Dynamic Article LinksC<Journal ofMaterials Chemistry

Cite this: J. Mater. Chem., 2011, 21, 4502

www.rsc.org/materials PAPER

Dow

nloa

ded

by S

ungk

yunk

wan

Uni

vers

ity o

n 11

Mar

ch 2

011

Publ

ishe

d on

07

Febr

uary

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0JM

0322

7AView Online

Humidity-assisted selective reactivity between NO2 and SO2 gas on carbonnanotubes

Fei Yao,a Dinh Loc Duong,a Seong Chu Lim,a Seung Bum Yang,b Ha Ryong Hwang,b Woo Jong Yu,a

Il Ha Lee,a Fethullah G€unesxa and Young Hee Lee*a

Received 26th September 2010, Accepted 26th November 2010

DOI: 10.1039/c0jm03227a

In spite of the technical importance of detecting environmental SOx and NOx gases, a selective

detection has not been realized because of their similar chemical properties. In this report, adsorption

and desorption of SO2 and NO2 gas on carbon nanotubes are investigated in terms of different

humidity levels at room temperature. A random-network single walled carbon nanotube (SWCNT)

resistor is constructed by a dip-pen method using a SWCNT/dichloroethane (DCE) solution. In the

case of SO2 gas adsorption, the resistance increases at high humidity level (92%) and shows no obvious

change at low humidity levels. On the other hand, in the case of NO2 gas adsorption, the resistance

always decreases independent of moisture levels. Our density functional theory (DFT) calculations

show that this selective behavior originates from cooperative charge compensation between the SO2–

nH2O complex and the p-type CNT resistor. The change of response time and recovery time with

different moisture levels is further investigated. This humidity-assisted gas reaction provides a simple

route to detect these two gases selectively.

Introduction

Accurate detection of sulfuric oxides and nitric oxides is an

important issue since both chemicals adversely affect human

health and the environment. Both of them exist as reactive gases

that primarily originate from human activities such as coal

combustion and petroleum refining, leading to the formation of

ground-level smog and acid rain. Detection, particularly of NO2

gas, with high sensitivity has been realized.1–8 Nevertheless,

commercialization is still limited due to the poor selectivity from

SO2 gas. To our knowledge, no single device with a simple

chemiresistor structure to selectively detect these two gases has

been realized yet.

A typical metal oxide for gas sensing with a reasonable

sensitivity usually needs to be operated at high temperature.

Nanomaterials such as carbon nanotubes (CNTs) which have

a large surface area to volume ratio can be a good candidate to

operate at room temperature with high sensitivity and also

provide less power consumption with size minimization.

Recently metal/metal oxide-decorated CNTs have been intro-

duced to modify CNT surfaces and thus control the strength of

the interaction with the gases.4–6 Polyethyleneimine (PEI) has

been suggested to decorate CNT surfaces where amine groups

aDepartment of Energy Science, BK21 Physics Division, SungkyunkwanAdvanced Institute of Nanotechnology, Sungkyunkwan University,Suwon, 440-746, Korea. E-mail: leeyoung@skku.edu; Fax: (+) 82-31-290-5954bR&D Division, Wise Control Incorporation, 199 Sanggai-dong, Giheung-gu, Yongin-si, Gyeonggi-do, 446-905, Korea

4502 | J. Mater. Chem., 2011, 21, 4502–4508

become an electron donor to convert a CNT channel to an n-

type, and thus enhance charge donation from CNTs to NO2.7

The related works are summarized in Table 1. On the other hand,

SO2 gas also behaves as an electron acceptor which is quite

similar to NO2 gas. Therefore, distinguishing these two gases has

been a great challenge.

The purpose of this paper is to design a sensor with both high

sensitivity and selectivity to NO2 and SO2 gas, which can be

operated under ambient pressure at room temperature. In this

work, the random network single walled (SW) CNTs resistor was

constructed by a dip-pen method using dichloroethane (DCE)-

dispersed surfactant-free SWCNTs. Selective detection of these

two gases was simply achieved by controlling the humidity level.

At low humidity levels, NO2 gas provided high sensitivity, while

SO2 gas showed negligible sensitivity. At a high humidity level

(92%), however, the channel resistance decreased with NO2 gas

but increased with SO2. This selective behavior, which was

explained by both experiments and density functional theory

(DFT) calculations, was attributed to the complicated electron

donation process between the SO2/H2O complex and the p-type

CNT resistor.

Experimental

Carbon nanotubes dispersion and device fabrication

The HiPCO SWCNTs (purity: �85 wt%) were purchased from

CNI company. The SWCNTs of 1 mg were dissolved in a 30 mL

DCE solution and sonicated for 12 h in a bath type sonicator

This journal is ª The Royal Society of Chemistry 2011

Table 1 Summary of previous results

Analyst Material Configuration/Method Sensing Mechanism Detect Limit Response/Recovery References

NO2 Bare CNTs ChemiFET NO2 is electron acceptor 2 ppm <600 s 1Chemiresistor NO2 is electron acceptor 10 ppb <120 s 2Chemicapacitor NO2 is electron acceptor 20 ppm / 3

Metal-decorated(Au/Pt) CNTs

Chemiresistor NO2 is electron acceptor 100 ppb <600 s (150 �C) 4

Metal oxidedecorated CNTs

ChemiFET NO2 is electron acceptor 25 ppm/500 ppb 4.5 min/3.5 min 5,6

Polymer-coated CNTs ChemiFET NO2 is electron acceptor 100 ppt <1000 s 7SWCNTs Spectroscopic analysis NO2 is electron acceptor 10ppb (150 K) 4 s 8,9Theory calculation / NO2 is electron acceptor / / 10–13

SO2 SWCNTs Spectroscopic analysis SO2 is electron acceptor 10 ppb (150 K) 4 s 8,9SWCNTs Chemiresistor SO2 is electron acceptor / <60 min 14MWCNTs Chemiresistor SO2 is electron donor 10 ppm <1200 s 15Other materials Amperometric sensor / 0.6 ppm / 16–19

H2O Theory calculation / H2O is electron donor;H2O is electron acceptor(graphene); no chargetransfer;

/ / 10,11,20–23

MWCNTs rope Chemiresistor / / �200 s 24MWCNTs film Chemicapacitor / 12–75.8% 20 s/10 s 25SWCNTs mat Chemiresistor/

ChemiFETH2O is electron donor

(type conversion)35–70% / 23,26

CNT/SiO2 composite Chemicapacitor H2O is electron donor 0–100% 2–3 min/h 27

Dow

nloa

ded

by S

ungk

yunk

wan

Uni

vers

ity o

n 11

Mar

ch 2

011

Publ

ishe

d on

07

Febr

uary

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0JM

0322

7AView Online

(Power Sonic 505). To remove large-size bundles, the

SWCNT solution was further centrifuged for 20 min at 21,380 g

(15 000 rpm). The supernatant of the SWCNT solution was

decanted and used for a dip pen method to fabricate the SWCNT

sensor. The next step is to construct electrodes for a sensor. An

interdigitated electrode of Pt (3000 �A) with an area of 200 � 200

mm2 was fabricated by using photolithography on a SiO2

(100 nm)/Si substrate (Fig. 1a). A gap distance of 2.5 mm was

maintained between the two electrodes. The SWCNT solution

was filled into a home-made dip pen machine with a contact area

of 3.14 mm2. The number of dip pen contacts and the CNT

density were optimized to give the best performance of the

sensor. Once the SWCNTs were transferred onto the designed

Fig. 1 (a) Schematic of an SWCNT sensor: CNTs on Pt electrodes on

SiO2/Si substrate, (b) comb-type interdigitated electrode, (c) SWCNTs

network between two electrodes, and (d) schematic of experimental setup

with measurement system.

This journal is ª The Royal Society of Chemistry 2011

electrode, the sample was heat-treated at 400 �C under Ar

ambient for 30 min to stabilize the final microstructure, improve

contact, and remove pre-adsorbed molecules.2,3

Measurement setup

NO2 (SO2) gas of 20 ppm was diluted to 2 ppm by mixing 450

SCCM of dry air and 50 SCCM target gas through an automatic

mass flow controller via a data acquisition card from National

Instrument, as shown in Fig. 1d. To control the humidity level,

air was flowed into a bubbler, where the humidity level was

detected by a humidity detector (Rotronic Hygroflex HF3x), or

directly into the chamber. The chamber whose length is 10 cm

and the quartz diameter is 2.5 cm, was maintained under ambient

pressure at room temperature. A DC bias of 0.1 V (Digital

Electronics Co. LTD 1083468) was applied to the electrode

through an external resistor to protect the SWCNT channel.

Current change was measured using a Keithley 2000 multimeter.

All of the data were recorded automatically through a general

purpose interface bus by a Labview interface.

Calculational method

Ab initio electronic structure calculations were performed to

further understand the experimental findings. The NO2 (SO2)

with or without H2O molecules were absorbed on (13, 0) nano-

tube to examine the adsorption behavior. The density functional

theory within the local density approximation (LDA) and

generalized gradient approximation (GGA) for the exchange–

correlation energy was adopted. Our model structures were

relaxed until the atomic forces on the atoms were smaller than

0.05 eV �A�1.A double numerical plus polarization basic set was

employed within the Dmol3 package with an orbital cut off of

4.0 �A. In the case of adsorption of NO2 onto the CNT, since

Dmol3 has some weak points for the calculations of the spin

system, the plane wave basic set implemented by PWscf with spin

J. Mater. Chem., 2011, 21, 4502–4508 | 4503

Fig. 2 (a) Resistance change with 2 ppm SO2 gas exposure at different

humidity levels. Shadowed regions indicate exposure of gas. Humidity

levels are indicated in the legend. Region I: gradual increment of

humidity levels, Region II: reproducibility of humidity levels. Before

exposing to gas in the next step, the system was allowed to recover close

to the original state. With this, the release time was varied, while the gas

exposure time was fixed. (b) Resistance change only at different humidity

levels. Humidity level is indicated in the legend.

Dow

nloa

ded

by S

ungk

yunk

wan

Uni

vers

ity o

n 11

Mar

ch 2

011

Publ

ishe

d on

07

Febr

uary

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0JM

0322

7AView Online

polarization functional was used. The kinetic cut-off for wave

functions was 35 Ry (475 eV) for all atoms and the convergence

threshold for calculation of self-consistent energy was �10�5 eV.

The structures were relaxed until the forces on the atoms were

smaller than 0.025 eV �A�1.

Results and discussion

Adsorption of SO2 with different humidity levels

Fig. 1 shows a schematic of comb-type interdigitated electrodes

which are composed of random-network SWCNTs and the

experimental setup. The performance of the device strongly relies

on the dispersion and CNT density. Small bundle size and

random distribution of SWCNTs in the electrodes are the key

factors for having a reasonable range of series resistance that

suppresses the dark current and enhances the on-current with

gases.28 Fig. 1b and 1c are the SEM images of the electrode and

typical SWCNTs network used in this study. Individual

SWCNTs or small-size bundles of SWCNTs were well distrib-

uted between two electrodes with a reasonable density. The

optimized series resistance was in a range of a few tens kU–MU.

Fig. 2a shows a 2 ppm SO2 response to humidity. When the

sensor was installed in the chamber, the chamber was purged with

dry air (humidity level is about 5%) for 100 s, followed by SO2

uptake. The resistance increased gradually as the humidity level was

incremented and the sensitivity was negligible at low humidity

levels. The gradual increase of the resistance at a low humidity level

originates mainly from humidity (Fig. 2b). The humidity depen-

dence became more obvious at higher humidity levels. With an

increasing humidity level, the resistance of the originally p-type

CNT channel increased, suggesting charge compensation. In other

words, electron donation from H2O compensates hole carriers in

the channel, congruent with previous works.23,24,26 The sensitivity of

SO2 gas increased at high humidity levels (92%). The resistance

increased more prominently with SO2 uptake and then decreased

upon release of SO2 gas. The gas uptake was maintained for about

500 s each time where the resistance approached saturation. This

implies that the reaction is still in an equilibrium state, where the

adsorption is dominated by energetics. This will be discussed in total

energy calculations later. The repeatability was checked in region II

with a low humidity level, where dry air was again introduced at the

last cycle. Negligible response was again observed, similar to the

initial stage in region I. The humidity-assisted reactivity of SO2 gas

is intriguing. We emphasize here that no type conversion is involved

with/without SO2 even at a high humidity level. As shown in Fig. 2b,

the resistance decreased at a moisture concentration of 92%. A high-

resistance device was used in this experiment but the response to

gases is expected to be similar to the low-resistance device in Fig. 2a.

This is in good contrast with previous reports which explained the

decrease of resistance at high humidity level by CNTs type

conversion from a p-type to an n-type channel.23,26,29 The similar gas

reaction behavior was observed in different devices with initial

resistances in a range from �kU to �MU.

Theoretical analysis for cooperative charge transfer between

SO2-nH2O complex and p-type CNT resistor

The underlying mechanism will be elucidated with DFT calcu-

lations. Fig. 3 describes the amount of electrons extracted from

4504 | J. Mater. Chem., 2011, 21, 4502–4508

intrinsic CNT (without charge doping) as a function of moisture

level with/without SO2 (squares) according to theory calcula-

tions. With an increasing moisture level, more electrons are

extracted from the CNT, i.e., water behaves like a p-type dopant

to intrinsic CNT. In the case of SO2 uptake, SO2 extracts elec-

trons from the CNT in the absence of humidity (solid squares).

The amount of electrons extracted from CNT is reduced when

moisture is introduced but electrons are still extracted from CNT

(Fig. 3b). However, this model cannot be applied to explain our

experimental observations, since pristine CNTs in experiment are

originally p-type due to ambient gas adsorption or substrate

effect.13

To make our theoretical model similar to the experimental

situation, +2 charges (or +0.0167 e/carbon atom) are introduced

in the calculation system, i.e., two electrons are extracted from

CNT originally to ensure a large Fermi energy shift (3.76 eV) to

simulate pristine p-type CNT. Even charges were chosen for

calculational convenience. As a consequence, the Fermi level is

downshifted below the highest occupied molecular orbital

(HOMO) (Fig. 3c). CNT is therefore p-type. It is quite intriguing

to observe that in this case electrons are donated to the CNT at

This journal is ª The Royal Society of Chemistry 2011

Fig. 3 (a) The amount of electron transfer from CNT to different adsorbents based on Hirshfeld charge analysis. Squares and circles indicate inter-

action between gas molecules and intrinsic/+2 charged CNT, respectively. Summaries of charge transfer in the case of intrinsic CNT (b) and p-doped

CNT (c).

Dow

nloa

ded

by S

ungk

yunk

wan

Uni

vers

ity o

n 11

Mar

ch 2

011

Publ

ishe

d on

07

Febr

uary

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0JM

0322

7AView Online

low humidity levels, as shown in Fig. 3a (circles), whereas at high

humidity levels, electrons are extracted from H2O–cluster formed

by hydrogen bonding again.

This theoretical prediction in fact explains experimental

observations of pure water adsorption in Fig. 2b. The increase of

resistance at low humidity levels is explained by charge

compensation from a water molecule to CNT. Thus, when SO2 is

introduced at a low humidity level, the effect of SO2 to resistance

is minor (Fig. 3a, solid circles) since both H2O and SO2 are

electron donors to p-type CNT. However, at a high humidity

level, the amount of electrons is slightly donated from SO2-nH2O

complex to CNT, which is the opposite direction of H2O–cluster

adsorption. In other words, the presence of SO2 compensates the

effect of H2O–cluster at a high humidity level. This explains why

a rather mild resistance change with SO2-nH2O was shown in

Fig. 2a at high humidity level compared to the rapid decrease of

the resistance in the absence of SO2 in Fig. 2b. Thus, all the

experimentally observed phenomena can be explained by charge

compensation rather than type conversion. This argument still

holds true even for the case of a lightly p-doped CNT. Intuitively,

electron affinity of positively charged (p-type) CNT increases

compared to neutral CNT. In neutral CNT, oxygen in water

molecules has a higher electron affinity and thus extracts elec-

trons from CNT independent of the number of molecules, as

shown in Fig. 3b. However, in the case of positively charged

CNT which has a higher electron affinity than a single water

molecule, the single water molecule loses electrons to CNT, while

a water cluster still extracts electrons from CNT, as shown in

Fig. 3c. The charge transfer direction will be competed by the

Fermi level shift of CNTs and humidity level.

This journal is ª The Royal Society of Chemistry 2011

Adsorption of NO2 with different humidity levels

Fig. 4 shows the resistance change with 2 ppm NO2 gas exposure

at different humidity levels. A similar procedure to SO2 was

conducted for NO2 gas. Unlike SO2 gas, the sensitivity was

highly independent of humidity level, although there was a slight

variation with different humidity levels. With an uptake of NO2

gas the resistance decreased, which is the opposite to that

observed for SO2 gas at high humidity levels. This was attributed

to the charge transfer from CNTs to NO2 molecules.1 The

repeatability was also confirmed in region II by alternating dry

air and 92% moisture. The sensitivity with dry air became lower

than that of the initial stage of region I (inset) due to the pre-

existing moisture in the previous cycle.

Dynamic response and recovery of NO2 and SO2 gas

Next we investigate the effect of moisture to response time and

recovery time of NO2 gas. The response (recovery) to NO2 gas

upon gas exposure (release) at different humidity levels was

extracted from Fig. 4 and redrawn in Fig. 5a and 5b. The resis-

tance change decreased (increased) exponentially and saturated

later. The response curve for adsorption (desorption) can be

fitted to DR ¼ DR0 � DR0 exp(�t/s). Here R0 is the final resis-

tance change at a long exposure time (recovery), t is time, and s is

the response (recovery) time constant. At small t, the curve can

be approximated to DR ¼ DR0(t/s). Therefore, s can be obtained

from the linear slope S (s ¼ DR0/S, S ¼ DR/t). R0 can be

extracted from exponential fitting and the slope can be measured

from the linear fitting. The linear fitting was plotted in the inset of

J. Mater. Chem., 2011, 21, 4502–4508 | 4505

Fig. 4 A similar experiment to SO2 was performed using 2 ppm NO2 gas

exposure. The inset shows the enlarged portion of NO2 input at

a humidity level < 5% in region I.

Dow

nloa

ded

by S

ungk

yunk

wan

Uni

vers

ity o

n 11

Mar

ch 2

011

Publ

ishe

d on

07

Febr

uary

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0JM

0322

7AView Online

Fig. 5a and 5b. Note that the reference resistance during

desorption was reset to zero when the gas was released.

The response time constant extracted from the linear slope in

Fig. 5c decreased gradually with increasing humidity level. The

response time is directly related to the activation barrier height,

as shown in the reaction energy curve in the inset of Fig. 5c. The

response of NO2 gas on CNTs in the presence/absence of

Fig. 5 (a) The resistance change with 2 ppm NO2 gas exposure at different hu

exposed. The inset indicates the linear curve fitting at an initial time as disc

desorption at different humidity levels. The reference resistance was taken righ

after NO2 desorption. (c) Response time constant and recovery time constant

pathway of an NO2 molecule. (d) Schematic of proposed reaction mechanism

4506 | J. Mater. Chem., 2011, 21, 4502–4508

moisture is expected to be endothermic. Thus, the activation

barrier height decreases with increasing humidity level. On the

other hand, the recovery time constant behaves differently from

the response time constant. The recovery time increased gradu-

ally with increasing humidity level and decreased at high

humidity level (92%). It is also noted that the recovery time is

much larger, by about 10 times, than the response time. This

implies that the binding energy is much larger than the activation

barrier height since gas desorption mainly relies on the binding

energy. It has been known that NO2 gas decomposes into NO

and NO3 by the metal oxide catalytic reaction.30 Since transition

metals such as Fe and Co are introduced during CNT synthesis,

they always remain even after sophisticated multistep purifica-

tion processes. In this study we used HiPCO SWCNTs with 85

wt% purity and therefore metal catalysts are likely to remain and

could be oxidized. Therefore, the dominant species that adsorbs

on the CNT surface is not only NO2 but also NO and NO3. The

binding energy of NO2, NO, and NO3 are 0.47, 0.44, and 1.08 eV,

respectively, as shown in Table 2.30 The formation of these

products depend on different humidity levels, and thus the

recovery constant will be different.

(i) At a low humidity level, a possible chemical reaction

pathway under the presence of metal oxide (M–O) on the CNT

surface is as follows:31

NO2(g) + O site / NO3(ad) (1)

NO2(g) + M site / NO(g) + M–O (2)

midity levels. The reference resistance R0 was set right before NO2 gas was

ussed in the manuscript. (b) Corresponding resistance change upon gas

t before NO2 gas was turned off. The inset shows linear curve fitting right

plots according to different humidity levels. The inset shows the reaction

for NO2 response on CNTs with catalysts.

This journal is ª The Royal Society of Chemistry 2011

Table 2 The binding energy calculations for different adsorption configurations. The binding energy is calculated by Eb(CNT-ads) ¼ Etot(CNT) +Etot(ads) � Etot(CNT-ads), where Etot is the total energy of a given system

Gas species NO NO3 NO2 SO2

Configuration SWCNT(11,0) -NO

SWCNT(11,0) -NO3

SWCNT(11,0) -NO2

SWCNT(13,0) -NO2

(SWCNT (13,0)) -SO2 (SWCNT (13,0)-H2O) -SO2

Bindingenergy (eV)

0.47 (LDA) a 1.08 (LDA) a 0.44 (LDA) a 0.09 (GGA) b 0.292(LDA) c 0.05(GGA) b 0.439(LDA) c 0.235(GGA) c

a Ref. 30. b Calculated by PWscf code with plane wave basic set. c Calculated by Dmol3 code with numerical atomic orbital.

Dow

nloa

ded

by S

ungk

yunk

wan

Uni

vers

ity o

n 11

Mar

ch 2

011

Publ

ishe

d on

07

Febr

uary

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0JM

0322

7AView Online

NO2(g) + M–O / M–NO3 (ad) (3)

Since the catalyst wt% is usually small, generation of NO3 (ad)

will be small and the dominant species will still be NO2, this is

indicated in the sketch of Fig. 5d (i).

(ii) At an intermediate humidity level, a possible chemical

reaction pathway will be:8,30

3NO2(g) + H2O / NO(g) + 2HNO3 (4)

2NO2(g) + 2NO(g) / 2NO3(ad) + N2(g) (5)

Following reaction (4), HNO3 is produced, which will form

C]O bonds on the CNT surface. This will enhance the reaction

(1) and generate more NO3(ad). Another consequence of reac-

tion (4) is further formation of NO3(ad) by following the

chemical reaction (5). As a result, NO2 and NO3 coexist at

intermediate moisture levels and NO3 becomes dominant with an

increasing moisture level. Therefore, the recovery time constant

increases due to the dominant NO3 that has a higher binding

energy than NO2, this is indicated in the sketch of Fig. 5d (ii).

(iii) At a high humidity level such as 92%, metal oxides and

C]O sites are deactivated by the abundant H2O molecules and

therefore we expect simply that the CNT surface is covered by

H2O molecules. In this case, NO2 is a dominant adsorption

species again without further catalytic decomposition. Thus the

recovery time constant is expected to decrease; this is indicated in

the sketch of Fig. 5d (iii).

However, in the case of SO2, the response/recovery time had

no such obvious tendency. This can be explained by the radical

nature of NO2 compared to SO2. The highest occupied molecular

orbital (HOMO) in NO2 is half-filled and appears to have

bonding interactions with the valence and conduction bands of

metal oxides.31 This induces the catalytic decomposition. SO2

which has fully occupied HOMO is stable by itself, making the

decomposition difficult to realize. Our calculation shows that the

binding energy of a SO2 molecule on the CNT surface is 0.05 eV

from GGA, which is less than the 0.09 eV observed for a NO2

molecule. The binding energy of SO2 on H2O-covered CNT is

0.235 eV (GGA) and 0.439 eV (LDA), stronger than that on

intrinsic CNT (Table 2). This also explains why reasonable

sensitivity with SO2 gas was observed at high moisture level only.

Conclusion

This study developed a new route to selectively detect NO2 and

SO2 gas by simply introducing humidity. In the case of NO2, the

device resistance decreased regardless of the humidity level, since

This journal is ª The Royal Society of Chemistry 2011

it behaves like a p-type dopant. However, the device resistance

increased upon SO2 exposure only at a high humidity level (92%),

while SO2 was rather insensitive at a low humidity level. This

different behavior is a consequence of different charge transfer

direction between a SO2–nH2O complex and a H2O–cluster upon

a p-type CNT. Our density functional theory calculations by

introducing p-type CNT provided consistent data with experi-

mental results. To date, this simple method of introducing

moisture provides a possibility to selectively detect NO2 and SO2

under ambient conditions for the first time.

Acknowledgements

This work was supported by the STAR-faculty program and

WCU (World Class University) program through the KRF

funded by the MEST (R31-2008-000-10029-0).

Notes and references

1 J. Kong, N. R. Franklin, C. W. Zhou, M. G. Chapline and S. Peng,Science, 2000, 287, 622.

2 L. Valentini, I. Armentano, J. M. Kenny, C. Cantalini, L. Lozzi andS. Santucci, Appl. Phys. Lett., 2003, 82, 961.

3 F. Yao, S. C. Lim, W. J. Yu, I. H. Lee, F. Gunes, H. R. Hwang,S. B. Yang, P. K. So, G. H. Han and Y. H. Lee, J. Phys. Chem. C,2010, 114, 3659.

4 M. Penza, G. Cassano, R. Rossi, M. Alvisi, A. Rizzo andM. A. Signore, Appl. Phys. Lett., 2007, 90, 173123–1.

5 G. H. Lu, L. E. Ocola and J. H. Chen, Adv. Mater., 2009, 21, 2487.6 C. Bittencourt, A. Felten, E. H. Espinosa, R. Ionescu, E. Llobet,

X. Correig and J.-J. Pireaux, Sens. Actuators, B, 2006, 115, 33.7 P. F. Qi, O. Vermesh, M. Grecu, A. Javey, Q. Wang, H. J. Dai,

S. Peng and K. J. Cho, Nano Lett., 2003, 3, 347.8 A. Goldoni, R. Larciprete, L. Petaccia, S. Lizzit and J. Am, J. Am.

Chem. Soc., 2003, 125, 11329.9 A. Goldoni, L. Petaccia, L. Gregoratti, V. Kaulich, A. Barinov,

S. Lizzit, A. Laurita, L. Sangaletti and R. Larciprete, Carbon, 2004,42, 2099.

10 J. J. Zhao, A. Buldum, J. Han and J. P. Lu, Nanotechnology, 2002, 13,195.

11 S. Peng and K. Cho, Nanotechnology, 2000, 11, 57.12 H. Chang, J. D. Lee, S. M. Lee and Y. H. Lee, Appl. Phys. Lett., 2001,

79, 3863.13 W. Kim, A. Javey, O. Vermesh, Q. Wang, Y. Li and H. J. Dai, Nano

Lett., 2003, 3, 193.14 S. G. Yu and W. Yi, IEEE Trans. Nanotechnol., 2007, 6, 545.15 J. Suehiro, G. Zhou and M. Hara, Sens. Actuators, B, 2005, 105, 164.16 D. R. Shankaran, N. Uehera and T. Kato, Sens. Actuators, B, 2002,

87, 442.17 H. Li, Q. Wang, J. Xub, W. Zhang and L. Jin, Sens. Actuators, B,

2002, 87, 18.18 G. Shi, M. Luo, J. Xue, Y. Xian, L. Jin and J. Y. Jin, Talanta, 2001,

55, 241.19 D. R. Shankaran and S. S. Narayanan, Sens. Actuators, B, 1999, 55, 191.20 R. Pati, Y. Zhang, S. K. Nayak and P. M. Ajayan, Appl. Phys. Lett.,

2002, 81, 2638.

J. Mater. Chem., 2011, 21, 4502–4508 | 4507

Dow

nloa

ded

by S

ungk

yunk

wan

Uni

vers

ity o

n 11

Mar

ch 2

011

Publ

ishe

d on

07

Febr

uary

201

1 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0JM

0322

7AView Online

21 O. Leenaerts, B. Partoens and F. M. Peeters, Phys. Rev. B: Condens.Matter Mater. Phys., 2008, 77, 125416–1.

22 F. Schedin, A. K. Geim, S. V. Morozov, E. W. Hill, P. Blake,M. I. Katsnelson and K. S. Novoselov, Nat. Mater., 2007, 6, 652.

23 P. S. Na, H. Kim, H. M. So, K. J. Kong, H. B. H. Chang, Ryu,Y. Choi, J. O. Lee, B. K. Kim and J. J. Kim, Appl. Phys. Lett.,2005, 87, 093101–1.

24 P. C. P. Watts, N. Mureau, Z. Tang, Y. Miyajima, J. D. Carey andS. R. P. Silva, Nanotechnology, 2007, 18, 175701.

25 J. R. Huang, M. Q. Li, Z. Y. Huang and J. H. Liu, Sens. Actuators, A,2007, 133, 467.

26 A. Zahab, L. Spina, P. Poncharal and C. Marli�ere, Phys. Rev. B:Condens. Matter, 2000, 62, 10000.

4508 | J. Mater. Chem., 2011, 21, 4502–4508

27 O. K. Varghese, P. D. Kichambre, D. Gong, K. G. Ong, E. C. Dickeyand C. A. Grimes, Sens. Actuators, B, 2001, 81, 32.

28 S. Y. Jeong, K. K. Kim, K. H. An, H. R. Hwang, C. S. Han,M. H. Yun and Y. H. Lee, Nano, 2006, 1(3), 235.

29 At a high humidity level, an additional conduction channel of a waterlayer might be formed on the surface of the CNT. However, theconductivity through the liquid channel is still much smaller thanthat of CNTs. Therefore, conduction through a liquid channelshould be negligible. See paper J. Am. Chem. Phys., 2002, 124,9058–9059.

30 S. Peng, K. Cho, P. F. Qi and H. J. Dai, Chem. Phys. Lett., 2004, 387, 271.31 J. A. Rodriguez, T. Jirsak, S. Sambasivan, D. Fischer and A. Maiti,

J. Chem. Phys., 2000, 112, 9929.

This journal is ª The Royal Society of Chemistry 2011