Transient Operation Techniques for Gas Sensor Applications

Transient Operation Techniques for Gas Sensor Applications

Available online at www.sciencedirect.com Procedia Engineering 47 (2012) 1466 – 1473 Proc. Eurosensors XXVI, September 9-12, 2012, Kraków, Poland T...

842KB Sizes 3 Downloads 113 Views

Available online at www.sciencedirect.com

Procedia Engineering 47 (2012) 1466 – 1473

Proc. Eurosensors XXVI, September 9-12, 2012, Kraków, Poland

Transient Operation Techniques for Gas Sensor Applications Roland Pohleaa* a

Siemens AG, Corporate Technology, Otto-Hahn-Ring , D-81739 Munich, Germany

Abstract This paper provides an update, how transient operation methods can improve the performance of gas sensors and sensor systems for specific applications. Even new applications can be enabled, if existing state-of-the-art sensors are operated in transient modes adapted to the needs of the application in mind. Several parameters are available for transient variation, depending on the particular sensing technology: Suspended Gate FET sensors deliver a simultaneous readout of work function and capacitive effects. By applying pulsed gate voltages these effects can be interpreted separately and used e.g. for compensation of humidity effects and improved selectivity. Modulation of the operating temperature of μ-machined metal oxide sensors is creating a type of virtual sensor array. In this case, the dynamic behavior of adsorption and desorption at changing temperature levels delivers a fingerprint for application relevant gas mixtures. The pulsed application of charge voltage on standard exhaust lambda probes and the recording of the related discharge characteristic over time provide an exhaust NOx sensor based on a standard sensor setup. The charging creates a misbalance in oxygen-related surface species not available under continuous operation conditions. The related discharge characteristic is strongly influenced by the interaction of these oxygen species with NOx. © ThePublished Authors. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Symposium Cracoviense ©2012 2012 by Elsevier Ltd. Sp. z.o.o.

Keywords: Gas sensor; transient operation; metal oxide, exhaust gas, Suspended Gate FET

1. Introduction In the last decades, gas sensors have been more and more established in our daily live. Nevertheless, only a few technologies are commercially used in large quantity applications. The most common ones are

* Corresponding author. Tel.: +49 89 636-48934 ; fax: +49 89 636-46881 . E-mail address: [email protected].

1877-7058 © 2012 The Authors. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Symposium Cracoviense Sp. z.o.o. doi:10.1016/j.proeng.2012.09.432

Roland Pohle / Procedia Engineering 47 (2012) 1466 – 1473

1467

zirconia-based lambda probes for car exhaust monitoring and resistive metal oxide sensors for natural gas alarms and air quality monitoring. High development efforts have been used to enable reliable and cost effective production of these sensors on an industrial level with several millions of pieces per year. In parallel, Suspended Gate FET sensors have been brought to a state of technological maturity, which allows this new low cost, low power gas sensor technology to enter large quantity markets. Beside these established technologies, enormous efforts are made on new sensing materials for existing transducer platforms and on innovative transducer principles. Nevertheless there are high costs and simultaneously high financial risks to consider in order establish a new sensor technology on the market, even if the results of the foregoing research and development are highly promising. Up to now, two main strategies are obvious to cope with this situation: - Avoid the need of new gas sensors This approach has been followed e.g. by predicting exhaust gas concentrations from signals available in the vehicle engine management system [1,2]. Often this approach fails, since the inherently high requirements for the quality and stability of the parameters used as input for the estimation can not be fulfilled. - Use existing sensors Following this path, additional costs have to be encountered related to the fact, that a sensor not designed for a specific application will not show the expected performance. A third approach emerges from the second, if the sensor is respect more as a system than as a component. Respecting the special physical, chemical and electrical properties and inherent functional mechanisms, existing sensor technologies can be operated in a way, that allows to fulfill criteria requisited by a targeted application. By transient manipulation of existing functional operational parameters of a gas sensor (e.g. temperature, electrode potential), the sensing characteristics can be altered dramatically. Since the control of these parameters is already implemented in the sensor itself and often also to in existing driving circuits, a very cost effective way is created to adapt established sensors for additional applications.

2. Gate pulsed readout of Floating Gate FET sensors The readout of gas induced work function changes via hybrid suspended gate field effect devices (HSGFET) is accepted as a promising technique for the realization of a versatile, low-cost sensor platform since several years [3]. The industrialization of the advanced floating gate FET (FGFET) device is already started by Micronas GmbH [4]. The freedom in choice of sensing materials is due to the fact that the hybrid setup enables to use sensing materials produced in independent technological steps not compatible to CMOS standards. Numerous sensing layers for a variety of gases have been developed creating a signal in the FGFET setup by different physical and chemical effects [5]. Commonly HSGFET and FGFET sensors are considered to read out changes in work function due to adsorption of gas molecules as a surface effect, which is only valid for conducting or thin sensing layers. In the case of relatively thick (several μm) isolating sensing layers, a significant part of the response is due to capacitive effects related to the volume of the sensing layer. Regarding surface induced signals, baseline instabilities can be induced by unintended surface conductivity and charge drifts within the transducer limiting the accuracy and the longterm stability of the sensor signal. By promoting the capacitive components of the sensor signal, surface related drift effects can be suppressed effectively. This approach is exemplified by means of the gate-pulsed readout of FGFET sensors [6]. Fig. 1a depicts the hybrid FGFET setup, which consists of a CMOS FET structure as readout device and a suspended gate including the gas sensitive layer. The floating gate electrode is prolonged forming a

1468

Roland Pohle / Procedia Engineering 47 (2012) 1466 – 1473

capacitive element with the suspended gate electrode on the backside of the sensitive layer. The air gap is of the size of a few microns to allow gas diffusion into the capacity. As depicted in the equivalent circuit diagram of an FGFET type sensor (fig. 1b), the capacitive coupling of the floating gate is determined by fixed capacities given by the CMOS process (Cwell, Cpass) and capacities related to the hybrid FGFET setup (Clayer, Cair gap) influenced by the gas atmosphere. Ugate suspended gate

floating gate

Clayer suspended gate electrode

gas sensitive layer air gap

Cair gap Readout FET

Cpass Cwell

capacitance well substrate (p-Si)

Usource

Udrain

readout FET

Fig. 1 (a). : Scheme of the Floating Gate FET (FGFET) [4]. (b) simplified equivalent circuit diagram of the FGFET. Clayer : Capacitance of the gas sensitive layer, Cair gap : Capacitance of the air gap, Cpass : Capacitance of the passivation layer on top of the floating gate. Cwell : Capacitance between floating gate and the capacitance well [6].

The influence of changing humidity levels on the gate pulse response of a FGFET with polyamide as sensing layer is shown in fig. 2a. The change in output voltage caused by the gate voltage pulse at 85 % relative humidity (% r.h.) is significantly higher compared to the response at 40 % rel. humidity. This correlates to a better capacitive coupling of the gate voltage to the FET and therefore to an increase in gate capacity. This increase can be caused by swelling of the polymer sensing layer or by an increase of the dielectric constant of the sensing layer. The response of a FGFET calibrated as described above exhibits an accurate measurement of relative humidity comparable to the reference sensor (fig. 2b).

voltage on suspended gate

1.0

FGFET signal

2.3 2.2 40

FGFET gate pulsed readout (% r.h.) r.h. reference (% r.h.)

80 1.5

2.5 2.4

90

0.5

relative humidity (% r.h.)

2.6

2.0 40% r.h. 85% r.h.

voltage on suspended gate (V)

2.7

gate pulse response at 40% rel. humidity

FGFET signal (V)

2.8

FGFET with polyamide sensing layer gate pulse response at 85% rel. humidity

2.9

70 60 50 40 30 20 10

50

60

(a)

time (s)

70

0.0

0

0

20

40

60

80

100

120

140

160

180

time (minutes)

(b)

Fig. 2. (a) Response curves of a FGFET sensor to gate voltage pulses at changing humidity levels. (b) comparison of the transient response of a calibrated FGFET humidity sensor in gate pulsed operation with a reference r.h. sensor at humidity levels from 10% to 85% r.h.

1469

Roland Pohle / Procedia Engineering 47 (2012) 1466 – 1473

Besides increased baseline stability the gate pulsed readout eliminates cross sensitivities to numerous gases as demonstrated for H2, NH3 and NO2 in fig. 3. This is again a validation, that continuous readout delivers readout mainly of work function changes due to gas adsorption on the sensor layer surface, while in pulsed readout volume effects are dominating the response due to capacitive changes. 50

r.h. reference (% r.h.) FGFET continuous readout FGFET gate pulsed readout

40 35 30 FGFET with polyamide sensing layer

25 100 10 1

720

780

840

50 T (°C)

H2

40

NH3

30

NO2

20

900

time (min)

960

1020

10

r.h. (%) T(°C)

[gas] (ppm)

rel. humidity (%)

45

1080

Fig. 3. Cross sensitivities of a FGFET humidity sensor to H2, NH3 and NO2 in continuous and gate pulsed readout.

3. Stepwise temperature modulation of micromachined metal oxide based gas sensors The modulation of the operation temperature of metal oxide gas sensors is already widely investigated as reviewed in [7] in order to increase selectivity and improve signal stability and has also been combined with adapted data evaluation algorithms for specific applications like fire detection [8]. Conventional metal oxide sensors have relatively long thermal time constants and are therefore limited to operation with relatively long temperature cycle times in the range of several ten seconds ending up in long response times and high power consumption. Therefore the combination of MEMS based sensors with thermal time constants in the millisecond range with temperature modulated operation is obvious. Commercial micromachined metal oxide based gas sensors (AppliedSensor type AS MLC) have been investigated using the temperature modulated operation as depicted in fig 4a: The sensor temperature is stepwise varied in the range from 100°C up to 400°C, while the sensor resistance is recorded in set of 90 data points with a time interval of 10 ms. Each data point corresponds to the sensor resistance at a specific sensor temperature with a specific gas response. In this way, a virtual sensor array is created, allowing the distinction of different gaseous substances. As an example, the signal evaluation of the temperature modulated response has been optimized for the suppression of NO2 cross sensitivity as depicted in fig 4b.

1470

Roland Pohle / Procedia Engineering 47 (2012) 1466 – 1473

2.0

500

10k

gas sensor temperature modulated operation

AppliedSensor sensor in temperature modulated operation

heater power = 5mWsec = 0.5 mW at 10sec sample rate

200

4k

100

2k

0 0.0

0.2

0.4

0 0.8

0.6

1.0

H2

CO

NO2

0.5 continuous readout evaluation of temperature modulated signal

100

70

r.h. (%)

H2 (ppm) CO (ppm) NO2 (ppm)

10 1

60 50 40

0.1

600

630

660

time (sec)

(a)

690

720

time (min)

750

780

30

810

(b)

Fig. 4. (a) example for the temperature profile and the related sensor resistance used for detection of human activities; (b) Suppression of NO2 cross sensitivity by adapted signal evaluation of the sensor response in temperature modulated operation.

The ability to recognize human activities with the help of gas sensors has been investigated with different sensor technologies and signal evaluation approaches [9,10]. For our approach we combine the MEMS metal oxide gas sensor, which shows excellent response to human induced Volatile Organic Compounds (VOCs) [11] with both temperature-modulated operation and statistic methods for data evaluation. The responses of several temperature modulated gas sensors installed in different places in a test apartment are shown in Fig. 5a. The sensors respond clearly to the number of people in the room, if windows are opened or by cooking. The evaluation of the sensor data regarding more abstracted activities of daily living (fig. 5b) demonstrates the capability of non invasive activity monitoring by temperaturemodulated metal oxide gas sensors.

gas sensor response in test appartment ceiling kit chen window near electric cooker desk couch

sensor resistance (Ohm)

open kitchen window

100000

4 persons in room

open all windows

prediction personal hygiene

sleeping

2 persons in room 1 persons in room

Prepare and drink coffee

Cooking and eating

10000

12:00

Increased physical activity

cooking

13:00

14:00

15:00

16:00

17:00

Increased physical activity

Cooking and eating

Prepare and drink coffee

sleeping

18:00

time

(a)

(b)

Fig. 5. (a) response of 5 metal oxide sensors in temperature modulated operation in different places of a test apartment; (b) prediction of activities of daily living based on data obtained from temperature modulated metal oxide sensors.

Personal hygiene

rel.humidity (%)

300

Signal (a.u.)

6k

[gas] (ppm)

Tsensor sensor resistance

Tsensor (°C)

sensor resistance (Ohm)

1.5

400

8k

1471

Roland Pohle / Procedia Engineering 47 (2012) 1466 – 1473

4. Pulsed Polarization of Platinum Electrodes on YSZ Electrochemical devices based on oxygen-ion-conducting yttria-stabilized-zirconia (YSZ) are appropriate for applications as oxygen sensor at temperatures higher than 600°C. Almost any automobile being powered by a gasoline combustion engine is equipped with at least one zirconia exhaust gas oxygen sensor (lambda sensor) for detection of the air-to-fuel-ratio or lambda. At lower temperatures, the chemical reactions on the surface of metal-oxide electrodes and electrolytes compete with electrochemical reactions. These simultaneous reactions limit the selectivity to different exhaust gas components. Regarding NOx detection, standard potentiometric sensors suffer under the opposite sign for the emf potential of NO and NO2 which makes it very hard to monitor total NOx [12]. Therefore, a wide range of complex material systems is under investigation in order to obtain reliable NOx detection [13, 14]. A main drawback of this approach is the insufficient knowledge about the long term stability of these material systems. Hence, our approach is to use standard lambda probes known as robust and reliable systems and to apply transient operation techniques to measure NOx concentrations in a reliable way. The transient measurement cycle is shown schematically in fig. 6a [15]: after a positive charging voltage is applied, the voltage supply is disconnected and the self-discharge voltage of the sensor is recorded. This procedure is then repeated used a voltage with opposite sign. A positive charging voltage is defined as a higher potential of the outer exhaust electrode with respect to the inner air electrode, which is exposed to a reference gas with defined pO2 (fig. 6b). This pulse sequence with electrode polarisations of opposite signs is applied permanently and the discharge curves are measured continuously. All relevant parameters as voltage, pulse duration and pause time have been varied in order to evaluated their influence of the discharge behavior.

porous detection layer

exhaust electrode

pO 2 pO 2ref

Uheater

US

YSZ solid electrolyte reference electrode rod-type heater

(a)

(b)

Fig. 6 (a): Schematic depiction of the measurement approach; (b) Schematic depiction of a thimble-type Lambda sensor.

The difference in gas response using continuous mixed potential readout (no external voltage applied) and the evaluation of the discharge curve described above is illustrated in fig. 7. Tests with the same gases and gas concentrations have been performed with both methods. By continuous readout of the mixed potential nearly no response to NO was obtained (fig 7c). In contrast, the response to NO is the most prominent compared to other gases if the pulsed discharge method with positive voltage is applied (fig 7a). By polarisation with negative pulses the response to reducing gases is comparable to the mixed potential measurement, while the response to NH3 and NO2 is slightly increased (fig 7b). In comparison,

1472

Roland Pohle / Procedia Engineering 47 (2012) 1466 – 1473

the response to ethanol is similar for continuous readout and by evaluation of the positive discharge curve respectively. Recent investigations indicate that changes in oxidation states of platinum induced by the charging/discharging process plays a major role in the related NOx detection mechanism [16].

(a) after positive pulse

(b) after negative pulse

(c) continuous voltage readout

Figure 7: Comparison of the characteristic curves related to NO, NH3, ethanol, H2 and hydrocarbon mixture (a) by evaluation of the positive discharge curve, (b) by evaluation of the negative discharge curve and (c) with continuous mixed potential readout.

5. Conclusion It has been demonstrated that transient operation is a powerful way to increase the performance of existing gas sensors. Very much depending on the sensor technology and the related application, improvements in stability, selectivity and sensitivity can be obtained without expensive development on the sensor element itself. For SGFET sensors, the separate use of surface and volume effects on the sensing layer is realized by evaluation of signal response with pulsed gate voltage. This method can be used either to suppress cross sensitivities or to extract two independent signals from one sensing element. Temperature modulated operation of commercial available metal oxide sensors has proved not only to increase selectivity, but also to differentiate complex gas mixtures arising from activities of daily living. If pulsed voltages are applied to standard zirconia-based automotive lambda probes and the self-discharge behavior in between the voltage pulses is investigated, a very selective response of the discharge characteristic to different NOx concentrations have been obtained. Despite these promising experimental results, detailed investigations of the physical and chemical effects related to the transient operation are crucial for optimization of the operation parameters for specific requirements.

Roland Pohle / Procedia Engineering 47 (2012) 1466 – 1473

1473

References [1] C. Liu, F. Yan, Radical basis function neural network-based NOx soft sensor technique, International Conference on Electrical and Control Engineering, p. 1152 – 1156, (2011) [2] A. Pawar, R. Kulkarni, Numerical Prediction of NOX in the Exhaust of a Compression Ignition Engine, Proceedings of World Academy of Science: Engineering & Technology (2008) vol. 45 p. 1-4 [3] M. Fleischer, Advances in application potential of adsorptive-type solid state gas sensors. 2008 Meas. Sci. Technol. 19 042001 [4] R. Pohle et al., Realization of a new sensor concept: improved CCFET and SGFET type gas sensors in hybrid flip-chip technology, Proceedings of the 12th International Conference on Solid-State Sensors, Boston, MA, USA, June 8–12, 2003. [5] R. Moos et al., Solid State Gas Sensor Research in Germany – a Status Report, Sensors 2009, 9(6), 4323-4365. [6] R. Pohle et. al., Gate pulsed readout of floating gate fet gas sensors, Procedia Engineering(2010) vol. 5 issue (Eurosensor XXIV Conference) p. 13-16 [7] A.P. Lee, B.J. Reedy, Temperature modulation in semiconductor gas sensing, Sens. Actuators B 60 (1999) 35–42 [8] T.Conrad, P. Reimann, A. Schütze, A hierarchical strategy for under-ground early fire detection based on a T-cycled semiconductor gas sensor, IEEE Sensors (2007) 1221 - 1224 [9] S. Kar, S. Varshney, Accurate estimation of indoor occupancy using gas sensors, in Proc. For International Conference on Intelligent Sensors, pp.355 – 360, 2007. [10] T. Hayashi, K. Haruhiko, T. Oyabu, Recognition of Human Activities by Gas Sensor Response Using Genetic Algorithm, Electronics and Communications in Japan, Part 3, Vol. 85, No. 1, 2002 [11] S. Herberger, et al., Detection of human effluents by a MOS gas sensor in correlation to VOC quantification by GC/MS, in Building and Environment 45, pp. 2430-2439, 2010. [12] F. Ménil, V. Coillard, C. Lucat, Critical review of nitrogen monoxide sensors for exhaust gases of lean burn engines, Sens. Actuators B 67 (2000) 1-23. [13] S. Zhuiykov, N. Miura, Development of zirconia -based potentiometric NOx sensors for automotive and energy industries in the early 21st century: What are the prospects for sensors?, Sens. Actuators B 121 (2007) 639-651. [14] Jeffrey W. Fergus, Materials for high temperature electrochemical NOx gas sensors, Sens. Actuators B 121 (2007) 652-663. [15] S. Fischer et. al., Method for detection of NOx in exhaust gases by pulsed discharge measurements using standard zirconiabased lambda sensors, Sens. Actuators B 147 (2010) 780-785 [16] S. Fischer et. al., Pulsed polarization of platinum electrodes on YSZ, Solid State Ionics (2012) in press