the role of morphology and crystallographic structure of metal oxides

Bạn đang xem bản rút gọn của tài liệu. Xem và tải ngay bản đầy đủ của tài liệu tại đây (4.69 MB, 39 trang )

The role of morphology and crystallographic structure of metal oxides
in response of conductometric-type gas sensors
G. Korotcenkov
a,b,
*
a
Korea Institute of Energy Research, Daejeon, Republic of Korea
b
Technical University of Moldova, Chisinau, Republic of Moldova
Available online 18 March 2008
Abstract
This review paper discusses the inﬂuence of morphology and crystallographic structure on gas-sensing characteristics of metal oxide
conductometric-type sensors. The effects of parameters such as ﬁlm thickness, grain size, agglomeration, porosity, faceting, grain network, surface
geometry, and ﬁlm texture on the main analytical characteristics (absolute magnitude and selectivity of sensor response (S), response time (t
res
),
recovery time (t
rec
), and temporal stability) of the gas sensor have been analyzed. A comparison of standard polycrystalline sensors and sensors
based on one-dimension structures was conducted. It was concluded that the structural parameters of metal oxides are important factors for
controlling response parameters of resistive type gas sensors. For example, it was shown that the decrease of thickness, grain size and degree of
texture is the best way to decrease time constants of metal oxide sensors. However, it was concluded that there is not universal decision for
simultaneous optimization all gas-sensing characteristics. We have to search for a compromise between various engineering approaches because
adjusting one design feature may improve one performance metric but considerably degrade another.
# 2008 Elsevier B.V. All rights reserved.
Keywords: Metal oxides; Polycrystalline; One-dimensional; Gas sensor; Sensor response; Morphology and crystallographic structure inﬂuence
Contents
1. Introduction . 2
2. Structural parameters of metal oxides controlling gas-sensing characteristics . . 3
2.1. The role of sensor geometry and contacts . 3
2.2. The role of dimension factors in gas-sensing effects . . 7

2.2.1. The inﬂuence of thickness . . 7
2.2.2. Grain size inﬂuence . . 11
2.3. The role of crystallographic structure of metal oxides. 16
2.3.1. Crystal shape . . . 16
2.3.2. Surface geometry 20
2.3.3. Film texturing . . 22
2.3.4. Surface stoichiometry (disordering) . . . 23
2.4. The role of morphology and porosity of metal oxides. 24
2.4.1. Grain networks, porosity, and the area of inter-grain contacts . . 24
2.4.2. Agglomeration . . 28
2.5. Peculiarities of one-dimensional structure characterization. . . 31
3. Concluding remarks 31
Acknowledgements 35
References . . 35
www.elsevier.com/locate/mser
A
vailable online at www.sciencedirect.com
Materials Science and Engineering R 61 (2008) 1–39
* Correspondence address: Korea Institute of Energy Research, Daejeon, Republic of Korea.
E-mail address:
0927-796X/$ – see front matter # 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.mser.2008.02.001
1. Introduction
Conductometric (resistive) metal oxide sensors comprise a
signiﬁcant part of the gas sensor component market. While
many different approaches to gas detection are available [1–
23], metal oxide sensors remain a widely used choice for a
range of gas species [1–5,15,24–34]. These devices offer low
cost, high sensitivity, fast response and relative simplicity,
advantages that should work in their favor as new applications

emerge, especially in the ﬁeld of portable devices. The working
principle of a typical resistive metal oxide gas sensor is based
on a shift of the state of equilibrium of the surface oxygen
reaction due to the presence of the target analyte. The resulting
change in concentration of chemisorbed oxygen is recorded as a
change in resistance of gas-sensing material. As an example,
reducing gases (CO, H
2
,CH
4
, etc.) lead to an increase of the
conductivity for n-type semiconductors and a decrease for p-
type material, respectively, whereas the effect of oxidizing
gases (O
3
, etc.) is vice versa. The sensor response (sensitivity)
of such devices, that is, the ability of a sensor to detect a given
concentration of a test gas (analyte), is usually estimated as the
ratio of the metal oxide electrical resistance (conductivity)
(S = R
gas
/R
air
,orR
air
/R
gas
) measured in air and in an atmosphere
containing the target gas. The rate of sensor response is
described in such parameters as the response or recovery times,

which characterize the time taken for the sensor output to reach
90% of its saturation value after applying or switching off the
respective gas in a step function.
Numerous materials have been reported to be usable for
metal oxide sensors design including both single and multi-
component oxides [15,31–33]. At that it has been established
that materials in different structural states can be used in those
resistive type gas sensors. These states include amorphous-like
state, glass-state, nanocrystalline state, polycrystalline state,
and single crystalline state. Each state has its own unique
properties and characteristics that can affect sensor perfor-
mance. However, in practice, nanocrystalline and polycrystal-
line materials have found the greatest application in solid-state
gas sensors [4,15,24,27,32,35–37]. Nanocrystalline and poly-
crystalline materials have the optimal combination of critical
properties for sensor applications including high surface area
due to small crystallite size, cheap design technology, and
stability of both structural and electro-physical properties.
Typically amorphous-like and glassy materials are not stable
enough for gas-sensing applications, especially at high
temperature [32,38]. Single crystalline and epitaxial materials
have maximum stability and therefore the use of materials in
these sta tes for gas sensors may improve the temporal stability
of the sensor. Unlike polycrystalline material, devices based on
epitaxial and single crystalline materials will not be plagued
with the problem of instability of grain size. However, the high
cost and technological challenges associated with their
deposition limit their general use in gas sensors.
One-dimensional structures, which are single crystalline
materials, can be synthesized using inexpensive, simple

technology [24,39,40]. Wide use of one-dimensional structures
is however impeded by the great difﬁculties required for their
separation and manipulation [41,42]. During the synthesis
process of one-dimension structures one may observe a
considerable diversity in their geometric parameters. In
polycrystalline and even nanocrystalline material we work
with averaged grain size, while using one-dimension structures,
each sensor is characteristic by the speciﬁc geometry of the
one-dimensional crystal. Therefore, reproducibility of perfor-
mance parameters for sensors based on one-dimension
structures would depend on the uniformity of those structures.
Unfortunately, the problem of separation, sizing, and manip-
ulations of one-dimensional structures is not resolved yet. To
achieve uniform sizing and orientation, new advanced
technologies will need to be implemented, and these would
be expensive and not accessible for wide use. There are a few
interesting proposals for controlling one-dimensional structures
[43], but they require further improvement for practical
implementation. Thus, gas sensors based on individual one-
dimension structures are not yet readily available commer-
cially. Further, the manufacturing cost of sensors based on one-
dimensional structures would far exceed that of polycrystalline
devices. Based on what was said above, it becomes clear that in
near future, polycrystalline materials would remain the
dominant platform for solid-state gas sensors.
Nano- and polycrystalline materials are very complicated
objects for study, because the electro-conductivity of those
materials depends on great number of factors [34,44–55].
Therefore, to specify optimal technologies for gas sensor
manufacturing on the basis of such mater ials, it is necessary to

expand our understanding of gas sensor mechanism in nano-
and polycrystalline oxides. For example, it is necessary to
establish the role of morphology and crystallographic structure
in gas-sensing effects, because there is a lack of real, detailed,
and integrated research establishing a connection between
structural parameters of oxides and parameters of sensor
response. One cannot ﬁnd a large number of good works in this
ﬁeld. There are the works of Yamazoe and coworkers in the
ﬁeld of ceramic type sensors [26,44,56–62], in which a direct
correlation between grain size of metal oxides and gas
sensitivity of conductometric sensors was established; Ega-
shira’s group [63–67] conducted a qualitative study of material
porosity inﬂuence on sensor response; Morante’s group
established a correlation between structural and gas-sensing
properties of metal oxides [68–73], and papers of Korotcen-
kov’s group conducted research in the ﬁeld of thin ﬁlm gas
sensors [34,74–84]. Korotcenkov’s works emphasized the need
for a broader approach for structural engineering of metal oxide
ﬁlms for solid-state gas sensors.
While a lot of reviews and book chapters describe the
working principle of metal oxide gas sensors in detail [1–
5,15,24–30,37,50], the aim of this review is to summarize the
results highlighting the correlation between material structure
and gas-sensing properties, and formulating some general
conclusions typical for metal oxides. Earlier assessments of
modeling morphological effects were made in Refs.
[34,47,79,85]. The results used in the present review were
obtained mainly with SnO
2
and In

2
O
3
-based gas sensors. These
materials are the most studied metal oxides for gas sensor
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–392
applications [24,27,32,33,36,55,79–81,86–89] as well as the
most commercially available. For example tin oxide is indeed
the most popular material for gas-sensing due to its relatively
low cost, its high sensitivity, and stability in different
environments .
The main focus in this review is on the analysis of undoped
material. Consideration of electro-physical and catalytic proper-
ties of a device with a second component would make the
analysis too complicated. The introduction of the second
component changes both the catalytic activity of base material
and the chemical composition. New compounds or solid
solutions with speciﬁc properties signiﬁcantly different from
the undoped material can be formed during metal oxide doping.
Additives could also inﬂuence grain size, the shape of
crystallites, bulk and surface stoichiometry, properties of
intercrystalline barriers, and bulk electro-physical properties
[34,74,87,90]. Additional possible effects of metal oxide doping
includeformation of p–njunctions, the appearance of transitional
areas and layers acting as catalytic ﬁlters, the changes in the
valency of metal state, and others [91–95]. The analysis of those
interrelated processes requires individual consideration. Some
important conclusions regarding the inﬂuence of the second
phase on structural, electro-physical and gas-sensing properties
of metal oxides can be found in Refs. [25,33,34,57,71,74,93,95–

99,]. More detailed information about the effect of additives in
metal oxide sensors can be obtained also from earlier reviews
[24,25,35,46,49,50,52,57,101].
2. Structural parameters of metal oxides controlling
gas-sensing characteristics
As it was indicated earlier, the fundamentals of resistive type
sensor operation are based on the changes in resistance (or
conductance) of the gas-sensing material as induced by the
surrounding gas. The changes are caused by various processes,
which can take place both at the surface and in the bulk of gas-
sensing material [24,34,35,48,51,52,100–105]. Possible pro-
cesses, which can control gas-sensing properties, are presented
in Fig. 1.
The possible consequences of these processes for surface
and electro-physical properties of metal oxides are shown in
Fig. 2.
Research has conﬁrmed that all processes indicated in Fig. 1,
including adsorption/desorption, catalysis, reduction/reoxida-
tion, and diffusion are relevant in gas sensors and inﬂuenced by
structural parameters of the sensor material. This afﬁrms that
gas-sensing effects are structurally sensitive as well. Taking
into account the complexity of the gas-sensing mechanism and
its dependence on numerous factors, it becomes clear that we
have to consider the inﬂue nce of a great number of various
structural parameters of metal oxide matrix on gas sensors’
parameters (see Fig. 3).
It has been shown in Refs. [25,46,47,50,85] that the
inﬂuence of the above-mentioned parameters on gas-sensing
characteristics takes place through the changes in the effective
area of inter-grain and inter-agglomerate contacts, energetic

parameters of adsorption/desorption, number of surface sites,
concentration of charge carriers, initial surface band bending,
coordination number of metal atoms on the surface, etc.
2.1. The role of sensor geometry and contacts
Fig. 4 shows some reported gas sensor electrode geometries.
To make measurements on a semiconductor gas-sensing
material it is possible to use a compressed pellet (see
Fig. 4a), which may or may not be sintered, with metal
electrodes on each face. This construction was used in Refs.
[106,107] to obtain fundamental information on the tempera-
ture dependence of conductance of tin dioxide. In a study of the
competition between water and oxygen adsorption in tin
dioxide [108], the electrode assembly consisted of two
concentric tantalum cylinders with powdered tin dioxide
Fig. 1. Diagram illustrating the processes, controlling the rate of sensor response.
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 3
between them (Fig. 4b). However, real devices usually have the
sensing material presented as a thin (e.g. sputtered, vacuum-
evaporated, or deposited as a result of chemical reactions) or
thick (e.g. screen-printed) ﬁlm on a substrate (Fig. 4 c–f) [109–
114]. Both electrodes can be fabricated together on the
substrate before (Fig. 4g) or after (Fig. 4h) the sensing ﬁlm is
deposited. This provides great ﬂexibility in the fabrication
process, as it need not be compatible with the sensing material.
At ﬁrst approximation the sensor geometric parameters of
length (L) and width (W) do not inﬂuence sensor response. The
L/W ratio inﬂuences only sheet conductivity of the gas sensor
(GS). As a rule, one needs to use inter-digital geometry of
contacts (Fig. 4e) with small distance between contacts (L)in
order to get small sheet conductivity. Appropriate adjustment of

these design parameters can achieve acceptable value of gas
sensor resistance suitable for further electronic processing.
However, in reality the situation might be signiﬁcantly
different. First of all, the purely geometric effect arises because
the ﬁlm conductance does not change instantly or uniformly
when the gas ambient changes: the gas must diffuse through the
ﬁlm, reacting with the particle surfaces as it does so. This leads
to variations in local ﬁlm conductance. A numerical simulation
indicated, for example, that where a sensor is highly sensitive to
the test gas, the sensitivity increased with electrode spacing
when the electrodes were underneath the ﬁlm, but decreased
with spacing when the electrodes were deposited on top of the
ﬁlm [111]. If electrode spacing was decreased to less than the
ﬁlm thickness, it was possible to detect a les s-reactive gas in the
presence of a more reactive one [111]. The possibility of
exploiting these effects to produce self-diagnostic sensors has
been considered in Ref. [110]: If two or more pairs of contacts
with different separations are made on the sensor, then the
Fig. 2. Diagram illustrating processes taking place in metal oxides during gas detection and their consequences for polycrystalline metal oxides properties.
(Reprinted with permission from Ref. [105]. Copyright 2007: Elsevier).
Fig. 3. Diagram showing structural parameters of metal oxides, which control gas-sensing properties.
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–394
conductance measured between any two pairs under a given set
of conditions will be related by a known function, even though
the individual values will, of course, change with test gas
concentration. Thus, if the relationship observed deviates from
this function, the sensor must be malfunctioning.
For thin ﬁlms discussed above-mentioned effect does not
work. However, even in this case the inter electrode distance
may be a strong inﬂuencing factor. For example, in Ref. [109] it

was shown that the decrease of the distance betwee n inter-
digitated electrodes from 400 mm to 200 mm may enhance the
CO response in ceramic type SnO
2
-based sensors. Even greater
differences in sensor parameters could appear when the
distance between measurement electrodes in a sensor becomes
less than some critical value. Such inﬂuences could be
connected with the following factors:
Electrode materials used (Pt, Pd, Au) are active catalysts
with speciﬁc catalytic properties. As a result, in the area close to
contact (spillover zones), electrode mater ials act as catalysts
able to increase activity of gas-sensing metal oxides
[97,98,115,116] (see Fig. 5). Spillover is a very important
term in catalysis [117]. It is used as a shorthand description of
the diffusion of adsorbed species from an active adsorbent to an
otherwise inactive support. For instance, this could be the
diffusion of atoms from active metal nanoparticles, where the
dissociation is non-activated, to a support, where the
dissociation directly from the gas phase is activated.
Experimentally, this process was determined for hydrogen
and oxygen [117]. Therefore, if the distance between contacts is
comparable with the width of spillover zone (see Fig. 6b), the
inﬂuence of geometric parameters of sensors on their gas-
sensing characteristics would become noticeable. The width of
spillover zone depends on the material and the nature of the
detected gas.
The inﬂuence of the contacts on sensor response with
decreased length of the sensitive layer could become stronger
because of another reason as well. At some distance the

contact’s resistance could be comparable in magnitude or more
than the resistance of the gas sensitive layer, especially in the
atmosphere of reducing gases (for n-type semiconduct ors). In
some cases, the potential barrier b etween the metal of the
electrode and the gas-sensing oxide could be comparable to the
potential barriers between the metal oxide grains. Under these
circumstances, the chemical reactions between gas and metal–
metal oxide interface could affect the total conductance of the
sensor, even without the inﬂuence of the spillover effect [119].
Experimental data conﬁrm both these effects [24,109,119–
123]. For example, Laluze et al. [120] found very large
differences in the operation of sintered SnO
2
sensors fabricated
using different electrodes. Fig. 7 illus trates how strong this
inﬂuence could be. One can see that the change of electrode
Fig. 4. Possible constructions of solid-state metal oxide sensors and topologies of measurement contacts. (Adapted with permission from Ref. [112]. Copyright 2005:
Elsevier).
Fig. 5. Schematic illustration of spillover effect at the SnO
2
surface at
T
oper
< 180–210 8C. (Adapted with permission from Ref. [118]. Copyright
2003: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 5
metal affects both the magnitude of the sensor response and the
temperature position of the maximum sensitivity.
Other studies also established that different electrode
materials can affect sensor behavior [72,109,119,125].For

example, in Ref. [125] the characteristics of SnO
2
based sensors
with Pt, Au, and Pt–Au contacts were compared. It was shown
that at approximately 550 8C, the conductance was about the
same and independent on the electrode material. However,
below 150 8C the conductance of the sensors with a Pt electrode
was about three orders of magnitude higher than for those with
Au electrodes. In Ref. [119] it was reported that for SnO
2
thick
ﬁlm sensors the conductance changes induced by H
2
and CO
were very different for Pt and Au electrodes. The sensor with Pt
electrodes was more sensitive to H
2
, whereas Au electrodes
seemed to provide a better response to CO. In these
experiments, an inter-digital electrode design with a 5 mm
gap was used. The same effect was observed in Refs. [124,126].
In Ref. [126] it was shown that a chlorine detector, made from
WO
3
and aluminum electrodes, had sensor response of about
400 for 1 ppm Cl
2
in air. The sensor response dropped to 1 with
Pt electrodes.
Response to humidity was also affected by the electrode

material. In Refs. [109,125], it was established that the
inﬂuence of water on the CO response of SnO
2
-based sensors is
greater in the case of Au contacts, and lower in the case of Pt
contacts.
That the metal–semiconductor junction may be the main
gas-sensing element responsible for the observed sensor
response was conﬁrmed in Ref. [127]. In this study, the effect
of gap size on the sensor response to dilute NO
2
was
investigated (see Fig. 8). Gap sizes in WO
3
microsensors were
varied from 0.1 to 1.5 mm. It was found that the response to
dilute NO
2
was unchanged for gap sizes larger than 0.8 mm,
whereas below 0.8 mm the sensor response tended to increase
with decreasing gap size. The sensitivity to 0.5 ppm NO
2
was as
high as 57 at a gap size of 0.11 mm. For an explanation of the
observed effect it was assumed that the contribution of
Fig. 6. Diagram illustrating the role of spillover zones in thin ﬁlm gas sensors.
Fig. 7. Inﬂuence of electrode material on gas-sensing characteristics of SnO
2
sensors, fabricated on the base of thin ﬁlms deposited by electrostatic spray
pyrolysis. (Adapted with permission from Ref. [124]. Copyright 1999: Else-

vier).
Fig. 8. Sensitivities to dilute NO
2
of WO
3
microsensors as a function of gap
size. WO
3
microsensors with micro-gap electrodes were fabricated by means of
MEMS techniques (photolithography and FIB) and suspension dropping
method. WO
3
powders were prepared by wet process. Powders were calcined
at 400 8C for 3 h. (Adapted with permission from Ref. [127]. Copyright 2005:
Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–396
resistance at the electrode–grain interface to the total sensor
resistance becomes larger when the gap size is decreased. It was
concluded that the resistance change at the electrode–grain
interface is much larger than that at the inter-grain boundary
when the microsensor is exposed to NO
2
. Thus, the sensitivity
is increased with the decreasing gap size.
It means that at small distances between contacts, the role of
contact material is essential and should be considered in the
design of gas sensors. Moreover, it is possible to control the
sensing properties of semiconductor gas sensors simply by
using different electrode materials. Optimum electrode
material and electrode geometry could also be used to enhance

the gas-sensing properties [25,111,119].
Using the electrodegeometry as a design parameter, onecould
probe the variation of sensor signal with electrode position within
the porous sensor body (see Fig. 4g and h). If the electrodes are
closely spaced, the current for conﬁguration shown in Fig. 4g
probes only the base of the sensor layer, while the current probes
the whole sensor layer for electrodes that are spaced sufﬁciently
widely. For conﬁguration shown in Fig. 4h we have other
situation. The current can be pushed out into the layer by using
narrow inter-digitated electrodes, and pulled down towards the
base of the layer by using wide electrodes. Such a measurement
could, for example, lead to a determination of the rate constant
for the surface-catalyzed decomposition, which should be a
characteristic parameter of the gas, the surface composition, and
the temperature. To some degree, such measurements can beused
to identify the gas [25]. Hoefer et al. [128] used an array of
electrodes of differing width and separation to examine contact
resistance effects in tin dioxide sensors. In its original form, they
used the transmission line method for measuring the total
resistance of a semiconductor sample as a function of electrode
separation. The linear relation obtained allowed a determination
of sheet resistance and contact resistance, while an additional
‘‘end resistance’’ measurement allowed an estimation of a further
parameter: The ‘‘modiﬁed sheet resistance’’ of the ﬁlm in the
vicinity of the electrode [129]. It was shown that the modiﬁed
sheet resistance displayed greater sensitivity to CO and NO
2
than
either the sheet resistance itself or the contact resistance [128].In
this case, wide electrodes with narrow spacing would produce the

most sensitive detection. An array of electrodes varying in width
and spacing (see Fig. 4f), but all using the same sensing material,
could be used to resolve a mixture of CO, CH
4
,NO
2
and water
vapor into separate measurements of each component by ﬁrst
determining the relative sensitivity of the total resistance of each
electrode pair to the individual gases [128]. Further, simulations
have shown that a poorly reactive gas can be detected in the
presence of a highly reactive gas if electrode placement and ﬁlm
thickness are chosen well [111]. In Ref. [107] it was found out
that the lower detection limit can be improved by reducing the
number of grains between the electrodes.
The electrode material also affects the stability of the gas
sensor. It was shown in Refs. [109,130] that Au electrodes are
less stable as compared to Pt electrodes. Scanning electron
microscopy (SEM) and resistance measurements, carrie d out in
Ref. [130], have shown that platinum on an adhesion layer of
titanium was stable up to 500 8C, while the changes in gold
ﬁlms with various adhesion layers were observed at noticeably
lower temperatures. For example, gold on chromium starts
degrading at as low a temperature as 250 8C. An increased
diffusion coefﬁcient and an inclination to form alloys are
probably the reasons for such behavior of Au electrodes. If
aluminum was to be used as an interconnect metallization, it
was found that the maximum stability had contacts with an
additional layer of platinum as the metal for making contact
with the sensor material (metal oxide) and a barrier layer of

titanium–tungsten between the aluminum and platinum. This
combination was also usable up to 500 8C. Other layer
structures show less thermal stability.
Other important aspect of length’s inﬂuence appears when
the distance between electrodes becomes less than the
crystallite’s size (see Fig. 9c and d). In this case we could
observe a situation, when the intercrystallite barriers stop
affecting the gas-sensing effects, which could induce sig-
niﬁcant changes in sensor performance parameters or even loss
of sensitivity. This implies that as the distance is decreased, the
gas sensor mechanism could change. At sma ll distances only
bulk grain effects would be present.
It seems that a realization of this condition is impossible in
near future for ﬁnely dispersed metal oxides. This principle can
be realized only with single crystals, epitaxial ﬁlms, and one-
dimensional structures, where the grains and inter-grains
boundaries do not exist. However, successful development of
new advanced technologies [131] may make it feasible to
produce sensors based on one individual grain. As it was shown
in Ref. [112], state-of-th e-art electron-beam techniques can
produce extremely narrow and closely spaced metallized lines
with features of less than 10 nm in size.
2.2. The role of dimension factors in gas-sensing effects
2.2.1. The inﬂuence of thickness
At present there are three main gas sensor design
approaches: (1) – ceramics; (2) – thick ﬁlm; and (3) – thin
ﬁlm [35]. Therefore, while analyzing the inﬂuence of thickness
on sensor parameters, it is necessary to remember that for
ceramics and thick ﬁlm sensors the grain size does not depend
on the thickness; rather it is determined by the conditions of

synthesis and the thermal treatment parameters. The situation
for thin ﬁlm senso rs is fundamentally different. Th e grain size is
determined directly by the thickness of the deposited ﬁlm. The
strength of that inﬂuence is shown in Fig. 10. The main
regularities of ﬁlm thickness inﬂuence on structural properties
of SnO
2
and In
2
O
3
deposited by spray pyrolysis were discussed
in Refs. [77–79,132].
The inﬂuence of ﬁlm thickness (d) on sensor response to
ozone and reducing gases for In
2
O
3
-based sensors fabricated
using thin ﬁlm technology is shown in Figs. 11–13.In
2
O
3
ﬁlms
in these experiments were deposited by spray pyrolysis. It is
seen that the change of ﬁlm thickness can lead to a change in
both the magnitude and temperature position of the sensor
response’s maximum. At that, the effect of thickness on gas-
sensing characteristics was most pronounced for oxidizing
gases. When the In

2
O
3
ﬁlm thickness increases from 20 to
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 7
Fig. 10. The inﬂuence of the thickness of In
2
O
3
ﬁlm deposited by spray
pyrolysis on the grain sizes measured by (1) XRD; (2) AFM; and (3) TEM
methods. (Reprinted with permission from Ref. [79]. Copyright 2004: Elsevier).
Fig. 11. Thickness inﬂuence on sensor response to (1 and 2) ozone and (3) H
2
of
In
2
O
3
thin ﬁlms deposited by (1 and 3) spray pyrolysis and (2) sputtering.
Results (2) were obtained immediately after In
2
O
3
photoreduction. For this
purpose the samples were directly irradiated in vacuum by mercury pencil lamp
for 20 min. (Adapted with permission from Refs. [79,81,133]. Copyright 2001
and 2004: Elsevier).
Fig. 9. Diagram illustrating the inﬂuence of grain size on potential distribution along sensor.
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–398

400 nm, the gas response to ozone drops by more than a factor
of 100 (Fig. 11). This drop in sensitivity can be rationalized by
an increase in grain size [77,79] and a decrease of gas-
permeability within the ﬁlm. Due to high activity, ozone
decomposition occurs on the top layer of the metal oxide ﬁlm.
Thus, thin ﬁlms designs should be used for effective detect ion
of oxidizing gases.
Regarding the detection of reducing gases, especially
hydrogen, the opposite effect occurs on the ﬁlms. In general,
thick ﬁlms work better for hydrogen and other reducing gases.
For example, thin In
2
O
3
ﬁlms, deposited from diluted solutions,
had lower sensitivity to hydrogen than thick ﬁlms (Fig. 11, curve
3). The same effect was observed earlier for SnO
2
ﬁlms [134],
prepared by the spin-coating method. An explanation of this
effect was presented in Refs. [59,64,110], where the diffusion-
reactive model of gas sensitivity was developed. According to
Ref. [64], the increased sensitivity to H
2
in thick ﬁlms arises
because H
2
has a much higher diffusion coefﬁcient than oxygen.
It is necessary to note that the sensitivity of sensors
fabricated by thick ﬁlm technology is dependent on ﬁlm

thickness as well. However, different authors have observed
signiﬁcantly different dependencies with thickness (see
Fig. 12). Some reports have observed an increase in sensitivity
to reduc ing gases with increased ﬁlm thickness [137], while
others have observed a loss in sensitivity [135], and one study
observed that the sensor response would reach a maximum at
some thickness [138]. Such disagreement demonstrates once
again that gas sensitivity of metal oxides is dependent on many
factors, which are hard to control.
Because the depth of penetration of various gases into the
oxide matrix depends on their diffusion coefﬁcient and activity,
the disposition of contacts (on top or below gas-sensing layer)
starts playing an important role for ceramics or thick ﬁlm
sensors. This effect was studied in detail in Refs. [110,123], and
was used for the determination of gas diffusion parameters into
tin oxide. Research has shown that sensor characteristics, in
particularly the gas nature inﬂuence on the temperature
dependence of sensor response, are strongly dependent on
the position of electrodes (see Fig. 14). Such a strong effect
might be used for deﬁnition of the nature of detecting gas. At
sufﬁcient thickness the top layer of the sensing material could
act as a ﬁlt er for certain gas molecules [139]. This effect could
also explain the conclusion made in Ref. [140] regarding the H
2
response of the SnO
2
-based sensors with two types of noble
metal (Au, Pt, and Pd) electrodes covering the surface of the tin
oxide nanohole arrays. It was found that the temperature
dependence of the sensor response differed between the sensors

equipped with a pair of electrodes on both surfaces and the
sensors equipped with a couple of inter-digital electrodes on
one side. At that, the H
2
response of the sensors equipped with a
pair of electrodes on both surfaces was much higher than that of
the sensors equipped with inter-digital electrodes on one side.
With increased ﬁlm thickness, problems arise in using
physical methods for the deposition of noble metals catalysts
onto metal oxides [70,118]. In Refs. [62,72,125], this problem
was studied for SnO
2
ﬁlm doping by Pd and Pt. Some results are
Fig. 12. Sensor response of SnO
2
-based sensors to reducing gases vs. ﬁlm thickness, determined in various laboratories: (a) devices were fabricated by dropping and
spinning of sol suspension over an alumina substrate attached with comb-type Au electrodes. SnO
2
powders were prepared by hydrothermal method; (b) SnO
2
ﬁlms
were deposited by (1) MOCVD method on alumina substrates with two Au electrodes, with following annealing at 600 8C for 15 h, and (2) by reactive DC sputtering
with following annealing at 600 8C for 10 h. (Adapted with permission from Refs. [70,134,135]. Copyright 1994, 2001, and 2003: Elsevier).
Fig. 13. SnO
2
ﬁlm thickness inﬂuence on normalized S(T
oper
) dependencies of
sensor response to reducing gas. Sensors were fabricated on the base of thin
ﬁlms deposited by spray pyrolysis from SnCl

4
–water solution. (Reprinted with
permission from Ref. [136]. Copyright 2001: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 9
shown in Fig. 15. It was concluded that the optimal technical
solution varied with the type of noble metal used as an additive
[72].
Other important aspect of the inﬂuence of ﬁlm thickness on
gas sensor performance pertains to response and recovery
times. The effect of ﬁlm thickness on the time constants of
sensor response to ozone and hydrogen are shown in Figs. 15
and 16. One can see that the time constants of sensor response
increase as the ﬁlm thickness increases. At that, the effect is
more pronounced with oxidizing gas than for reducing gases.
Response and recovery times for reducing gas exposure on
In
2
O
3
-based thin ﬁlm gas sensors increased nearly 10-fold as
the ﬁlm thickness changed from 20 to 400 nm (Fig. 16).
Response times during ozone detection in a dry atmosphere
changed almost two orders of magnitude (Fig. 17), although in
humid atmosphere, this effect was weaker (see Fig. 17).
Thus, gas sensor designers should decrease thickness to
improve the sensing characteristics of metal oxide-based gas
sensors. The use of thin ﬁlms assures fast response and
recovery. As shown empirically, there is not any diffusion
limitation in response kinetics in thin-ﬁlm devices [81].
The same conclusion was made by the authors of Ref. [143],

studying gas-sensing properties of ZnO sensors fabricated by
magnetron sputtering. It was found that sensors with minimal
ﬁlm thickness in the range 65–390 nm had the maximum
response to CO and minimum response time. These results
indicate that thin ﬁlm gas sensors (d < 100 nm) will always be
faster than thick (d > 100 nm) ﬁlm gas sensors [63]. Another
Fig. 14. Inﬂuence of electrode position (a) on gas sensitivity of SnO
2
-based thick ﬁlm sensors (d $ 250–300 mm) loaded with 1.0 wt% of Pt or Pd to (b) H
2
and (c)
CH
4
. Porous thick ﬁlm sensors having (2) interior and (1) surface electrodes were fabricated on a porous mullite tube of 2 mm enter diameter and 1.7 mm inner
diameter. SnO
2
powders had a surface area (S
surf
)of75m
2
/g). (Adapted with permission from Ref. [64]. Copyright 1998: Elsevier).
Fig. 15. Scheme of the three doping methods (a), and inﬂuence of doping methods on the gas response of SnO
2
-based sensors (d $ 200 nm) to 100 ppm CO
(T
oper
= 400 8C, RH = 40%). SnO
2
ﬁlms and catalytic additives, (b) Pt and (c) Pd were deposited by reactive DC sputtering. 1, 2, and 3 correspond to methods of
doping shown in ﬁgure. (Adapted with permission from Ref. [72]. Copyright 2003: Elsevier).

G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–3910
conﬁrmation of this statement can be found in Ref. [142], where
thick (d ! 1.0 mm) ﬁlm In
2
O
3
-based ozone sensors were
described. Because of optimization of the ﬁlm structure
(increased ﬁlm porosity), gas sensor sensitivity to ozone was
improved. However, the authors failed to decrease both t
res
and
t
rec
. Even at operating temperatures of approximately 390 8C
the t
res
exceeded 6 min.
Another important advantage of thin ﬁlm sensors, especially
those designed in epitaxial and 1D structure versions, is better
temporal stability (see Fig. 18) at high operating temperatures.
The change of grain size in thin ﬁlms during thermal treatment
is not as strong as for thick ﬁlms.
However, the diminution of ﬁlm thickness has limitations,
controlled by the properties of material itself and the technical
difﬁculty in forming a continuous layer. For example a
thickness of 10–15 nm is the limit for standard thin-ﬁlm
technologies usually used for gas sensor fabrication. For thick
ﬁlm technology, a minimal thickness of about 50–150 nm is
required to form a continuous ﬁlm [113].

Besides slow response and recovery processes in thick ﬁlm-
and ceramics-type sensors, it is necessary to take into account
that above a certain thickness, ceramic and thick ﬁlms have an
inclination to cracking [87]. An example of such a process is
shown in Fig. 19. As a result of cracking, the ﬁlm electro-
conductivity and gas-permeability could change fundamen-
tally. For example, both response and recovery times in such
sensors could decrease due to improved gas-permeability.
However, the process of cracking is hard to control, because the
magnitude of the sensor parameters change is dependent on the
depth and length of the cracks. Probably therefore the authors of
[113], considering the inﬂuence of ﬁlm thickness on sensor
response, did not take into account the results, obtained for
ﬁlms with thickness more than 200 nm.
2.2.2. Grain size inﬂuence
At present either the ‘‘grains’’ model, or ‘‘necks’’ models (see
Figs. 9 and 20) are applied to rationalize the electro-physical
properties of polycrystalline materials, which are dependent
strongly on their microstructure [24,46,47,50,144,145]. It has
been established that the grain size and the width of the necks are
the main parameters that control gas-sensing properties in metal
oxide ﬁlms as well. Moreover in the frame of modern gas sensor
models the inﬂuence of grains size and necks size on sensor
response may be attributed to the fundamentals of gas sensor
operation [24,46,47,57,93,144–146]. Usually it is displayed
through the so-called ‘‘dimension effect’’ e.g., a comparison of
Fig. 16. SnO
2
ﬁlm thickness inﬂuence on response time during detection (1)
CO (T

oper
= 340 8C) and (2) CH
4
(T
oper
= 430 8C). SnO
2
ﬁlms were deposited by
spray pyrolysis method. (Reprinted with permission from Ref. [281]. Copyright
1999: Elsevier).
Fig. 17. Dependencies of time constants of In
2
O
3
response to ozone on
thickness of porous ﬁlms (T
pyr
= 475 8C): (1) recovery time; (2) response time
in dry atmosphere; (3) response time in wet atmosphere. Dry air corresponds to
1–5% RH, wet air corresponds to 40–50% RH. (Reprinted with permission from
Ref. [102]. Copyright 2007: Elsevier).
Fig. 18. Inﬂuence of annealing temperature on average grain sizes in In
2
O
3
ﬁlms estimated on the base of AFM measurements. In
2
O
3
ﬁlms were deposited

by spray pyrolysis method: (1) T
pyr
= 520 8C; d $ 200 nm; (2) T
pyr
= 520 8C;
d $ 50 nm; (3) T
pyr
= 400 8C; d $ 50 nm. (Reprinted with permission from
Ref. [77]. Copyright 2005: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 11
the grains size (d) or necks width (X) with the Debye length (L
D
)
L
D
¼
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
ekT
2pe
2
N
r
; (1)
where k is the Boltzmann constant, T is the absolute tempera-
ture, e is the dielectric constant of the material, and N is the
concentration of charge carries.
The distribution of the potential along polycrystalline oxides
for different d, X and L is presented in Figs. 9 and 20. It is clear
that the width of the necks determines the height of the potential
barrier for current carriers, while the length of the necks

determines the depletion-layer width of the potential barrier. It
is necessar y to note that the increase of the necks length
increases the role of necks in the limitation of metal oxide
conductivity, and correspondingly in gas-sensing effects. The
grain size would determine the depth of valley on the potential
distribution within grains. Corresponding potential diagrams
for one-dimension structures are given in the Fig. 9c and d. The
distribution of the potential looks similar to those observed in
individual grains except that at the boundary of a one-
dimensional structure with Ohmic contact, the potential barrier
would be considerably lower than the potential barrier between
grains.
In brief, for explanation of the ‘‘dimension effect’’ on the
gas-sensing effect it is possible to provide the following
Fig. 19. Cracking inﬂuence on morphology of In
2
O
3
and gas penetrability of metal oxides. (Adapted with permission from Refs. [87,134]. Copyright 2001 and 2004:
Elsevier).
Fig. 20. Diagram illustrating the role of necks in the conductivity of polycrystalline metal oxide matrix and the potential distribution across the neck. (Reprinted with
permission from Ref. [32]. Copyright 2007: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–3912
argument (see Figs. 9 and 20) [50,146]. For large crystallites
with grain size diameter d ) 2L
s
(Fig. 9b), where L
s
is the
width of surface space charge ðL

S
¼ L
D
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
eV
2
S
=kT
p
Þ, and for a
small width of necks (d < L
S
), the conductance of both the ﬁlm
and ceramics usually is limited by Schottky barriers (V
S
) at the
grain boundary. In this case the gas sensitivity is practically
independent of d.
In the case of d $ 2L
s
every conductive channel in necks
between grains is overlapped (see Fig. 20). If the number of
long necks is much larger the inter-grain contacts, they control
the conductivity of the gas-sensing material and deﬁne the size
dependence of the gas sensitivity.
If d < 2L
s
, every grain is fully involved in the space-charge
layer (Fig. 9a) and the electron transport is affected by the
charge at the adsorbed species. In Ref. [147], it was

demonstrated that, when the grain size becomes comparable
to twice the Debye length, a space-charge region can develop in
the whole crystallite. The latter case is the most desirable, since
it allows achievement of maximum sensor response. More
detailed descriptions o f models used for explanation of both
grain size and necks inﬂuence on gas-sensing effects may be
found in Refs. [26,50,56,57,157].
It is necessary to note that the applicability of ‘‘grains’’ or
‘‘necks’’ models depends strongly on the technological routes
used for metal oxides synthesis or deposition and sintering
conditions [57]. Usually the appearance of necks is a result of
high temperature annealing (T
an
> 700–800 8C). Taking into
account processes which take place at inter-grain interfaces
during high temperature annealing, one can assume that the
forming of long necks in inter-grain space is a consequence of
mass transport from one grain to another. According to Ref.
[144] for metal oxide samples after high temperature annealing,
the neck size (X) is proportional to grain size (d) with a
proportionality constant (X/d) of 0.8 Æ 0.1. However, it is
necessary to note that this constant depends on the sintering
parameters and may be different. For thin metal oxide ﬁlm and
ceramics, which were not subject to high temperature
treatments, the gas-sensing matrix is formed from separately
grown grains. Therefore, in such metal oxides, the necks
between grains are very short or are absent. It means that for
description of their gas-sensing properties we can use the
‘‘grains’’ model. According to this model, between grains there
are Schottky type contacts with the height of potential barrier

depending on the surrounding atmosphere. In the frame of such
approach the grain boundary space charge or band bending on
inter-grain interfaces are the main parameters controlling the
conductivity of nanocrystalline metal oxides.
The adequacy of above-mentioned model was estimated on
the base of results obtained during the impedance spectroscopy
of metal oxides. At present, impedance spectroscopy is the
most effective method for experimental determination of
factors that limit the conductivity of metal oxides. For example,
in Ref. [148], impedance spectroscopic measurements of SnO
2
ﬁlms with grain size equaling 5–14 nm showed that at low
operating temperatures (25–300 8C), both the grain and grain
boundary contribute to the conductivity. At higher operating
temperatures (above 300 8C), the grain boundary contribution
for the conductivity is dominant. The same results was obtained
in Ref. [149] for SnO
2
ﬁlms (d $ 0.6–1.0 mm) during
interaction with H
2
S and NH
3
at 250 8C. The impedance of
these ﬁlms was mainly contributed by the potential barriers at
grain boundaries. Labeau et al. [150] also have found that the
main contribution to the sensor impedance of polycrystalline
SnO
2
sensors during interaction with CO and C

2
H
5
OH arises
from grain boundaries, although a small contribution from bulk
was also seen at high frequencies. As was shown in Ref. [151],
the grain boundaries are limiting elements in conductivity of
WO
3
thin ﬁlms (d $ 50 nm) under dry air or/and ozone in the
temperature range from 150 to 375 8C as well.
In Ref. [152] using the same impedance method it was
shown that the contribution of the grain boundary in the total
conductivity of a metal oxide depends also on the grain size.
The impedance spectra of SnO
2
ultra dispersed ceramics
(d $ 3–43 nm) have been investigated in the temperature range
25 8C < T < 300 8C under a dry oxygen atmosphere. It was
established that the grain boundaries give the major contribu-
tion to the electric transport for the samples with d less than
25 nm. For the samples with larger grain size, the contributions
of the grain volume and grain boundaries to the conductivity are
of the same order. The authors of Ref. [152] assumed that grains
with size smaller 24 nm are completely depleted by charge
carriers.
Regarding experimental conﬁrmation of grain size inﬂuence
on sensors response one can say that the dramatic increase in
sensitivity for metal oxides with grain size smaller than a Debye
length has been demonstrated many times for various materials,

such as SnO
2
[26,34,56,60,85,153],WO
3
[154,155] and In
2
O
3
[79,144,156]. This effect for In
2
O
3
and SnO
2
-based sensors is
illustrated in Figs. 21 and 22. For example, in Ref. [30] it was
established that the sensitivity of sensors based on tin oxide
Fig. 21. (1–5) Theoretical and (6–8) experimental dependencies of SnO
2
sensor
response to reducing gas on grain size for (7) undoped SnO
2
ceramics and
ceramics doped by (6) Sb and (8) Al: (1) N
d
=10
17
cm
À3
; 2–3 Â 10

17
cm
À3
;3–
10
18
cm
À3
; 4–3 Â 10
18
cm
À3
; 5–10
19
cm
À3
. Porous sensor elements were
fabricated on an alumina tube with Pt wire electrodes. Elements were sintered
at 700 8C for 4 h. SnO
2
powders were synthesized by conventional hydrolysis of
SnCl
4
. Foreign oxides (5 at.%) were added by an impregnation method.
Sensitivity to H
2
(800 ppm) was estimated at 300 8C. (Adapted with permission
from Ref. [34,56]. Copyright 1991 and 2005: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 13
nanoparticles dramatically increased when the particle size was

reduced down to 6 nm. Below this critical grain size, the sensor
sensitivity rapidly decreased. As the calculated Debye length of
SnO
2
is L
D
= 3 nm at 250 8C [147], the highest sensitivity was
actually reached when the particle diameter corresponded to
2 nm.
However, it is necessary to note that the threshold value of
grain sizes determined in Ref. [30] is not an invariable constant
of SnO
2
. Wang et al. [157] considered grain size effects in
particulate semiconductor gas sensors, which contained
mixtures of necks and grain boundary contacts, and concluded
that the gas sensitivity would increase sharply for particle
diameters below about 35 nm. Studies on tin dioxide-based
hydrogen sensors carrie d out in Ref. [159] indicated that
devices produced using 20 nm particles were around 10 times
more sensitive than devices made from 25 to 45 nm particles.
Lu et al. [158] found that the SnO
2
-based sensor response to
500 ppm CO increases drastically if the particle diameter
becomes smaller than 10 nm.
Because the Debye length depends on the concentration of
free charge carriers (see Eq. (1)), it becomes clear that the
observed variation of threshold value of crystallite’s size is
valid. Moreover, this parameter must be dependent on the

material’s properties and the doping. For example, the Debye
length in SnO
2
($3 nm) estimated in Ref. [147] corresponds to
a donor concentration equal to n
d
= 3.6 Â 10
24
m
À3
. The
decrease of free charge carriers’ concentration, for example,
through metal oxide stoichiometry improvement induced by
annealing or doping, can considerably increase this threshold
value [144]. Thus, Al-doped SnO
2
(see Fig. 21, curve 8) shows
high sensitivity with increasing grain size even at d above
20 nm, while Sb-doped SnO
2
(see Fig. 21 curve 6) is insensitive
in the whole d region (see Figs. 19 and 20). If d ) L
D
, the
increase of sensor response with grain size decrease is not so
signiﬁcant [160].
The effect of grain size on the sensitivity of chemoresistive
nanocrystalline metal-oxide gas sensors was evaluated also in
Ref. [145] by calculating the effective charge carrier
concentration as a function of the surface state density for a

typical sensing material, SnO
2
, with different grain sizes
between 5 and 80 nm. These calculations demonstrated a sharp
decrease in the charge carrier concentration when the surface
state density reached a critical value that corresponds to a
condition of fully depleted grains, namely, when nearly all the
electrons are trapped at the surface. Assuming that the
variations in the surface state density are induced by surface
interactions with ambient gas molecules, authors of Ref. [145]
simulated the response curves of nanocrystalline gas sensors.
The simulations showed that the conductivity increases linearly
with decreasing trapped charge densities, and that the
sensitivity to the gas-induced variations in the trapped charge
density is proportional to 1/d, where d is the average grain size.
However, experimental results presented in Ref. [79] shown
that this dependence can be stronger. As it is seen in Fig. 22 the
dependence of sensor signal on grain size for In
2
O
3
-based
sensors obeys the correlation S $ t
À3
for ozone detection in dry
atmosphere. It means that the authors of Ref. [145] probably did
not take into account all factors inﬂuencing the gas sensitivity
of solid-state sensors.
It is necessary to note that relationship to grain size is
dependent on the type of metal oxide, detection mechanism,

and the analyzed gas. For example, in In
2
O
3
-based gas sensors,
the role of crystallites’ size is appreciably lower for the
detection of reducing gases in comparison with oxidizing gases
[79–81].InRef.[81],itwasfoundthatforthecertain
deposition conditions, the sensor signal to the reducing gases
could either increa se or decreas e wi th increasing gr ain size. In
other words, for reducing gases, the In
2
O
3
grain size is less
important than in the case of SnO
2
-based gas sensor [81].The
sensitivity of In
2
O
3
-based gas sensors to reducing gases may
have an acceptable value even when th e crystal lite size is
bigger than 80–100 nm. With such crystallite sizes the
response of In
2
O
3
sensors to oxidizing gases will be minimal.

In the case of SnO
2
ﬁlms, the increase in grain size leads to
decrease of the sensor response to both oxidizing and reducing
gases [45,56]. For explanation of the observed effect in Re f.
[100] it was assumed that in In
2
O
3
-based sensors during
reducing gases detection due to high surface unstoichiometry,
the resistance of the ﬁlms was not controlled by inter-grain
barriers. X-Ray photoelectron spectroscopy (XPS) data and the
absence of dependences of ﬁlm resistance normalized to ﬁlm
thickness (RÁd)onthegrainsizecharacteristic for conductivity
limited by the resistance on inter-grain barrier s were the basis
for such a conclu si on [81,104]. For comparison in ozone
atmosphere such strong dependence of RÁd on grain size has
appeared.
For In
2
O
3
it was found also that sensors fabricated on the
base of thin ﬁlms with minimal crystallites’ size had minimal
response time (see Fig. 23) in addition to maximum sensor
response (see Fig. 22). For example, the decrease in grain size
from 60 to 80 nm to 10–15 nm decreased t
res
during ozone

detection in dry air by a factor of 50–100 times. In a humidiﬁed
Fig. 22. The inﬂuence of In
2
O
3
ﬁlm grain size on the sensor response to ozone
(T
pyr
= 475 8C; T
oper
= 270 8C): (1) ﬁlms deposited by spray pyrolysis from
0.2 M InCl
3
–water solution; (2) 1.0 M InCl
3
solution; (3) extrapolated curve
according S $ t
À3
dependence. (Reprinted with permission from Ref. [79].
Copyright 2004: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–3914
atmosphere, the established correlation was signiﬁcantly
weakened [79,80].
Besides, the surface reactivity of particles is known to
rapidly increase with the increase of the surface-to-bulk ratio
because the strong curvature of the particle surface generates a
larger density of defects, which are the most reactive surface
sites [161]. This high reactivity has largely been taken
advantage in catalysis, where ultraﬁne particles have been
used for decades. When properly processed during the

fabrication of chemical semiconductors sensors, these nano-
particles are sufﬁciently reactive to make the use of catalytic
additives (such as Pt or Pd) unnecessary and to decrease the
working temperature of the sensors without any loss of
sensitivity [162,163].
However, the decrease in grain size cannot be unlimited. At
some critical dimension, the number of free electrons in the
grain could become zero even at V
S
= 0. This leads to a grain
resistance that would not be dependant on the changes in the
surrounding atmosphere. For charge carrier concentration in
metal oxides of 10
21
cm
À3
, the critical crystallite size is 1 nm.
The use of ﬁnely dispersed small crystallites can also have a
deleterious effect on the temporal stability of the sensor [34,76].
It was shown that an excessive decrease of grain size leads to a
loss of structural stability [34,77], and, as a consequence, to
change both surface and catalytic properties of the material
[164]. For example, it was shown that for SnO
2
with a grain size
of about 1–4 nm, the grain-growth process begins already at
temperatures equal to $200–400 8C (see Fig. 24) [77,165].In
contrast, SnO
2
crystallites with average sizes ranging from 1.7

to 4.0 mm were stable up to 1050 8C.
AccordingtoRefs.[169–171], there are two main reasons
for grain growth during thermal treatments. This instabi lity
may arise from a known defect in the bulk state, such as faulty
stoichiometry, or due to the ﬁnitesizeofthegrains.Recent
theoretical simulations have conﬁrmed this statement [170].A
quantum mechani cal study of the stability of SnO
2
nanocrys-
talline grains has shown that the increase of both the grain size
in the range 0.3–4.0 nm, and the oxygen content in SnO
2
increased the stability of SnO
2
grains. The fact that the size has
a strong inﬂuence on the melting temperature of thin metallic
ﬁlms is another direct conﬁrma tion of this statement. It was
established that the decrease of Ag, Bi, Sn, and Pb ﬁlms’
thickness to 5 nm was accompanied by temperature decrease
of crystallization up to 0.6–0.7 of the melting temperature of
bulk samples [172]. As it was established in Refs. [166–168],
the decrease of grain size is accompanied by lowering of the
melting temperature of the semiconductor nanocrystals as
well.
However, the principle of surface and interface energy
minimization explains only the appearance of a driving force
for the recrystallization of polycrystalline ﬁlms during their
annealing. This principle cannot account for the threshold
nature of the grain size changes observed [77] , i.e. the presence
of a threshold temperature (T

st
) below which the crystallites
with ﬁxed size remain stable, and the absence of t
1/2
type grain
size dependencies on time during thermal annealing.
Because the process of coalescence starts through the
breaking of Me-atoms bonds with the lattice of metal oxides, in
Ref. [77] it was assumed that the formation energy of surface
and bulk vacancies of Me-atoms ( V
In
, V
Sn
) could be such a
critical energy, characterizing the temperature threshold of
structural stability. The more the energy of V
In
and V
Sn
formation is, the more stable is the lattice, i.e. the temperature
of possible grain coalescence is higher. In spite of the fact that
the observed process of grain growth in both In
2
O
3
and SnO
2
ﬁlms takes place through diffusion of Me-atoms (In, Sn)
following their incorporation in the lattice of a bigger
crystallite, this process is not controlled by the coefﬁcient of

self- or surface Me-atom diffusion. From our point of view, this
process is controlled by certain energy parameters, for example
Fig. 23. Dependencies of response time of In
2
O
3
sensors during gas detection
in dry atmosphere on the grain size of In
2
O
3
ﬁlms deposited from 0.2 M InCl
3
–
water solution: (1) ozone ($1 ppm) detection (T
oper
= 270 8C); (2) CO
($1000 ppm) detection (T
oper
= 270 8C). (Reprinted with permission from
Ref. [79]. Copyright 2004: Elsevier).
Fig. 24. Inﬂuence of annealing temperature on grain size in SnO
2
thin ﬁlms,
thick ﬁlms and ceramics, fabricated using different manufacturing methods.
(Reprinted with permission from Ref. [77]. Copyright 2005: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 15
the energy of vacancy formation, characterizing the thermo-
dynamic stability of the crystallite.
Theoretical calculations conducted for NiO [173] and Cu

2
O
[174] have shown that the energies of point defect formation in
polycrystalline ceramics are really smaller when compared
with the bulk. Such expectations, for example, have been
experimentally conﬁrmed for Cu
2
O [175]. The obtained results
have shown clearly that the point defect concentration in ﬁne-
grained metal oxides signiﬁcantly exceeds the concentration of
point defects in large-grained materials.
It thus becomes clear, that the presence of a ﬁnely dispersed
fraction with a grain size smaller than 2–5 nm will lead to some
structure instability of the metal oxide matrix at moderate
operating temperatures (T < 600 8C). This is true even for ﬁlms
with average grain sizes greater than 100 nm. Therefore, future
design methods of nano-scale devices, which assure grain size
stabilization during long-term operation at high temperature,
will gain priority over the design of methods producing nano-
scaled materials with minimal grain size. The production of
both metal oxide powders and deposition of ﬁlms with a small
dispersion of grain sizes is an effective method for improve-
ment of the temporal stability of solid-state gas sensors as well.
Utilizing thin ﬁlms in gas sensors also leads to an improved
stability of the gas-sensing matrix. Research has shown that the
grain size in thin ﬁlms during the annealing process changes
considerably less than in thick ﬁlms (see Fig. 18 ). Therefore,
unless there is a pressing need to improve detection limit s
(sensitivity), it is not advisable to design sensors with
excessively reduced grain sizes. This problem is especially

serious for operation in atmospheres of reducing gases.
As it was discussed earlier, some studies [34,162] have
considered the possibility of introducing micro additives to
limit the mobility of adatoms and stabilize grain size. This is an
interesting area of research. It is well known in metallurgy that
trace impurities or second phase particles are very effective in
altering and inhibiting grain growth. In accordance with results
of research carried out in this ﬁeld, additives such as K, Ca, Al,
or Si salts may be such grain-growth inhibitors [27,98].
Both experimental research and theoretical simulations have
established that in the presence of a second phase, grain growth
is inhibited when the average domain size is comparable to the
average inter-particle distance. Therefore, the controlled
addition of selected impurities may be one way to achieve a
particular grain size, morphology, and structural stability of
metal oxides necessary for practical applications. However,
even small quantities of additives used for grain size
stabilization can affect catalytic and adsorption surface
properties of the gas-sensing material. As a result, though
grain size may be stabilized, there may be a strong change in
both the electro-physical and the gas-sensing properties of the
metal oxide [34]. This research is still in the early stages;
therefore detailed studies conformably to metal oxides are
needed to develop real technology for the stabilization of nano-
scaled materials
As the grain size decreased, one more interesting effect was
established. With decreased grain size, a signiﬁcantly increased
sensitivity to humidity was observed [34] (see Fig. 25). The
same effect was observed experimentally in Ref. [176].
Williams and Coles found that alumina-based humidity sensors

also increased their sensitivity by an order of magnitude when
prepared from 13 nm particles as compared to 300 nm particles.
It means that the humidity inﬂuence of sensor parameters is
structure dependent and may be controlled through structural
engineering of the metal oxides used. In Refs. [75,177–179],it
was conﬁrmed that water adsorption on the surface of metal
oxides depends on its crystallographic structure. At that, the
difference of water attachment on different crystallographic
SnO
2
planes appeared not only in concentration, but also in the
type of bonds between hydroxyl groups and the SnO
2
surface.
The OH-groups participate in gas detection reactions, and water
adsorption/desorption processes can control the kinetics of gas
response [24,87,141,180,181]. Thus, if minimum sensitivity to
humidity is required, the size of crystallites in sensor material
should be the maximum permissible.
2.3. The role of crystallographic structure of metal oxides
2.3.1. Crystal shape
At present, the affect that grain shape and faceting of
crystallites has on gas-sensing is not being analyzed. However,
there are reasons to conclude that the role of these parameters is
undeservedly understated. The following statements canbe made
based on the results given in Refs. [34,36,75,85,132,177,182–
186]:
(i) The external planes of nanocrystals participate in gas-solid
interaction, and theref ore these very planes determine the
gas-sensing properties of nanostructured materials. Even

spherulities have micro-planes and facets. The nanocrystal
shape may determine such parameters as crystallo graphic
planes, inter-grain contacts, area of inter-grain contact s,
gas-permeability, and so on.
(ii) Every crystallographic crystal form has its own combina-
tion of crystallographic planes, framing the nanocrystal.
Fig. 25. Inﬂuence of pyrolysis temperature on (2) grain size and (1) sensitivity
to air humidity during H
2
(1000 ppm) detection by In
2
O
3
-based sensors
(d $ 60–80 nm). Sensitivity to air humidity was estimated as the ratio of sensor
response at T
oper
= 370 8CtoH
2
measured in dry and wet atmospheres. In
2
O
3
ﬁlms were deposited from 0.2 M InCl
3
solution.
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–3916
Every crystallographic plane has its own combination of
surface electron parameters, which include surface state
de nsity, energetic position of the levels, induced by

adsorbed s pecie s, adsorption/desorption energies of
interacted gas molecules, concentration of adsorption
surface states, the energetic position of surface Fermi
level, activation energy of native point defects, and so on.
This implies that t he chemisorption characteristics
change noticeably from the crystal surface orientation
to another. Thus, there is a large surface dependence at
the atomic level for chemi cal bonding of the adsorbed
particles.
(iii) As chemical processes, the adsorption/desorption have
activation energies. The parameters controlling these
processes are orientation and grain size dependent. The
decrease in crystal size in the nm range notably strengthens
the crystallite shape inﬂuence on the adsorption properties.
Both the shape and the size of nanocrystals have a
profound inﬂuence on the concentration of adsorbed
species and on the type of bonding to the surface that takes
place. It is known that depending on the type of bonding,
some chemical species may have preferred adsorption on
either the edge/corner sites or on the plane facet. For
example in Ref. [182], it was shown that monodentate
Fig. 26. SEM images of SnO
2
ﬁlms deposited by spray pyrolysis using different technological parameters. (Reprinted with permission from Ref. [78]. Copyright
2005: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 17
adsorption on MgO and CaO is preferred on the edge/
corner sites, whereas the bidentate adsorption is favored by
smooth planes.
Thus, depending on external form of the nanocrystallites, a

nanostructured gas-sensing matrix will have a unique
combination of structural, electronic, and adsorption/deso-
rption process parameters.
Detailed studies [34,78,132] conducted on SnO
2
ﬁlms
deposited by a spray pyrolysis method conﬁrmed the
appropriateness of this approach for modiﬁcation of gas-
sensing properties. For example, it was established that the
growth of grains, especially in the range from nm to mm, during
which the transition from spherulites to nanocrystallites and
from nanocrystallites to nanocrystals and crystals takes place, is
accompanied by the change s in both the size and the external
shape of crystallites [78,132]. Possible morphologies of SnO
2
ﬁlms deposited by spray pyrolysis are shown in Fig. 26.
In Ref. [78] it was shown that for SnO
2
ﬁlms deposited by
spray pyrolysis, the shape of crystallites with different
crystallographic planes such as (1 1 0), (1 1 1), (2 0 0),
(0 1 1), ð
¯
1
¯
12Þ, and (2 1 0) is dependent on the deposition
parameters and ﬁlm thickness. Changing the facet of
nanocrystalls as their size and growth conditions are modiﬁed
has been observed in other metal oxides as well [187].
The predominance of different crystallographic planes

causes corresponding changes in atomic and electronic
properties along with surface energy parameters. Examples
of some crystallographic planes observed in SnO
2
ﬁlms
deposited by spray pyrolysis are shown in Fig. 27. Consistent
with the results reported in Refs. [78,132,188–190], the (1 1 0)
and (1 0 1) planes of the SnO
2
crystal are F (plane) faces, while
the (1 1 1) plane is K (kinked). This means that the (1 1 1) plane
has a much rougher surface than both (1 1 0) and (1 0 1) planes.
According to the periodic bond chains (PBC) theory [189],
surface atoms of F-faces are strongly bound to each other in
directions parallel to the face. Accordingly, these faces have a
small tendency to react with the arriving atoms. In reality this
statement is not entirely correct, because besides ‘‘in-plane’’
oxygen at the (1 1 0) SnO
2
surface, there are also ‘‘bridging’’
oxygen ions located at equal distance above and below the
surface plane. However, the statement that the (1 1 0) surface
plane is the most stable facet, is correct. Indeed, the faces
parallel to only one PBC (stepped S-faces) or to none (kinked
K-faces) are highly reactive because there are many unsaturated
bonds cutting their surfaces.
The catalytic activity of atomic planes to a large extent is
determined by the surface concentration of non-saturated
cations and weakly bounded bridging oxygen. The following
rank of catalytic activity (CA) of ideal atomic planes can be

proposed: CA
(110)
< CA
(001)
,CA
(100)
< CA
(101)
. Simple esti-
mations show that dissociative chemisorption on the surface of
SnO
2
is orientation dependent as well. Various crystallographic
planes have different distances between Sn atoms, which form
the following series d
(110)
$ d
(100)
< d
(101)
< d
(001)
[132].Tin
atoms are centers of oxygen chemisorption, and therefore the
change of indicated distance must inﬂuence the rate of
dissociative oxygen chemisorption, which in many cases is a
controlling factor of gas-sensing phenomena [4,24,28,29,46].
The nature of the bonding of an adsorbed gas molecule with
the metal oxide is anot her important factor [37]. For example,
the authors of Ref. [182] observed that monodentate adsorption

on MgO and CaO is preferred on the edge/corner sites. This is
unique to small nanocrystallites (spherulites). In contrast
bidentate adsorption is favored by ﬂat planes, which are more
prevalent on the larger nanocrystals. This feature of surface
species adsorption indicates that the control of surface
roughness and grain shape can be exploited for improvement
of gas-sensing characteristics such as absolute magnitude and
selectivity. Such control may be as effective as the use of
catalytic additives. This last factor is an important one for the
stability improvement of solid-state gas sensors [34].
Results of experimental research presented in Refs.
[132,177,191] and theoretical simulations discussed in Refs.
[75,177] gave the same conclusion. For example, research
carried out in Ref. [177] has shown that the (1 1 0), (1 1 0) and
(1 0 1) SnO
2
planes have different surface energies, different
phase transition conditio ns, and different energy spectra of
surface states generated during their reduction. Different
crystallographic planes also have different peculiarities of
interaction with water [75,186]. In Ref. [191], it was established
that SnO
2
ﬁlms with different texturing have different catalytic
activity for selective oxidation of CH
4
. In thes e investigations,
one type of SnO
2
ﬁlm had predominant orientation in the

Fig. 27. The models of unrelaxed rutile SnO
2
surfaces: (a) (1 1 0); (b) (1 0 0); (c) (0 0 1); (d) (1 0 1). The light large balls represent tin atoms and the dark small balls
oxygen atoms. (Reprinted with permission from Ref. [76]. Copyright 2005: IOP Publishing Ltd.).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–3918
{1 1 0} crystallographic direction and another in the {2 1 1}
and {3 0 1} directions.
So, the determination of crystallographic planes with
optimal combinations of adsorption/desorption and catalytic
parameters, and the development of methods for grains
deposition with indicated faces can be considered as a main
contemporary goal for thin ﬁlm technology as applied to metal
oxide gas sensors. Research, presented in Ref. [177] could be
considered as the ﬁrst step in this direction.
A proper choice for crystallite deposition technology or
synthesis with a necessary grain facet can be also one method to
decrease humidity effects in gas sensors. For example, in Ref.
[75], using a Mulliken population analysis it was shown that the
chemisorption of OH-groups on the (1 1 0) face is accompanied
by the localization of negative charge to a greater extend than
the chemisorption of OH-groups at the (0 1 1) surface of SnO
2
.
This means that adsorption/desorption processes and surface
reactions with water vary with different SnO
2
crystallographic
planes.
This consideration rationalizes the large dispersion of both
the temperature position of sensor response maximum and the

half-width of the temperature dependence of conductivity
response observed for SnO
2
ﬁlms deposited using various
technological parameters (see Fig. 28). As indicated earlier,
every structure of a polycrystalline ﬁlm has its own
combination of crystallographic planes that can participate
in the gas-sensing reaction. Since a large variety of crystal
structures can be obtained during metal oxide deposition, a
variety of shapes for the S(T) dependencies, reﬂec ting the
speciﬁc of adsorption/desorption (A/D) processes on these
planes will inevitably be observed. Experimental S(T)
dependencies are a superposition of individual chemical
reactions taking place on the individual surface planes.
Understandably, the preparation of polycrystalline metal
oxides with necessary grain faceting is difﬁcult to control. But,
it is achievable for one-dimensional sensors and should be a
high-priority area of research. One-dimensional structures are
crystallographically perfect and have clear faceting with a ﬁxed
set of planes. Research has shown that the planes and faceting in
one-dimensional structures depend on the parameters of
synthesis. For example, in Ref. [192] it was reported that
using solid–vapor phase deposition, it was a possibility to
synthesize In
2
O
3
nano-belts with {1 0 0} and {1 2 0} growth
directions. Both types of nano-belts had the top and bottom
surfaces being (0 0 1), while the {1 0 0} nano-belts had side

surface of (0 1 0) and a rectangular cross-section. The {1 2 0}
nano-belts had a parallelogram cross-section. Furthermore, it
was reported in Refs. [193,194] that In
2
O
3
nano-belts might
have growth directions of {1 1 1} and {1 1 0} with side and top
surfaces being (1 0 0). In Ref. [195] it was found that In
2
O
3
nano-belts were enclosed by (1 0 0) and (0 1 0) planes and that
their growth direction was parallel to {0 0 1}. As discussed
earlier in this review, the possibility to control faceting planes
by growth conditions gives addi tional opportunities for
controlling the sensor’s performance parameters.
The crystallographic geometry of other metal oxide one-
dimensional nanostructures is presented in Table 1.Itis
important to note that, for example, the SnO
2
nanostructures are
not enclosed by (1 1 0) planes, which are the most stable
crystallographic planes in the SnO
2
lattice. It means that
crystallographic planes, which have never been analyzed by
methods of theoretical simulation, participate in gas-sensing
effects.
It is necessary to note that semiconducting one-dimension

metal oxide structures with well deﬁned geometry and perfect
crystallinity could represent a perfect model-material family
for systematic experimental study and theoretical under-
standing of the fundamentals of gas-sensing mechanisms in
metal oxides.
In analyzing the opportunities of one-dimensional structures
of various types for their practical application in gas sensors, it
is necessary to note that nano-belts (nano-ribbons) probably
could be the most demanding one-dimensional structure to
exploit for gas-sensing applications. Nano-belts are thin and
plain belt-type structures with rectangular cross-section (see
Fig. 29). At present, nano-belts have been obtained for nearly
all oxides used in gas sensors. There is considerable data
pertaining to the synthesis of nano-belts for SnO
2
,In
2
O
3
,ZnO,
Ga
2
O
3
,TiO
2
, etc. [39,192,195,199–205]). Typical nano-belts
have widths of 20–300 nm, and lengths from several mmto
hundreds, or even some thousands of mm [195,200,206]). The
typical width-to-thickness ratio for nano-belts ranges from 5 to

10. For comparison, for nanowires (or nanorods) this ratio
equals 2–5 [195]. Synthesis of nano-belts coul d be done with
various methods [195,199,207–210] which provide consider-
able opportunities for research on such nano-size materials.
Nano-belts do not have the mechanical strength of
nanotubes. However, they have structural homogeneity and
crystallographic perfectio n. It is well known that crystal-
lographic defects may destroy quantum-size effects. Because of
the zero-defects of nano-belts, structural defects will not be a
problem as observed for nanowire type structures. It is
necessary to emphasize that that suitable geometry (see
Fig. 30), high homogeneity of the structure, and long length
are important advantages of nano-belts for mass manufacturing
Fig. 28. Temperature dependencies of sensor response to H
2
for undoped SnO
2
ﬁlms deposited at different parameters of spray pyrolysis and corresponding
deconvoluted peaks (Reprinted with permission from Ref. [85]. Copyright
2001: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 19
of gas sensors. Besides, nano-belts have ﬂexible structures, and,
therefore, they could be curved up to 1808 without being
damaged. As it is known, nanotubes do not have such
properties. This fact gives additional advantages to nano-belts
for sensor designs.
2.3.2. Surface geometry
Unfortunately, the role of surface geometry in gas- sensing
effects, which includes the role of surface steps, terraces,
crystallographic defects, and corners, has not been studied in

detail, in spite of the fact that the inﬂuence of surface faceting in
gas adsorption phenom ena and catalysis is a well known
phenomenon [211,212]. The concept of an ‘‘active site’’ is one
of the funda mentals in heterogeneous catalysis. This notion
reﬂects experimental evidence that, in general, catalytic
surfaces are not uniformly active. Rather, only sites,
distinguished by a special arrangement of surface atoms
(including defects), or by a particular chemical composition are
actually reactive. For example, numerous experiments have
shown that in many cases monoatomic steps are highly
preferential sites for many surface reactions. To address this
issue, Zambelli et al. [214] studied the dissociative chemisorp-
tion of NO on the (0 0 0 1) surface of ruthenium, which is
known to be the most sel ective catalyst. The surface was
exposed to a small dose of NO (0.3 Langmuir,
1L=10
À6
Torr s) at room temperature (RT), and subsequently
scanning tunnel microscopy (STM) images were recorded.
From the distribution of product atoms, observed in STM
micrographs, it was inferred that adsorbed NO molecules
diffuse rapidly across surface terraces until they meet a step,
where they are observed to dissociate with high probability.
Another good example of the structural sensitivity of a gas
interaction with a solid-state surface is hydrogen’s behavior at
the surface of SnO
2
. In Refs. [37,115], it was shown that H
2
molecules are not activated on smooth SnO

2
surfaces of single
crystals. The activation means excitation of a bond and, as a
possible consequence, ionization, dissociation, or formation of
radicals. The same effect was observed for WO
3
and V
2
O
5
.
However, on the rough surface of sintered SnO
2
, activation and
reaction to H
2
O occurs at 470 K. In contrast to H
2
, as it follows
from Refs. [37,115], methane can be activated on a smooth
SnO
2
surface, while it forms a CH
3
radical.
The important role of surface steps in diffusion processes
was established for perovskites as well. In Ref. [213] it was
found that the in-diffusion of surface oxygen vacancies mostly
takes place at the step edges. In Ref. [215], by a Monte Carlo
simulation of the deposition process and a mean-ﬁeld theory

model of the SrTiO
3
surface structure, it was established that,
during deposition by laser molecular beam epitaxy, the
concentration of oxygen vacancies close to step edges is larger
than that on ﬂat terraces and remains stable. It was suggested
that oxygen vacancies diffusing on the surface tend to
Table 1
Crystallographic geometry of one-dimensional oxide nanostructures
Nanostructures Crystal structure Growth direction Top surface Side surface Ref.
ZnO-belt Wurtzite
[0001]or½01
¯
10Æð2
¯
1
¯
10Þ or Æð2
¯
1
¯
10ÞÆð01
¯
10Þ or Æ(0 0 0 1)
[195]
Ga
2
O
3
-belt Monoclinic [0 0 1] or [0 1 0] Æ(1 0 0) or Æ(1 0 0)

Æ(0 1 0) or Æð10
¯
1Þ
[195]
Ga
2
O
3
-sheet Monoclinic [1 0 1] (normal) Æ(1 0 0) Æ(0 1 0)
Æð10
¯
1Þ and Æð21
¯
2Þ
[195]
t-SnO
2
-belt Rutile [1 0 1]
Æð10
¯
1ÞÆ(0 1 0) and Æð10
¯
1Þ
[195]
SnO
2
-belt Rutile [1 0 0] Æ(0 0 1) [196]
t-SnO
2
-wire Rutile [1 0 1]

Æð10
¯
1Þ
Æ(0 1 0) [195]
SnO
2
-belt
(Zigzag-initial) Rutile [1 0 1] Æ(0 1 0)
Æð10
¯
1Þ and Æ(1 0 0)
[197]
(Zigzag-ﬁnal) Rutile [1 0 1] Æ(0 1 0) Æ(1 0 0) [197]
a-SnO
2
-wire Orthorhombic [0 1 0] Æ(1 0 0) Æ(0 0 1) [195]
SnO
2
-diskette Tetragonel Æ[1 0 0] and Æ[1 1 0] Æ(0 0 1) Æ(1 0 0) and Æ(1 1 0) [195]
SnO
2
-ribbon Rutile [1 0 1]
ð10
¯
1Þ=ð
¯
101Þ (0 1 0)/ð0
¯
10Þ
[198]

SnO
2
-ribbon
(Sandwich) Rutile/orthorhom.
[1 1 0]
o
/½6
¯
53
t
Æ(1 0 0)
o
/Æ(2 3 1)
t
Æ(0 0 1)
o
/Æð10
¯
1Þ
t
[195]
Fig. 29. Schematic diagram of the geometrical conﬁguration of SnO
2
nano-
belts.
Fig. 30. Schematic illustration of (a) nanowires and (b) nano-belts position on
bonding pad and typical sizes of nano-belts used in gas sensor design.
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–3920
accumulate near step edges due to their slow in-diffusion rate
there, and that this in-diffusion dominates the oxidation of the

as-deposited ﬁlm. The step edges act as a route for oxygen
vacancies on the surface to move into the ﬁlm. The authors of
Ref. [215] concluded that the in-diffusion rate of oxygen
vacancies is limited by step edges. No matter how large the
surface diffusion rate is, the oxidation process is dominated by
the in-diffusion of oxygen vacancies near step edges.
These observations illustrate that the adsorption properties
of many gases are very sensitive to the microscopic surface
geometry of metal oxides. Therefore, by expanding our
understanding of these processes along with the development
of necessary technologies, this phenomenon could be used for
improvement of the selectivity of solid-state gas sensors. For
example, it is possible that the increased sensitivity to nitrogen
oxide observed after high temperature annealing and subse-
quent mechanical milling of SnO
2
[58,68,216] is a consequence
of appearance of micro-steps formed as a result of the
mechanical crushing.
Within the context of the surface morphology role in sensor
performance, the results discussed in Ref. [178], deserve
special attention. Analyzing the interaction of water with a
SnO
2
(1 0 1) surface, the authors of Ref. [178] have shown that
the water weakly interacts with the reduced surface; the small
amount of water at the SnO
2
surface for T > 130 K may be
connected with water adsorption at surface defect sites such as

step edges [177]. This implies that surface geom etry may
control water inﬂuence on gas response as well.
The effect of noble metal surface clustering, which has not
been studied in the gas-sensing ﬁeld, could be directly
connected with surface geometry as well. There are statements
[62,217–219] indicating that the dispersion state of surface
catalyst particles, which can be characterized by particle size
and population, can favorably affect gas-sensing character-
istics. It has been shown that improved sensitivity can be
achieved if the aggregate distribution of noble metals such as Pd
and Pt in the ﬁlm is characterized by very small particle size
coupled with a high number density [220]. However, it was
impossible to ﬁnd in the literature any correlation between the
size of surface clusters and gas sensit ivity.
At the same time the most recent research has shown that the
process of surface clustering is structurally sensitive [213], and
therefore the consequences of surface modiﬁcation would
depend on the surface structure of the used metal oxides. The
most important consequences pertain to such structural
parameters as surface geometry, i.e. the presence of steps,
terrace and facets, size of planes, degree of surface reduction,
and others [221,222].
Recent works [221,223,224] have provided a dir ect
experimental conﬁrmation tha t fol lowing therm al tre at men ts,
noble metal clusters accumulate at the step edges of metal
oxides (see Fig. 31). Due to this behavior o f noble metals, an
increase of terrace area should be accompanied by an increase
in the size of clust ers, which can be accumulated at the step
edges of this terrace. It means that for the same degree of
surface coverage by noble metals, the number of clusters will

be smaller, while the distance between the clusters will be
bigger on the surface with the bigger terrace area (see
Fig. 32).
From this it follows that the appearance of extended
atomically ﬂat surface planes facilitates the process of cluster
growth. At the same time the presence of atomically stepped-
like surfaces will provide with high probability the conditions
for atomic dispersion of noble metals at the surface of metal
oxides. In Ref. [222], it was established that Au clusters had
higher densities on highly reduced TiO
2.
It was suggested that
reduced Ti sites act as active sites for the nucleation and growth
of Au clusters.
The size of clusters also depends on the nature of deposited
metal. For example, in Ref. [225], analyzing the behavior of Pt
and Pd atoms on a ZrO
2
surface, it was found that the Pd-atoms
have higher surface mobility and more easily form larger metal
clusters than Pt. The higher mobility of the Pd agrees with other
experimental data [225]. It was established that the cluster size
distribution of Pd on the TiO
2
(1 1 1) surface readily moves
towards larger cluster sizes after annealing at 700 K.
Alternatively, the size distribution of Pt is unchanged aft er
annealing, with smaller clusters bein g stable at higher
temperatures. Furthermor e, it was found that Pd shows a
higher probability to adsorb during deposition on the both

terrace and step sites, while Pt is characteristically observed on
the steps (see Fig. 33)
Fig. 31. STM image of Au/TiO
2
(1 1 0)–(1 3 1) after 120 min of CO:O
2
(2:1)
exposure at 10 Torr (300 K). The Au coverage was 0.25 ML, and the sample
was annealed at 850 K for 2 min before the exposures. (Reprinted with
permission from Ref. [223]. Copyright 1998: Science).
Fig. 32. Schematic illustration of terrace size inﬂuence on the size of noble
metal clusters. (Reprinted with permission from Ref. [34]. Copyright 2005:
Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 21
The thickness of deposited noble metal layers appreciably
inﬂuences the cluster size as well. As a rule, for maximum
sensor response, the concentration of noble metal additives in
the bulk of metal oxides should not exceed 1–2 wt%
[69,220,226]. According to Ref. [220] Pt-clusters in SnO
2
have a mean-size smaller than 1.5–2.0 nm. Experimental
results indicate that the thi ckness of the deposited noble metal
layer for ﬁlms prepared by rheotaxial growth and thermal
oxidation (RGTO) [227] should not exceed 2.5 nm to attain the
same effect during surface doping. Estimations, given in Refs.
[118,228,229], indicate that for thin metal oxides this thickness
should be even less. This research has shown that for the
formation of an atomically dispersed Pd layer at the SnO
2
surface, the Pd thickness should not exceed 1 ML. The

deposition of several ML equivalents of Pd leads to the
formation of three-dimensional Pd particles with an average
size in the range 3–5 nm [228,229]. Approximately the same
size of Pd-clusters on SnO
2
grain surfaces was observed at bulk
doping, when the concentration of doping additives was 5 wt%
[69]. Similar results were also obtained in Ref. [225]. Auger
spectroscopic studies on Pd and Pt supported on ZrO
2
show that
both Pd and Pt have a layer-by-layer growth at low metal
coverage and temperatures. At elevated temperatures, or
following annealing, the Pd-clusters are more raft-like (2-
dimensional), while the Pt-clusters show three-dimensional
islands [225].
Based on the catalytic activity of gold (Au) clusters
[223,230–232], one can judge how important it is to form and
stabilize noble metal clusters with a speciﬁed size. This
precious element has been known for a long time as being
essentially chemically inert (and therefore catalytically
inactive) in its bulk form [233]. Quite recently, however, it
was found that, when dispersed in the form of nano-scale
clusters and supported on transition metal oxide surfaces, such
as TiO
2
[223,231], Au exhibits very high catalytic activity at
low-temperature for the partial oxidation of CO, hydrocarbons,
hydrogenation of unsaturated hydrocarbons, and reduction of
nitrogen oxides [230]. It was shown that the catalytic properties

of Au nanoparticles depend speciﬁcally on the support, the
preparation method and, critically, on the size of the Au
clusters, which are most active when their diameter is smaller
than 3.5 nm [223,234] (see Fig. 34). Cluster diameters below
3 nm lead to a decrease in reactivity.
It is known that the activity of palladium substantially
increases with a decrease in cluster size. For example, in Ref.
[235] it was reported that the catalytic activity towards
electrochemical proton reduction is enhanced by more than two
orders of magnitude as the diameter of the palladium particles,
parallel to the support surface, decreases from 200 to 6 nm.
Thus, during fabrication of surface modiﬁed metal oxide gas
sensors, one should learn not to create clusters of speciﬁed size
and to stabilize their size in the desired range.
It is important to note that each surface catalyst has its own
optimal size for providing maximum response for each analyte.
The results obtained in Refs. [25,62,71,97,115,118] provide
good example of this effect. It is important to note that the size
of noble metal clusters that are optimal for gas sensitivity does
not coincide with the optimal size for heteroge neous catalysis
[217]. It therefore follows that the surface noble metal clusters
in gas sensors and heterogeneous catalysis perform different
functions.
2.3.3. Film texturing
This structural parameter has not been considered earlier
because in ceramics all grains are oriented arbitrarily. Only
since the transition to ﬁlm technology has ﬁlm texture started
attracting attention. For example, in Ref. [236] it was
established that the highly preferred orientation of homo-
geneous ZnO ﬁlms obviously decreased the electrical

resistance of the ZnO ceramics at room temperature. The
authors of Ref. [236] have found that textured ZnO ﬁlms have
more low-angle grain boundaries in comparison with the ones
without texture. Another interesting effect was observed in
Refs. [79,80]. It was found that for ozone detection using In
2
O
3
-
based sensors the recovery time increases as the ﬁlm texture
increases (see Fig. 35). For reducing gases this dependence is
observed for both response (t
res
) and recovery times (t
rec
).
Moreover, the change of t
res
and t
rec
in the senso r response to
reducing gases was subject to the same regularity of T
pyr
and
ﬁlm thickness inﬂuence as was the change in t
rec
for ozone
detection. In other words, the faster response times were
obtained from structures formed from arbitrarily oriented
crystals

Fig. 33. Possible positions of (a) Pd and (b) Pt-clusters on the surface of ZrO
2
.
(Reprinted with permission from Ref. [225]. Copyright 2004: Elsevier).
Fig. 34. CO oxidation turnover frequencies (TOFs) at 300 K as a function of the
average size of the Au clusters supported on a high surface area TiO
2
support
(7). The Au/TiO
2
catalysts were prepared by deposition–precipitation method,
and the average cluster diameters were measured by TEM. The solid line serves
merely to guide the eye. (Reprinted with permission from Ref. [223]. Copyright
1998: Science).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–3922
It has been assumed that the inﬂuence of the ﬁlm’s texture on
gas-sensing characteristics takes place through the change of
porosity of the gas-sensing matrix. In Ref. [24],In
2
O
3
ﬁlms
studied were textured in the (0 0 1) direction, perpendicular to
the substrate. The degree of texture increases when both the
deposition temperature and the ﬁlm thickness increase. One can
assume that the indicated parameters of In
2
O
3
ﬁlm deposition

promote more compact crystallite packing. The decrease of ﬁlm
porosity and gas-permeability may be the result of such
changes in the In
2
O
3
ﬁlm structure.
An attempt to estimate the inﬂuence of ﬁlm texture on the
In
2
O
3
conductivity response to NO
2
was made in Ref. [237].It
was found that In
2
O
3
ﬁlms with preferential orient ation in the
{2 1 1} direction had higher sensor response to NO
2
than the
layers with preferential {2 2 2} texture. Films studied in Ref.
[24] had a different texture than those prepared in [237].
Therefore, it was not possibl y to compare results discussed in
Refs. [24,237]. As a result, we can neither conﬁrm nor refute
the above-mentioned conclusions. However, the existence of
such an effect in an addition to the ﬁndings presented in Refs.
[77,78], allows us to conclude that the modiﬁcation of the ﬁlm

texture of polycrystalline ﬁlms as well as the decrease of grain
size and ﬁlm thickness, may open a new way for the
optimization of thin-ﬁlm gas sensor parameters.
2.3.4. Surface stoichiometry (disordering)
Surface stoichiometry (disordering) of the metal oxide
grains may be as important a parameter in gas-sensing as ﬁlm
thickness and grain size. This parameter determines the
adsorption ability and the surface charge through the number
of oxygen vacancies, and as a result, controls the initial surface
band bending (eV
s
) of the metal oxide and the change of eVs at
replacement of the surrounding gas.
Considering that the sensing prope rties of thin ﬁlm sensors
such as sensitivity, selectivity, and stability are strongly related
to their microstructures and to the exact stoichiometry of their
surfaces, an accurate control of these parameters is extremely
important for the production of sensors with reproducible
behavior [77,238]. For example, the change in the surface
oxygen composition of SnO
2
is accompanied by the change in
the electronic structure of the surface Sn from a valency of (IV)
for the stoichiometric surface to a valency of (II) for the reduced
surface. This change in the composition also has a strong
inﬂuence on its chemical and gas-sensing properties. In Ref.
[177] for the adsorption of water it was showed that it
dissociates on the stoichiometric surface but adsorbs weakly on
the reduced surface. While this easy reduction of the SnO
2

surface can play an important role in the gas-sensing behavior
of this material, it is unlikely to be of general importance for
other gas-sensing metal oxides. The particular behavior of the
SnO
2
surfaces arises from the dual valency of Sn. For other gas-
sensing materials like e.g. ZnO no variation in the lattice-
oxygen composition is expected.
It is known that the excessive lattice disordering of the
surface layer could be accompanied by a signiﬁcant growth of
native surface states. Theoretical simulations presented in Refs.
[46,47,85] indicate that the increase of these surface defects
could lead to pinning of the surface Fermi level position, and
correspondingly to the drop of response of gas sensors. This
means that if we want to maintain effective operation of solid-
state gas sensors, the concentration of those states as well as the
lattice disordering of the surface layer of metal oxide grains
should be minimized. Only in this case the surface Fermi level
will not be pinned, because the charge of native surface states
becomes comparable or less than the charge of chemisorbed
particles. The indicated surface property creates a condition for
modulation of surface band bending of semiconductors with the
change of the surrounding atmosphere. The change of surface
stoichiometry of metal oxide ﬁlms may result from changes in
deposition and annealing temperature.
High resolution transmission electron microscopy (HRTEM)
images of SnO
2
grains, presented in Ref. [239] (see Fig. 36),
indicate that this effect could occur in metal oxides. In many

cases the particles of SnO
2
consist of a well-crystallized core
covered with an amorphous layer of tin oxide. Further, such
amorphous layers could appear in undoped tin dioxide fabricated
with low temperature technological routes, or in doped material
at superﬂuous concentrations of doping additives.
The same conclusion was made in Ref. [240] based on an
analysis of the complete Rama n spectrum of nanometric SnO
2
particles. The authors of Ref. [240] found that for grains smaller
than 7 nm, two bands appeared in the high-frequency region of
spectrum. They proposed that these bands were due to a surface
layer of nonstoichiometric SnO
2
with a different symmetry than
SnO
2
. The thickness of this layer was calculated to be $1.1 nm,
which is about two to three unit cells.
One can conclude that the presence of such a layer on the
surface of small grains is a main reason for lower thermal
stability of the metal oxide structure and the enhanced
sensitivity to air humidity of sensors fabricated on the base
of ﬁne-dispersed material with grain size smaller than 5–7 nm.
The presence of an unstoichiometric surface layer can also
be responsible for a strong interaction between noble metals
Fig. 35. The inﬂuence of In
2
O

3
ﬁlm texture on the recovery time during ozone
detection. Films with thickness 40–300 nm were deposited by spray pyrolysis
method from 1.0 M InCl
3
–water solution (T
oper
= 270 8C). (Reprinted with
permission from Ref. [79]. Copyright 2004: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 23
and metal oxides. As it is known, the role and activity of noble
metal catalysts incorporated on solid-state sensors are
determined by ether their chemical state, aggregation form,
or interaction with metal oxide [219]. The formation of alloys is
the main reason for the indicated dependence [241].
Prior studies [242] assumed that the surface stoichiometry
of metal oxides, through a charge transfer induced diffusion
mechanism for O
2
molecules adsorbed on a metal oxide
surface, can also control the surface diffusion of oxygen
molecules. Time-resolved scanning tunneling microscopy of
the TiO
2
(1 1 0) surface has shown that the O
2
hopping rate
depended on the number of surface donors (oxygen
vacancies), which determines the density of conduction
band electrons. The authors of Ref. [242] assumed that the

metal oxides act as reservoirs for oxygen; and the O
2
diffusion may be a rate-limiting step in oxidation processes
on these metal oxides. Diffusion of oxygen molecules
on a metal oxide surface plays a vital role in gas-sensing
effects and therefore those results may h ave implications for
the understanding of their nature. This mechanism is
expected to be an important one for such reducible oxides
as TiO
2
,Fe
2
O
3
,SnO
2
, and ZnO, where shallow donor states
provide a rise to a high density of electrons in the conduction
band.
2.4. The role of morphology and porosity of metal oxides
2.4.1. Grain networks, porosity, and the area of inter-grain
contacts
From an analysis of the numerous param et ers which can
affect chemo-resistance sensors, one can conclude that in order
to achieve maximum gas sensi tivity it is necessary either to
increase the role of surface c onductivity or to increase the
contribution of inter-crystalline barriers by decreasing
the contact area or the width of the neck [46,47,50] (see
Figs. 9 and 20).
Fig. 37. Schematic representation of material density inﬂuence on gas penetrability of gas-sensing matrix.

Fig. 36. HREM images of a SnO
2
nanoparticles showing the crystallized tin oxide core and the amorphous tin oxide shell: (a) SnO
2
:Pd; (b) undoped SnO
2
. (Adapted
with permission from Ref. [239]. Copyright 2000: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–3924
At present there is no doubt that the simplest method to
achieve maximum sensitivity independently on gas-sensing
material is to increas e the ﬁlm’s porosity [65,137,243–247].In
the case of compact layers, the interaction with gases takes
place only at the geometric surface (see Fig. 37).
In porous material the volume of the layer is accessible to the
gases and theref ore the active surface is much higher than the
geometric area and the sensor response is higher [24]. A greater
porosity results in a smaller number of contacts with the necks
that are not being overlapped under interaction with the
surrounding gas. Experimental results obtained from a study of
gas-sensing characteristics of SnO
2
ﬁlms synthesized by the
RGTO technique [227,248] are in good agreement with this
conclusion. The ﬁlms had high sensitivity, in spite of the high
level of agglomeration. Further, it was observed that ﬁlms with
minimal contact area between agglomerates showed maximum
sensitivity. This was clearly illustrated with the SEM images
presented in Refs. [227,248]. Increased porosity also decreased
the probability of forming so-called‘‘capsulated zones’’ in the

volume of the gas-sensing layer. Capsulated zones are isolat ed
from contact with atmosphere, and therefore their resistance
does not depend on the surrounding gas [106]. Density,
porosity, pore size and gas-permeability can be determined by
the corresponding measurement techniques described in Refs.
[249,250,253].
It is necessary to note that the speciﬁc surface area
parameter, which is rarely used for gas-sensing material
characterization, deserves more attention. The authors of Ref.
[251] have the same opinion. Experiments, discussed in Refs.
[61,245,251,252], have shown that the use of material with high
speciﬁc surface area could lead the development of high-
sensitive metal oxide gas sensors (see Fig. 38).
Speciﬁc surface area has a more universal meaning than
generally used parameters, because it combines such para-
meters as porosity and grain size. From the data of speciﬁc
surface areas, the average grain size may be calculated using the
equation [251,254]
d
SA
¼
6
rA
; (2)
where d
SA
is the average grain size of spherical particles, A is
the surface area of the powder, and r is the theoretical density of
SnO
2

. Moreover, in many cases the speciﬁc surface area shows
better correlation with gas response than does crystallite size
[93,245].
It is important to note that, as a rule, the increase of speciﬁc
surface area of a gas-sensing material is accompanied by the
shift of the sensitivity maximum into the range of lower
operating temperatures (see Fig. 39). One can assume this effect
is caused by the thermally activated nature of gas diffusion
inside the gas-sensing matrix [100].
Speciﬁc surface area data, estimated by classical multipoint
Brunauer, Emmett, and Teller (BET) adsorption techniques,
and grain size, calculated by using X-ray diffraction (XRD)
data, provide a more complete and reliable description of the
gas-sensing material. The difference between grain sizes,
determined by those methods (see Table 2), may be used for
porosity characterization [251]. However, the speciﬁc surface
area should be determined only after conducting all techno-
logical operations, used during the sensor’s fabrication.
Thus, as the porosity and active surface area of the gas-
sensing material increases, the sensor response increases as
well (see Figs. 38 and 39). This is consistent with the
conclusion, made in Ref. [220], that the porosity and speciﬁc
surface area are basic factors that inﬂuence the solid/gas
interactions and ultimately the material performance in gas
detection. This is an important conclusion. However, it is
necessary to know that in some cases dense, nonporous metal
oxides may compensate their lower sensitivity and other
shortcomings by having highe r temporal and thermal stability
[255,256].
The results presented in Refs. [79,137,257] show that gas

sensors with higher porosity have faster response. In Ref. [137]
Fig. 38. Relationship between surface areas of the SnO
2
sensors and their
sensitivity to 500 ppm of H
2
and CO at 300 8C. Sensors were fabricated by
pressing and annealing at 450 8C calcined SnO
2
powders into 14 mm diameter
and 1 mm thickness pellets with two Pt electrodes. SnO
2
for these experiments
were prepared by a surfactant-templating method. The as-synthesized SnO
2
samples were calcined at 450 8C. (Adapted with permission from Ref. [252].
Copyright 1999: Elsevier).
Fig. 39. Inﬂuence of surface area of SnO
2
-based thick ﬁlm sensors on response
to 500 ppm of H
2
: 1–3 m
2
/g; 2–54 m
2
/g; 3–69 m
2
/g; 4–77 m
2

/g; 5–92 m
2
/g; 6–
99 m
2
/g. Tested sensors were fabricated using technology described in Fig. 38.
(Adapted with permission from Ref. [252]. Copyright 1996: Elsevier).
G. Korotcenkov / Materials Science and Engineering R 61 (2008) 1–39 25

the role of morphology and crystallographic structure of metal oxides

Tài liệu liên quan

Tài liệu bạn tìm kiếm đã sẵn sàng tải về