13
ELSEVIER Sensors and Actuators B 31 (1996) 71-75
The effect of Pt and Pd surface doping on the response of
nanocrystalline tin dioxide gas sensors to CO
M. Schweizer -Berber ich a,*, J .G. Zheng a, U. We imar a, W. G6pe l a, N. B~san b, E. Pent ia b, A. Tomescu b
alnstitute of Physical and Theoretical Chemistry, Center of Interface Analysis and Sensors, University of Tlibingen,
Auf der Morgenstelle 8, D-72076 Ttibingen, Germany
blnstitute of Physics and Technology of Materials, PO Box MG-7, Bucharest-Magurele, R-76900, Romania
Abstract
SnO 2 sensors were prepared by precipitating Sn(OH) 4 from an aqueous solution of SnCl4 or by evaporating SnO 2 in the UHV.
Subsequent X-ray and TEM measurements show grain sizes in the nanometer range after annealing at different temperatures or
annealing times. The effect of Pt and Pd doping upon the response to CO is studied. Parameters, which we focused on, are sensor signal
dependent on the operating temperature and dependent on the gas concentration. The results for ceramic sensors are compared with
those for thin film sensors.
Keywords: Tin dioxide; Gas sensor; Pt and Pd doping
1. Introduction
Solid-state gas sensors based on tin dioxide are the
predominant solid-state devices for gas alarms used in a
wide variety of domestic and industrial applications.
Their performance has been developed over the years,
mostly on an empirical basis. Their current status has
been recently reviewed in several text books [1-3]. In
contrast to their widespread application, the present un-
derstanding of the mechanism of gas detection is still im-
mature, although many papers have been published on
these fundamental aspects in recent years [4-6]. The main
reason for this unsatisfactory situation is the fact that the
structural characterization and in particular the influence
of nanosized particles on the overall conduction is not
understood well enough for real samples.
Nanocrystalline materials with controlled composition
are of increasing interest in gas sensing [7]. Compared
with the "classical" tin dioxide sensors, this new class of
device offers an increased sensitivity and selectivity. In
addition, it is a new and exciting subject of fundamental
research. This paper deals with the aspects of perform-
* Corresponding author.
0925-4005/96/$15.00 © 1996 Elsevier Science S.A. All fights reserved
SSDI 0925-4005(95)01775-4
ance improvements by using metallic additives. The ap-
proach is different from the heuristic one, normally used,
because our purpose was not to obtain the best sensors by
trying lots of additives, concentrations and calcination
temperatures, but to understand the mechanisms of per-
formance improvements. In this respect we focused not
only on the sensitivity of our samples, but also on the
changes in the detection mechanism.
2. Experimental
We used two types of transducers: Firstly the classical
Taguchi type was chosen, consisting of a ceramic tube
(alumina) with a platinum heater inside. The electrodes
were made from platinum or gold. Secondly a 100/~m
interdigitated platinum comb structure on sapphire was
chosen, containing a platinum heater meander and a
thermocouple (Pt/10%RhPt) on the back.
The individual sensors were mounted in a teflon test
chamber. The resistance was measured with a scanner
multimeter (Keithley DMM199, Munich). Gas concentra-
tions of carbon monoxide were adjusted by computer
controlled mass flow controllers (Tylan 260, Munich). All
sensor measurements were performed in humidified (50%
relative humidity) air.
72 M. Schweizer-Berberich et al./ Sensors and Actuators B 31 (1996) 71-75
40
30
E
20
c_
10
(
annealing at 700°C in synth, air
y=232 + 41 " In ( t + 0 .004)
(
, I , I I , I , I , I , I
0 20 40 60 80 100 120 140
time t in h
Fig. 1. Dependence of the grain size of thin film SnO 2 on the annealing
time at 700°C in air. 20 nm SnO 2 were vacuum deposited on NaC1
substrates, were annealed at 700°C and grain sizes were determined by
TEM.
the thin SnO 2 films (20 nm) start from nominally zero
(amorphous films) and increase upon annealing. A loga-
rithmic time dependence fits the experimental data best
(Fig. 1, T = 700°C). In future work it has to be checked
whether the crystallite growth stabilizes at a limited grain
size. As the sensor operation temperature is much lower
(maximum at 400°C), it is expected to adjust the desired
grain size at 700°C with sufficient stability. The chosen
steps for the preparation of thin and thick films ensures a
comparable grain size in both cases.
In the following figures typical sensor response curves
are shown. Fig. 2 presents the temperature dependence of
the conductance of thick film (pure, Pt-doped and Pd-
doped sensors) and thin film (Pt-doped) sensors in clean
air (50% relative humidity). In Fig. 3 the temperature
dependence of the sensor response (as defined by the ratio
of the conductance G in 100 ppm CO to the conductance
G O in pure 50% humidified air) is given. The dependence
of the conductance G on the CO concentration (in the
100-500 ppm range) may formally be described by
The nanocrystalline SnO 2 materials were prepared by
two different methods. Ceramic material was prepared by
a wet chemical method [8]. First Sn(OH)4 was precipi-
tated by adding ammonia to an aqueous SnCI4 solution.
Tin dioxide with different grain sizes was obtained by
controlled annealing (4-15 h) at elevated temperatures
(450, 800 and 1000°C). The SnO2 paste was deposited
onto the cylindrical substrates. The solvent (1,3-propane-
diol) was burnt out at 7000C for 5 min. This second tem-
perature treatment did not influence the grain sizes, as
revealed by TEM measurements. The surface doping was
performed by a powder impregnation using Pt and Pd
chlorides for 20 h at room temperature. The doped pow-
ders were annealed again at 4500C for 1-4 h in order to
convert the chloride to the metal (or the metal oxide). The
concentration of the dopants in the tin dioxide powder
was 0.2% wt. The thicknesses of the ceramic materials
were between 0.1 and 1 mm.
Thin film sensors were prepared by a PVD method.
50 nm amorphous tin dioxide was deposited onto the sap-
phire substrates by electron beam deposition from a SnO2
target in high vacuum (Edwards 306A). The grain size,
adjusted by a controlled annealing procedure, depends on
the annealing temperature and the annealing time; usually
700°C in synthetic air (20% 02/80% N2) was used. Pt
doping (approximately 1 and 2 monolayers) was per-
formed by a sputtering process (Balzer SCD 050).
3. Results and discussion
X-Ray diffraction (XRD) and transmission electron
microscopy (TEM) show grain sizes in the nanometer
range for both preparation procedures [9]. Grain sizes of
the ceramic tin dioxide powder are in the 10-30 nm range
after the annealing procedure at 450°C for 8 h. Those of
G = const, x P~o (1)
for all sensor types and at all heating voltages. The value
of the parameter n is related to the chemical surface reac-
tion between CO and chemisorbed oxygen [10]. Fig. 4
describes the correlation between the operation tempera-
ture and the value n. For sensors shown in this report n
varies between 0.002 and 0.9, with the lower values for
thin films and the higher values for thick film (ceramic)
sensors.
The experiments and the following discussion are fo-
cused on two aspects: (1) the effect of metallic additives
as surface dopants (Pt, Pd), on materials prepared by the
same thick film technology and (2) the influence of the
preparation technology for sensors (i.e. thick or thin
1000
03
=" 100 .E
C0
8
oo 10
r~~u ~
r " l [ ]
_ o ~ 2
o
O' A / ~
A~- - -A - - Pt doped
---O.-- Pt, thin film
I i i i , i i i i , I i i
250 300 350
temperature T in °C
Fig. 2. Dependence of the conductance on the operation temperature of
pure and doped tin dioxide sensors (ceramics calcinated at 450°C and
thin films at 700oC) monitored in humid air (50% relative humidity).
M. Schweizer-Berberich et al. / Sensors and Actuators B 31 (1996) 71-75 73
E
.E
E
40
30
20
10
ceramic
- - [ ] - - pure . ©
O Pt ,
- -A - - Pd
A ~. ,X
thin film
,, Ptl
\, - -O - - Pt 2
©
/O
~zx jzx
5
4
E
..=
3"6
E
2
150 200 250 300 350 400
temperature in °C
Fig. 3. Dependence of the sensor signal of pure and differently doped
sensors to 100 ppm CO on the operation temperature. The sensor signal
was defined as Ggas IGai r. Ceramic materials were calcinated at 450°C;
thin films at 700°C.
films) with comparable properties and surface dopant
(Pt).
In the first case we examined the effect of surface do-
pants on the three sensor parameters: conductance (resis-
tance) in clean (50% relative humidity) and CO contami-
nated (100 ppm) air, sensitivity and parameter n of the
power law. The way in which the surface dopants influ-
ence the gas response characteristics is still a subject of
debate. The two mechanisms usually considered are the
spill-over and the Fermi-level mechanism. In the spill-
over mechanism, the surface dopant contributes to the
loading of tin dioxide surface with reacting species such
as CO and O2 in our case. The gas detection takes place at
the tin dioxide surface, but the reaction rates are affected
by the surface dopant. In the Fermi-level mechanism, the
gas detection reaction takes place at the surface where the
dopants, which are metallic clusters, change the electro-
static potential due to their different electronic affinities.
The conductance of the tin dioxide sample varies accord-
ing to the variation of the pinning of the tin dioxide Fermi
level by the Fermi level of the surface dopant. As surface
dopant, Pd is considered typical for the Fermi level
mechanism [11], whereas Pt is considered typical for the
spill-over mechanism.
Concerning the gas interaction with CO more gener-
ally, the effects of the addition of Pd or Pt to surface of
the differently annealed sensors are as follows.
- A decrease of the conductance in clean air by one to
three orders of magnitude. The amount of decrease
either depends on the operating temperature. At
higher temperature (about 400°C) the decrease is
not as dramatic as at lower temperature (about
275 °C).
- A shift of the temperature corresponding to the
maximum of sensitivity to detect CO from 350°C
and higher for the pure undoped material to less
than 300°C. For comparable coverages Pd shifts the
maximum to lower temperatures (average around
240°C) than Pt (average around 250°C), but the
platinum doped material shows the higher sensitiv-
ity to detect CO.
- The shift of the temperature corresponding to the
maximum value of factor n from about 350°C for
the pure material to approximate 250°C for Pd and
280°C for Pt. The shape of the curve also depends
on the calcination temperature.
The main difference between thin and thick film sam-
pies is related to the active area; for thin films, which are
more compact (especially after the annealing process,
where the amorphous SnO2_, is oxidized), the active area
for gas detection is nearly the geometric surface area. For
thick films, which are more porous, the active area can be
orders of magnitudes higher than the geometric surface.
This is shown by both TEM pictures and specific surface
measurements, which indicates for our preparation condi-
tions values of 18 m 2 g-1 [2]. As a result, the impact of
surface phenomena on the total conductance is higher in
thick films if compared to dense thin films. The differ-
ence between thin and thick film samples (Pt doped) as
revealed by our measurements are striking: The maxi-
mum sensitivity is up to 25 times higher for thick films in
comparison with thin films. The maximum value of the
parameter n does not exceed 0.25 for thin films and it can
be close to 1 for thick films. The general shape of the
temperature dependence of sensitivity and factor n is
qualitatively the same for both preparation procedures.
The difference in sensitivity between the differently
doped thick film samples can be explained by the well
known catalytic activity of Pt in CO adsorption. We ex-
1.0
0.8
0.6
,9.0
t~ 0.4
0.2
O
[]
/i-7•
/
[3
[]
50nm thin film
Pt 1
Pt 2
© O
ceramic, Pt doped
- . [ ] Tcalc. = 450°C
O Tcai¢= 1000°C
I i
150
0.0 , ~ i ' eS . ,
200 250 300 350 400
temperature in °C
Fig. 4. Dependence of the parameter n of Pt-dopcd sensors on the op-
elation temperature. Ceramic sensors were calcinatcd at 450°C and
1000°C, thin film sensors with two thicknesses of Pt (approximately 1
and 2 monolayers) at 700°C. For the definition o fn sec Eq. (1).
74 M. Schweizer-Berberich et aL / Sensors and Actuators B 31 (1996) 71-75
30
== 20
fit o1 simulated data: G = a'e n simulated conductance Re Ra
R d / R a a n G a = const, G a = 1 + c os -----1 i----1 p - -
[ ] 10 1.47 0.07 G =k* (Ga*Gd)/(Ga+Gd)
0 1 2.01 0,29
A 0.1 1,40 0.44 . ~ / J ~ . ~ ' ~
0.01 1.29 0.46 ~ - -
~-/0 0 0 0 0 0 0 0 0
0
D-- ~ ~ ~ -~-E]--- B -D- - -On E}~ -IS]
0 i I ~ I i I ~ I = I
0 100 200 300 400 500
concentration c in ppm
Fig. 5. Simulation (symbols) and fit (line) of the sensor conductance as
a function of the concentration of a reducing gas. The simulation is for
a series of a contact ('dead') resistance (Rd) and a sensor resistance
(Rs).
pect, in the case of Pt doping, an increase in the efficiency
of chemisorbed oxygen reduction by CO and as a conse-
quence an increase in the sensitivity. This type of Pt sen-
sitization of tin dioxide based gas sensors is well docu-
mented. The surprising effect is the difference between
the values of parameter n for thick an thin film sensors.
This difference indicates that there is a change of the con-
duction mechanism associated with Pt doping. An expla-
nation could be the fact that the loading with chemisorbed
oxygen in pure air is so high (Fig. 2), that the crystallites
are completely depleted and a conduction mechanism in
which there is also a dependence of mobility on the sur-
face phenomena becomes possible [12]. Such a mecha-
nism is possible only if one accepts the spill-over mecha-
nism for the sensitization by Pt surface doping.
For Pd surface doping we cannot yet discriminate be-
tween the two mechanism. Both could explain the
changes in conductance and the temperature shift. In fu-
ture work, we will examine the dynamical characteristics
of the gas response of Pd doped samples in order to ob-
tain additional information.
The difference between thin and thick films can be
explained if one takes into account the role of contacts as
monitored by AC impedance spectroscopy [13]. The re-
sults indicate that the contact contribution plays an impor-
tant role for the total conductance for thin film samples;
the contact resistance between the electrodes and the tin
oxide (or between the separated grains) can vary over a
wide range of the total resistance of the thin film sensor.
If one of the resistances in series to the tin oxide resis-
tance is not sensitive, new effects arise: The effect of such
a 'dead' resistance in series with the gas sensitive resis-
tance on both sensitivity and parameter n (Fig. 5) is obvi-
ous. A qualitative model for the simulation of the conduc-
tance dependent on the concentration uses a series of
contact ('dead') resistance and sensor resistance. Eq. (1)
with a parameter n of 0.5 is used to calculate the conduc-
tance of the sensitive element. A scaling factor (k) adjusts
the calculated conductance to experimental data. A fit of
the simulated data with Eq. (1) (as is used normally for
experimental data) results in a factor n which is depend-
ent on the ratio of 'dead' and sensitive resistance. The
sensitive part becomes dominant if the 'dead' resistance is
only 10% of the active one. Otherwise it will be masked
by the 'dead' resistance.
Thick film sensors contain a large number of inter-
granular regions. These regions show surface band bend-
ings of the same order of magnitude as at the electrode-
tin dioxide interface. Consequently the contact resistance
has no practical importance. In Fig. 6 there is a plot of the
impedance of such a thick film sample as a function of
frequency. The equivalent circuit which fits such a plot is
a simple series combination of the resistance of the tin
oxide layer with the contact part in parallel to the capaci-
tance of the substrate. The conclusion is that the contri-
bution of the contact resistance to the overall resistance is
below 10% in this case.
2
N,
o measurement lOOkHz c
simulation
Cg: 0.979pF
- :- - -y _ Ra: 18820R
- cR--c-; - t t - Re: 1240R
Cc: 0.13 nF
lOOHz
16 18 20
Zrea= in kf~
Pig. 6. Impedance plot of a thick film sensor (annealed at 800*(2, nn-
doped, with Pt-electrod~). The capacity of the clean substrat= cone-
sponds to C s of the model. R c is found to be less than lO~ ofR I.
4. Conc lus ion
Differently doped SnO 2 sensors prepared in thick film
and thin film technology were examined. The results
show that both the sensitivities and parameter n of a
power law of thin and thick film sensors vary. Thick film
sensors (especially the doped ones) show higher sensi-
tivities to detect CO due to the more porous structure of
the material and the parameter n (of a power law fit)
reaches higher values.
If the contact resistance is high compared with the
sensitive resistance the factor n can be masked.
The factor n can also vary due to another reaction
mechanism because of changes in the operating tempera-
ture. Pt and Pd doping results in a shift of the maximum
M. Schweizer-Berberich et al. / Sensors and Actuators B 31 (1996) 71-75 75
of sensitivity and factor n to lower operating tempera-
tures. The sensitivity itself is enhanced (dramatically for
ceramic sensors).
Acknowledgements
This work was supported by the BMBF project
'Intelligente Gas Sensoren' (IGS, contract no. 13MV-
0297). One of the authors (N.B.) is grateful for support
from the DAAD.
References
[1] W. G6pel, J. Hesse and J.N. Zemel (eds.), Sensors - A Compre-
hensive Survey, Vol. 2: Chemical and Biochemical Sensors (Parts
1 and 2), VCH Weinheim, New York, 1991.
[2] K. Ihokura and J. Watson, The Stannie Oxide Gas Sensor - Prin-
ciples and Applications, CRC Press, Boca Raton, FL, 1994.
[3] G. Sberveglieri (ed.), Gas Sensors - Principles, Operation and
Developments, Kluwer, Dordrecht, 1992.
[4] D. Kohl, Surface processes in the detection of reducing gases with
SnO2-based devices, Sensors and Actuators B, 18 (1989) 71-114.
[5] K.D. Schierbaum, U. Weimar, R. Kowalkowski and W. G6pel,
Conductivity, work function and catalytic activity of SnO2-based
sensors, Sensors and Actuators B, 3 (1991) 205-214.
[6] N. B~irsan and R. Ionescu, The mechanism of interaction between
CO and SnO 2 surface - the role of water vapour, Sensors and
Actuators B, 12 (1993) 71-75.
[7] W. Gfpel and K.D. Schierbaum, SnO 2 sensors: current status and
future prospects, Sensors and Actuators B, 26 (1995) 1.
[8] N. B~u'san, R. Ionescu and A. Vancu, Calibration curve for SnO 2-
based gas sensors, Sensors and Actuators B, 18-19 (1994) 466-
469.
[9] A. Dieguez, A. Romano-Rodriguez, J.R. Morante, U. Weimar, M.
Schweizer-Berberich and W. G6pei, Morphological analysis of
nanocrystalline SnO 2 for gas sensor applications, Sensors and
Actuators, B31 (1996) 1-8.
[10] S.R. Morrison, The Chemical Physics of Surfaces, Plenum Press,
New York, 1977.
[11] N. Yamazoe, New approaches for improving the semiconductor
gas sensors, Sensors and Actuators B, 5 (1991) 7-19.
[12] N. B~rsan, Conduction models in gas-sensing SnO 2 layers