thermal treatment effects on the material and gas-sensing properties of room-temperaturetungstenoxidenanorodsensors

Sensors and Actuators B 137 (2009) 297–304
Contents lists available at ScienceDirect
Sensors and Actuators B: Chemical
journal homepage: www.elsevier.com/locate/snb
Thermal treatment effects on the material and gas-sensing properties of
room-temperature tungsten oxide nanorod sensors
Yong Shin Kim
∗
Department of Applied Chemistry, Hanyang University, Ansan 426-791, Republic of Korea
article info
Article history:
Received 5 September 2008
Received in revised form 7 November 2008
Accepted 27 November 2008
Available online 9 December 2008
Keywords:
Gas sensor
Tungsten oxide nanorod
Metal oxide semiconductor
Thermal treatment
abstract
Room-temperature tungsten oxide sensors were prepared by using a solution containing single-
crystalline and monodispersed WO
2.72
nanorods with an average 75 nm length and 4 nm diameter.
Thermal treatment-dependent gas-sensing characteristics of the sensors were examined for achieving a
sensor with good performance. They were explained and discussed with their material properties probed
by SEM, XRD, XPS and Raman spectroscopy. Optimized thermal treatment was found to be an annealing
process at around 400
◦
C under the flow condition of inert N

2
or Ar gas. This treatment leads to the partial
oxidation of nonstoichiometric W
5+
states into the fully oxidative W
6+
without any noticeable change in
morphology or crystalline structure. These changes in material properties result in a great improvement
in detection and recovery times with only a slight sacrifice of detection response.
© 2008 Elsevier B.V. All rights reserved.
1. Introduction
Metal oxide semiconductors (MOS) have been utilized as gas-
sensing active materials for half a century [1,2]. One of the most
promising solid-state MOS chemosensors is a tungsten oxide-based
gas sensor. Several studies have proved that this sensor could be
used for the detection of nitrogen oxide (NO and NO
2
), ammonia
vapors, hydrogen sulfide, and hydrocarbons [3–7]. In the last few
years, the nanostructures of tungsten oxides have been found to
be more effective sensing materials due to their high surface-to-
volume ratio and small grain size [8–12]. They have demonstrated
novel sensing properties such as high sensitivity, fast response time,
and low operation temperature. These properties are unattainable
by using classical MOS materials consisting of submicrometer-sized
polycrystalline particles.
MOS sensors usually operate in the temperature range of
200–50 0
◦
C. This operation temperature results in high electrical

power consumption, which limits the use of MOS sensor as a sens-
ing element in battery-powered portable devices. To overcome
such a drawback, there have been many investigations of novel
one-dimensional (1-D) MOS nanostructure sensors able to oper-
ate at room temperature. These include carbon nanotubes, SnO
2
nanowires or nanobelts, In
2
O
3
nanowires, and WO
2.72
nanorods
[11–16]. Among these 1-D MOS, tungsten oxide are regarded as
an encouraging material for achieving low operation temperature
∗
Tel.: +82 31 400 5507; fax: +82 31 400 3949.
E-mail address:
since tungsten oxide has lower intergrain energy barrier than SnO
2
or TiO
2
[17,18].
In this work, we prepared tungsten oxide films by drop-coating
single-crystalline, size-controlled WO
2.72
nanorod solution, and
then investigated the dependence of their material and gas-sensing
properties on thermal treatment conditions. The thermal treat-
ments were performed with the variation of annealing temperature

in the range of 200–700
◦
C under the ambient conditions of nitro-
gen, Ar, oxygen or dry air. Our previous works reported that WO
2.72
nanorod sensors had highly sensitive detection capability even at
room temperature for various reducing and oxidizing compounds
[11,12]. The purpose of this study is to find thermal treatment con-
ditions that can optimize gas-sensing performance in a tungsten
oxide nanorod system. In addition, thermal treatment-dependent
gas-sensing characteristics are explained and discussed in terms of
observed material properties.
2. Experimental
Tungsten oxide nanorods with an average 4 nm diameter and
75 nm length were synthesized in massive quantity by the colloid-
based synthetic approach [19]. Diluted HCl solution was added into
the colloid solution to precipitate WO
2.72
nanorods stabilized by
amine-based surfactants. The resultant precipitates were separated
by a centrifuge, and further purified by dissolving in toluene and
consecutive centrifuging three times. Tungsten oxide films were
deposited by casting isopropyl alcohol solution of WO
2.72
nanorods.
The isopropyl alcohol solution was severely agitated to disperse
the nanorods homogenously just before the deposition process. As-
0925-4005/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.snb.2008.11.037
298 Y.S. Kim / Sensors and Actuators B 137 (2009) 297–304

deposited blue films were dried at 100
◦
C under vacuum for more
than 10 h to remove residual organic compounds. Additional ther-
mal treatments were performed for 3 h at a temperature i n the
range of 200–70 0
◦
C under the flow conditions of N
2
,O
2
,Arordry
air.
Tungsten oxide films prepared on Si were utilized to character-
ize their material properties through various methods of analysis.
Thickness was measured by using a mechanical profiler (Alpha-Step
500, KLA-Tencor), and surface microstructures were measured by
scanning electron microscopy (SEM; Sirion, FEI). X-ray diffraction
(XRD; D/MaxRC, Rigaku) having a Cu K␣ anode and Raman spectra
excited at 514 nm were utilized for probing a crystalline structure.
In addition, atomic constituents and chemical bonding states were
evaluated through X-ray photoelectron spectroscopy (XPS; SCALAB
200R, VG Scientific). Curve fittings of XPS spectra were performed
with the freeware program of XPSPEAK 4.1 using the sum of Gaus-
sian and Lorentzian functions as the modelfunctions. During fitting,
the spin-orbit splitting energy between W 4f
7/2
and W 4f
5/2
was

fixed at 2.1 eV. The fitting was started with the initial W 4f
7/2
bind-
ing energies of 35.5 and 34.3 eV for the W
6+
and W
5+
oxidation
states, respectively [20].
Tungsten oxide sensors were fabricated by using Au-patterned
glass substrates. The sensor substrate was prepared by the consec-
utive e-beam depositions of 5 nm Cr and 100 nm Au layers through
a shadow metal mask. It possesses two interdigitated Au elec-
trodes with a sensing area of 3 mm × 10 mm and an electrode gap of
300 ␮m. Sensing characteristics were measured by using the flow
injection system as previously described [21]. They were carried
out by placing a gas sensor in a detection chamber and blowing
analyte vapors over it with a flow rate of 1000 ml/min. Analyte con-
centrations were regulated with the relative flow ratio between
the carrier dry air and diluted analyte vapor. The sensing mea-
surements were performed upon exposure to the four different
analytes of ethanol, hexane, benzene, and ammonia. Measured ana-
log signals were delivered to the digital interface (NI, DAQ 6062E)
connected to a laptop PC through a shielded cable. Data acquisition
was performed by the user software programmed under LabVIEW
(NI) environments with a typical sampling rate of 5 Hz.
3. Results and discussion
3.1. Thermal treatment effects on the material properties
As-deposited tungsten oxide films were measured to have a
thickness of 3–7 ␮m by the mechanical profiler. Their surface mor-

phology observed by SEM displayed a porous appearance resulting
from randomly arranged agglomerates, favorably formed by par-
allel alignment of high anisotropic WO
2.72
nanorods [11]. This
appearance is almost identical to the SEM image of a N
2
-annealed
sample at 500
◦
C (see Fig. 1A), suggesting no noticeable change
in surface morphology. In fact, we were not able to observe any
microstructural change in SEM images of samples annealed at
below 500
◦
C in ambient N
2
conditions. The surface appearance
however begins to change significantly for N
2
-annealed samples
at above 550
◦
C (see Fig. 1B–D). In spite of the similarity in over-
all size, the surface appearance of a sample annealed at 550
◦
C
exhibits that tungsten oxide nanorod agglomerates are no longer
basic constitutional components. Instead, new larger, collapsed
crystalline particles have appeared. They seem to form through the

recrystallization of nanorod aggregates. As the anneal temperature
increases, the surface morphology reveals more dramatic changes.
Above600
◦
C, the newly born crystals must undergo severe arrange-
ment, resulting in a favorable columnar crystal growth along the
vertical direction. Moreover crystal and void sizes become larger
when the anneal temperature increases from 600 to 700
◦
C. The
average surface crystal size of N
2
-annealed sample at 700
◦
C finds
Fig. 1. Typical surface SEM images of tungsten oxide nanorod films annealed under the ambient nitrogen conditions at the four different temperatures of (A) 500
◦
C, (B)
550
◦
C, (C) 600
◦
C and (D) 700
◦
C.
Y.S. Kim / Sensors and Actuators B 137 (2009) 297–304 299
Fig. 2. Typical surface SEM images of tungsten oxide nanorod films annealed under
the ambient oxygen conditions at the two different temperatures of (A) 400
◦
C and

(B) 500
◦
C.
to be around 100 nm, which is two times larger than that of 600
◦
C.
Similar changes in surface morphology were also examined under
the thermal treatments with different ambient conditions of dry
air, O
2
or Ar. An identical result was observed in the Ar-annealing
condition: the onset of recrystallization is around 550
◦
C and the
columnar crystals grow larger when an anneal temperature reaches
above 600
◦
C. On the other hand, dry air and O
2
anneal treatments
result in a different onset temperature. Fig. 2A and B shows surface
SEM images of O
2
-annealed WO
2.72
films at 400 and 500
◦
C, respec-
tively. They distinctly demonstrate that crystal growth takes place
at a temperature between 400 and 500

◦
C. Since the average crystal
size of a 500
◦
CO
2
-annealing sample is observed to be larger than
that of 600
◦
CN
2
-annealed, oxygen environments must encour-
age the coalescence processes among the aggregated nanorods. An
identical enhancement in crystal growth was also observed for air-
annealed samples. The dependence of an onset temperature on the
ambient environments suggests that oxygen molecules can act as
a crystallization enhancer. Due to the nonstoichiometric nature of
WO
2.72
nanorods, oxygen infiltration might easily initiate the oxida-
tive recrystallization into the most stable WO
3
state. Even under
the flow condition of inert N
2
or Ar, the oxidation process seems to
proceed at higher temperature due to a small number of residual
oxygen-containing compounds within a furnace.
Fig. 3 shows the XRD patterns of the N
2

-annealed samples
shown in Fig. 1. The deflection pattern of 500
◦
C sample shown
in Fig. 3A displays no discernable change from the as-deposited
films, which reported broad background peaks due to the small
nano-sized dimension of nanorods and the very weak (0 1 0) mon-
oclinic peak at the diffraction angle of 23.3
◦
assigned to the growth
direction [12]. The asterisk-designated peak at around 33
◦
comes
from a used Si substrate. However, there are many new peaks
for the samples annealed at over 550
◦
C, indicating the forma-
Fig. 3. X-ray deflection spectra of the same tungsten oxide samples displayed in
Fig. 1. They were annealed under the ambient nitrogen condition at the four different
temperatures of (A) 500
◦
C, (B) 550
◦
C, (C) 600
◦
C and (D) 700
◦
C. The Miller indices
are determined from the monoclinic structure (JCPDS no. 05-0363), and the peak
indicated with an asterisk results from a used Si substrate.

tion of new crystalline phases. These peaks have higher intensity
and narrower FWHM as temperature increases from 550 to 700
◦
C.
It is well coincident with previous observations in SEM that the
higher temperature-annealed sample shows a larger crystal struc-
ture. Stoichiometric WO
3
crystals have known to have monoclinic
or triclinic structures at room temperature [22]. The most inten-
sive three diffraction peaks at the angle of 22–25
◦
correspond to
the pseudo-cubic reflections originating from the slight distortion
of the ideal cubic {100} lattice planes. These three reflections can-
not be used solely to determine whether the crystalline structure
is triclinic or monoclinic, since their position and relative inten-
sity is very similar. On the other hand, the intensity distribution
of the diffraction peaks in the range of 32–35
◦
can provide a clue
to distinguish between triclinic and monoclinic structures [23]. Our
intensity distributions were found to agree better with that of mon-
oclinic WO
3
, so that the newborn peaks were assigned based on
monoclinic structure (JCPDS no. 05-0363).
The temperature-dependent crystalline changes were also
investigated for the samples annealed under dif ferent ambient con-
ditions. All observed XRD peaks were well interpreted with the

monoclinic WO
3
structure even though there is a little difference in
preferential crystal structure. They exhibited an identical tendency
as observed in SEM: N
2
and Ar anneal treatments displayed the
same temperature dependence while the dry air and O
2
environ-
ments induced the recrystallization at the lower onset temperature
of less than 500
◦
C.
Raman shift spectra are shown in Fig. 4A and B for the films
annealed under ambient N
2
and O
2
environments, respectively.
Each consists of two spectra obtained at two different anneal
temperatures, which correspond to before and after the onset of
an extensive crystal growth. The two spectra obtained at 400
◦
C
observed to be similar to that of as-deposited sample [12]. All the
spectra display four distinct bands at 270, 327, 712, and 807 cm
−1
except for the lattice mode band below 200 cm
−1

and a small
instrumental artifact around 520 cm
−1
. The four bands fall very
300 Y.S. Kim / Sensors and Actuators B 137 (2009) 297–304
Fig. 4. Raman shift spectra of tungsten oxide films annealed at two different tem-
peratures under the ambient conditions of (A) nitrogen and (B) oxygen. They were
observed with an Ar-ion laser excitation at 514 nm.
close to the wavenumbers of the strongest modes of monoclinic
tungsten oxide. The two bands at 270 and 327 cm
−1
had been
assigned to O
W O bending modesof the bridging oxide, while the
712 and 807 cm
−1
bands are the corresponding stretching modes
[24,25]. Moreover, the band intensities of the samples annealed
at higher temperature increase by more than one order in mag-
nitude compared with those at lower temperature. Consequently,
these results give further evidence that the monoclinic crystals
are grown and become larger as the anneal temperature goes
higher.
Main constituent elements of as-deposited films were observed
to be tungsten and oxygen atoms from wide-scan XPS measure-
ments, except for additional minor peaks resulting from carbon and
Si elements. The appearance of Si 2s and 2p peaks can be explained
with photoelectrons ejected from a Si substrate due to the highly
porous nature of nanorod films. Small amount of carbon impuri-
ties were confirmed to exist within the film probably due to the

incorporation of carbon atoms originated from carbon-containing
chemicals used in the synthesisprocesses, together with a consider-
able amount of surface-adsorbed carbon moieties. However, there
was no other discernable impurity except for the carbon atoms.
Narrow-scan XPS measurements were performed to quantify
the amount of carbon impurities in the binding energy range of
240–295 eV. Fig. 5 shows spectra obtained from the samples with
different thermal treatments: (A) no annealing, (B) N
2
annealing
at 400
◦
C, (C) N
2
annealing at 600
◦
C, (D) O
2
annealing at 300
◦
C
and (E) O
2
annealing at 500
◦
C. In the case of N
2
annealing treat-
ment, the intensity of C 1s peak begins to decrease considerably at
600

◦
C under the condition of maintaining the W 4d peak inten-
sities (see Fig. 5A–C). It implies that carbon contaminants are not
removed noticeably for theN
2
400
◦
C annealing while they decrease
by about one third of as-deposited films for the 600
◦
C annealing.
A similar decrease in carbon content was also observed in the O
2
annealing treatments: the C 1s intensity decreases with ascend-
ing O
2
annealing temperature (see Fig. 5D and E). Judging from the
slight decrease in C 1s intensity even for the sample annealed at
Fig. 5. XPS spectra of tungsten oxide samples in the binding energy range of
240–295 eV. They have undergone dif ferent thermal treatments: (A) no annealing,
(B) N
2
annealing at 400
◦
C, (C) N
2
annealing at 600
◦
C, (D) O
2

annealing at 300
◦
C and
(E) O
2
annealing at 500
◦
C.
300
◦
C, carbon-containing impurities could be favorably eliminated
in ambient O
2
conditions, probably through oxidative reaction
pathways.
The W 4f XPS spectrum of an as-deposited blue WO
2.72
film was
previously reported to decompose into two components resulting
from W
5+
and W
6+
oxidation states [12]. Since the nanorod has
crystallographic shear planes relative to a ReO
3
-type structure, the
cations will have the W
5+
formal oxidation state at shear plane

boundaries in which a WO
6
octahedron has three corner-sharing
and two edge-sharing octahedral neighbors. Each component con-
sists of W 4f
7/2
and W 4f
5/2
doublet peaks with the spin-orbit
splitting energy of 2.1 eV. Fig. 6A and B shows W 4f XPS spectra
of N
2
annealed samples at the two temperatures of 400 and 600
◦
C,
respectively. The two dashed lines correspond to best-fitted curves
for the W
6+
and W
5+
oxidation states while the solid line displays a
sum spectrum of the two components. The sum spectrum displays
fairly good agreement with the raw data displayed by open circles.
In addition, the W 4f XPS spectra obtained from 300 and 500
◦
CO
2
annealed samples are shown in Fig. 7A and B, respectively. These
spectra are also well interpreted with the two components corre-
sponding to W

6+
and W
5+
oxidation states. Table 1 summarizes the
parameters use d in the deconvolution of W 4f XPS peaks, i.e. the
binding energy of the W 4f
7/2
, FWHM and relative contribution of
the two oxidation states. All the samples exhibit identical binding
energies of around 35.7 and 34.2 eV for the W
6+
and W
5+
oxida-
tion states, respectively, which are well consistent with the values
in the literatures [20,26,27]. This observation suggests that chemi-
cal bonding characteristics of the two oxidation states do not alter
noticeably according to the thermal treatments. However, the rela-
tive W
5+
population steadily decreases as N
2
annealing temperature
increases: 19% for the as-deposited sample, 12% at 400
◦
C, and 6% at
600
◦
C. It must be attributed to the oxidation of partially-reduced
W

5+
states into fully oxidized stable W
6+
states. This interpreta-
tion is consistent with the SEM and XRD results indicating the
extensive monoclinic WO
3
crystal growth at above 600
◦
C. It is
Y.S. Kim / Sensors and Actuators B 137 (2009) 297–304 301
Table 1
Parameters used in the deconvolution of W 4f XPS peaks into the two possible oxidation state components.
Annealing conditions W
6+
component W
5+
component
BE
a
(eV) FWHM
a
(eV) Population (%) BE
a
(eV) FWHM
a
(eV) Population (%)
None
b
35.7 1.9 81 34.2 1.4 19

N
2
,400
◦
C 35.7 1.7 88 34.3 1.5 12
N
2
,600
◦
C 35.7 1.6 94 34.2 1.2 6
O
2
,300
◦
C 35.8 1.9 85 34.3 1.4 15
O
2
,500
◦
C 35.7 1.6 94 34.3 1.4 6
a
BE and FWHM are abbreviations for binding energy and full width at half maximum of the main W 4f
7/2
peak, respectively.
b
These results were obtained from our previous work [12].
worth mentioning that the oxidation significantly occurs for the
400
◦
CN

2
-annealing sample with the same surface morphology
of as-deposited films. It suggests that the nanorod elements first
undergo partial oxidation at relatively low temperature and then
agglomerate to form a larger crystal above the onset temperature.
Furthermore, an identical trend is observed for the O
2
-annealing
samples: the relative W
5+
populations are 15% at 300
◦
C and 6% at
500
◦
C. Judging from the relative W
5+
populations, the oxidation
process can be activated more easily in the O
2
-containing ambient
conditions.
Fig. 6. Narrow-scan W 4f XPS spectra of tungsten oxide samples N
2
-annealed at the
two different temperatures of (A) 400
◦
C and (B) 600
◦
C. The open circle symbols

show raw data, and the solid line corresponds to a sum spectrum of the two dashed
lines which are best-fitted curves for the W
6+
and W
5+
oxidation states.
3.2. Thermal treatment effects on the gas-sensing properties
Gas-sensing measurements were performed for WO
2.72
nanorod
sensors anneale d at different temperatures under ambient N
2
con-
ditions. Fig. 8 shows annealing temperature-dependent sensor
response profiles at room temperature as a function of a detec-
tion time. The sensors were exposed to air-diluted ethanol vapors
of 1000 ppm with the concentration variation of step function. The
lowest plot in Fig. 8 displays the expected time profile of ethanol
concentration for the sequence of 5 min injection and 20 min recov-
ery period.In our detection chamber, the elapsed timesof filling and
removing ethanol vapors were observed to be less than 10 s from
Fig. 7. Narrow-scan W 4f XPS spectra of tungsten oxide samples O
2
-annealed at the
two different temperatures of (A) 300
◦
C and (B) 500
◦
C. The open circle symbols
show raw data, and the solid line corresponds to a sum spectrum of the two dashed

lines which are best-fitted curves for the W
6+
and W
5+
oxidation states.
302 Y.S. Kim / Sensors and Actuators B 137 (2009) 297–304
Fig. 8. Time-dependent response profiles of WO
2.72
nanorod sensors having the dif-
ferent N
2
anneal temperatures in the range of 300–700
◦
C. No annealed sample is
also given for comparison. All sensors were measured under the condition of flow-
injecting air-diluted ethanol vapors of 1000 ppm as presented in the lowest curve.
The vertical arrow corresponds to the magnitude of 10 in sensor response.
separated measurements using a fast response sensor. The length
of the vertical arrow corresponds to the magnitude of 10 in sen-
sor response, which was defined by a relative percentage change
of a sensor resistance R with respect to a stabilized initial value
R
0
, i.e. response =100 × (R–R
0
)/R
0
=100× R/R
0
. The response pro-

files exhibit an increase change in resistance upon exposure to
the reducing ethanol vapors, corresponding to p-type response.
This phenomenon was previously explained with the competi-
tion adsorption between ambient molecular oxygen and analyte
vapors on the surface of WO
2.72
nanorods instead of the conven-
tional mechanism observed at above 200
◦
C in MOS sensor, i.e.
the reaction between ionosorbed oxygen moieties and analytes
[11]. The adsorbed analyte molecules may act as active scatter-
ing centers, thus suppressing the electrical conductance of free
electron carriers in n-type tungsten oxide system. This scattering
results in the resistance increase for both oxidizing and reducing
analytes.
The response of a no-annealed sample steadily increases for
the ethanol injection period and eventually reaches to a maximum
response of 9.1. It slowly decreases into the initial position for a fol-
lowing recovery time of 20 min. The recovery time of no-annealed
sensor was observed to be around 30 min which is too long to use
as a practical MOS gas sensor. Such a long recovery time was previ-
ously observed in other MOS sensor systems operated at ambient
temperature [13–15]. It might be attributed to the low opera-
tion temperature leading to slow desorption rate of pre-adsorbe d
moieties. As an annealing temperature increases until 500
◦
C, the
response profile becomes closer to a square shape, indicating that
the sensor has fast response and short recovery time. In the case

ofa400
◦
C annealed sensor, the response abruptly increases to the
90% level of a maximum value within a time of less than 1 min, and
the recovery time becomes two times shorter compared with the
no-annealed sample. However, there is a slight decrease in maxi-
mum response: a maximum response of the 400
◦
C annealed sensor
corresponds to 85% of the no-annealed one. As a result, N
2
anneal
treatment below 500
◦
C leads to a great improvement in response
and recovery times with a slight sacrifice of response. This behavior
might be explained in the light of the observed material character-
istics as stated below. The material properties of N
2
-annealed films
at below 500
◦
C have been confirmed to be identical to those of
as-deposited samples except for the population difference in W
6+
and W
5+
oxidation states. Therefore, the temperature-dependent
response variation might result from the difference in sensing abil-
ity of the two oxidation states. Fully oxidized WO

3
is the most
stable tungsten oxide compound so that the W
6+
state may be
more inert to the sensing-related surface reactions than the W
5+
state. The slight decrease in response with the increase in the
anneal temperature can be interpreted as the local oxidation of
more reactive W
5+
states to W
6+
as probed by XPS. Furthermore,
the W
5+
states have a stronger interaction with analyte molecules
so that the long recovery time of as-deposited sensor can be also
understood with the high W
5+
population. Consequently, N
2
ther-
mal treatment of WO
2.72
nanorod films below 500
◦
C results in
the partial oxidation of nonstoichiometric W
5+

states without any
noticeable change in morphology or crystalline structure, which
gives a chance to modulate systematically gas-sensing characteris-
tics such as response magnitude and detection time by means of
careful control of the relative population ratio between W
5+
and
W
6+
states.
The sensing characteristics become suddenly worse as the N
2
annealing temperature increases beyond 600
◦
C at which point
the nanorods begin to coalesce and develop large, columnar WO
3
crystals. Considering that porous film made of small crystalline
nanorods with a high aspect ratio are favorable for achieving a high
sensitive MOS sensor, the deterioration in sensing capability can be
explained by the formation of larger crystals with a columnar crack.
Response profiles were also evaluated for the sensors annealed
in ambient O
2
conditions at a temperature range of 30 0–600
◦
C.
On the whole, O
2
-annealed sensors had exhibited bad sensing

properties. The most serious problem was their weak operation
durability: the O
2
-annealed sensors easily broke down in spite of
material properties similar to N
2
-annealed analogues. This propen-
sity was also observed for the sensors annealed in ambient air
conditions. The thermal treatment under the O
2
-containing ambi-
ence therefore gives adverse effects on gas-sensing characteristics.
Even though there has been no direct clue to explain this effect,
it seems to be ascribed to delicate differences in microstructures
induced by O
2
thermal treatment.
Additional measurements were performed at room tempera-
ture upon exposure to other volatile vapors for evaluating the
sensing characteristics of our nanorod sensors in terms of sensi-
tivity and selectivity. Fig. 9 shows time-dependent sensor response
profiles of 400 and 600
◦
CN
2
-annealing samples upon exposure
to the three different analytes of 1000 ppm hexane, 1000 ppm
benzene, and 10 ppm NH
3
. The bottom square pulses display the

time-dependent profiles of supplying air-diluted analyte vapors.
The middle profiles were obtained from the sensor annealed at
400
◦
C while the uppers correspond to the sample annealed at
600
◦
C. The upper curves were shifted upward with respect to
the middle ones in order to distinguish easily between them. The
scale in sensor response is, however, identical for the two cases
and the vertical arrow size matches up to the magnitude of 2.
The relative detection magnitudes in sensor response exhibit great
differences for the three different analytes: the 400
◦
C sensor is pos-
sible to achieve feasible detection of hexane and benzene while
the 600
◦
C sensor can detect NH
3
vapors more sensitively. In fact,
the detection responses of the 400
◦
C sensor are fond to be 1.2 for
hexane and 2.1 for benzene. These values are much larger com-
pared with the 600
◦
C response of <0.1. On the other hand the
NH
3

detection response of 600
◦
C sensor is about six times larger
Y.S. Kim / Sensors and Actuators B 137 (2009) 297–304 303
Fig. 9. Time-dependent response profiles of WO
2.72
nanorod sensors for the three
different analyte exposures of 1000 ppm hexane, 1000 ppm benzene, and 10 ppm
NH
3
. The top profiles are sensor responses of the 600
◦
CN
2
-annealing sample while
the middles correspond to those of the 400
◦
CN
2
-annealing. The bottom square
pulses exhibit the time profiles of supplying three different analytes. The vertical
arrow corresponds to the magnitude of 2 in sensor response.
than that of 400
◦
C. As a result these response variation according
to anneal temperatures implies that carefully controlled thermal
treatments could provide a chance to improve detection selectivity
which is regarded as obstacles in a tungsten oxide nanorod sys-
tem for achieving a good performance sensor operated at room
temperature.

4. Conclusion
We have investigated the dependence of gas-sensing character-
istics on thermal treatment conditions in a tungsten oxide nanorod
system that demonstrated the facile detection of various analytes at
ambient temperature. As the annealing temperature increases, the
surface morphology of as-deposited films consisting of randomly
arranged WO
2.72
nanorod agglomerates were observed to develop
larger WO
3
particles with a columnar monoclinic crystal structures,
probably through a recrystallization process of nanorod aggregates.
The onset temperatures for the recrystallization were found to
be around 550
◦
C for the inert N
2
or Ar annealing conditions and
450
◦
C for the O
2
-containing ambient, which suggests that oxygen
molecules act as a crystallization enhancer. Since such a crystal-
lization process leads to large crystals and many cracks between
grain boundaries, the thermal treatments above the onset temper-
ature were found to deteriorate the detection capability of WO
2.72
sensors. In addition, thermal treatments under the O

2
-containing
active environments were found to result in bad reproducibility in
sensor response compared with the inert N
2
conditions. As a result,
the recommendable thermal treatment conditions for WO
2.72
sen-
sors were an annealing temperature of 300–500
◦
C under inert N
2
or Ar ambient. The annealing temperature within the range could
be utilized as a parameter to regulate a relative population ratio
between W
5+
and W
6+
states without any noticeable change in mor-
phology or crystalline structure. We can have an opportunity to
optimize gas-sensing properties such as detection selectivity and
a recovery time through the careful control of relative populations
between the two oxidation states.
Acknowledgement
This work was supported by the research funds of KOSEF (R01-
2008-0 00-20460-0) and SMBA (S708002511 and S6070381211).
References
[1] N. Barsan, U. Weimar, Conduction model of metal oxide gas sensors, J. Electro-
ceram. 7 (2001) 143–167.

[2] Y. Shimizu, M. Egashira, Basic aspects and challenges of semiconductor gas
sensors, MRS Bull. 24 (1999) 18–24.
[3] L.G. Teoh, I.M. Hung, J. Shieh, W.H. Lai, M.H. Hon, High sensitivity semiconductor
NO
2
gas sensor based on mesoporous WO
3
thin film, Electrochem. Solid-State
Lett. 6 (2003) G108–G111.
[4] C. Cantalini, H.T. Sun, M. Faccio, M. Pelino, S. Santucci, L. Lozzi, M.Passacantando,
NO
2
sensitivity of WO
3
thin film obtained by high vacuum thermal evaporation,
Sens. Actuators B-Chem. 31 (1996) 81–87.
[5] E. Llobet, G. Molas, P. Molinàs, J. Calderer, X. Vilanova, J. Brezmes, J.E. Sueiras,
X. Correig, Fabrication of highly selective tungsten oxide ammonia sensors, J.
Electrochem. Soc. 147 (2000) 776–779.
[6] C. Bittencourt, R. Landers, E. Llobet, G. Molas, X. Correig, M.A.P. Silva, J.E. Sueiras,
J. Calderer, Effects of oxygen partial pressure and annealing temperature on the
formation of sputtered tungsten oxide films, J. Electrochem. Soc. 149 (2002)
H81–H86.
[7] R. Ionescu, A. Hoel, C.G. Granqvist, E. Llobet, P. Heszler, Low-level detection of
ethanol and H
2
S with temperature-modulated WO
3
nanoparticle gas sensors,
Sens. Actuators B-Chem. 104 (2005) 132–139.

[8] J. Polleux, A. Gurlo, N. Barsan, U. Weimar, M. Antonietti, M. Niederberger,
Template-free synthesis and assembly of single-crystalline tungsten oxide
nanowires and their gas-sensing properties, Angew. Chem. Int. Ed. 45 (2006)
261–265.
[9] C.S. Rout, M. Hegde, C.N.R. Rao, H
2
S sensors based on tungsten oxide nanos-
tructures, Sens. Actuators B-Chem. 128 (2008) 488–493.
[10] C. Balázsi, L. Wang, E.O. Zayim, I.M. Szilágyi, K. Sedlacková, J. Pfeifer, A.L. Tóth,
P I. Gouma, Nanosize hexagonal tungsten oxide for gas sensing applications, J.
Eur. Ceram. Soc. 28 (2008) 913–917.
[11] Y.S. Kim, S C. Ha, K. Kim, H. Yang, S Y. Choi, Y.T. Kim, J.T. Park, C.H. Lee, J.
Choi, J. Paek, K. Lee, Room temperature semiconductor gas sensor based on
nonstoichiometric tungsten oxide nanorod film, Appl. Phys. Lett. 86 (2005)
213105.
[12] Y. S. Kim, K. Lee, Material and gas-sensing properties of tungsten oxide nanorod
thin-films, J. Nanosci. Nanotechnol. doi:10.116 6/jnn.2008.SE32.
[13] J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, M. Meyyappan, Carbon nanotube sensors for
gas and organic vapor detection, Nano Lett. 3 (2003) 929–933.
[14] M. Law, H. Kind, B. Messer, F. Kim, P. Yang, Photochemical sensing of NO
2
with
SnO
2
nanoribbon nanosensors at room temperature, Angew. Chem. Int. Ed. 41
(2002) 2405–2408.
[15] Y. Wang, X. Jiang, Y. Xia, A solution-phase precursor route to polycrystalline
SnO
2
nanowires that can be used for gas sensing under ambient conditions, J.

Am. Chem. Soc. 125 (2003) 16176–16178.
[16] D. Zhang, Z. Liu, C. Li, T. Tang, X. Liu, S. Han, B. Lei, C. Zhou, Detection of NO
2
down to ppb levels using individual and multiple In
2
O
3
nanowire devices, Nano
Lett. 4 (2004) 1919–1924.
[17] C. Malagù, V. Guidi, M.C. Carotta, G. Martinelli, Unpinning of Fermi
level in nanocr ystalline semiconductors, Appl. Phys. Lett. 84 (2004)
4158–4160.
[18] C. Malagù, M.C. Carotta, S. Gherardi, V. Guidi, B. Vendemiati, G. Martinelli, AC
measurements and modeling of WO
3
thick film gas sensors, Sens. Actuators
B-Chem. 108 (2005) 70–74.
[19] K. Lee, W.S. Seo, J.T. Park, Synthesis and optical properties of colloidal tungsten
oxide nanorods, J. Am. Chem. Soc. 125 (2003) 3408–3409.
[20] D. Barreca, S. Bozza, G. Carta, G. Rossetto, E. Tondello, P. Zanella, Structural and
Morphological analyses of tungsten oxide nanophasic thin films obtained by
MOCVD, Surf. Sci. 532 (2003) 439–443.
[21] Y.S. Kim, S C. Ha, H. Yang, Y.T. Kim, Gas sensor measurement system capable
of sampling volatile organic compounds (VOCs) in wide concentration range,
Sens. Actuators B-Chem. 122 (2007) 211–218.
[22] P. Woodward, A. Sleight, T. Vogt, Structure refinement of triclinic tungsten tri-
oxide, J. Phys. Chem. Solids 56 (1995) 1305–1315.
[23] E. Cazzanelli, C. Vinegoni, G. Mariotto, A. Kuzmin, J. Purans, Low temperature
polymorphism in tungsten trioxide powders and its dependence onmechanical
treatments, J. Solid State Chem. 143 (1999) 24–32.

[24] M.F. Daniel, B. Desbat, J.C. Lassegues, B. Gerand, M. Figlarz, Infrared, Raman
study of WO
3
tungsten trioxides and WO
3
·xH
2
O tungsten trioxide hydrates, J.
Solid State Chem. 67 (1987) 235–247.
[25] C. Santato, M. Odziemkowski, M. Ulmann, J. Augustynski, Crystallographically
oriented mesoporous WO
3
films: synthesis, characterization, and applications,
J. Am. Chem. Soc. 123 (2001) 10639–10649.
[26] .
[27] S.C. Moulzolf, S. Ding, R.J. Lad, Stoichiometry and microstructure effects
on tungsten oxide chemiresistive films, Sens. Actuators B-Chem. 77 (2001)
375–382.
304 Y.S. Kim / Sensors and Actuators B 137 (2009) 297–304
Biography
Yong Shin Kim received a PhD in chemistry from Korea Advanced Institute of Sci-
ence and Technology (KAIST) in 1997. After his degree, he had worked as a senior
research member at Electronics and Telecommunications Research Institute in the
field of developing flat panel displays, i.e. thin-film electroluminescent devices and
miniaturized electronic nose system. Since 2007, he has been employed in Hanyang
University. Now his research activities are focused on the development of minia-
turized, smart chemical sensor systems and novel nanomaterials for the sensor
applications.