Sie sind auf Seite 1von 6

Sensors and Actuators B 114 (2006) 507–512

An ozone monitoring instrument based on the tungsten


trioxide (WO3) semiconductor
S.R. Utembe a,∗ , G.M. Hansford a , M.G. Sanderson b , R.A. Freshwater a ,
K.F.E. Pratt c , D.E. Williams d , R.A. Cox a , R.L. Jones a
a Chemistry Department, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, UK
b Met Office, Fitzroy Road, Exeter EX13PB, UK
c Department of Chemistry, University College London, 20 Gordon Street, London WC1H 0AJ, UK
d Unipath Ltd., Priory Business Park, Bedford MK44 3UP, UK

Received 1 February 2005; received in revised form 11 April 2005; accepted 11 April 2005
Available online 10 August 2005

Abstract

Novel sensors based on tungsten trioxide (WO3 ) semiconductors have been found to hold much promise as a cheaper alternative for ozone
monitoring. The sensitivity of the metal oxide to ozone is owed to the presence of surface oxygen vacancies. The filling of oxygen vacancies
by the decomposing ozone traps free charge carriers with consequent measurable increase in resistance. A Langmuir-type model simulating
the fundamental surface reactions has been found to reproduce the measured sensor response. Using a temperature-cycling technique in which
the temperature of the sensor is alternated between 600 and 530 ◦ C representing scrubbing and measurement resistances, respectively, ozone
measurements have been obtained from several urban sites. Results from eight sensors show that the sensors have adequate sensitivity and
response time with measurements comparable to ozone measured by UV absorption spectrometer (r.m.s. of variances up to 7 ppbv). Problems
of apparent changes in sensor sensitivity in the course of field measurements have been explained by inadequate ‘burn-in’ period prior to
calibration and field deployment. The sensors should be run in ambient conditions for about a week before calibration.
© 2005 Elsevier B.V. All rights reserved.

Keywords: WO3 ; Ozone; Temperature cycling; Pre-measurement calibration; Post-measurement calibration

1. Introduction sensor based on the WO3 semiconductor has been found


to exhibit excellent sensitivity and selectivity [3]. Recent
The monitoring of ground-based ozone is normally done research in the use of tungstic trioxide semiconductor has
by using spectroscopic methods. While these techniques have focused on characterising its detailed response to the pres-
been found to be stable and reliable, their applicability is ence of ozone [4]. In this paper, we present a comparison of
somewhat limited by the fact that the instruments are usually the performance of the WO3 sensor instrument with a con-
both expensive and bulky. There is now increased interest ventional UV technique.
in alternative cheaper and more portable instruments [1]. In
this paper, we present results from a portable and lightweight
ozone monitoring instrument which utilises changes in the 2. Characterisation and understanding of the sensor
well established electronic properties of WO3 [2]. An ozone response


There has been good progress in establishing whether
Corresponding author at: Department of Environmental Science and
Technology, Imperial College, Silwood Park, Ascot, Berkshire SL5 7PY,
the response of the WO3 sensor can be characterised
UK. Tel.: +44 207 5942466; fax: +44 207 5942308. quantitatively in terms of fundamental reactions at the
E-mail address: s.utembe@imperial.ac.uk (S.R. Utembe). gas/semiconductor interface. The basic model that emerges

0925-4005/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.snb.2005.04.049
508 S.R. Utembe et al. / Sensors and Actuators B 114 (2006) 507–512

is that the conductivity is controlled by the (temperature-


dependent) conductivity of the bulk WO3 , but also by oxygen
vacancies which can form at the interface. Vacancies are gen-
erated thermally (1) and can be destroyed by reaction with
both O3 and O2 (via (2) and (3), respectively) as shown by
reactions below, where V represents a vacancy:
k−1
Oo −→V + 0.5O2 (1)
k1
V + O3 −→Oo + O2 (2)
k2
V + O2 −→Oo + 0.5O2 (3)
Under steady state conditions (constant O2 and O3 ), the
coverage of surface vacancies can be represented using the
Langmuir isotherm [5]. However of more relevance to the
detection of ambient O3 concentrations is the time-dependent Fig. 2. Comparison of model (solid lines) and observed sensor response to
sensor response. Considering only O3 and assuming zero 43 ppb ozone at operating temperatures of 500 and 530 ◦ C shows differ-
initial coverage, casting the Langmuir isotherm in a time ent sensitivities and sensor responses. The observed sensor resistance is a
dependent manner leads to the following generic expression response following a switch in temperature from 600 ◦ C (scrubbing phase)
to the operating (measurement phase) temperatures of 500 and 530 ◦ C while
for surface vacancy coverage, θv (t), following the introduc-
exposing the sensors to air samples of constant ozone concentrations.
tion of an O3 concentration [O3 ]:
k−1 + k1 [O3 ]e−(k−1 +k1 [O3 ])t microstructure on the response characteristics: the interior of
θv (t) = (4) the tungsten oxide crystallites that comprise the sensor acts as
k−1 + k1 [O3 ]
a gas-insensitive shunt that damps the response [6]. In Fig. 1
The time dependent resistance of the device is then given is shown a comparison of the model (solid) with measure-
by: ments (symbols) using a WO3 sensor operated at 530 ◦ C for
Ω(t) = (Ci + 2θv (t))−1 (5) O3 concentrations between 13 and 68 ppbv, representative of
typical surface ozone concentrations, using each rate constant
where Ci is the (temperature dependent) intrinsic carrier con- k1 , k−1 and k3 as a global adjustable parameter independent
centration. This formalism can be readily expanded to include of O3 concentration.
the effects of O2 by inclusion of additional terms of the form The model reproduces well the effects of different O3 con-
k2 [O2 ] in Eq. (4). centrations measured at a sensor operating temperature of
In practice, to reproduce in detail the observed response, 530 ◦ C. Lowering the operating temperature (for example to
which shows a slow response superposed on a more rapid 500 ◦ C, see Figs. 2 and 3) has the effect of increasing sensitiv-
one (see Fig. 1), it is necessary to include two distinct rate ity although the response time is now too slow (∼300 s). The
constants. The effect has its origin in the effects of sensor reasons for the apparent changes in sensitivity and response
time with operating temperature are fully explained in [3]

Fig. 1. Comparison of model and observed sensor response at 530 ◦ C. Model


is represented by solid lines (see text for details). The observed sensor resis-
tance is a response following a switch in temperature from 600 ◦ C (scrubbing
phase) to the operating (measurement phase) temperature of 530 ◦ C while Fig. 3. Sensor resistance (at different operating temperatures) against ozone
exposing the sensors to air samples of constant ozone concentrations. shows increased sensitivity for the lower temperature of 500 ◦ C.
S.R. Utembe et al. / Sensors and Actuators B 114 (2006) 507–512 509

and [4]. These measurements allowed a WO3 band gap of


−200 kJ/mol to be determined, comparable to literature val-
ues (264 kJ/mol, see [7]).

3. Instrument design

The design goal was to develop a small, lightweight and


weatherproof instrument capable of making frequent ground-
level measurements with minimal regular maintenance. The
instrument consists of an air sampling line and control elec-
tronics housed in a tough plastic/aluminium box, conforming Fig. 4. Picture of the sensing element (WO3 ) deposited on a gold electrode
structure (used by permission of City Technology Ltd., UK).
to the IP66 protection standard and to EMC regulations [8].
Each instrument weighs approximately 1.5 kg and has dimen-
sions 25 cm × 18 cm × 9 cm. Two ozone sensors are used 3.2. Temperature cycling
within each instrument to improve reliability.
The flow line was made up of Teflon tubing and nylon The two-temperature mode of operation has been found
fittings to minimise the loss of ozone on exposed surfaces, to be the best way of operating the sensor. In this mode, the
and an inverted funnel was attached to the inlet to prevent operating temperature of the sensor is cycled between a high
ingress of rain droplets. Air was drawn through the line by temperature of 600 ◦ C and a low temperature of 530 ◦ C repre-
a miniature rotary pump (ASF Thomas model G12/02-4E) senting ‘scrubbing’ and ‘measurement’ phases, respectively.
at a rate of approximately 3 lpm. The flow was split in two During the high temperature phase (at which the sensitivity
and passed over the two sensors, each mounted in a Teflon of the sensor to ozone is insignificant), the surface of the sen-
chamber, before recombining and flowing over a thermistor sor is ‘reset’ in readiness for the subsequent low temperature
and capacitative humidity sensor, also mounted in a Teflon phase during which the ozone measurement is conducted.
housing. The temperature and humidity measurements were The observed sensor resistance (shown in Figs. 1 and 2) is
made when the sensor heaters were off to minimise interfer- a response following a switch in temperature from 600 ◦ C
ence. The pump was positioned downstream of the sensors, (scrubbing phase) to the operating (measurement phase) tem-
and the exhaust air was vented on the opposite side of the box peratures of 500 ◦ C or 530 ◦ C while exposing the sensors to
to the inlet. air samples of constant ozone concentrations.
The details of the design of the ozone sensor control elec-
tronics are very similar to those published before [9]. One 3.3. Calibration and use of the WO3 sensor
difference is that potentiometers were used in the Wheatstone
bridge. The temperature sensitivity of these components is The calibration procedure for the sensors has been done
less important for ground-based instruments, which experi- by flowing given constant levels of ozone over the sensors
ence much smaller temperature ranges than the balloon-borne and recording the resistance at 2 min in the low temperature
devices. Signal voltages were digitised with a Burr-Brown measurement phase. The ozone used to calibrate the sensors
ADS7825 a/d-convertor, and a TDS2020 Triangle Digital was generated by a Dasibi Multi-Gas Calibrator (series 5008)
Services embedded computer was used for data logging and manufactured by Dasibi Environmental Corporation Ltd. The
control. For a typical logging cycle in which one measure- calibration was repeated three times for each ozone concen-
ment is made every 10 min, there is sufficient onboard flash tration.
memory for unattended operation up to one month. In the
event of a power failure, the instrument will automatically
restart when power is restored. 4. Comparison of ozone measured by WO3 sensor
and UV absorption spectrometer
3.1. The WO3 sensor
The WO3 sensor has now been used by a Cambridge Uni-
The sensing element (City Technology Ltd., UK) con- versity (UK) research group in a wide range of studies in the
sists of a WO3 film mounted on one side of a 2 mm × 2 mm, urban planetary boundary layer.
250 ␮m thick alumina substrate. A platinum heater and gold Fig. 5 is a correlation plot showing typical performance of
electrode are screen-printed onto the substrate. The platinum the WO3 sensor compared to a commercial UV absorption
element is used for heating the sensor to a specified elec- spectrometer operated in such a mode. The measurements
tronically controlled temperature since the sensor operates were made for over 30 days at a site in Sandwell (UK).
at temperatures above ambient conditions. A picture of the The level of agreement obtained, with an r.m.s. deviation of
WO3 sensing element deposited on a gold electrode structure around 5 ppbv, is excellent, illustrating clearly the potential
is shown in Fig. 4. of the WO3 sensor methodology.
510 S.R. Utembe et al. / Sensors and Actuators B 114 (2006) 507–512

Fig. 5. Comparison of performance of a WO3 sensor with UV absorption Fig. 7. Comparison of performance of eight WO3 sensors with UV absorp-
spectrometer for ozone measurements that were collected for a period of tion spectrometer for measurements that were collected for a period of over
over 30 days in urban conditions at Sandwell (UK). 30 days in urban conditions in the West Midlands (Central UK).

Fig. 6 shows the corresponding time series plotted along better than another is not obvious, but it may be caused by
with the differences in ozone measured by the two techniques. some non-uniformity in the sensor production process. These
Clearly, the WO3 sensor is responding well to the ozone struc- sensitivity and scrubbing-phase resistance changes can be
ture caused by diurnal variations. corrected for by conducting another sensor calibration after
A more general assessment of the WO3 sensor is shown the field measurement period.
in a correlation plot of data collected from eight sensors from
various urban sites (Fig. 7). The data is over a 30-day period 4.1. Pre-measurement and post-measurement
and is taken from a NERC funded field campaign held at calibration concept
several urban sites in the West Midlands (central UK). Given
that the ozone measurements were taken for a period of over In this concept, an additional calibration is conducted after
30 days for each sensor, this level of correlation is acceptable a measurement campaign. This data can then be used to anal-
(r.m.s. of the ozone differences is around 7 ppbv). Neverthe- yse the measurements. Fig. 9 shows a time series of ozone
less, the relatively high degree of scatter in the correlation plot measured by one sensor but analysed with both the pre- and
can be improved. One of the causes of the high scatter is the post-measurement calibration data.
apparent change in sensor sensitivity (for some sensors) with Obviously, for the first few days, the pre-measurement
time and drifting sensor scrubbing-phase resistance. Fig. 8 calibration data gives better ozone comparison (with ozone
shows differences in sensor calibration curves for two WO3 measured by a UV absorption spectrometer) than the post-
sensors, which were obtained immediately before and after measurement calibration data. After about a week of mea-
a 30-day measurement period. Why one sensor should be surements, the sensitivity of the sensor has changed so that

Fig. 6. Time series of ozone measured by a WO3 sensor and an UV absorp- Fig. 8. Calibration curves from WO3 sensor calibrations conducted before
tion spectrometer in urban conditions at Sandwell (UK). and after a 30-day measurement period shows changing sensor sensitivity.
S.R. Utembe et al. / Sensors and Actuators B 114 (2006) 507–512 511

helps to correct for any changes in sensitivity and scrubbing-


phase resistance.

Acknowledgements

The authors gratefully acknowledge the support of Bob


Appleby of Birmingham (UK) City Council Environmental
Services for his helpful information on measurement sites
and UV absorption ozone measurements, and the UK Nat-
ural Environment Research Council (NERC) for financial
support.

Fig. 9. Time series of WO3 sensor ozone measurements calculated by cal- References
ibrations conducted before (solid thick line) and after (solid thin line) a 30
day field measurement campaign. [1] PORG, Ozone in the United Kingdom, Fourth Report of the United
Kingdom Photochemical Oxidants Review Group, Department of the
Environment, London, 1997.
[2] J.M. Berak, M.J. Serenko, Effect of oxygen deficiency on electronic
after this period, the calibration data that gives the best results transport properties of tungsten trioxide crystal, J. Solid State Chem.
is the one that is done after the measurement period. 2 (1970) 109–133.
This observation has been found to be the case for most [3] D.E. Williams, Semiconducting oxides as gas-sensitive resistors, Sens.
of the sensors used in this study. Thus, it has been neces- Actuators B 57 (1999) 1–16.
[4] S.R. Aliwell, J.F. Halsall, K.F.E. Pratt, J. O’Sullivan, R.L. Jones, R.A.
sary to use only the post-measurement calibration data to
Cox, S.R. Utembe, G.M. Hansford, D.E. Williams, Ozone sensors
calculate the ozone concentrations from the measured sensor based on WO3 : a model for sensor drift and a measurement correction
resistances. method, Meas. Sci. Technol. 12 (2001) 684–690.
The fact that the sensor sensitivity seems to change so soon [5] I. Langmuir, The evaporation, condensation and reflection of
after the start of the field measurements seems to suggest that molecules and the mechanism of adsorption, Phys. Rev. 8 (1916)
149.
it is necessary to burn-in the sensors before commencement of
[6] D.E. Williams, S.R. Aliwell, K.F.E. Pratt, D.J. Caruana, R.L. Jones,
the field measurements, if the post-measurement calibration R.A. Cox, G.M. Hansford, J. Halsall, Modelling the response of tung-
is to be used. We recommend about a week of burning-in as sten oxide semiconductor as a gas sensor for the measurement of
a minimum. During this period, the sensors should be run in ozone, Meas. Sci. Technol. 13 (2002) 923–931.
ambient ozone concentrations before calibration. [7] P.A. Cox, Transition metal oxides: an introduction to their electronic
structure and properties, Clarendon Press, 1995.
[8] Electromagnetic Compatibility (EMC): Guide to the Application of
Directive 89/336/EEC, 1997 ed. Office for Official Publications of
5. Conclusion the European Commission, European Commission DG III-Industry,
Luxembourg.
An ozone instrument based on the WO3 sensor is showing [9] G.M. Hansford, R.A. Freshwater, R.A. Bosch, R.A. Cox, R.L. Jones,
K.F.E. Pratt, D.E. Williams, A low cost instrument based on a solid
a lot of potential as a cost-effective alternative to conventional state sensor for balloon-borne atmospheric O3 profile sounding, J.
ozone measuring instruments. Based on the fundamental Environ. Monit. 7 (2005) 158–162.
reactions occurring at the surface of the oxide, a Langmuir-
type model is able to characterise the response of the sensor
to changes in ozone concentrations. Biographies
For the first time, urban ozone measurements made by a
WO3 solid-state sensor have been compared with those mea- Steven Robert Utembe: Steven Utembe’s honours degree in chemistry
was awarded at Chancellor College, University of Malawi in 1996. His
sured by a conventional UV absorption instrument. Results PhD, which was obtained in 2000 at the University of Cambridge involved
from over 30 days of data obtained from eight WO3 sensors both tropospheric modelling studies and measurement studies of urban
have shown good correlation with ozone measurements from ozone concentrations using a novel WO3 sensor. His major interests are
a UV absorption spectrometer, with an r.m.s. of the variance in modelling studies of tropospheric chemistry. He is currently conducting
of about 7 ppbv. The results have shown that the performance research work in photochemical pollution modelling studies at Imperial
College London.
of the sensor depends on the sensor behaviour in the course
of the measurement period. Apparent changes in sensor sen- Graeme Hansford: Graeme Hansford obtained a BA degree in natu-
sitivity can be accounted for by calibrating the sensors after ral sciences (chemistry), and a PhD in infrared laser spectroscopy at
the University of Cambridge, UK. Currently working at the Centre for
approximately 1 week of ‘burn-in’ in ambient ozone condi- Atmospheric Science, Chemistry Department, University of Cambridge,
tions. Alternatively, it has been found that conducting another research interests include instrumentation for balloon- or aircraft-borne
calibration immediately after a field measurement campaign operation, particularly for the measurement of ozone and water vapour.
512 S.R. Utembe et al. / Sensors and Actuators B 114 (2006) 507–512

Michael Sanderson: Michael Sanderson, Bsc(Hons) in chemistry, in air pollution research, firstly at the UKAEA Harwell laboratory and
awarded 1991, and DPhil in chemistry, awarded 1995, both from the since 1995 at the department of chemistry. During that time Dr. Cox
University of York. He is currently at the Met Office. His interests are has made many pioneering laboratory studies of the kinetics and mech-
how trace gases, particularly surface ozone levels, are affected by climate anisms of reactions involved in stratospheric and tropospheric chemistry,
change, and atmosphere–vegetation interactions. and measurements of ozone and other atmospheric pollutants. Dr. Cox
has acted as co-ordinator of the CEC Combustion Kinetics Programme
Ray Freshwater: Ray Freshwater obtained his BSc with honours in between 1988 and 1990, and several EC projects in the atmospheric
physics and electronics at Brunel University in1986. He is currently chemistry area, including the current fifth Framework project: ‘Tropo-
employed as a technical officer in the Department of chemistry at the spheric Halogens—Effect On Ozone—(THALOZ)’. Since joining Cam-
University of Cambridge. His current field of interest is in aircraft and bridge University nine years ago he established new facilities for studies
balloon borne atmospheric instrumentation for ozone and water vapour of heterogeneous atmospheric processes and a website containing the
measurements. IUPAC evaluation of kinetic data for atmospheric chemistry.
Keith Pratt: Keith Pratt’s first degree was in chemistry with computer
based technology, at the University of Warwick. His PhD in electro- Roderic Lewis Jones: reader in atmospheric science, department of
chemistry was awarded at the University of Southampton in 1994. He chemistry, University of Cambridge. Dr. R.L. Jones has over 25 years
subsequently worked on gas sensor R&D, both in industry and academia, experience in atmospheric science research. He obtained his D Phil at the
for 10 years. He is currently Principal Scientist at City Technology Ltd. (then) department of atmospheric physics, Oxford for studies of the struc-
working on novel gas sensor technologies. ture and composition of the stratosphere from satellites, which provided
some of the early insights on atmospheric circulation and the strato-
David Edward Williams: David Williams is currently chief scientist spheric water vapour budget. In 1984 he moved to the UK Meteorological
at Inverness Medical Innovations, based at Unipath Ltd., Bedford, UK. Office where he became involved with observational studies of tropopheric
Previously (1991–2002) he was professor of physical chemistry and head composition using aircraft and the pioneering airborne investigations of
of department of chemistry at University College London. He started stratospheric ozone loss in polar regions. In 1990 he took up his cur-
his career in chemical sensor research and development in 1981 at the rent post where he has been involved with a wide range of observational
Harwell Laboratory of the UK Atomic Energy Authority. studies of atmospheric composition and the development of a range of
Richard Anthony Cox: adjunct reader in atmospheric chemistry, Univer- measurement methods including lidars, cavity ringdown spectroscopy and
sity of Cambridge. Dr. R.A. Cox, has had 36 years research experience solid state gas sensors.

Das könnte Ihnen auch gefallen