis, properties and applications of carbon nanotubes, Annual Review of Materials Research 33 (2003) 419–501. [12] B. Mahar, C. Laslau, R. Yip, Y. Sun, Development of carbon nanotube-based sensors—a review, IEEE Sensors Journal 7 (2007) 266–284. [13] E. Bekyarova, M.E. Itkis, N. Cabrera, B. Zhao, A. Yu, J. Gao, R.C. Haddon, Electronic properties of single-walled carbon nanotube networks, Journal of the American Chemical Society 127 (2005) 5990–5995. [14] B.Y. Lee, M.G. Sung, J. Lee, K.Y. Baik, Y.-K. Kwon, M.-S. Lee, S. Hong, Universal parameters for carbon nanotube network-based sensors: can nanotube sensors be reproducible, ACS Nano 5 (2011) 4373–4379.

147

[15] M.C. LeMieux, M. Roberts, Y.W. Jin, J.M. Kim, Z. Bao, Self-sorted, aligned nanotube networks for thin-film transistors, Science 321 (2008) 101–104. [16] W. Wongwiriyapan, S. Honda, H. Konishi, T. Mizuta, T. Ohmori, Y. Kishimoto, T. Ito, T. Maekawa, K. Suzuki, H. Ishikawa, T. Murakami, K. Kisoda, H. Harima, K. Oura, M. Katayama, Influence of the growth morphology of single-walled carbon nanotubes on gas sensing performance, Nanotechnology 17 (2006) 4424–4430. [17] S. Rajaputra, R. Mangu, P. Clore, D. Qian, R. Andrews, V.P. Singh, Multi-walled carbon nanotube arrays for gas sensing applications, Nanotechnology 19 (2008) 345502. [18] M. Penza, R. Rossi, M. Alvisi, G. Cassano, M.A. Signore, E. Serra, R. Giorgi, Ptand Pd-nanoclusters functionalized carbon nanotubes networked films for subppm gas sensors, Sensors and Actuators B 135 (2008) 289–297. [19] Y. Kim, S. Lee, H.H. Choi, J.-S. Noh, W. Lee, Detection of a nerve agent simulant using single-walled carbon nanotube networks: dimethyl-methylphosphonate, Nanotechnology 21 (2010) 495501. [20] M. Lee, J. Lee, T.H. Kim, H. Lee, B.Y. Lee, J. Park, Y.M. Jhon, M.-J. Seong, S. Hong, 100 nm scale low-noise sensors based on aligned carbon nanotube networks: overcoming the fundamental limitation of network-based sensors, Nanotechnology 21 (2010) 055504. [21] J. Kong, M. Chapline, H.J. Dai, Functionalized carbon nanotubes for molecular hydrogen sensors, Advanced Materials 13 (2001) 1384–1386. [22] T.H. Kim, B.Y. Lee, J. Jaworski, K. Yokoyama, W.-J. Chung, E. Wang, S. Hong, A. Majumdar, S.-W. Lee, Selective and sensitive TNT sensors using biomimetic polydiacetylene-coated CNT-FETs, ACS Nano 5 (2011) 2824–2830. [23] Z. Kuang, S.N. Kim, W.J. Crookes-Goodson, B.L. Farmer, R.R. Naik, Biomimetic chemosensor: designing peptide recognition elements for surface functionalization of carbon nanotube field effect transistors, ACS Nano 4 (2010) 452–458. [24] A.R. Krause, C. Van Neste, L. Senesac, T. Thundat, E. Finot, Trace explosive detection using photothermal deflection spectroscopy, Journal of Applied Physics 103 (2008) 094906. [25] L.A. Pinnaduwage, T. Thundat, J.E. Hawk, D.L. Hedden, P.F. Britt, E.J. Houser, S. Stepnowski, R.A. McGill, D. Bubb, Detection of 2,4-dinitrotoluene using microcantilever sensors, Sensors and Actuators B 99 (2004) 223–229. [26] L.A. Pinnaduwage, V. Boiadjiev, J.E. Hawk, T. Thundat, Sensitive detection of plastic explosives with self-assembled monolayer-coated microcantilevers, Applied Physics Letters 83 (2003) 1481–1483. [27] J.H. Lee, R.T.J. Houk, A. Robinson, J.A. Greathouse, S.M. Thornberg, M.D. Allendorf, P.J. Hesketh, Investigation of microcantilever array with ordered nanoporous coatings for selective chemical detection, SPIE 7679 (2010) 767927. [28] M.A. Urbiztondo, I. Pellejero, M. Villarroya, J. Sese, M.P. Pina, I. Dufour, J. Santamaria, Zeolite-modified cantilevers for the sensing of nitrotoluene vapors, Sensors and Actuators B 137 (2009) 608–616. [29] I. Pellejero, J. Agusti, M.A. Urbiztondo, J. Sesea,d, M.P. Pina, J. Santamaria, G. Abadal, Nanoporous silicalite-only cantilevers as micromechanical sensors: fabrication, resonance response and VOCs sensing performance, Sensors and Actuators B 168 (2012) 74–82. [30] P. Xu, X. Li, H. Yu, M. Liu, J. Li, Self-assembly and sensing-group graft of premodified CNTs on resonant micro-cantilevers for specific detection of volatile organic compound vapors, Journal of Micromechanics and Microengineering 20 (2010) 115003. [31] N.W. Piekiel, Richard E. Cavicchib, M.R. Zachariaha, Rapid-heating of energetic materials using a micro-differential, scanning calorimeter, Thermochimica Acta 521 (2011) 125–129. [32] V.H. Carreto-Vazquez, A.K. Wojcik, Y.-S. Liu, D.B. Bukur, M.S. Mannan, Miniaturized calorimeter for thermal screening of energetic materials, Microelectronics Journal 41 (2010) 874–881. [33] L. Wang, B. Wang, Q. Lin, Demonstration of MEMS-based differential scanning calorimetry for determining thermodynamic properties of biomolecules, Sensors and Actuators B 134 (2008) 953–958. [34] C.-C. Chen, W.W. Bannister, A. Viswanathan, Micro-thermal analysis for airport detection of energetic materials, Proceedings. 27th Conference of the North American Thermal Analysis Society 67–72 (1999). [35] L.A. Pinnaduwage, A. Gehl, D.L. Hedden, G. Muralidharan, T. Thundat, R.T. Lareau, T. Sulchek, L. Manning, B. Rogers, M. Jones, J.D. Adams, A microsensor for trinitrotoluene vapour, Nature 425 (2003) 474. [36] L.A. Pinnaduwage, A. Wig, D.L. Hedden, A. Gehl, D. Yi, T. Thundat, R.T. Lareau, Detection of trinitrotoluene via deflagration on a microcantilever, Journal of Applied Physics 95 (2004) 5871–5875. [37] L.A. Pinnaduwage, H.F. Ji, T. Thundat, Moore’s law in homeland defense: an integrated sensor platform based on silicon microcantilevers, IEEE Sensors Journal 5 (2005) 775–785. [38] D. Yi, A. Greve, J.H. Hales, L.R. Senesac, Z.J. Davis, D.M. Nicholson, A. Boisen, T. Thundat, Detection of adsorbed explosive molecules using thermal response of suspended microfabricated bridges, Applied Physics Letters 93 (2008) 154102. [39] L.R. Senesac, D. Yi, A. Greve, J.H. Hales, Z.J. Davis, D.M. Nicholson, A. Boisen, T. Thundat, Micro-differential thermal analysis detection of adsorbed explosive molecules using microfabricated bridges, Review of Scientific Instruments 80 (2009) 035102. [40] L.A. Pinnaduwage, D. Yi, F. Tian, T. Thundat, R.T. Lareau, Adsorption of trinitroluence on uncoated silicon microcantilever surfaces, Langmuir 20 (2004) 2690–2694. [41] R.E. Cavicchi, G.E. Poirier, N.H. Tea, M. Afridi, D. Berning, A. Hefner, J. Suehle, M. Gaitan, S. Semancik, C. Montgomery, Micro-differential scanning calorimeter for combustible gas sensing, Sensors and Actuators B 97 (2004) 22–30.

148

W. Ruan et al. / Sensors and Actuators B 176 (2013) 141–148

[42] W. Ruan, Z. Wang, Y. Li, L. Liu, In-situ heat capacity measurement of carbon nanotubes using suspended microstructure-based microcalorimetry, IEEE Transactions on Nanotechnology 11 (2012) 367–373. [43] J.J. Zhao, J.P. Lu, J. Han, C.K. Yang, Noncovalent functionalization of carbon nanotubes by aromatic organic molecules, Applied Physics Letters 82 (2003) 3746–3748. [44] Y. Zhou, Z. Wang, Q. Zhang, W. Ruan, L. Liu, A front-side released single crystalline silicon piezoresistive microcantilever sensor, IEEE Sensors Journal 9 (2009) 246–254. [45] W. Ruan, Z. Wang, L. Li, L. Li, K. Jiang, Synthesis of carbon nanotubes on suspending microstructures by rapid local laser heating, IEEE Sensors Journal 11 (2011) 3424–3425. [46] L.A. Currie, Nomenclature in evaluation of analytical methods including detection and quantification capabilities, Pure and Applied Chemistry 67 (1995) 1699–1723. [47] Y. Chang, J.S. Oh, S.H. Yoo, H.H. Choi, K.-H. Yoo, Electrically refreshable carbonnanotube-based gas sensors, Nanotechnology 18 (2007) 435504.

Biographies Wenzhou Ruan received the B.E. degree in electronics science and technology from Xi’an Jiaotong University in 2007. He is currently working toward the Ph.D. degree in the Institute of Microelectronics, Tsinghua University. His research interests are carbon nanotube microsensors. Yuanchao Li received his B.E. degree in Computer Science and Technology from Shandong University, Jinan, China, in 2010. He is currently pursuing the M.S. degree

in the Institute of Microelectronics, Tsinghua University. His research interests are carbon nanotubes and their application in sensors. Litan Liu received the B.S. degree in electronic engineering from Tsinghua University, Beijing, China, in 1970. He is currently a Full Professor at the Institute of Microelectronics, Tsinghua University. He has authored and coauthored more than 150 peer-reviewed journal and conference papers, and supervised 20 Ph.D. students. His research interests include semiconductor devices, integrated sensors, and MEMS. Kaili Jiang was born in China in 1972. He received the B.S., M.S. and Ph.D. degrees in Physics in 1995, 1998, and 2006, all from Tsinghua University, Beijing, China. Currently, he is a Full Professor at Department of Physics, Tsinghua University. His research interests include growth mechanisms, controlled synthesis, physical properties, and applications of carbon nanotubes. He has published more than 60 peer-reviewed journal and conference papers. Zheyao Wang was born in China in 1972. He received the B.S. degree in mechanical engineering in 1995, and the Ph.D. degree in mechatronics in 2000, both from Tsinghua University, Beijing, China. From 2000 to 2002, he was a Postdoctoral Research Fellow at the Institute of Microelectronics, Tsinghua University, where he worked on silicon micromachining for microsensor applications. In 2002, he joined DIMES, Delft University of Technology, Delft, The Netherlands, as a postdoctoral researcher, and worked on silicon micromachining for 3-D packaging. Currently, he is a Full Professor at Tsinghua University. His research interests include microsensors, micromachining, and MEMS. He has published more than 60 peer-reviewed journal and conference papers, and serves as a technical program committee member for IEEE Sensors Conference since 2008.