Gas Sensors

This invention relates to gas sensors. In particular to hydrogen sensors.

Extensive recent research into devices for the detection of inflammable or toxic gas in air has been centred on tin(IV) oxide (SnO.sub.2) based thick film or sintered powder type sensors, these are commonly referred to as SnO.sub.2 gas sensors. The mechanism of operation of such transducers relies on the conductivity changes experienced by the n-type semiconducting material when surface chemisorbed oxygen reacts with reducing gases such as carbon monoxide (CO) or methane (CH.sub.4) at elevated temperatures. For carbon monoxide the overall reactions occurring on the SnO.sub.2 surface can be written simply:

where e.sup.- represents a conduction band electron. In the absence of reducing gas (e.g. CO), electrons are removed from the semiconductor conduction band via the reduction of molecular oxygen, leading to a build-up of O- species and consequently the SnO.sub.2 becomes very resistive. When CO is introduced, it undergoes oxidation to CO.sub.2 by surface oxygen species and subsequently electrons are re-introduced into the conduction band leading to a decrease in this resistance.

The main advantages of SnO.sub.2 based sensors include high sensitivity, low cost, fast response speed and low power consumption. However, there are also significant drawbacks such as long term drift and ambient humidity and temperature effects associated with such devices. Attempts have been made to overcome this last effect by operating these devices under very accurately temperature-controlled conditions though it is important to recognise that it is the surface and not the bulk temperature which controls the response. Another problem with SnO.sub.2 based sensors is their relative lack of selectivity, since the chemisorbed oxygen (responsible for controlling surface conductivity) reacts with a wide range of reducing gases. Several approaches have been investigated with a view to enhancing specificity, including the use of filters or specific surface additives. These, along with other methods outlined briefly below have been reviewed recently by Morrison (Sensors and Actuators, 14 (1988) 19-25).

A certain degree of selectivity can be introduced by operating a SnO.sub.2 sensor at different temperatures: Firth et al (Ann. Occup. Hyg., 18 (1975) 63-68) observed that a temperature of 300.degree. C. resulted in response to the presence of CO but none to CH.sub.4, while temperatures over 600.degree. C. favoured CH.sub.4 detection. Experiments have been performed by the applicants to see if this effect can be used by operating sensors in a temperature modulated mode. While good discrimination could be obtained for CO and CH.sub.4 by operating the sensor at different temperatures, the presence of hydrogen (H.sub.2) interfered. Not only did the sensors respond to H.sub.2 at all temperatures but the response was erratic. There was also some evidence that gases were being adsorbed and then desorbed in subsequent tests (i.e. the sensors show a "memory effect").

By far the most popular method of achieving specificity is by the addition of catalysts or promoters. Researchers have reported that inclusions such as thorium dioxide confer CO selectivity, while the presence of silver improves H.sub.2 response although not removing sensitivity to CO and CH.sub.4. Such H.sub.2 sensors incorporating silver have been reported by Yamazoe et al [(J.Chem. Sec. Japan them. Lett. (1982), 1899-1902) and (Sensors and Actuators, 4 (1983) 283-289)]in report showing apparently impressive sensitivity and selectivity for hydrogen. However the sensors described have several drawbacks.

Firstly they are not as sensitive and selective for hydrogen as appears at first sight from the graphs of results since, for example, H.sub.2 concentration of 0.8% are being compared with CO concentrations of 0.02%, a forty-fold difference in concentration that dramatically effects the presentation of results.

Secondly peak hydrogen sensitivity is shifted to a temperature of .apprxeq.100.degree. C. which is very low for such sensors which are usually held at in excess of 200.degree. C. At such low temperatures as 100.degree. C. problems can arise with condensation of volatiles on the sensor, even moisture being a problem. Additionally at such low temperatures the response time for the sensors would be very high. Response times are quoted in the above mentioned reports of 20 seconds to obtain a 90% response at 200.degree. C.

Thirdly the response is not linear with concentration (see FIG. 3 of the Sensors and Actuators paper). and so complex calibration would be required.

The same authors have studied the effects of catalysts such as manganese, nickel, cobalt or copper on sensor response to various gases including CO, CH.sub.4, H.sub.2 and propane, while in most modern commercial sensors the presence of trace quantities (0.5-5% w/w) of palladium or platinum is an essential prerequisite.

There has also been described (Sensors and Actuators, 7 (1985) 89-96) the modification of tin oxide based gas sensors by inclusion of .about.15% w/w bismuth oxide (Bi.sub.2 O.sub.3) to give CO selectivity or 36% w/w aluminium silicate and 1.5% w/w palladium chloride (PdCl.sub.2) for CH.sub.4 selectivity. The CO selective sensor described was very sensitive to the amount of Bi.sub.2 O.sub.3 present. Below 15% w/w the sensor was CH.sub.4 sensitive; above 17% w/w the sensitivity to CO began to fall effectively disappearing at between 20-30% w/w.

U.K. Patent Application 2149123A describes sensors including, inter alia, Bi.sub.2 Sn.sub.2 O.sub.7 and a gas sensitive material. Although, in discussing this material, the U.K. patent application describes gas sensitivity to O.sub.2, CH.sub.4, CO, H.sub.2, C.sub.2 H.sub.4, and NH.sub.3 no indication is given as to relative sensitivities and selectivities for these gases.

Accordingly the present invention provides a method of measuring concentrations in gases comprising the use of a hydrogen sensor of the SnO.sub.2 type comprising a layer of tin (IV) oxide (Sno.sub.2) incorporating bismuth (III) oxide (Bi.sub.2 O.sub.3) characterised in that the Bi.sub.2 O.sub.3 is present in an amount less than 35% w/w based on the total amount of SnO.sub.2 and Bi.sub.2 O.sub.3 and sufficient that the sensor has hydrogen sensitivity and selectivity over the gases CO and CH.sub.4.

More particularly, in accordance with the present invention there is provided a method of measuring hydrogen concentrations in gases comprising the use of a sensor of the SnO.sub.2 type comprising a layer of tin (IV) oxide (SnO.sub.2) incorporating bismuth (III) oxide (Bi.sub.2 O.sub.3) wherein the Bi.sub.2 O.sub.3 is present in an amount in excess of 17% by weight and less than 35% by weight based on the total amount of SnO.sub.2 and Bi.sub.2 O.sub.3 and sufficient that the sensor has hydrogen sensitivity and selectivity over the gases CO and CH.sub.4.

In another aspect of the invention, the layer of tin (IV) oxide incorporating bismuth (III) oxide is deposited on a substrate from a slurry.

In accordance with another aspect of the invention, there is provided a hydrogen sensor for use in the above methods which contains a catalyst selected from one or more of the metals Ir, Pt, Ag, Ru, Au or Pd, and in which the amount of Bi.sub.2 O.sub.3 is 22.5% by weight or above based on the total amount of SnO.sub.2 and Bi.sub.2 O.sub.3.

Further features of the invention will be made evident in the claims and the following description with reference to the drawings in which:

FIG. 1 is a schematic drawing of a gas sensor substrate;

FIG. 2 illustrates sensor response as a function of increasing Bi.sub.2 O.sub.3 ;

FIG. 3(a) and 3(b) shows the response characteristic of a particular preferred form of sensor:

FIG. 4(a) and 4(b) shows response characteristics of a further particular form of gas sensor;

FIG. 5 shows comparative results of the gas sensors referred to in relation to FIGS. 3(a) 4(b) and 4(a) and 4(b)

FIG. 1 shows a SnO.sub.2 gas sensor substrate comprising a plate 1 of ceramic material (e.g. alumina), sensor leads 2 connected with platinum contact pads 3 on the plate 1, and heater leads 4 connected with a platinum heater circuit (not shown) applied to the back face of the plate 1.

For experimental purposes hydrogen selective sensors were prepared from an aqueous slurry containing SnO.sub.2, Bi.sub.2 O.sub.3 (17%-30% w/w) and, in some cases, a metal catalyst (usually added in the form of a chloride salt, 1.5% w/w). A bead of this paste was then applied across the contact pads 3 of the alumina substrate as shown in FIG. 1 (alumina plate 1 supplied by Rosemount Engineering Ltd.), left to dry in air and then sintered at 1000.degree. C. for 2 hours (using a furnace heating rate of 500.degree. C./h). Some were sintered instead at 800.degree. C. for comparative times.

In certain cases, sensors were fabricated from SnO.sub.2 /Pt powders prepared by deposition of a finely divided metal colloid. This procedure ensures that the sensor material is not contaminated by chloride ions which are released into the SnO.sub.2 bead when the metal chloride additive is reduced to the metal itself.

The colloid was prepared by refluxing a solution containing 150mg of H.sub.2 PtCl.sub.6 (Pt content=40% w/w), 150 cm.sup.3 of a 1% w/v sodium citrate solution and 450 cm.sup.3 of distilled water for a period of 5h. A very stable Pt/citrate colloid, black in colour, resulted.

The calculated concentration of metal in the colloidal system was 0.1 gdm.sup.-3.

Deposition of the colloidal metal onto the SnO.sub.2 surface was carried out via the following procedure. 1 gram of tin(IV) oxide powder was suspended in 100 cm.sup.3 of the colloid and 11.6 g of NaCl was then added with continuous stirring. The resulting conditions of high ionic strength lead to destabilisation of the metal particles followed by precipitation (a process which appears to be complete within seconds). The solution was then left stirring for 1 h before filtering off the SnO.sub.2 /Pt powder and washing thoroughly with distilled water to remove Na.sup.+, Cl.sup.- and citrate ions from the material.

A SnO.sub.2 +Ag powder was also prepared via the method outlined by Yamazoe et al, J. Chem. Soc. Japan Chem. Lett. 1982, 1899-1902 which consists of impregnating the stannic oxide with aqueous silver nitrate followed by evaporation to dryness and calcination at 600.degree. C.

The H.sub.2 sensing characteristics of these materials were then investigated after mixing into an aqueous slurry with Bi.sub.2 O.sub.3 (25% w/w) and applying to the surface of an alumina substrate via the procedure described above.

FIG. 2 illustrates the sensor response as a function of increasing Bi.sub.2 O.sub.3 (15-30% w/w) composition in the presence of 1% v/v gas mixtures of H.sub.2, CO or CH.sub.4 in air. There appears to be a gradual decline in sensitivity to Co upon increasing Bi.sub.2 O.sub.3 doping from 15% to 25% w/w whereas the loss of H.sub.2 response under the same conditions is only very slight. At compositions of >30% w/w virtually all CO signal is lost while H.sub.2 sensitivity remains reasonably substantial. However, a study of sensitivity to lower concentrations of H.sub.2 and CO (100 and 1000 ppm) indicates that results gained using 1% v/v contaminant gas mixtures do not give a true overall picture. Data listed in Table 1 show a gradual decrease in H.sub.2 sensitivity with increasing Bi.sub.2 O.sub.3 content to both gas concentrations in an analogous manner to CO. Interestingly though, a comparison of the ratios of H.sub.2 and CO sensitivities for a specified gas concentration and Bi.sub.2 O.sub.3 composition (i.e. S(H.sub.2)S(CO)) leads to the observation that H.sub.2 response disappears at a slower rate than the CO signal upon increasing % Bi.sub.2 O.sub.3. From the results listed in Table 1 it can be shown that S(H.sub.2)/S(CO) rises from 2 to 5 and from 1.8 to 9.5 for contaminant gas concentrations of 100 and 1000 ppm respectively when % Bi.sub.2 O.sub.3 is increased in the range 15 to 30% w/w. Close scrutiny of FIG. 2 shows that an SnO.sub.2 -based sensor doped with 30% Bi.sub.2 O.sub.3 (w/w) exhibits H.sub.2 selectivity in the presence of CO and CH.sub.4. However, results also indicate that such a sensor is not very responsive in the 0-1000 ppm H.sub.2 concentration region although exhibiting relatively linear changes in conductance upon exposure to 0.1-1.0% v/v H.sub.2 /air mixtures.

Attempts to enhance the sensitivity of this H.sub.2 sensor to low levels (100-1000 ppm) of contaminant gas were centred on reducing slightly the Bi.sub.2 O.sub.3 content and including a metal catalyst to confer improved selectivity.

Table 2 shows the results obtained upon doping a SnO.sub.2 /Bi.sub.2 O.sub.3 mixture of fixed proportion (23.5% w/w Bi.sub.2 O.sub.3) with various metal catalysts (1.5% w/w added as chloride salt--final amount in sensor about 1% w/w). On inspection of the above table it seems that most of the metal additives improve sensor selectivity to H.sub.2 compared with the undoped material, although this appears to be at the expense of some sensitivity. As a consequence the most promising sensors from those listed above (in terms of selectivity) are only suitable for detection of H.sub.2 inclusions in the 500-10,000 ppm region e.g. sensors 3 and 7 (Ag and Ir) in Table 2. FIGS. 3(a) and 3(b) illustrate the response characteristics of a SnO.sub.2 /Bi.sub.2 O.sub.3 (23.5% w/w) mixture doped with Iridium. Although the sensor exhibits a slight resistance drop in a 1% v/v air/CO mixture, FIG. 3(b) shows that in the 500-7000 ppm concentration range, the CO response is negligible. In contrast, H.sub.2 response is substantial and linear in this concentration region. Optimisation of the constituent proportions of a SnO.sub.2 /Bi.sub.2 O.sub.3 mixture with added platinum yielded a further sensor exhibiting enhanced sensitivity (compared to the sensor described above) as well as H.sub.2 selectivity. A "fine-tuned" Bi.sub.2 O.sub.3 composition (an empirically determined compromise between selectivity and sensitivity), along with a platinum salt results in a H.sub.2 selective SnO.sub.2 -based sensor whose response characteristics are shown in FIGS. 4(a) and 4(b). This comprised 25% w/w Bi.sub.2 O.sub.3 (a compromise between selectivity and sensitivity) with 1.5% w/w K.sub.2 PtCl.sub.4. The source of platinum did not appear to affect sensor response greatly, if anything the inclusion of other salts or finely divided colloidal platinum in place of the original salt served only to diminish sensitivity to a small extent. The linearity of H.sub.2 response for both types of sensor described above is demonstrated clearly over a very wide concentration range (up to 2 orders of magnitude), by a plot of S versus [H.sub.2 ] using log-log axes (FIG. 5). Results obtained with Bi.sub.2 O.sub.3 and Pt doped sensors are in conflict with Lee et al (Sensors and Actuators 12 (1987), 441-7) who observe enhanced CO sensitivity over H.sub.2 when incorporating a platinum promoter in their thick film transducer. These workers report S(=.DELTA.G/G.sub.o) to be 1 and 0.35 for 200 ppm levels of CO and H.sub.2 respectively at optimum operating temperature, while in our work S is 0.03 and 1.3 respectively for 1000 ppm of these gases.

A characteristic of the mixed SnO.sub.2 /Bi.sub.2 O.sub.3 sensors is their high resistivities, and consequently they may be prepared using a substrate with an interdigitated contact array in order to reduce base resistances to easily measurable values. Alternatively, Sb.sub.2 O.sub.3 (0-2% w/w) can be incorporated into a H.sub.2 sensor. This will reduce the resistivity by a factor of up to x 50 without sacrificing sensitivity or selectivity.

Another point worth noting is the difference between sensors sintered at 800.degree. C. and 1000.degree. C. If sensors are sintered at the higher of the two temperatures they tend to become "poisoned" (i.e. experience irreversible resistance drops) when exposed to H.sub.2 and in some cases CO. This problem can be easily rectified by high power "cleaning", applying power to the heater leads at about 2 W for 1 minute and such a procedure should be borne in mind should this type of sensor be incorporated into a measuring device or alarm circuit.

One of the remarkable characteristics of this sensor is its linear response over a very wide (200-10,000 ppm) concentration range.

Additionally all of the SnO.sub.2 /Bi.sub.2 O.sub.3 sensors studied had fast response times; most having a response time of less than 5 seconds to show a 100% response.

Although only single metal catalysts have been investigated it will be clear that mixed metal catalysts are within the scope of the invention.