WO2001033205A1

WO2001033205A1 - Semiconductor gas sensors

Info

Publication number: WO2001033205A1
Application number: PCT/GB2000/004235
Authority: WO
Inventors: Keith Francis Edwin Pratt; Charles Edmund King
Original assignee: Capteur Sensors And Analysers Limited
Priority date: 1999-11-04
Filing date: 2000-11-03
Publication date: 2001-05-10
Also published as: GB9926174D0

Abstract

A semiconductor gas sensor has a gas sensing element and an elongate channel leading to or through the element for the passage of gas to be sensed. The sensor may be a substrate with the element and channel formed thereon as a film, e.g. the screen printing and firing, and with a gas-impervious layer covering the element and film except at a small opening therein, e.g. at a side edge of the substrate.

Description

SEMICONDUCTOR GAS SENSORS

Preamble Gas sensors based on semiconducting metal oxides generally have poor selectivity

Selectivity may be imparted by adding dopants to the gas sensing mateπal or modifying its microstucture, or by the use of a separate filter, typically activated charcoal to adsorb mterferent gases. Doping the sensing mateπal and optimising its microstructure have relatively small effects on the gas selectivity of the mateπal, and the use of protective filters is undesirable as they add cost to the sensor and have a finite adsorption capacity. Existing technology relevant to the present invention

The gas sensitive mateπals used in semiconducting oxide gas sensors are, by virtue of their functionality, catalytically active for combustion of the gases that they respond to This combustion process, coupled with the fact that thick film sensors consist of a porous gas sensitive layer with tortuous paths for gas access, results in concentration gradients of gases within the sensing layer, whereby the gas concentration decreases with depth into the sensing layer. The magnitude of this concentration gradient is determined by the rate of gas combustion reaction relative to the rate of mass transport or diffusion of the gases through the sensing layer Vaπous types of sensor design have been developed which take advantage of this effect. Sensors with electrodes aπanged to probe the gas concentration as a function of depth within the sensing film have been descπbed ',",'" which show that the concentration gradient through the sensor depends on the type of gas used, as well as the operating conditions of the sensor, and that discπmination between gases can be achieved by measurement of the resulting concentration gradients. The steepness of the concentration gradient through the mateπal can be defined " by the parameter kh²/D, where k is the rate constant for combustion of the target gas (assumed here to be first order for simplicity), D is the effective diffusion coefficient for the target gas, which depends also on the microstructure of the mateπal, and h is the height of the sensing layer as shown in Figure 1. Theoretical shapes of concentration profiles through the sensing layer as a function of the parameter kh²/D are shown in Figure 2. Theoretically " there should also be a transient vaπation of the concentration profile on initial application of the target gas, but in practice this is not observed since the concentration profile reaches equihbnum more quickly than the response time of the gas sensitive mateπal. The diffusion time constant is given by the parameter h /D and hence is also dependent on the height of the sensing layer. Practical screen pπnted structures can have heights, h, up to several hundred micrometres Larger values of h can be achieved using pressed pellets, where the faces of the pellet are covered by alumina discs so that gas may only access via the peπmeter surface ^1V The parameter h then coπesponds to the diameter of the pellet. Steeper concentration gradients may then be obtained using pellet devices where h is equivalent to several millimetres. It has been demonstrated that a concentration gradient of gas may be produced between electrodes at the peπmeter of a circular pellet and those at the centre, ^1V This approach has also been demonstrated for the production of a disc sensor sandwiched between alumina substrates with integral heater ^v Sensors fabπcated by third type of approach are impractical for mass production for a number of reasons Firstly the use of a pressed pellet of gas sensing mateπal generally results in a relatively fragile structure requiπng additional reιnforcement,^{lv v} secondly the application of physically separate plates to the surfaces of the gas sensing mateπal does not result in perfect sealing, especially under varying thermal conditions Devices fabπcated by such approaches are also physically large since the overall dimensions of the device determine the cπtical dimension h Due to the radial diffusion conditions, the effective value of h is less than half of the pellet diameter ^1V

The steepness of concentration gradient through the sensing mateπal may be increased by adding a catalytically more active mateπal to it,¹" though this increases the susceptibility to poisoning Also, concentration gradients may be generated across an additional catalytic layer above the gas sensitive mateπal, thereby reducing the concentration of more reactive gases reaching the sensing mateπal ^vι The additional layer may also selectively adsorb unwanted interferent gases Protective layers of this type are routinely used on catalytic bead type sensors. ^v" Invention

One embodiment of the present invention is shown in schematic form in Figure 3 Here the sensing mateπal is pπnted in the form of a long naπow channel The sensing mateπal is then encased in the glass dielectπc such that only one end of the channel is accessible to gas, which must diffuse along the channel as shown in the figure. The example shows two pairs of sensing electrodes; (a) is nearer to the gas access point than (b). Thus gas reaches electrodes (a) more easily than (b). As shown in Figure 2, a combustible gas would have a lower concentration in the region of electrodes (b) than at (a), so the former would give a lower response. Also a delay may be observed due to the time taken for the gas to diffuse along the channel with the result that electrodes (a) would respond to the gas more quickly than (b) A single pair of electrodes may also be used at position (b) only. Since gas access is hoπzontally rather than vertically, the electrode structure shown in Figure 4 may be used without causing additional restπction to gas access. This idea can be taken further m that the gas impermeable coating itself may be compπsed of an electncally conducting mateπal such as gold or platinum, thereby acting as a common electrode, as shown in Figure 5. Many vaπants of this basic design can be conceived. It is not necessary for the sensing regιon(s) and channel to be compπsed of the same mateπal One approach to this is shown in Figure 6 In the example shown, the mateπal labelled (2) is catalytically active for decomposition of the target gas whereas 1 and 3 are not. This results in concentration profiles of the type shown in Figure 7. In addition, as suggested in Figure 6, the sensing mateπals in regions 1 and 3 need not be the same mateπal. Clearly, this approach can be extended to include multiple regions extending inwards from the gas entry port

The patterning ability of screen pπntmg means that complex geometπes can be achieved Figure 8 shows that by pπntmg a serpentine pattern, a channel may be made with a path length greater than the dimensions of the sensor substrate. Figure 9 shows another method of using a different mateπal for the sensing layers and channel Here, sensing chambers are connected to vaπous points along the channel.

Screen pπntmg can also be used to produce complex multilayered structures. For example, Figure 10 shows how alternate layers of gas permeable mateπal and gas impermeable dielectπc may be used to produce a long serpentine channel. It can be seen that, by using the ideas in figures Figure 8 and Figure 10, channel lengths of several cm could be produced on sensor chips a few mm square.

Figure 11 shows a practical implementation of a sensor with multiple electrode connections along the length of the channel, using the vertical electrode structure from Figure 4. The channel has a total length, h, of 4.3mm

The geometry of the channel itself will affect its behaviour. This is explained in figure Figure 12. The narrower channel in Figure 12a will result in more restπction of gas access to the sensing region than the larger channel in Figure 12Z?. Therefore, by varying the dimensions of a channel along its length, its properties may be modified. This can be achieved by vaπation of the screen pπnt thickness or by design of the screen pattern. A practical implementation of this is shown in Figure 13, whereby the width of the channel from Figure 11 is reduced between the sensing regions to give more restπcted gas access. The concentration gradients and diffusion timescales in such a device can be made significantly higher than if the channel had a constant width. The actual performance of such a design can be modelled using a multidimensional computational simulation ^ιv

Another implementation of the sensor is shown m Figure 14. In this example different mateπals may be used for the sensing mateπal and the channel. In addition to this, the sensing region is large in compaπson to the width of the channel, so that the channel will give a greater restπction to gas access than would be the case if the channel and sensing region had similar dimensions, le the effective path length, h, is longer than the actual track length, the latter being approximately 2mm In this design, only a single pair of electrodes is used for detection. Examples

Figure 15 and Figure 16 show the behaviour of the type of sensor shown in Figure 3, with a channel composed of chromium titanium oxιde^vl" with approximate dimensions 3mm long, 200μm wide and 20μm thick, when exposed to carbon monoxide and ammonia respectively For both gases, the response on the inner pair of electrodes is smaller than that on the outer pair showing that consumption of the gas is occurπng within the channel, producing a concentration gradient The concentration gradients of the two gases are different hence they could be distinguished using methods we have descπbed before.' " '" ^1V In addition to this, it can be seen that the transient behaviour towards the two gases differ. The speed of response to ammonia is the same on both pairs of electrodes, however m the case of carbon monoxide, the inner pair of electrodes responds more slowly than the outer pair. This indicates that diffusion of carbon monoxide through the mateπal is slower than that of ammonia An explanation for this may be that carbon monoxide undergoes adsorption/desorption on the channel mateπal, which will result in a reduction of the apparent diffusion rate.'" Figure 17 shows responses to a range of gases for the sensor type shown in Figure

14, with chromium titanium oxide as both sensing layer and channel mateπal. The sensor is compared to a conventional planar screen pπnted chromium titanium oxide sensor, as descπbed in Capteur product literature for GL07 sensor It can be seen that the responses to ethanol and carbon monoxide are almost completely removed, and those to heptane and hydrogen are significantly diminished Of the gases tested, only propane gives a significant response on the new sensor.

The use of different mateπals for the channel is exemplified in Figure 18, where chromium titanium oxide is still used as the sensing mateπal but the channel is composed of either chromium oxide, tungsten oxide or platinised tin oxιde Different relative responses to the vaπous gases tested show that the selectivity of the sensors can be tailored by choice of channel mateπal and its microstructure, even while using the same sensor mateπal. The sensing mateπal may also be changed to further vary the behaviour. Methods of operation

The examples so far have shown sensors operating at constant temperature, which show that the sensors allow significant improvements in discπmination towards target gases compared with existing technology

The transient response differences between gases, shown in Figure 15 and Figure 16 can also be used to further discπminate between the gases. The difference in transient behaviour between ammonia and carbon monoxide can be visualised more clearly by plotting the ratio of outer to inner electrode response as shown m Figure 19 and Figure 20. The response towards carbon monoxide is then characteπsed by a peak in the response ratio" as seen in Figure 19. Discπmination of gases using this approach would be more complicated if the gas concentration was changing with time This could be overcome by using a sampling system which uses an external valve to switch between clean air and sample gas, so that gas is applied to the sensor at a predefined time, and the sensor is allowed to recover back to its baseline after the gas exposure. An alternative approach to sampling the gas takes advantage of the thermal properties of the gas and the channel/sensing mateπal. If the sensor is taken to a temperature significantly higher than its normal operating temperature, then adsorbed gases within the sensor channel would be desorbed into the air within the sensor channel, and combustible gases decomposed to species such as carbon dioxide and water. In addition to this, volume expansion of the gas within the channel will pump out the desorbed gases. Without this volume expansion occurπng, desorbed gases within the chamber would simply re-adsorb on returning to the lower measuπng temperature. Figure 21 shows the same sensor exposed to carbon monoxide while its temperature is cycled between 400 and 475° C. At the higher temperature, where the baseline resistance is lower, there is no detectable response on the inner electrode pair, and only a small response on the outer pair. On returning to the lower temperature, the transient behaviour seen in Figure 15 and Figure 19 is again observed showing that the channel has been effectively purged at the high temperature. The behaviour is repeatable on repeated cycles. Since the start of the transient is now defined by the time at which the temperature is switched from high to low value, the full shape of the transient can readily be used to discπminate the type of gas applied Figure 23 to Figure 28 show the sensor operated in the same mode, when exposed to ammonia, ethanol and propane. For these three gases, the peaked response characteπstic of carbon monoxide is not observed since there is no discernible delay time between the response on the two electrode pairs, but differences between the gases can be seen in terms of the relative responses on outer and inner electrodes at the low temperature For propane, there is a significant response on the inner electrode, even at the high temperature. This is consistent with the data in Figure 17, and shows that propane is not significantly decomposed within the channel. This lack of effective purging of the gas at the high temperature also results in drift in the repeated temperature cycles in Figure 27. Analysis of the shape of the response transients can now be used to distinguish carbon monoxide from the other gases.

The effects of thermal expansion / contraction of gas with temperature can be further enhanced by the design of the device. Figure 29 shows schematically how this may be achieved. A chamber is produced at the inner end of the channel, whose dimensions may be larger than the channel itself. It may contain a porous mateπal, or may be a void created by burning out a sacπficial support ¹ In either case, thermal expansion and contraction of gases within the chamber which occur on changing the operating temperature can then be used to pump gases through the channel / sensing layer. This results in faster gas flow through the porous channel, thereby allowing faster measurement times than can be achieved if mass transport is by diffusion alone, while still maintaining a long channel length to give good separation. Another more subtle difference with this design is that the timescales and concentration gradients of gases within the channel will be less dependent on their diffusion coefficients. If the sensor design in Figure 29 is dropped from a high purge temperature to a suitably low temperature, the pressure chamber would rapidly fill with gas until it equilibrated with the atmospheπc pressure. After this, slow leakage of gas into and out of the chamber would only occur by diffusion through the channel. The device could then be used as a 'sample and hold' system. On increasing the device to a suitable measurement temperature, the pressure increase within the chamber then pumps the sampled gas through the channel where separation and detection can occur as before. The increase to the measurement temperature may be a step change, or a ramp.

This idea can be further developed by incorporating an adsorbent mateπal in the pressure chamber, as shown in Figure 30, or in between the pressure chamber and the channel as shown in Figure 31 The adsorbent mateπal could, for example, be chromium titanium oxide as this has been shown above to selectively adsorb carbon monoxide, palladium to adsorb hydiogen, zeolites or any other type of adsorbent compatible with the thick film fabπcation process and operating temperatures. As with the separation channel, it is not necessary for the mateπal to be sintered together but may be present as a loose powder which is contained by the gas impermeable support. In addition to acting as an adsobent, the mateπal of either the adsorbent bed or the channel may act as a catalyst to convert the adsorbed gas into a species which may be more readily detected by the sensor. The designs shown in Figure 30 or Figure 31 could then be used as follows . A high purge temperature is initially used to clean out the system and heat the air in the pressure/adsorption chamber. The device is then lowered to a suitable temperature for adsorption of the target gas(es) The temperature is then stepped to an intermediate temperature suitable for detection, whereby gas desorbs from the adsorbent, and is pumped or diffuses out through the channel and detection electrodes. The presence of a separation channel and/or multiple electrodes allows discπmination of gases based on their transport, sorption and reaction properties through the channel, which acts as a form of gas chromatograph. An alternative operation method is that, following the adsorption step, the operating temperature is ramped upwards at a controlled rate, which should allow the detection and discπmination of species adsorbed on the bed based on their desorption temperatures, in a similar manner to conventional temperature programmed desorption. In this latter case, separation may be achieved without requiπng the channel and multiple electrodes. The temperature programs are indicated in Figure 32. Figure 33 shows a practical implementation of the ideas shown in Figure 29 and Figure 30.

Benefits of the invention

The basic invention allows control of the transport of a gas along a defined channel of porous mateπal. The length of the channel may be many times longer than the physical dimensions of the device, thus path lengths many times longer than achieved in a pellet device of dimensions ~lcm can easily be achieved m a screen pπnted device on a substrate of dimensions of the order of 2x2mm. The effective diffusion length, h, may also be longer than the physical length of the channel by careful design of its geometry. This reduction in size is beneficial both from the view of power consumption and also because the reduced thermal mass allows more rapid changes in operating temperature to be achieved thus facilitating sophisticated temperature programs to be used. The use of screen pπnting allows mass production of the devices without any of the complexities involved in producing pressed pellets of micromachined devices. Adhesion of the different layers is improved, and the mechanical properties of the mateπals can be tailored

Only relatively small quantities of the channel mateπal are required to give a long path length, which is useful if the mateπal is expensive or hazardous.

As the devices are encapsulated in a glassy dielectπc mateπal their physical strength is greatly improved relative to conventional sensors whether produced by screen pπnted or other techniques, and contamination on storage is minimsed thus reducing burn-m time. This encapsulation means that the active mateπals do not need to be physically robust or self-supporting, and hence fπable mateπals or even loose powders may be used. The properties of the channel/adsorbent mateπals can then be chosen to optimise performance without regard to effects of thermal mismatch, adhesion, cohesion etc. Figure Legends

Figure 1 A schematic cross sectional view of a conventional planar screen pπnted thick film sensor, showing the concentration profile of a gas within the porous sensing layer Figure 2 Theoretical equilibrium concentration profiles for gases within a porous thick film for indicated values of the parameter kh²/D

Figure 3 Schematic view of top(a) and sιde(b) of simplest implementation of present invention

Figure 4 Schematic view of invention with vertical electrode structure

Figure 5 Schematic view of invention with vertical electrode structure where the common electrode also functions as the gas impermeable coating

Figure 6 Schematic view of invention with different materials for channel and sensing regions

Figure 7 Schematic concentration profiles for the design in Figure 6, for the situation where significant reaction of the target gas only occurs in region 2

Figure 8 Schematic view of invention where length of channel is increased bv the use of a serpentine printed pattern In this example the channel also functions as the gas sensitive material

Figure 9 Schematic view of invention where length of channel is increased by the use of a serpentine pπnted pattern In this example the gas detecting regions are separate from the channel, and may be of different mateπal(s)

Figure 10 Schematic view of invention where length of channel is increased by the use of a multiple layered printed structure

Figure 1 1 Practical implementation of the ideas in Figure 4 and Figure 8, on a chip with overall dimensions 3x3mm

Figure 12 Schematic diagram to explain the effects of the geometry of the device on its behaviour The narrow channel in (a) gives more restriction to gas diffusion than that in channel (b) Figure 13 Practical implementation of a device, based on Figure 11 , where additional restriction to gas flow between the electrodes is achieved as described in Figure 12

Figure 14 Practical implementation of a device on a 2x2mm substrate, with a single pair of detecting electrodes corresponding to electrodes (b) in Figure 3, using a separate sensing region and channel as in Figure 7, with restricted gas access as in Figure 12 The sensing mateπal is printed over an interdigitated electrode structure (not shown) and surrounded by the first dielectπc layer This is overprinted with the channel, and the whole covered with the second dielectric layer which allows gas to enter only via the gas entry port shown The volumes of the sensing region and channel may be independently varied by printing multiple layers

Figure 15 Response of the sensor design in Figure 3 and the text, to 2000ppm carbon monoxide in air, when operated at a temperature of 400° C

Figure 16 Response of the sensor from Figure 15, to 200ppm ammonia in air, when operated at a temperature of 400° C

Figure 17 Response to a range of gases, of a sensor as shown in Figure 14, using chromium titanium oxide for both sensing and channel materials, operated at 500° C Figure 18 Response to a range of gases, of sensor as shown in figure 14, using chromium titanium oxide as the sensing mateπal and a range of channel matenals, operating at 400° C and 500° C

Figure 19 The response data from Figure 15, expressed as the ratio of response between the outer and inner electrode pairs, showing a characteristic peak for carbon monoxide Figure 20 The response data from Figure 16, expressed as the ratio of response between the outer and inner electrode pairs, showing the lack of a peak in the response to ammonia

Figure 21 The response of the above sensor to 2000ppm carbon monoxide, where the sensor is cycled between 400 and 475° C, with 5 minutes at each temperature The data is normalised to the baseline resistances at 400°C Figure 22 The data from Figure 21 , expressed as the ratio of response between the outer and inner electrode pairs

Figure 23 The response of the above sensor to 200ppm ammonia, where the sensor is cycled between 400 and 475° C, with 5 minutes at each temperature The data is normalised to the baseline resistances at 400° C Figure 24 The data from Figure 23, expressed as the ratio of response between the outer and inner electrode pairs

Figure 25 The response of the above sensor to 500ppm ethanol, where the sensor is cycled between 400 and 475° C, with 5 minutes at each temperature The data is normalised to the baseline resistances at 400° C Figure 26 The data from Figure 25, expressed as the ratio of response between the outer and inner electrode pairs

Figure 27 The response of the above sensor to 2000ppm propane, where the sensor is cycled between 400 and 475° C, with 5 minutes at each temperature The data is normalised to the baseline resistances at 400° C Figure 28 The data from Figure 27, expressed as the ratio of response between the outer and inner electrode pairs

Figure 29 Schematic diagram of an implementation of the invention whereby a chamber at the inner end of the channel is used to enhance the effects of gas volume changes on changing the operating temperature Figure 30 Schematic diagram of an implementation of the invention whereby a chamber at the inner end of the channel contains an adsorbent mateπal

Figure 31 Schematic diagram showing a combination of the inventions in Figure 29 and Figure 30

Figure 32 Example temperature programmes used with the sensor designs in Figure 29 - Figure 31

Figure 33 Practical implementation of a device using the inventions in Figure 29 and Figure 30 The first dielectric has apertures for the chamber and sensing mateπals, the latter connecting to the gold electrodes(a) These are overprinted with the channel (b) which is electrically isolated from the electrodes The upper dielectric (c) seals the whole structure allowing gas entry and exit access only via the port shown The chamber may be filled with an adsorbent, an inert porous filler or a sacrificial material which burns out to leave a void The volumes occupied by the sensing materials, channel and chamber may be independently varied by printing multiple layers References

¹ UK Patent GB2 256 055, GAS SENSOR INCORPORATING MALFUNCTION TESTING, D.E. Williams, 1992; UK Patent GB2 218523, SENSING THE COMPOSITION OF GAS, C.P. Jones, P.T. Moseley, D.E. Williams, 1989

" 'Theory of Self-diagnostic Sensor Array Devices using Gas-sensitive Resistors", J. Chem. Soc. Faraday

Trans. 91(13), 1961-1966 (1995).

¹¹¹ "Resolving Combustible Gas Mixtures using Gas Sensitive Resistors with Arrays of Electrodes" , J. Chem.

Soc. Faraday Trans., 92(22), 4497-4504 (1996).

^,v "Reaction-Diffusion Effects and Systematic Design of Gas-Sensitive Resistors based on Semiconducting

Oxides", J. Chem. Soc. Faraday Trans., 91(23), 4299-4307 (1995). ^v US Patent No. US4911892, Apparatus for simultaneous detection of target gases, Grace et al, 1987 ^vι US Patent No. US6012327: Gas sensor and method for manufactuπng the same, M. Seth,

M. Fleischer and H. Meixner, 1997

™ UK Patent No. GB2083630, Catalytic combustible gas sensors, D.W. Dabill, S.J. Gentry, N.W. Hurst, A.

Jones and P.T. Walsh, 1982

^V," EP-A-0656111

'" "Modelling of Gas-sensitive Conducting Polymer Devices", IEE Proc. Circuits Devices and Systems,

142(5), 321-333 ( 1995)

" EP-A-0757791

Claims

1. A semiconductor gas sensor having a gas sensing element and an elongate channel leading to or through said element for the passage of gas to be sensed.

2. A sensor according to claim 1 wherein the element is formed as the channel.

3. A sensor according to claim 1 having separate element and channel materials.

4. A sensor according to any preceding claim wherein the element and channel are each a film mounted on a substrate, the element and channel being covered by a gas-impervious layer except at a small opening therein.

5. A sensor according to claim 4 wherein the opening is at a side edge of the substrate.

6. A sensor according to any preceding claim having a single electrode pair in contact with the element at an inner end of the channel.

7. A sensor according to any of claims 1 to 5 having multiple electrode pairs along the length of the channel.

8. A sensor according to any preceding claim wherein the element and/or the channel is a semiconducting oxide.

9. A sensor according to claim 8 wherein the sensing element and/or the channel is of tin dioxide, tungsten trioxide, or chromium titanium oxide with or without additional dopants.

10. A sensor according to any preceding claim having a chamber for gas in communication with said element.

11. A sensor according to claim 10 wherein the chamber is a void.

12. A sensor according to claim 10 wherein the chamber contains a porous material.

13. A sensor according to claim 12 wherein the porous material is capable of adsorbing one or more gases.

14. A method of operating a sensor as claimed in any of claims 1 to 13 so as to discriminate between a target gas and an interferent gas, the discrimination being obtained by the comparison of responses between at least two pairs of electrodes on the sensor.

15. A method according to claim 14 wherein the comparison is obtained from the speed of response of the pairs of electrodes.

16. A method according to claims 14 or 15 which comprises heating the sensor to provide a high temperature purge of gas followed by a lower measuring temperature.