WO1995004926A1

WO1995004926A1 - Gas sensing apparatus and sensors therefor

Info

Publication number: WO1995004926A1
Application number: PCT/GB1994/001727
Authority: WO
Inventors: Peter Mcgeehin; Patrick Timothy Moseley; David Edward Williams
Original assignee: Capteur Sensors & Analysers Ltd.
Priority date: 1993-08-05
Filing date: 1994-08-05
Publication date: 1995-02-16

Abstract

A gas sensor comprises a sensing element in the form of a porous, thick, semiconducting metal oxide layer (38). The sensor (20) is suspended by fine wires (24, 26) in a housing (22), which is part of an apparatus that includes a processing circuit for measuring sensor resistance and a circuit for controlling the working temperature of the sensor. These circuits, and characteristic of the sensor, are such that: baseline sensor resistance is repeatable to less than ± 5 %; the output signals of all sensors of a given type bear a common simple mathematical relationship to target gas concentration; the resistance/gas concentration characteristic is repeatable to less than ± 15 %; and operating temperature is consistent within ± 30 % for a sensor population having the same zero resistance. The processing circuitry is adapted for a wide variety of different sensor types.

Description

GAS SENSING APPARATUS AND SENSORS THEREFOR

This invention relates to gas sensing apparatus comprising a sensor, or sensing device, comprising a sensor element made of gas sensitive metal oxide semiconductor material, together with processing means for processing signals received from the sensor; and to such sensors p_eχ sfi.

It is well known that the electrical conductivities of metal oxide semiconductor materials are sensitive to the presence of various gases or vapours and can be used in sensors to detect their presence. See for example the documents GB-A-2 149 120; GB-A-2 149 121; GB-A-2 149 122; GB-A-2 149 123; GB-A-2 166 244 and GB-A-2 218 523; "The Tin Oxide Gas Sensor and its applications", J. Watson, Sensors and Actuators

54(1984) 29-42; "The Detection and Measurement of CO using ZnO Single Crystals", B. Bott et al, Sensors and Actuators 5(1984) 65-73; "The Role of Catalysis in Solid State Gas Sensors", S. R. Morrison, Sensors and Actuators 12(1987) 425-440; and "Electrical Conduction in Solid State Gas Sensors", J. W. Gardner, Sensors and Actuators 18(1989) 373-387.

All such gas sensors rely on the gaseous medium under observation impinging on a surface of a body of the semiconducting metal oxide material, and then undergoing some reaction with it which affects the conductance of the semiconducting metal oxide material. This conductance is detected by means of at least one pair of electrodes which are formed upon the body of semiconducting metal oxide material.

Such sensors, whilst they are inexpensive, light and robust, suffer a reputation for: irrepeatability in baseline, calibration and response law between devices; unacceptable drift of both baseline and calibration; unacceptable sensitivity to variations in relative humidity; and cross-sensitivity to different gases. Although widely used in alarm applications, they are usually considered unsuitable for applications that call for reliable quantitative results. Furthermore, it is also generally considered that devices of this type have unacceptably high power drain for extended use in a portable instrument.

We have found that the above drawbacks of gas sensors based on metal oxide semiconducting materials can be substantially reduced.

The invention provides gas sensing apparatus comprising a sensor, or sensing device, comprising a sensor element made of gas sensitive metal oxide semiconductor material, together with processing means for processing signals received from the sensor, and further provides such sensors pe se for use in such apparatus, in which at least one of the features (i) to (xvii) listed below is present, the selection of features being such that:

- the repeatibility of baseline resistance or conductance as between one sensing device and another is better than ±5%;

- the output signal of the sensors bears a simple mathematical relationship to the gas concentration for all sensors of a given type;

- the slope of the characteristic curve of the said output signal as a function of gas concentration is repeatable as between one sensing device and another to better than ±15%; and

the operating temperature of a population of sensors with uniform zero resistance is consistent within ±30°C.

The output signal, representing resistance or conductance, varies with a function of the gas concentration, this relationship being represented by the characteristic curve. The said function consists of the gas concentration raised to a power which is a ratio of low integers (e.g. , 1, ι ~ ) _<■ ^as ^^s characteristic of chemical reaction kinetics. Thus, for example, the curve may be linear or not.

The said features are as follows:

(i) The sensor element has a porous metal oxide layer of thickness greater than 50 .m, with thickness uniform across the device to better than about ±20% and repeatable between devices to better than ±20%.

(ii) The porosity of the said layer is in the inclusive range 30-60%.

( iii) The said layer is free from macroscopic flaws (e.g. cracks and/or bubbles) larger than 5 times the mean pore size in the layer.

(iv) The metal oxide particle size is preferably less than 5 itm, with no isolated particles being of dimension greater than 5% of the layer thickness.

(v) The sensor element has a metal oxide substrate, in electrical contact with the said metal oxide layer and carrying a pair of interdigitated metallic electrodes, which are made of a material (such as platinum or gold) that is appropriate both for the composition of the metal oxide layer and for the target molecule being detected. These electrodes are spaced uniformly apart by a distance in the inclusive range 1 - 300 n,m such that the electrical resistance between the said electrodes lies in a range convenient for measurement. This distance is maintained with an accuracy better than ±5%.

(vi ) The said layer is bonded to the electrodes and/or the substrate by sintering.

(vii) The substrate is of alumina, the characteristic dimensions of which are in the range 0.2 to 3 mm.

(viii) The said processing means includes a resistance measuring circuit electrically connected with the said electrodes.

( ix) The apparatus includes sensor heating means and heating control means connected with the heating means.

(x) The heating means comprises an electrically isolated heating element in electrical contact with the said metal oxide layer of the sensor.

( i) The heating control means are arranged to control the temperature of the metal oxide layer to better than ±0.1°C, and to maintain the temperature across the said layer at a uniformity of ±30°C.

(xii) The apparatus includes a housing substantially enclosing the sensor, the housing being made of material such that reaction, decomposition and irreversible adsorption of the target gas (i.e. the gas to be detected) are effectively absent, the housing having means for allowing free access of the gas to the sensor and defining an included volume in the inclusive range 30 - 3000 times the volume occupied by the sensor.

(xiii) The housing is such that any heated portion of the sensor is more than 3 mm away from the inner surface of the housing.

(xiv) The housing material is preferably impervious to the target gas.

(xv) The apparatus includes sensor support means, such that the sensor is suspended in free air by fine wires, having a typical diameter of 100 um or less, which serve both to supply power to the heater and to provide connection for the resistance measuring circuit, such wires having and electrical resistance no greater than 5% of the heater resistance; their length and diameter are such that no more than 30% of the power dissipated in the heating means is conducted down any one of the wires.

(xvi ) The apparatus includes means for measurement of the resistance of the sensor layer such that the electrical current through the metal oxide is less than 5^,_A.

(xvii) The apparatus includes means for, if necessary, linearising the response of the sensor resistance to change of the gas concentration.

The invention enables the following to be achieved: - improvement of repeatability between devices in baseline resistance, calibration, response law, response time, and recovery time;

- stabilisation against baseline and calibration drift;

- minimisation of power drain; and

- minimisation of the influence of variations in relative humidity on sensor output.

The invention will be further discussed, by way of example only including some specific examples, and with reference where appropriate to the accompanying drawings, in which:-

Figure 1 consists of two diagrams (a) and (b), showing variations in the electrical resistance of gas sensors in air and in a target gas;

Figure 2, which again consists of two diagrams (a) and (b) , shows variation of the resistance of a sensor the resistance of which increases with gas concentration, plotted against two functions, respectively, of the gas concentration;

Figure 3 is similar to Figure 2, but for a sensor the resistance of which decreases with gas concentration;

Figure 4 consists of three diagrams (a), (b) and (c), showing the effects of relative humidity changes on sensor response at low gas concentrations;

Figure 5 is a diagram showing the variation in resistance, with temperature, of heaters used in preferred forms of apparatus according to the invention; Figure 6 is an electrical circuit diagram for a heater driver circuit in a preferred form of apparatus according to the invention;

Figure 7 is a circuit diagram showing one example of a simple general electronic interface circuit for use in apparatus according to the invention;

Figures 8 and 9, which are diagrammatic and not to scale, show a typical sensor according to the invention with its support means and electrical connections, Figure 9 being a cross section on IX-IX in Figure 8;

Figure 10 is a diagram showing a response curve for a chromium titanium oxide sensor in the presence of propane; and

Figure 11 is a diagram showing a response curve for a tin oxide sensor in the presence of carbon monoxide.

Semi-conducting oxide gas sensors work by exhibiting electrical conductance controlled by molecular reactions at the oxide surface. It follows that a sensor required to offer a perfectly stable conductance in an atmosphere of constant composition and to respond sensitively when the composition changes must have a carefully crafted microstructure. The microstructure must not evolve thermally at the operating temperatures (by sintering leading to conductance change). It must also be controllable in terms of gas access, diffusion and compositional profiles of reactant and product species.

In the absence of the target gas, the resistance of the sensor depends upon its operating temperature, and is of the form: R_τ = R₀ exp (E_A/kT) 1

where k is Boltzmann's constant, T is the absolute temperature, and E_A is the energy for the thermal activation of charge carriers. R_Q is the resistance at any reference temperature. This behaviour is typical of a thermally activated conduction process, and in the scientific literature log R_τ is often plotted against 1/T or 1000/T to give a straight line.

The behaviour of the sensor resistance in the presence of the target gas is superimposed on the above temperature effect. The details differ for different sensor materials and target gases. In Figure 1 the variation of sensor resistance R with temperature T in air, at a single concentration of the target gas, is illustrated schematically. There are two types of response. For some sensors, exposure to the target gas results in a resistance increase, Figure 1(a), whilst for others the resistance in the target gas falls, see Figure 1(b).

In Figure 1, the curves A are those of resistance in air. In Figure 1(a), curve B is that of resistance in a fixed concentration of the target gas, and curve C in Figure 1(b) is again that of resistance in the target gas. Maximum sensitivity is indicated at D, so that the optimum sensor operating ter perature is T_Q. In both cases, the resistance falls as the temperature increases according to the above equation.

The difference between the resistance of the sensor in air and the resistance in a given concentration of the target gas is also a function of temperature. This can be seen from Figure 1, where in both cases the response can be seen to rise with increasing temperature to a maximum, after which a further increase in temperature results in a decrease in this resistance change, which finally reaches zero again. Typically, a sensor responds reasonably over a temperature range of 100°C or more. At the temperature (T ) of maximum response, the increase or decrease in resistance of the sensor over its normal operating target gas concentration range is usually by a factor in the range from 2 to more than 10.

It is known in the art that gas-sensitive resistors show a power-law dependence of the response on the gas concentration. For example, P. K. Clifford and D. T. Tuma, Sensors & Actuators 3 (1983) page 233, give the following relationship for the response of sensors made of tin dioxide ( CT denotes conductivity, P partial pressure) :

β(

with the exponent β varying in the range 0.25 to 0.55. Some theoretical discussion has been given of this equation (by D. E. Williams in P. T. Moseley & B. C. Tofield (Ed.) "Solid State Gas Sensors", Adam Hilger, 1987); and it has been hypothesised that the variability of β arises from a microstructural heterogeneity in the device. Clearly, such a complex response law, coupled with the variability of the exponent, would mean that devices of this type were strictly limited in their application, and of no use at all for quantitative work. This indeed represents considered opinion in the current state of the art. We have now found that, by the careful control of operating temperature (e.g. 440°C for tin dioxide) and by the use of careful and optimised sensor fabrication methods, the response law is greatly simplified, and the variability in the exponent is eliminated. We have also found that it is possible to describe a general response law which is applicable to many gases and sensor materials.

In Figure 2, the resistance R of a properly fabricated sensor is plotted as a function of the concentration of the target gas for a fixed temperature. The general form of resistance variation follows a square root law. Thus at the very lowest concentrations, the resistance change for a change in gas concentration from 0 - 100 ppm (parts per million) is greater than from 100 - 200 ppm, which in turn is greater than the resistance change for the range of concentration 200 - 300 ppm, and so on. This behaviour is true at all operating temperatures for which the sensor has a response, and not just at the temperature of maximum response.

Sensors for a particular target gas have distinct operating ranges defined by the range of target gas concentration for which the change in resistance is substantial. In Figure 2(a), the responses of a pair of sensors are shown, namely a response E for a sensor designed to operate over a range of low concentrations, with the steeper characteristic; and another, shallower, characteristic F for a sensor which is designed to operate over a wider target gas concentration range. The square root behaviour means eventually that the change in sensor resistance gets smaller as the concentration increases, and sensors are designed so as to avoid operation in the flattest part of the characteristic where they are least sensitive. This is illustrated schematically in Figure 2(b), in which the curves for the same sensors are again E and F respectively.

Two additional points can be made about Figure 2. Firstly, the effect of moving the temperature away from optimum is to reduce the slope of the "square root" characteristic of Figure 2(a), with the actual values of the various resistances following the curves in Figure 1(a). Secondly, in the first 20% of the concentration range in Figure 2(b), the resistance varies approximately linearly with target gas concentration. The resistance change is defined mathematically by the equation:

where Ra is the resistance of the sensor in air and Rg the resistance in the target gas at a concentration c. The constant K depends on the units of measurement of c and the operating temperature of the sensor only. This equation can be rearranged, because at fixed temperature Ra is constant, to show that:

RgoC^c

(neglecting the intercept when c = 0)

Figure 3 shows that similar behaviour is evident for sensors G and H, in which resistance decreases as concentration increases, eventually flattening at high concentrations C of the target gas, see Figure 3(b). The main difference from Figure 2 is that for a linear "square root" plot, the reciprocal of the sensor resistance R should be used, see Figure 3(a). Here, employing the same symbols as previously, with the constant K taking a different numerical value but with the same reliance on temperature and concentration units as before, the response is given by:

Ra-Rg = K / c ... 4 Rg ^γ

This can again be rearranged, and neglecting the intercept at c = 0, it is found that:

The gas sensitive materials in thick film gas sensors have in the past been fabricated by screen printing techniques, but these are really only convenient for preparing porous sensors up to a thickness of 10 - 20 microns. Even then, they are prone to micro-structural inhomogenuity due to solid settling in the printing vehicle. An alternative process, ideally suited for the preparation of oxides with thicknesses up to 300 microns, leads to stable performance in gas sensor applications, because it is characterised by a high degree of control over the factors that influence access of gas to the electrodes. This process consists in milling the powder to a uniform and suitable particle size; mixing this with a plastic, a plasticiser, a dispersant and a solvent in the appropriate ratio in a pot containing hard oxide media; and casting the mix under a doctor blade to form plastic-ceramic sheet. Sensor elements are formed by cutting pieces from the plastic sheet and adhering these down across a pair of interdigitated electrodes on an alumina substrate, followed by firing in a controlled atmosphere such as dry air so as to achieve a suitable final porosity, strength and adhesion.

This process, implemented with optimised process parameters, gives semiconducting gas sensors which have both a minimised response to changes in relative humidity, and maximum stability. In addition, control of the final firing stage permits a tight control over the sensitivity of the sensor. Sensor performance is also improved by the avoidance of additives or sintering aids such as frits, which are commonly employed when thick film printing techniques are employed, and which are employed in the manufacture of gas sensors of inferior performance.

The document WO92/21018 illustrates that, besides the material and its method of fabrication and microstructure, another factor that influences its performance in a practical sensor is the geometry of the sensing material and the sensor element itself. The shape of the sensing element (by way of example whether it is cylindrical, tubular or planar), the spacing of the electrodes with respect to the dimensions of the sensor material (e.g. its thickness), and the nature of the electrodes (reactive materials such as platinum, or passive ones such as gold), are also influential in the ultimate performance of the sensor.

Besides the relationship between sensor resistance and gas concentration, there are several other parameters 14

important in the performance of the sensor which similarly depend upon the nature of the material, its method of fabrication and resultant microstructure, and the configuration and nature of the electrodes making contact with it. Of practical importance are: sensor operating temperature and its stability; effects of relative humidity; and the rate of response of the resistance of the sensor to a step function change in the target gas concentration.

The mounting of the sensor in a housing is important for its thermal management, and we have also found that the construction of the housing can be influential in the response time of the sensor, particularly where strongly adsorbing gases such as chlorine and hydrogen sulphide are involved. The most important single factor appears to be the material of construction of the housing, and we have found that engineering plastics materials in unfilled form, or filled with either glass fibre or glass beads, and based upon polyphenylene sulphide, have excellent thermal stability and do not attract such adsorbing gases.

The effect of relative humidity is particularly important. In the case where the sensor is operating at constant temperature, it determines the accuracy of measurement in practice, and therefore also the usefulness of the device. Figure 4 shows the effects of change in relative humidity (RH) on the sensor resistance R at low gas concentrations; R is plotted against time t. Baseline resistance at 0% RH is indicated as R_Q in Figure 4(a) and (c), and at 100% RH as

^{n F}Ϊ9^ure 4(b). Three different patterns of behaviour are illustrated, according to the sign of the 15

gas response.

If the sensor resistance decreases with increasing target gas concentration, the effect of increasing relative humidity is invariably to reduce the resistance, as in Figure 4(c). This plot shows the effect on sensor resistance of successive exposures to dry air (0% RH) , followed by exposure to a low concentration of the target gas (at I ) before returning to dry air. The sensor is then exposed to air saturated with water at room temperature (100% RH) during which time it is exposed to the same low target gas concentration as before. The sensor is then returned to dry air and exposed successively to a high concentration of the target gas in backgrounds of dry and then wet air (at J), before finally returning to dry air.

In Figure 4(a) and (b), exposure to low and high concentrations of gas are again indicated at I and J respectively. In Figure 4(a), the same time sequence of gas exposures is applied to a sensor which shows the inverse behaviour, that is to say a resistance increase with increasing levels of both the target gas concentration and the relative humidity.

The response in Figure 4(b) is anomalous, in that the sensor response to the target gas is a resistance increase, and the effect of increasing relative humidity is a resistance decrease. In this case, the sensor baseline can be considered to be water saturated air (100% RH), rather than the dry air of Figure 4(a) and (c) .

Sensor operating temperature is important not only in its influence on response time, but also because many applications employ battery powered instruments. It is advantageous therefore that the sensor operating temperature should be as low as possible commensurate with these other factors. Besides being as low as is practical, the stability of the temperature of the sensing element is a major contributor to the stability of the resistance output of the device. Activation energies (E_A in equation 1 ) of many useful materials are in the range 0.5 to 1 eV, and thus for a stability in resistance of a per cent or so, in the absence of any target gas, the temperature of the sensor needs to be controlled to within ±1°C. Such a degree of temperature stabilisation cannot be achieved easily in small devices which are susceptible to small changes in their thermal environment (such as might arise from a change in the surrounding atmosphere or draughts) .

Temperature fluctuations can be compensated by use, for example, of a thermistor. We however prefer the use of a heater with a significant temperature coefficient of resistance, for example as shown in Figure 5, which shows the variation in resistance of three heaters having a resistance at room temperature of either 6, 14 or 20 ohms. All have an approximately constant temperature coefficient of resistance, which results in a doubling of the heater resistance for a temperature rise of about 400°C.

A constant voltage heater power supply permits less perfect compensation for ambient temperature variations than does constant resistance excitation of the heater. Constant current excitation is unsuitable because it tends to magnify changes in the ambient temperature, resulting in fluctuation of the sensor baseline. The temperature of a sensor with such a heater is uniquely defined by the value of the resistance adopted by the heater when it is excited. This allows simple electronic principles to be employed to control the sensor temperature, by including its heater in a Wheatstone bridge arrangement (half or whole bridge).

An example of such a circuit is shown in Figure 6. The sensor heater 10 forms part of a Wheatstone bridge, and the current through it is controlled so as to maintain the sensor heater at constant resistance. The value of this resistance is determined by the setting of a potentiometer VR2. With the values of resistors R4 and R5 shown in Figure 6, VR2 allows the heater to be set at any value between about 15 times and twice that of the wire-wound resistor R3. Effectively, therefore, the potentiometer VR2 sets the temperature of the sensor, which is controlled by an amplifier ICA and a field effect transistor FET1 to better than ±l^βC.

The supply voltage needs to be stable and about one volt above that specified, in order to achieve the correct sensor temperature. A small ripple in the sensor output resistance can sometimes be observed, with a period of about 20 seconds, which indicates that the sensor temperature is being stabilised by the heater driver.

Whilst this "constant resistance" excitation principle can be realised in a number of different ways, the circuit shown in Figure 6 has been shown to give excellent performance with sensor heaters having a resistance in the range from 6 to 50 ohm. Alternative components can be employed. For example, the amplifier ICA may be of the type ICL 7612, and the transistor FET1 may be of the type IRF D120 or IRF D520. The resistor R3 should be wire wound and have a power rating of 1.5 watts, although generally it dissipates less than a tenth of this. For 14 ohm and 20 ohm heaters, a supply voltage of 10V is used; for heaters of 6 ohms and less the supply voltage should be 5V.

This circuit is suitable for sensors that have a power consumption of less than 1 watt, and to a large degree it removes the dependence of the heater excitation on the immediate thermal surroundings of the sensor (including its housing). Additionally, output resistance values of the sensor can be adjusted by trimming the potentiometer VR2. This alters the operating temperature, and therefore the zero resistance: and the resistance in a fixed concentration of the target gas will change.

Drift of the sensor indication is another practical weakness of the current state of the art. We have found that the drift in sensor indication is strongly related to the current passed through the device in order to address the resistance changes. If the measurement current is kept below 10mA, and preferably less than 5mA, drift of the sensor indication becomes negligible.

Taken together, these multiple observations permit the use of an electronic signal processing circuit to deliver an output which is unambiguously related to the gas concentration, converting semiconductor gas sensors into devices for quantitative measurement, whereas they have been employed hitherto primarily for indication or alarm purposes.

An example of such a circuit, given in Figure 7, is an interface circuit that includes a low-voltage rail-to- rail operational amplifier Ul (e.g. of the type TLC 2272 CP).

This circuit can be used for converting the output resistance change of semiconductor gas sensors into a "standard" voltage change. The output of the circuit, at 12 in Figure 7, is in the approximate range 0.5 to 5 volts, and it varies linearly with the square root of the target gas concentration. This output can be used directly for alarm applications. It can also give the actual gas concentration, either by digital techniques (e.g. use of a "look-up" table) or by analogue techniques. In the latter case, the zero offset is removed, and the signal is then squared using techniques standard in instrumentation electronics.

In the second stage of the circuit, shown in Figure 7, the amplifier U1B has a nominal gain of 15. To obtain an output in the desired range, an input voltage change of 200 mV is required as the sensor resistance changes. This stage is provided with a zero adjustment potentiometer R5 and a gain adjustment potentiometer R8. Only three resistor positions in the first stage are involved in matching a very wide range of sensors to the electronics, namely positions A, B and R3. Sensors for which the resistance increases in the target gas are placed in position B, and a fixed resistor, chosen to match the sensor output characteristic, is put in position A. R3 is generally 10 kilohm. Sensors for which the resistance decreases are placed in position A; a fixed resistor (generally 39 kilohm) is located at position B, and a resistor selected to match the sensor is positioned at R3.

In the case where the sensor resistance increases, a resistor Rl and a diode Dl establish a stable referenec voltage of 2.45 V, and, through fixed resistors R2 and R3, they establish a drive voltage of 222 mV at the non-inverting input of the operational amplifier Ul. By feedback around the operational amplifier, the inverting input is maintained at the same voltage. This causes a constant current i to flow through the fixed resistor at A. The required 200 mV change due to the passage of this current through the sensor at B develops a voltage signal, on top of the driving voltage, which is proportional to sensor resistance. The current i should not exceed 5 μA, and is given by:

i = 2£Ω Rg max - Rg min

where the units here and in the equations below are , mV and kΩ.. Rgmax and Rgmin are the maximum and minimum values of the sensor resistance at the maximum and zero concentrations of the target gas respectively. The voltage of 222 mV at the non-inverting input of Ul allows the value R_A of the resistor at A to be determined:

R_A = 220/i.

With R3 = 10kJ , typical values of R_A are given in the following table. Sensor range (kΛ) Rgmin to Rgmax R_A (k-0-)

10 to 100 100

20 to 120 110

30 to 300 300

40 to 400 390

80 to 180 110

80 to 250 180

100 to 250 160

100 to 1000 (1MΛ) 1000

In the case where the sensor resistance decreases as the target gas concentration increases, the sensor is now in position A, and the driving voltage causes a variable current to flow through this sensor. A fixed resistor, which is used to develop a voltage from this current, is in position B. The resistance at R3- needs to be adjusted in order to limit the current through the sensor. In this configuration, the output voltage is (as required) proportional to the reciprocal of the sensor resistance. The determination of the resistor values is a two-step procedure, using the same units as previously and with Rgmax and Rgmin the maximum and minimum values of the sensor resistance, this time at the zero and maximum concentrations of the target gas respectively. The drive voltage V^ at the non- inverting input of Ul, for a maximum current of 5 I A through the sensor at its minimum resistance, is:

V^ = 5.Rgmin.

The resistance R3 is then calculated as follows

(bearing in mind that Dl develops 245mV and R2 is 100kJ ) : R3 = 100

( 2450/V_d-l ) .

The resistance R_β of the gain resistor at position B, to obtain the desired 200mV change, is given by:

R_B = 2ΩΩ.

(5-V_d/Rgmax)

In practice, the required value of R_B does not vary very much: and a 39kfl resistor will generally suffice to generate the required voltage of about 200 mV. Values of the drive voltage V_d of less than 25mV should be avoided. Suitable values of R3 are tabulated below for different sensor resistance ranges:

Sensor range (kΛ) R3 (kjft)

Rgmax to Rgmin

100 to 10 2

200 to 20 4.7

300 to 30 7.5

500 to 50 13

500 to 20 4.7

The flexibility of circuits such as that described above make it possible to interface all semiconductor gas sensors with a single circuit board such as a printed circuit board (PCB) . Sensors are connected to the PCB at either position A or B, according to the sign of their response. For sensors at position A (resistance decreasing), only the value of R3 need generally be changed; it depends upon the lowest value of the resistance achieved by the sensor at the maximum expected target gas concentration. For sensors connected at position B, the resistor at position A is approximately equal to the change in the resistance of the sensor between the zero and maximum expected target gas concentrations. For sensors of a given type, made and operated according to the present invention and responding to a particular target gas and concentration range, only a single value of either R3 or R_A is required.

Reference is now made to Figures 8 and 9, which are purely diagrammatic and in no way limiting. A gas sensing apparatus comprises at least one sensor 20 having electrodes 22 for conveying the output signal from the sensor representing its electrical resistance, together with suitable signal processing means and suitable means for supporting the sensor. The processing means preferably comprises a resistance measuring circuit, for example as shown in Figure 7 and described above.

The apparatus preferably includes a housing 22, substantially enclosing the or each sensor 20 and made of a material that substantially prevents reaction, decomposition and irreversible adsorption of the target gasl, to which it is also preferably impervious. The housing (or each housing) enables gas to reach the associated sensor 20 freely, and has an included volume in the inclusive range of 30 - 3000 times the volume of this sensor. It may comprise a header.

Figure 8 envisages the housing 22 in the form of an open-ended tube or duct, in which the sensor 20 is supported by being suspended in the duct in free air, by means of wires 24, 26. There may be any convenient number of these wires; and in this example they also provide electrical connections for the sensor.

In Figure 8, the sensor comprises a substrate 28 in the form of an alumina tile, on which two metallic electrodes 30, a heater element 32, and two pairs of contact pads 34, 36 are applied before a porous, semiconducting metal oxide layer 38 is applied to the substrate by sintering so as to be bonded to the electrodes 30 or substrate 28 or both. The characteristic dimensions of the substrate, i.e. its thickness here, are in the range 0.2 to 3 mm. The electrodes 30 are preferably interdigitated, but are shown here as simple rectangles for clarity. The spacing between them has a predetermined value, maintained during manufacture within a tolerance of

±5%. This value is in the inclusive range 1 - 300 am.

The layer 38 leaves the contact pads 34 36 exposed as indicated in Figure 8. It constitutes the sensing element of the sensor 20, and preferably has one or more of the following features:

- its thickness, greater than 50 Urn, is uniform within ±20% across the layer;

- its porosity is 30 - 60% inclusive; and

- it has no macroscopic flaws larger than 5 times the mean pore size of the layer.

In addition, the particle size of the oxide of the layer 38 is preferably less than 5 jtΛm, with no particle having a dimension greater than 5% of the thickness of the layer. The heater element 32 (shown in stylized form in Figure 8) is electrically isolated from the layer 38 (e.g. by a suitable insulating material 40), but is in heat transfer relationship with the latter so as to heat the

« layer 38. The heater 32 and signal electrodes 30 are connected to the pads 36 and pads 34 respectively.

There is a spacing of at least 3 mm between any heated portion of the sensor 20 (in particular the layer 38) and the inner surface of the housing 22.

The wires 24 and 26 are connected to respective pins 42 carried by the housing 22, these pins being joined outside the latter to electrical leads 44 or 46. The leads 44 connect the pins 42 associated with the signal elements 30 to the processing circuit, while the leads 346 connect those associated with the heater 32 to a heater driver, or control, circuit which is preferably of the kind described with reference to Figure 7.

The total resistance of the wires 24, 26 is preferably no more than 5% of that of the sensor 20. Their length and diameter are preferably such that no more than 30% of the power dissipated in the heater 32 is conducted along any one of the wires. In this example the maximum diameter of each of these wires is 100 Aim.

We have employed this systematic approach to gas sensor design, fabrication and operation to a variety of different materials, with different electrical properties and sign of response, as in the following examples, though it is to be emphasised that the approach is general and not dependent on the composition of particular materials quoted herein.

Example 1 A propane sensor was prepared from Cr-^_gTig_^2°3 according to the following procedure.

30 g pre-calcined ceramic powder of 39 ' s purity grade, 1.05 g hypermer KD1 dispersant, 25 cc of 111 trichloroethylene, 4.4 g Butvar binder, and 1.1 g din- butyl phthalate plasticiser, were mixed together in a pot with zirconia media and milled for 48 hours, before being cast under a doctor blade 800 microns above a release paper. Sections were cut from the ensuing tape and fired onto a substrate consisting of gold electrodes 200 lAm apart, printed onto an alumina tile.

The firing process followed a regime in which the temperature was raised at 1°C per minute up to 500^βC, and then at 10°C per minute to 900°C, which was held for 8 hours. Four leads were attached to contact pads, previously printed 200 J m apart on the 3 mm square by 0.6 mm thick ceramic substrate, in order to connect the sensor material and the heater to external circuits. These were attached to pins in the base of a two-part, moulded, glass fibre-filled, polyphenylene sulphide header assembly. The external extensions of the pins were connected to a heater driver circuit and a circuit for measuring the resistance of the sensor. Over a range of different operating temperatures around 480^CC, the sensor was then exposed to test gases comprising various concentrations of propane diluted by ambient air.

A typical sensor response curve is shown in Figure 10, in which sensor resistance R, in ohms, is plotted against time in seconds, under conditions going from

50% relative humidity (RH) to dry air (0% RH) and back to 50% RH, followed by exposure to the propane concentrations indicated. Figure 10 shows, in particular, a rapid response and recovery time, and a very modest influence of relative humidity on the zero resistance of the device. The long term stability of this sensor, when tested over a 70-day period, was shown by a zero drift which was consistently less than 2.5%.

Example 2

Chromium oxide and titanium oxide were mixed together to give the composition Cr^ 8^Ti0 2°3 1_" This oxide was milled and made into a paste for printing on to alumina substrates. The firing process in air followed a regime consisting of a 15"C per minute temperature rise up to 850°C, which was held for 2 hours. Four leads were attached to previously printed contact pads 20 AΛJΠ apart on the 3 mm square by 0.6 mm thick ceramic substrate, in order to connect the sensor material and the heater to external circuits. These were attached to pins in the base of a two-part, moulded, glass fibre-filled, polyphenylene sulphide header assembly.

The external extensions of the pins were connected to a heater driver circuit and a circuit for measuring the resistance of the sensor. Over a range of different operating temperatures around 480°C, the sensor was then exposed to test gases comprising various concentrations of propane diluted by ambient air. The performance of the sensor was similar in all material respects to Example 1 above. Example 3

For sensors employing tin oxide as the gas sensitive material, metastannic acid was fired at 800^βC in air for a period of 8 hours and then milled to a particle size of less than 10 microns with zirconia media, before being made into a paste which was screen-printed onto the substrate. The purity of the tin oxide was better than 99.85%, the impurities present being as follows:

- less than 0.05%: As, Bi, Co, Cu, Fe, In, Ni, Pb, Sb

- less than 0.01%: Ag, Pd, Mg, Mn, Zn.

The firing process in nitrogen followed the regime of a 15°C/min. temperature rise up to 850°C, which was held for 90 minutes. Four leads were attached to contact pads, which were printed prior to the deposition of the tin oxide on the ceramic substrate, in order to connect the sensor material and the heater to external circuits. The substrate was again 3 mm square by 0.6 mm thick. The contact pads were attached to pins in the base of a two-part, moulded, glass fibre-filled, polyphenylene sulphide header assembly. The external extensions of the pins were connected to a heater driver circuit and to a circuit for measuring the resistance of the sensor. Over a range of different operating temperatures around 450°C, the sensor was then exposed to test gases comprising various concentrations of carbon monoxide diluted by ambient air.

Figure 11 shows a typical response curve for this sensor, demonstrating in particular low humidity effects and rapid response and recovery time. Long- term stability was estimated to be equivalent to a zero drift of less than 5% per year.

An iron niobate sensor can be used for chlorine or NO_χ as target gas; and a chromium titanium oxide sensor can be used for ammonia or hydrogen sulphide as target gas. Preparation of these sensors will be within the competence of a person skilled in this field, given the above Examples and/or the remainder of this disclosure.

Claims

1. A gas sensor comprising a sensing element of semiconducting material for giving an electrical output signal dependent on the effect of a target gas on the resistance of the sensing element, characterised in that the sensing element comprises a porous metal oxide layer having at least one of the following features:

(a) a thickness greater than 50 /u.m, uniform across the element to within ±20%;

(b) a porosity in the inclusive range 30 - 60%; and

(c) freedom from macroscopic flaws larger than 5 times the mean pore size of the layer,

the sensor being such that, when it is one of a ^• plurality of nominally identical sensors having a substantially uniform baseline value of resistance in the absence of the target gas,

- the variation in said baseline value bewtween the sensors is less than ±5%;

- the output signals of all the sensors bear a common simple mathematical relationship to target gas concentration;

- the said output signals define a characteristic curve, as a function of gas concentratiobn, which varies between the sensors by less than ±15%; and

- the optimum sensor operating temperature varies between the sensors by less than 30°C.

2. A sensor according to Claim 1, characterised in that the metal oxide has a particle size less than 5 m, isolated particles of dimension greater than 5% of the thickness of the said layer being absent.

3. A sensor according to Claim 1 or Claim 2, characterised by a metal oxide substrate in electrical contact with the said layer, and a pair of interdigitated metallic electrodes carried by the substrate so that the said resistance of the element is the resistance between the electrodes, the latter being spaced apart by a distance, which is within ±5% of a predetermined value in the inclusive range 1 - 300 Llm .

4. A sensor according to Claim 3, characterised in that the said layer is bonded to the electrodes or the substrate, or to both, by sintering.

5. A sensor according to Claim 3 or Claim 4, characterised in that the substrate is of alumina having characteristic dimensions in the range 0.2 to 3 mm.

6. Gas sensing apparatus comprising: at least one sensor having electrodes for conveying the output signal from the sensor; signal processing means for processing the output signal or signals; and sensor support means, the electrodes of the or each sensor being connected to the processing means by electrical connecting means, characterised in that the or each said sensor is a sensor according to any one of the preceding Claims.

7. Apparatus according to Claim 6, characterised in that the processing means comprise a resistance measuring circuit.

8. Apparatus according to Claim 6 or Claim 7, characterised in that it further includes sensor heating means for heating at least part of at least one said sensor, and heating control means connected with the heating means.

9. Apparatus according to Claim 8, characterised in that the heating means comprise an electrically isolated heating element in heat transfer relationship with the oxide layer of the sensor or of at least one said sensor.

10. Apparatus according to Claim 8 or Claim 9, characterised in that the heating control means are adapted for controlling the temperature of the said layer to an accuracy of less than ±1°C, and to maintain the temperature across the said layer to an accuracy within ±30°C.

11. Apparatus according to any one of Claims 6 to 10, characterised in that it further includes a housing, substantially enclosing the or each sensor and made of a material such as substantially to prevent reaction, decomposition and irreversible adsorption of the target gas, the or each housing having means for permitting free access of gas to the associated sensor, the or each housing defining an included volume in the inclusive range 30 to 3000 times the volume of the associated sensor.

12. Apparatus according to Claim 11 when dependent on any one of Claims 8 to 10, characterised in that the or each said housing has an inner surface spaced more than 3 mm away from any heated portion of the associated sensor.

13. Apparatus according to Claim 11 or Claim 12, characterised in that the or each housing is impervious to the target gas.

14. Apparatus according to any one of Claims 6 to 13, characterised in that the sensor support means comprise wires of diameter 100 lκm at most, whereby the or each sensor is suspended in free air, at least some of the said wires constituting the said electrical connecting means for conveying the sensor output signal.

15. Apparatus according to Claim 14 when dependent on any one of Claims 8 to 10, characterised in that the said wires include wires that connect the sensor heating means electrically with the heating control means, the resistance of the said wires being at most 5% of that of the associated sensor, with their length and diameter being such that no more than 30% of the power dissipated in the heating means is conducted along any one said wire.

16. Apparatus according to any one of Claims 6 to 15, characterised in that the processing means include means for measuring the resistance of the said oxide layer of the or each sensor, being such as to limit the current through the metal oxide to less than 5/.A.

17. Apparatus according to any one of Claims 6 to 16, characterised in that the processing means include means for linearising the response of the sensor resistance to changes in the concentration of the target gas.