Professional Documents
Culture Documents
General Characteristics of Thermally Cycled Tin Oxide Gas Sensors 1989
General Characteristics of Thermally Cycled Tin Oxide Gas Sensors 1989
General Characteristics of Thermally Cycled Tin Oxide Gas Sensors 1989
- Editors’ Choice—Review—Solid-State
Electrochemical Carbon Dioxide Sensors:
To cite this article: W M Sears et al 1989 Semicond. Sci. Technol. 4 351 Fundamentals, Materials and Applications
Suresh Mulmi and Venkataraman
Thangadurai
0268-1242/89/050351
$02.50 + 09 @ 1989 IOP Publishing Ltd 351
W M Sears et a/
tage. Generally, a 5 V bias is applied to the heater coil Figure 1. The conductance of a tin oxide gas sensor as
which corresponds to a sensor operating temperature a function of time under different cycledheater
between 300 and 400 "C [ 1,191. This has been esti- voltages in 200 PPM ethanol. Maximum heater voltage
is: (A) 5.0 V; (B) 6.2 V; (C) 7.2 V. Curve (D): sinusoidally
mated by direct measurements of tin oxidesensor varying heater voltage, normalised to 1.0. '
conductance in a small oven equipped with a thermo-
couple.
The sensor surface temperature is affected not only
3. Results
by the heater voltage but also by Joule heating caused
by the current used to measure the conductance and by
3.1. General effects
the exothermic nature of inflammable gases the sensor
is detecting. This makes knowledgeof the exact surface In figure 1 we see the effect of the maximum heater
temperature under all conditions that much more diffi- voltage and therefore the maximum temperature on the
cult. response of the Taguchi sensor to 200 PPM of ethanol
The electrical leads are mounted to metal pins in a vapour. The heater voltage for the three curves shown
resin container which can be plugged intostandard was varied from a value close to zero to a maximum of
sockets. The container has upper and lowerstainless 5.0, 6.2,or 7 . 2 V . Thebrokencurve showingthis
steelmeshto allow gas to diffuse in buttoprevent sinusoidal variation has been normalised, and only one
flamesfromspreading outunderconditions of high period (50 S) of the waveform is shown, as the curves
concentration of inflammable gas. representasteady-stateresponse which is attained
After applying the heater voltage the sensor takes a within two cycles, after the injection of gas. After this
while to come to an equilibrium conductance. There is the curves are quite reproducible until the gas concen-
an initial surge in conductance as organic contaminates, tration is changed again. One may note that the right
adsorbed gases, and water vapour are burned or driven and leftmost values on the figure match.
off. After a few minutes a lower equilibrium value is The most obvious trend in the three curves is the
attained. The sensors tested here were all allowed a few separation into two peaks, one on the heating and one
hours to come to equilibrium before testing. on the coolingsides of the sinusoidalheating curve.
According to Figaro [18], the sensor is designed as The valley at about 20s represents loss of response at
an alcohol or toxic gas detector (e.g. CO) but it will higher temperatures. This loss is related to the catalytic
respond to a number of gases as shown below. reaction rate of the vapour [ll,20,211. At lower tem-
Inourexperimentstheheater was driven by a peratures the reaction rate is low, but rising, and the
HewlettPackardfunctiongenerator(model 3310B) partialpressure of the reducingagent in the local
which allowed theapplication of variouswaveforms, atmosphere is unchanged. This means that the sensor
frequenciesand DC offsets to the heater. The heater response (conductance) is limited by the reaction rate.
voltage was continuously cycled up and down in tem- At the higher temperatures the reaction rate is much
perature in a periodic fashion. For example, an offset higher and this keepsthe local pressure of reducing
sinusoidal voltage was applied to vary the heater vol- vapour low. In this case the sensor response is limited
tage from 0.2 to 6.2 V every 50 S (figure 1). by the diffusion of gas towards the tin oxide. It is also
The changes in conductance of thesensor were possible that at high temperatures the reducing vapour
measured at an applied bias of 5 V across the tin oxide. begins to burn (combine with oxygen in the air) slightly
Digital electrometers were used to simultaneously removed from the tin oxide surface. This would have
measure the voltage and current and this allowed the the same effect of decreasing the amount of vapour
real time measurement of conductance changes using a available to react with chemi-adsorbed oxygen on the
computer data acquisition system. tin oxide surface.
The sensorwas tested inside a lOOL box where gases The conductance of thesensorreachesan initial
and vapours were tested by direct injection. Laboratory peak at approximately the same temperature for the
air was used as the clean air baseline. threecurves:atabout 5.0 V forthe 5 V maximum
352
General characteristics of thermally cycled Sn02gas sensors
353
W M Sears er a/
4 <
- The ethanol response occurs at slightly lower tem-
0
v
2 - peratures (3.9V heating, 3.1V cooling and 3.5V
2 average as seen before) and there is a shoulder on the
1
heating peak
notseen
for
the
other twogases.
0 Interestingly, this shoulder is at the same position as
0 100 200 300 4 00 theheatingpeaks of hydrogenandacetone.These
Time I s 1
features can serve as selection criteria in the use of tin
Figure 4. As figure 3 with (A) 500 PPM hydrogen; oxide gas sensors.
(B) 200 PPM ethanol; (C)200 PPM acetone; The type of waveform used in the thermal cycling of
(D) sinusoidal heater voltage, 6.2 V maximum.
the gas sensor can also have a significant effect on the
selective nature of theresponsealthoughthe main
7.2 Vis used and reflects the high temperatures needed featuresjust discussed remain. For
propanethe
to oxidise these gases. There is some structure, how- reponseafterthe initialheating tendsto follow the
ever, consisting of a shoulder at about 5.2 V in the two shape of the heater voltagewaveform.Forasquare
gases. Forpropane,thisshoulder is of thesame wave (0.2 to 6.2V switching, 50s period) we observe
strengthastherest of thepeak giving anear flat thatafteran initialheatingperiod of about 8 S the
response over a range of temperatures. response becomes flat and remains so until the heater is
Carbon monoxide (500 PPM) produces more struc- turnedoff.Thestructure,admittedlysmall,seenfor
ture (figure 3, curve C) namely a heating peak at 2.0 V the sine wave largely vanishes. This flatness is due to
and a cooling peak at 3.1 V. This gives an average of the high reaction temperatures needed for the complete
2.6 V and represents the lowest temperature (highest oxidation of propane combined with the uniform tem-
reaction rate) of all gases tested. This is also reflected perature of the square wave over much of its cycle. A
by the fact that the conductance of the sensor at its triangular waveform shows much the same response as
lowest temperature is only slightly lower thanthe the sine wave but with a well defined peak for propane
cooling peak which is in sharp contrast to all the other at the maximum temperature in the cycle.
gaseswhoseresponsealmostvanishesas the sensor For the gases and vapours reacting at lower tem-
cools toroomtemperature.The tin oxidesensor, peraturesthechangesduetothewaveformare less
havingjust beenthermallycleaned, will respond to dramatic. As the sine wavehas no sharp edges, the
CO at very low temperatures (near room temperature temperature can more easily follow the heater voltage
in fact) for a very long time, but eventually adsorbed (the relationship is, of course, not linear) and is more
gases and water vapour will poison the surface. Thus useful for the discrimination of identifying features. It
the sensor cannot be used reliably at low temperature might be useful to design a waveform that would give a
for extended times and many commercial C O sensors linear ramp in temperature, by compensating for the
involve thermal cycling. non-linearrelationshipbetweenheatervoltageand
There are also features at 4.4 and 1.7 V of the CO sensor temperature. The square wave has the advan-
response. This complexity must relate to the changing tage of ease of generation and usefulness in the selec-
nature of the catalytic reaction of CO on the tin oxide tivedetection of gasessuchas propaneandcarbon
surface,and possibly tointeractions with adsorbed monoxide.
water vapour.
In figure 4 we have three very similar curves, which
3.2. Concentration dependence
imply a similar degree of catalytic reaction for hydro-
gen (500 PPM)andtheorganicvapoursinvolved, Figures 5 to 9 display the concentration dependence of
200 PPM ethanol and 200 PPM acetone. They all have sensor response to different reducing gases. The heater
strong well defined heating and cooling peaks. voltage (broken curve) is varied sinusoidally between
For hydrogen these peaks occur at 4.7 and 3.8V 0.2 and 6.2V. The responsesrepresenttheperiodic
with an average of 4.2 V. Acetone gives almost identi- (50 S period)butsteady-statevariation of thecon-
cal positions at4.7 and 3.7 V with an average also about ductance of the sensor in the presence of the various
4.2 V . The most significant difference between them, concentrations stated. Curve Ain the five figures repre-
besidestheirabsolute sensitivity ontheTaguchi gas sents the conductance due to laboratory air, before test
sensor, is the relative response at the maximum heater gas has been injected. Table1 lists conducatance values
temperatureandtheintensityratiobetweenheating takenfromthese figures, andothersnotshown,for
and cooling peaks.Thehydrogenresponseremains variousgasesandconcentrations.Laboratoryair is
relatively high between the peaks suggesting that there listed as zero concentration (G,)).
354
General characteristics of thermally cycled SnO, gas sensors
0 to 5 s 7 to 12 S 18 to 22 S 27 to 32 S 43 to 47 S
Concentration
(PPM) AG
AGIG,,
AG AGIGo AG AGIGo AG AGIGoAGf AG Go
Propane
0 0.18 - 22.3 - 36.1 - 6.2 - 0.01 -
10 0.0 0.0 0.6 0.03 0.2 7.6 4.3 0.7 0.0 0.0
50 0.38 2.1 24.6 1.1 28.4 0.8 14.0 2.3 0.0 0.0
200 1.46 8.1 63.5 2.9 66.9 1.9 34.2 5.5 0.09 9.0
1000 4.45 25 131 5.9 146 4.1 74 12 0.40 40
CO
0 0.1 - 5.0 - 18.0 - 3.0 - 0.1 -
10 2.5 25 42 8.4 18.0 1.0 4.8 1.6 0.0 0.0
50 45 450 100 20 39 2.2 23.5 7.8 6.4 64
200 167 167037 186 78 4.3 84 28 64 640
500 317 317059 295 115 6.4 157 52 142 1420
H?
0 0.1 - 5.0 - 18.0 - 3.0 - 0.1 -
10 4.9 49 87 17.4 43 2.4 10.230.7 0.5 5.0
50 38.3 383 203 40.6 106 5.9 117 39 9.0 90
200 106 106082 408 241 13.4 306 102 35 350
500 166 1660 651 130 426 24 555 185 62 620
1000 223 2230 815 163 588 33 767 256 90 900
Ethanol
0 0.1 - 5.0 - 18.0 - 3.0 - 0.1 -
1 0.3 3 25 5.0 12 0.7 3.0 1.0 0.0 0.0
5 5.2 52 78 15.6 32 1.8 14.3 4.8 0.25 2.5
20 40 400 179 35.8 60 3.4 49 16.3 3.5 35
100 208 2080 656 131 139 7.7 218 73 35.1 351
200 340 3400 886 177 192 10.7 339 113 66.7 667
500 680 6800 1897 379 321 17.8 839 280 172 1750
Acetone
0 0.1 - 5.0 - 18.0 - 3.0 - 0.1 -
1 0.0 0.0 20 4.0 11.0 0.6 2.0 0.7 0.0 0.0
5 0.7 7 92 18.4 33 1.8 14.0 4.7 0.0 0.0
20 5.8 58 249 49.8 68 3.8 50 16.7 0.9 9
100 33.2 332 743 149 144 8.0 193 64 7.8 78
200 75 750 1153 231 195 10.8 350 117 20 200
500 139 1390 1580 316 319 17.7 665 222 46 460
Note: Go is the base conductance at zero concentration of reducing gas and AG is the increase in conductance
observed when gas is added. The units used in the table are ,USand represent the maximum value in the cycle
time range shown.
Thetable splits thethermal cycle upinto five muchlarger in the wings duetothe very low con-
regions or windows. The conductancesused in the ductance of the sensorin laboratory air in this region of
calculation of variables in the table represent themaxi- the cycle.
mumvaluewithineachwindow. AG is listed as the Figure 6 shows the response to concentration varia-
change in conductance (conductance at a set concen- tions of carbon monoxide. In contrast to propane, the
tration minus the value at zero) for the various concen- build-up of response in the wings, the areas of negli-
trations except zero where G,, is listed instead. AG/Go gible clean air response, is dramatic for carbon monox-
is the ratio of this change to the zero concentration ide with a significant change in shape with concentra-
value. tion. This shift in shape makes identification of carbon
Examining the sensor response for propane (figure monoxide difficult at low concentrationsalthoughat
5 and table 1) we see that there is very little change in higherconcentrationsthe large low-temperature res-
shape of the conductancecurvefrom low to high ponse is areliableindication. The rapidincreases in
concentrations.Theresponse follows the increasing responsefortheheatingpeakas well asthe low-
temperature,asrepresented by theheatervoltage temperature regions is reflected in table 1 in the very
curve, with a slight shift to the cooling side. There is large sensitivities shown, as high as a conductance ratio
also a widening of the response; the wings are increas- of 3000. This means that thermally cycling the Taguchi
ing faster than the centre. The latter can be seen more sensorcangreatlyincreasethesensitivity(measured
clearly in table 1 although the effect is slight for pro- early in the cycle) and the selectivity (measured late in
pane. the cycle).
Althoughtheabsoluteconductancechangesare Theconcentrationvariation of hydrogen gas in
larger at peak response, the sensitivity ratio AG/Go is figure 7, also shows enhanced sensitivity in the wings.
355
W M Sears et a/
20
0
0
Time ( S ) Time l S )
Figure 5. The conductance of a tin oxide gas sensor Figure 8. As figure 5 for ethanol. (A) clean air;
as a function of time (50 S period) for various (B) 20 PPM; (C) 100 PPM; (D) 500 PPM; (E) heater voltage.
concentrations of propane under an applied sinusoidal
heater voltage (6.2 V maximum). (A) clean air;
(B) 50 PPM; (C) 200 PPM; (D) 1000 PPM; (E) heater
voltage. the response due to carbon monoxide. This problem in
selectivitycan bealleviatedsomewhat by theuse of
responseratios of theconductancesat two or more
points in the thermal cycle.
Figures 8 and 9, giving the response of ethanol and
5 i 3
acetone respectively, are very similar. They both have
* strongheatingpeaksfollowed by deep valleys and
significantly weaker cooling peaks. The difference in
W
g 2 overall shape caused by the rapid rise of the cooling
c
U
peak with concentration is also common to both. Thisis
v
0
-< reflected in the sensitivities listed in table 1.
U
2 -
1 The maindifference is thelowertemperature of
1 reaction of ethanol vapour which results in higher wing
sensitivities. The significantly higher low-temperature
0 response of ethanol also means that it will interfere
0 10 20 30 40 50
Time l S )
more seriously with the selective detection of carbon
monoxide in that region.
Figure 6. As figure 5 for carbon monoxide. (A) clean In figure 10 we examine in more detail the issue of
air; (B)50 PPM; ( C ) 200 PPM; (D) 500 PPM; (E) heater
voltage. sensitivity
variation
fromregion to region in the
thermal cycle. Plotted is the log to the base ten of the
change in conductance to added ethanol vapour versus
Although the response atlow sensor temperatures (the
the log of the concentration of the added vapour. This
45 S mark) is relatively low compared to the rest of the
is done as it is often found that the conductance change
response, at high hydrogen concentrations it can rival
I I I l 17
Ll
i6 20 I
-
I
,
I *
E
-\
'
I I I 17
*, \
P 6
I \
0
/ \ 1
d l.
0
c
v
S 4
0
U
2
1
0
0 10 20 30 40 50
Time (S)
Tlme Is1
Figure 7. As figure 5 for hydrogen. (A) clean air;
(B) 50 PPM; (C) 200 PPM; (D) 1000 PPM; (E) heater Figure 9. As figure 5 for acetone. (A) clean air;
voltage. (B) 20 PPM; (C) 100 PPM; (D) 500 PPM; (E) heater voltage.
356
General characteristics of thermally cycled SnO, gas sensors
I I I 4. Discussion
357
W M Sears et a/
We assume this reaction is also irreversible, with a rate voltagedecreasesbecause of arelatively high series
constantk2, as theproduct(RO) will desorb. In resistance this effect is less certain).
equation (3) the reducing gas removes chemi-adsorbed This affects our equation as the rate constants are a
oxygen,frees upanelectroniccarrier,andthereby function of temperature. We can write
increases the conductivity of thesemiconductor gas
sensor. The two reactions (equations (2) and (3)) there- k = ko exp ( - E,/ KBT) (9)
fore act in opposite directions and for a given concen- where ko is the value at infinite temperature, E , is an
tration of reactants will reach a steady state, involving activation energy, K B is the Boltzmann constant and T
the continuous consumption of oxygen andreducing is the temperature. This leads to
gas.
Using the law of mass action we can rewrite these A Ink E,
equations as currents
" --
(10)
AT KBT'
11= k,Nt[O,l (4)
which shows that with increasing temperature the rate
of change of the log of the rate constantwith respect to
temperaturedecreases,although it always remains
I , = 21, (6) positive. Given that we suspect a direct, although non-
and therefore linear dependence of the temperature on the reducing
vapourconcentration,therate of change of k with
respect to [R] should also decrease with increasing [R]
and should always be positive.
So given that the rate constants are function
a of [R]
whereNe is theconcentration of electronsandthe we can write
square brackets denote the concentration of enclosed
compounds.Wecandirectlyidentify Ne with the A log k, A log k,
conductance change A G (within a constant of propor- n=o*5(Alog[R]-Al~g[R]+1)' (11)
tionality and assuming no changes in carrier mobility)
as theabovetreatment gives zerocarriersforzero From the above argument, with increasing [R] the first
concentration of reducing gas. [R] can also be directly term increases n but the second tends to decrease it.
related to volume gas concentration expressed as parts Assuming that the temperature variation of the rate
per million. constantduetothe reducingagent dominates, we
We will now examine the effect of equation (7) on obtainslopes greaterthan 0.5 and possibly much
the log-log slope of conductance versus concentration greaterthan 1. Duetothe effect of equation (lo),
in equation (1) (A log N J A log[R] = n ) under various however, at higher temperatures the slopemay start to
conditionsandassumptions.First we assume that k, decrease,although still remainhigherthan 0.5. This
and k, are constant (fixed temperature), that [O-] is effectshouldbegreatestfor high [R] but low or
constant (relatively low concentration of [R]), and that moderate temperature and therefore should explain the
[O,] is constant (the atmosphere represents an infinite high slopes seen in the wings of the conductance curves.
source). For this case we find that n = 0.5. This repre- Withincreasing
concentration of reducing
gas
sents the situation for the valley section of the ethanol another effect is seen. Because of the higher reaction
data (figure 10, curve C). rate the concentration of the reducinggas at the tin
If we allow [O-] to be a function of the reducing oxide surface begins to drop with increasing tempera-
vapour concentration, which we expect to happen at ture, as the gas is consumed faster thanit can diffuse in;
very high concentrations due to the rapid consumption it is also possible thatthe gas begins toburnsome
of chemi-adsorbed oxygen, we obtain distance removed from the surface. In both cases the
effect is to decrease the amountof vapour able to react
with chemi-adsorbed oxygen on the tin oxide surface at
the higher temperatures. Therefore we can write
358
General characteristics of thermally cycled SnO, gas sensors
than at low gas concentrations. This will have a slight [6] Morrison S R 1987 Sensors and Actuators 11 283-7
[7] Clifford P K and Tuma D T 1982-83 Sensors and
tendency to reduce the log-log slope. Actuators 3 233-54
We thus see that as well as the obvious advantages [8] Clifford P K and Tuma D T 1982-83 Sensors and
of thermal scanning with respect to selectivity (shape) Actuators 3 255-81
thereare also greatadvantages of sensitivity if the [9] Heiland G 1982 Sensors and Actuators 2 343-61
measurement is performed at the correct point in the [lo] Morrison S R 1987 Sensors and Actuators 12 425-40
cycle. [Ill Yamazoe N, Kurokawa Y and Seiyama T 1983 Sensors
and Actuators 4 283-9
[l21 Lalauze R, Bui N D and Pijolat C 1983 Proc. lnt.
Meeting Chemical Sensors at Fukuoka, Japan
Acknowledgments (Analytical Chemistry Symp. Ser. 17) (Amsterdam:
Elsevier) pp. 47-51
We wish to thank Halitec Industries for their sponsor- [l31 Le Vine H D 1975 US Patent no 3 906 473
ship of this research and in particular Vern Moen for [l41 Eicker H 1977 US Patent no 4 012 692
many valuable and encouraging suggestions. We also [l51 Owen L J 1980 US Patent no 4 185 491
[l61 Advani G N, Beard R and Nanis L 1983 US Patent no
acknowledge the financial support of an IRAP Grant 4 399 684
from the National Research Council of Canada. [l71 Bukowiecki S, Pfister G , Reis A , Troup A P and Ulli
H P 1986 US Patent no 4 567 475
[l81 Figaro Engineering Inc. Semiconductor Gas Sensors
References (technical information)
(191 Watson J 1984 Sensors and Actuators 5 29-42
[l] Oyabu T, Ohta Y and Kurobe T 1986 Sensors and [20] Baresel D, Gellert W, Sarholz W and Scharner P 1984
Actuators 9 301-12 Sensors and Actuators 6 35-50
[2] Torvela H , Romppainen P and Leppavuori S 1988 [21] Gentry S J and Jones T A 1986 Sensors and Actuators
Sensors and Actuators 14 19-25 10 141-63
[3] Lee D D, Sohn B K and Ma D S 1987 Sensors and I221 Lantto V, Romppainen P and Leppavuori S 1988
Actuators 12 441-7 Sensors and Actuators 14 149-63
[4] Windischmann H and Mark P 1979 J . Electrochem. [23] Pijolat C and Lalauze R 1988 Sensors and Actuators 14
SOC. 126 627-33 27-33
[5] Thornton E W and Harrison P G 1975 J . Chem. Soc. [24] Matsuura Y, Takahata K and Ihokura K 1988 Sensors
Faraday Trans. I 71 461-72 and Actuators 14 223-32
359