Chemical Physics Letters 383 (2004) 617622

www.elsevier.com/locate/cplett

Sensors for inorganic vapor detection based on carbon nanotubes

and poly(o-anisidine) nanocomposite material
a,*
L. Valentini , V. Bavastrello b, E. Stura b, I. Armentano a, C. Nicolini b, J.M. Kenny a

a
Materials Science and Technology Center, INSTM Unit, Universita di Perugia, 05100 Terni, Italy
b
Department of Biophysical M&O Science and Technologies, University of Genoa, 16132 Genoa, Italy
Received 24 September 2003; in nal form 15 October 2003
Published online:

Abstract

A gas sensor, fabricated by selective growth of aligned carbon nanotubes (CNTs) by pulsed plasma on Si3 N4 /Si substrates
patterned by metallic platinum, is presented for inorganic vapor detection at room temperature. Poly(o-anisidine) (POAS) depo-
sition onto the CNTs device was shown to impart higher sensitivity to the sensor. Upon exposure to HCl the variation of the CNTs
sensitivity is less than 4%, while the POAS-coated CNTs devices oer a higher sensitivity (i.e. 28%). The extended detection ca-
pability to inorganic vapors is attributed to direct charge transfer with electron hopping eects on intertube conductivity through
physically adsorbed POAS between CNTs.
2003 Elsevier B.V. All rights reserved.

1. Introduction Nevertheless, several outstanding issues remain. Firstly,

The high surface area, size, hollow geometry and
chemical inertness remarkable properties of CNTs make
them attractive for demanding applications in the eld of
gas sensing. To date studies on possible applications of
CNTs have been focused either on individual single-
walled carbon nanotubes as sensitive materials towards
O2 , NO2 and NH3 [14] or on multi-walled carbon na-
notubes (CNTs) mats as NH3 , CO, CO2 humidity and O2
gas sensors [57]. More recently we have reported on the
preparation of CNTs thin lms by radio frequency plas-
ma enhanced chemical vapor deposition on Si3 N4 /Si
substrates, provided with interdigital Pt electrodes, for
NO2 monitoring at low concentrations (10100 ppb in air)
[8].
Nanotube sensors oer signicant advantages over
conventional metal-oxide-based electrical sensor materi-
als in terms of sensitivity and small sizes needed for min-
iaturization and construction of massive sensor arrays.
* complex respect to other CP. This fact is due to their
Corresponding author. Address: Department of Civil and Envi-
ronmental Engineering, University of Perugia, Pentima Bassa, 05100
Terni, Italy. Fax: +39-0744492925.
E-mail address: mic@unipg.it (L. Valentini).
for sensing purposes, it is desirable to reliably obtain de-
vices consisting of semiconductor CNTs operating at
room temperature [1,4,9]. Secondly, molecular sensing
requires strong interactions between a sensor material
and target molecules. This is also the case for nanotubes.
We have found that nanotubes are not sensitive to many
types of molecules (i.e. CO, H2 , HCl), indicating an ap-
parent lack of specic interactions between nanotube and
these molecules. We believe that nanotube sensors with
molecular specicity can be obtained through rational
chemical and/or physical modication of nanotubes.
Chemical modication may include sidewall functional-
ization by desired molecular groups [10] while physical
modication may involve simple deposition of suitable
species on the nanotubes. Here, we present our recent
results in tackling the issues above.
Polyaniline (PANI) and its derivatives have been
deeply studied for their good electric properties, easy
methods of synthesis and high environmental stability
[1114]. The chemistry of polyanilines is generally more
dependence on both the pH value and the oxidation
states, described by three dierent forms known as leu-
coemeraldine base (LB) (fully reduced form), emeraldine

0009-2614/$ - see front matter 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.cplett.2003.11.091
618 L. Valentini et al. / Chemical Physics Letters 383 (2004) 617622

H H
N N N N

OCH3 OCH3 OCH3 OCH3

- 2 H+ + 2 H+

H H H H
N N N N
+. +.

OCH3 OCH3 OCH3 OCH3

Fig. 1. Schematic representation of POAS in emeraldine base and

emeraldine salt states.

base (EB) (50% oxidised form), and pernigraniline base

(PB) (fully oxidised form). The most important is the
EB form and its protonation by means of H ions
generated from protic acids gives the emeraldine salt
form (ES), responsible of the strong increment of
conducting properties [15] (see Fig. 1). This process is
reversible and it is possible for the presence of imine
groups basic sites located along the conducting polymer
backbone [16,17]. The remarkable fact that the chemi-
calphysical properties of PANI and its derivatives are
pH sensitive has led to the study of these materials as
sensors [18,19].
The present work focuses on CNTs thin lms pre-
pared by radio frequency pulsed plasma enhanced
chemical vapor deposition glow discharge system for the
detection of HCl. In this work we show that excellent
molecular HCl sensors can be enabled by adding on
CNTs poly(o-anisidine) (POAS). The POAS modied
CNTs sample exhibit signicant electrical conductance
modulation upon exposure to small concentrations of
HCl in air. Importantly, these advanced sensing char-
acteristics are obtained at room temperature. Further-
more, we demonstrate that CNTs and poly(o-anisidine)
nanocomposites [20] can be easily scaled up with en-
semble of self-assembling of CNTs made by a simple
plasma deposition growth approach.
2. Experimental details
The CNTs thin lm was grown using a radio fre-
quency pulsed plasma enhanced chemical vapor depo-
sition (RF PECVD) system. Prior to the nanotube
growth, a Si3 N4 /Si substrate was patterned with plati-
num lm (60 nm thick) by vacuum deposition through
shadow masks, containing rectangular stripes 30 lm
wide and a back deposited thin lm platinum heater
commonly used in gas sensor applications [3]. A thin
lm (3 nm) of Ni catalyst was deposited onto the Si3 N4 /
Si substrates using thermal evaporation (Fig. 2a). The
CNTs lm was obtained by pulsed RF-PECVD depo-
sition on Si3 N4 /Si substrates as reported previously [21].
Fig. 2. Nanotube device structure. (a) SEM image of platinum pat-
terned Si3 N4 /Si substrate. (b) High resolution SEM images of aligned
CNTs. The inset is the top view with the Pt electrode region high-
lighted and the as-grown aligned structure of CNTs on a Si3 N4 /Si
substrate.
POAS was chemically synthesized by oxidative po-
lymerization of the monomers by using (NH4 )2 S2 O8
under controlled conditions. For the synthesis the
monomer, previously distilled, was dissolved in HCl 1 M
aqueous solution. The oxidant was thus added slowly
maintaining the temperature at 04 C by means of an
ice bath. The reaction was continued for about 12 h. The
dark green precipitate emeraldine salt form recovered
from the reaction vessel was ltered and washed with
distilled water. The precipitate was subsequently treated
with a diluted solution of NH4 OH for 6 h to obtain the
crude materials in the undoped form (blue in color). The
precipitate recovered after ltration was washed again
with water and treated rstly with methyl alcohol, then
with ethyl ether for 24 h to remove the oligomers con-
stituted by few monomer units. The material obtained
from the last ltration was heated at 80 C in a vacuum
system to remove the traces of solvents. The nal
product contained the emeraldine base form of POAS
that was completely soluble in chloroform.
In order to check the eective improvement generated
by the presence of CNTs inside the polymeric matrix,
measurements of specic conductivity of both the
 L. Valentini et al. / Chemical Physics Letters 383 (2004) 617622
nanocomposite and the pure conducting polymer were
carried out. Specic resistance calculations were per-
formed basing on current/voltage (I=V ) characteristics
measured with 6517 electrometer (Keitley, USA) driven
by computer. Samples for these measurements were
deposited onto glass substrates by LS technique. Con-
tacting was set up with a silver paint. V =I characteristics
revealed linear behavior for all samples demonstrating
that Ohmic contact was realized. Thus, dierent num-
bers of monolayers (1, 10, and 30) of both the material
were deposited onto glass substrates by using the LS
technique [20].
For gas sensing measurements a POAS lm (30
monolayers) was then deposited onto the planar resistor
coated with CNTs.
The scanning electron microscopy investigation was
performed on a eld emission scanning electron micro-
scope LEO 1530 operated at 5 kV. The transmission
electron microscopic (TEM) investigation was per-
formed on a conventional 200 kV electron microscope.
For TEM analysis the as grown aligned CNTs lm was
deposited on Si substrate and was immersed into a HF/
H2 O solution. After 10 min, peeled nanotubes were
dispersed on a copper carbon-microgrid.
Raman scattering spectra were recorded by a Jobin
Yvon micro-Raman LabRam system. All the spectra
reported in the work were measured in backscattering
geometry using 514.5 nm laser excitation wavelengths.
To characterize the temperature dependence of the
electrical transport property, CNTs and polymer/
nanotubes composite deposited onto the planar resistor
were held in a LEYBOLD COLD-HEAD cryodyne re-
frigerator, and the temperature was controlled by 1901
temperature controller.
The electrical resistance of the lm was measured in
owing air, using a volt-amperometric technique with a
Keitley 236 multimeter. HCl mixture was prepared in
order to have vapor concentration of 100 ppm. Electri-
cal measurements were performed by xing the tem-
perature of the lm at 25 C.
3. Results and discussion
High-resolution eld emission SEM images of CNTs
deposited with pulsed plasma are plotted in Fig. 2b. In
Fig. 2b the edge zone between the Pt electrode and the
sensitive CNTs lm is shown. The reason why CNTs did
not grow on Pt, as shown in Fig. 2b, can be attributed to
the lack of fragmentation of the Ni lm on the platinum
electrode. This fact can be due to a diusion of Ni atoms
through Pt during the annealing process at 650 C,
eventually leading to the formation of a binary NiPt
alloy. From Fig. 2b, it is possible to observe that the
nanotubes are aligned perpendicularly to the substrate,
as the pulsed plasma technique well allows. The tubular
619
structure of the nanotubes was veried using TEM mi-
croscopy, as shown in Fig. 3. The nanotubes generally
consist of graphitic shells 150200 nm long. Fig. 3 rep-
resents the lattice image of graphite planes parallel to
the axis. However, long-range lattice images are not
formed due to the relatively low growth temperature.
This was also conrmed in Raman scattering measure-
ment that a relatively large defective peak at 1350 cm
1
was observed with the main graphitic G-peak at 1600
cm
1 [22,23].
Fig. 4a shows the resistance of both CNTs and POAS
coated devices plotted logarithmically against T
1=4
while in Fig. 4b is reported the resistance against T
1=2
over the higher temperature range for CNTs device. One
important thing in our CNTs based system is that the
data of resistance follow the general hopping law ex-
pression in lnR=R298 K / 1=T x with an exponent x
changing from 1/4 to 1/2. In particular it was found that
the attempts to t the R=R298 K data for CNTs by using a
dierent reciprocal temperature exponent in the hopping
law expression are markedly poorer. Only the T
1=2
plots yield the closest linear dependence over tempera-
ture, thus suggesting the Coulomb gap in this tempera-
ture range; transport is due to localized carriers
originating from metallic CNTs [2426]. The CNTs de-
vice shows a nearly temperature dependence of relative
resistance of metallic conductors in a range from 298 to
100 K. It suggests that the resulting nanotube mainly
contains metallic tubes, while some semiconducting
tubes cannot be ruled out. When the temperature de-
creases the semiconducting tubes become insulating,
whereas the metallic tubes will maintain their electronic
structure.
Table 1 shows the results obtained from the specic
conductivity measurements of POAS and POAS-CNTs
in the undoped (in air) and doped (exposed to HCl)
forms, respectively. By the analysis of the data it can be
seen that, for thin lms, the close vicinity to the sub-
strate determines the rather high level of the conduc-
tivity due to the presence of some defects resulted from
Fig. 3. TEM micrograph of CNTs obtained by plasma pulsed mode.
620 L. Valentini et al. / Chemical Physics Letters 383 (2004) 617622

(a) 7.0 of the same order of magnitude for all samples of dif-
ferent thickness. This result highlights the impossibility
6.5 POAS coated CNTs of enhancing the specic conductivity of POAS by in-
creasing the number of lm monolayers on the substrate
6.0 even when the conducting polymer is in its better con-
ln(R) [Ohm]

ducting form. Interesting results were obtained in the

5.5 3.0
POAS coated CNTs case of POAS-CNTs nanocomposite where the incre-
2.5
ment of monolayers results in a signicant improvement
R /R298K
5.0 2.0

1.5 of the specic conductivity. In this case the CNTs seems

4.5 1.0 to have close interactions with the conducting mecha-
0 50 100 150 200 250 300
Temperature [K] nism of the polymer demonstrating that the polymeric
4.0 chains are not only wrapped around the CNTs.
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
-1/4 -1/4
Fig. 5 shows two distinct groups of IV curves re-
T [K ] sulting from nanotubes and polymer POAS coated na-
notubes. Each group consists of results from several
(b) sensors. It can be seen from Fig. 5 that the conductivity
6.8
CNTs (slope) is almost the same for CNTs and POAS-coated
6.4
CNTs, but it varies for the device when exposed to HCl.
6.0 In particular, Fig. 5 shows that the conductance of na-
ln(R) [Ohm]

notubes slightly increases (inset of Fig. 5), while the

5.6
1.8 POAS coated device shows a signicant conductance
CNTs
5.2 1.6
1.4
increment. It is interesting to note (Fig. 5b) that the
R/R298K

1.2 response transient of POAS coated device is a few sec-

4.8 1.0
0.8
4.4 0.6
4.0
0.1 0.2 0.3 0.4 0.5
-1/2
T [K
Fig. 4. (a) Temperature dependence of the POAS coated device resis-
tivity plotted as lnR vs. T
1=4 . The inset shows the temperature de-
pendence of scaled resistivity. (b) Temperature dependence of
resistivity of the CNT device plotted as lnR vs. T
1=2 . The inset shows
the temperature dependence of scaled resistivity in current direction
perpendicular to the tube axis.
the interaction of the monolayer with the substrates.
On the contrary, when the thickness of the lm in-
creases, the electrical properties are determined by the
POAS itself, which is insulating in the undoped form. In
other words the poor specic conductivity of the con-
ducting polymer in the undoped is predominant. It can
be also seen that the presence of CNTs in the polymeric
matrix seems to decrease the mere insulator capability of
the undoped form, but is not able to generate a wide
improvement of the conducting properties. In the case of
POAS and POAS-CNTs doped form it can be seen that
for the pure polymer the specic conductivity is almost
onds, while sensor based on resistance changes of POAS
0 50 100 150 200 250 300
exhibits a poorer response time. The circumstance to
Temperature [K] maintain lm resistance below 1 kX, which is signicant
0.6 0.7 0.8 0.9 lower with respect to that reported for POAS sensors
-1/2
] (50 MX), makes POAS coated CNTs lms integration
in electronic circuitry easier and cheaper, since lower
DC voltages are required to drive the sensor response.
The doping process of polyanilines is always associated
to conformational modications of the polymer chains,
due to the local distortions created by the addition of
H ions to the basic sites and usually provides stable
systems [27]. It means that the conducting polymer in
the doped form can be maintained in this state for long
periods of time till the material reacts with basic re-
agents and strongly changes its chemicalphysical
properties. In other words, the reversibility of the pro-
cess is not spontaneous.
From Fig. 5c and Table 2, if we dene sensor sensi-
tivity (S) as the ratio S RA
RG =RA   100, where
RA represents the resistance in air and RG the resistance
in vapor, the gas sensitivity increases from S 3:0% to
27.9%. It reveals that by selecting proper polymer
functionalization sensor sensitivity to HCl may be
improved.

Table 1
Specic conductivity parameters of POAS and POAS-CNTs composite LS lms as a function of dierent numbers of layers
Undoped form specic conductivity (S/cm) Doped form specic conductivity (S/cm)
1 monolayer 10 monolayers 30 monolayers 1 monolayer 10 monolayers 30 monolayers

5
6
7
3
3
POAS 9.3  10 4.1  10 6.6  10 1.7  10 1.4  10 1.1  10
3
POAS-CNTs 1.6  10
3 7.6  10
5 2.8  10
5 7.0  10
3 2.2  10
1 2.0
 L. Valentini et al. / Chemical Physics Letters 383 (2004) 617622 621

(a) Table 2
POAS coated CNTs
POAS coated CNTs exposed to HCl 100 ppm Sensor gas sensitivity to 100 ppm of HCl of (a) as-grown CNTs and (b)
-2
POAS coated device at 25 C
1.0x10
As-grown Exposed to HCl Sensitivity

5.0x10
-3 Resistance [X] Resistance [X] S [%]
CNTs 58.2  0.5 56.4  0.5 3.0
I [mA]

0.0 16 CNTs POAS-coated 61.5  0.5 44.3  0.5 27.9

12 CNTs exposed to HCl 100 ppm
8 CNTs
4

I [mA]
-3 0
-5.0x10
-4
-8
-12
-1.0x10
-2
-16 and H2 . Thus, the decreasing of the POAS coated CNTs
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6
Voltage [V] sensor resistance with HCl and substantial low sensi-
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 tivity for CNTs sensor represents a key feature of this
Voltage [V] material. Moreover, the possibility of arranging arrays
of sensor based on metal oxide semiconductor gas sen-
(b) POAS coated device exposed to HCl 100 ppm
POAS exposed to HCl 100 ppm
sors and POAS coated CNTs materials with an inverse
response to HCl, the high sensitivity of POAS coated
1.0
device to HCl together with suitable electrical response
processing algorithms, may represent a practical solu-
0.8 tion which improves selectivity.
Qualitatively, the CNTs can be thought of as being
R/Rt0

0.6
composed of conductive rods [29]. Hence, the resistance
of the material indicates that strong scattering occurs at
the tube boundaries as a result of intertube energy
0.4 barriers, so that the tube to tube contacts act as static
defects, limiting the mean free path of the electrons. The
0.2 resistivity (or resistance) behavior then reects the
0 10 20 30 40 50 mean-free-path perpendicular to the tubes. The elec-
Time [s] trons may localize on the individual tubes and intertube
electron transport is thermally activated, requiring
(c) CNTs exposed to HCl
40 POAS coated CNTs exposed to HCl
electrons to thermally hop across intertube energy
barriers [30].
35 The transport phenomena from the mat of aligned
nanotubes can be understood using the following model.
30
The network of CNTs is connected to CNTCNT
junctions. The cross-junctions between the CNTs or
S [%]

ropes importantly act as a gate for the carriers to move

5 In order to explain our observations regarding the
100 150 200 250 300
HCl [ppm]
Fig. 5. (a) IV curves at room temperature of POAS coated devices to
HCl vapor. The inset shows IV curves of the CNTs device at room
temperature to HCl 100 ppm. (b) The time-dependence change of the
normalized resistance (Rt0 is the initial resistance of the sample) of
POAS coated device and CNT lm at room temperature to HCl 100
ppm. (c) Sensitivity vs. HCl concentrations of POAS coated device and
CNT lm at room temperature.
Metal oxide semiconductor gas sensors reported to
date, are known to increase their resistance when in-
teracting with oxidizing gases like NO2 , Cl2 and O3 [28]
and decrease their resistance with reducing ones, like CO
in the mat.
change in the lm resistivity when the polymer is added
onto the tubes (Figs. 4 and 5), we suggest that the most
important mechanism involved is the eect on the
350 400 tunnelling resistance between tubes. The current ow in
our samples is strongly inuenced by the tunnelling
within the nanotube separations. All nanotubes share
the same graphene structure, hence their work function
is expected to be nearly the same, and the Fermi level
of the metallic tubes is expected to align the mid-gap of
the semiconducting energy gap. The insertion of the
polymer between CNTs (Figs. 4 and 5) modies the
density of states introducing an impurity-like level near
the onset of the valence band of the nanotube. When
the interacting polymer is suciently high, the CNTs
Fermi level is shifted to lower energies. Therefore the
barrier oered to charge transport is lowered with re-
spect to the as deposited nanotubes and the resistivity
622 L. Valentini et al. / Chemical Physics Letters 383 (2004) 617622
of the whole CNTs layer, as experimentally observed,
decreases.
In other words, the adsorption occurs in the intersti-
tial space between CNTs to form a nanotubepolymer
nanotube junction, leading to a hopping mechanism for
intertube charge transfer between CNTs and an intertube
modulation of the CNTs network in lieu of a conduc-
tivity change. This occurs for both the metallic and
semiconducting CNTs. The intratube modulation is
similar to that of the interaction between semiconductor
metal oxides and donor or acceptor types of molecules
[31] while the intertube modulation is similar to that
of the interaction between conductive polymers and
physically adsorbed molecules. The above phenomeno-
logical model explains our experimental observations;
i.e. in HCl detection, the intertube modulation plays a
key role.
4. Conclusions
In conclusion, CNTs thin lms prepared by pulsed
RF PECVD demonstrated their potentiality as a new
class of materials for HCl detection for environmental
applications. Moreover, polymer functionalization en-
hances the sensitivity to these devices. The advances
made here shall pave the way for future work in devel-
oping CNTs sensor arrays for highly sensitive and spe-
cic molecular detection and recognition in gases and in
solutions.
Acknowledgements
This work was supported by MIUR through the
project FIRB. One of the authors (I.A.) gratefully ac-
knowledge the nancial support from the National In-
stitute of Materials Science and Technology. We are
grateful to Dr. Jenny Alongi (Dipartimento di Chimica e
Chimica Industriale University of Genova) for access
to transmission electron microscopy as well as technical
support. The authors would like to thank Professor S.
Santucci for helpful discussions and the free supply of
the sensor layout.
References
[1] J. Kong, N.R. Franklin, C. Zhou, M.G. Chapline, S. Peng, K.
Cho, H. Dai, Science 287 (2000) 622.
[2] P.G. Collins, K. Bradley, M. Ishigami, A. Zettl, Science 287 (2000)
1801.
[3] J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, M. Meyyappan, Nano Lett.
3 (2003) 929.
[4] P. Qi, O. Vermesh, M. Grecu, A. Javey, Q. Wang, H. Dai, Nano
Lett. 3 (2003) 347.
[5] K.G. Ong, K. Zeng, C.A. Grimes, IEEE Sensors J. 2 (2002)
82.
[6] O.K. Varghese, P.D. Kichamber, D. Cong, K.G. Ong, E.A.
Grimes, Sens. Actuators B 81 (2001) 32.
[7] S. Chopra, A. Pham, J. Gaillard, A. Parker, M. Rao, Appl. Phys.
Lett. 80 (2002) 4632.
[8] L. Valentini, C. Cantalini, L. Lozzi, I. Armentano, J.M. Kenny,
S. Santucci, Appl. Phys. Lett. 82 (2003) 961.
[9] J. Kong, H. Dai, J. Phys. Chem. 105 (2001) 2890.
[10] P. Boul, J. Liu, E. Mickelson, C. Human, L. Ericson, I. Chiang,
K. Smith, D. Colbert, R. Hauge, J. Margrave, R. Smalley, Chem.
Phys. Lett. 310 (1999) 367.
[11] A.G. MacDiarmid, J.C. Chiang, M. Halpen, W.S. Huang, S.L.
Mu, N.L.D. Somasiri, W. Wu, S.I. Yaniger, Mol. Cryst. Liq.
Cryst. 121 (1985) 173.
[12] E.W. Paul, A.J. Ricco, M.S. Wrighton, J. Phys. Chem. 89 (1985)
1441.
[13] A.J. Epstein, A.G. MacDiarmid, in: H. Kuzmany, M. Mehring,
S. Roth (Eds.), Electronic Properties of Conjugated Polymer,
Springer Verlag, Berlin, 1989.
[14] T. Kobayashi, H. Yoneyama, H. Tamura, J. Electroanal. Chem.
177 (1984) 281.
[15] C. Chiang, A.G. MacDiarmid, Synth. Met. 13 (1986) 193.
[16] A.J. Epstain, Synth. Met. 18 (1987) 303.
[17] E.M. Genies, M. Labkowski, J. Electroanal. Chem. 236 (1987)
199.
[18] A.A. Karyakin, O.A. Bobrova, L.V. Lukachova, E.E. Karyakina,
Sens. Actuators B 33 (1996) 34.
[19] C.A. Lindino, L.O.S. Bulhoes, Anal. Chim. Acta 334 (1996)
317.
[20] V. Bavastrello, M.K. Ram, C. Nicolini, Langmuir 18 (2002)
1535.
[21] L. Valentini, I. Armentano, J.M. Kenny, L. Lozzi, S. Santucci,
Mater. Lett. 57 (2003) 3699.
[22] L. Valentini, C. Cantalini, I. Armentano, L. Lozzi, J.M. Kenny, S.
Santucci, J. Vac. Sci. Technol. B 21 (2003) 1996.
[23] C.J. Lee, D.W. Kim, T.J. Lee, Y.C. Choi, Y.S. Park, Y.H. Lee,
W.B. Choi, N.S. Lee, G. Park, J.M. Kim, Chem. Phys. Lett. 312
(1999) 461.
[24] Y. Lee, D. Kim, H. Kim, B. Ju, J. Appl. Phys. 88 (2000) 4181.
[25] Y. Yosida, I. Oguro, J. Appl. Phys. 86 (1999) 999.
[26] X. Wang, Y. Liu, G. Yu, C. Xu, J. Zhang, D. Zhu, J. Phys. Chem.
B 105 (2001) 9422.
[27] Z.T. de Oliveira Jr., M.C. dos Santos, Solid State Comm. 114
(2000) 49.
[28] C. Cantalini, M. Pelino, H.T. Sun, M. Faccio, L. Lozzi, M.
Passacantando, S. Santucci, Sens. Actuators B 3536 (1996)
112.
[29] A. Thess, R. Lee, P. Nikolaev, H. Dai, P. Petit, J. Robert, C. Xu,
Y. Lee, S.G. Kim, A.G. Rinzler, D.T. Colbert, G.E. Scuseria,
D. Tomanek, J.E. Fischer, R.E. Smalley, Science 273 (1996)
483.
[30] L. Langer, L. Stockman, J.P. Heremans, V. Bayot, C.H. Olk, C.V.
Haesendonck, Y. Brugnseraede, J.P. Issi, J. Mater. Res. 9 (1994)
927.
[31] J. Gardner, P.N. Bartlett, Electronic Nose: Principles and
Applications, Oxford University Press, New York, 1999.