You are on page 1of 13

Sensors and Actuators B 107 (2005) 587–599

A rate equation approach to the gas sensitivity of thin


film metal oxide materials
S. Ahlersa,∗ , G. Müllera , T. Dollb
a Corporate Research Centre, EADS Deutschland GmbH, D-81663 München, Germany
b IMM Institut für Mikrotechnologie Mainz GmbH, Mainz, Germany

Received 14 July 2004; received in revised form 26 October 2004; accepted 12 November 2004
Available online 6 January 2005

Abstract

Thin film metal oxide materials exhibit a bell-shaped variation of the gas sensitivity with sensor operation temperature. With respect
to the temperature TM at which a sensitivity maximum occurs, the distribution of the gas sensitivity is asymmetric exhibiting a relatively
steep increase below TM and a more moderate drop-off above TM . In this paper a rate equation approach is described, which successfully
reproduces temperature- and gas-concentration dependent sensitivity distributions S(T, cgas ) experimentally determined for a number of
reducing analyte gas molecules. We show that such distributions are determined by two energetic parameters, which are specific for the
special adsorbate/adsorbent system involved. These are (i) the strength of adsorption of neutral analyte gas molecules Eads and (ii) the kinetic
barrier Ea that needs to be overcome to induce a surface combustion event involving an adsorbed analyte gas molecule and a surface oxygen
ion.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Metal oxide; Modeling; Thin film; Adsorption

1. Introduction with the kinetics of adsorption and desorption on the sensor


surface. These models yield functional relationships between
Metal oxide gas sensors sensitively respond to a wide the sensor signal and the analyte gas concentration applied,
range of oxidising and reducing analyte gases via conduc- i.e. the calibration curve for a particular kind of analyte
tivity changes. The gas response profiles S(T, cgas ) vary gas at a fixed sensor operation temperature. The variation
with sensor operation temperature T and analyte gas con- of the gas sensitivity with sensor operation temperature,
centration cgas in a manner specific for the particular adsor- on the other hand, has been more explicitly considered in
bate/adsorbent system involved. In general, such profiles vary a second group of papers, which is concerned with porous
in a bell-shaped manner with temperature, exhibiting a sensi- thick-film materials. In this latter group of publications
tivity maximum SM at an analyte-specific temperature TM and [5–10] diffusion–reaction theory [11,12] is invoked to
a lower sensitivity above and below this temperature [1–3]. calculate temperature-dependent gas penetration profiles
The variation with gas concentration is universally found to within a porous thick-film sensing layer. These latter models
be sublinear [4]. successfully reproduce the observed bell-shaped variation of
Recently, there have been several attempts at explaining the gas sensitivity with sensor operation temperature. With
the peculiarities of metal oxide gas-sensing materials. Those regard to the temperature TM of maximum gas sensitivity
included in the review of Barsan et al. [4] are concerned SM , the low-temperature drop-off of the gas sensitivity
is explained by a decreasing reactivity of the analyte gas
∗ Corresponding author. Tel.: +49 89 607 21074; fax: +49 89 607 24001. molecules and a concomitantly deeper gas penetration. The
E-mail address: simon.ahlers@eads.net (S. Ahlers). high-temperature drop-off, on the other hand, is explained

0925-4005/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.snb.2004.11.020
588 S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599

by an increasing reactivity and an increasingly shorter gas


penetration into the sensing layer. The intriguing aspect
with these latter models is that they predict a continuously
increasing gas sensitivity with sensor operation temperature
in the limit of very thin films. As such an effect is not
observed in reality, it is suggested that the basic reasons
for the high-temperature drop-off of the gas sensitivity
are more likely due to adsorption–desorption rather than
diffusion–reaction phenomena.
In the present paper, we present gas sensitivity measure-
ments on thin-film tin dioxide films as a function of sensor
operation temperature and analyte gas concentration. These
measurements confirm that both films with smooth and gran-
ular surface morphology exhibit qualitatively the same kinds
of bell-shaped gas sensitivity profiles as porous thick-film Fig. 1. Side view of a compact tin dioxide film generated by e-beam evap-
materials. We then proceed to develop a simple rate equa- oration of SnO2 material.
tion approach for the thin-film gas sensitivity that builds on
Aiming at modelling gas sensor behaviour, it is advisable
early ideas of Windischmann and Mark [13] and later elab-
to reduce the number of poorly controlled parameters with
orations by Barsan and Weimar [14]. The model proposed
a wide statistical distribution as far as possible. Sensor mor-
describes the sensitivity towards reducing gases in terms of
phology clearly is such a problem parameter as the size of the
adsorption–desorption processes involving both analyte gas
metal oxide grains and the shape of sintering necks between
and oxygen molecules. We show that once the sensor baseline
adjacent grains cannot be controlled tightly enough (note ex-
resistance has been fixed by assuming appropriate rate con-
cept [28]), or even determined experimentally. As compact
stants for the oxygen adsorption and desorption, the exper-
tin oxide layers do not exhibit such complicated morphol-
imentally observed gas sensitivity profiles can be described
ogy, such films are best suited to serve as a model system for
in terms of two energetic parameters that are specific for the
studying gas surface interactions.
particular analyte gas species. These are: (i) the strength of
The method used to deposit compact sensing layers was
adsorption of the analyte gas molecules on the sensor surface
electron beam evaporation (PVD) of pure SnO2 pellets. The
and (ii) the activation energy for triggering a surface com-
resulting material turns out to be slightly oxygen-deficient,
bustion reaction between an adsorbed analyte molecule and
which results in a finite electrical conductivity, which is
a surface oxygen ion. By leaving out some of the complex-
needed for supporting a gas sensing effect. In order to put
ity of the Barsan model [14], we are able to gain values for
the gas response of such compact layers into perspective with
the two energy parameters from fits to the experimental data.
“typical” porous gas sensing films, precursor films of metal-
An encouraging feature is that both energy parameters, and
lic tin were evaporated and oxidised afterwards to form tin
in particular, the activation energy are clearly lowered upon
dioxide by thermal annealing at 600 ◦ C in ambient air.
introducing catalyst impurities.
The following two SEM images (Figs. 1 and 2) reveal the
With the values of these two energy parameters being de-
morphology of the two different types of sensing layers.
termined it is also possible to generalise our model to describ-
Granular tin dioxide films were also used to study the
ing porous thick-film materials. This generalisation, which
impact of noble-metal dopants on the gas sensitivity distri-
includes diffusion–reaction processes in addition to the more
basic surface interactions, will be presented elsewhere [15].

2. Preparation of thin-film tin oxide materials

Throughout the literature available today metal oxide lay-


ers, which are considered for gas sensor applications, exhibit
a very porous morphology, almost regardless of the prepa-
ration method employed [16–21]. Porous metal oxide layers
are of interest for two reasons: firstly, a sharp increase in the
gas sensitivity is observed as the grain size is reduced below
about 30 nm [22–27]; secondly, small grains give rise to a high
degree of porosity which enhances the penetration of gases
into thick-film sensing layers, i.e. into those materials that
are almost exclusively employed in present-day commercial Fig. 2. View onto a porous tin oxide layer as obtained by e-beam evaporation
metal oxide sensor devices. of Sn and subsequent thermal oxidation in laboratory air.
S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599 589

Fig. 3. PVD deposition of noble metal doped thin film sensors. A sandwich
consisting of 10 nm tin and 5 nm catalyst layers is first evaporated onto the
substrate. Subsequently the layer stack is subjected to thermal oxidation
in ambient air, resulting in a catalytically enhanced tin oxide sensor with Fig. 4. Maximum response SM of a compact (dark grey) and a porous (light
granular morphology. grey) tin dioxide layer to reducing gases. Whereas a compact film morphol-
ogy reduces the gas sensitivity, it is more convenient to analyse theoretically.
The operating temperature of the sensors is not fixed in this graph, rather the
butions. In order to obtain such doped material, multi-layer optimum is chosen for each gas species.
stacks of tin and dopant materials were evaporated as indi-
cated in Fig. 3 and annealed in ambient air after deposition to charge regions. Such modulation effects, however, are much
obtain SnO2 material containing agglomerates of dispersed more pronounced at the sintering necks of a porous layer
noble metal. because of the associated small cross sections of the current
paths. This matter has already been discussed in considerable
detail in the literature [23,30–32].
3. Results of gas measurements In the following, only results on compact layers will be
considered. Relevant results on such layers are presented in
In the following we will use the terms “response” and Figs. 5–7. Granular layers will be considered in Section 4.2.
“sensitivity” synonymously. Mathematically the response S The sensitivity profiles reported in Figs. 5–7 exhibit sev-
is defined as follows: eral characteristics, which are common to a wide variety of
R0 other metal oxide sensing layers: (i) the sensitivity at room
S= −1 (1)
Rgas temperature and up to about 100 ◦ C is small for all gases and
gas concentrations considered; (ii) with rising temperature
where R0 is the sensor resistance in clean air and Rgas is the the response increases up to a certain temperature TM where
sensor resistance under the influence of a reducing gas.
This formula is used because it produces reasonable values
ranging from 0 for clean air to high positive values for strong
interactions with reducing gas species. The values generated
are consistent with those obtained by the common formula
S = Rgas /R0 − 1 for oxidising gases.
Turning to the results of our gas sensing experiments, we
first compare the gas response of tin dioxide films with a com-
pact and a granular morphology. This response was measured
for various reducing gases over a range of sensor operation
temperatures. Specifically the gases ethene (C2 H4 ), carbon
monoxide (CO) and hydrogen (H2 ) were considered as sum- Fig. 5. Gas response of a compact tin dioxide layer to hydrogen. The bell-
marised in Table 1. shaped variation of the gas sensitivity with sensor operation temperature is
Fig. 4 displays the response of both kinds of layers under typical of metal oxide gas sensors. The maximum response to 10.000 ppm
identical conditions of gas exposure. (=1%) H2 is comparably weak.
Obviously, the response of a porous tin dioxide layer is
larger than that of a compact layer [29]. In both cases – and
independent of the film morphology – gas reactions at the ac-
cessible surface area cause a modulation of the surface space

Table 1
Test gases used
Gas Concentration (ppm)
C2 H4 20–5000
CO 2–500
H2 100–10000 Fig. 6. Gas response of a compact tin dioxide layer to ethene (C2 H4 ).
590 S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599

Fig. 7. Gas response of a compact tin dioxide film to carbon monoxide CO.
Fig. 9. Variation of the C2 H4 sensitivity with sensor operation temperature
for tin dioxide samples with granular morphology containing different types
a sensitivity maximum SM occurs. This temperature depends of noble metal impurities. The test gas mixture applied was 5000 ppm C2 H4
both on the kind and on the concentration of analyte gas ap- in synthetic air.
plied; (iii) at sensor operation temperatures higher than TM
the response drops again, however, more slowly than below
TM . In particular, a vanishing gas response is not reached,
even at the highest temperatures applied.
Considering these results, it is not very surprising to find
that the temperature of maximum response TM is gas-species
dependent. It is perhaps more surprising that TM also depends
on the analyte gas concentration. This is not at all obvious
at first sight. The theoretical considerations in the next sec-
tion, however, indicate that, in principle, TM should exhibit
a much stronger gas concentration dependence than actually
observed.
Fig. 10. Variation of the CO sensitivity with sensor operation temperature
As will be discussed below, the relatively weak depen- for tin dioxide samples with granular morphology containing different types
dence of TM on the gas concentration is evidence for a of noble metal impurities. The test gas mixture applied was 500 ppm CO in
concentration-dependence of the thermodynamic and kinetic synthetic air.
parameters that characterise a particular adsorbent/adsorbate
system. lysts see [3,33–40]. We will see in the discussion below that
Turning to SnO2 layers with a granular surface morphol- the strong effects of noble metal catalysts on the response to
ogy, we demonstrate the effect of incorporating noble metal some gases is reflected in the physical parameters that govern
impurities. Figs. 8–10 display the gas response of granular the sensing behaviour. Especially the lowering of the reac-
SnO2 films with regard to H2 , C2 H4 and CO. In addition, tion activation energy (introduced in Section 4.1.1), which
these plots contain data for materials that were doped by the is commonly associated with catalysts, can be directly ob-
addition of several percent of platinum (Pt) and gold (Au). served.
These latter data vividly demonstrate that noble metal doping
can have a very severe impact on the magnitude of the gas
response S and on the temperature TM at which the sensitivity 4. Modelling the sensing behaviour
maximum SM occurs. For phenomenological and theoretical
considerations about the effects of adding noble metal cata- 4.1. Compact films

Throughout this section, we consider our metal oxide lay-


ers as being compact slabs as displayed in Fig. 11. This means

Fig. 8. Variation of the H2 sensitivity with sensor operation temperature for


tin dioxide samples with granular morphology containing different types of Fig. 11. The assumed metal oxide layer geometry. The layer is compact and
noble metal impurities. The test gas mixture applied was 1% H2 in synthetic as gas/surface interactions are confined to the free surface, gas penetration
air. effects do not need to be considered.
S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599 591

that gas diffusion into the sensing layer can be neglected and
that all gas interactions take place at the free surface of the
slab. This also means that there is only one single surface
depletion layer, which can become modulated in response to
the gas/surface interactions. Because of the simple geometry,
the depletion layer can be described by a single parameter,
i.e. the space charge width WSCR .
Changes in the depletion layer width modulate the cur-
rent through the undepleted bulk very much in the same way
as in a depleted thin film transistor. As metal oxide surfaces
can interact with an adjoining gas phase in a chemically re-
versible way, the arrangement of Fig. 11 basically represents
a chemically sensitive thin film transistor. The properties of
such transistors are discussed in more detail below. Fig. 12. Exchange of electronic charge across the surface barrier. In addition
to thermionic emission over the surface barrier hopping through localised
4.1.1. Surface combustion of reducing molecules donor states may occur in the limit of high oxygen vacancy concentrations
Under clean air conditions, the surface of a metal oxide is and thin surface barriers. More details are given in Appendix A.
covered by adsorbed oxygen. Depending on the temperature
of the sensor, there are different forms of adsorbate possi-
ble [4,41]. At typical sensor operation temperatures of above
150 ◦ C the prevailing adsorbate is O− . The surface coverage In this spirit the following rate equation can be written:
with the latter creates a space charge region and thus con-
 
trols the baseline of the sensor resistance. The surface band d 2(EC + qVs − EF bulk )
bending qVs associated with this space charge region is: NO = κf0 pO2 NC exp −
2
dt kB T
 
q2 NO2 2(EC − EO minus )
qVs = (2) − κr0 NO2 exp − (4)
2εε0 nD kB T
where NO is the surface density of surface oxygen ions; nD
the donor density; qVs the surface band bending; q the ele- where NO is the surface density of surface oxygen ions; pO2
mentary charge; ε0 and ε the absolute and relative values of is the oxygen partial pressure; κf0 and κr0 are the kinetic
the dielectric constant. parameters for adsorption and desorption.
In case reducing analyte gases prevail in the ambient at- Considering the fact that Eq. (4) contains the band bending
mosphere, the areal density of oxygen ions on the surface qVs on the right-hand side and that the band bending in turn
is changed. In this case, surface combustion reactions take depends on the surface ion density NO through Eq. (2), it is
place, transforming the reducing gas species and a corre- clear that (4) is an implicit equation that can only be solved
sponding number of surface oxygen ions into CO2 and H2 O by numerical means.
molecules, which are subsequently desorbed into the gas We now generalise this equation to the case that reducing
phase again [35,42–44]. The electrons formerly trapped at analyte gas molecules are present. In this case reactions take
the surface oxygen ions are returned to the metal oxide con- place which reduce the density of surface oxygen ions NO .
duction band, which causes the space charge layer thickness In order for those reactions to take place, the following steps
to shrink and the sensing layer resistance to decrease. need to be considered: first, reducing gas molecules need to
In order to arrive at a rate-equation formalism able to adsorb on the metal oxide surface. This will lead to a certain
describe gas detection via surface combustion processes, surface coverage with this new kind of adsorbate. A frac-
the rate of adsorption and desorption of oxygen ions at a tion of the adsorbed molecules in turn may exchange charge
metal oxide surface needs to be considered. Starting from the with the metal oxide bulk, but these adsorbates will then be
ionosorption reaction fixed on the surface and therefore likely to be unable to travel
close to one of the surface oxygen ions [43]. Those analyte
O2(g) + 2e− ↔ 2O(s) − , (3)
molecules, which did not exchange charge with the metal ox-
it is seen that ionosorption requires two electrons to be ther- ide surface, are able to diffuse along the surface until they run
mally emitted from the Fermi energy across the surface bar- into a surface oxygen ion, where they are likely to undergo
rier (EC + qVs − EF bulk ) to become trapped at a pair of ad- a catalysed surface combustion event. Triggering such a re-
sorbed oxygen atoms. As the surface oxygen ions formed action, a kinetic barrier, represented by an activation energy
give rise to energy levels whose positions are fixed relative Ea , needs to be overcome. The reaction products – usually
to the position of the band edges, desorption of the surface CO2 or H2 O – will then leave the surface and the electrons
oxygen ions requires re-emission of two electrons across an formerly trapped at surface oxygen ions are released to the
energy barrier of height (EC − EO minus ) (Fig. 12). conduction band.
592 S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599


In this spirit Eq. (5) is modified by a reaction term Re to κf0
NO air = pO n 2 (12)
reflect the additional possibility of surface combustion: κr0 2 D
 
d 2(EC + qVs − EF bulk ) with the value of NO being determined, the width of the space
NO = κf0 pO2 NC exp −
2
dt kB T charge layer can be calculated:
 
2(EC − EO minus ) NO
− κr0 NO exp −
2
− Re (5) WSCR = (13)
kB T nD
  With the total film thickness being D, the actual resistance
Ea
Re = LNO k0 exp − (6) of the metal oxide layer RMOX turns out to be:
kB T
1 Ls
The term Re itself contains a number of important factors: RMOX = (14)
σMOX Bs (Ds − WSCR )
first, a Langmuir relative surface coverage L, which reflects
the adsorption of the reducing gas without charge transfer. σMOX = µn qns (15)
L in turn contains several important parameters [45]:
where Ls is the contact distance; Bs is the contact width; Ds
pgas is the thickness of the metal oxide; WSCR is the width of
L= (7)
pgas + p0 the space charge region; nS is the conduction electron con-
  centration (≈nD at elevated temperatures); µn is the electron
kB T Eads
P0 = exp − (8) mobility.
VQ kB T As one is usually not so much interested in the baseline
 1.5 resistance of the sensor itself, but in its gas response, we use
2πh̄
VQ = (9)
Mgas M0 kB Tgas Rair
S= −1 (16)
Rgas
where VQ is the quantum volume of the reducing gas; pgas
is the partial pressure of the reducing gas; Tgas is the gas to deduce from Eq. (14) a final formula for the gas response:
temperature (300 K); T is the sensor operation temperature;
Eads is the binding energy of adsorbate; M0 is the atomic mass DS − WSCR gas
S(T, cgas ) = −1 (17)
unit (1.67 × 10−27 kg); Mgas is the relative atomic mass of the DS − WSCR air
reducing gas (e.g. 28 for CO and 2 for H2 ).
where WSCR air is the space charge region width in clean air;
Assuming that a situation of dynamical equilibrium holds,
WSCR gas is the space charge region width in the presence of
the time rate of change of the oxygen ion density vanishes
reducing gas; Ds is the metal oxide layer thickness.
and Eq. (5) can be solved to obtain the density of surface
In this latter formula the electrical parameters µn and nD
oxygen ions both under clean-air as well as under reducing
of the sensing layer cancel out.
analyte gas conditions. Again Eq. (5) is an implicit equation
Combining Eqs. (7)–(9), (11), (13) and (17), an explicit ex-
that can only be solved by numerical means. A more straight-
pression S(T, cgas ) for the gas response of a metal oxide layer
forward analytical solution, however, can be obtained when
can be given. This expression is very lengthy and not very
more restrictive assumptions are made. For reasons detailed
enlightening. Rather than presenting it, simulated response
in Appendix A, we will consider the barriers at the surface to
patterns are displayed in Section 5 below (see Fig. 15).
be transparent for tunnelling by electrons. In this latter case,
It is relevant to note that the bell-shaped variation of the
Eq. (4) reduces to the form put forth by Barsan et al. in [4],
gas sensitivity with sensor operation temperature is deter-
either (Eq. (2) in [4]) without a reaction term, or (Eq. (31) in
mined by the cooperation of the Langmuir adsorption and
[4]) with a reaction term.
the first-order reaction terms in Eq. (6). This latter fact is
Assuming steady-state conditions, an equation for the
illustrated by the data plotted in Figs. 13 and 14. Whereas
areal density of surface oxygen ions can then written:
the first plot shows how these terms vary with temperature
 
Ea individually, the second plot illustrates how both terms co-
0 = κf0 pO2 nD − κr0 NO − LNO k0 exp −
2 2
(10) operate in reproducing the peculiar form of the gas response.
kB T
The striking fact brought out by Fig. 14 is that due to the
This latter equation can be analytically solved to obtain the properties of the Langmuir adsorption term, the sensitivity
surface oxygen ion density both under reducing analyte gas maximum should shift towards higher sensor operation tem-
and clean air conditions (clean air conditions are represented
peratures once higher gas concentrations are applied. We will
by cgas = 0, and therefore L = 0):
see in Section 5 that this latter effect is not observed in reality
    2 and that the absence of a major shift in the sensitivity maxima
Lk0 exp − kEB aT + Lk0 exp − kEB aT + 4κr0 κf0 pO2 n2D
is due to a decrease of the adsorption energy Eads of analyte
NO = (11)
2κr0 molecules with increasing surface coverage.
S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599 593

It is our intention in this paper to arrive at a description of


the sensor with which it is possible to fit experimental results
to the mathematical formulae and get an impression of the
values for some of the physical parameters that cannot be
obtained otherwise. The grain diameter can be estimated by
XRD or REM imaging. However, the neck diameter is not as
easily accessible, but can be just as important.
Therefore, instead of using a formalism that we cannot
justify by experimental evidence, we make a simplification
here by simply introducing a factor that allows an enhance-
ment of the sensitivity of a granular metal oxide film over a
Fig. 13. Variation of the first-order reaction term (broken line) and of the compact one. This takes away some of the sophistication that
Langmuir adsorption terms (full lines) with sensor operation temperature. is already present in published models [14], but allows on the
Higher analyte gas pressure causes the decrease of the Langmuir adsorption other hand fitting experimental data, which is a valuable step.
term to shift towards higher sensor operation temperatures.

5. Extraction of adsorbate/adsorbent-specific
parameters

In this final section, we apply the mathematical expres-


sions derived above to the analysis of thin-film gas sensor
behaviour. Before we start, we recall that these expressions re-
late the temperature- and concentration dependence of the gas
sensitivity S to the energies Ea and Eads , which are analyte-
gas-specific and to the kinetic parameters κf0 and κr0 , which
Fig. 14. The cooperative effect of the first-order reaction term and the Lang-
determine the sensor baseline resistance in ambient air. All
muir adsorption term accounts for the bell-shaped variation of the gas sensi- other parameters relate to material properties, which can ei-
tivity with sensor operation temperature. Concentration-independent values ther be extracted from measurements or estimated using lit-
of the kinetic barrier and Langmuir adsorption energies cause the temper- erature values. For the sake of clarity we summarise these
ature of maximum gas response to shift towards higher sensor operation parameters in Table 2 below.
temperature.
As the ratio κf0 /κr0 determines the sensor baseline resis-
tance, which is independent of the type and concentration of
4.2. Granular thin films analyte gas considered, this latter parameter as well as all
other constants in Table 2 have been kept constant during
In the case of compact films, the geometry is simple and all fitting procedures described below. Making such a choice
the influence of the gases on the conduction straightforward. only leaves two parameters to be determined from data fit-
With granular films, this situation becomes more difficult. ting. We now go on showing how the two analyte-gas-specific
To a first approximation the thickness of the space charge parameters Ea and Eads determine the gas sensing behaviour
region does not change, but it now applies to grains and necks of thin film metal oxide sensing layers.
between grains instead of the compact layer. This fact can be The sensitivity enhancement factor A simplifies the effects
accounted for by exchanging the layer thickness with the of granularity. As this factor is fixed for any single sensor, we
average neck diameter, arriving at the following equation: consider it is merely a morphology-related constant that is
dneck − 2WSCR gas common to all fits relating to different analyte species and
S= −1 (18) analyte gas concentrations.
dneck − 2WSCR air
As a first example we consider thin compact layers of tin
The factor of 2 reflects the fact that now the gas–surface- dioxide. As described in Section 2, these films were formed
interactions can take place on both sides of a grain or neck, by evaporating SnO2 powder and by performing a subse-
instead of only the top surface as in the compact case. Unfor- quent annealing step in ambient air afterwards. While the
tunately, with small grains and, by implication, small necks sensitivity of these films towards H2 , CO and ethene (C2 H4 )
the space charge region may extend over most or all of a is displayed in Figs. 5–7 in Section 3; Fig. 15 shows response
grain. If this is the case, the influence on the conductivity values towards CO in combination with fits to these data. In
of the film is changed not only in quantity, but also in qual- the latter figure measured data are indicated by symbols and
ity. This has been suggested in [20,26] and further discussed the bell-shaped curves interpolating through the experimen-
in [14]. The models discussed there clearly show different tal data are fits to Eq. (17). Comparing these fits it is seen that
regions of grain sizes where different mathematical formula- sensitivity curves relating to different analyte gas concentra-
tions should be used. tions cgas exhibit sensitivity maxima SM that tend to shift to
594 S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599

Table 2
Parameters entering the fit equation S = f(T, cgas ) which stands for the sensor response as a function of sensor operation temperature and gas concentration
pgas Reducing gas partial pressure Fit variable
T Sensor operation temperature Fit variable
Ea Reaction activation energy; analyte-gas-dependent Fit parameter
Eads Binding energy of reducing gas upon chemisorption; Fit parameter
analyte-gas-dependent
κf0 , κr0 Unknown; ratio κf0 /κr0 fixed by clean air oxygen Fit parameter
coverage; independent of reducing gas type
NO clean Clean air oxygen coverage 2.6 × 1012 cm−2 [46]
PO2 Atmospheric oxygen partial pressure 2 × 104 Pa
Tgas Gas temperature 300 K
Mgas Relative mass of the reducing gas MH2 = 2; MCO = MC2 H4 = 28
ma Atomic mass unit (AMU) 1.67 × 10−27 kg
k0 Surface phonon frequency, reaction attempt frequency 1013 Hz [43]
nD Doping concentration of the metal oxide bulk 1018 cm−3 [4,47]
kB Boltzmann constant 8.617 × 10−5 eV/K
A Sensitivity enhancement factor Fit parameter in porous layers,
fixed to 1 for the compact layers

Fig. 15. Gas response of a compact tin oxide layer to carbon monoxide Fig. 16. Fitted values for the strength of adsorption (top) and the reaction
(CO) as a function of the sensor operation temperature. The symbols stand activation energy (bottom) for the pure tin dioxide film and 20–5000 ppm
for measured sensitivity values; the lines are fits to the function S(T, cgas ) ethene as reducing gas. The magnitude of both energies drops with increasing
(Eq. (17)). analyte gas concentration.

slightly higher temperatures TM as larger gas concentrations particular, the temperature TM , at which the sensitivity maxi-
are applied. Sticking to a constant sensor operation temper- mum SM occurs, was shown to coincide with the temperature
ature close to TM , it is further revealed that the gas response at which the Langmuir isobar starts to fall from a relative
S increases in a sub-linear manner with increasing cgas . All surface coverage of Θ = 1 towards Θ = 0. As, due to the prop-
these characteristics are quite generally observed in all kinds erties of the Langmuir isobar this temperature is strongly
of metal oxide materials. concentration-dependent, the maximum of S(T, cgas ) is ex-
From the fits above, values for the energy parameters Ea pected to experience a considerable shift towards higher sen-
and Eads can be obtained. The next series of graphs display sor operation temperatures as long as constant values of Eads
those values of Ea and Eads that were determined from these are assumed (see Fig. 14). In view of the relatively constant
fits. The most striking feature in these data is that both energy values of TM that emerge from the data of Figs. 5–7, sim-
parameters turn out to be concentration-dependent. On the
whole these data show that reaction thresholds with surface
oxygen ions are lowered and that the strength of adsorption
of the analyte gases is reduced as higher analyte gas concen-
trations are applied (Figs. 16–18).
This concentration-dependence of the two energy param-
eters deserves more attention. In Section 4.1, where the func-
tion S(T, cgas ) was derived, it was shown that the bell-shaped
variation of the gas sensitivity arises from the cooperation of
two opposing effects: (i) an increasing probability of trigger-
ing detection reactions as the sensor operation temperature is
raised and (ii) an increasing probability of adsorbed analyte Fig. 17. Fitted values for the strength of adsorption (top) and the reaction
gas molecules to desorb prior to suffering a detection reaction activation energy (bottom) for the pure tin dioxide film and 2–500 ppm CO
as the sensor operation temperature is further increased. In as reducing analyte gas.
S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599 595

A possible alternative reason for a seeming concentration-


dependence of Ea and Eads could originate also in part from
the mathematical simplifications that had been introduced
into the derivation of the simplified rate Eq. (10). In order to
obtain a simple explicit equation for NO , the assumption of
a constant surface electron density ns needed to made. The
validity of this assumption will break down in case surface
barrier profiles are no longer transparent to tunnelling. In this
latter case, a concentration-dependence of ns will arise as in-
creasing concentrations of analyte gases are applied. Even in
Fig. 18. Fitted values for the strength of adsorption (top) and the reaction this latter case the assumption of a constant (although low-
activation energy (bottom) for the pure tin dioxide film and 100 ppm to 1% ered) ns is correct as long as one is dealing with a small-signal
H2 as reducing analyte gas. situation, i.e. with very low analyte gas concentrations. In this
spirit, we prefer to interpret as true analyte-specific energies
those values of Ea and Eads that can be extracted from the
ple mathematical reasons demand concentration-dependent fitted values by extrapolation towards zero analyte gas con-
values of Ea and Eads to be inferred to obtain fits to the ex- centration.
perimental data. After these initial considerations we should like to turn
Supporting evidence for concentration-dependent energy our attention to metal oxide materials which are more sim-
parameters comes from the literature on heterogeneous catal- ilar to commercially applied materials. Such films exhibit a
ysis. In this field it is well known that heats of chemisorp- granular morphology and these are also likely to be doped
tion generally depend on the surface coverage Θ. Turning with catalytic noble metal materials such as Au and Pt. As
to the subject of chemisorption of CO and its catalytic con- an approximation to such materials we consider thin SnO2
version to CO2 via reaction with adsorbed oxygen species, films with a granular surface morphology, which had been
it is found that adsorption energies are strongly coverage- formed using the two-step approach also described in Sec-
dependent. Dulaurent et al. [48], for instance, find that the tion 2. In this approach metal evaporation is followed by
heat of adsorption of CO on a Ru/Al2 O3 catalyst decreases subsequent annealing in ambient air to form films consist-
from about 1.8 eV at Θ = 0 down to 1.2 eV at Θ = 1. In [49] the ing of nanometer-sized metal oxide grains. Into such mate-
authors find that there seems to be no competition between rials, it is easy to introduce noble metal impurities such as
oxygen and CO in finding adsorption sites on a Pt/SiO2 cata- Au and Pt. The data in Figs. 8–10 (Section 3) have already
lyst because the two species do not tend to adsorb at the same demonstrated that such noble metal doping can have a very
sites. Considering this strong coverage-dependence of Eads severe impact on the magnitude of the gas response and on
on catalytic surfaces, on the one hand, and the very moderate the temperature TM at which the sensitivity maximum SM
concentration dependence of Eads on SnO2 sensor surfaces, occurs.
on the other hand, it is suggested that the surface coverage CO Tables 3–5, in turn, show how such changes in the gas
on a SnO2 surface is already close to Θ = 1, even at low CO response are reflected in the fitted values of the activation
concentrations in the ambient air. In this way, the very moder- energy Ea and the adsorbate binding energy Eads .
ate decrease in Eads from 1.33 to 1.18 eV can be explained as A first inspection of the experimental data shows that in
the CO concentration ranges from 2 to 500 ppm. Compared the case of pure SnO2 the temperature TM is quite high for all
to this amount of variation, the concentration-dependence of three reducing gases investigated. From the literature one can
Ea is much smaller, amounting to 3–6 meV only. We there- see that such sensors generally exhibit high reaction tempera-
fore do not attempt to provide a physical explanation for this tures for reducing gases such as CO, H2 , hydrocarbons and in
latter effect. particular CH4 . Somewhat lower temperatures are observed

Table 3
Measured and fitted results for 1% H2
1% H2 Temperature of max- Maximum sens- Reaction activ- Adsorption energy Eads (eV)
imum sensitivity TM (◦ C) itivity SM at TM ation energy Ea (eV)
Pre tin oxide 400 15 0.44 1.07
(smooth)
Pre tin oxide 370 85 0.57 0.96
(granular)
Tin oxide:Pt 220 2100 0.31 0.66
Tin 370 43 0.42 0.89
oxide:Au
Catalyst doping influences the directly measurable parameters SM and TM . Changes in these parameters are reflected in the fitted energy parameters Ea and
Eads .
596 S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599

Table 4
Measured and fitted results for 0.5% ethene (C2 H4 )
5000 ppm Ethene Temperature of maximum Maximum sensitivity Reaction activation Adsorption energy
sensitivity TM (◦ C) SM at TM energy Ea (eV) Eads (eV)
Pure tin oxide 400 4 0.53 1.3
Pure tin oxide (granular) 490 14 0.78 1.4
Tin oxide:Pt 290 314 0.52 1.0
Tin oxide:Au 490 24 0.61 1.4
The influence of the catalysts on the ethene response resembles that on the H2 one.

Table 5
Measured and fitted results for 500 ppm CO
500 ppm CO Temperature of maximum Maximum sensitivity Reaction activation Adsorption energy
sensitivity TM (◦ C) SM at TM energy Ea (eV) Eads (eV)
Pure tin oxide 330 2 0.52 1.3
Pure tin oxide (granular) 410 0.8 1.45
Tin oxide:Pt 420 1 0.87 1.3
Tin oxide:Au 550 4 0.82 1.7
The influence of the catalyst doping on the CO response differs strongly from that on the C2 H4 and H2 one.

in the case of NO2 and O3 , which are less stable oxidising flected in concomitant changes in the two energy parame-
gases [50]. ters.
Considering undoped material first and comparing the data The case of CO is somewhat different. In this latter case,
obtained for the three reducing gases, it is seen that the sensi- the effect on the experimentally accessible quantities SM and
tivity profiles are quite similar, exhibiting different absolute TM is marginal (Pt doping) or even counter-productive in the
values of the gas sensitivity S only. This similarity is reflected sense that detection reactions are inhibited rather than be-
in the fitted energy parameters, which are similar in all three ing promoted (Au doping). As far as the consistency of our
gases. According to these, the gas with the highest response model is concerned, the changes in the measurable quanti-
value (H2 ) also exhibits the lowest activation barrier with ties SM and TM are also reflected in concomitant changes in
regard to detection. the fitted values of the energy parameters Ea and Eads . In the
Noble-metal-doped sensors, on the other hand, exhibit case of Au doping, in particular, the inhibition of detection
quite different behaviour. The data in Tables 3–5 clearly re- reactions is reflected in increased values of Ea and Eads . At
veal a similar impact on the two energy parameters both for this point, however, it should be noted that the two energy
H2 and for C2 H4 . Both catalysts vastly enhance the reaction values do not simply reflect material properties of Pt or Au,
with hydrogen, thereby reducing the maximum temperature respectively. Considering the fact that significant enhance-
TM and raising the maximum sensitivity SM . The ethene sen- ments in the CO sensitivity can be obtained by evaporation
sitivity is also promoted, although this effect is not as strong of dispersed clusters of Au onto granular SnO2 surfaces [54],
as in the case of H2 . it is demonstrated that the fitted values of Ea and Eads are
Pt is the noble metal catalyst most widely used in industrial more representative of the manner of introducing, dispersing
applications. Examples are automobile exhaust gas cleaning, and binding the catalytic impurities at the metal oxide sen-
ammonia oxidation and petrochemical applications like fab- sor surface than representing simple material properties of Pt
rication of plastics from natural oil [50,51]. On a Pt surface or Au.
H2 molecules tend to dissociate into single hydrogen atoms.
Such single atoms are more likely to react with other adsor-
bents, which reduces the activation energy Ea . Concomitantly
the adsorption energy Eads is also reduced. This latter reduc- 6. Conclusions
tion is very important for the functioning of a catalyst: in
order to be able to promote heterogeneous surface reactions, A new approach has been presented that is able to explain
reactant molecules must neither be bound too loosely, as this the gas sensitivity of thin film metal oxide materials towards
would not mediate any reactivity, nor too strongly, as this reducing gases.
would slow down surface diffusion and thus inhibit any fur- This approach satisfactorily explains the bell-shaped vari-
ther reaction. This is also known as the Sabatier principle ation of the gas sensitivity S with the sensor operation temper-
[52,53]. ature T. In particular, the derived function S(T, cgas ) explains
Although Au as a catalyst is not widely used in indus- the asymmetric deformation of the sensitivity profile with
try, it nevertheless acts as a reaction promoter in SnO2 . The respect to the sensitivity maximum SM occurring at the tem-
effects of Au are not as strong as those of Pt, but still TM perature TM as well as the sub-linear variation of S with the
for the H2 detection is lowered somewhat, which is also re- analyte gas concentration cgas .
S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599 597

The experimentally accessible parameters SM and TM , was that under equilibrium conditions, mobile donors should
which characterise the sensitivity distributions, are related give rise to a narrowing of the space charge zones. More quan-
to the microscopic energy parameters Eads and Ea . The first titatively these authors propose that the usual parabolic band
of these parameters describes the strength of the analyte gas bending profile should be replaced by a logarithmic one. In
binding on the metal oxide surface and the second the ki- this way, the distance to the surface, that needs to be traversed
netic barrier that needs to be overcome to induce a surface by a tunnelling electron, is reduced.
combustion event.
Fits to experimental gas sensitivity data for the gas species A.2. Fermi-energy dependence of the oxygen vacancy
CO, C2 H4 and H2 indicate that Eads and Ea tend to depend on concentration
the analyte gas concentration and thus on the surface coverage
of analyte molecules on the sensor surface. Additionally, the concentration of oxygen vacancies is un-
A comparison of undoped and noble-metal-doped SnO2 likely to be constant throughout the space charge zone. Start-
films illustrates that the addition of catalysts, in general, pro- ing out from the undepleted bulk material, an upward band
motes surface reactions by changes in both energy parame- bending causes the Fermi energy to retreat more and more
ters. from the conduction band edge as the free surface is ap-
proached. According to Hellmich [56] a change in the Fermi
energy relative to the band edges causes the vacancy concen-
Acknowledgement tration to increase. In [56] the reason for this increase has
been traced to the fact that part of the formation energy of an
Part of this work has been financed by the Bundesminis- oxygen vacancy–interstitial pair is regained by statistically
terium for Education and Research (BMBF) under the con- dropping the two valence electrons trapped at an oxygen va-
tracts 16SV1129/5 (MISSY) and 16SV 1532 (IESSICA). cancy onto the Fermi energy. As an upward band bending
lowers the Fermi energy with respect to the conduction band
edge, the density of oxygen vacancies should steeply increase
Appendix A. Efficiency of electron tunnelling as the free surface is approached. This effect in turn reduces
through metal oxide surface barriers the extent of the space charge region, so that in an equilibrated
state, it is likely to extend only a few nm into the metal oxide
The function for the temperature- and concentration- bulk.
dependent gas response of metal oxide materials S(T, cgas )
presented in Section 4, was derived by assuming tunnelling A.3. Tunnelling through dopant states
transparency through the adsorption-induced surface band
bending profile. In this Appendix A we should like to present Oxygen vacancies act as the main donors in tin oxide
more supporting evidence in favour of this assumption. Our and other pure metal oxides. These vacancies give rise to
arguments rely on three items, which are specific for metal localised levels below the conduction band edge of about 30
oxide semiconductors: and 150 meV [46]. The shallower one of these can be ex-
pected to be emptied at common operating temperatures of
A.1. Narrowed band bending due to oxygen vacancy more than 300 ◦ C (about 50 meV thermal energy). When the
mobility band bending near the surface rises above 0.15 eV the deeper
state will also become emptied, resulting in a situation as
The surface band bending in a semiconductor with bound sketched in Fig. 12 above.
surface charges can be determined by solving Poisson’s This latter figure displays a situation of a highly doped
equation. The standard solution to this problem results in metal oxide layer. In a typical metal oxide gas sensor, an
a parabolic band bending profile extending from the surface electron can travel from one state to another via thermally
into the undepleted bulk material. The assumptions made to activated hopping to arrive at the free surface [57]. Fig. 12
arrive at this standard solution are: zero mobile charge den- shows that much smaller amounts of thermal energy need
sity within the space charge region and a fixed concentration to be expended to arrive at the free surface by proceeding
of doping atoms within the semiconductor crystal. The first through a series of hops over localised bandgap states than
assumption is generally not true, especially in the case of by direct thermionic emission over the surface barrier. The
shallow band bending profiles. The second is usually correct, rate of charge transfer between bulk and surface will there-
but in the special case of metal oxides, where donor impu- fore depend much less on temperature in the case of hopping
rities are relatively mobile oxygen vacancies, the validity of as in direct thermionic emission over the barrier. In order to
this latter assumption is likely to break down. arrive at mathematically tractable equations we therefore de-
The subject of oxygen vacancy motion has been thor- cided to neglect the smaller temperature dependence of the
oughly investigated by Kamp [55]. Rantala et al. [47] have hopping transport with regard to the much larger impact that
investigated the influence of this effect on the shape of the reaction activation and adsorption energies have on the reac-
near-surface band bending profile. The conclusion reached tion kinetics and the sensitivity of a metal oxide gas sensor.
598 S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599

References [22] N.S. Baik, G. Sakai, N. Miura, N. Yamazoe, Hydrothermally treated


sol solution of tin oxide for thin-film gas sensor, Sens. Actuators B
[1] W. Göpel, K.D. Schierbaum, SnO2 sensors: current status and future 63 (2000) 74–79.
trends, Sens. Actuators B 26–27 (1995) 1–12. [23] J. Tamaki, Z. Zhang, K. Fujimori, M. Akiyama, T. Harada, N. Miura,
[2] D.E. Williams, Conduction and gas response of semiconductor gas N. Yamaze, Grain-size effects in tungsten oxide-based sensor for
sensors, in: Solid State Gas Sensors, Adam Hilger, Bristol, 1987, nitrogen oxides, J. Electrochem. Soc. 141 (8) (1994) 2207–2210.
pp. 71–123. [24] W.Y. Chung, G. Sakai, K. Shimanoe, N. Miura, D.-D. Lee, N. Yama-
[3] K. Ihokura, J. Watson, The Stannic Oxide Gas Sensor—Principles zoe, Preparation of indium oxide thin film by spin-coating method
and Applications, CRC Press, Boca Raton, USA, 1994. and its gas-sensing properties, Sens. Actuators B 46 (1998) 139–145.
[4] N. Barsan, M. Schweizer-Berberich, W. Göpel, Fundamental and [25] A. Diéguez, A. Romano-Rodriguez, J.R. Morante, Morphological
practical aspects in the design of nanoscaled SnO2 gas sen- analysis of nanocrystalline SnO2 for gas sensor applications, Sens.
sors: a status report, Fresenius J. Anal. Chem. 365 (1999) 287– Actuators B31 (1996) 1–8.
304. [26] C. Xu, J. Tamaki, N. Miura, N. Yamazoe, Grain size effects on
[5] S. Ahlers, Diploma Thesis, Technical University München, Germany, gas sensitivity of porous SnO2 -based elements, Sens. Actuators B 3
2000. (1991) 147–155.
[6] G. Sakai, N. Matsunaga, K. Shimanoe, N. Yamazoe, Diffu- [27] A. Gurlo, N. Bârsan, M. Ivanovskaya, U. Weimar, W. Göpel, In2 O3
sion equation-based analysis of thin film semiconductor gas and MoO3 –In2 O3 thin film semiconductor sensors: interaction with
sensor—sensitivity dependence on film thickness and operating tem- NO2 and O3 , Sens. Actuators B 47 (1998) 92–99.
perature, in: Proceedings of the Eurosensors XV, Munich, Germany, [28] M.K. Kennedy, F.E. Kruis, H. Fissan, Tailored nanoparticle films
2001. from monosized tin oxide nanocrystals: particle synthesis, film for-
[7] Th. Becker, S. Ahlers, C. Bosch-v.Braunmühl, G. Müller, O. mation, and size-dependent gas-sensing properties, J. Appl. Phys. 93
Kiesewetter, Gas sensing properties of thin- and thick-film tin-oxide (1) (2003) 551–560.
materials, Sens. Actuators B 77 (2001) 55–61. [29] P. Althainz, L. Schuy, J. Goschnick, H.J. Ache, The influence of
[8] G. Sakai, N.S. Baik, N. Miura, N. Yamazoe, Gas sensing proper- morphology on the response of iron-oxide gas sensors, Sens. Actu-
ties of tin oxide thin films fabricated from hydrothermally treated ators B 25 (1995) 448–450.
nanoparticles. Dependence of CO and H2 response on film thickness, [30] N. Bârsan, Conduction models in gas-sensing SnO2 layers: grain-
Sens. Actuators B 77 (2001) 116–121. size effects and ambient atmosphere influence, Sens. Actuators B 17
[9] F. Hossein-Babaei, M. Orvatinia, Analysis of thickness dependence (1994) 241–246.
of the sensitivity in thin film resistive gas sensors, Sens. Actuators [31] P. Romppainen, V. Lantto, The effect of microstructure on the height
B 89 (2003) 256–261. of potential energy barriers in porous tin dioxide gas sensors, J. Appl.
[10] P. Boeker, O. Wallenfang, G. Horner, Mechanistic model of diffu- Phys. 63 (10) (1988) 5159–5165.
sion and reaction in thin sensor layers—the DIRMAS model, Sens. [32] M.C. Horillo, J. Gutiérrez, L. Arés, J.I. Robla, I. Sayago, J. Getino,
Actuators B 83 (2002) 202–208. J.A. Agapito, Hall effect meaurements to calculate the conduction
[11] J.W. Gardner, A non-linear diffusion-reaction model of electrical control in semiconductor films of SnO2 , Sens. Actuators A 42 (1994)
conduction in semiconductor gas sensors, Sens. Actuators B 1 (1990) 619–621.
166–170. [33] N. Yamazoe, Y. Kurokawa, T. Seiyama, Effects of additives on semi-
[12] J.W. Gardner, A diffusion-reaction model of electrical conduction conductor gas sensors, Sens. Actuators 4 (1983) 283–289.
in tin oxide gas sensors, Semicond. Sci. Technol. 4 (1989) 345– [34] A. Cabot, J. Arbiol, J.R. Morante, U. Weimar, N. Bârsan, W. Göpel,
350. Analysis of the noble metal catalytic additives introduced by im-
[13] H. Windischmann, P. Mark, A model for the operation of a thin- pregnation of as obtained SnO2 sol-gel nanorystals for gas sensors,
film SnOx conductance-modulation carbon monoxide sensor, J. Elec- Sens. Actuators B 70 (2000) 87–100.
trochem. Soc.: Solid-State Sci. Technol. 126 (1979) 672. [35] A. Cabot, A. Vilà, J.R. Morante, Analysis of the catalytic activity
[14] N. Barsan, U. Weimar, Conduction model of metal oxide gas sensors, and electrical characteristics of different modified SnO2 layers for
J. Electroceramics 7 (2001) 143–167. gas sensors, Sens. Actuators B 84 (2002) 12–20.
[15] S. Ahlers, G. Müller, Factors influencing the sensitivity of metal ox- [36] S.J. Gentry, T.A. Jones, The role of catalysts in solid-state gas sen-
ide gas sensing layers, Encyclopedia of Sensors, American Scientific sors, Sens. Actuators 10 (1986) 141.
Publishers, 2005. [37] P.J. Shaver, Activated tungsten gas detectors, Appl. Phys. Lett. 11
[16] G. Sberveglieri, G. Faglia, S. Gropelli, P. Nelli, R.G.T.O: a new (1967) 255.
technique for preparing SnO2 sputtered thin films as Gas sensors, [38] J.C. Loh, French Patent 1545292 (1967).
Proc. Transducers (1991) 165–168. [39] G.S.V. Coles, G. Williams, Selectivity studies on tin oxide-based
[17] W. Hellmich, C. Bosch-v.Braunmühl, G. Müller, G. Sberveglieri, M. semiconductor gas sensors, Sens. Actuators B 3 (1991) 7–14.
Berti, C. Perego, The kinetics of formation of gas-sensitive RGTO- [40] J. Wöllenstein, H. Böttner, M. Jaegle, W.J. Becker, E. Wagner, Ma-
SnO2 films, Thin Solid Films 263 (1995) 231–237. terial Properties and the influence of metallic catalysts at the surface
[18] J.S. Suehle, R.E. Cavicchi, M. Gaitan, S. Semancik, Tin oxide gas of highly dense SnO2 films, Sens. Actuators B 70 (2000) 196–202.
sensor fabricated using CMOS micro-hotplates and in situ process- [41] S.C. Chang, Oxygen chemisorption on tin oxide: correlation between
ing, IEEE Electron Device Lett. 14 (1993) 118–120. electrical conductivity and EPR measurements, J. Vac. Sci. Technol.
[19] C. Cantalini, W. Wlodarski, Y. Li, M. Passacantando, S. Santucci, E. 17 (1) (1980) 366–369.
Comini, G. Faglia, G. Sberveglieri, Investigation on the O3 sensitivity [42] D. Kohl, Surface processes in detection of reducing gases with SnO2 -
properties of WO3 thin films prepared by sol-gel, thermal evaporation based devices, Sens. Actuators 18 (1989) 71–113.
and r.f. sputtering techniques, Sens. Actuators B 64 (2000) 182–188. [43] T. Doll (Ed.), Advanced Gas Sensing: The Electroadsorptive Effect
[20] H. Steffes, C. Imawan, F. Solzbacher, E. Obermeier, Fabrication pa- and Related Techniques, Kluwer, 2003.
rameters and NO2 sensitivity of reactively r.f. sputtered In2 O3 films, [44] T. Becker, S. Mühlberger, C. Bosch-v. Braunmühl, G. Müller, T. Zie-
Sens. Actuators B 68 (2000) 249–253. mann, K.V. Hechtenberg, Air pollution monitoring using tin-oxide-
[21] C. Nayral, E. Viala, V. Collière, P. Fau, F. Senocq, A. Maisonnat, B. based microreactor systems, Sens. Actuators B 69 (2000) 108–119.
Chaudret, Synthesis and use of a novel SnO2 nanomaterial for gas [45] M. Henzler, W. Göpel, Oberflächenphysik des Festkörpers, 2. Aufl.,
sensing, Appl. Surface Sci. 164 (2000) 219–226. Teubner, Stuttgart, 1994.
S. Ahlers et al. / Sensors and Actuators B 107 (2005) 587–599 599

[46] T. Rantala, V. Lantto, T. Rantala, Computational approaches to the [51] S.R. Morrison, Chemical Sensors, in: S.M. Sze (Ed.), Semiconductor
chemical sensitivity of semiconducting tin oxide, Sens. Actuators B Sensors, Wiley, New York, 1994.
47 (1998) 59–64. [52] P. Sabatier, Berichte derDeutschen Chem, Gesellschaft 44 (1911)
[47] T.S. Rantala, V. Lantto, T.T. Rantala, Effects of mobile donors on 1984.
potential distribution in grain contacts of sintered ceramic semicon- [53] R.I. Masel, Principles of Adsorption and Reaction on Solid Surfaces,
ductors, J. Appl. Phys. 79 (12) (1996) 9206–9212. second ed., Wiley, New York, 1996.
[48] O. Dulaurent, M. Nawdali, A. Bourane, D. Bianchi, Heat of adsorp- [54] P. Nelli, G. Faglia, G. Sberveglieri, E. Cereda, G. Gabetta, A.
tion of carbon monoxide on a Ru/Al2 O3 catalyst using chemisorption Dieguez, A. Romano-Rodriguez, J.R. Morante, The aging effect on
equilibrium conditions at high temperatures, Appl. Catalysis A: Gen. SnO2 –Au thin film sensors: electrical and structural characterization,
201 (2000) 271–279. Thin Solid Films 371 (2000) 249–253.
[49] A. Bourane, D. Bianchi, Oxidation of CO on A Pt/Al2 O3 catalyst: [55] B. Kamp, Beiträge zur Sensorik redox-aktiver Gase, Ph.D. Thesis,
from the surface elementary steps to light-off steps. I. Kinetic study University of Stuttgart (2002).
of the oxidation of the linear CO species, J. Catalysis 202 (2001) [56] W. Hellmich, Dissertation, Universität der Bundeswehr, Germany,
34–44. 1996.
[50] Hollemann, Wiberg, Lehrbuch der Anorganischen Chemie, de [57] N.F. Mott, Phil. Mag. B 19 (1969) 835.
Gruyter, 101 Auflage (1995).

You might also like