Conductance, work function and catalytic activity of SnO2-based gas sensors

Conductance, work function and catalytic activity of SnO2-based gas sensors

Sensors and Actuators B, 3 (1991) 205-214 Conductance, sensors K. D. Schierbaum, work function 205 and catalytic activity of SnO,-based gas U. W...

909KB Sizes 0 Downloads 5 Views

Sensors and Actuators B, 3 (1991) 205-214

Conductance, sensors K. D. Schierbaum,

work function


and catalytic activity of SnO,-based


U. Weimar and W. Giipel

Institute of Physical and Theoretical Chemistry of the University of Tiibingen, D-7400 Tiibingen (F.R.G.)

R. Kowalkowski CSEY, S-A., CI-I-Zoo0 Neuch&tel (Switzerland) (Received April 5, 1990, in revised form January 5, 1991; accepted February 7, 1991)

Abstract We have studied the geometric microstructure as well as the elemental composition and distribution of Sn02-based polycrystahine sensors with X-ray diffraction and surface spectroscopic techniques. Electrical properties and responses against reducing gases are characterized by a.c. and d.c. measurements of resistances and/or conductances. The models explaining the high sensitivity of polycrystalline SnO, against reducing gases in terms of changes of the intergrain conductivity have been completed. The Schottky-barrier mechanism of the electron transport across the grain boundaries is valid only for SnOr grains larger than the Debye length of electrons. For grains smaller than the Debye length, the band bending at the surface can be neglected compared with the overall shift of the Fermi level in the grains during gas exposure. From combined measurements of conductances, work function changes and catalytic activities as functions of temperature, CO and HZ0 partial pressures, we deduce that OH dipoles, which do not influence the oxidation kinetics of carbon monoxide, are formed during interaction with water at 673 K.

1. Introduction

Polycrystalline SnO,-based Taguchi sensors are practically important gas-sensing devices for detecting reducible gases [l] such as hydrogen (Hz), methane (CHJ, butane (C,H,,) and carbon monoxide (CO) with high sensitivities. The selectivity for a certain gas may be enhanced to some extent by the proper choice of operation temperature or chemical modification 121. The main disadvantage of these devices is the influence of humidity on the magnitude and the stability of sensor signals, i.e., on their changes with time. Several qualitative models have been proposed in the literature which explain the interaction of reducing gases in the presence of water [3,

tial pressures of CO, CH, and Hz0 in air for systematically varied temperatures. With additional information from our earlier spectroscopic results [5], these data will make it possible to describe in detail the elementary steps during sensor/gas interaction. The practical implications of these simultaneous measurements of the three independent sensor parameters have been discussed before. The quantitative determination of several partial pressures of different gases in a mixture is possible by measuring the above-mentioned parameters of the same sensor and applying pattern-recognition methods [6]. This is an alternative approach to evaluating measurements of the same sensor parameter for chemically different sensors [7].


The aim of this study is to develop models which explain this interaction quantitatively for the most commonly used gases. We therefore measured simultaneously the dependence of conductances, work functions and catalytic activities of Taguchi sensors on par-

2. Experimental Scanning electron microscopy (SEM), scanning Auger microscopy (SAM), and scanning energy-dispersive X-ray microanaZysti (EDX)

0 1991 -

Elsevier Sequoia, Lausanne


were performed in a scanning Auger multiprobe system (PHI 600, Perkin Elmer) with an EDX accessory. For Auger electron spectroscopy (AES) and EDX measurements, we used 5 and 25 keV excitation energy at 30 and 5 nA emission current. X-ray diffraction measurements were performed with Cu K, radiation at 28=64.58” and a wavelength of A= 1.5415 x 10-i’ m by using the Au(220) reflection as reference. Impedances g(v), i.e.-reciprocal a.c. conductances (admittances Y) of the sensors were measured at frequencies 5 I v< 13 X lo6 Hz with an impedance analyser (4192A LF, Hewlett-Packard) and independently with a frequency response analyser (1255 HF, Solartron-Schlumberger). The EQUIVALENT CIRCUIT program (B.A. Boucamp, Faculty of Chemical Technology, University of Twente, The Netherlands) was used to fit the experimental g(v) values with equivalent RC circuits. D.c. conductivity measurements were performed at temperatures between 570 and 720 K at different partial pressures of CO, H,, H,O and CH,. Work function measurements were performed under the same conditions with a piezo-driven Kelvin probe (Kelvin Control 05, Besocke, KFA Jiilich) and a molybdenum reference electrode. We checked the stability of this inert reference electrode in measurements of a gold sample for the different CO and H,O partial pressures chosen in this study. We finally measured the catalytic activity of the sensor to produce CO, and H,O molecules upon exposure to CO and CH, in synthetic air. These reaction products were monitored directly by means of a Fourier transform infrared spectrometer (FTIR IFS 48, Bruker) and indirectly by measuring the CO and H,O partial pressures before and after passing an electrochemical CO (TN 70 253, Bieler & Lang) or a capacitive HZ0 (hygrotest 6400, Testoterm) sensor.

3. Results 3.1. Structure determination A typical electron micrograph of an SnO,based Taguchi sensor is shown in Fig. 1. The elemental composition and the distribution

Fig. 1. Scanning electron micrograph of an Sn02-based sensor (TGS 812). At characteristic positions labelled 1 and 2 the surface chemical compositions were determined by Auger electron spectroscopy. The elements Si, C, Sn, 0 and Al (the last only at point 2) were detected within the resolution depths 5 and 12 nm. The bulk chemical composition as determined by energy-dispersive X-ray analysis (EDX) shows Sn, Al, Si and 0. For quantitative details, see Tables 1 and 2.

of the SnO,, of the dopant Pd, of the substrate a-A1203 (the latter was added to the SnO,paste during fabrication of the sensors [8]) and of the polymerized ethyl silicate binder were investigated with surface-sensitive Auger spectroscopy and bulk-sensitive energy-dispersive X-ray analysis for an undoped SnO, sensor (TGS 812) and a Pd-doped SnO* sensor (TGS 813). Results are given in Tables 1 and 2. From scanning EDX, we conclude the 10 to 50 pm large A1,03 crystals in Fig. 1 to be covered by small SnO, grains with mean diameters in the micro- and submicrometer range (fine structure in Fig. 1, e.g., at spot 1). By means of SAM the existence of SnO, was identified below the resolution limit of SEM at the surface of the A&O3 crystals (e.g., spot 2, see also Table 2). The polymerized binder is distributed homogeneously and forms a Si-0 network which improves the long-term stability of the electrical sensor properties [9]. From the approximately identical linebroadening B(20) = 0.016 * 0.1 (in radians) of the (110) and the (011) reflections at the Bragg angles 8= 13.35” and 16.95”, respectively, in the X-ray diffraction diagrams (Fig. 2) we calculate the statistical average size (referred to the number of crystallites) of SnO, crystallites to be T= 9.4 nm. For this


1. Relative

100 fimXlO0

TGS 812 TGS 813



intensities of SnOr-based

sensors (TGS

812 and TGS 813) as determined

by integration



13.7 13.3

2. Relative

7.8 1.5

SAM peak-to-peak

intensities 1(1)/I(2)


at points 1 and 2 in Fig. 1






0.19 0.11

0.14 0.17


Point 1 Point 2

2.2 3.0

W-kLL) m&LL) limit

0.05 0.12

I -400


“0 = ‘2





Fig. 2. Section of an X-ray diffraction diagram of a Taguchi 812 sensor indicating broad SnOr(Ol1) and SnOr(110) reflections and sharp cr-AIrOs(211) and a-A1r03(110) reflections.

evaluation we assumed t=KA/[cosl9XB(2e)]

(1) with B(28) as the linewidth, 0 the reflection angle, h the wavelength and 0.84 1Ks0.89 the proportionality factor related to the crystallite form and the evaluation method for B(28) [lo, 111. We use the full width at half maximum for B(28) and neglect the small contribution of the instrumental broadening. Contributions from internal stress in the crystals can be neglected because of the absence of a characteristic tg(0) dependence in the line-broadening. The grain sizes can be adjusted by specific metal dopants during the preparation of SnO, sensors [12]. 3.2. Sensorproperties 3.2.1. Compkx impedances A typical result of impedances 2’(v) of TGS 812 sensors measured at T= 673 K in synthetic air (50% r.h.) only and in synthetic air (50%








Fig. 3. Impedances i(v) with imaginary (Im 2) and real (Re 2) parts of a Taguchi 812 sensor in synthetic air (50% r.h.) and in synthetic air (50% r.h.) with pco=50 ppm at T=670 K. The semicircles are fitted by equivalent RC units. The physical origin of a very small partial-pressure dependence of the resistance RZ (2.4 X ldsR, ~3.2 X 103 II)could not be clarified because of relatively high errors in its determination (R,BRJ.

r.h.) with carbon monoxide (PC0 =50 ppm) for frequencies 1s vs 10’ s- ’ 4 shown in Fig. 3. The experimental data Z(V) can be fitted by simple equivalent circuits consisting of a pa-dependent resistance RI and a capacitance C, = 15 pF, which is to a good approximation independent of pc-,. We observed an increase of C1 during exposure to HP As a typical value, we find C, = 40.2 pF for pw=300 ppm. 3.2.2. D. c. cortductances Figure 4 shows the time dependence of the d.c. conductance G (which shows completely ohmic behaviour) measured in synthetic humid air (50% r.h.) during repeated exposure to CO with different partial pressures 30 spcO 5300 ppm. The conductance



w CO exposure

300 I

100 -


1Pco~Ppm)l 50 n

30 -






300 t





Fig. 4. Time-dependent changes of the conductance G of a Taguchi 812 sensor in synthetic air (50% r.h.) and during exposure to CO with different partial pressures pco at constant temperature. A small baseline drift is shown for

observed during variations of the water partial pressure from nominally dry air (i.e., for pHzO < 360 ppm under our experimental conditions) to humid air (e.g., topHzo 2 8000 ppm in Fig. .5), if the temperature was kept constant at a value T<800 K. Figures 6 and 7 show typical results on relative changes G/G,, of conductances as functions of CO and Hz0 partial pressures with reference to the conductance G, measured under ‘standard conditions’. The latter are defined by ‘standard partial pressures’ pi.0 of the gas component i (30 ppm CO and 13 860 ppm H20, i.e., 50% relative humidity). The relative changes GIG,, as obtained in our experiments may to a good approximation be described by

PC0 = 0.

with nCO,G and It&O,G as parameters determined from the slopes in Figs. 6 and 7. Above a critical value of pHzO (2 15 000 ppm), the characteristic parameter nco,G k COnStaId and independent of pHzO. In line with our earlier studies [7] and for simplification of the fol-

PC0 [ppml

Fig. 5. Conductance of partial pressures T=760 K.

G of a Taguchi of CO and

812 sensor as a function HZ0 at a temperature

G is a state function in the thermodynamic sense [7] if variations of the partial pressures are kept within well-defined ranges, as shown in Fig. 5 for a typical example with 30 spco I 400 ppm and 8000 spHIO I 18 700 ppm. In contrast to the state function characteristics of G of an individual sensor, the conductances G (and the corresponding frequency-dep_end_ent conductances, i.e., the admittances Y=Z- ‘) of different sensors show large differences. As indicated in Fig. 4, low drifts of G and hence deviations from the state function characteristics are observed only in pure synthetic air, i.e., for pco =O. In contrast, complete reversibility of G was

_L? 4.5 4”




c.3 -1











(.!k.-1 pco.0

Fig. 6. Relative conductance changes (G/G,) of a Taguchi 812 sensor as a function of relative CO partial pressure changes @co/pco~o) in a log/log scale for different HZ0 partial pressures at a sensor temperature of 670 K with the parameter ncO,c as defined in eqn. (2).




‘wml 20




1’ 3.53-





Fig. 7. Relative conductance changes (G/G,) of a Taguchi 812 sensor as a function of relative Hz0 partial pressure changes (~+~,$n~~,~) in a log/log scale for different CO partial pressures at a sensor temperature of 670 K with the parameter nH2c.c as defined in eqn. (2).

lowing description, we always refer the G,, conductance value to 30 ppm CO and 13 860 ppm H,O (i.e., 50% relative humidity in air). __I

3.2.3. Work


( ) PH.0 PHp.0

9. Relative work function changes ([email protected]/A4$,) of a Taguchi 812 sensor as a function of relative changes in Hz0 partial pressures (pm/pm,,) in a linilog scale for different CO partial pressures at a sensor temperature of 670 K determined for pmo> 15 000 ppm (compare eqn. (3)).











b0 0



-is ‘8





2 N

o!? c



0 1.L





Fig. 10. Catalytic reaction rate rw of CO oxidation of a Taguchi 812 sensor monitored by the partial pressure pcor as a function of temperature for different CO partial pressures in a In/T-’ plot. Also indicated are activation energies E,, as determined from the slopes (compare eqn. (4)).

formally be described by the reaction rate r to produce CO*:

(4) Here kO is a pre-exponential factor which includes the activation entropy [13] as well as parameters describing the flow and diffusion conditions through the porous SnOz and the sticking coefficients of CO and Hz0 at the SnOz surface. The activation energy is denoted by EA, and the reaction orders with respect to CO and Hz0 by ncOSat and respectively. Typical results of the nHzO,cat, dependence of rcoz on temperature and CO partial pressure are shown in Figs. 10 and 11, respectively.

4. Discussion 4.1. Conductances and grain-size effects Under ambient air conditions, different chemisorbed donor- and acceptor-type oxygen-related species (such as (OH-),,, and (O-),,) are present at the sur(0,- )a& face of Sn02. These species (except (OH-),,) are involved in the catalytic oxidation of reducing gases in general. The specific ex-

3 Fig. 11. Reaction rates rm monitored byp- of a Taguchi 812 sensor as a function of CO partial pressures in a In/ In scale for different sensor temperatures Ts between 543 and 726 K. Also indicated is the mean reaction order %o,CUas deduced from the slopes (corresponding to eqn. (4)).

ample of CO oxidation will be discussed below. The concentration of acceptor- and donortype chemisorbed species is determined by the temperature- and time-dependent reactions [14] +O,+e(O,-),,

s +e-


(O*- ),a e


2(0-L =


(6) +e-


which establish the steady-state concentration of electronic surface states related to these equilibria. For polycrystalline Taguchi sensors with different grain sixes, the different proposed models [15] for the conduction mechanism are shown schematically in Fig. 12. For simplification, only one acceptor-type species is shown here. In ambient air, surface states lead to depletion layers and hence to a Schottky-barrier-limited electronic conductance (Fig. 12(a)), a surface-trap-limited conductance (Fig. 12(b)), or a grain-controlled conductance (Fig. 12(c) and (d)) across the semiconducting grains. During exposure to reducing gases such as CO, donor-type chemisorption complexes such as CO,,+ are formed, thereby leading to changes in the


Lo< 112

Lo -= 112





Fig. 12. The effect of O2 (. . . . -) and subsequent additional CO (-- -) exposure on grain-boundary barriers as measured by changes in conductances AC and work functions [email protected] Also illustrated are corresponding changes in the electron concentration n and in the energy E of conduction band electrons over the conduction path P through the grains. E; denotes the conduction band edge in SnOr, eAVp the band bending at the grain boundaries in air and Er the Fermi energy. After chemisorption of acceptor-type molecules, the following simplified cases are indicated: case (a) holds for a Schottky-barrier-controlled conductance in non-sintered SnOr. Here, changes of the band bending by a value eAV, at the grain boundaries occur without variation of the bulk value of EC. i.e., A(Ec-EF)b=O and AQ=O. Case (b) holds for an ohmic-contact-controlled conductance through sintered SnOr with grains larger than the Debye length of the electrons (Ln<@. Here, EE changes only at the necks and A(Ec-EF)b= 0 holds. Cases (c) and (d) hold for grains smaller than the Debye length of the electrons (Ln>@2) for non-sintered and sintered SnO, grains, respectively. Here, no difference exists between the bulk and surface position of EC, i.e., A(& -EF)b = A(& -Er)s holds. The concentration of electrons nb does not differ between pressed and sintered small grains.

conductivity through the necks (Fig. 12(b)) and/or within the SnO, grains (Fig. 12(c) and (d)) and/or through the Schottky barriers at the Sn02 grain boundaries (Fig. 12(a)). The electron conduction through necks with an average size of 70-90% of the crystallite size [12] is determined by the extension of the depletion layer, i.e., the Debye length L,, of electrons:

centration of the conduction path within the intergrain region. In -contrast to this, for SnO, crystallites with I < LD the electron concentration changes drastically and homogeneously within the grains upon changes in the occupation of surface states and we expect an exponential dependence of G on the temperature according to G = PO, exp( - [EC - E,],/kT)


(8) within individual single Sn02 grains. Here, E denotes the dielectric coefficient, lo the vacuum permittivity, k the Boltzmann constant, T the temperature, e the elementary charge and nb the bulk concentrations of the electrons. For grains with diameters f large compared to the Debye length LD (1 > L,), the depletion layer affects sensitively the local electron con-


This value is influenced by changes in the difference between the Fermi level EF and the conduction band edge EC in the bulk upon gas exposure. In this case, flat-band conditions are fulfilled to a good approximation (see Fig. 12(c)) and a0 in eqn. (9) denotes the temperature-dependent bulk conductivity. The latter is determined by concentrations and ionization energies of bulk donor and acceptor states as well as mobilities of conduction electrons. The proportionality constant P in eqn. (9) characterizes the sample


geometry and the optimum percolation paths expected theoretically for T-, ~0 for conduction of electrons through the sample. In our model these percolation paths are assumed to, be independent of the temperature and the partial pressures of the gas components. For Schottky-barrier-controlled conduction (I> L,) we expect a characteristic dependence of G on the band bending eAV, [15, 161 according to G = PgO exp( - eAVJk7)

AOlmeV 70




For Sn02 crystallites with diameters Z-
[email protected] IAQ Fig. 13. Logarithm of relative changes of conductance (log(G/G,) as a function of relative work function changes ([email protected]/AQ determined from results in Figs. 6 and 8.


of the Fermi

level Er according


[email protected]= -eAV,+Ax+A(E,-E,), (II) The concentration of surface charges determines eV, and/or (EC--E,), and is directly linked to the conductivity. To a first approximation we therefore assume eAV, or A(E,-E,), to be zero for a constant relative conductance G/G0 (Fig. 13, broken line). We then estimate changes in Ax from [email protected] for changes in the water partial pressure between 360 (dry air) and 9000 ppm. These changes result from the formatiorrof OH dipoles which influence only the local electrical fields at the grain boundaries on the atomic scale without affecting the conductivity. At higher values of pHzO, the OH groups interact with each other. This leads to partial compensation of effective dipoles and hence to lower changes in Ax upon further increased water pressure at p&O 2 9000 ppm. 4.3. Catalysis and kinetic behaviour For a brief discussion of catalytic effects, we focus ‘on one example only. We discuss the catalytic oxidation of CO as characterized of the rate by reaction rates rco2. Parameters eqn. (4) may be explained by assuming pressure-independent constant concentrations of


oxygen species at the surface. We also deduce Henry’s law to hold for CO, i.e., we find the concentration of adsorbed CO to increase in proportion to the partial pressure pco. From the slopes of the Arrhenius-type plot ln[Pco2] versus T- ’ in Fig. 10, we determine activation energies E, of CO oxidation of about 0.16 eV in the partial pressure range 30400 ppm. The reaction order nCOscatwith respect to CO is close to one (Fig. 10). In contrast, the corresponding value is zero, i.e., it is independent of CO partial pressures. This indicates that OH groups do not affect the catalytic activity for CO oxidation but strongly influence the electrical characteristics of the electron transfer across the grain boundaries in polycrystalline SnO, sensors. We conclude, therefore, that the CO molecules are oxidized only by ionic oxygen species at the surface with the CO chemisorption as the rate-limiting step.

gases with SnOr-based

in the detection of reducing devices, Sensors and Actuators,

28 (1989) 71-113. 5 K. D. Schierbaum,

H. D. Wiemhofer and W. Gopel, Defect structure and sensing mechanism of SnO, gas sensors: comparative electrical and spectroscopic studies, Solid State Ionics, 28-30 (1988) 1631-1636. 6 U. Weimar, K. D. Schierbaum and W. Gopei, Pattern recognition methods for gas mixture analysis: application to sensor arrays based upon SnOs, Sensor and Actuators, BI (1990) 93-96. 7 K. D. Schierbaum, U. Weimar



10 11

and W. Gopel, Multicomponent gas-analysis: an analytical chemistry approach applied to modified SnOr sensors, Sensors and Actuators B, 2 (1990) 71-78. K. Ihokura, Tin oxide gas sensor for deoxidizing gas, New Maten’aLs and New Processes in Electrochemical Technology, Vol. 1, 1981, pp. 43-50. S. Yasunaga, S. Sunahara and K. Ihokura, Effects of tetraethyl orthosilicate binder on the characteristics of an SnO, ceramic type semiconductor gas sensor, Sensors and Actuators, 9 (1986) 133-145. A. Baiker, Experimentelle Methoden der Katalysatorcharakterisierung, Chimiu, 35 (1981) 4401146. C. R. Adams, H. A. Benesi, R. M. Curtis and A. G. Meisenheimer, Particle size determination of supported catalytic metals: platinum on silica gel, J. Cut., 1 (1962) 336-344.

5. Outlook

12 N. Yamazoe, New approaches for improving semiconductot gassensors,Proc. 3rd. Int. Meet. ChemicalSensors,

We have demonstrated that measurements of complex impedances, conductances and work functions as well as catalytic activities are tools to investigate sensitively the electronic and chemical processes during gas/ sensor interactions. These measurements may also serve as independent signals for pattern recognition [6] to detect different gas components in air with the same sensor.

Acknowledgement We would like to acknowledge the financial support of Land Baden-Wiirttemberg (Forsch. Schwerpunkt 39) and Fond der chemischen Industrie for this work.


Techn. Mess., 52 (2) (1985) 47; (3) (1985) 92, (5) (1985) 175. N. Yamazoe, Y. Kurokawa and T. Seiyama, Effects of


4 D. Kohl, Surface processes

on semiconductor

gas sensors,

Actuators, 4 (1983) 283-289. S. R. Morrison, Semiconductor and Actuators 2 (1982) 329-341.

Sensors and

gas sensors,





1990, pp. 3-8.

13 W. Gopel, Chemisorption and charge transfer at semiconductor surfaces: implications for designing gas sensors, Frog. Suq? Sci., 20 (1985) 9. 14 V. Lantto and P. Romppainen, Electrical studies on the reactions of CO with different oxygen species on SnOr surfaces, Sulj: Sci, I92 (1987) 243-264. 15 J. F. McAIeer, P. T. Moseley, J. 0. Norris and D. E. Williams, Tin dioxide gas sensors: aspects of the surface chemistry revealed by electrical conductance variations, J. Chem. Sot., Faraday Trans. I, 83 (1987) 1323-1346. 16 J. F. McAleer, P. T. Moseley, J. 0. Norris and D. E. Williams, Factors affecting the performance of SnOr as a gas sensor, Proc. Brit. Ceram. Sot., 36 (1985) 89-105.

Biographies Klaus-Dieter Schierbaum obtained his Ph.D. in chemistry in 1987 in the field of chemical sensors and interface analysis. He is now appointed as assistant professor at the Institute of Physical and Theoretical Chemistry at the University of Tiibingen, F.R.G. He is in charge of the Center of Interface Analysis and Sensors. Udo Weimar received his diploma in physics at the University of Tiibingen in the field of


chemical sensors and pattern recognition in 1988. He is now working on his Ph.D. thesis in the same field. Worfang Giiper received his Ph.D. from the University of Hannover in 1971. After visiting scientist positions at the Xerox Palo Alto (CA), Xerox Webster (NY) and IBM Watson Research Center (NY), he was appointed as full professor of physics at the Center of Surface and Submicron Analysis, Bozeman

(MT). Since 1983 he has been director of the Institute of Physical and Theoretical Chemistry at the University of Tubingen, with research interests in interface properties of new materials for chemical sensors, catalysts and microelectronic devices. Rainer Kowalkowski received his Ph.D. in chemistry in 1987 at the University of Tubingen in the field of metal oxide gas sensors. He is now working at CSEM, S.A., NeuchAtel (Switzerland).