FEASIBILITY OF USING OSCILLATORY CATALYTIC OXIDATION PHENOMENON FOR SELECTIVE CARBON MONOXIDE SENSING HEIKKI TORVELA

Tin dioxide based sensors with different additives were constructed and tested in air environment containing carbon monoxide. Conductance oscillations were observed in samples containing palladium but not in those without. Oscillations occurred at temperatures ranging from 150℃ to 320℃. Within this temperature region the range of CO concentrations at which oscillations appeared became higher as the test temperature increased. The lowest CO concentration at which oscillations were observed was 200 ppm and the highest 10000 ppm.


INTRODUCTION
Small amounts of CO together with oxygen have been found to cause conductance oscillations in tin dioxide based semiconductor gas sensors1'2'3.These oscillations have been observed when sensors contain noble metal catalysts.The domain of appearance has been in the temperature range of 150C to 300C and in the CO concentration range of 500 to 8000 ppm.
It has also been noted that at any particular temperature the period and the amplitude of the oscillations depend on CO concentration.These observations have evoked the question as to whether this phenomenon could be put to use as the operation principle of a selective CO sensor.
In this study the applicability and reproducibility of SnO.-based sensors is evaluated so as to establish the feasibility of using the oscillatory oxidation phenomenon as the basis of CO sensing.For this purpose sensors of different compositions were manufactured, some with addition of palladium and some without.The sensors were fabricated using thick film technology.The base material was SnO2 to which Pd was added as the catalyst.Oscillation characteristics of these sensors were investigated.Comparisons were made between sensors with the same composition but sintered at different temperatures.The ranges of CO concentration in which oscillations occurred at different sensor operating temperatures were determined.Measurements were carried out in synthetic and ambient air environments so as to evaluate the importance of the environment on the oscillation response of the sensors to CO.
The oscillations observed can be attributed to the oscillatory catalytic oxidation of CO.This aspect is discussed in terms of CO oxidation on supported noble metal catalyst.

EXPERIMENTAL PROCEDURE
The constituents of the sensing material were mixed with organic component to form a paste capable of being screen printed.One composition was similar to that used by S. Kanefusa et alwhich was SnO2 containing 1 wt% PdC12, 1 wt% Mg(NO3)2, 5 wt% ThO2 and 2 wt% SiO2.Two other compositions were similar except that one of them was prepared without ThO2 and the other without PdC12.The fourth composition contained only SnO2 with 2 wt% SiO2.
Sensors were fabricated using thick film technology4.The final sintering of the gas sensitive layer was carried out in a belt furnace at peak temperatures of 600C and 850C.
The sintering time was 15 min.
The current responses of the sensors were measured at temperatures ranging from 150C to 320C.In the case where temperature control was accomplished by a platinum resistor printed on the reverse side of the substrate supporting the sensor, the sensors were placed in a glass chamber.A temperature-controlled oven with a CrNi-Ni thermocouple was used to calibrate the heating resistors.This oven was also used in some response measurements to ensure identical thermal conditions for the samples being compared.
The concentration of the gas was adjusted by mixing 1 vol% of CO in N2 with synthetic air in a gas blender.Synthetic air was used to eliminate the influence of water vapour.The water content in the synthetic air used was around 10 ppm.In some experiments normal ambient air from the compressed air delivery system was used so as to evaluate the effect of water vapour.
In order to record responses and oscillations the sensors were connected, 4 to 7 at a time, with a 10 V DC voltage source and series resistors over which the voltages were recorded.

RESULTS
Oscillations occurred with sensors containing Pd at temperatures from 150C to 320C.Oscillatory behaviour varied widely with gas environment and temperature.Oscillations were found to be statistical in nature.Some variation around the mean value was observed in frequency and amplitude.The oscillatory behaviour seemed to be associated with the presence of Pd since oscillations were not observed when sensors did not contain Pd.Other additives did not greatly influence sensing characteristics.The oscillatory behaviour with and without ThO2 were very similar.
For the sensor sintered at 600C (Sample 1) at the test temperature of 150C the oscillations occurred at the concentration of 200 ppm CO in synthetic air.The amplitude was 2.5/A, or 33% of the maximum value of the oscillating current, and the period was ---3 min.For the sensor sintered at 850C (Sample 2) oscillations occurred in the CO concentration range of 200-400 ppm.Only at a CO concentration of 400 ppm were the oscillations regular with a period of ---1 min and amplitude of 0.6/A, or 75% of the maximum value of the oscillating current.At the temperature of 170C the range where the oscillations occurred was 200-400 ppm of CO in synthetic air with Sample 1 and 400-800 ppm with Sample 2. The periods at 400 ppm of CO were ---3 min and 1 min in Samples 1 and 2, respectively, and amplitudes as indicated in Table I and II.The oscillation waveform of Sample 2 at a CO concentration of 800 ppm is shown in Figure 1, which also shows the cessation of oscillation when the CO concentration is increased.
Figure 2 shows the current values obtained at 200C.The recordings were made at CO concentrations of 1500 and 2600 ppm.The oscillation frequency is lower for the sensor sintered at the lower temperature.The periods at 1500 ppm of CO were 25 s and 12 s, respectively, in sensors sintered at 600C and 850C.The corresponding amplitudes were 3.8 /A and 3/A or 37% and 60% of the maximum value of the oscillating current.The oscillation patterns were different (Figure 2a).
The oscillations ceased at a CO concentration of -2500,ppm for the sensor sintered at 600C.At higher concentrations this sensor showed a steady current reading which increased as the CO concentration increased.The period of the oscillations of the sensor sintered at 850C was 13 s at a CO concentration of 2600 ppm (Figure 2b).For this sensor the oscillations continued up to a CO concentration of 4500 ppm with the period and amplitude 15 Z FIGURE Oscillation waveform of sample 2 at 170C in synthetic air.There is a cessation of the oscillation when CO concentration is increased from 800 to 1000 ppm.
increasing.This can be seen in Figure 3 which shows the dependence of the oscillation amplitude and period on CO concentration for Samples 1 and 2.
Generally it was noticed that the frequency of the oscillation was higher at higher test temperature (Table III and IV).Also the range of concentrations in which the oscillations appeared became higher with increasing test temperature as shown in Figure 4.For the sensor sintered at 600C, for example, oscillations at the test temperatures of 300C and 320C occurred at the CO concentration of 10000 ppm or 1%, which was the highest concentration available with the experimental set-up.These oscillations were very quick, the shortest period being half a second.The waveforms of the oscillation varied from a gentle gradient as seen in curve 1 in Figure 2a to a steep cut-off and onset.Even if the period was long the start-up and decline of the impulse could happen very quickly as is apparent from Figure 1.Between these two types a smooth oscillating waveform appeared, an example of which is shown in Figure 2b.
The oscillations recorded when compressed ambient air was the carrier gas had clearly higher frequencies than those with synthetic air as is apparent from Tables III and IV.The amplitudes of the oscillations remained essentially the same.The difference in frequency is believed to be caused by the influence of water vapour.The sensors without Pd addition showed a steady current reading the value of which in ambient air was about twice that in synthetic air.
The values of the test temperatures and the ranges of the concentration of CO in synthetic air in which oscillations appeared have been outlined in Figure 4 for two types of sensors with the same composition but different final sintering temperatures.It,can be concluded that the oscillations tend to occur at lower concentration ranges for the sensor sintered at 600C.At all temperatures and concentrations the oscillation period was longer for this sensor compared to that for the sensor sintered at 850C as can be seen from Tables III and IV.The oscillatory responses to CO in air were different for sensors from different fabrication batches with different sintering temperature even if the compositions were the same.Within the same batch the results of both frequency and amplitude measurements between two sensors appeared to be similar with an accuracy of about 10%.
The reproducibility of the repeated experiments on successive days was generally within 25%.

DISCUSSION
The catalytic oxidation of CO on supported noble metal catalyst is characterized by dual dependence of reaction rate on CO concentration.This rate has a maximum value at a certain CO concentration, perhaps some percent of CO in 02.When CO concentration increases over that concentration the reaction order with respect to CO changes from positive to negative.
Oscillations can occur when the ratio of CO and oxygen concentrations is in the vicinity of that critical value.
According to E. McCarthy et al the rate Ro of catalytic oxidation of CO on supported Pt catalyst can be described as follows: 1/k(CO) + (CO)/k2 where (CO) is the CO concentration, kl is the rate constant for the reaction CO + O(a CO2, k2 is the rate constant for the adsorption of oxygen and O(al denotes the active chemisorbed form of oxygen on Pt.
This expression shows the maximum as a function of CO concentration.It also includes two distinct rate constants that dominate at different CO concentrations so that the rate determining step of the overall reaction depends on CO concentration.Of the rate constants k can be sensitive to crystallite size of the catalyst metal.
The rate of CO oxidation on supported Pt catalyst at different temperatures as a function of CO concentration was determined by E. McCarthy et al5.Their rate curves had maxima in the temperature range of 200C-250C approximately.This is comparable to the range in which oscillations were observed in this study, namely about 150C-300C.
It has also been observed that the maximum value of the CO oxidation rate is obtained at a higher CO concentration when the substrate temperature becomes higher5'6'7.A similar effect was found in this study-the CO concentration ranges in which the oscillations occurred were higher at higher temperatures as shown in Figure 4.In the oscillatory oxidation reaction of CO the adsorption and desorption steps can be critical.For example in the theory described above the dissociative adsorption of oxygen on noble metal catalyst is required to start the reaction.These adsorption and desorption steps can be influenced by other species adsorbed on the surface.In ambient air there is water vapour always present.The influence of water vapour can be the reason why the oscillations were quicker with CO in ambient air than with CO in synthetic iir.Water vapour is known to increase the rate of CO oxidation on a solid catalyst.Water vapour also greatly enhances the sensitivity of semiconductor gas sensors to CO8.
A typical feature of sensors manufactured in this study was the variation of the oscillatory response to CO in frequency and amplitude and in some cases also in CO concentration where the oscillations appeared between different batches.This could be attributed to different particle and surface structures of the Pd catalyst.The difference in the oscillation frequency, for instance, was very clear between sensors sintered at different peak temperatures which were 600C and 850C.
How the oscillatory oxidation of CO by hetereogeneous catalysis on sensor surface manifests itself as the oscillation of conductivity of the SnO2 base material has not yet been unambiguously explained.In the present study of current oscillations the sharp decline of the current pulse and the quick onset of the current increase appear to be salient features.They mean quick decrease and increase in sensor conductance.
It has been suggested that the conductance oscillations could be due to the variation of temperature caused by the catalytic reaction.In the context of the oscillatory catalytic oxidation reaction of CO the thermal nature of the conductance variation must be excluded because this oxidation reaction has been pointed out to be isothermal under relevant conditions 5,9.
The rapid changes observed in the conductance cannot be explained by any slow process like electron transfer between surface species at particle boundaries and the material.Such transfers can change the conductivity by changing the heights of the energy barriers caused by surface charges at boundary surfaces between particles.But these charge transfer processes are much too slow at the test temperatures used in this study.
The barrier energies and consequently the conductance could, however, be changed even quickly by changes in the donor concentration of the SnO2 material.The donor con- centration could in turn be affected by changes in the stoichiometry due to oxidation or reduction reactions.Yet the quick oxidation of the SnO2 surface by oxygen adsorbed directly from the atmosphere is not expected to happen, because this type of oxidation of the surface of SnO2 has been found to be a slow process1.
One explanation for the quick decrease of the conductance could be the fast oxidation of the surface caused by dissociated active oxygen.This form of oxygen could be formed in a spill-over reaction over Pd (or any other noble metal) catalyst.After this step the oxidation of CO would start rapidly and result in the formation of CO2 residing adsorbed at the surface.The reacting CO would partly reduce the SnO2 surface.After the initial stage the rate of the CO oxidation reaction would decrease, as seen in Figure 1.When the saturation condition of surface CO2 is reached quick desorption of CO2 would happen giving way to a fast reoxidation step.
That the dissociated form of oxygen plays a role in the oscillation process has been demonstrated by quickly dropping the temperature of SnO2 gas sensors from 380C to 125C11.These sensors did not contain any noble metal catalyst.After the drop in temperature current oscillations appeared at 125C and decayed away slowly.This thermal treatment is believed to leave dissociated oxygen ions (O-) formed on the surface at 380C even at 125C.This dissociated form of oxygen slowly vanishes which manifests itself as the slow decay of the oscillation.
Gases other than CO which show oscillatory behaviour include hydrogen and ammonia.
According to Kanefusa et al 2 these oscillations occur around temperatures of 90C (H2) and 370C (NH3) which are outside the range of oscillations due to CO.As the sensor compositions were also different in both these cases there should be no interference by these gases when CO is being sensed using this technique.

CONCLUSIONS
Conductance oscillations in the presence of CO were observed in sensors containing Pd at temperatures between 150C and 320C.The range of CO concentration in which conductance oscillations occurred increased with increasing test temperature.At the lowest the concentration range was around 200 ppm CO and at the highest from 8000 to 10000 ppm.
By controlling the sensor operating temperature different ranges of CO concentration could be utilized for detection.
The presence of water vapour seems to make oscillations rapid and clear.Even if it changes the frequency of oscillations the concentration ranges of the occurrence of oscillations seem to remain roughly the same.
Due to the statistical nature of the oscillations the surface and particle structure of the catalytic additive is probably critical.This means strict control of processing conditions is essential.On the other hand it allows the control of the sensing properties like the concentration range and oscillation frequency.
CO sensors utilizing the conductance oscillation phenomenon might find use, for example, in threshold value indicators or alarm systems.Better understanding of the phenomenon and control of fabrication and operation conditions could lead to feasible use of this type of sensor for more exact measurements.The specificity of the phenomenon seems to indicate that it is possible to obtain good selectivity under appropriate conditions.

FIGURE 3
FIGURE3 The dependence of oscillation amplitude and oscillation period on CO concentration in synthetic air.Test temperature was 200C and sensors were sintered at (a) 600C and (b)850C.

FIGURE 4
FIGURE 4 Ranges of CO concentrations in which oscillations occurred at different temperatures in synthetic air.

TABLE
Oscillation amplitudes of Sample at different temperatures and CO concentrations in synthetic air.The amplitude value (AV) and the maximum value (MV) of the oscillating current in #A are given.

TABLE II Oscillation
amplitudes of Sample 2 at different temperatures and CO concentrations in synthetic air.The amplitude value (AV) and the maximum value (MV) of the oscillating current in pA are given.

TABLE III
Periods of the oscillations in seconds in Sample at different test temperatures and CO concentrations.Synthetic air (S) and ambient air (A) environments are indicated.

TABLE IV
Periods of the oscillations in seconds in Sample 2 at different test temperatures and CO concentrations.Synthetic air (S) and ambient air (A) environments are indicated.