Effect of processing on the properties of tin oxide-basedthick-film gas sensors

Aftab Ahmad*, Jennifer Walsh, Thomas A. WheatMaterials Technology Laboratory CANMET/NRCan, 405 Rochester Street, Ottawa, Ont., Canada K1A 0G1

Abstract

Semiconducting tin oxide-based powders were synthesized via three different chemical precipitation methods and also obtained from a

commercial source. The precursor powders were characterized using X-ray diffraction (XRD), particle size and surface area measurement

techniques. Sensor elements were produced from calcined powders using thick-film screen printing technology. The influence of various

powder processing methods, sintering conditions, microstructure and moisture on the carbon monoxide gas detection capabilities, and the

response and recovery times of the sensors produced from the above four powders are discussed.

Crown Copyright # 2003 Published by Elsevier Science B.V. All rights reserved.

Keywords: Semiconducting oxide; Gas sensors; Materials processing; Characterization

1. Introduction

Intelligent systems based on sensors and controls offer an

opportunity to increase the efficiency of energy usage in

various combustion systems and reduce the emission of

greenhouse gases (GHG’s) that cause global warming, smog

and acid rain. In addition, sensors (S’s) and controls (C’s)

can be used for process control to improve productivity and

product quality in various industrial sectors. Applications

exist in air quality monitoring and control, energy efficiency,

industrial process control, transportation, health and safety

(e.g. to detect toxic, flammable and explosive gases in

mines, industrial/residential environments), medical diag-

nostics, material recycling, climate change, etc. There is

an ongoing need for the development of fast, sensitive,

rugged and reliable sensors and controls for applications

in harsh (high temperature/corrosive) environments found in

automotive, metallurgical processing (e.g. steel-making,

metal casting, etc.), aerospace, glass, ceramics, pulp and

paper, energy production/generation, etc., industries. The

automotive industry is an excellent example, wherein on-

line oxygen sensors and a three-way catalyst in conjunction

with a microprocessor-based control circuit provide >10%

increase in fuel efficiency and commensurate reduction in

GHG’s.

It has long been known that the electrical properties of a

semiconducting material change as a result of adsorption

and/or reaction with the gases in the atmosphere. This

property of semiconducting materials is typically exploited

in gas sensing measurements. Semiconducting oxides such

as ZnO, TiO2, Fe2O3, WO3, and SnO2 are commonly

investigated materials for gas sensing applications. Among

these, tin oxide (SnO2) is perhaps the most widely investi-

gated. It generally behaves as an n-type semiconductor (i.e.

electrons are majority charge carriers) with a wide band-gap,

DE ¼ 3:6 eV due to slight non-stoichiometry arising from

the variable oxidation states of tin [1]. The major drawback

with tin dioxide-based semiconducting gas sensors is their

non-selectivity, i.e. the inability to provide clearly distin-

guishable response signals when exposed to a mixture of

reducing gases, e.g. CO, CH4, C3H8, H2, etc. Also, environ-

mental factors such as variations in temperature and humid-

ity can often adversely influence the sensor response [2–6].

These problems can often be resolved by using higher

operating temperatures, by adding dopants or other metal

oxides into the tin oxide or else, by changing the base

semiconducting oxide [7–13].

The gas sensing properties of these materials are known to

be influenced by the material formulation, precursor raw

materials, as well as various fabrication and processing

conditions that affect the microstructure of the precursor

powders and the gas sensing elements. This report provides

data on the tin oxide-based gas sensors prepared from

precursor powders synthesized via three different precipita-

tion routes and also those fabricated using commercial tin

oxide powders. Precursor powders have been synthesized

using both high- and low-concentration solutions of chemi-

Sensors and Actuators B 93 (2003) 538–545

* Corresponding author.

E-mail address: aahmad@nrcan.gc.ca (A. Ahmad).

0925-4005/03/$ – see front matter. Crown Copyright # 2003 Published by Elsevier Science B.V. All rights reserved.

doi:10.1016/S0925-4005(03)00238-7

cal reagents (i.e. SnCl4�5H2O and NH4OH) used for pre-

cipitation reactions. In addition, a homogeneous precipita-

tion method that involve precipitation of SnCl4�5H2O using

a urea solution has also been used for the synthesis of SnO2

precursor powders. The effect of reagent/solution concen-

tration on the physical properties of precipitated and as-

calcined powders as well as on the sintered thick-film sensor

elements produced from these powders has been investi-

gated. To our knowledge, no systematic study on the influ-

ence of chemical reagents/solution concentrations on the

properties of tin oxide precursor powders and sintered sensor

elements has been previously reported. Also, even though

several semiconducting materials (including SnO2) have

been synthesized via the homogeneous precipitation method

to provide high surface area precursor powders, reports on

their utilization for gas sensors applications appear to be

absent. The paper also gives detailed information on the

synthesis and processing of tin oxide powders via three

different chemical processing methods to produce high

surface area tin oxide precursor powders. Such details are

often missing in the literature. Here, data is provided about

the influence of various powder synthesis and processing

conditions on the surface area/particle size, phase composi-

tion and crystal size (as determined by X-ray diffraction

(XRD) technique) of both uncalcined and calcined materi-

als. The influence of precursor powders, sintering tempera-

ture, sintering time, humidity, as well as the concentration of

the CO gas on the electrical conductance, sensitivity and

response/recovery times of the thick-film gas sensors pro-

duced from these powders is also discussed.

2. Experimental

2.1. Preparation of precursor powder

The SnO2 precursor powders were synthesized via three

different precipitation routes and also obtained from a

commercial supplier (Strem Chemicals, USA). The first

precipitation method designated here as Sn-HC used rela-

tively higher concentrations of chemical reagents. For this

purpose, 0.52 mol/L aqueous solution of SnCl4�5H2O was

slowly added to 5.85 mol/L NH4OH solution with constant

stirring. A white gelatinous precipitate was formed. The

mixture was centrifuged at �1500 rad/s for 10 min and the

supernatant liquid was decanted. The precipitate was

washed with distilled water and re-centrifuged; again the

supernatant liquid was decanted. This process was repeated

several times to remove traces of chloride ions. The presence

of chloride ions in the supernatant layer was tested with

silver nitrate solution. The precipitate was then suspended in

isopropanol, stirred and rotary-evaporated to dryness to

produce a soft, agglomerated powder. The powder was dried

at 120 8C, hand-ground using a mortar and pestle and

calcined at 550 8C for 2 h in an alumina crucible to provide

the tin oxide powder product.

The second precipitation process (Sn-DL) involved the use

of dilute NH4OH (�0.4 mol/L) and SnCl4�5H2O (0.1 mol/L)

solutions. For this purpose, 0.1 mol/L aqueous solution of

SnCl4�5H2O was added dropwise into 1 L of ammonia solu-

tion (pH ¼ 10) over a period of 1.5 h with constant stirring. A

fluffy, cloudy white precipitate was observed. The beaker was

topped up to the 4 L mark with distilled water and the mixture

was stirred overnight. The stirring was then stopped to allow

the precipitate to settle. The supernatant liquid was siphoned

off and discarded. The remaining wet precipitate was cen-

trifuged for 15 min at 1500 rad/s and the supernatant layer

was decanted. The precipitate was washed with distilled

water with stirring, re-centrifuged for 5–10 min and dec-

anted. This was repeated several times to remove traces of

chloride ions as confirmed via the silver nitrate test. The

precipitate was suspended in isopropanol, stirred and rotary-

evaporated to remove the solvent. The powder was dried at

120 8C, hand-ground using a synthetic corundum (Diamo-

nite) mortar and pestle and calcined at 550 8C for 2 h.

The third precipitation process (Sn-HM) is also known as

the homogeneous co-precipitation method [14]. In this

process, the precipitation occurs via a highly uniform

increase in the pH of the solution and this uniformity is

achieved by thermal decomposition of urea according to the

following reaction:

ðNH2Þ2CO þ 3H2O ! CO2 þ 2NH4þ þ 2OH� (1)

For the homogeneous co-precipitation process, 0.5 mol/L

urea solution was added dropwise to 0.025 mol/L

SnCl4�5H2O solution, with stirring over a 10 min period.

The mixture was then heated slowly to 80 8C and allowed

to react for 4 h. The urea decomposes slowly during the

release of ammonia and carbonate ions into the solution. The

gradual and uniform rise in pH results in nucleation and

growth of uniformly-sized and shaped particles of the metal

oxy-basic carbonate. The mixture was then cooled in an ice

bath for 3 h. The liquid layer above the white, gel-like

precipitate was then discarded, the precipitate was centri-

fuged for 10 min and the supernatant discarded. The pre-

cipitate was washed several times with water to remove any

traces of chloride ions. Once all the chloride was removed and

the liquid siphoned off, the product was suspended in iso-

propanol which was removed by rotary-evaporating. It was

then dried at 120 8C, hand-ground and calcined in an alumina

crucible at 550 8C for 2 h to form tin oxide powders. Com-

mercial tin oxide powders (Sn-CM) were also calcined at

550 8C for 2 h before being used to fabricate sensor elements.

2.2. Fabrication of sensor elements

The calcined powders obtained by the above four methods

were first ground using a mortar and pestle to achieve

consistent and fine particles. The sensor fabrication proce-

dure included the sequential screen-printing of inter-digital

gold electrodes onto an alumina substrate (2:54 cm2:54 cm in dimension, laser-scribed into nine equal squares)

A. Ahmad et al. / Sensors and Actuators B 93 (2003) 538–545 539

followed by the superposition of a gas-sensitive layer using

an Accu-Coat 3200 screen printer. The gold inter-digital

electrodes (having comb-type structure) were printed on an

alumina substrate using gold paste, 3066 Au, obtained from

Ferro Inc. These electrodes were fired at 850 8C for 10 min.

The pastes used for the screen-printing of a gas-sensitive

layer consisted of the calcined powders produced by the

above four methods. These powders were mixed with an

organic vehicle in the ratio of 65 weight percent powder, 35

weight percent organic vehicle. The paste was made using a

three-roller mill from Ross Mixers Inc. A statistical design

for experiments (SDE) methodology developed by Taguchi

[15] was used for the sintering of the sensor elements using

an L-9 Taguchi orthogonal array [16]. This fractional

factorial design methodology utilizes specially designed

orthogonal array matrices to generate SDE. The sensing

elements produced from each of the four precursor powders

were sintered at 725, 775 and 850 8C, each for periods of 10,

20 and 30 min.

2.3. Materials/sensors characterization

X-ray diffraction measurements were conducted to deter-

mine crystal structure or phase composition of the materials.

The crystallite sizes (Rx) of the calcined and sintered mate-

rials were calculated from the XRD peak broadening of the

(1 1 0) peak at a diffraction angle of �26.558 2y using

Scherrer’s equation [17]:

Rx ¼0:9lb cos y

(2)

where l is the wave length (1.5406 A for Cu Ka radiation), bthe full-width-half-maximum (FWHM) of a peak in radians

and y the diffraction angle. The specific surface area of

powders was measured by BET method by nitrogen adsorp-

tion using a Quantasorb instrument from Quantachrome

Corp. The mean particle size Rm (nm) was calculated from

surface area data using the following equation:

Rm ¼ 6

rtSa(3)

where rt is the true density (g/m3) and Sa the surface area of

the powder (m2/g).

Gas-sensing experiments were carried out in a stainless

steel sample holder apparatus equipped with a programma-

ble temperature controlled heater. The sample holder can

accommodate up to eight samples for simultaneous elec-

trical resistance measurements. An inlet and an outlet were

provided to enable the selected test gases to flow into and out

of the sample holder. For resistance measurements, the

samples were heated up to �500 8C for about 1 h and

electrical data was collected on the cooling cycle using

a cooling ramp of 2 8C/min. The gas flow rate over the

sensor was kept constant at 120 standard cm3/min during the

entire test. The sample/test gases used were in the range of

100–500 ppm of CO in air. A load resistance (RL) was

connected in series with the sensor and a circuit voltage

(� 5 V) was applied across them. The resistance of the

sensor RSample, was obtained by measuring the voltage drop

across the standard load resistance using the following

equation:

RSample ¼Vc

Vout� 1

� �RL (4)

where RL denotes the load resistance value (ranging from

10 kO to 1 MO), Vc is the applied voltage (5 V) and Vout the

voltage drop across the load resistance. The conductance of

the sample in air or in carbon monoxide is the inverse of the

sample resistance, RSample.

A Hewlett-Packard data acquisition system, HP 34970A

combined with a multiplexer HP34901A and HP Benchlink

data logger software was used to collect the data. The gas

flow was controlled using mass flow controllers and a four-

channel readout from MKS instruments. A type K thermo-

couple placed close to the sensor chip indicated the operat-

ing temperature. The electrical resistance of each sensor

element was measured in air (Ra) and in the sample gas (Rg),

(CO in this case), over a temperature range of 500–200 8C.

Gas sensitivity, S, at a particular temperature was evaluated

using the relationship, S ¼ Ra/Rg. For gas sensitivity mea-

surements, the sample was heated to a chosen temperature

where it remained for the duration of the test. At this

particular temperature, the resistance was measured at three

different concentrations of carbon monoxide in air: 100, 250,

and 500 ppm. First, air was run for about 10–20 min to

establish a baseline by measuring the sample resistance

every 30 s. Thereafter, 100 ppm CO in air was run for a

fixed length of time (5 min) while resistance of the samples

was measured every 30 s. This was followed by resistance

data measurements in air every 30 s for a period of 10 min.

This procedure was repeated for 250 ppm CO in air as well

as for 500 ppm CO in air. A constant gas flow rate of

120 sccm was maintained and the flow rates were adjusted

for both air and CO accordingly to obtain the desired CO

concentrations.

3. Results and discussions

3.1. Phase evaluation and particle size of samples

The X-ray diffraction patterns of tin oxide powders

obtained via four different processing routes and calcined

at 550 8C for 2 h, are shown in Fig. 1. The calcined powders

exhibit only the cassiterite phase without any measurable

impurity phase. Compared to the powders from the com-

mercial source, the XRD patterns of the powders prepared

in-house by the three precipitation methods displayed rela-

tively broader diffraction peaks, indicative of smaller grain/

crystallite size. The crystallite size (Rx values) of the cal-

cined tin oxide powders calculated from the [1 1 0] XRD

peak using the Scherrer formula is given in Table 1. The data

540 A. Ahmad et al. / Sensors and Actuators B 93 (2003) 538–545

clearly indicates that the tin oxide powders synthesized via

chemical processing routes provide a relatively finer grain/

crystallite size (Rx values) compared to those obtained from

the commercial source. For surface-conducting type sensors,

a finer particle size is desirable as it provides a greater

surface-to-bulk-volume ratio resulting in increased adsorp-

tion of gases at the sensors surface leading to enhanced

sensitivity of the gas sensors. The chemically precipitated

powders (Sn-DL and Sn-HM) synthesized using dilute

solutions of precipitating agents (NH4OH and urea, respec-

tively) provide finer powders with grain sizes of 10.06 and

10.77 nm, respectively, whereas Sn-HC powders prepared

using higher concentration NH4OH as a precipitating agent,

produced slightly larger grain size powders (12.32 nm in

size). Compared to the chemically processed materials,

powders obtained from a commercial source exhibited much

larger crystallite size (30.15 nm). Although all of the pow-

ders in this study were calcined at 550 8C for 2 h, the

relatively large crystallite sizes obtained for the commercial

powders indicate that synthesis and processing conditions

different from those used in-house for the production of

precursor powders may have been used by the commercial

supplier. For example, the supplier may have subjected

commercial powders to higher heat treatment conditions.

The sensor elements were printed using pastes that were

made from the calcined powders produced via the above four

methods. The sensor chips were sintered in air at 725 8C for

10 min. The calculated Rx values for the sensor chips

sintered at 725 8C for10 min are also given in Table 1. As

expected, the high sintering temperature used for the firing

of the sensor chips (725 8C for 10 min) resulted in enlarging

the grain size of the calcined powders. Compared to calcined

powders, the sintered chips exhibit 10–30% increase in grain

sizes. The XRD patterns of the sintered sensor chips also

displayed XRD peaks arising from the alumina substrate.

This occurred because the sensor element did not fully cover

the alumina substrate and a small part of the alumina was

exposed to XRD.

Surface area measurements were performed using the

BET method on powders produced via the above four

processing routes. Both the calcined and uncalcined/as-dried

(at 135 8C for 12 h) powders were used for surface area

measurements. Data on the powder surface area measure-

ments and the mean grain size calculated using Eq. (3) are

provided in Table 2. Both the calcined and as-dried powders

from the commercial source displayed fairly large grain size

>120 nm and small surface area (6.41 and 7.01 m2/g, respec-

tively) due presumably to the different processing conditions

used by the commercial supplier as discussed above. Pow-

ders produced via chemical precipitation routes and calcined

at 550 8C for 2 h exhibited relatively larger surface areas

Fig. 1. XRD patterns of tin oxide powders calcined at 550 8C for 2 h.

Table 1

Crystallite size (Rx) calculated from [1 1 0] X-ray diffraction peaks

Sample Rx (nm) calcined

powders

Rx (nm) sintered

chips

Sn-CM 30.15 35.46

Sn-DL 10.06 13.42

Sn-HC 12.33 15.48

Sn-HM 10.77 14.38

Table 2

Surface area and particle size data for uncalcined (uc) and calcined (c)

powders

Powder Surface area (m2/g) Particle size (nm)

Sn-CM (uc) 7.0 122.44

Sn-CM (c) 6.4 133.92

Sn-DL (uc) 286.6 2.99

Sn-DL (c) 37.3 22.97

Sn-HC (uc) 191 4.48

Sn-HC (c) 27.7 30.94

Sn-HM (uc) 268 3.19

Sn-HM (c) 33.4 26.22

A. Ahmad et al. / Sensors and Actuators B 93 (2003) 538–545 541

(>27 m2/g) and smaller grain size (<27 nm). The uncalcined

powders Sn-DL and Sn-HM precipitated from dilute solu-

tion exhibited very large surface area (268 m2/g) and very

small particle size (<3.5 nm). The data also indicates that the

calcination of these powders at 550 8C for 2 h results in

almost an order of magnitude reduction in both the surface

area and the particle size of the powders. The precipitated

powders from the concentrated solution Sn-HC provided

relatively smaller surface areas (191 and 27.7 m2/g) for the

uncalcined and calcined powders respectively.

3.2. Electrical conductance and CO sensitivity data

A typical response plot in air and CO for thick-film gas

sensors produced from the above four precursor powders is

shown in Fig. 2. The sensitivity ‘‘S’’ of the gas sensors was

obtained from the ratio of the resistance in air to that in CO

(S ¼ Ra/RCO ) using response plots similar to Fig. 2. The

sensitivity data at 4258C for gas sensors sintered at 725 8Cfor 10 min, as a function of CO gas concentration is shown in

Fig. 3. Here, sensors produced from Sn-DL and Sn-HM

powders provide relatively higher sensitivity to carbon

monoxide. As expected, the sensitivity of the sensors

increased with the increase in CO concentration. These

powders also exhibited higher surface areas and smaller

particle size values. In some cases, sensitivity of the sensors

produced from Sn-DL powders was more than twice as large

as those produced from Sn-HC and Sn-CM. It is not clear

why sensors produced using Sn-HC powders displayed

lower sensitivity values despite the fact that the precursor

powders for these samples also exhibited substantially

higher surface area and smaller particle size compared to

Fig. 2. Typical response plot and sensitivity data for SnO2 sensors.

Fig. 3. Sensitivity at 425 8C vs. CO concentration for various SnO2 sensors.

542 A. Ahmad et al. / Sensors and Actuators B 93 (2003) 538–545

commercial powders. Sensors produced from Sn-DL pow-

ders and sintered at 725 8C for 10 min exhibited the highest

sensitivity (5.64) to CO gas at 425 8C.

The sensitivity data for these sensors as a function of

operating temperature is shown in Fig. 4. The sensitivity of

the sensors increased with the increase in temperature and

peaked at �400 8C, except for the sensor Sn-DL where the

peak sensitivity was found to be around 425 8C. The sensi-

tivity of the gas sensors decreased with further increase in

temperature. Sensors produced from Sn-DL powders also

displayed a decrease in sensitivity to CO with the increase in

sintering temperature and sintering time (Fig. 5). No dis-

cernable trends between sintering conditions and the sensi-

tivity data were obtained for sensors produced from the other

three powders. The sensitivity data in CO at 425 8C for Sn-

HC sensors sintered at 725 8C for 10 min under dry and

humid (� 85% RH) 500 ppm CO in air is shown in Fig. 6. It

is clear that the sensitivity of the gas sensors is substantially

increased in a wet CO environment. A similar increase in gas

sensitivity in a humid environment has previously been

observed by others [18,19] and is generally attributed to

the formation of H2 due to a surface-catalyzed water-gas

shift reaction:

COðgÞ þ H2O $ CO2 þ H2

The free energy change for the above reaction is negative at

the test temperature.

The response and recovery times at 350 and 425 8C for

these sensors when exposed to 500 ppm CO were calculated

from response plots similar to that shown in Fig. 2. The

Fig. 4. Sensitivity vs. temperature data at 500 ppm CO in air for sensors sintered at 725 8C for 10 min.

Fig. 5. Sensitivity vs. sintering temperature at 425 8C for Sn-DL in 500 ppm CO in air.

A. Ahmad et al. / Sensors and Actuators B 93 (2003) 538–545 543

response time is defined here as the time it takes for the

conductance of the gas sensor to increase to 90% of the

maximum conductance when 500 ppm of CO gas is intro-

duced into an environment of air. The recovery time is the

time required for 90% reduction in conductance when CO is

turned off and air is re-introduced into the chamber (i.e. time

to reach 10% conductance value in CO). The conductance

data for these experiments was recorded every 10 s. As

shown in Table 3, both the response and recovery data for

these sensors show strong dependence on the method of

processing precursor powders as well as the sensor’s oper-

ating temperature.

Sensors (Sn-DL and Sn-HM) produced from powders

synthesized via precipitation of dilute solutions displayed

shorter or faster response and recovery times, whereas

sensors produced from commercial powders (Sn-CM) and

those precipitated using high concentration reagents (Sn-

HC) generally exhibited longer response and recovery times.

Also, an increase in the operating temperature of the sensors

from 350 to 450 8C resulted in substantial reduction in

response and recovery times (>50% in many cases). The

response and recovery times for these samples in a humid

environment (�85% RH) of 500 ppm CO in air were �6–10

times longer than those observed in a dry environment of

500 ppm CO in air.

4. Conclusions

The three precipitation methods used in this study to

synthesize precursor powders provided materials having

very high surface areas (190–286 m2/g) and small particle

sizes (3–4.5 nm). Calcination of these powders at 550 8C for

2 h provided single-phase tin oxide with almost an order of

magnitude reduction in surface area and particle size values.

However, these calcined precipitated powders exhibit sur-

face area values that are still several times larger than those

exhibited by commercial powders. The relatively larger

particle/crystallite size and smaller surface area observed

for the commercial powders suggests that probably different

processing or heat treatment conditions than those applied in

this study were used by the commercial supplier to produce

their powders. Sensors produced from the powders synthe-

sized via precipitation of dilute solutions (Sn-DL and Sn-

HM) displayed substantially higher sensitivity to CO gas

than those precipitated from concentrated solutions (Sn-HC

powders) or those made from the commercial powders (Sn-

CM). Sn-HC samples in a wet CO environment exhibited a

large increase in sensitivity. The response and recovery data

for these sensors displayed a strong dependence on the

powder processing method, sensor operating temperature

and humidity. The response times for samples produced

from powders synthesized via precipitation of dilute solu-

tions were two to three times shorter than those produced via

precipitation of concentrated solutions. Overall, the data

clearly indicates that the physical characteristics of tin

oxide-based semiconducting gas sensors are strongly influ-

enced by the powder processing methods as well as operat-

ing conditions. For example, Sn-DL samples sintered at

850 8C for 10 min exhibited �20% higher sensitivity to CO

gas compared to those sintered at 850 8C for 30 min. These

samples also exhibited >100% higher sensitivity to 500 ppm

CO in air at 425 8C compared to the samples produced from

Fig. 6. Sensitivity at 400 8C vs. concentration in dry and wet (�85% RH) CO in air for Sn-HC, sintered at 725 8C for 10 min.

Table 3

Ninety percent response and recovery time data for SnO2 sensors in

500 ppm dry CO in air

Sensor Response time (s) Recovery time (s)

350 8C 425 8C 350 8C 425 8C

Sn-DL 10–20 10–15 50–60 30–35

Sn-HM 20–30 10–20 70–80 45–50

Sn-CM 50–55 10–20 150–160 50–55

Sn-HC 70–75 30–35 190–200 70–75

544 A. Ahmad et al. / Sensors and Actuators B 93 (2003) 538–545

commercial powders and also those produced from powders

synthesized via the precipitation of concentrated solution

(i.e. Sn-HC samples). All of the samples displayed an

increase in sensitivity to CO gas in a humid environment;

however, compared to dry CO, the response time in wet

(� 85% RH) CO was about an order of magnitude greater.

Acknowledgements

The authors gratefully acknowledge the financial support

for this work from the Natural Resources Canada Program

on Energy Research and Development (NRCan-PERD). The

technical support provided by Mr. George McDonald,

Senior Research Technologist, NRCan/MTL is also very

much appreciated.

References

[1] S. Samson, C.G. Fonstad, Defect structure and electronic donor

levels in stannic oxide crystals, J. Appl. Phys. 44 (1973) 4618–4621.

[2] J.F. McAleer, P.T. Moseley, J.O.W. Norris, D.E. Williams, Tin

dioxide gas sensors, J. Chem. Soc., Faraday Trans. Part 1 83 (1987)

1323–1346.

[3] N. Yamazoe, J. Fuchigami, M. Kishikawa, T. Seiyama, Interactions

of tin oxide surface with O2, H2O and H2, Surf. Sci. 86 (1986) 335–

344.

[4] J. Tamaki, M. Nagaishi, Y. Teraoka, N. Miura, N. Yamazoe,

Adsorption behavior of CO and interfering gases on SnO2, Surf.

Sci. 221 (1989) 83–196.

[5] P. Van Geloven, M. Honore, J. Roggen, S. Leppavuori, T. Rantala,

The influence of relative humidity on the response of tin oxide gas

sensors to carbon monoxide, Sens. Actuators B 4 (1991) 185–188.

[6] M. Egashira, M. Nakashima, S. Kawasumi, T. Seiyama, Temperature

programmed desorption study of water adsorbed on metal oxides. 2.

Tin oxide surfaces, J. Phys. Chem. 85 (1981) 4125–4130.

[7] G.S. Devi, S.V. Manorama, V.J. Rao, SnO2/Bi2O3: a suitable system

for selective carbon monoxide detection, J. Electrochem. Soc. 145 (3)

(1998) 1039–1044.

[8] G.S.V. Coles, G. Williams, B. Smith, Selectivity studies on tin oxide-

based semiconductor gas sensors, Sens. Actuators B 3 (1991) 7–14.

[9] G.S.V. Coles, S.E. Bond, G. Williams, Selectivity studies and oxygen

dependence of tin(IV) oxide-based gas sensors, Sens. Actuators B 4

(1991) 485–491.

[10] Zhou, Y. Xu, Q. Cao, S. Niu, Metal-semiconductor ohmic contact of

SnO2-based ceramic gas sensors, Sens. Actuators B 41 (1997) 163–

167.

[11] K. Nomura, H. Shiozawa, T. Takada, H. Reuther, E. Richter,

Gas-sensor properties of SnO2 films implanted with gold and iron

ions, J. Mater. Sci.: Mater. Electron. 8 (1997) 301–306.

[12] H.W. Cheong, J.J. Choi, H.P. Kim, J.M. Kim, J.M. Kim, The role of

additives in SnO2 semiconductor gas sensors, IEEE (1991) 154–156.

[13] J.L. Solis, V. Lantto, Gas-sensing properties of different alpha-

SnWO4-based thick films, Phys. Scripta T69 (1997) 281–285.

[14] K.C. Song, Y. Kang, Preparation of high surface area tin oxide

powders by a homogeneous precipitation method, Mater. Lett. 42

(2000) 283–289.

[15] G. Taguchi, Introduction to quality engineering, Asian Productivity

Organization, Distributed by American Supplier Institute Inc.,

Dearborn, MI, 1981.

[16] M.S. Phadke, Quality Engineering Using Robust Design, Printice-

Hall, London, 1989.

[17] B.D. Cullity, Elements of X-ray Diffraction, second ed., Addison-

Wesley, Massachusetts, 1978.

[18] L.N. Yannopoulos, Antimony-doped stannic oxide-based thick-film

gas sensors, Sens. Actuators A 12 (1987) 77–89.

[19] J.F. Boyles, K.A. Jones, The effect of CO, water vapour and surface

temperature on the conductivity of SnO2 gas sensors, J. Electron.

Mater. 6 (6) (1977) 717–732.

Biographies

Aftab Ahmad received his BS and MS degree in physical chemistry from

University of Karachi and his PhD degree in photophysical chemistry from

the University of New Brunswick in Fredericton, New Brunswick, Canada.

He joined the Materials Technology Laboratory of the Ministry of Natural

Resources Canada in 1980, where he is currently working as a senior

research scientist and functional group leader in the Advanced Materials

Group. He has been involved in the research and development of a variety

of electronic ceramic materials such as gas/chemical sensors, piezoelectric/

electrostrictive sensors and actuators, fuel cells, solar cells, thermistors,

varistors, etc., for over 20 years.

Jennifer Walsh received her BS degree in chemistry from Dalhousie

University in Halifax, Nova Scotia, Canada. She joined the Materials

Technology Laboratory of the Ministry of Natural Resources Canada in

2001, where she is currently working as a research technologist in the

Advance Materials Group. She is also working towards her MS degree in

chemistry with specialization in the development of catalytic materials for

gas sensor applications.

Thomas Wheat received his BSc degree in chemistry and PhD degree in

ceramics from the University of Leeds in UK. He joined the Mineral

Sciences laboratory (MSL) of the Ministry of Natural Resources Canada as

a Research Scientist in 1968. He was the Head of the Ceramics Group at

MSL where he conducted R&D work both in the area of Functional and

Structural Ceramics. He is currently an emeritus scientist at the Materials

Technology Laboratory of the Ministry of Natural Resources Canada.

A. Ahmad et al. / Sensors and Actuators B 93 (2003) 538–545 545