Japanese Journal of Applied Physics Vol. 47, No. 9, 2008, pp. 7440–7443 #2008 The Japan Society of Applied Physics
Highly Sensitive, Room-Temperature Gas Sensors Prepared from Cellulose Derivative Assisted Dispersions of Single-Wall Carbon Nanotubes Annamalai K ARTHIGEYAN, Nobutsugu M INAMI, and Konstantin I AKOUBOVSKII Nanotechnology Research Institute, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Ibaraki 305-8565, Japan (Received March 19, 2008; accepted May 24, 2008; published online September 12, 2008)
We have developed highly sensitive gas sensors based on single-wall carbon nanotube (SWNT) networks prepared from aqueous hydroxypropylcellulose-assisted dispersions. Gas responses were monitored at room temperature for different concentrations of NO2 . The stable baseline and its recovery (after NO2 exposure) were achieved by ultraviolet (UV)-induced photodesorption, for which a compact and low-power UV light emitting diode can be used as a light source. The sensors are capable of detecting 25 ppb or lower concentrations of NO2 and 5 ppm ammonia, and show almost no baseline drift after multiple NO2 exposures. The simple and low-cost fabrication process, reproducible performance and room-temperature operation bode well for industrial mass production and broad uses. [DOI: 10.1143/JJAP.47.7440] KEYWORDS: carbon nanotube, gas sensor, cellulose, hydroxypropylcellulose, dispersion, NO2 , ammonia, photodesorption, UV light emitting diode
1.
Introduction
2.
The exceptional properties of single-wall carbon nanotubes (SWNTs) as gas sensing materials1) have generated growing interest among gas sensor researchers. Fast response, high sensitivity, room-temperature operation and small size make this material very attractive for the application of monitoring toxic air pollutants endangering environment and human health. It is expected that they surpass the existing metal oxide based sensors2) that suffer from high operation temperature (above 200 C) and poor long-term stability. Sensors with individual SWNTs1) or its ensembles3–8) (networks) were reported as sensing elements for detecting various gases, but they suffer from major drawbacks involving complex and expensive fabrication processes, irreproducibility and low yield. They rely on some of sophisticated techniques such as electron beam evaporation, chemical vapor deposition (CVD), reactive ion etching (RIE), and e-beam/optical lithography for the patterning of catalytic metals and SWNT growth. Hence the successful industrial application of this exotic material will depend on minimizing such complex and costly techniques through the realization of an innovative and cost-effective fabrication approach. In the present work, we demonstrate SWNT-network based gas sensors fabricated on quartz substrates by a simple, mass production compatible method, capable of sensing NO2 below 25 ppb at room temperature. The networks were prepared from cellulose derivative assisted dispersions of SWNT in a very simple and cost-effective way, a rational extension of our previous work on SWNT thin film fabrication.9) Our method eliminates the tedious procedures of growing, locating, assembling, and aligning individual SWNTs or their networks for sensor manufacturing. Moreover, the use of a cellulose derivative — a cheap, safe, waterprocessable, and environmentally benign dispersant — makes the present approach industrially feasible.
E-mail address:
[email protected]
Experimental Procedure
For the formation of the sensing layer, 18 mg of HiPco SWNT (Carbon Nanotechnologies) was mixed with 15 mL of 5 wt % aqueous solution of hydroxypropylcellulose (HPC; Klucel Type L) using a bath sonicator. The diameter of HiPco SWNT distributes in the range 0.8 –1.1 nm. The thus prepared dispersion was further sonicated for 15 min using a horn sonifier (Branson 450) and then centrifuged at 45,000 rpm (150;000g) for 5 h to remove impurities and large bundles of nanotubes using an ultracentrifuge (Hitachi CS100GXL). After centrifugation, the upper 80% of the supernatant was collected as the final dispersion, which was then spin cast onto quartz substrates and dried at room temperature overnight. The spin coated substrates were subjected to a vacuum heat treatment at different temperatures between 300 and 800 C for 1 h. Atomic force microscopy (AFM) analysis was carried out to monitor the removal of HPC and the formation of exposed nanotube networks. Interdigitated gold electrodes (IDEs) with 100 mm width fingers and gaps were thermally evaporated in vacuum onto the nanotube networks through a metal mask. For gas sensing, a synthetic air generator was used for continuous supply of purified air. Controlled concentrations of NO2 gas were generated with a permeator (GASTEC Model PD-1B) by dilution with the purified air as a carrier gas. Further, the second gas dilution was carried out to obtain different concentrations of NO2 using a combination of mass flow controllers. The current through the sensors, at a constant applied voltage of 1 V, was monitored using a computer-controlled Keithley 2602 source meter. The sensor experiments were performed in a stainless-steel test chamber equipped with an ultraviolet (UV) transparent optical window. The measurements were conducted in a constant flow rate (200 mL/min) of the total gas. To recover the baseline after gas exposure, we apply UV induced photodesorption,10) for which the sensor surface was illuminated with UV light from a deuterium lamp through a focusing lens (wavelength range: 200 – 400 nm, power density: 2 mW/cm2 at the sensor surface). We also used a UV light
7440
Jpn. J. Appl. Phys., Vol. 47, No. 9 (2008)
A. KARTHIGEYAN et al.
1.25x10
-6
1.00x10
-6
7.50x10
-7
5.00x10
-7
2.50x10
-7
G (A/V)
(a)
NO2 OFF UV ON
base line set by UV
100 ppb
50 ppb
(a)
0.00
(b)
0
30
60
90
120
1.0x10
-5
8.0x10
-6
6.0x10
-6
4.0x10
-6
2.0x10
-6
(b)
emitting diode (UV-LED) as a light source for photodesorption, whose emission wavelength was 280 nm and power density 0.4 mW/cm2 at the sensor surface. Note that the compactness and low power consumption of the UVLED can be crucial for practical applications of gas sensors. All the gas sensing experiments were carried out in dry air at room temperature.
180
NO2 OFF
G (A/V)
UV ON base line set by UV
0.0 3 ppm
Results and Discussion
An AFM image of a spin coated SWNT/HPC film treated at 300 C in vacuum shows a sponge-like structure, indicating that the SWNT network is still embedded in HPC [Fig. 1(a)]. However, a treatment at 350 C yielded a welldispersed and clean SWNT network as evident from Fig. 1(b), confirming the removal of HPC. This temperature was chosen because it corresponds to the maximum decomposition rate of HPC as observed by thermo-gravimetry. Note that further increase in the treatment temperature up to 400 C resulted in only a slight change in AFM images or in the sensor behavior, above which temperature the response of the sensor deteriorated probably due to the degradation of SWNT. We stress that the successful formation of the SWNT nano-scale network structure best suited for gas sensor applications is made possible by the proper choice of the dispersant, HPC, which combines SWNT-dispersing ability with excellent burning behavior, namely, a very small amount of residue. While, among various cellulose derivatives, carboxymethylcellulose is known to possess excellent SWNT-dispersing ability,9) it is unsuitable for the present approach because of a large amount of residue after burning. Note that there still remain a number of small particles [Fig. 1(b)], which persisted at even higher temperatures. At present, their influence on the gas sensing abilities is unknown, which should be a subject of further studies. Electrical measurements on IDE-deposited SWNT networks showed non-linear and symmetrical current–voltage (I–V) characteristics probably deriving from the contact between the electrodes and SWNTs and temperaturedependent resistance, indicating the semiconductor-like behavior of SWNTs. The effect of UV irradiation on the conductance of the SWNT networks was preliminarily checked using a deuterium lamp. Under UV irradiation, the conductance of an air-stabilized sample decreased and leveled off in 20 min, attributable to photodesorption of
150
Time (min)
Fig. 1. (Color online) AFM images of spin coated SWNT-dispersed films after heat treatments in vacuum at (a) 300 and (b) 350 C.
3.
1000 ppb NO2 ON UV OFF
500 ppb
300
5 ppm
330
360
15 ppm
390
30 ppm
420
50 ppm NO2 ON UV OFF
450
480
Time (min)
Fig. 2. Room-temperature response (conductance change) of the SWNT networks to NO2 exposure in the range of (a) 50 to 1000 ppb and (b) 3 to 50 ppm. For photodesorption, UV light was turned on 5 min after the termination of each gas exposure and then turned off immediately before the next exposure. The same sequence as indicated for 1000 ppb (a) and 50 ppm (b) is repeated for all the exposures.
oxygen, a p-dopant for SWNT. After turning off UV light, the sample was exposed to NO2 (here a 100 ppm concentration was used as a trial), a strong p-dopant; we observed a considerable increase in conductance, after which photodesorption tests were carried out. Different wavelengths from the deuterium lamp were used by successively changing cut filters. We found that wavelengths shorter than 300 nm were the most effective in NO2 photodesorption, in good agreement with an earlier report.10) We examined the responses of the sensor to different concentrations of NO2 . The initial baseline (before introducing a test gas) and recovery to the baseline (after removal of a test gas) was achieved by UV photodesorption using unfiltered light from the deuterium lamp. Figure 2 displays the dynamic response of the SWNT networks (annealed at 350 C) to different concentrations of NO2 ranging from 50 ppb to 50 ppm. UV illumination was stopped immediately before NO2 gas introduction. After 5 min gas exposure, the samples were purged in purified air for 5 min. Then UV illumination was re-started to accelerate gas desorption from the surface that lasted for 30 min. Upon restarting UV illumination, the conductance returned to the baseline within 2 – 5 min. Defining the sensor response as S ¼ ðG G0 Þ=G0 , where G0 is the conductance in air before exposing to NO2 and G the maximum conductance after 5 min exposure, we obtain sensitivity response of around S ¼ 0:5 for 50 ppb NO2 . Sensor response as a function of NO2 concentration
7441
A. KARTHIGEYAN et al.
24 22 20 18 16 14 12 10 8 6 4 2 0
G (A/V)
Sensor Response S (ΔG/G0)
Jpn. J. Appl. Phys., Vol. 47, No. 9 (2008)
5.0x10
-4
4.0x10
-4
3.0x10
-4
UV-LED on UV-LED on UV-LED on UV-LED on
2.0x10
-4
1.0x10
-4
25 ppb NO2
0
10
20 30 NO2 Conc. (ppm)
40
50 ppb 100 ppb UV-LED off UV-LED off
200 ppb UV-LED off
0.0
50
0
30
60
90
120
150
180
Time (min)
Fig. 3. Sensor response (S) of the SWNT networks as a function of the NO2 concentration at room temperature.
plotted in Fig. 3 shows a non-linear coverage of NO2 molecules at low concentrations and saturation at higher concentrations. We checked that different sensors prepared by the same procedures showed reasonably good reproducibility in these sensitivity curves (within 50% variance). The gas sensor response can be explained as the following. The strong oxidizing reaction of NO2 on the sensitive layer produces large increase in conductance due to an increased hole concentration in the nanotubes, indicating p-type semiconductor behavior of the networks. The increased hole concentration is attributed to the expected charge transfer (CT) from SWNT networks to chemisorbed NO2 .11) When considering the observed conductance increase by CT, we should note that the SWNT networks consist of a mixture of metallic and semiconducting nanotubes and so do any conducting pathways connecting the both electrodes (gap: 100 mm). They are likely to form different combinations of junctions among them. Work function differences form junction barriers across tube–tube and tube–electrode contacts.12) It is known that metal–semiconductor junctions have higher resistances than metal–metal or semiconductor– semiconductor junctions due to the presence of Schottky barriers,13) and such junctions could be strongly modulated by CT. For these reasons, we believe that NO2 -induced change in conductance mostly originates from dopinginduced barrier changes rather than from variations in the carrier concentration. To compare with previous works, a sensor made of a few individual SWNTs bridging the source–drain gap (5 mm) of an FET structure produced a conductance change of 80% upon 10 min exposure to 100 ppb NO2 in Ar.3) While the sensitivity is comparable to our sensors, the fabrication process involved multiple, sophisticated and expensive techniques such as lithographical electrode patterning, RIE, etc., and thus to the detriment of industrial application. Some of CVD-grown SWNT thin films showed NO2 sensitivities comparable to ours, but only after electrical or thermal pre-treatments.4,5) In addition, a heating/cooling cycle or measurements at high temperature were required for satisfactory sensor performance. Solution-processed SWNT networks were also used to prepare NO2 sensors, but the measurements were performed mostly in the ppm range.6,7) We believe that heavily bundled SWNTs used in some of
Fig. 4. Demonstration of the effective use of a UV-LED (280 nm) for photodesorption of NO2 adsorbed from a concentration ranging from 25 to 200 ppb.
these studies, instead of well-dispersed ones, should have limited the detection sensitivity and the response time. Compared with these works, the present approach is distinguished by combined advantages of high sensitivity and simple and low-cost fabrication. This owes much to the use of HPC as a dispersant as well as to our accumulated experience in polymer-assisted dispersion and thin film fabrication of SWNTs.9) It is noteworthy that the original baseline is reached even after multiple NO2 exposures of the sensing film. This indicates no poisoning effect on the sensitive surface, probably reflecting the chemical stability of SWNT surface and a reversible oxidizing reaction of chemisorbed NO2 under UV illumination. Moreover, room-temperature operation of our sensor is advantageous for long-term stability, considering the reported signal drift and surface poisoning in metal oxide sensors due to grain growth and the nature of gaseous reaction during high temperature operation.14) The use of the deuterium lamp for photodesorption can be an impediment to the practical application of the present gas sensors. To address this issue, we have introduced the compact UV-LED as an alternative light source. The results are shown in Fig. 4, where sensor responses were measured for 25 to 200 ppb NO2 . Although somewhat slower than that for the deuterium lamp, probably due to the lower power, the photodesorption effect is sufficient for the baseline recovery purpose. We believe that the small size (8 mm 7 mmH) and low power consumption of the UV-LED show great advantage for practical, battery-powered, SWNT gas sensor systems. We also performed detection of an electron-donating gas. A 30% conductance decrease was observed upon exposure to 5 ppm NH3 at room temperature (Fig. 5), a rather high NH3 detection sensitivity for a carbon nanotube gas sensor. The conductance decrease induced by an electron donor again confirms the p-type nature of the SWNT networks. We are now studying the suitability of the sensor for sub-ppb NO2 detection and selectivity against different gases by appropriate chemical modifications. In conclusion, we have demonstrated a highly sensitive NO2 sensor consisting of SWNT networks prepared by HPC-assisted dispersion and thin film fabrication. The
7442
Jpn. J. Appl. Phys., Vol. 47, No. 9 (2008)
2.0x10
-4
1.8x10
-4
1.6x10
-4
1.4x10
-4
1.2x10
-4
1.0x10
-4
8.0x10
-5
A. KARTHIGEYAN et al.
Acknowledgment
G (A/V)
5 ppm NH3
320
The authors thank Yeji Kim for her contribution in an early stage of the work.
NH3 OFF
330
340
350
360
370
Time (min)
Fig. 5. Response of the SWNT networks to 5 ppm NH3 in dry air at room temperature.
sensor is capable of detecting NO2 concentrations of 25 ppb or less at room temperature, showing almost no baseline drift after multiple NO2 exposures. The simple and low-cost fabrication and reproducible results bode well for the mass production and practical uses. The demonstration that a compact and low-power UV-LED can be used as a photodesorption light source should enhance this expectation. Furthermore, this approach can easily be implemented on micro and macro electronic platforms for the fabrication of various electrical devices based on SWNT networks for different applications.
1) J. Kong, N. R. Franklin, C. Zhou, M. G. Chapline, S. Peng, K. Cho, and H. Dai: Science 287 (2000) 622. 2) I. Simon, N. Baˆrsan, M. Bauer, and U. Weimar: Sens. Actuators B 73 (2001) 1. 3) P. Qi, O. Vermesh, M. Grecu, A. Javey, Q. Wang, H. Dai, S. Peng, and K. J. Cho: Nano Lett. 3 (2003) 347. 4) W. Wongwiriyapan, S. I. Honda, H. Konishi, T. Mizuta, T. Ikuno, T. Ito, T. Maekawa, K. Suzuki, H. Ishikawa, K. Oura, and M. Katayama: Jpn. J. Appl. Phys. 44 (2005) L482. 5) L. Valentini, I. Armentano, J. M. Kenny, C. Cantalini, L. Lozzi, and S. Santucci: Appl. Phys. Lett. 82 (2003) 961. 6) J. Suehiro, G. Zhou, H. Imakiire, W. Ding, and M. Hara: Sens. Actuators B 108 (2005) 398. 7) J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, and M. Meyyappan: Nano Lett. 3 (2003) 929. 8) J. P. Novak, E. S. Snow, E. J. Houser, D. Park, J. L. Stepnowski, and R. A. McGill: Appl. Phys. Lett. 83 (2003) 4026. 9) N. Minami, Y. Kim, K. Miyashita, S. Kazaoui, and B. Nalini: Appl. Phys. Lett. 88 (2006) 093123. 10) R. J. Chen, N. R. Franklin, J. Kong, J. Cao, T. W. Tombler, Y. Zhang, and H. Dai: Appl. Phys. Lett. 79 (2001) 2258. 11) H. Dai: Acc. Chem. Res. 35 (2002) 1035. 12) J. Zhao, J. Han, and J. P. Lu: Phys. Rev. B 65 (2002) 193401. 13) M. S. Fuhrer, J. Nygard, L. Shih, M. Forero, Y. G. Yoon, M. S. C. Mazzoni, H. J. Choi, J. Ihm, S. G. Louie, A. Zettl, and P. L. McEuen: Science 288 (2000) 494. 14) B. Ruhland, Th. Becker, and G. Mu¨ller: Sens. Actuators B 50 (1998) 85.
7443
